EMBRYOPSIDA Pirani & Prado

Gametophyte dominant, independent, multicellular, initially ±globular, not motile, branched; showing gravitropism; glycolate oxidase +, glycolate metabolism in leaf peroxisomes [glyoxysomes], acquisition of phenylalanine lysase* [PAL], flavonoid synthesis*, microbial terpene synthase-like genes +, triterpenoids produced by CYP716 enzymes, CYP73 and phenylpropanoid metabolism [development of phenolic network], xyloglucans in primary cell wall, side chains charged; plant poikilohydrous [protoplasm dessication tolerant], ectohydrous [free water outside plant physiologically important]; thalloid, leafy, with single-celled apical meristem, tissues little differentiated, rhizoids +, unicellular; chloroplasts several per cell, pyrenoids 0; centrioles/centrosomes in vegetative cells 0, microtubules with γ-tubulin along their lengths [?here], interphase microtubules form hoop-like system; metaphase spindle anastral, predictive preprophase band + [with microtubules and F-actin; where new cell wall will form], phragmoplast + [cell wall deposition centrifugal, from around the anaphase spindle], plasmodesmata +; antheridia and archegonia +, jacketed*, surficial; blepharoplast +, centrioles develop de novo, bicentriole pair coaxial, separate at midpoint, centrioles rotate, associated with basal bodies of cilia, multilayered structure + [4 layers: L1, L4, tubules; L2, L3, short vertical lamellae] (0), spline + [tubules from L1 encircling spermatid], basal body 200-250 nm long, associated with amorphous electron-dense material, microtubules in basal end lacking symmetry, stellate array of filaments in transition zone extended, axonemal cap 0 [microtubules disorganized at apex of cilium]; male gametes [spermatozoids] with a left-handed coil, cilia 2, lateral, asymmetrical; oogamy; sporophyte +*, multicellular, growth 3-dimensional*, cuticle +*, plane of first cell division transverse [with respect to long axis of archegonium/embryo sac], sporangium and upper part of seta developing from epibasal cell [towards the archegonial neck, exoscopic], with at least transient apical cell [?level], initially surrounded by and dependent on gametophyte, placental transfer cells +, in both sporophyte and gametophyte, wall ingrowths develop early; suspensor/foot +, cells at foot tip somewhat haustorial; sporangium +, single, terminal, dehiscence longitudinal; meiosis sporic, monoplastidic, MTOC [= MicroTubule Organizing Centre] associated with plastid, sporocytes 4-lobed, cytokinesis simultaneous, preceding nuclear division, quadripolar microtubule system +; wall development both centripetal and centrifugal, 1000 spores/sporangium, sporopollenin in the spore wall* laid down in association with trilamellar layers [white-line centred lamellae; tripartite lamellae]; plastid transmission maternal; nuclear genome [1C] <1.4 pg, main telomere sequence motif TTTAGGG, KNOX1 and KNOX2 [duplication] and LEAFY genes present, ethylene involved in elongation; chloroplast genome with introns (not: Mesostigma), close association between trnLUAA and trnFGAA genes [precursors for starch synthesis], tufA, minD, minE genes moved to nucleus; mitochondrial trnS(gcu) and trnN(guu) genes +.

Many of the bolded characters in the characterization above are apomorphies of more or less inclusive clades of streptophytes along the lineage leading to the embryophytes, not apomorphies of crown-group embryophytes per se.

All groups below are crown groups, nearly all are extant. Characters mentioned are those of the immediate common ancestor of the group, [] contains explanatory material, () features common in clade, exact status unclear.


Sporophyte well developed, branched, branching dichotomous, potentially indeterminate; hydroids +; stomata on stem; sporangia several, terminal; spore walls not multilamellate [?here].


Sporophyte long lived, cells polyplastidic, photosynthetic red light response, stomata open in response to blue light; plant homoiohydrous [water content of protoplasm relatively stable]; control of leaf hydration passive; plant endohydrous [physiologically important free water inside plant]; PIN[auxin efflux facilitators]-mediated polar auxin transport; (condensed or nonhydrolyzable tannins/proanthocyanidins +); borate cross-linked rhamnogalactan II, xyloglucans with side chains uncharged [?level], in secondary walls of vascular and mechanical tissue; lignins +; roots +, often ≤1 mm across, root hairs and root cap +; stem apex multicellular [several apical initials, no tunica], with cytohistochemical zonation, plasmodesmata formation based on cell lineage; vascular development acropetal, tracheids +, in both protoxylem and metaxylem, G- and S-types; sieve cells + [nucleus degenerating]; endodermis +; stomata numerous, involved in gas exchange; leaves +, vascularized, spirally arranged, blades with mean venation density ca 1.8 mm/mm2 [to 5 mm/mm2]; sporangia in strobili, sporangia adaxial, columella 0; tapetum glandular; sporophyte-gametophyte junction lacking dead gametophytic cells, mucilage, ?position of transfer cells; MTOCs not associated with plastids, basal body 350-550 nm long, stellate array in transition region initially joining microtubule triplets; archegonia embedded/sunken [only neck protruding]; embryo suspensor +, shoot apex developing away from micropyle/archegonial neck [from hypobasal cell, endoscopic], root lateral with respect to the longitudinal axis of the embryo [plant homorhizic].


Sporophyte growth ± monopodial, branching spiral; roots endomycorrhizal [with Glomeromycota], lateral roots +, endogenous; G-type tracheids +, with scalariform-bordered pits; leaves with apical/marginal growth, venation development basipetal, growth determinate; sporangium dehiscence by a single longitudinal slit; cells polyplastidic, MTOCs diffuse, perinuclear, migratory; blepharoplasts +, paired, with electron-dense material, centrioles on periphery, male gametes multiciliate; nuclear genome [1C] 7.6-10 pg [mode]; chloroplast long single copy ca 30kb inversion [from psbM to ycf2]; mitochondrion with loss of 4 genes, absence of numerous group II introns; LITTLE ZIPPER proteins.


Sporophyte woody; stem branching axillary, buds exogenous; lateral root origin from the pericycle; cork cambium + [producing cork abaxially], vascular cambium bifacial [producing phloem abaxially and xylem adaxially].


Growth of plant bipolar [plumule/stem and radicle/root independent, roots positively geotropic]; plants heterosporous; megasporangium surrounded by cupule [i.e. = unitegmic ovule, cupule = integument]; pollen lands on ovule; megaspore germination endosporic, female gametophyte initially retained on the plant, free-nuclear/syncytial to start with, walls then coming to surround the individual nuclei, process proceeding centripetally.


Plant evergreen; nicotinic acid metabolised to trigonelline, (cyanogenesis via tyrosine pathway); microbial terpene synthase-like genes 0; primary cell walls rich in xyloglucans and/or glucomannans, 25-30% pectin [Type I walls]; lignin chains started by monolignol dimerization [resinols common], particularly with guaiacyl and p-hydroxyphenyl [G + H] units [sinapyl units uncommon, no Maüle reaction]; roots often ≥1 mm across, stele diarch to pentarch, xylem and phloem originating on alternating radii, cork cambium deep seated, gravitropism response fast; stem apical meristem complex [with quiescent centre, etc.], plasmodesma density in SAM 1.6-6.2[mean]/μm2 [interface-specific plasmodesmatal network]; eustele +, protoxylem endarch, endodermis 0; wood homoxylous, tracheids and rays alone, tracheid/tracheid pits circular, bordered; mature sieve tube/cell lacking functioning nucleus, sieve tube plastids with starch grains; phloem fibres +; cork cambium superficial; leaf nodes 1:1, a single trace leaving the vascular sympodium; leaf vascular bundles amphicribral; guard cells the only epidermal cells with chloroplasts, stomatal pore with active opening in response to leaf hydration, control by abscisic acid, metabolic regulation of water use efficiency, etc.; branching by axillary buds, exogenous; prophylls two, lateral; leaves with petiole and lamina, development basipetal, lamina simple; sporangia borne on sporophylls; spores not dormant; microsporophylls aggregated in indeterminate cones/strobili; grains monosulcate, aperture in ana- position [distal], primexine + [involved in exine pattern formation with deposition of sporopollenin from tapetum there], exine and intine homogeneous, exine alveolar/honeycomb; ovules with parietal tissue [= crassinucellate], megaspore tetrad linear, functional megaspore single, chalazal, sporopollenin 0; gametophyte ± wholly dependent on sporophyte, development initially endosporic [apical cell 0, rhizoids 0, etc.]; male gametophyte with tube developing from distal end of grain, male gametes two, developing after pollination, with cell walls; embryo cellular ab initio, suspensor short-minute, embryonic axis straight [shoot and root at opposite ends], primary root/radicle produces taproot [= allorhizic], cotyledons 2; embryo ± dormant; chloroplast ycf2 gene in inverted repeat, trans splicing of five mitochondrial group II introns, rpl6 gene absent; ??whole nuclear genome duplication [ζ/zeta duplication event], 2C genome size (0.71-)1.99(-5.49) pg, two copies of LEAFY gene, PHY gene duplications [three - [BP [A/N + C/O]] - copies], 5.8S and 5S rDNA in separate clusters.

VII. ANGIOSPERMAE Lindley / MAGNOLIIDAE Takhtajan - Back to Main Tree

Lignans, O-methyl flavonols, dihydroflavonols, triterpenoid oleanane, apigenin and/or luteolin scattered, [cyanogenesis in ANA grade?], lignin also with syringyl units common [G + S lignin, positive Maüle reaction - syringyl:guaiacyl ratio more than 2-2.5:1], hemicelluloses as xyloglucans; root cap meristem closed (open); pith relatively inconspicuous, lateral roots initiated immediately to the side of [when diarch] or opposite xylem poles; epidermis probably originating from inner layer of root cap, trichoblasts [differentiated root hair-forming cells] 0, hypodermis suberised and with Casparian strip [= exodermis]; shoot apex with tunica-corpus construction, tunica 2-layered; starch grains simple; primary cell wall mostly with pectic polysaccharides, poor in mannans; tracheid:tracheid [end wall] plates with scalariform pitting, multiseriate rays +, wood parenchyma +; sieve tubes enucleate, sieve plates with pores (0.1-)0.5-10< µm across, cytoplasm with P-proteins, not occluding pores of plate, companion cell and sieve tube from same mother cell; ?phloem loading/sugar transport; nodes 1:?; dark reversal Pfr → Pr; protoplasm dessication tolerant [plant poikilohydric]; stomata randomly oriented, brachyparacytic [ends of subsidiary cells ± level with ends of guard cells], outer stomatal ledges producing vestibule, reduction in stomatal conductance with increasing CO2 concentration; lamina formed from the primordial leaf apex, margins toothed, development of venation acropetal, overall growth ± diffuse, secondary veins pinnate, fine venation hierarchical-reticulate, (1.7-)4.1(-5.7) mm/mm2, vein endings free; flowers perfect, pedicellate, ± haplomorphic, protogynous; parts free, numbers variable, development centripetal; P = T, petal-like, each with a single trace, outer members not sharply differentiated from the others, not enclosing the floral bud; A many, filament not sharply distinguished from anther, stout, broad, with a single trace, anther introrse, tetrasporangiate, sporangia in two groups of two [dithecal], each theca dehiscing longitudinally by a common slit, ± embedded in the filament, walls with at least outer secondary parietal cells dividing, endothecium +, cells elongated at right angles to long axis of anther; tapetal cells binucleate; microspore mother cells in a block, microsporogenesis successive, walls developing by centripetal furrowing; pollen subspherical, tectum continuous or microperforate, ektexine columellate, endexine restricted to the apertural regions, thin, compact, intine in apertural areas thick, orbicules +, pollenkitt +; nectary 0; carpels present, superior, free, several, spiral, ascidiate [postgenital occlusion by secretion], stylulus at most short [shorter than ovary], hollow, cavity not lined by distinct epidermal layer, stigma ± decurrent, carinal, dry; suprastylar extragynoecial compitum +; ovules few [?1]/carpel, marginal, anatropous, bitegmic, micropyle endostomal, outer integument 2-3 cells across, often largely subdermal in origin, inner integument 2-3 cells across, often dermal in origin, parietal tissue 1-3 cells across, nucellar cap?; megasporocyte single, hypodermal, functional megaspore lacking cuticle; female gametophyte lacking chlorophyll, four-celled [one module, egg and polar nuclei sisters]; ovule not increasing in size between pollination and fertilization; pollen grains bicellular at dispersal, germinating in less than 3 hours, siphonogamy, pollen tube unbranched, growing towards the ovule, between cells, growth rate (ca 10-)80-20,000 µm h-1, tube apex of pectins, wall with callose, lumen with callose plugs, penetration of ovules via micropyle [porogamous], whole process takes ca 18 hours, distance to first ovule 1.1-2.1 mm; male gametophytes tricellular, gametes 2, lacking cell walls, ciliae 0, double fertilization +, ovules aborting unless fertilized; fruit indehiscent, P deciduous; mature seed much larger than fertilized ovule, small [<5 mm long], dry [no sarcotesta], exotestal; endosperm +, ?diploid [one polar nucleus + male gamete], cellular, development heteropolar [first division oblique, micropylar end initially with a single large cell, divisions uniseriate, chalazal cell smaller, divisions in several planes], copious, oily and/or proteinaceous, embryo short [<¼ length of seed]; plastid and mitochondrial transmission maternal; Arabidopsis-type telomeres [(TTTAGGG)n]; nuclear genome [2C] (0.57-)1.45(-3.71) [1 pg = 109 base pairs], ??whole nuclear genome duplication [ε/epsilon event]; ndhB gene 21 codons enlarged at the 5' end, single copy of LEAFY and RPB2 gene, knox genes extensively duplicated [A1-A4], AP1/FUL gene, palaeo AP3 and PI genes [paralogous B-class genes] +, with "DEAER" motif, SEP3/LOFSEP and three copies of the PHY gene, [PHYB [PHYA + PHYC]]; chloroplast IR expansions, chlB, -L, -N, trnP-GGG genes 0. - orders, families, (223,300-)352,000(-422,127) species.


Note: In all node characterizations, boldface denotes a possible apomorphy, (....) denotes a feature the exact status of which in the clade is uncertain, [....] includes explanatory material; other text lists features found pretty much throughout the clade. Note that the precise node to which many characters, particularly the more cryptic ones, should be assigned is unclear. This is partly because homoplasy is very common, in addition, basic information for all too many characters is very incomplete, frequently coming from taxa well embedded in the clade of interest and so making the position of any putative apomorphy uncertain. Then there are the not-so-trivial issues of how character states are delimited and ancestral states are reconstructed (see above).

On occasion, I use the term "basal" below, but this is in the context of a ladderized tree and refers to the branch in which there has subsequently been less diversification than in its sister clade, or branch(es) immediately below the clade that is currently being discussed. Thus Equisetum, Amborella, Acorus, Enkianthus, Humbertia, and many other clades can all be called "basal" from this point of view. (At one level, it is difficult to talk about branching of a phylogenetic tree; there is no trunk, and no "branch", i.e., one of a pair of sister clades, can be basal to the other.) However, note that this has only the topological connotations I have just mentioned, and there is certainly no implication that these basal taxa necessarily lack apomorphies or are "primitive". Indeed, the terms "primitive" and "advanced" are heavily loaded and I have tried not to use them. "Plesiomorphic", and "derived" or "apomorphic", its opposite, refer to the evolutionary status of individual characters.

Age. Age estimates of crown-group angiosperms vary considerably, although many are in the range (210-)150-140(-130) Ma (e.g. J. A. Doyle 2001; Sanderson & Doyle 2001; Wikström et al. 2001; Aoki et al. 2004; Davis et al. 2004a; Sanderson et al. 2004; Bell et al. 2005; Leebens-Mack et al. 2005; Moore et al. 2007; Soltis et al. 2008: a variety of estimates; Moore et al. 2010; S. A. Smith & Brown 2018; Lutzoni et al. 2018; Y. Yang et al. 2020: Suppl. Fig. 22). Bell et al. (2010: note topology) suggest ages of (199-)183(-167) and (154-)147(-141) Ma (see also Magallón 2009) and Iles et al. (2014) ages of (167.7-)158.7(-151) Ma. Some recent estimates based on molecular data tend to be substantially older, Magallón (2008 and references) and Magallón & Castillo (2009) noting ages of 182-158 Ma and 130 or 242 Ma respectively, i.e. mostly Lower to Middle Jurassic or older, indeed, S. A. Smith et al. (2010: see esp. Table S3) suggested that crown-group angiosperms were (270-)228, 217(-182) Ma, and there are ages of 275-216 Ma in Magallón (2010), (280-)246, 209(-186) Ma in Zeng et al. (2014; see also Rothfels et al. 2015b), 267-247 Ma in J. W. Clark and Donoghue (2017), and ca 279 Ma in Z. Wu et al. (2014). Yet again, other ages are somewhat younger, e.g. (240-)205(-175) Ma in Clarke et al. (2011: other dates), (256-)198(-163) Ma in N. Zhang et al. (2012, similar in Xue et al. 2012), (257.9-)208.7-193.7(-157.7) Ma in Magallón et al. (2013) for this clade and around 195.4-185.3 Ma in Naumann et al. (2013); see also Schneider et al. (2004). (152-)144(-133) Ma is the estimate in Silvestro et al. (2015), ca 244.7 Ma in Tank et al. (2015), (253-)221, 206(-176) Ma, or as much as ca 242 or as little as 161-154 Ma in Foster et al. (2016a, q.v. for caveats), and 246.5-197.5 Ma in Morris et al. (2018). Magallon et al. (2015) estimate the age of crown-group angiosperms to be around (141-)139.4(-136) Ma by taking the fossil record as an indicator of their age; they argue that since quite a number of fossils are known soon after this date, their absence before is perhaps real; dates of clades within angiosperms from this paper should be interpreted accordingly (see also Sanderson 2015). Beaulieu et al. (2015) suggest that some older ages for crown-group angiosperms should be treated with caution, finding ages of around 140-130 Ma where (270-)228-217(-182) Ma was the previous estimate (S. A. Smith et al. 2010), although Beaulieu et al. themselves were agnostic about what any "real" age might be. Ca 130 Ma is the estimate in Laenen et al. (2014), (237-)196(-161) Ma in Foster & Ho (2017), while the preferred ages in Salomo et al. (2017: 22 possible fossil calibrations) are (342-)284(-226) Ma (20 fossil calibrations), although other estimates using different combinations of fossil and seed plant age constraints, sampling, etc., ranged from (397-)363...257(-202) Ma, while 256-149 Ma was suggested by Barba-Montoya et al. (2018), a post-Jurassic origin being rejected, and 263-234 Ma by L. Zhang et al. (2019). See also Guindon (2018); Sauquet and Magallón (2018) have some comments on this rather extraordinary situation - it is amusing to see a graph of suggested angiosperm ages against the publication dates of these ages...

Indirect estimates of the age of angiosperms are interesting. Thus angiosperms are optimized as the ancestral host plants for sawfly larvae - and sawflies are estimated to be Triassic in age (Schneider 2016: see caveats there). Along the same lines, Kawahara et al. (2019) estimated that the age of Angiospermivora, a lepidopteran clade that includes nearly the entire group and almost all the caterpillars of which eat angiosperms, at (277-)256(-234) Ma, late Permian.

Evolution: Divergence & Distribution. For ages of clades throughout angiosperms, see e.g. Hedges and Kumar (2009: summary of early dates), Durka and Michalski (2012: the European flora), Zanne et al. (2014), Foster et al (2016a), Harris and Davies (2016: Table S4), S. A. Smith and Brown (2018) and Janssens et al. (2019/2020: rbcL, matK (inc. trnK), crown-group angiosperms ca 200 Ma). I have not tried to incorporate all these articles, some of which have hundreds of dates, below, also, check topologies when using these publications. See also Hedges et al. (2006), Kumar et al. (2017) and the TimeTree and the caveats on the Home page there. Also, note that as of xii.2019 onwards, nuclear genome/transcriptome/microsynteny analyses are increasingly suggesting a reevaluation of relationships immediately basal to the rosid and asterid clades, at the base of the asterids/campanulids/lamiids, etc. - some of our ideas of relationships, based as they have been largely on data from the plastome, will need adjustment.

For discussions on evolution and diversification, see e.g. Hasebe (1999), D. Soltis et al. (2005a, b), Laenen et al. (2014), Magallón et al. (2018), etc.; hundreds of diversification shifts within angiosperms are identified by Landis et al. (2018). Sánchez-Reyes et al. (2017) look at ages, diversification rates and age/richness correlations for a number of clades and Harris and Davies (2016) and especialy Henao Diaz et al. (2018/2019a, see also Wiens & Scholl 2019, Henao Diaz et al. 2019b) examine the relationship between clade age and diversity. Note, however, that it is more difficult to talk sensibly about evolution than we had thought ()...

A net diversification increase possibly associated with the ε/epsilon nuclear genome duplication of angiosperms (see that node) is placed at the Mesangiosperm node (Tank et al. 2015), i.e., there is a considerable lag between inital cause and effect. P. Soltis and Soltis (2016) briefly discuss the evolution of the flower in the context of this duplication. At 50 Ma, the lag period suggested by J. W. Clark and Donoghue (2017) between the ε duplication (319-297 Ma) and angiosperm diversification (267-247 My) is at the upper limits of Tank et al.'s estimates. Furthermore, if Clark and Donoghue (2017) are right in their estimate of the age of the duplication - 319-297 Ma - and authors like Magallón et al. (2015) are right in their estimate for the crown-group age of angiosperms - ca 139 Ma - the lag period is around 170 Ma. In any event, stem-group angiosperms have been evolving for anything from 84 to 273 My (Barba-Montoya et al. 2018; see also H.-T. li et al. 2019), and this is discussed further under both stem- and crown-group angiosperms. As we have seen above, estimates of the age of crown-group angiosperms are all over the map.

Be such dates as they may, Simonin and Roddy (2018) see genome downsizing in angiosperms as critical to their rise to global dominance, in that it enabled smaller nuclei, and hence short stomatal guard cells, high stomatal density, high venation density and ultimately higher rates of gas exchange and photosynthetic rates. However, Amborella, etc., have small genomes, so like genome duplication there is a lag period between "cause" and effect. ANA grade angiosperms have rather low rates of speciation if perhaps slightly higher rates of genome size evolution (rate of evolution, rather than size per se, seems to be the important thing), a situation that does not really change until the evolution of monocots and core eudicots (Puttick et al. 2015; see also Leitch & Leitch 2008). In any event, the clade size imbalance at the base of the angiosperm tree should give us pause for thought when thinking about floral characters, genome duplications, etc., and their relationship to diversification (see also Lunau 2004). Whatever any global drivers of angiosperm diversification might be, more locally-operating features and clade diversifications greatly help in understanding overall angiosperm diversity (e.g. Crepet & Niklas 2009; P. Soltis et al. 2019; Onstein 2019). Crepet and Niklas (2009) noted that episodic high diversification rates had persisted in angiosperm evolution, while although such rates had been high initially in gymnosperms and ferns, they decreased. Angiosperms may be distinguished by the continued evolution of morphological and reproductive innovations, evolutionary "reinvention", associated with evolutionary variation involving much parallelism on the new themes that had evolved (Crepet & Niklas 2009; Onstein 2019). Whether such "'trait flexibility' and 'trait-dependent diversification'" are themselves heritable traits (Onstein 2019: p.6) is perhaps a separate issue.

Where exactly to place many characters on the angiosperm tree is unclear. This is in part because some taxa basal to the [magnoliid + monocot + eudicot] group have been surprisingly little studied and there is also considerable homoplasy as well as variation within and between families of the ANA grade in particular for several characters. For example, if reticulate-perforate pollen is optimized to the [Austrobaileyales + other angiosperms] node (see Friis et al. 2009b for a discussion), it effectively makes the pollen morphology of the common ancestor of all angiosperms ambiguous. 34/44 origins of perianth differentiation occur in the ANA grade, Magnoliidae, and monocots (Reyes et al. 2018: there are also 117 reversals all told). Other features such as a nucellus only one (Nymphaeales) to three cells across above the embryo sac and a stylar canal lacking an epidermal layer that appear to be plesiomorphous for basal grade angiosperms (Williams 2009), however, where on the tree a thicker nucellus and a stylar epidermal layer are acquired has not been indicated. For other features such as details of sugar transport in the phloem, which has both ecological and phylogenetic correlations, see below; their placement on the tree is frankly speculative. Different topologies of ANA-grade angiosperms (see below) may have little effect when thinking about the evolution of characters like those of habit/habitat (c.f. Barkman 2000b; Drew et al. 2014), but together the patterns of variation of characters, problems with the delimitation of character states and their optimization (see above) should cause some concern, furthermore, Hydatellaceae in particular are highly derived (see also Endress & Doyle 2015) and relationships immediately above the ANA grade are unclear. The morphology of those gymnosperms - currently largely unknown - that are on the angiosperm stem clade and when angiospermy itself evolved will also affect where apomorphies placed at the crown-group-angiosperm node are to be pegged. "Key evolutionary innovations" are placed on a tree by Z.-H. Wang et al. (2017) in odd places, see also the tree topology, also odd.

The pentatomy in the main tree here above the ANA grade makes life very difficult, to say the least. Note that one topology used by Endress and Doyle (2015) on which to optimize characters is [ANA-grade angiosperms [[Chloranthales + Ceratophyllales] [[magnoliids + monocots] eudicots]], in another the relationships are [ANA-grade angiosperms [[Chloranthales + magnoliids] [monocots [Ceratophyllales + eudicots]]] (see the chloroplast gene analyses of Jansen et al. 2007; Moore et al. 2007). Zeng et al. (2014) optimised the distribution of a large number of characters on a tree with a topology [monocots [magnoliids [[Ceratophyllum + Chloranthaceae] [eudicots]]] (see also below). Details of the exact position and magnitude of changes in characters like leaf venation density and pollen tube growth rate are provisional (see Boyce et al. 2008; Williams 2008: Reese & Williams 2019 for more details).

For possible floral features of the ancestral angiosperm, see e.g. Doyle and Endress (2000), Endress (2001a), Endress and Doyle (2015), Gottsberger (2016a) and in particular Sauquet et al. (2017). Sauquet et al. (2017) suggested that that the ancestral condition for both the perianth and androecium was to be 3-merous and at least 4-whorled, the gynoecium being spirally arranged, and that much subsequent floral evolution, e.g. of the pentapetalous flower, was by merging of adjacent whorls, rather different from evolution from an ancestor with spirally-arranged parts, the conventional view (for spiral→whorled, see also Ronse de Craene 2018). However, support for their preferred hypothesis was not strong (Sauquet et al. 2017: Fig. 1) and may well be the result of the approaches used: "MP... has the attractive property that no assumption is made on the relationship between change and branch length; ... Conversely, ML and rjMCMC results are conditional on the assumption that rates of morphological change are constant through times and across lineages." (Sauquet et al. 2017: Supplementary Methods). Despite this apparent attractive property of MP, the behaviour of ML and rjMCMC methods was seen as an "advantage" (ibid.) and they used the results of the rjMCMC analyses as an initial guide to their thoughts about floral evolution. (Such tensions are evident in several other morphological analyses/character optimizations that have appeared in the last few years and that use similar methods.) Since molecular clocks are somewhat out of favour these days, the reasons for invoking a morphological clock, as in ML and rjMCMC methods, are unclear to me. In any event, one of the key aspects of this reconstruction is that a whorled androecium abuts a spiral qynoecium, although the likelihood of this has been questioned by Sokoloff et al. (2017b; see also De-Paula et al. 2018: comments on characters); who found no convincing examples of such phyllotactic shifts between androecium and gynoecium and very few between the perianth and androecium in extant angiosperms, and thought that the ancestral flower was likely to be entirely whorled or entirely spiral. However, the present is not necessarily a good guide to the past, as Sauquet et al. (2018) rightly observed - the pull of the present - but all bets would currently seem to be off (Ledford 2018). Sauquet et al. (2017) also noted that in a few cases phyllotactic shifts occurred during floral development, and in such cases they scored the taxon as it came to appear - and this would make it comparable to more of the literature that they included. However, emphasizing the earliest stages is surely equally if not more defensible (e.g. Stull et al. 2018; c.f. Vasconcelos et al. 2016).

Amborella itself is more or less dioecious (Anger et al. 2017), while taxa in Nymphaeales and Austrobaileyales have a variety of breeding systems. Primary polyandry in which stamens develop centripetally from separate primordia, is common in the ANA grade, Magnoliids, etc.; for further discussion of polyandry, see Pentapetalae and euasterids.

Protogyny is very common in basal angiosperms (Routley et al. 2004: Amborellaceae are ± dioecious), while protandry is common in eudicots and above Petrosaviales in monocots (see also Bertin & Newman 1993; Endress 1994b). For connections between protogyny and protandry and self compatability and self incompatability, see Routley et al. (2004 and references). Armbruster et al. (2002), Hristova et al. (2005) and Sage et al. (2009) suggest that the basic condition of the angiosperm stigma may have been dry, the incompatability reaction being at the stigmatic surface, and the pollen tubes growing intracellularly, rather than wet, the incompatability reaction occurring down the style, and the pollen tubes growing in the stigmatic secretion (see also Allen & Hiscock 2008; Thien et al. 2009); dry stigmas are often associated with sporophytic self incompatability, the probable ancestral state for angiosperms (Lau et al. 2017). For differences in pollen tube growth in Pinaceae and angiosperms see Rounds and Benzanilla (2013).

The "standard view" is that the ancestor of flowering plants had monosulcate pollen (e.g. Walker & Walker 1984; Doyle 2013), but pollen of Amborella is anaulcerate. However, if reticulate-perforate pollen is optimized to the second node on the tree (see Friis et al. 2009 for a discussion), it makes the pollen morphology of the common ancestor of all angiosperms ambiguous, and the exine of Amborella has been variously interpreted (J. A. Doyle 2009). For exine development and sporopollenin production, see e.g. Heslop-Harrison (e.g. 1968; papers in 1971), Ariizumi and Toriyama (2011), Chebli et al. (2012) and Quilichini et al. (2015), and for tapetum type and orbicule production in particular, see Verstraete et al. (2014 and references). Sporophytic tapetal cells play an important role in pollen wall development (Heslop-Harrison 1968; see also Blackmore et al. 2009 for a good introduction). For pollen characters of angiosperms of the ANA grade and the magnoliids, see also under the respective orders, Sampson (2000), Doyle (2005: c.f. topologies of trees used, 2008a, 2008b). The pollen morphology of Amborellaceae is not well understood and there is much variation in microsporogenesis and pollen morphology in Nymphaeales, Amborellaceae, etc. (e.g. Furness et al. 2002; Taylor et al. 2015). However, Wortley et al. (2015), Luo et al. (2015), L. Lu et al. (2015) and others - note character state delimitation, sampling, treatment of variation, optimization, etc. - are placing pollen variation on the angiosperm tree, and their work should be consulted by those interested. Mander (2016) looked at the result of pollen evolution in angiosperms by placing pollen form in a combinatorial morphospace and found much of the latter to be unoccupied. Pollenkitt of many angiosperms includes hydroxycinnamic acid (HCAA) and flavonol α-1,2-linked diglycosides (3-O-sophorosides) of tapetal origin that may be antioxidants protecting the pollen against UV radiation (Fellenberg & Vogt 2015: "tryphine"; see also below). Pacini and Hesse (2005) discuss possible functions of pollenkitt.

J. H. Williams (2007, 2012a, esp. 2008, 2009, 2012a, 2012b; Williams et al. 2016) look at pollen tube/male gametophyte development, etc., and Harder et al. (2016) at different patterns of pollen tube growth in the style, basically, the population ecology of male gametophytes. Pollen tubes of angiosperms generally grow very quickly, although a number of Fagales are exceptions (Germain 1994; Sogo & Tobe 2005d), and the sperm nuclei in some Liliaceae with much DNA in their nuclei may take 3-8 days to fuse with the female gamete after penetrating it (Bennett 1972). Perez di Giorgio et al. (2018) discuss aspects of the control of pollen tube growth, aquaporins novel to angiosperms being involved in the tranport of water and non-ionic compounds across tube membranes. As Williams showed, in angiosperms the male gametophyte is subject to novel interactions with the stigma and to competition with other male gametophytes as the pollen tubes grow down the stye/compitum. For the roles of the obturator, integuments and synergids in directing the final stages of pollen tube growth, see Lora et al. (2018 and references).

Armbruster et al. (2002) and X.-F. Wang et al. (2011) discuss the evolution of syncarpy and the complex patterns of gains and losses of various kinds of compita (see also Endress 2001a; Endress & Igersheim 2000; etc.). Staedler et al. (2009; see also Wang et al. 2011) thought that the presence of an extragynoecial compitum could be an apomorphy of angiosperms, and it would then have to be lost at least twice. See also Endress and Igersheim (1997 and references), Armbruster et al. (2002), and Endress (2015) for where to place compitum characters on the tree, but this will depend on the optimization procedure used and of course resolution of relationships along the backbone of the tree. For gynoecial morphology and carpel closure, about which much has been written, see e.g. Endress (2015), Sokoloff et al. (2013), Sauquet et al. (2017), etc.. In Amborellaceae and some other ANA-grade angiosperms, including Hydatellaceae, the stigma has multicellular papillae. Endress (2001a) notes i.a. that the carpels (?of the ancestral angiosperm) may have uniseriate hairs, and in Trimeniaceae and Nymphaeales the apical cell of these hairs is elongated and tanniniferous. For the evolution of placentation "types", see Ickert-Bond et al. (2014c). Features such as parietal tissue (= nucellus) two to three cell layers across above the embryo sac and a stylar canal lacking an epidermal layer are plesiomorphous for ANA angiosperms (J. H. Williams 2009), but where higher up on the tree they might change is unclear.

Friedman (2006; c.f. Tobe et al. 2000) described the very distinctive embryo sac of Amborella: A third synergid cell arises from a cell division that also produces the female gamete/egg, and the result is a 9-nucleate embryo sac. In nearly all other angiosperms the polar nuclei are sister to the egg nucleus (at one end) and the central chalazal nucleus (at the other), and the egg is produced by a nuclear division. Friedman and Ryerson (2009; see also Xi et al. 2014) discuss the evolution of the angiosperm embryo sac in detail, and that of Amborella may be derived, but it will depend on how characters are optimized. Porsch (1907) early took the view that the micropylar and chalazal ends of the eight-nucleate embryo sac were identical, so the eight- (or nine-)nucleate embryo sac would then result from the duplication of the 4-nucleate unit of Nymphaeales and Austrobaileyales. Tobe (2016) recently described embryogenesis and fertilization in the odd genus Cardiopteris (Aquifoliales-Cardiopteridaceae) where it seems that cells that would be antipodal cells in the normal Polygonum embryo sac form the egg apparatus, so there may be more flexibility in the development of the embryo sac than has been allowed. Indeed, the development of the embryo sac is a complicated operation, Tekleyohans et al. (2016: p. ) bemoaning the fact that "the remarkable ability of all FG [female gametophyte] cells, to be basically able to adopt any other FG cell identity, has complicated the interpretation of mutant phenotypes.".

A number of changes in the gametophytic phase of the angiosperm life cycle result in its usually being notably shorter than that of extant gymnosperms (J. H. Williams 2012b). The pollen tube growth rate in angiosperms is much higher than that of extant gymnosperms. Figures are 80-600 µm/hour in ANA-grade angiosperms, overall in angiosperms 10-20,000 µm/hour, with Fagaceae at the low end of the spectrum, versus 10-20 µm/hour in gymnosperms with Gnetum at the high end of the spectrum, and there is no correlation with genome size as there is in gymnosperms (Hoekstra 1983; esp. J. H. Williams 2008, 2009; Reese & Williams 2019; also Rudall & Bateman 2008). Fertilization occurs within about 24 hours of pollination in most angiosperms as compared to seven days or often far more in extant gymnosperms (Williams 2008). The development of callose plugs and callose in the pollen tube wall probably facilitates this faster pollen tube growth, and callose tubes and fast growth rates are both apomorphies of angiosperms. Although callose synthase genes are expressed in at least some gymnosperm pollen, the pollen tube there is cellulose-based (Williams 2008; Parre & Geitmann 2005: mechanical properties of callose; Abercrombie et al. 2011). Harder et al. (2016 and references) discuss different patterns of pollen tube growth in the style, basically, the population ecology of male gametophytes. There is much recent research in this whole area (see J. H. Williams & Mazer 2016 and refences).

In gymnosperms, there is much growth of the female gametophyte after pollination but before fertilization, often quite a lengthy period, and although growth is less in Gnetum, even there the ovule increases appreciably in size after pollination. The ovule itself grows little after fertilization since provisions for the developing embryo have already been sequestered in the massive female gametophyte. In most gymnosperms there is substantial loss to the plant if ovules have to abort/are not fertilized, however, as partial insurance, ovule growth in gymnosperms may depend on pollination, at least (Little et al. 2014).

On the other hand, there is usually little or no increase of angiosperm ovule size after pollination but before fertilization. The gametophyte remains very reduced, and since few reserves are committed to angiosperm ovules with their tiny mature female gametophytes, little is lost when unfertilized ovules abort. Resources start to be channelled to the developing embryo in a transfer mediated by the evolutionarily novel endosperm tissue that develops only after fertilization; maternal perisperm tissue may also be involved (Haig & Westoby 1991; Friedman 2001b; Baroux et al. 2002; Leslie & Boyce 2012; Sakai 2013; Little et al. 2014; Lafon-Placette & Köhler 2014; L. Yuan et al. 2018). Thus production of seed reserves is a post-fertilization event, and the pre-fertilization ovules are small, representing little investment by the plant, and many ovules can easily be accomodated within a single carpel (L. Yuan et al. 2018). Itagaki et al. (2019) summarized the literature on minimizing costs of ovule production, and found that there seems to be little correlation of ovule size with plant size.

For details of the fertilization process - the two male gametes are identical (but not in Plumbago zeylanica, a derived feature there), see Kawashima and Berger (2011); one gamete fuses with the egg cell, producing a zygote, the other fuses with a haploid or diploid maternal nucleus to produce endosperm, tissue involved in the nutrition of the embryo. Endosperm usually has a diploid maternal and a haploid paternal contribution and is unique to angiosperms, but little is known about its evolution (e.g. Baroux et al. 2002; Friedman & Williams 2004; Nowack et al. 2007). Interestingly, L. Yuan et al. (2018) recently found that in Ginkgo biloba the CKl1 gene, expressed in the neck canal cell, sister to the egg cell (and initially in other cells of the female gametophyte), can partly rescue a mutation of a gene in Arabidopsis involved in specifying the endosperm precursor polar central cell, near sister of the egg nucleus there. The ventral canal cell is sister to the egg cell, while in angiosperms one of the polar nuclei is in the same immediate lineage as the egg cell - and if the female gametophyte is built up of two 4-nucleate modules, then both polar nuclei may have similar origins.

Bachelier and Friedman (2011) discuss female gametophyte competition within a single ovule and its implications for angiosperm evolution. Whether or not a triploid endosperm is a synapomorphy for all angiosperms or only for those angiosperms above the ANA grade is unclear (Friedman 2001a, b, 2006; Baroux et al. 2002), but Friedman et al. (2003a, esp. b) and Friedman and Williams (2003, 2004) incline towards the latter hypothesis. If diploid endosperm is the ancestral condition, triploid endosperm will have evolved in parallel in Amborella and again in the mesangiosperm clade (see also J. H. Williams & Friedman 2004 and references; Xi et al. 2014). Why there should be variation in embryo sac development in the ANA grade and sporadically elsewhere, too, that affects the balance of maternal and paternal genes in the endosperm is unclear. However, a higher ratio of paternal genes in diploid compared to triploid endosperm (1:1, vs 1:2) may lead to more "selfish" behaviour of individual endosperm tissues as they scavenge nutrients at the expense of other ovules in the carpel (e.g. Friedman et al. 2008; esp. Friedman & Ryerson 2009) - hence perhaps the rather low ovule number (per carpel) of many ANA-grade angiosperms which have this higher ratio. Nymphaeales have only a slight amount of diploid endosperm and it probably functions as transfer tissue betweem embryo and perisperm (Friedman et al. 2012), the latter being tissue of maternal origin. Chalazal endosperm is involved in nutrient transfer from the mother to the embryo, and chalazal endosperm cells have larger nuclei caused by nuclear fusion or endoreduplication, processes that can be disturbed by genomic imbalance of the endosperm, resulting in the death of the embryo (Gehring & Satyaki 2017). Povilus et al. (2018) found that in Nymphaea thermarum increasing the relative male contribution to developing seed increased endosperm growth alone, while increasing the female ploidy level decreased seed growth in general, including that of the endosperm, irrespective of the paternal contribution; this was consistent with what happens in interploidy crosses in angiosperms. They interpreted this as showing how the female parent could control seed development (see also Haig 2013b). Female control is commonly exserted through the genomic imbalance of the triploid endosperm (female gene dosage relatively increased) or by maternal imprinting of the endosperm (Povilus et al. 2018); see also Gehring and Satyaki (2017) for the complexities of male and female genome control of the endosperm. Interestingly, in crosses involving parents of different ploidy levels, maternal genome excess is associated with early endosperm cellularization and paternal excess with delayed cellularization (Gehring & Satyaki 2017), and failure of cellularization - impaired cellularization timing - results in embryo arrest/death (e.g. Hehenberger et al. 2012; Lafon-Placette & Köhler 2014). One might wonder if there are connections with general endosperm evolution (cellular → nuclear) in angiosperms. However, in crosses between wild tomatoes, where endosperm is cellular, but derived, the endosperm does not start to break down until the early globular stage of the embryo. This is pretty much the same stage as in Arabidopsis, with nuclear endosperm (Roth et al. 2018), hexose levels increasing, although they normally decrease at this time, sucrose increasing (Hehenberger et al. 2012: the heart stage critical). For more on interactions between the embryo, the maternal parent, and triploid endosperm tissue, see Geist et al. (2019) and Lafon-Placette (2020), but to say the whole issue is complex and not well understood is an understatement.

Seed size and embryo length:seed length in ANA angiosperms are summarized in Losada et al. (2017). For the possibility that morphophysiological dormancy is the ancestral condition for angiosperms, see Willis et al. (2014b) and Fogliani et al. (2017).

The whole angiosperm life cycle is speeded up (Stebbins 1965, 1981), the evolution of carpels faciltating maximum seed production, seed dispersal, and seedling survival (Stebbins 1981). Seedlings of angiosperms grow faster than those of gymnosperms; again, time to maturity is reduced (e.g. Bond 1989; Coiffard et al. 2006). Overall, angiosperms tend to become mature at a younger age than do gymnosperms (Bond 1989; Verdú 2002). However, within gymnosperms, some species of Ephedra (Gnetales) are mature by about 7 years, although 20-100 years are the ages for most other gymnosperms (B. Wang et al. 2015).

The triterpenoid oleanane is widely distributed in angiosperms, in the long-extinct seed plant groups giganopterids and Bennettitales, and also in some ferns (Taylor et al. 2006; Feild & Arens 2007 and references; note that the triterpenoid isoarborinol is synthesized by a marine bacterium - Banta et al. 2017). For woodiness and wood anatomy, see S. Kim et al. (2004a) and Herendeen et al. (1999: a useful table), for changes in phyllotaxy, Ronse De Craene et al. (2003); chloranthoid leaf teeth - three veins converging at or below the apical gland of the tooth - may be a feature of all angiosperms (J. A. Doyle 2007; Doyle & Upchurch 2014). For a survey and evaluation of floral morphology and systematics, see Matthews and Endress (2012), Endress and Doyle (2015: ANA angiosperms in particular), Hickey and Taylor (1995), Rudall (2013) and Claßen-Bockhoff (2016b), also Taylor and Kirchner (1995: carpel evolution), Wing and Boucher (1998: ecology), Donoghue (2004: general), Whitney (2009: stronger selection for divergent flower than fruit morphology). For the evolution of microsporogenesis, see especially Furness et al. (2002b) and Taylor and Osborn (2006). What goes on in the male gametophyte in the period from pollination to fertilization is of much current interest (J. H. Williams & Mazer 2016 for literature). For variation in embryo size, see Verdú (2006). For various other aspects of early angiosperm evolution, see below.

Ecology & Physiology. See Early Cretaceous Evolution below for the ecology of ANA-grade angiosperms and of that of early angiosperms in general.

The mycorrhizal condition of members of the ANA grade is largely unknown, as is that of Canellales, Piperales, and most Laurales. Mycorrhizae are absent in Nymphaeales and Ceratophyllales, as might be expected for aquatic groups (Landis et al. 2002; B. Wang & Qiu 2006). However, the ancestor of Magnoliophyta may have had vesicular-arbuscular mycorrhizae, and this condition may have been a feature of the common ancestor of all seed plants, if not to be placed still deeper in the tree (see Maherali et al. 2016, also Werner et al. 2018).

The presence of vessels is placed at the [Austrobaileyales + the rest] node here. The distinction between vessels and tracheids can be hard to make (e.g. Carlquist 2012a, c), Amborella may have vessels only two cells long (Feild et al. 2000b), vessels in Nymphaeaceae and Cabombaceae are quite distinctive (Carlquist & Schneider 2009; Schneider & Carlquist 2009), and vessels are absent in Hydatellaceae. Overall, vessels of one sort or another may have arisen some three times in the basal part of the angiosperm tree, and vessels in taxa of the basal angiosperm branches may be less efficent in water transport than tracheids in Pinales, for example (e.g. Sperry 2003; Pitterman et al. 2005, 2010; Hacke et al. 2005, 2015). Along the same lines, note that the venation density of the lamina in ANA-grade angiosperms is similar to that of gymnosperms, bryophytes, etc., not like that of most Pentapetalae, etc., even if genome size is quite small. The physiological changes that may in part be responsible for angiosperm success seem not to have occurred in the basal branches of the angiosperms (e.g. Feild et al. 2009a, 2011a; Simonin & Roddy 2018 - see also Genes & Genomes below).

An arsenite transporter controlled by the single-copy ACR3 gene and promoting arsenic tolerance is found throughout land plants other than flowering plants (Indriolo et al. 2010), so its loss is another apomorphy.... Boyce at al. (2004) noted that the lignification of the primary cell wall was heavy in ANA-grade angiosperms (and in Drimys, a magnoliid) and gymnosperms, less in eudicots, and this had implications for ion-mediated xylem flow. However, the sampling is very preliminary. Compared with conifer lignin, primarily made up of guaiacyl (G) units, the syringyl(S)-rich lignins of angiosperms are less dense and less highly condensed and the polymer units are smaller and have more β-ether inter-unit linkages. Wood fibres are rich in S units, while in vessel elements G lignin predominates (Wagner et al. 2015); any implications this might have for fungus-mediated breakdown of lignin is unclear. p-hydroxybenzaldehyde, a component of many lignins, is apparently absent from broad-leaved angiosperms - at least from magnoliids and most eudicots (Towers & Gibbs 1953), but it is present in monocots and in some living gymnosperms (and also some Myrtaceae, etc.).

Eight families of Late Embryogenesis Abundant (LEA) genes that are expressed in response to stress - dessication in particular - are known from throughout flowering plants, including Amborella, but some are lost in aquatics (Artur et al. 2018). They are commonly expressed in seed development, but also in the adult stage of resurrection plants. The LEA_2 gene family is by far the largest, being ten times the size of all other families bar one, and interestingly, members of this family are both involved in stress responses (e.g. Graether & Boddington 2014: dehydrins) and also in plant hypersensitive responses after infections; tandem gene duplications seem to be involved in the diversification of this family (Artur et al. 2018).

Genes & Genomes. The Amborella Genome Project (2013) suggested that there had been a genome duplication, the ε duplication event, somewhere in the ancestry of angiosperms (see also Jiao et al. 2011), and J. W. Clark and Donoghue (2017) date this event to 319-297 Ma, S. Kim et al. (2004b) to ca 260 Ma and Hernández-Hernández et al. (2006) to ca 290 Ma (see also Karlgren et al. 2011; Landis et al. 2018). However, the timings of such early events have been questioned by Ruprecht et al. (2017) - they will have to be at least as old as the crown-group age of the clade with which they are associated - so although the ε duplication has been tentatively placed as an apomorphy for angiosperms above, be warned (see also Zwaenepoel & Van de Peer 2019). As will quickly become evident, there have been numerous genome duplication/polyploidy events and associated reductions in genome size, genome rearrangements and change in gene content pretty much throughout the angiosperm tree (Wendel 2015; see also below; Murat et al. 2017; Landis et al. 2018: Fig. 1, Appendix S1). Of course one of the problems with such early events is knowing exactly where they are to be placed - e.g. see the discussion about the γ genome duplication.

The base chromosome number of the ancestral angiosperm may be x = 5–7 (Stebbins 1971; Raven 1975), or more specifically x = 7 (Oginuma et al. 2000); for an estimate of the genome size of the ancestral angiosperm above, see above (Puttick et al. 2015). For 1C genome values, see e.g. Plant DNA C-values Database, consulted vi.2013 and subsequently (also Bennett & Leitch 2005, 2010). Genome size in many angiosperms is small, 1-1.4 picograms being the estimated ancestral nuclear C value (Masterson 1994; Leitch & Bennett 2005; Leitch et al. 2005; Puttick et al. 2015; Pellicer et al. 2018). However some Proteaceae, Liliales and Asparagales in particular have very large genomes. Extant gymnosperms have large genomes, although the sizes of those of Ephedra in particular are very variable and that of Gnetum notably small for a gymnosperm (Wan et al. 2018), and there genome size and ploidy level are quite closely connected (Leitch et al. 2005; Nystedt et al. 2013; Ickert-Bond et al. 2015a; Pellicer et al. 2018; see also below).

T. Zhao and Schranz (2019), looking at genome order across angiosperms, found that overall there was a low proportion of syntenic genes in the eudicots and even lower in monocots (the overall comparison was with mammals), but Brassicaceae and many Poaceae, however, had a notably higher proportion of such genes. On the other hand, angiosperms had many multicopy and/or lineage-specific microsynteny clusters, and again there was a phylogenetic signal, with a number of families or other clades having clusters specific to them (Zhao & Schranz 2019: over half the 107 species included were in four families).

Gossmann et al. (2016) compared genes expressed in the gametophyte and those expressed in the sporophyte and found that there was a higher proportion of rapidly evolving and young genes in the former, perhaps connected with factors like the low tissue complexity of the gametophyte (genes are developmentally less linked) and haploidy. Cui et al. (2015) suggested that new sporophytic genes for plant defence and the like might result from the competition between male gametophytes, and there may be a similar pattern in the moss Physcomitrella (O'Donoghue et al. 2013)...

A PCA analysis of functional protein domains suggested that angiosperms were rather different from the other vascular plants in the study, although Amborella was perhaps intermediate (Wan et al. 2018: sample size small). For B-function floral genes, etc., see S. Kim et al. (2004b) - synonymy: AP3 and PI with DEF and GLO respectively. The numbers of LATERAL ORGANS BOUNDARIES DOMAIN genes in Amborellales are around double those in extant gymnosperms, while elsewhere in the angiosperms numbers may be (much) higher again (Chanderbali et al. 2015). Ma et al. (2015) noted that some genes in the RNA-directed DNA methylation pathway (RdDM) pathway were restricted to angiosperms and that sRNA 24 nucleotides long is abundant there; in other land plants the peak was at 21 nucleotides, that peak being less obvious in angiosperms. Interestingly, Dicer-like 2 (DCL 2), a protein that makes sRNAs, is present in Pinaceae, Ginkgo and angiosperms, but not in the two Gnetales examined (?loss) (Ma et al. 2015: ?Cycadales). For the number and distribution of 5S and 35S rDNA sites along the chromosomes, see Roa and Guerra (2012) and especially Garcia et al. (2016). Thus there tend to be fewer 35S sites in angiosperms than in gymnosperms, and although the modal number in both is two per genome, four is close in angiosperms, but certainly not in gymnosperms; where possible the sites are at the end of the short arms of chromosomes (always in the terminal region in holocentric chromosomes), and they are at the ends of the chromosomes when there is only a single site. However, Pinaceae tended to differ from other gymnosperms in having notably more interstitial positions for both 5S and 35S rDNA sites, interestingly, Pinaceae (= Pinales) were more similar to angiosperms in the overall distribution of 5S rDNA positions while gymnosperms minus Pinales were more similar when it came to 35S rDNA positions (Garcia et al. 2016: Fig. 1). Rosato et al. (2016) discuss the distribution of 45S RNA in bryophytes s.l. in the context of bryophytes being a basal grade of land plants, and they found that the 5S and 45S genes were linked and overall there was little variation.

There are two quite extensive surveys (Corriveau & Coleman 1988; Q. Zhang et al. 2003) of plastid inheritance in angiosperms that have since been put in a phylogenetic context (Birky 1995: the first paper; Y. Hu et al. 2008: both papers). Perhaps 20% of all angiosperms have substantial biparental inheritance (bpi) of plastids, but very low levels are common; plastid inheritance is most often maternal, while strictly paternal inheritance is rare (e.g. Birky 1995; Snijder et al. 2007; Hu et al. 2008; Greiner et al. 2015) - basically a continuum. Active duplication of plastids in the generative cells of male gametes of taxa like Weigela floribunda that show bpi has been demonstrated (Hu et al. 2008 and references). Bpi is especially evident in cases of interspecific but not intraspecific hybridization (Hansen et al. 2007; see also Ruhlman & Jansen 2018). Barnard-Kubow et al. (2016) suggest that bpi helps rescue plants from cytonuclear incompatability, and it is noticeable that a number of clades with chloroplast genomes that are distinctive in one way or another (e.g. Fabaceae-IRL clade; Geraniaceae, Campanulaceae) include taxa with bpi (see also Shrestha et al. 2019). This may reflect general destabilization of the chloroplast genomes, so the existence of chloroplast variants may help here, but as Barnard-Kubow et al. (2016) note, bpi is known from taxa with apparently ordinary chloroplast genomes. The transmission of plastids, indeed, organelle genomes in general, are commonly, if unstably, of maternal inheritance (Greiner et al. 2014), although Q. Zhang and Sodmergen (2010) suggest that bpi . In any event, bpi seems to show a fair bit of phylogenetic signal, if for the most part it characterizes shallow nodes (Hu et al. 2008; c.f. in part Birky 1995); overall, sampling is still poor and I have rarely flagged "plastid inheritance biparental" as an apomorphy.

The basic angiosperm plastome contains 113 genes - 79 protein-coding genes, 30 transfer RNA genes, and 4 ribosomal RNA genes (Wicke et al. 2011a). For several chloroplast genes whose losses are synapomorphies for angiosperms, see Jansen et al. (2007), and for general chloroplast genome evolution, see Kua et al. (2012), and it will be clear that there have been extensive parallelisms in the evolution of the chloroplast genomes of echlorophyllous parasitic and mycoheterotrophic plants. Hein and Knoop (2018) discuss RNA editing in the chloroplast; variation here may have some phylogenetic signal. For the evolution of the IR/LSC junction in the chloroplast genome, see R.-J. Wang et al. (2008), Mower and Vickrey (2018), etc.. Extensive evolution in the NADH dehydrogenase-like (NDH) complex of genes is associated with this node (Ruhlman et al. 2017).

W. Guo et al. (2016a) summarize information on angiosperm mitochondrial genomes, and suggest that ancestrally there were 41 protein genes and 25 introns, the genome was slowly evolving, and so on.

Chemistry, Morphology, etc.. Xylans are more common than glucomannans in the cell wall (Zhong et al. 2019), as also in Gnetales. Magnoliales, at least, have glucoronosyl units every (6) 8 (10. 12) xylosyl residues, and they are acetylated, but there are no α-arabinosyl units (Busse-Wicher et al. 2016). For mannans, etc., see also Popper and Fry (2004): Austrobaileya has mannans, although two other members of the order sampled lack them, Nymphaea also has mannans, but other Nymphaeales and Amborella were not sampled. Pribat et al. (2010) discuss the distribution of a folate-dependent phenyanaline hydrolase, which may be correlated with major plant groups when sampling is improved. Triterpenoids are produced by a diversity of subgroups of CYP716 enzymes in angiosperms, especially in eudicots, but they are absent in monocots (Miettinen et al. 2017). It is possible that duplication of the CYP73 - cinnamate 4-hydroxylase - gene, involved in the second step of the phenolic network synthesising phenylpropanoids, occurred somewhere around here, and although Renault et al. (2017) are inclined to place class II CYP73 genes as an apomorphy of seed plants, they are known only from Taxaceae in gymnosperms.

For root traits, see Valverde-Barrantes et al. (2017), for root hairs see L. Huang et al. (2017), Hwang et al. (2017), Salazar-Henao et al. (2016) and references, also the discussion above. During the first year leaf traces make connections only with xylem (Tomlinson et al. 2006); c.f. Pinales. In plants that have rhizomes the hypodermis is often more or less suberized and has a Casparian strip (e.g. Perumalla et al. 1990b). Stomatal morphology and development in many members of the ANA grade is notably variable (e.g. Upchurch 1984; Rudall & Knowles 2013). For the plate meristem, submarginal meristematic tissue in the lamina made up of parallel layers of cells dividing anticlinally, so perpendicular to the surface of the lamina, and how it relates to venation density, see Sack et al. (2012); is the plate meristem an apomorphy of angiosperms? The filiform apparatus is developed at the micropylar ends of the synergid cells (Zini et al. 2016).

Phylogeny. Evidence currently favours the hypothesis that angiosperms are sister to a clade that includes all extant gymnosperms. For further information on the major seed plant groups, see above, Cupressales, Cycadales, Ginkgoales, Gnetales and Pinales, and for discussion about their relationships, see also Angiosperm History I, conifers in general, and extant seed plants in general.

Within early-diverging angiosperms, Donoghue and Mathews (1998) listed 16 different hypotheses of relationships that involved the first three nodes. However, Amborellaceae alone have often been found to be sister to other angiosperms (not an hypothesis that Donoghue and Mathews included), Nymphaeales are sister to the rest, and then Austrobaileyales, the three making up the ANA grade (this used to be called the ANITA grade) (e.g. Mathews & Donoghue 1999, 2000; Qiu et al. 1999, 2001: checked for long-branch attraction - none, 2006b [some analyses], 2007; P. Soltis et al. 1999, 2000; Parkinson et al. 1999; Zanis et al. 2002; Magallón & Sanderson 2002; Kim et al. 2003; Borsch et al. 2003, 2005; Hilu et al. 2003; Nickerson & Drouin 2004; Aoki et al. 2004; P. Soltis & D. Soltis 2004; Müller et al. 2006a; Hansen et al. 2007; Duarte et al. 2008; McCoy et al. 2008; E. K. Lee et al. 2011; Zeng et al. 2014; Drew et al. 2014; Wickett et al. 2014: transcriptome analyses; Sun et al. 2014: chloroplast and nuclear data; etc.). For other studies, see S. W. Graham et al. (2000), Cai et al. (2006), Bausher et al. (2006), Chang et al. (2006), Jansen et al. (2006b, 2007), Ruhlman et al. (2007), Moore et al. (2007, 2011: position not as stable as one might like), Wu et al. (2007), Huang et al. (2010: c.f. rooting), Iles et al. (2014), Zhong and Betancur-R (2017) and Gitzendanner et al. (2018a). These relationships were also recovered by H.-T. Li et al. (2019) in a 2,881 plastid genome analysis that includes all orders and 85% of the families, surely close to the last word in chloroplast-based relationships. In general, the topology of the tree they found, the PPA (plastid phylogenomic angiosperm) tree, is similar to that on which APG IV (2016) was based, and H.-T. Li et al. are mentioned below only when there are differences from the APG IV tree or from family-level relationships that have been accepted here, and also sometimes to emphasize a point (see especially H.-T. Li et al. 2019 Suppl. Fig 3).

On the other hand, Goremykin et al. (2003a), using complete chloroplast sequences, but for only 10 angiosperms, suggested the relationships [[Amborellaceae + Calycanthaceae] [eudicots + monocots]], but poor taxonomic sampling (monocots were represented by Poaceae alone) with resultant long-branch attraction may be responsible for these results (D. Soltis & P. Soltis 2004; Jansen et al. 2004; Stefanovic et al. 2004; Degtjareva et al. 2004b; D. Soltis et al. 2004). Grasses themselves are highly derived monocots (see Kuhl et al. 2004 for the very distinctive genome of Poaceae). Goremykin et al. (2004) found the same general result when adding Nymphaea; it linked with Amborella - which was still not sister to all other angiosperms. Indeed, even when looking at complete chloroplast sequences of just a few flowering plants, the inclusion of Acorus, breaking up the long branch leading to Poaceae, had a major effect (Stefanovic et al. 2004), although questions remained about the models used in the analyses (Lockhart & Penny 2005; Goremykin et al. 2005).

An [Amborellales + Nymphaeales] clade has quite often been recovered. This topology was perhaps particularly prominent in some analyses of mitochondrial genes, as in Qiu et al. (2006b), although there were several unexpected if poorly supported relationships elsewhere in their preferred tree (Qiu et al. 2010; see also Qiu et al. 2000, 2005, 2006a). Sun et al. (2014) also found this topology in most of their mitochondrial analyses, as did Barkman et al. (2000: genes from all three compartments) and Evkaikina et al. (2017: chloroplast genes). Soltis et al. (2007; data from D. Soltis et al. 2000) found that the relationships obtained depended on the method of analysis; Bayesian analysis favoured [Amborellaceae + Nymphaeaceae], while parsimony yielded [Amborellaceae + The Rest] (see also Bausher et al. 2006: ML compared with MP; Mardanov et al. 2008: support weak). Goremykin et al. (2012) suggested that a wrongly specified substitution model would produce a topology [Amborellaceae + The Rest]. What kinds of characters are analysed may be important. Goremykin et al. (2009b) found an [Amborella + Nymphaea] clade after removing a relatively few (500) highly variable positions from the analysis (see also Goremykin et al. 2015), and although criticized by Drew et al. (2014) who found that the characters removed gave the [Amborellaceae + The Rest] topology, studies hint more or less strongly that rate heterogeneity may be an issue (Barkman et al. 2000a; Stefanovic et al. 2004; Leebens-Mack et al. 2005; Finet et al. 2010; Goremykin et al. 2013). In a comprehensive series of analyses, Xi et al. (2014) thought that the problem had to do with the rate of gene evolution, not amount of data. Concatenation results tended to give a clade [Amborellaceae + The Rest], but if fast-evolving sites were removed an [Amborella + Nymphaea] clade was found, as in all their coalescent analyses. In a series of analyses of fairly well sampled transcriptome data, Wickett et al. (2014) found very little support for the [Amborella + Nymphaea] clade whatever the analytical method used. Simmons and Gatesy (2015) returned to the analysis of Xi et al. (2014) and suggested that one of the reasons that their preferred topology was found was that Selaginella, with a very long branch, was used for rooting and some of the coalescent methods they used were susceptible to misrooting caused by its inclusion, and also that the slowly evolving genes in some of their analyses had very little phylogenetic signal. There might be somewhat more support for an [Amborella + Nymphaea] clade in the plastid data, but that suggestion, too, was questioned (Simmons & Gatesy 2015 and references). Edwards et al. (2015) also tended to recover an [Amborellaceae + Nymphaeales] clade, again using sites that were parsimony-informative but evolving only slowly (see alo Xi et al. 2014; Goremykin et al. 2015), but their results were questioned by Simmons (2016); Shen et al. (2017: remove one gene!) evaluate the support for the [Amborellales + Nymphaeales] clade. The [Amborellaceae [Nymphaeales [Austrobaileyales...]]] topology would on balance seem to be the preferred hypothesis (see also L. Zhang et al. 2019; O.T.P.T.I. 2019; Y. Yang et al. 2020: no Austrobaileyales).

Sampling and analytical strategies are critical, the latter particularly in cases like this when there are - and can only be - relatively few taxa, but each can have very large amounts of data (e.g. Jarvis et al. 2014). In some cases large amounts of data may indeed be the solution, in others, perhaps quite surprisingly little data per taxon but improved sampling seems to do the trick (e.g. Rokas et al. 2005; Hedtke et al. 2006). The discovery that Hydatellaceae are sister to other Nymphaeales (Saarela et al. 2006) unexpectedly did allow sampling in this area of the tree to be improved, but its topology was not affected. Fiz-Palacios et al. (2011) suggested a number (25+) of "non-conventional" relationships in their study on land plant diversification, but this, too may be a taxon sampling issue; I rarely mention these relationships here. More data are not always an unmixed blessing, thus Barrett et al. (2012) used whole chloroplast genomes of monocots, but found that adding data did not result in the gradual stabilization of support values, rather, these continued to fluctuate even when there was quite a lot of data and relatively small further amounts of data were added. Finally, some kinds of DNA data may be positively misleading when it comes to understanding relationships (Duvall & Ervin 2004; Qiu et al. 2005; Duvall et al. 2006, 2008b; G. Petersen et al. 2006b). Thus mitochondrial data tend to give odd topologies, and horizontal transfer is notably common in mitochondrial genomes (Sanchez-Puerta et al. 2008, 2011; Hao et al. 2010; W. Wang et al. 2012; c.f. Cusimano et al. 2008; Rice et al. 2013). The merits and demerits of coalescence and concatenation approaches have been much debated, and the former in general seems to have advantages (e.g. Buddenhagen et al. 2016; Mello 2018; L.-N. Zhang et al. 2019; Gonçalves et al. 2019: chloroplasts). Walker et al. (2019) and Gonçalves et al. (2019) critically examine the bahviour of plastid genes, the former group suggesting that the relatively little used rpoC2 gene individually performs best, while the latter i.a. noting particular areas where there were changes from earlier studies. With genome and transcriptome data now being accumulated for considerable numbers of plants (e.g. H.-T. Li et al. 2019: 2,881 plastid genomes, angiosperms; O.T.P.T.I. (2019) and Mandel et al. 2019: ca 1000 nuclear genes, 207 genera, green plants), issues surrounding how best to analyse massive amounts of data become central.

There are convenient summaries of the copious literature on relationships between the major angiosperm clades in e.g. P. Soltis & D. Soltis (2004), D. Soltis et al. (2005b, 2016) and Qiu et al. (2005). For information on broader patterns of relationships, see especially the notes at the mesangiosperm node; for relationships higher up in the tree, see especially monocots, eudicots, Pentapetalae/core eudicots, asterids and euasterids. As work based on analyses of various aspects of nuclear genomes becomes more rotine, in quite a few places it will be clear that there is still substantial uncertainty about relationships. For discussion on the positions of Ceratophyllum and Chloranthaceae, two groups that branch off somewhere in the monocot/eudicot/magnoliid area, see Mesangiospermae.The main tree here is more conservative than that in A.P.G. IV (2016, c.f. in particular O.T.P.T.I. 2019), and in places this latter tree is more conservative than A.P.G. III (2009).

Classification. The classificatory framework, i.e. family and above, for angiosperms follows that of the Angiosperm Phylogeny Group (A.P.G. 1998, II 2003, III 2009, IV 2016); some infra-familial classifications are included, and these are slowly increasing in number. For other embryophytes, I follow what seem to be the best supported hypotheses. The classifications used are phylogenetic classifications, and for the principles that underly it, see above. For full bibliographic information on all names above genus in angiosperms, see Reveal and Chase (2011), Reveal (2012) and especially Reveal's Index nominum supragenericorum plantarum vascularium (2000 onwards); the latter is an invaluable resource that I had hoped would be kept up-to-date, although it seems not to be.

Synonymy: Alismatidae Takhtajan, Arecidae Takhtajan, Aridae Takhtajan, Asteridae Takhtajan, Bromeliidae C. Y. Wu et al., Burmaniidae Heintze, Caycanthidae C. Y. Wu et al., Caryophyllidae Takhtajan, Ceratophyllidae Doweld, Chloranthidae C. Y. Wu et al., Commelinidae Takhtajan, Cornidae Reveal, Dillenidae Reveal & Takhtajan, Ericidae C. Y. Wu et al., Hamamelididae Takhtajan, Iliciidae C. Y. Wu et al., Juncidae Doweld, Lamiidae Reveal, Lauridae C. Y. Wu et al., Liliidae J. H. Schaffner, Loranthidae Tieghem, Malvidae C. Y. Wu et al., Myrtidae J. H. Schaffner, Nelumbonidae Takhtajan, Nymphaeidae Takhtajan, Orchididae Heintze, Piperidae Reveal, Plumbaginidae C. Y. Wu et al., Polygonidae C. Y. Wu et al., Ranunculidae Takhtajan, Rosidae Takhtajan, Rutidae Doweld, Theidae Doweld, Triuridae Doweld, Winteridae Doweld, Zingiberidae Cronquist - Magnoliophytina Reveal - Magnoliophyta Reveal

EVOLUTION AND DIVERSIFICATION OF THE ANGIOSPERMS (This section remains under construction, and always will)

The first three sections below are more introductory, while the others attempt to summarise particular aspects of angiosperm evolution and diversification. Sections 2 and 3 include aspects of the evolution of insects and fungi respectively; their associations with angiosperms have been of great importance for both parties. Some of the issues raised here are taken up later and more from the point of view of the plant. Our knowledge of the evolution of stem-group angiosperms remains rudimentary (section 4). The evolution of flowers and fruits figure prominently in narratives of angiosperm evolution and success (see sections 4, 5A-C, 6A, 7), and they are indeed important, although not in any simple sense. In sections 5D and 5E in particular I discuss some physiological-ecological dimensions to angiosperm evolution, emphasizing aspects that seem to have had a major hand in shaping the global environment over the last 100 Ma or more, and in sections 6B-D I look at other aspects of angiosperm evolution, in particular at the evolution of lowland tropical rain forest and the myriad linkages of plants and animals that today characterize it. In section 8 I turn to asymmetries in evolution, emphasizing relatively small groups of both plants and animals that seem to have had a disproportional (in terms of their species numbers) effect at scales from the global environment to the local community. When thinking of plants in a more ecological context - in section 8B in particular - measures like primary productivity, biomass accumulation, and the like can be used as indicators of importance, and species-rich clades in the gentianids then barely have a walk-on part. In section 9 I attempt a summary. Needless to say, many very important topics are ignored.

1. Important Caveats.

2. Angiosperms and Insects.
2A. Insects, Plants and Herbivory.
2B. Diversification of Phytophagous Insects.
2C. Diversification of Pollinating Insects.

3. Angiosperms and Fungi.
3A. Early Plant-Fungal Relationships.
3B. Mycorrhizae.
      Ecto- and Ericoid mycorrhizae.
      Mucoromycotes, Including Fine Root Endophytes.
3C. Endophytic Fungi.
3D. Mycorrhizae and Endophytes.
3E. Further Complexities.

4. Angiosperm History I: Evolution in Stem Group Angiosperms.
4A. Relationships.
4B. Pollination and Seed Dispersal.

5. Angiosperm History II: Cretaceous (or much earlier?) Origins, Subsequent Cretaceous Diversification.
5A. Introduction.
5B. Early Cretaceous Evolution - to the end of the Albian, ca 113 Ma.
5C. Later Cretaceous Evolution - Albian to Maastrichtian.
5D. Venation Density, Stomatal Size, and Vascular Evolution:
5E. Wood and Litter Decay.

6. Angiosperm History III: Caenozoic Diversification.
6A. The K/P Boundary Event.
      Land Plants.
6B. Flowering Plants.
6C. Latitudinal Gradients of Diversity.
6D. Gene and Genome Duplication and Genome Size.
6E. Diversification of other Plant and Animal groups associated with Flowering Plants.
6F. Discussion.

7. Flowers and Pollination.
7A. Flowers, Pollination and Fertilization.
7B. Major Clades With Monosymmetric Flowers.
7C. Major Clades With Wind-Pollinated Flowers.
7D. A Cautionary Note.

8. Asymmetries in Evolution:
8A. Plant-Animal interactions.
      8A1. Pollination.
      8A2. Seed Dispersal (much more to do here).
8B. Carbon Sequestration.
      Major Players.
      C4 Photosynthesis, and Grasses and Grasslands.
      Ectomycorrhizal Plants.
      Seagrasses, Mangroves, and Tidal Saltmarshes.
      Attempt at a Synthesis.

9. In Conclusion.

1. Important Caveats.

When thinking about evolution in general, a well-supported phylogeny is of course essential, yet there are several parts of the Main Tree as well as of many ordinal trees where we know less than we would like. The huge data sets being developed are likely to yield some unexpected topologies, especially as nuclear and mitochondrial data become more widely used (e.g. Rydin et al. 2017: mitochondrial data cause conniptions in Rubiaceae), indeed, evidence for hybridization is becoming ever more pervasive, with potentially serious effects for analyses based on results from single genomic compartments. Smedmark et al. (2014) emphasized how difficult it was to reconstruct biogeographical events if more than one topology was allowed in places where support was weak, so using a single tree for such reconstructions was inadvisable, yet one still sees trees of different topolgies for the one group used in the discussions of relationships, classification and biogeography, which makes very little sense, and phylogenetic uncertainty similarly compromises our understanding of diversification (Peña & Espeland 2015). Indeed, there are critical issues of dating, working out diversification rates, optimising characters on trees (see elsewhere), etc., that need to be understood, and I discuss some of these briefly below. The primary literature should be consulted for details, and C. S. P. Foster (2016) is a very readable account that covers many of these issues, while Sauquet and Magallón (2018) identify six major questions bearing on angiosperm macroevolution.

1. The relationships of angiosperms to other seed plants, other than a sister group relationships with extant gymnosperms, are still unclear, and thus so are the whens, whys and hows of their initial diversification (see Davies et al. 2004b; Friis et al. 2005, 2011; Frohlich & Chase 2007; Pennisi 2009; Lee et al. 2011; Herendeen et al. 2017; Meyer-Berthaud et al. 2018); Buggs (2017) discusses the historical dimension of this problem. Here we need to distinguish between the origin of the clade of which angiosperms are the only extant representative, i.e. stem angiosperms ("origin 1"), the origin of plants with carpels, tepals, and a heterosporangiate strobilus, i.e. the evolution of plants with flowers ("origin 2"), and finally, the origin of crown-group angiosperms, i.e. extant flowering plants and their immediate common ancestor ("origin 3"). Stem angiosperms presumably are early Carboniferous in age or even older, 350±35-305-275±35 Ma old, if the angiosperm clade is sister to the clade including all living gymnosperms (e.g. Savard et al. 1994; Crane et al. 1995; Crane 1999; Magallón & Castillo 2009; Clarke et al. 2011), perhaps to a younger bound of Permian in age (J. A. Doyle 1998a). Even if crown-group angiosperms are as much as 270 to 182 Ma (S. A. Smith et al. 2010, but c.f. Beaulieu et al. 2015), they will still have a substantial stem history. For the bulk of this some 100 My+ plants along the angiosperm stem probably had naked seeds, lacked flowers, etc. - for further discussion, see below.

Current evidence suggests that extant gymnosperms are monophyletic, but when including fossil taxa extant gymnosperms are paraphyletic with respect to angiosperms - some variant of the anthophyte hypothesis. When integrating extant taxa and fossils in attempts to understand the relationships of angiosperms, for example, morphological analyses may yield rather less information than might be supposed (Puttick et al. 2017; see also O'Reilly et al. 2016). Overall, little progress has been made over the last fifty years or more in identifying plants that can be placed somewhere between angiosperm origins 1 and 3 (E. L. Taylor & Taylor 2009; J. A. Doyle 2012).

2. Estimating ages is critical step, but how best to do this remains a subject of intense discussion (e.g. Magallón and Sanderson 2001; Graur & Martin 2003; Pirie et al. 2005; Renner 2005b; Bell & Donoghue 2005; Magallón & Sanderson 2005; Rutschmann et al. 2007; Sanderson et al. 2004; H. Wang et al. 2009; S. A. Smith et al. 2010; Magallón 2009; Milne 2009: sampling; Burleigh 2012; Sauquet et al. 2012; Magallón et al. 2013; Magallón 2014; Heath et al. 2014; Sytsma et al. 2014; Warnock et al. 2014; Clarke & Boyd 2014; Bell 2015; Schenk 2016: secondary calibrations; Foster et al. 2016a: quite a spread for different methods, increased/more representative sampling needed, also methodological changes, new information from fossils, i.e., just about everything; Matschiner et al. 2016: cichlids get around by l.d.d., not continental drift; Saladin et al. 2017: value of fossils; Landis 2017: dating using palaeogeographic events; Warnock et al. 2017: molecular clock and fossil calibrations; Foster et al. 2017; Foster & Ho 2017: clock partitioning; Salomo et al. 2017: fossil constraints; Barba-Montoya et al. 2018; Guindon 2018). See also Hedges et al. (2006), Kumar et al. (2017) and the TimeTree, also Mello (2018). Dates based on molecular, tectonic, and paleontological data are often in conflict, and the first two often give substantially older ages than the last (c.f. some Nymphaeales; Wilf & Escapa 2014 and subsequent correspondence). In all too many cases there are wildly different estimates for the same event. For instance, compare Wikström et al. (2001, 2004), Clarke et al. (2011) and Z. Wu et al. (2014) for angiosperm ages, Wikström et al. suggest a crown-group age in the Cretaceous, Clarke et al. an age in the Jurassic or earlier, and Wu et al. an age in the mid-Permian. In a study of different ways of calibrating, and two different ways of dating, Sauquet et al. (2011) found that the ranges of the means alone for each node varied by up to a factor of 10 (see also Parham et al. 2011). Relaxed ages are of course often substantially older than constrained ages - for example, the relaxed crown-group age for angiosperms is about 242 Ma, and the constrained age about 130 Ma (Magallón & Castillo 2009). See also Sauquet et al. (2012) for Nothofagus, Crisp and Cook (2011), Martínez et al. (2012) and Condamine et al. (2015) for cycads, Barreda et al. (2010b, 2012a) and Heads (2012) for Asteraceae (and Waters et al. 2013 for a critique of the latter), Crisp et al. (2014) for Asphodelaceae-Xanthorrhoeoideae, and Franzke et al. (2016) for Brassicaceae. Beaulieu et al. (2015) suggested caution for some claims of angiosperm ages while J. W. Brown and Smith (2017/18) outlined another set of problems such that, with articles like these, it becomes difficult to believe much about dating at all - and of course that then affects many other aspects of evolutionary biology. As Sauquet and Magallón (2018: p. 1172) observed of the range of estimates for the age crown-group angiosperms, it was "staggeringly broad" - indeed - while Bromham (2019: p. 484) concluded her review of six questionable assumptions often made when trying to get dates from molecular data by saying "Are molecular dates becoming more reliable? The awkward truth is that, unless we already know the truth, we just don't know.".

Fossil evidence is central to dating. However, fossils are usually more or less incomplete, and their identity, especially that of older fossils, always needs to be confirmed. Indeed, some studies have questioned what had previously seemed to be quite well established fossil identifications (e.g. Cook & Crisp 2005; Nothofagus; Biffin et al. 2010b: Araucariaceae). Recent developments in leaf identifications using a sparse code learning approach may completely overhaul this aspect of the business, although the study in question used cleared leaves, which fossil plants are remiss in not providing (Wilf et al. 2016a). The composition of cuticle waxes of fossils may also help in identifications (Vajda et al. 2017; McElwain 2017: potentially very important). Fossils cannot be expected to be simply "ancestral", rather, they may lack the apomorphies of the crown-group and/or they may have evolved distinctive features of their own, and these may suggest links with unrelated groups in morphological analyses, or the fossils may be assignable to more than one node. Not all parts of the organism fossilize equally easily/well or tell the same story, and in the zoological literature in particular there are attempts to tease apart differences in the signal provided by different kinds of characters - and this also depends on the group in question (Sansom & Wills 2017 and references). But fossils, treated with care, can help in the calibration of molecular trees (e.g. Gandolfo et al. 2004; A. Graham 2010; Clarke et al. 2011; Parham et al. 2011; Ronquist et al. 2012: total evidence; Warnock et al. 2014, esp. 2017; Grimm et al. 2015: comparing Bayesian approaches; Salomo et al. 2017; etc.), and their role here can be quite sobering (Wilf & Escapa 2014; c.f. Q. Wang & Mao 2015, but see Wilf & Escapa 2015); good, well-identified fossils of crown groups give minimum ages (Donoghue & Benton 2007) and under some circumstances can even be included in morphological/molecular analyses with extant taxa (Pyron et al. 2011), even if the results of such analyses should be treated with great care. However, if referred to a stem group in particular their significance is ambiguous since stem groups can persist after the evolution of the relevant crown group. A database for reliably-dated fossils has recently been developed (Ksepa et al. 2015, see also Xing et al. 2016 for other fossil databases). The fossil record is sure to have surprises (indeed - see Barreda et al. 2015; Wilf et al. 2017a), although reports of Jurassic flowers (e.g. Z.-J. Liu et al. 2015; Han et al. 2016 and references) need confirmation, indeed, Herendeen et al. (2017) critically review some of them, and none passes the muster (see also ), while Schönenberger et al. (2020) describe the development of an angiosperm-wide data base that allows the suggested positions of fossil angiosperms to be tested against molecular backbones, although the latter leave a little to be desired. Sauquet and Magallón (2018) offer a number of suggestions for a "fossil revolution", including the description of fossils whose relationships are unclear (so making their odd morphologies generally accessable), the use of total evidence approaches to dating, etc..

Poaceae, q.v. for more details, present particular problems. Fossils identified as Poaceae-Poöideae (Poinar 2004, 2011) from Cretaceous amber from Myanmar are dated to ca 98.8 Ma (Shi et al. 2012). Such dates would imply that many other dates for flowering plant clades are too young and cause a general rethinking of angiosperm evolution. But yet more: “With the discovery of a Claviceps-like fossil infecting a floret of an Early–Mid Cretaceous Asian grass, we propose that the progenitor of Claviceps evolved in Asia among early grasses sometime in the Mid- to Late Jurassic” (Poinar et al. 2015: p. 17), and similarly, it has been suggested that the stem-group age of Simplicia, a member of Poeae endemic to New Zealand, was late Jurassic-Cretaceous, i.e. around 140 Ma (Heads 2018c), so crown-group angiosperms would probably be Triassic. Even the ca 66 Ma age of crown Oryzoideae phytoliths from India (e.g. Prasad et al. 2011, accepted by Iles et al. 2015) will have a substantial knock-on effect on other age estimates, and hence on diversification rates, biogeography, etc. (Christin et al. 2014).

A clade restricted to an island would seem to have to be younger than that island, but there are several examples suggesting the contrary in these pages (see also Heads 2011, 2018a: metapopulation vicariance; c.f. Franzke et al. 2016); perhaps the most extreme example is Mankyua (Ophioglossaceae), a clade about 194.8 M years old, growing on Jejudo Island, ca 2 Ma, the only place from which it is known (Gil & Kim 2018). Island ages cannot be used as a maximum age constraint for a clade without there being supporting evidence, indeed, there is much discussion about the geological history of islands like New Zealand and the ages of clades on those islands (e.g. Knapp et al. 2005; Cook & Crisp 2005; He et al. 2016a; Wallis & Jorge 2018; McCulloch & Waters 2019); clade ages on New Caledonia have also occasioned some discussion (e.g. Condamine et al. 2016; He et al. 2016a; Grandcolas 2017; Nattier et al. 2017; Heads 2018a, b; Wallis & Jorge 2018). Along the same lines, even if there were no other problems with dating, ages of clades that are being associated with particular dispersal or vicariant events may be overestimates because of patterns of extinction (Barraclough 2010: Fig. 4). Vicariance aside, the suggested age of a clade that is thought to have moved to Madagascar from Africa (for example) will be a maximum age since formation of the clade and its dispersal are not necessarily contemporaneous events. In particular, if all the evidence one has for for the age of a species/clade on an island (or anywhere) is the age of its common ancestor with a sister taxon (e.g. Hillebrandia sandwicensis/Begonia, although one could argue that we know a little more in this particular case), then that age is indeed a maximum age for the species/clade, but of itself conveys precisely zero information as to how long it has been where it is (see also Ho et al. 2015; Swenson et al. 2019). (Note that the argument here is similar to that in using sister-group comparisons to estimate the effect of the acquisition of a particular feature on subsequent diversification - see Käfer & Mousset 2014 and 7, Diversification below.)

Many ages for clades are given on these pages, and dates from older literature, not all mentioned here, are conveniently assembled in Hedges and Kumar (2009). However, it should be abdundantly clear that all dates should be treated with extreme caution, since a very large number of dates in the literature, even if recently published, must be more or less seriously wrong - or, if with a large standard error (for example), of little real use. The original literature should be consulted for details of methods used, the actual node to which the date refers, i.e. the node immediately subtending the species in the family (for example) that have been sampled (I have tried to be accurate), the range of dates suggested, and the topology of the tree being dated. Here I refer to crown-group ages for the most part (but c.f. ages for taxa on islands above), although stem ages can be worked out by looking at the crown age of the next node down; in a number of taxa there is a prolonged period between origination and diversification, the so-called phylogenetic fuse (Cooper and Fortey 1998).

3. Distributions are not easy to interpret. First of all, there is ever-increasing evidence that the present and past distributions of many plant and animal groups are very different. Thus early in the Caenozoic the distributions of a number of tropical taxa like Nypa and Cyclanthaceae that are today rather restricted were much wider (e.g. Plaziat et al 2001; S. Y. Smith et al. 2008), while in the Oligocene the ancestors of hoatzins, hummingbirds and parrots were flying around in Europe (Mayr 2002, 2004, 2009). Genera and families continue to be added to the list of groups in which past and present distributions are very different (e.g. Friis et al. 2011: numerous examples; Stull et al. 2012; Manchester et al. 2012; Grímsson et al. 2013; Hofmann et al. 2015; Sadowski et al. 2015; Grimsson et al. 2017b). Many taxa now restricted to Southeast Asia grew in Europe and North America in the Caenozoic (e.g. Ferguson et al. 1997; Manchester et al. 2009) and many currently Australasian taxa are found in Eocene forests of Patagonia (Merkhofer et al. 2015), although such changes in longitudinal distributions are easier to understand.

Secondly, as with estimating ages, pattern→process arguments are dangerous, and independently-derived ages are essential when interpreting distribution patterns. Patterns that that seemed to reflect vicariance caused by plate tectonic events may be better explained by more recent dispersal/migration events (e.g. Sanmártin and Ronquist (2004), Renner 2005b; de Queiroz 2005, esp. 2014; Wen & Ickert-Bond 2009: summaries, also Higgins et al. 2003; Nathan 2006; Yoder & Nowak 2006; Carpenter et al. 2010b; Gillespie et al. 2012a; Baker & Couvreur 2012a, b; Christenhusz & Chase 2012; Hauenschild et al. 2018b). Even Lars Brundin's hitherto iconic chironimid midge drift-determined distributions may need reinterpretation from this point of view (Krosch et al. 2011). Pure vicariance ages, i.e. ages based largely on distribution patterns alone, tend to be rather different from those suggested for other reasons (e.g. c.f. Heads 2014 and de Queiroz 2017; see also Grehan 2017; Heads 2018b), and metapopulation vicariance (e.g. Heads 2017, 2018a) also makes thinking about ages complicated. However, if the ages for some crucial fossils are upheld, vicariance may have to be revisited (see in part Ladiges & Cantrill 2007; Heads 2008, etc.). Thus Wilf and Escapa (2014) questioned the ages of a number of Patagonian fossils and hence the dispersal-type explanations that were based on these ages (see also Wilf et al. 2013, 2016; Barrera et al. 2015).

A vicariance approach to biogeography downplays the importance of long distance dispersal (LDD) events in achieving plant ranges, however, Nathan et al. (2008; see also Nathan 2006) suggested situations in which LDD might be promoted, while Viana et al. (2016 and references) examined the gut contents of birds cached by Eleonora's falcon on Alagranza (the Canaries), and this allowed them to suggest dispersal distances of propagules in those birds of a minimum of 80 or 170 km, depending on the species of propagule recovered (see also Ridley 1930; Berg 1983). If LDD is suspected, the modes of diaspore dispersal become of considerable interest, although it is difficult to know what went on in any particular instance given that such events are very unlikely to be observed directly. For recent discussions about bipolar and amphitropical distributions and their causes, see e.g. Guilliams et al. (2017), Schenk and Saunders (2017), Johnson and Porter (2017) and Villaverde et al. (2017). For more on how organisms achieve the ranges that they have, see e.g. Van der Pijl (1982), Gehrke and Linder (2009), Schurr et al. (2009), Tamme et al. (2014), Chacón et al. (2017) and also the individual family accounts.

4. The apparently simple issue of species numbers is in fact not that simple even when discussing the size of extant clades. There are two aspects to this - what is really the clade of interest?, and, how many species does it contain?

A. I take it as axiomatic that comparisons between taxa simply because they have the same hierarchical rank are ill-advised, putting it mildly. Simplistic "major clade"-type comparisons are of little value (e.g. S. A. Smith et al. 2011; Igea et al. 2015 - but see Ricotta et al. 2012 and Laenen et al. 2014: Fig. 1), and that all Poaceae, for example, are assigned to tribes (see below) does not thereby make comparisons between these tribes of any biological value or interest. There are many examples of extreme clade size imbalance throughout the tree in which categorical ranks are less than informative about evolution and diversification. Thus an emphasis on genera and genus size (e.g. Frodin 2004) is inappropriate; the proper units of comparison for a genus are its sister taxon and/or other clades of the same age, not other genera (c.f. Givnish 2016: Brocchinia is to be compared with all other Bromeliaceae) - although I do mention genus size on occasion, it is simply the size of a clade for which there is an accepted name at generic rank. Adjustments may have to be made if one is thinking about characters and diversification since the ages of clades and characters associated with that clade may differ (Käfer & Mousset 2014: see below).

Orchidaceae, often considered to be very diverse - in terms of numbers of species - when compared with other families, are a good example of the problems faced. Since Orchidaceae are sister to all other Asparagales, the disparity in species number, although considerable, is only four-fold (ca 27,800 vs. 7,100), furthermore, Asparagales as a whole, with ca 34,900 species, are sister to commelinids, with some 24,500 species. Within Orchidaceae the highly speciose and mostly epiphytic-CAM crown Epidendroideae include around 19,560 species (figures from Pridgeon et al. 2005, 2009, 2014), i.e., about two thirds of the species in the whole family (e.g. Gravendeel et al. 2004). So answers to the questions, "Are orchids particularly diverse, and if so, why?", are not straightforward. Perhaps Epidendroideae, or a clade within them, are the hyperdiverse group (see below for another example).

B. Estimates of the number of extant species of flowering plants vary by a factor of about two - 422,127 (Govaerts 2001) to 223,300 (Scotland & Wortley 2003) - and perhaps add 20% (Joppa et al. 2010); when necessary, I use an intermediate figure of 352,000 species of flowering plants (see Paton et al. 2008); c.f. Nic Lughaddha et al. (2016) for comments on this and other estimates. However, new estimates come out every other year or so, e.g. Christenhusz and Byng (2016), where, at 295,383 species, the number is lower, but using their predictions of the rate of description of new species they will have the same number as Paton et al. around about 2045 AD... Interestingly, links to the numbers in Christenhusz and Byng were immediately posted on Wikipedia, however, a still more recent estimate is, at 369,434 species, very different (Nic Lughadha et al. 2016). In some groups uncertainty over species numbers is particularly great, thus estimates of species numbers in groups that one would have thought were well known vary widely, two examples being Narcissus (estimates in the last five years range from 15-81 species: Marques et al. 2017) and Ophrys (16-252 species: e.g. Bateman et al. 2011a; Vereecken et al. 2011; Alibertis 2015: photographs!, see also below). Species numbers increasingly depend on the kinds of data collected and how they are analysed and evaluated, and this is a very active area of research. Furthermore, as Mallet (2013: p. 690) noted, "species counts over large areas of space and time [i.e., those in these pages] represent only a sketchy measure of biodiversity, a measure that owes more to taxonomic and metaphysical fashion than to science", and be grateful that land plants represent a clade that is relatively small and stable in numbers in the great scheme of things (c.f. Mahé et al. 2017; Larsen et al. 2017: 1,000,000-7,000,000< species overall...). Treat species (and genus) numbers below with a grain of salt; where estimates of species numbers in Christenhusz and Byng (2016) differ from those given here by more than 15% or so, the former are included in parentheses.

5. Many characters seem to come and go on the tree almost willy-nilly, so making character optimisation a distinctly hazardous undertaking. Thus using either Bayesian methods or maximum likelihood, making apparently reasonable assumptions about the relative weighting of gains versus losses, or just using the simple models of evolution explicit in ACCTRAN or DELTRAN often affects the position of synapomorphies on trees, and hence our ideas of evolution (e.g. Donoghue & Ackerley 1996; Cunningham et al. 1998; Omland 1997, 1999; Ree & Donoghue 1999; Polly 2001; Webster & Purvis 2001; Ronquist 2004; Crisp & Cook 2005; Sannier et al. 2007, 2011; Remizowa et al. 2010b; Cohen 2012; O'Meara 2012; Sokoloff et al. 2013d; Sundue & Rothfels 2013; Gascuel & Steel 2014; Kriebel et al. 2014; Wright & Hillis 2014; Wortley et al. 2015; Sauquet et al. 2017; Parins-Fukuchi 2017; Stull et al. 2018). Although genomic data are becoming more extensive, resolving relationships in some areas nevertheless remains very difficult; hard polytomies are evident. In such situations thinking about character evolution is very tricky (an understatement); Koenen et al. (2019) grasp this particular bull firmly by the horns. And even when a character is confidently associated with a node, then understanding when it evolved is difficult - does one use stem age or crown age (e.g. Pausas et al. 2018)? In sister-group comparisons to see if features like latex acquisition affect diversity, the same question arises (but see below under 7, Diversification). Syme and Oakley (2012) suggest that tree-based and node-based methods give very different results when it comes to allowing reversals. And of course the very definitions and circumscriptions of the features of interest are all too often problematical. There is further discussion on this very important set of issues in the Introduction.

6. Understanding the palaeoecological context of the evolution of angiosperms is a challenge. Ecological contexts change over time, and that of the early Caenozoic diversification of angiosperms is likely to be quite different from those of the origins of stem- and crown-group angiosperms; the past and the present are not immediately comparable. The Caenozoic context is initially connected with the bolide impact at the K/P boundary and the eruptions that produced the massive Deccan Traps, although much of angiosperm diversity - and that of the animals associated with it - as we now appreciate it seems to be a phenomenon of the later Caenozoic, and again the ecological context has changed; Meseguer et al. (2014b) attempt to deal with the problem. As we go back in the past, the proportion of no-analogue communities, i.e. those that we cannot understand by extrapolating from the present to the past, increases.

Palaeontologists face this problem on a daily basis. Thus Rothwell et al. (2000) reconstructed the palaeoecology of the small, probably short-lived conifer Aethophyllum using a combination of evidence from the fossil, the palaeoenvironment, etc. (see also e.g. Strömberg 2006). Although it is tempting to read the ecology of early angiosperms from that of extant taxa of the ANA grade, this is hazardous (e.g. Wheeler & Baas 1991, 1993; Uhl 2006; Philippe et al. 2008). Little et al. (2010) challenged the reliability of the use of aspects of leaf morphology, especially the presence of teeth, as palaeoclimatic indicators (for which, see e.g. Wolfe 1978). However, Royer and Wilf (2006; see also Zohner et al. 2019) suggest there maybe a connection between teeth and temperature, teeth enabling photosynthesis the more the growing season was short. In any event, since the immediate relatives of angiosperms are unclear, working out how the ancestral crown angiosperm might have functioned will for now have to be a top-down process (Feild & Arens 2005, 2007), even if the present is an imperfect key to the past.

7. With time, the tree, distributions, apomorphies, and numbers of species, we can begin to think about "diversification". Although mentioned frequently below, diversification and the related "adaptive radiation", "key innovation" and "success" are both imprecise and difficult to estimate and to interpret (e.g. Heard & Hauser 1995; Bengtsson 1998; Hunter 1998; Sanderson 1998; Davies et al. 2004; Ricklefs 2007; Olson & Arroyo-Santos 2009; Ackerly 2009; Wertheim & Sanderson 2010; Yoder et al. 2010b; Stadler 2011a, b; Drummond et al. 2012; Givnish 2015b; Tucker et al. 2016: more ecological, c.f. richness, divergence, regularity; Rabosky 2017a; Jantzen et al. 2019; Avolio et al. 2019, etc.). There are at least three major issues here, the first two of which have already been mentioned: A, how to make estimates of species numbers and how these numbers change over time, B, linking character changes at a node to other changes, and thinking about the causal role of such change(s) in aspects of the subsequent history of that clade, and C, how to think about diversification - is it a matter of species numbers, and/or changes in ecological roles/moving into a new adaptive zone, and/or changes in such features as dominance, biomass production or net primary productivity of clades?

A very important literature deals with the proper use of phylogenetic comparative methods that allow us to make reasonable suggestions that a particular organismal/intrinsic or extrinsic feature can usefully be linked to the subsequent history of the clade it characterises (Maddison & FitzJohn 2014; Beaulieu & O'Meara 2016; Uyeda et al. 2017/2018 and references). In this respect a potential "key innovation" is like any other character: How is it to be linked to a particular node?

In curves showing diversity in clades over time, what can seem like an abrupt radiation, with rapid diversification after a period when there was little apparent diversity - the "broom and handle" and "stemmy" patterns evident in many clades - may in fact be the result of extinction, diversification after the extinction event resuming at a rate similar to that before the event (e.g. Crisp & Cook 2009). Extensive sampling (>80%) may be needed if accurate estimates of slowdowns in diversification are to be made (Cusimano & Renner 2010), extinction is hard to estimate (Rabosky 2010c), and diversification rates will automatically tend to increase towards the present. Chronograms may suggest different interpretations of phylodiversity than phylograms (Elliott et al. 2018; Jantzen et al. 2019). Simple experiments estimating future extinctions showed that these might affect estimates of clade size imbalance at nodes of up to ca 50 Ma (Clarke et al. 2011). In general, estimating clade size imbalance and the rate of change of clade size over time are remarkably tricky operations, especially in the near absence of fossils, the usual situation (e.g. Magallón & Sanderson 2001; Tarver & Donoghue 2011; Rabosky 2010a, b, c.f. Meyer & Wiens 2017, then c.f. Rabosky 2017b...; Morlon 2014: review; Marshall 2017: v. useful cautionary notes). Even when there is an excellent fossil record, strategies like removing recently-radiating clades may be needed if one is to detect diversity loss in other clades (Morlon et al. 2011: whales, etc.; see also Stadler 2011b).

The use of sister taxa when estimating the effect of putatively important features on diversification, an apparently straightforward operation (e.g. Sargent 2004), in fact needs some care, since if the feature of interest is apomorphic then the clade with that feature is ceteris paribus likely to be smaller than its sister clade - it first had to acquire the feature of interest, and so it must have had less time for diversification (Käfer & Mousset 2014; Lamont et al. 2018b - see also 2, Age above). In any case, looking for increased diversification rates may not be the best way to think about the acquisition of a key innovation (Rabosky 2017a). Note that both speciation and extinction rates may depend on the age of the group being examined - the younger the group, the faster the rates (Henao Diaz et al. 2018); if this relationship is confirmed, much literature will have to be re-examined. Sun et al. (2019) provide a critique of methods used to ascertain diversification rates, focussing on the rosids, and they found i.a. that common procedures used in phylogenetic analyses like representative (as compared to random) sampling, etc., were problematic.

Having found features of interest in the diversification of a clade and worked out changes in clade sizes over time, associating environmental/extrinsic features with these changes is not easy, although Bouchenak-Khelladi et al. (2015; see also Beaulieu & Donoghue 2013; Donoghue & Sanderson 2015; Givnish 2015b, and others) make the attempt. And when they are connected, is the diversity at that node to be labelled as adaptive radiation, with members of a clade "doing" different things, more an ecological concept (e.g. Simpson 1953; Stroud & Losos 2016), or diversification, more a species number issue? Other measures such as morphological disparity (Minelli 2016; Oyston et al. 2016; Nürk et al. 2019), dominance, biomass production and net primary productivity can all be evaluated at a variety of phylogenetic and ecological scales. Thus Avolio et al. (2019) discuss some of the different uses of "dominant", they prefer a definition that includes both commonness and the effect of the species on community and ecosystem; see also below.

8. Finally, angiosperms and the organisms with which they associate form complex symbiotic systems at a variety of levels. Angiosperm physiology is mediated by fungi and bacteria associated with the plant growing both in the surrounding soil and in the plant itself, and this has shaped and continues to shape the environment at all scales. Most plants are mycorrhizal, and they also have a variety of often poorly understood fungal and bacterial endophytes. These may affect the growth of the plant, whether by fixing nitrogen, detering pathogens, improving growth rate, or the like (Kembel et al. 2014; Weiß et al. 2016 for references). Features ascribed to plants may be the result of interactions between plants, fungi, and/or bacteria (Friesen et al. 2011), and this goes far beyond the ancient endosymbiotic events that resulted in chloroplasts and mitochondria. An individual plant is a microcosm or some kind of complex chimaera, as Herre et al. (2005) noted, referring to tropical plants and their endophytic fungi in particular (see below), and this idea is elaborated by, for instance, Bordenstein and Theis (2015), F. M. Martin et al. (2017) and Tripp et al. (2017a).

2. Angiosperms and Insects.

Direct associations between plants and insects, whether the latter are parasites, herbivores, detritivores, gall-formers, seed-dispersers or pollinators, are ubiquitous. The diversification of angiosperms is broadly contemporaneous with the massive diversification of many insect groups that are now more or less dependent on them, or the two are co-dependent, although there is some argument as to just how close the linkages are (see also above). Ehrlich and Raven (1964; see also Brues 1924; Fraenkel 1959) provide an early statement of the idea of co-evolution that centred on the relationship between angiosperms and the insects that eat them. Although no clear co-evolutionary mechanism was articulated there, Marquis et al. (2016) attempt to link herbivory, changes in secondary metabolites, and speciation, ideas that have been further developed by Maron et al. (2018) in a critique of this paper (see also Janzen 1980; Schemske 1983; Futuyma 2000; Brouat et al. 2001; Futuyma & Agrawal 2009; Kato et al. 2010; Fordyce 2010; Janz 2011; de Vienne 2013). Co-evolution has now come to include anything from situations in which changes in the two members of the association are not directly linked, to those in which changes in one member of the association is loosely linked to changes in the other (cospeciation need not be involved), to cospeciation where speciation of one member of an association effectively entails speciation of the other. However, strictly reciprocal evolutionary change and diversification of co-evolving plant and insect or other animal groups is at best uncommon (Suchan & Alvarez 2015), even in Ficus (Satler et al. 2019), and usually involves vertical inheritance of parasites/endophytes (de Vienne et al. 2013), although there is quite close co-evolution in some examples of herbivory and parasitism (e.g. Winkler & Mitter 2008; Althoff et al. 2012; Endara et al. 2018). Some possible connections between plant and animal diversification are discussed in the individual order pages, also in the discussion of the Cretaceous and Caenozoic diversification of angiosperms, the role of some pollinators, etc.. Host switching is often associated with radiation of a herbivore on a new host, the "escape and radiate" hypothesis (e.g. Ehrlich & Raven 1964; Fordyce 2010), and such radiation is quite common (de Vienne et al. 2013). For radiation of insects after plant radiations, see e.g. Stirman et al. (2010), Leppänen et al. (2012); dating the diversification of both partners is critical (but see above).

2A. Insects, Plants and Herbivory. Some details of plant-insect relationships are mentioned after individual orders and families. Most insects eat only at most a few species of plants, generalist herbivores being relatively uncommon (Forister et al. 2010; Boulain et al. 2018). What attracts an egg-depositing insect to one plant and prevents it laying eggs on another is often some aspect of plant chemistry that is detected by the insect (see e.g. Bernays & Chapman 1995 and Fernandez & Hilker 2007: Chrysomelidae; Suzuki et al. 2018: butterflies and gustatory receptor genes), furthermore, plants have evolved a variety of mechanical defences against herbivory. In general, more related plants have more similar animals eating them (Weiblen et al. 2006; Futuyma & Agrawal 2009 for literature), simply because they will tend to taste similar, having similar secondary metabolites. However, related plants may well show greater than expected diversity of traits involved in defence as they try to escape the herbivores eating their congeners growing in the same area (e.g. Becerra 2007; Becerra et al. 2009; Kursar et al. 2009); for the coexistence of members of plant species groups, herbivory, and interspecific diversity in secondary metabolites, see Sedio et al. (2017).

Plant tissues are for the most part rather nutrient-poor, and plant cell walls are made up of cellulose, rather indigestible, and lignin, still more indigestible. Lignin and cellulose digestion in termites occurs via their association with protozoa or the fungi they cultivate (Ni & Tokuda 2013), while some other insects and other arthropods are able to break down cellulose walls independent of any mutualistic association with micro-organisms, and this has implications for the evolution of land plants and their associated insects (Calderón-Cortés et al. 2012 for literature). In other cases, lignin and cellulose yield little of value to the insect. Yet although plants may not seem to offer much to insects and other herbivores, herbivory is often very extensive. As a result, the plant may commit a major part of its resources to defense. Thus phenolic defences in the young leaves of two species of Inga (Fabaceae) were ca 30% dry weight, or around 50% dry weight flavonols when a tissue-bound element was included (Lokvam & Kursar 2005; see also Salazar et al. 2018). Protieae with higher diversity in metabolites that affected herbivores, either positively or negatively, committed less to defence, and herbivore species richness was negatively correlated with metabolite richness. Those metabolites that reduced herbivory were more conserved across the plant phylogeny, but there was no particular correlation between the metabolite composition of the plant and herbivore phylogeny, as might be expected for a system such as this where the herbivores are largely generalists (Salazar et al. 2018). However, it should not be forgotten that it is not easy to measure how an insect affects the plant, since ultimately one major negative effect will be on the fitness of the plant, and this is rarely measured directly (Erb 2018).

p style="text-indent: 30px;">Protective secondary metabolites may be found in latex, or they may be translocated via the vascular tissue, or there may be other specialised tissues involved. Herbivorous insects that eat plants with such defences may show distinctive trenching or vein-cutting behaviours which stop the supply of any protectants to the plant tissue distal to the trench or cut and enable the insect to eat it, although such behaviours are also seen on plants like Passiflora and Pelargonium not noted for their latex, although the latter at least has glandular hairs (see e.g. Dussourd & Eisner 1987; McCloud et al. 1995; Becerra et al. 2001; Dussourd 2009). As Dussourd (2016) suggests, what we know about trenching behaviour is probably just the tip of the iceberg.

But protective secondary metabolites may become something of a double-edged sword. Some insects eat only plants with particular defences that they will coopt for their own defence (e.g. Termonia et al. 2001 for chrysomelid leaf beetles). Herbivorous insects may sequester secondary metabolites from the plant in the larva and/or adult stages, ensuring some measure of protection by so doing; they often have warning colouration, i.e., they are aposematic (Bowers 1993 for aposematic caterpillars). They may also use plant metabolites for pheromones to attract mates, or these metabolites may simply act as oviposition cues (Brower & Brower 1964 on butterflies; Nishida 2002 for a review). Secondary metabolites quite often benefit herbivores, especially generalists but also specialists (Smilanich et al. 2016), and although some may protect against herbivores, others attract/are necessary for herbivores (e.g., glucosinolates in Brassicales: specialists are attracted). Some herbivorous insects effectively track plant secondary metabolites and are found on whatever plant has a particular metabolite, independent of the phylogeny of the plant groups concerned (e.g. Winkler et al. 2009); glucosinolates and some alkaloids are examples. Glucosinolates are found in both Putranjivaceae (Malpighiales) and Brassicales, as are the pierid butterflies that are attracted to glucosinolates, while swallowtail butterflies (Papilionidae-Papilioninae) are found on Rutaceae and Lauraceae, the two having similar alkaloids, and on Rutaceae and Apiaceae, which both have furanocoumarins (Berenbaum 2001).

Within herbivores, there is a general decrease in host specificity both in temperate and tropical regions that follows the sequence: granivores > leaf miners > fructivores > leaf chewers = sap suckers > wood eaters > root feeders (Novotny & Basset 2005). Specialization in weevil-plant associations is similar: fruit and seed > wood > root and stem eaters (McKenna et al. 2009), while Novotny et al. (2010: around 24 guilds listed) found specialisation greatest in four guilds, leaf suckers, larval leaf chewers, leaf miners, and fruit chewers, guilds like root and phloem chewers showing less specificty. How insect larvae feed, i.e., whether they are internal feeders like stem borers and whether they can tolerate raphides, or latex, etc., may be more conserved than associations between larvae and particular groups of plants or other types of feeding behaviours (e.g. Powell 1980; Peigler 1986; Powell et al. 1999: associations with latex-containing plants; Konno 2011: chemistry; Farrell & Sequiera 2001; Lopez-Vaamonde et al. 2003, 2006). Phylogenetic conservatism may be greater in groups in which the adults tend to remain close to plants in/on which they grew up, as with beetles, compared to the situation where the adult may fly away, as in many lepidoptera (Berenbaum & Passoa 1999).

Extant angiosperms show a correlation between woodiness and tannin frequency and a negative correlation between tannins (generalized defence) and alkaloids and other secondary metabolites (specific defence). Plants that were obvious or apparent to herbivores were longer lived and had large amounts of generalized defences, while less apparent plants were more short-lived and had lower amounts of more specific defences (e.g. Feeney 1976; Silvertown & Dodd 1996; see also Levin 1976; Mole 1993; Endara & Coley 2010). Apparent plants were late successional, non-apparent plants early successional, although herbs in general came to be thought of a non-apparent (see Smilanich et al. 2016). (Insects that were specialized on their hosts ate the first group, other insects the second, and also the first - Endara & Coley 2010.) "Quantitative" defences like polyphenolics are more generalised, and butterflies such as Lycaenidae are the herbivores, while "qualitative" defences are highly toxic and butterflies like Nymphalidae are specialized herbivores (Fiedler 1996). Consistent with the apparency theory, the level of herbivory in woody plants is higher than that in herbivorous plants (Turcotte et al. 2014: see caveats about analyses).

The nature and amount of the defensive compounds produced can also be explained, and perhaps more satisfactorily, by the resource availability hypothesis, in which herbivore defence is thought of from a cost:benefit point of view (Endara & Coley 2010 for a summary). In particular, the growth rate of the plant affects the nature and amount of defences. Fast-growing plants need less in the way of defence since their leaves are short-lived and would soon be replaced even if they had not been eaten (Endara & Coley 2010). Deciduous plants in general, with their rather thin leaves, will tend to be eaten by insects more than plants with long-lived xeromorphic leaves (e.g. Coley & Barone 1996; Arnold et al. 2001; Wilf et al. 2001; Lewinsohn et al. 2005). When there are low concentrations of available nutrients, growth is slow, the leaves are long lived, and defences are laid down (Coley et al. 1985; Endara & Coley 2010). Along the same line, herbaceous light-demanding taxa are more often attacked by biotrophic fungi, fungi needing living plant tissue (therefore probably not loaded with toxic substances) to prosper, whereas woody taxa, and especially those growing in shaded conditions, were more likely to be infected by necrotrophic fungi, fungi which first kill plant tissue before digesting it (García-Guzmán & Heil 2013). Ali and Agrawal (2013) summarized the curent state of knowledge about the generalist/specialist paradigm, noting how little was really known about details of herbivore:plant interactions. Indeed, condensed tannins can be found in massive amounts in leaves and are sometimes thought to be an antiherbivore device - except that it has been suggested that they may not affect the activities of herbivores much, rather, they increase the tolerance of plants to herbivory, facilitating plant recovery after herbivore damage by increasing nitrogen uptake from litter, frass, etc. (Madritch & Lindroth 2015; see Barbehenn & Constabel 2011 for a review of a difficult literature). Although tannins may be involved in protein precipitation in an insect's gut, the oxidative capacity of tannins in more alkaline portions of the gut, perhaps most evident in some ellagitannins, may also affect the animal (Salminen & Karonen 2011).

There is much discussion about possible correlations between geography, species richness, herbivory and defensive metabolites, and some geographical dimensions of these correlations are discussed later. Menken et al. (2009) suggest a correlation between species numbers, host specificity and feeding habits in Lepidoptera in particular. Internal plant feeders tend to be small, have higher host specificity, and show relatively little diversification, while external feeders tended to be larger, show less host specificity, and formed more speciose clades. Nylin et al. (2014) thought that transient polyphagy in nymphalid butterflies facilitated diversification and subsequent host-plant shifts (see also Ehrlich & Raven 1964; Marquis et al. 2016), however, recent work suggests that in some butterflies, at least, there is no simple connection between host-plant shifts and diversification (Hamm & Fordyce 2015). The plesiomorphic condition in Lepidoptera is small size and internal feeding such as burrowing, and these are features of caterpillars of the basal lepidopteran clades, the adults of which have jaws, and those of basal Glossata (the adults have probosces) such as Eriocraniidae (Menken et al. 2009; Imada et al. 2011) - overall the change in host plant preferences has been from specialist to generalist (Menken et al. 2009).

Dipteran cecidomyiid gall midges diversified in the later Cretaceous some 100-80 Ma, and crown ages for the major groups are Tertiary (Dorchin et al. 2019: fig. 4). In some cases, diversification of plants can be linked to the development of particular defences, but this does not happen in any simple fashion; the mechanism by which insect diversification increases when feeding on angiosperms is also unclear (Janz 2011). However, a recent idea is to apply ideas from island biogeography to the problem. With cecidomyiid gallers, at least, if host species are closer, insect diversity may increase because it is easier to switch hosts, and if ranges of insects are large and the plants are structurally complex, diversity also increases; clade age has little to do with it (Joy & Crespi 2012; for age, c.f. Brändle & Brandl 2002 [in part]; Farrell & Mitter 1994).

To summarize: The impact of insect/plant associations on plant diversification is still poorly understood (e.g. Futuyma and Agrawal 2009: also other papers in Proc. National Acad. Sci. 106(43)). In addition, mutualists, particularly bacteria or viruses, associated with the insect, or parasites of the insect, or even viruses of parasites of the insect, may all affect interactions between the insect and plant, and sometimes in surprising ways (Poelman et al. 2011; Frago et al. 2012; Zhu et al. 2014, 2018; Tan et al. 2018). We need to know more about both the timing of diversification and patterns of phylogenetic relationships in both insect and plant, and evidence for the former in particular is often lacking or unclear (de Vienne et al. 2013), worse, there can be uncertainty in both areas; this is discussed further below.

2B. Diversification of Phytophagous Insects. There is still considerable uncertainty about overall arthropod numbers, although they are likely to be substantially below the 20,000,000 or so once suggested; current estimates are between 1.6 and 7.4 million (e.g. Hamilton et al. 2013 and references, but see Forbes et al. 2018). Phytophagous insects make up about one quarter of all described species, including over half the beetles (Janz et al. 2006: over half; Farrell 1998: ca one third of beetles; Hunt et al. 2007; J. Wu and Baldwin 2010: nearly half of almost 1,000,000 species; Wiens et al. 2015), nearly all Lepidoptera, etc., although there are also very species-rich beetle clades that are neither herbivores nor decomposers (e.g. Barraclough et al. 1998). There may be around (0.9-)1.5(-2.1) million beetles alone (Stork et al. 2015); for a comprehensive phylogeny, see Hunt et al. (2007). However, there may be (many) more hymenoptera than beetles, largely because of the diversity of parasitoids in Parasitica that parasitize beetles and other insect groups, including galling insects; they are quite often specific as to their hosts, and are themselves subject to hyperparasitism by yet more wasps (Smith et al. 2008; especially Forbes et al. 2018). Clades of phytophagous insects may be more speciose that their non-phytophagous sister groups, ectophagous clades more diverse than their endophagous sister taxa, and clades that eat angiosperms more speciose compared to those that eat other plants (Mitter et al. 1988; Winkler & Mitter 2008; Wiens et al. 2015), although in the first case, at least, the asymmetry may not be significant in beetles (Hunt et al. 2007; Wiens et al. 2015) - indeed, Rainford and Mayhew (2015) find no evidence of diet switches, specifically, of switches to herbivory, being connected with clade richness. Species estimates for various insect groups are given below.

Weevils include some 62,000 described species and perhaps 220,000+ species altogether. McKenna et al. (2009) suggest that crown-group diversification of major angiosperm-associated weevil clades was underway by the Aptian 125-112 Ma, and there was a "massive diversification" as angiosperms became common. Basal Curculionidae show strong associations with monocots, but there is little evidence that early monocots were either particularly abundant or ecologically successful (Crane et al. 1995; Friis et al. 2004; J. A. Doyle et al. 2008; c.f. McKenna et al. 2009). Scolytinae, Cossoninae, and Platypodinae are the three major clades of wood-boring (endophagous) weevils, a habit that originated independently in the three (Haran et al. 2013); for Scolytinae, see under Pinaceae.

Chrysomelidae or leaf beetles, including the bruchids or seed beetles, are another very speciose herbivorous clade. Their origin has been dated to (86-)79-73(-63) Ma, well after the origin of the angiosperms (e.g. Gómez-Zurita et al. 2007).

There are at least 150,000 species of butterflies and moths (Lepidoptera) (Roe et al. 2009; Mitter et al. 2016 for summaries; I follow the latter for nomenclature), and it is estimated that larvae of about two thirds of these are herbivores, most being mono- or oligophagous (Bernays & Chapman 1994). A clade that include many macrolepidoptera, butterflies and some groups of moths, are well embedded in the paraphyletic microlepidoptera. Caterpillars of many of the more basal microlepidopteran clades, i.e. not most Apoditrysia (see below), are leaf miners, stem borers, etc, i.e., they are internal feeders and do not feed exposed on the plant. Adults of the very basalmost clades of Lepidoptera have jaws, and they include Micropterigidae (their larvae often eat hepatics), Agathiphagidae (they eat Agathis), and Heterobathmiidae (they eat Nothofagus), and they may have diverged as early as the end-Triassic. Agathiphagidae may be sister either to Micropterigidae, Heterobathmiidae, or, perhaps most likely, to all other lepidoptera (Wahlberg et al. 2013; see esp. Regier et al. 2015; Kristiansen et al. 2015; Heikkilä et al. 2015; Mitter et al. 2016). However, recent work supports the position of Micropterigoidea as sister to all other lepidoptera, crown-group Lepidoptera being (312.4-)299.5(-276.4) Ma, and [Agathiphagoidea + the remainder] being (297.0-)280.6(-257.4) Ma (Kawahara et al. 2019). [Heterobathmioidea + other leps], = Angiospermivora, are (276.7-)257.7(-234.5) Ma (Kawahara et al. 2019; see also Regier et al. 2015; Mitter et al. 2016). Glossata, i.e. these other lepidoptera, have probosces, and nectar feeding may have evolved (261.1-)241.4(218.9) Ma (Kawahara et al. 2019). Within Monotrysia - the females have a single opening for mating and laying eggs - are the ca 850 species (2,000-2,500 in fact?) of tiny Nepticulidae, mostly leaf miners of woody core eudicots (Doorenweerd et al. 2016) sister to rest? Van Eldijk et al. (2018) speculate about the feeding habits of caterpillars of non ditrysian Glossata 212 Ma scales of which they found in Germany, suggesting that perhaps they ate gymnosperms. Relationships among major groups of Ditrysia, sister to Monotrysia, the ca 152,000 species of Glossata in which the females have separate openings for mating and laying eggs and which include the majority of macro- and microlepidoptera, have been unclear (Mutanen et al. 2010; Heikkilä et al. 2015). However, it seems likely that "Tineoidea" (Tinaeidae), with ca 3,000 species (Sohn et al. 2015), are strongly paraphyletic at the base of Ditrysia, interestingly, they are fungus and detritus feeders (Regier et al. 2013, esp. 2014). This might suggest that the internal feeding habits of other ditrysian lepidoptera are derived, not plesiomorphous, although optimization of this character will clearly be difficult. These others include a clade made up of Yponomeutoidea, which have ca 1,800 species and in which there have been several shifts between internal and external feeding (Sohn et al. 2013), and the around 2,000 species of Gracillarioidea, largely leaf miners although including Epicephala, a large genus that is distinctive in being a pollinator/seed predator on Phyllanthus s.l. in Malpighiales-Phyllanthaceae (e.g. Kawahara et al. 2016). This clade is sister to the ca 144,500 species of Apoditrysia, which consists of a lerge polytomy including a number of very small clades, also Tortricoidea, the leaf roller moths, and a very large clade containing the remaining leps (Mitter et al. 2016). Within Tortricoidea, the family Tortricidae include ca 11,000 species, mostly feeding on Rosales and Asterales, and Fagua et al. (2017) provide a dated phylogeny that is complete at the tribal level, allowing them to think about connections between moth and angiosperm diversification (see below). Kawahara and Breinholt (2014) link large moths including the megadiverse Noctuoidea s.l. with much smaller Pyraloidea, etc., the combined clade being sister to a clade that includes the monophyletic butterflies, Papilionoidea (see also Regier et al. 2013). Cho et al. (2011 and literature; see also Mutanen et al. 2010; Wahlberg et al. 2013; Regier et al. 2013) also suggested that butterflies and large moths were are not sister groups. Papilionoidea are part of a clade (or of a large polytomy - Mitter et al. 2016) that also includes Thyridoidea, the picture-winged leaf moths, and Copromorphoidea, tropical fruitworm moths (Regier et al. 2013). Papilionoidea themselves (Wahlberg et al. 2013) are perhaps (110.3-)98.3(-86.9) Ma (Kawahara et al. 2019) and include a small clade made up of the mostly night-flying Hedylidae, the American moth-butterflies, and relationships may be [Papilionidae [[Hedylidae + Hesperidiidae] [Nymphalidae [Pieridae [Riodinidae + Lycaenidae]]]], although these relationships are not set in stone (Heikkilä et al. 2011; Regier et al. 2013; Kawahara & Breinholt 2014: pierids not included; Breinholt et al. 2018; Kawahara et al. 2019: [Nymphalidae [Riodinidae + Lycaenidae]]). Immediately downstream from this whole [butterfly + some moths] clade are groups like Gelechioidea, with over 5,000 saprophagous species (Sohn et al. 2015), and Zygaenoidea (Regier et al. 2013).

However, clade ages in Lepidoptera are quite uncertain. Thus crown-group Lepidoptera may be 250 Ma (Condamine et al. 2016), ca 215 Ma (Wahlberg et al. 2013) or ca 140 Ma (Misof et al. 2014) or ca 299.5 Ma (Kawahara et al. 2019). Grimaldi (1999) and Grimaldi and Engel (2005) thought that diversification of Glossata began in the mid- to upper Jurassic, Labandeira et al. (1997), Wahlberg et al. (2013) and Kawahara et al. (2019: ca 241.4 Ma) suggesting somewhat older dates. Similarly, Ditrysia may have originated ca 160 Ma (Wahlberg et al. 2013) or ca 100 Ma (Misof et al. 2014). Within Ditrysia, the crown-group age of the butterfly clade is perhaps 104 Ma (Wahlberg et al. 2013) or (110.3-)98.3(-86.9) Ma Kawahara et al. (2019). The recent discovery of lepidopteran scales in deposits from the late Triassic 212 and 201 Ma in Germany may upset some apple carts (van Eldijk et al. 2018). Some of these scales have been linked with the basal grade of jawed moths (but not Agathiphagidea), while others are hollow and may have a serrated apex, and have been linked to Glossata (van Eldijk et al. 2018). However, given the uncertainty in divergence times within angiosperms it is difficult to say much about possible linkages between the diversification of lepidoptera and that of flowering plants, although stem-group angiosperms, at least, have been around for a long time... (Kawahara et al. 2019; see also Regier et al. 2015 and below).

There are over 4,400 species of aphids (Hemiptera-Sternorrhyncha-Aphidoidea-Aphididae) feeding on plant sap, and they tend to be monophagous. Their diversity is greatest in temperate areas, although myrmecophilous species are commonest in the tropics (Bristow 1991: Stadler & Dixon 2005; Iluz 2011). Diversification may be Late Cretaceous/early Caenozoic (von Dohlen & Moran 2000) or somewhat earlier (R. Chen et al. 2016). Hemiptera-Sternorrhyncha-Coccoidea, with around 8,000 species (Iluz 2011; Burns & Watson 2013), and Psylloidea (jumping plant lice) also tend to be very host-specific.

Crown-group diversification of major angiosperm-associated weevil clades may have been underway by the Aptian 125-112 Ma, with a "massive diversification" of Curculionidae - now ca 90% of all weevils - 112-93.5 Ma during the Cretaceous Terrestrial Revolution (KTR: McKenna et al. 2009), and the evolution of obligate associations between weevils (ambrosia beetles) and ascomycete fungi are discussed further below; see also Pinaceae. Plant-sawfly associations may have become reorganised around the KTR, interestingly, basal clades of sawflies are associated with angiosperms, while within monilophytes there is no association between the phylogenetic position of the plant with that of the sawfly larvae that eat it (Schneider 2016). Initial divergence within butterflies s.l. (Papilionoidea) may have been ca 110-95 Ma (Heikkilä et al. 2011), (110.3-)98.3(-86.9) Ma (Kawahara et al. 2019) or (143-)119(-91 Ma (Espeland et al. (2020: other dates from the literature).

Galls, often with very distinctive morphologies, result from close associations between plant and insect (see Shorthouse & Rohfritsch 1992; Raman 2011; Redfern 2011; Ferreira et al. 2019 for good introductions). There are anywhere from (21,000-)132,930(-211,000) species of of galling insects (e.g. Espiritó-Santo & Fernandes 2007), the estimates are based on extrapolation and so partly depend on the numbers of flowering plants because of the specificity of many galler/plant associations.

The dipteran gall midges (Cecidomyiidae-Cecidomyiinae, see Roksam 2005 for a phylogeny) are the largest group of galling insects and make up perhaps 25% of all galling insects in North America (Abrahamson & Weiss 1997). They are worldwide in distribution and pretty much host specific; Yukawa and Rohfritsch (2005) suggested that they showed no particular patterns of host associations, but c.f. Gagné and Jaschoff (2017) and Dorchin et al. (2019), and see below for geography. Fungus-eating may be plesiomorphic in cecidomyiids (Ferreira et al. 2019), and transition to plant feeding may have occurred twice in the Upper Cretaceous 100-80 Ma (Dorchin et al. 2019: fig. 3). Ambrosia galls (Dorchin et al. 2019), in which the insect larvae eat the mycelium of fungi growing between plant cells in the gall, represents a secondary association of midge larva and fungus. The hymenopteran gall wasps (Cynipidae) may comprise as many as 50% of gallers locally and are north temperate; they are almost entirely restricted to eudicots (Ronquist & Liljeblad 2001; Csóka et al. 2005; Ács et al. 2009 for species limits). Smaller groups of gallers include psyllids (jumping plant lice, hemipterans) which are particularly common in Australia, and also on Fabaceae, at least in the Neotropics (Fernandes & Price 1991; Crespi et al. 2004; Espiritó-Santo & Fernandes 2007; Raman et al. 2005) and a couple of hundred species of aphids, also hemipterans (Wool 2004; J. Chen et al. 2013 for the phylogeny of hormaraphidine gallers). Some other insects, including a few Lepidoptera, are also gallers, as are some acarines and nematodes, the latter galling either roots or stems, and acarines and nematodes, as well as Thysanoptera and a number of Hemiptera, may live in colonies inside the galls they stimulate (Ferreira et al. 2019). The morphology of particular gall-plant associations is often very distinctive, and even galls made by different species of the one genus of gallers on different species of the one genus of plants (the Nothotrioza-Psidium system) and that look very similar may differ at the cellular and molecular levels (Carneiro et al. 2015).

In general, gall-inducing insects are commonest on sclerophyllous plants growing on poor soils in warm climates between 25 and 45o N and S, or perhaps more generally in species-rich communities, whether dry or wet, but not necessarily in tropical climates (Price et al. 1998; Yukawa & Rohfritsch 2005; see Price et al. 1987 for galling in an adaptive context). Other organisms may be directly involved in functioning gall communities, such as fungi in cecidomyiid ambrosia galls (Rohritsch 2009 and references; Dorchin et al. 2019). Here the fungus may get its nutrients from the plant and is eaten by the midge larvae, indeed, cecidomyiids may originally have been fungivorous, other gallers are herbivorous (Roksam 2005; Dorchin et al. 2019). A complex network of parasites, hyperparasites and predators is all more or less dependent on gall larvae (Redfern 2011; see also figs and fig wasps).

2C. Pollinating Insects (see also butterflies above).

The evolution of bees is of particular importance, given the close involvement of many of them with angiosperm pollination (for bees and pollen, see Westerkamp 1996; for an account of all bee groups, see Michener 2007). Apoidea includes the spheciform wasps, bees evolving from within the wasps, a group that feeds their larvae with insects (e.g. B. R. Johnson et al. 2013), and they evolved from within the digger wasps, "Crabronidae", wildly polyphyletic and to be split into eight families or so (Peters et al. 2017a; Branstetter et al. 2017a; Sann et al. 2018). Social behaviour, communal nesting and eusociality, occurs in this clade (Sann et al. 2018). The immediate sister group of bees, (Anthophila), may be Ammoplanidae (Sann et al. 2018).

The basic phylogenetic structure within Anthophila/Apiformes, the clade that includes all bees, is [Dasypodainae [[Meganomiinae + Melittinae] [[Andrenidae [Halictidae [Stenotritidae + Colletidae]]] [Apidae + Megachilidae]]]] (Cardinal & Danforth 2013; Peters et al. 2017a), thus the old mellitids, here represented by Dasypodainae and [Meganomiinae + Melittinae], are paraphyletic, although they are monophyletic in Hedtke et al. (2014). Recent suggestions are that the age of stem-group bees is some (182-)149(-119) Ma, in line with some estimates of ages for the origin of angiosperms, with crown-group Megachilidae, a major clade including the leaf-cutting bees, dated at (154-)126(-100) Ma (Litman et al. 2011); see also Cardinal et al. (2018) for ages and relationships. Colletidae, a group of generalist bees, showed no obvious burst of Tertiary diversification (Almeida et al. 2011). Cardinal and Danforth (2013) suggest that crown-group Apiformes are some (132-)123(-113) Ma, all families having diverged by the K/P boundary, Sann et al. (2018) date stem-group Anthophila at (148-)128(-108) Ma, and Cardinal et al. (2018) estimate that crown-group anthophila are (156-)125(-?107) Ma old.

Stem Apidae are some 135-120 Ma (Grimaldi & Engel 2005), with their initial diversification apparently occurring in the early- to mid-Cretaceous 112-100 Ma in association with the evolution of angiosperms (Grimaldi 1999, see also Engel 2000; Michez et al. 2009, 2012: discussion of fossils purporting to be bees; Grimaldi & Engel 2005; Almeida & Danforth 2009; Cardinal & Danforth 2013; c.f. Renner & Schaefer 2010). Crown-group Apidae are dated to (95-)87(-78) Ma (Cardinal & Danforth 2011). Within Apidae, Xylocopinae entered the Caenozoic as four clades that had diverged about 20Ma before, but in the early Caenozoic diversification increased considerably; before the Caenozoic there is likely to have been extinction in these clades (Rehan et al. 2013). The somewhat over 1,000 species of primitively eusocial corbiculate bees, Apinae, have the relationships [Centridini [Euglossini [Apini [Meliponini + Bombini]]]]] (Bossert et al. 2018), or [[Euglossini + Apini] [Meliponini + Bombini]] (e.g. Cardinal et al. 2010; Cardinal & Danforth 2011; Danforth et al. 2013; Martins et al. 2014a), although morphology-based trees suggest relationships [Euglossini [Bombini [Apini + Meliponini]]] (Canevazzi & Noll 2015 and references). Corbiculate bees are estimated to be (89-)77(-66) Ma (Martins et al. 2014a, q.v. for other estimates - Ma Cardinal & Danforth 2011) or around 62 Ma, i.e. early Palaeocene (Peters et al. 2017a). Within the corbiculate clade, the crown group of the stingless, rather speciose and highly eusocial meliponines is dated to (61-)58(-56) Ma and (56-)51(-48) Ma, that of the euglossine orchid bees to (35-)28(-21) Ma and (38-)26(-17) Ma, of bumble bees (Bombini) to (31-)21(-12) Ma and (48-)26(-14) Ma, and of honey bees (Apini) to (30-)22(-16) Ma and (29-)22(-17) Ma (estimates from Cardinal & Danforth 2011 and Martins et al. 2014a respectively). There are other age estimates. Crown-group meliponine bees are known from the Late Cretaceous Raritan Amber ca 91 Ma (Dehon et al. 2014). However, a fossil from the Late Cretaceous (96-74 Ma) New Jersey amber was assigned to the extant genus Trigona, a highly derived eusocial stingless bee (Meliponini: Michener & Grimaldi 1988); both its age (now estimated at 70-65 My) and its relationships (it is placed in Cretotrigona, a stem meliponine) have been re-evaluated (Engel 2000). An estimate of the age of crown-group euglossines is 42-27 Ma (Ramírez et al. 2010) and the fossil euglossine Euglossopteryx biesmeijeri from Green River, Utah, is estimated to be ca 47 Ma (Dehon et al. 2014). An age for crown-group Bombini is (49-)44.6, 27.9(-25.4) Ma (Hines 2008: Table 1, highlighted areas). For relationships within Bombini, see S. Cameron et al. (2007; also Hines 2008), and for their classification, see P. H. Williams et al. (2008).

These general relationships are consistent with the appearance of bees in the fossil record. Both Apidae and Megachilidae, derived long-tongued bees, are known from Baltic amber of Eocene age (Danforth et al. 2006 and references). An early fossil, perhaps sister to other Apoidea, was found in amber of Upper Albian (probably Early Cenomanian age (99.4-)98.8(-98.2) Ma (Shi et al. 2012) from Myanmar (Poinar & Danforth 2006), however, it may really be a predatory wasp (Ohl & Engel 2007; c.f. Danforth & Poinar 2011). The corbiculate fossils are described as , and Bombus cerdanyensis ca 10 Ma deposits from Spain, Bombus s. str. ca 18 Ma They provide new information on the distribution and timing of particular corbiculate groups, perhaps global Eocene-Oligocene cooling-induced extinctions (Dehon et al. 2014). Anthophorine now 350 spp. Protohabropoda pauli ca 25 Ma Late Oligocene France reinforces previous hypotheses of anthophorine evolution in terms of ecological shifts by the Oligocene from tropical to mesic or xeric habitats. Lastly, the Andrenid, Andrena antoinei the today widespread genus Andrena ca 25 Ma in France; Apis 10 Ma (Dehon et al. 2014).

3. Angiosperms and Fungi. (See F. Martin 2017, Tedersoo 2017a, and Brundrett and Tedersoo 2018 for important entries into the literature, and keep an eye on the New Phytologist.)
3A. Early Plant-Fungal Relationships.
3B. Mycorrhizae in Extant Embryophytes.
      Ecto- and Ericoid mycorrhizae.
      Fine Root Endophytes.
      On mycorrhizal networks.
3C. Endophytes
3D. Mycorrhizae and Endophytes in General.
3E. Further Complexities.

3A. Early Plant-Fungal Relationships. Embryophytes and fungi established associations very early in the Silurian/Devonian (e.g. Selosse & Tacon 1998; Nebel et al. 2004). In some extant "bryophyte" clades Mucoromycotina are associated with the gametophytes, Endogone-like fungi forming mycorrhizae with liverworts like Treubia and Haplomitrium (Field et al. 2012, 2014, 2015d) and with hornworts (e.g. Bidartondo et al. 2011; Desirò et al. 2013; see also Rimington et al. 2019). Fungi in these liverworts are found in the rhizoids and thallus, and the relationship between plant and fungus seems to be one of mutualism (Field et al. 2014). Members of a clade of the basidiomycete Sebacinales-Serendipitaceae are also associated with liverworts (Weiß et al. 2016), but this is likely to be a fairly recent connection. Mosses - but not Takakia - usually lack mycorrhizal associations (Read et al. 2000; Kottke & Nebel 2005; Duckett et al. 2006b; Ligrone et al. 2007; Wickett & Goffinet 2008; Stenroos et al. 2010; Pressel et al. 2008, 2010; Rimington et al. 2014, 2017). Some early vascular plants of the 407 Ma Rhynie Chert formed associations with both mucoromycotes and Glomeromycotina (Remy et al. 1994; Strullu-Derrien et al. 2014, 2017; also Field et al. 2019 and references; Rimington et al. 2019). Since the nature of the plant-fungus association in many non-seeding plants can be rather different from the classic ectomycorrhizal (ECM) (AM) or vesicular arbuscular mycorrhizal associations, it has been suggested that such early plant-fungus relationships are best called paramycorrhizal associations (Kenrick & Strullu-Derrien 2014). There is discussion as to whether mucoromycotes or glomeromycotes were initially involved in these associations, some recent papers suggest the latter were the first plant symbionts, although Pirozynski and Malloch (1975: as oomycetes, chytrids, etc.) had early suggested that fungi that are now included in glomeromycotes were involved. However, the issue is now more complicated because relationships between bryophytes and vascular plants are currently being re-evaluated, the two perhaps being reciprocally monophyletic (for further details, see elsewhere), and fine root endophytes, previously included in Glomus, are in fact mucoromycotes (see below).

3B. Mycorrhizae in Extant Embryophytes. The evolution and ecological significance of mycorrhizae have been widely discussed (see Malloch et al. 1980; papers in Allen 1992; Read et al. 2000; Landis et al. 2002; Egger & Hibbett 2004; L. L. Taylor et al. 2009; Field et al. 2012, 2014, etc.), as has the morphology of the plant/fungus interface (e.g. Peterson & Massicotte 2004; Peterson 2013 and references) and how the fungus uses the 10% or more of photosynthesate that it gets from the plant (Leake et al. 2004). For the general economics of the exchange between the two partners - mostly sugars from the plant, but a greater variety of metabolites from the fungus, see Wyatt et al. (2014). For a comparison of carbon cycling in ECM and AM dominated communities, with a focus on subarctic alpine Sweden, see Soudzilovskaia et al. (2015) - soils in the latter are associated with organic horizons in which there is higher respiration, lower extractable C, etc.. For organic nutrient uptake by mycorrhizae and the amount of carbon in the soil, see Orwin et al. (2011: kind of mycorrhizae not specified). For comprehensive surveys of mycorrhizal associations, see Brundrett (2008: updated online resource, 2009), Akhmetzhanova et al. (2012) in particular, also Garbaye (2013) and numerous papers in the New Phytologist that have appeared over the years.

This is a complex story. Being mycorrhizal is not a simple either/or matter, furthermore, one species of plant can have different kinds of fungal associations (Field & Pressel 2018 and references; Rimington et al. 2019) and one species of fungus may enter into different kinds of associations with different plants (see also below). Inded, not all mycorrhizal associations are mutualistic, i.e. with positive results for both sides, but depending on the circumstances the benefits of a mycorrhizal association may change, and the fungus may end up as parasitic on the plant (e.g. N. C. Johnson et al. 1997; M. D. Jones & Smith 2004). Furthermore, the genome of the glomeromycote Rhizophagus irregularis has a number of similarities with those of Mucoromycotina (Tisserant et al. 2013), the two groups being more closely related to each other than to dikaryotic fungi (K. Lin et al. 2014: other zygomycotes not included), and both groups have Mollicutes-related endobacteria (Desirò et al. 2014); relationships between the two groups need clarifying (Field et al. 2015d). An additional complication is that fine root endophytes, fungi with hyphae ca 1.5 µm across that produce fan-like arbuscules and small vesicles and quite common in vascular plants, had been included in Glomus, but their SSU 18S ribosomal RNA gene groups with those of mucoromycotes, not with Glomus and relatives (Orchard et al. 2016; Hoysted et al. 2017 - see below). A possible complication in the story of the evolution of mycorrhizal relationships is the association of fungi with the gametophytic generation in liverworts, hornworts and even monilophytes, but with the sporophytic generation in seed plants (Desiró et al. 2013), although in some lycophytes and monilophytes the same fungus is found in both generations (Winther & Friedman 2008; Field et al. 2015). Finally, the basal Agaricomycete order Sebacinales are involved in important interactions with land plants from liverworts to orchids and oaks, with both chlorophyllous and echlorophyllous plants, and in associations that range from endophytic to ectomycorrhizal. Most of the over 1,000 species in the order are undescribed, and only recently has it been divided into two families. One is the largely ectomycorrhizal Sebacinaceae (= Group A Sebacinales of earlier studies), also found on basal Ericaceae, some orchid mycoheterotrophs, and even a few endophytes (Selosse et al. 2009; Weiß et al. 2016), and the other is the ecologically very diverse Serendipitaceae (= Group B Sebacinales), of which only four species have been described and of which "fruiting bodies"/basidiomes had never been seen until recently (Weiß et al. 2016).

Mycorrhizae are supposed to be uncommon in epiphytic taxa (Janos 1993; see other papers in Mycorrhiza 4(1). 1994; Desirò et al. 2013; Kato & Tsutsumi 2013), but c.f. Ericaceae and Orchidaceae, two of the major epiphytic angiosperm groups (e.g. Kottke et al. 2008; Martos et al. 2012). Ericoid mycorrhizae will be discussed along with ectomycorrhizae, since although the nature of the two is rather different, their effects on the biosphere are similar/complementary (see also elsewhere, but orchid mycorrhizae (q.v.) - basidiomycetes, including Sebacinales-Serendipitaceae are commonly involved (Weiß et al. 2016) - will not be mentioned further here, although in fungal associations they are quite similar both to ecto- and ericoid mycorrhizae (Toju et al. 2016; c.f. in part Imhoff 2009). Aquatic flowering plants, hardly surprisingly, often lack mycorrhizae (see Safir 1987 and Radhika & Rodrigues 2007 and references for records; de Marins et al. 2009). Mycorrhizae are also often absent in many Proteales, Caryophyllales, Brassicales, and the like. Such plants fall into two groups, those that grow on relatively fertile soils with exchangeable P, e.g. Brassicales, or those that grow on soils that are either low in P or where the P is sorbed onto the soil, e.g. by goethite, as in Proteales (Lambers et al. 2015c; see also Werner et al. 2018). Indeed, other than Proteales, many non-mycorrhizal plants are annuals, e.g. in Lamiales (Trappe 1987; Wilson & Hartnett 1998; Brundrett 2017b; Brundrett & Tedersoo 2018 for summaries, c.f. in part Werner et al. 2018 and Wipf et al. 2019) and are i.a. effective colonizers of disturbed habitats (e.g. Pirozynski & Malloch 1975). Mycorrhizal associations are proportionally somewhat less common in island plants than in plants from the mainland, and also in plants that grow at higher latitudes (Delavaux et al. 2019 and references, see also below), indeed, Bueno et al. (2017) found that non-mycorrhizal species made up to ca 30% of the total at higher latitudes in Europe. Overall some 8% of vascular plants lack mycorrhizae, and a further 7% are only facultatively mycorrhizal (Brundrett & Tedersoo 2018). Although normally non-mycorrhizal species may on occasion be associated with AM fungi, vesicles sometimes being produced but arbuscules only rarely, and there may even be some movement of carbon from plant to fungus (Lekberg et al. 2015), however, there are several reports of the growth of non-mycorrhizal plants being reduced when growing with AM plants (see Veiga et al. 2013: experiments on Arabidopsis thaliana). E. I. Jones et al. (2015) discuss the general problem of cheating in mutualistic situations like those mentioned here, as do Gomes et al. (2019) - mycoheterotrophy may be one outcome.

Endomycorrhizae or arbuscular mycorrhizae (AM) (no distinction between AM and vesicular-arbuscular mycorrhizae - VAM - is made below) are very widespread, being found in about 70% of seed plants, 72% of flowering plants, 67% of ferns, 80% of all land plants (?correct), and 92% of plant families (Blackwell 2011; Brundrett 2009; Brundrett & Tedersoo 2018), perhaps 200,000 or so species of plants being involved (Rinaldi et al. 2008). AM associations are of long standing, and if they were not to be found in the common ancestor of all land plants (Parniske 2008), they characterize at least a major subset of vascular plants (see above). Mycorrhizae are uncommon in fossil gymnosperm roots, although AM have been described in Upper Permian Glossopteris, the Triassic voltzialean Notophytum, etc. (Harper et al. 2013, 2015 and references). Just how many times AM associations evolved in land plants is unclear; in part, it depends on the very definition of mycorrhizae and on the relationships of bryophytes. Indeed, bryophytes are probably monophyletic (e.g. Puttick et al. 2018), while the recent realisation that fine root endophytes, although initially described as Glomus tenue, is/are in fact mucoromycotes, and that such fine root endophytes are quite widespread in land plants (Orchard et al. 2016, 2017; see also below), means that caution is needed when reading the earlier literature on the evolution of ecophysiological interactions between AM and land plants (Field et al. 2019).

Glomeromycota are often the fungi involved in AM associations (Schüßler et al. 2001; but see above). The origin of glomeromycotes may be dated at (715-)659(-606) Ma, their diversification beginning (529-)484(-437) Ma, roughly contemporaneous with the early diversification of embryophytes (485-)482(-473) Ma; mucoromycotes, their sister group, possibly also involved in early AM symbioses (as fine root endophytes), are ca 406 Ma, based on the divergence of Endogone (Lutzoni et al. 2018).

Some 290 species of Glomeromycota have been described (Öpik et al. 2010; Merckx et al. 2012), and all form AM associations; overall, individual species are widely distributed and show low host specificity (references in Schappe et al. 2017). However, some surveys suggest that the dispersal of the fungi may be limited, partly because their spores are relatively large and not dispersed by wind (Tedersoo et al. 2014b, but c.f. Geml 2017), so Glomeromycota diversity may be considerably underestimated (see also Kottke & Kovács 2013). Thus in a survey in an AM forest in New Zealand, Martínez-García et al. (2014) could associate only 8 out of the 113 glomeromycote OTUs found there with names, and Gorzelak et al. (2017) found a number of unidentified AM fungi in northern Thuja plicata-dominated forests. But even if there are around 1,500 species of glomeromycotes (Rosendahl 2008; N. C. Johnson 2009; Kivlin et al. 2011; Mathieu et al. 2018: Pickles & Pither 2013 for cautionary comments), that is still far fewer than the some 200,000 species or so of AM plants (Rinaldi et al. 2008; Kottke & Kovács 2013; E. Chen et al. 2018a). Indeed, other work suggests that glomeromycotes show very little local endemism, 93% of the taxa (virtual taxa) being known from more than one continent and one third found on six (Davison et al. 2015: Antarctica not included; also Öpik et al. 2016; Pärtel et al. 2016; Savary et al. 2017; see comments by Bruns et al. 2016) - not necessarily at odds with the findings of Martínez-García et al. (2014). There is some evidence for host specificity or at least host preferences of AM fungi (e.g. Gosling et al. 2013; Martínez-García et al. 2014). López-Garcí et al. (2017) found that the phylogenetic over-/underdispersion of AM communities depended on the life form of the plant with which they were associated - and on when the soil was collected...

There may indeed be relatively few AM species, but differences in AM taxa associated with two species of the palm, Howea, growing on different soil types on Lord Howe Island, off Australia, may be connected with the coexistence of these closely related species (Osborne et al. 2017). Pärtel et al. (2017) emphasized that the extent and linkages of tropical grasslands and savannas during the last glacial maximum was associated with the largest species pools of AM fungi. Wipf et al. (2019) even suggested that annual species had a greater diversity of AM fungi than pernnials, and that half the identified AM fungi are found only on one species (see Torrecillas et al. 2011, 2012)... Wenner et al. (2014) found that fungi in related species (= same genus, subfamily) of Asteraceae tended to be most similar (here Glomeromycota were in a minority); overall, AM fungal diversity was not correlated with that of its hosts (Tedersoo et al. 2014b), and in Panamanian forests composition of AM fungi was affected more by soil properties than tree identity, that of other fungi (no ECM fungi there) was affected by both (Schappe et al. 2017). Many plants form associations with several species of fungus, and ecological specialists and generalists may form associations with different fungi (Öpik et al. 2009, 2010). A single plant can also form different associations sequentially (van der Heijden et al. 2006) or unrelated species of plants may be colonized by the one fungus (Kottke et al. 2008; Walder et al. 2012). Some evidence suggests that fungal associations in AM networks are nested, that is, specialist fungi, AM fungi forming particular associations with plants, are frequently found along with generalist fungi that are also found on other AM plants (Toju et al. 2016 and references, c.f. 2018; Rimington et al. 2019). Thinking about specificity from a different point of view, gypsum-derived soils may have some distinctive AM fungi (Torricellas et al. 2014). AM associations predominate in the tropics, AM associations with woody plants being much less common towards the Arctic (see also below), and Veresoglou et al. (2019) suggest connections with the light budget; Veresoglou et al. (2018) examine latitudinal differences in glomeromycote diversity in China.

The association between plant and AM fungus is obligate (Kohler & Martin 2017). Glomeromycote hyphae are aseptate and intracellular, often forming vesicles and/or branching structures (the arbuscules) within the cells; morphological details of the fungus-plant association vary, Paris-type associations being largely intracellular and with intracellular coils, the commoner Arum-type being intercellular, but producing intracellular arbuscules (e.g. F. A. Smith & Smith 1997; Torti et al. 1997; Peterson & Massicotte 2004), and there are vesicles, etc.. However, the distinction between the two mycelial types depends in part, at least, on whether or not the host plant has intracellular spaces, and there are cases where the one species of fungus produces different mycelial types in different species of plant (Smith & Smith 1996 and references; Sýkorová 2014). For details of the fungus-plant interface focussing on the nature of the periarbuscular membrane synthesised by the plant that surround the fungal arbuscules, see Ivanov et al. (2019). AM fungi never enter meristems or the vascular cylinder (Sýkorová 2014).

The basic biology of glomeromycote fungi is poorly understood. The spores are large and multinucleate with hundreds to thousands of nuclei (Kamel et al. 2017). It has even been suggested that the nuclei in any one spore do not have an immediate common ancestor and that the units of selection were the individual nuclei (Jany & Pawlowska 2010). The hyphae are coenocytic, and hyphae from different mycelia can fuse, making the nuclear mix potentially yet more complex (Giovannetti et al. 2004; Simard et al. 2012 and references). However, K. Lin et al. (2014; see also Bruns et al. 2018) found little variation between different nuclei from the one spore. The phenotype of a particular species of AM fungus also depends on its host plant (Mathieu et al. 2018 for references), and Sanders (2018 and references) noted the great variation in fungal traits and in the growth of the plants with which they were associated (see also e.g. Koch et al. 2017; Ezawa & Saito 2018). This is in spores from the same fungus with no obvious sex (but see later), and if the pangenome - a concept designed to encompass population-level genomic variation in microbes - is large as it seems to be (see E. Chen et al. 2018a), then interactions between the fungus and different species of plants may well be complex. Indeed, infraspecific genome diversity in at least some AM fungi is very great. Thus Chen et al. (2018a), focussing on six laboratory-grown isolates of Rhizophagus irregularis coming from the same place in Switzerland, found a huge pangenome of around 150,000 genes (it is likely to become still larger), about four times the size of the genome of Arabidopsis thaliana and around ten times larger than that of some other fungi; around 50% of the genes in any one isolate were "dispensable or accessory genes with known molecular functions" (ibid. p. 1168), and there was also considerable variation in transposable elements (TE) between isolates (see also Mathieu et al. 2018; X. Sun et al. 2018: genome of Diversispora epigaea) and overall very high level of TE content (Morin et al. 2019). This also raises the issue of the nature of species in glomeromycotes (e.g. Sanders 2018; Bruns et al. 2018). Furthermore, there are still unanswered questions over how glomeromycotes reproduce. Thus R. irregularis has a number of genes that elsewhere would be involved in sexual reproduction (Tisserant et al. 2013 and references), and such genes have beeen found in other taxa (Morin et al. 2019). Ropars et al. (2016) found that some isolates were heterokaryotic, and sexual reproduction and accomoanying inter-nucleus genetic recombination in at least some taxa does seem likely (E. Chen et al. 2018b). Overall, rather little is known about the relationships between pangenome diversity, mode of reproduction, species limits and AM establishment and functioning.

Finally, glomeromycotes have a variety of bacterial endosymbionts (references in Wipf et al. 2019), but bacterial-AM fungal interactions are poorly understood (. Transmission of these bacteria may be vertical, and Burkholderia-related endobacteria seem to have a positive effect on their hosts, Gigasporaceae (X. Sun et al. 2018). Mollicutes/Mycoplasma-related endobacteria occur more widely in glomeromycotes, and horizontal transfer of genes occurs in both directions (Sun et al. 2018).

Some of the genes involved in the establishment of AM interactions are the same as those involved in establishment of nodulation (both by Rhizobium and the actinorhizal Frankia) in the nitrogen-fixing clade - there is a common symbiotic signalling pathway (CSSP) which may have originated around about 430 Ma in the Silurian (e.g. Gherbi et al. 2008; Gutjahr et al. 2008; Maillet et al. 2010; Svistoonoff et al. 2014; F. M. Martin et al. 2017; Barker et al. 2017; Gough & Bécard 2017 and references, see also Fabales), and there is also recent evidence of its involvement in some ectomycorrhizal associations (Cope et al. 2019); note that any connection with endophytic associations is unknown. This CSSP may well be ancestral in land plants (e.g. Martin et al. 2017). Delaux et al. (2014) distinguish between these CSSP genes, which also have other functions, and endomycorrhizal-specific genes. The establishment of AM associations and a variety of aspects of the root nodulation process starting with root hair curling are associated in Fabaceae and can be linked to an autophagy-related kinase, precursors for all these activities being produced by autophagy (Estrada-Navarrete et al. 2016). Initial attraction of the fungus to the plant, and also hyphal branching, is mediated by strigolactones secreted by the root (Akiyama 2010 and references). Lipochito-oligosaccharides are also involved in the signaling between plant and fungus (Maillet et al. 2011), and a karrikin receptor complex plays a role in the initial interaction between plant and fungus (karrikins and strigolactones both have a butenolide element), and this complex is found widely in embryophytes (Gutjahr et al. 2015). The invasion of plant tissue by the fungus involves appressorium formation in reponse to the production of cutin monomers and is similar to the establishment of parasitism (Bonfante & Genre 2010), especially by oomycetes (e.g. E. Wang et al. 2012, see also Brassicaceae and Fabaceae), although the overall effect of AM fungi on gene expression of the host is much less than that of endophytic or parasitic fungi (Dupont et al. 2015). For the set of "ancestral" AM genes in Selaginella, see Bravo et al. (2016). Interestingly, many important plant genes involved in AM associations have been detected in streptophyte clades immediately basal to the embryophytes and in basal clades of land plants (Delaux et al. 2012, 2015); for instance, strigolactones may initially have been involved in rhizoid elongation in the gametophyte (Delaux et al. 2012). Genes involved in signalling pathways are abundant in glomeromycotes, but in general it would be good to know more about what is going on in mucoromycotes... (Strullu-Derrien et al. 2018).

In AM associations, nutrient uptake by the plant - especially of phosphorus (P), and recent work adds nitrogen (N) (e.g. Herman et al. 2011: soil microbial decomposers benefit from AM, N to plant increases considerably; Nehls & Plassard 2018; Field & Pressel 2018) - is increased. The growth of the plant is stimulated (Sýkorová 2014) and water uptake is improved (Read 1991; Allen 1992; D. L. Jones et al. 2004: interaction of AM fungi and root exudates produced by plant; Govindarajulu et al. 2005; Leigh et al. 2009 and references; Tian et al. 2010; Bonfante & Genre 2010; S. E. Smith et al. 2011, 2015; Xue et al. 2018; Ezawa & Saito 2018; Wipf et al. 2019). Under low phosphate conditions root branching often increases, hence facilitating the increase of AM fungi, and the plant enables/pays for colonization of the fungi by diverting sucrose in particular to the roots to the tune of 20% or more of the total photosynthesate (Jakobsen & Rosendahl 1990: cucumber; S. E. Smith & Read 2008; Fusconi 2014; Sheldrake et al. 2017), interestingly, Tisserant et al. (2013; see also Bitterlich et al. 2017 and references) noted that the colonization of plants by AM might result in a 20% net increase in their photosynthesis (Kaschuk et al. 2009: crop Fabaceae). AM fungi effectively scavenge nutrients that become available as leaves, etc., decompose, but they cannot access nutrients in the complex humic substances in the soil organic matter that is left behind, thus organic P in temperate deciduous forests is in a less available pool than the P (as phosphate) where AM plants predominate (Rosling et al. 2015; see below). Most soil P is to be found in soil microbes, and AM efficiently scavenge soluble phosphate, but this becomes progressively more difficult to access as soils age (Turner et al. 2012). Gosling et al. (2013) found that the diversity, but often not colonization rate, of AM fungi was affected by the concentration of soil P, while Sharda and Koide (2010) found that high P levels were associated with lower levels of AM associations, or the associations might even become detrimental to the plant (see also N. C. Johnson 2009). Interestingly, soil P was found to affect species distributions in Panamanian l.t.r.f. (Condit et al. 2013), while endangered AM herbaceous plants in Eurasia persisted under low P conditions, but were lost when P was high (Wassen et al. 2005). Root hairs or an AM association may be alternative ways for a plant to obtain P when it is in short supply (Schweiger et al. 1995 and references).

Glomeromycotes are unable to break down and utilize complex biopolymers (Tisserant et al. 2013), a loss that has been dated to the end-Carboniferous ca 310 Ma (which has implications for glomeromycotes in bryophytes s.l.) and they obtain carbohydrates from the host plant (Helber et al. 2011; see also Walder et al. 2012; Kaiser et al. 2014; Strullu-Derrien at al. 2018). Recent work adds fatty acids, AM fungi having lost some fatty acid synthase genes (Kamel et al. 2017; Luginbuehl et al. 2017; Rich et al. 2017; Xue et al. 2018). Plant invertases convert sucrose that is moved to the roots to glucose and fructose which are then taken up by the fungus, and plastids in the plant initiate lipid biosynthesis, glycerol-3-phosphate acyltransferase being involved (Rich et al. 2017); carbon is stored in fungal spores, for example, as lipids (Rich et al. 2017). Much of the carbon transferred to the fungus is used up in respiration, furthermore, hyphal turnover time is commonly around as little as five days, so much C from the plant soon moves into the soil and thence back into the atmosphere (Staddon et al. 2003). Now rather old estimates suggest that around 5 gigatons of C may move from plants to AM fungi annually (Bago et al. 2000; Field & Pressel 2018).

A meta-analysis suggested that the increase of biomass in the host plant after inoculation with AM fungi was 65.7% ± 8.2 SE (Hoeksema et al. 2019). Overall, most AM fungi were associated with a variety of plant groups, although Funneliformis was particularly widely distributed, while the clade sister to other glomeromycotes, a clade that includes Archaeospora, was notably uncommon. There were few obvious patterns that linked with host phylogeny, although Funneliformis tended to have relatively lower beneficial effects than other taxa, except on Poaceae, but there Gigaspora had particularly low beneficial effects (Hoeksema et al. 2019).

AM also have other other effects on their hosts and the environment. They may improve soil structure - they produce large amounts of glycoproteins (L. L. Taylor et al. 2009; Garbaye 2013) - and drainage and so affect weathering; they may also help the plant in dealing with water stress (Augé 2001; Lehto & Zwiazek 2011: reviews). Indeed, AM fungi often improve plant growth during droughts, i.a. improving P uptake by the plant (Augé 2001) and also increasing stomatal conductance, more so in C3 than C4 plants (Augé et al. 2015; see also Worchel et al. 2013). Maherali and Klironomos (2007) found that ecosystem functioning was improved if all three major types of Glomeromycota were in the one community. Under certain conditions AM counteract negative effects of seedling damage, although overall rare species were more likely to benefit from the fungi - i.e. the Janzen-Connell effect was supported (Bachelot et al. 2016). The effect that parasitic species of Pedicularis with differing nutrient requirements had on their host, Trifolium repens, depended on the association of the latter with AM fungi and rhizobia, and in some combinations inoculation with the fungus might benefit both host and parasite, in the former alleviating damage caused by the parasite (Sui et al. 2018; see also Sui et al. 2014). Because hyphae of AM fungi from different plants may fuse, a potentially quite large number of plants from the same or different species may be put in indirect contact with each other (Giovannetti et al. 2004). As a result, the estimation of the costs and benefits accruing to the parties involved becomes complicated (Walder et al. 2012). Goes where: How changing carbon dioxide concentrations in the atmosphere might affect the soil carbon storage activities of AM is unclear (Verbruggen et al. 2012 and references).

There are additional dimensions to AM-plant associations. The jasmonic acid pathway, involved i.a. in defence against microorganisms that kill plant tissue before getting nutrients from it and against chewing insects, is upregulated in AM symbioses (Hause et al. 2002; Harley & Grange 2008; Jung et al. 2012; Wasternack & Hause 2013). There is anatagonism between this pathway and the salicylic acid pathway that causes the down-regulation of the latter, thereby perhaps making the plant more vulnerable to attacks by microorganisms that utilize living tissues and by sap-sucking insects (e.g. Thaler et al. 2012; Bastias et al. 2017). Furthermore, oomycete infection is connected with fungal associations. The RAM2 - Reduced Arbuscular Mycorrhization - locus is lost; RAM2 is involved in the production of cutin monomers recognized both by glomeromycotes during the establishment of AM associations and by oomycetes during the initiation of parasitism (E. Wang et al. 2012; Geurts & Vleeshouwers 2012). Hence loss of the ability to form AM associations is linked to the development of resistance to oomycete infestations in Brassicales. Indeed, this loss of ability has happened in parallel in various seed plants - perhaps around 25 times or so (Werner et al. 2018; see also e.g. Delaux et al. 2014; Selosse et al. 2015; Maherali et al. 2016; Kamel et al. 2016). In many cases AM associations have been replaced by ecto- or ericoid mycorrhizae, fungal associations with orchids, adoption of carnivory, etc. (Werner et al. 2018: Brassicaceae/Brassicales the major non-mycorrhizal group mentioned, apart from Orobancheae not muvh happening in the core asterids).

For other articles on AM, see Botany 94(6). 2016 and papers in F. Martin (2017).

Ecto- (ECM) and ericoid (ERM) mycorrhizae. ECM-plant relations are surveyed by Itoo and Reshi (2013); Corrales et al. (2018) outline what is known about ECM associations in the tropics. ECM/ERM seed plants are generally woody, although there are some herbaceous ECM taxa, particularly in Arctic-Alpine environments (e.g. Newsham et al. 2009: Polygonum, Carex [esp. the old Kobresia]). ECM fungi are associated with relatively few seed plants, although estimates vary: 2,500-3,000 species (S. E. Smith & Read 2008), 5,600 species of angiosperms + 285 species of gymnosperms (Brundrett 2009), 6,000-7,000 species in 250-300 genera (Tedersoo & Brundrett 2017) and ca 8,000 species (Rinaldi et al. 2008), both including gymnosperms; also F. Martin et al. (2016).

There are about 30 origins of the ECM habit in seed plants (Tedersoo & Brundrett 2017), and these origins show a strong phylogenetic signal (e.g. Alexander & Lee 2005; L. L. Taylor et al. 2009, 2011); this is discussed further below. The ECM habit predominates in only a few clades, a total of ca 8,500 species in 335 genera in 30 clades being a recent estimate, ca 2% of plant species (Brundrett & Tedersoo 2018). ECM plants include Fagales (some 1000 species), Pinaceae (210 spp.), Dipterocarpaceae (680 spp.) and relatives (together 915 spp), and Fabaceae-Detarioideae (250 spp. - Brundrett 2009; 450 spp. - B. Mackinder pers. comm. viii.2012), while some Salicaceae (Salix, Populus, 485 spp.) and other small clades are also ECM (see also Tedersoo & Brundrett 2017; Tedersoo 2017; etc.). Finally, there are almost 4,000 species of ERM Ericaceae; these are discussed further below.

When did these ECM-plant associations develop? ECM associations with Pinaceae, dating to perhaps 200 Ma or more, are likely to be the oldest; crown-group Fagales, ancestrally an ECM clade, have been dated to 120-62 Ma, which is not very helpful. Another major ECM clade is in Malvales, [[Pakaraimaea + Cistaceae] [Sarcolaenaceae + Dipterocarpaceae]], and this has been dated to as little as (25-)23(-21) Ma (Wikström et al. 2001) or over 88 Ma (Ducousso et al. 2004). Estimates of the age of crown-group Fabaceae-Detarioideae, in which ECM associations are notably common, range from ca 29.2 Ma (Lavin et al. 2005) to 68-64 Ma (de la Estrella 2017). For other estimates, see Tedersoo and Brundrett (2017) and Tedersoo (2017b) in particular.

ECM plants, unlike AM plants, are notably common in temperate to Arctic and montane habitats (Steidinger et al. 2019), although Dipterocarpaceae are largely Indian-South East Asian-Malesian and Fabaceae-Detarioideae in particular dominate large areas of Africa. In the Arctic, the distribution of ECM fungi is that of their hosts, mostly Pinaceae, Salicaceae and Rosaceae, so they are to be found quite far north, if fewer at the very highest latitudes (Newsham et al. 2009; Veresoglou et al. 2019: importance of light). Indeed, all boreal tree species are associated with ECM (Smith & Read 2008). AM are found to 82o N while ECM and ERM are found to only 79o N, being limited by the ranges of their mostly woody hosts (Newsham et al. 2009). However, AM fungal species were not found beyond 74o N in the Canadian Arctic (Olsson et al. 2004) and were very uncommon just below the Arctic Circle in Alaska, and this contrasted strongly with the diversity of other ecological groupings of fungi, although the disparity in terms of numbers of clones was somewhat less (D. L. Taylor et al. 2013). The diversity of ECM fungi increases in mid to high northern latitudes and that of ERM fungi also increases towards the poles, in both cases consistent with the distribution of their seed plant associates (Wardle & Lindahl 2014; Tedersoo et al. 2014b), although some early work suggests that the story in tundra habitats may not be so clear (Gardes & Dahlberg 1996). The relative abundance of ECM/ERM fungi in northern Temperate to Arctic areas is part of the reason why fungi as a whole tend not to have a bell-shaped latitudinal diversity curve, as is discussed later. Other groups of soil-dwelling organisms show a similar pattern that is perhaps connected with the substantial vertical stratification of soils in ECM-dominated communities (Tedersoo et al. 2012; see also Delgado-Baquerizo et al. 2016). ECM fungal diversity in dipterocarp forests and in detarioid savanna seems to be unexceptional. Although ECM fungi are found in white sand vegetation in tropical South America (Roy et al. 2016), their diversity there is unclear.

Tedersoo et al. 2012: Tedersoo et al. 2014b and Pickles & Pither 2013 for the care needed when estimating the diversity of ECM fungi). Indeed, a single individual of aspen, Populus tremula, may host as many as 122 ECM species (Bahram et al. 2011; see also Gardes & Dahlberg 1996; Kennedy et al. 2012 Walker et al. 2011; Timling & Taylor 2012). Serendipitaceae (and a number of other fungi) also form associations with Orchidaceae (Setaro et al. 2012; Yagame et al. 2016), but they are not discussed further here. Of course, until the advent of molecular methods, identification of the fungi depended on their being cultivable and undergoing sexual reproduction, and although Sebacinales, for instance, can now be identified using molecular data, some remain unculturable and most are undescribed (Tedersoo et al. 2010b; Weiß et al. 2016). Fungi growing in tropical white sand vegetation have recently been tabulated (Roy et al. 2016).

The ability to form ECM associations has evolved perhaps 82-86 times in fungi, especially in ascomycetes and basiodiomycetes, but also in Mucoromycotina, where Endogone, the fungus involved, is known as ECM on both Pinaceae and Fagaceae (Hoysted et al. 2018; Cope et al. 2019; Rimington et al. 2019 for references), and more origins are likely to be discovered especially in tropical and south temperate areas (F. Martin et al. 2010; Tedersoo & Smith 2013, 2017; see Wurzburger et al. 2016 for clades).

Dates of the origins of these fungal ECM clades, including those associated with Pinaceae, are split about equally between Late Cretaceous (e.g. Amanita) and Cenozoic (e.g. Hebelomateae, Tricholomatineae) (Ryberg & Matheny 2012; Tedersoo et al. 2014a and references; Sánchez-García & Matheny 2017: movement on to Pinus). Bonito et al. (2013) looked at the evolution of truffles (ascomycetes), and they suggested that the age of the clade that included Helvellaceae and Tuberaceae, all ECM fungi, was (184.7-)160.8(-137.4) Ma. In another estimate, the stem age of Cantharelles and Sebacinales, used as a proxy for the evolution of the ECM habit, was given as ca 229 Ma, somewhat after the early diversification of gymnosperms if well before the diversification of Pinaceae at (233.4-)183.3(-150) Ma (Lutzoni et al. 2018). The m.r.c.a. of Tuberaceae was dated to some (179.1-)156.9(-134.5) Ma, and Bonito et al. (2013) thought that its host was likely to have been an angiosperm. Kohler et al. (2015) suggest that ECM associations with plants developed in the last ca 175 Ma, and a clade of ECM, brown rot, and one white rot fungus is dated to ca 115 Ma. Augusto et al. (2014) dated confirmed ECM symbioses in both angiosperms and gymnosperms to the mid-Cretaceous, some 115 Ma, probable ECM symbioses in gymnosperms might be over 200 Ma in the Late Triassic, and possible ECM symbioses over 250 Ma, as early as the Permian, although Augusto et al. warn about extrapolating from the ecophysiological proclivities of modern gymnosperms to those of their early relatives. For the phylogeny of mushrooms, see Varga et al. (2019).

ECM fungus clade size varies greatly. Thus the ascomycete Cenococcum geophilum, the commonest ECM fungus (altough not that well represented in the study by Hoeksema et al. 2019), is the only ECM member of the 19,000+ species of the Dothidiomycetes (Peter et al. 2016), although this fungus in particular is unlikely to be able to do much to soil organic matter (Pellitier & Zak 2017). The fungi include basidiomycetes like Boletales and Agaricales in particular, also Pezizales and other ascomycetes like Cenococcum, and also some Zygomycota.

Garcia et al. (2015) and Cope et al. (2019) emphasize that differences in how ECM associations are established and maintained reflects their highly polyphyletic origin in seed plants and fungi. Thus ECM clades have quite a diversity of degradative enzymes (Martin et al. 2016). Hess et al. (2018) looked at the evolution of the ECM habit in Amanita but found no particular way of distinguishing the genomes of ECM members of the clade they studied from asymbiotic members, and both ECM and asymbiotic taxa might lose their genes for PCWDEs, indeed, the loss of some cellulose, hemicellulose, etc., breakdown genes might mean that the plant could not detect the fungus (see also Perotto et al. 2018). Hess et al. (2018) even toy with the idea that saprotrophic taxa might have some rudimentary capacity to form symbioses with plants, hence the recurrent evolution of ECM symbioses in particular (see also Hess & Pringle 2014). On the other hand, Schneider-Manoury et al. (2020) suggest that the endophytes (see below) may have been ancestral to the ECM habit. Along similar lines, although the common symbiotic signalling pathway (CSSM) is involved in the establishment of ECM in Populus with Laccaria bicolor (Agaricales-Hydnangiaceae), even other species of Laccaria have a rather different sets of symbiosis-specific proteins, and the CSSM may not be involved in the establishment of ECM associations in at least some Pinaceae-Pinoideae (Cope et al. 2019).

In ECM associations the fungal hyphae form a complete sheath investing the fine roots, the Hartig net, and hyphae penetrate between the epidermal and also cortical cells, the latter in gymnosperms; it is this sheath, not the root/root hair system (root hairs are often absent - Nehls 2008), that forms the interface between the plant and the soil (e.g. Clowes 1951; L. L. Taylor et al. 2009). Although some ECM Nyctaginaceae do have root hairs, the ECM association there may be fairly recent and it is certainly somewhat different from such associations in other groups (Haug et al. 2005). Tuberculate ECM, clusters of roots surrounded by hyphae, are another ECM variant (Paul et al. 2007: Pinus; M. E. Smith & Pfister 2009: Quercus). The fungi also form rhizomorphs of various morphologies that extend sometimes many metres from the plant, a distance that depends on the particular association (Agerer 2001; Hobbie & Agerer 2010; Águeda et al. 2014 and references), and these rhizomorphs are involved in foraging for nutrients (see below). Usually the Hartig net involves only the outermost layer of cells of the roots, but sometimes, as in Pinaceae, the hyphae penetrate more deeply (Brundrett 2004). The fungal hyphae are septate and are inter- but not intracellular.

Individual species of ECM fungi may be quite widespread, for example, the ascomycete Cenococcum geophilum is very common and occurs in various successional stages of temperate forests - interestingly, no spores of any kind are known from this fungus (Meyer 1964; Visser 1995; Obase et al. 2017). Overall local fungal diversity is often high and depends on the phylogenetic diversity of the host plants (Nguyen et al. 2016). Fungi in some tropical lowland ECM associations are as diverse as those in more temperate climates (Henkel et al. 2012: Dicymbe-Fabaceae; Brearley 2012: Dipterocarpaceae), and specificity is often low here, too (e.g. Vasco-Palacios et al. (2018). However, in some tropical communities where ECM plants although present, are uncommon, the specificity of the ECM fungus for the host may be quite high (Bruns et al. 2002; Timling & Taylor 2012; M. E. Smith et al. 2013; Corrales et al. 2018), although Nguyen et al. (2016) found specificity became apparent only when comparing fungi associated with different orders or higher groups of seed plants. African and Malagasy ECM woodlands showed relatively less fungal diversity, the same ECM fungi being found on unrelated species and at different successional stages (Tedersoo et al. 2011). On the other hand, there can be a distinct fungal succession as in regrowth of jack pine forests after burning (Visser 1995), or, more generally, as the forest gets older (Hagenbo et al. 2108). Specificity was reported to be highest in the early stages of other successional situations when fertility was lowest (Reverchon et al. 2012: also saprophytic fungi; Hayward et al. 2015). Tedersoo et al. (2013: Salicaceae, also sometimes AM) found that the phylogenetic relationships of the host had a strong effect on both ECM richness and community composition (see also Pölme et al. 2013). Indeed, the ECM species that associate with a particular plant species may depend on the genotype of the latter, thus the ECM associates of seedlings of pinyon pines (Pinus edulis) varied, but resembled those of their parents, which were either drought-tolerant, the basidiomycete Rhizopogon roseolus being prominent in their ECM community, or drought-intolerant, two species of the ascomycete Geopora then being prominent (Gehring et al. 2017a). This is quite a widespread phenomenon that potentially affects plant-microbe associations in general, and is further affected by the form of N in the soil, etc. (see Schweitzer et al. 2004; Sthultz et al. 2009a, b; Gehring et al. 2014; Gallart et al. 2017). Extensive shifts between unrelated plant hosts can occur within small clades of fungi - thus the ECM Strobilomyces/Afroboletus clade (Boletaceae), with perhaps 50 species, is found on five major unrelated clades of seed plants from pines to peas and eucalypts (Sato et al. 2016). However, suilloid basidiomycetes (Boletales) are pretty much restricted to Pinaceae, although they have commonly been introduced (Sato & Toju 2019; Vellinga et al. 2009). Indeed, looking at ECM introductions in general, the introduced fungus only sometimes moves on to native plants, and sometimes it may be replaced by a local ECM fungus (Vellinga et al. 2009). Interestingly, neither ECM nor ERM fungal networks are nested, that is, specialist fungi, forming associations with only a few species of plants, often do not also grow on generalist ECM plants in the same area - c.f. some studies on AM plants (see Toju et al. 2016, 2018 and references). In a broad survey of ECM specificity in European forests, van der Linde et al. (2018) found that overall 7% of the fungi were restricted to a single species of plant, otherwise, fungal species tended to associate either with members of Pinaceae or with broad-leaved angiosperms. However, it may not be that easy to interpret the significance of fungal host distributions that include both angiosperms and gymnosperms (Lofgren et al. 2018); for an overall review of the specificity of ECM symbioses, see Hoeksema et al. (2019). Finally, it may be noted that there is considerable taxonomic and geographic specificity in the association between parasitic Monotropoideae and the ECM fungi on which they depend (Bidartondo & Bruns 2001). Of course, until the advent of molecular methods, identification of the fungi depended on their being cultivable and undergoing sexual reproduction, and although Sebacinales, for instance, can now be identified using molecular data, some remain unculturable and most are undescribed (Tedersoo et al. 2010b; Weiß et al. 2016).

ECM may help the plant acquire N and P from organic material as diverse as pollen and dead nematodes (by breaking down chitin, chitinases are common in such fungi), and also from the weathering of rocks; ECM hyphae produce extracellular enzymes like acid phosphatases and are often common in the humus layer of the soil and the mineral layers immediately below that (e.g. E. A. Hobbie & Hobbie 2008; L. L. Taylor et al. 2009, 2011; Lehto & Zwiazek 2011; Koele et al. 2012; Habib et al. 2013; Averill et al. 2014; Shah et al. 2016; Rosling et al. 2016; Fernandez et al. 2015; Hawkins & Kranabetter 2017; Nehls & Plassard 2018; Steidinger et al. 2019). 61-86% of the N in plants growing in Alaskan Arctic tundra came from their ECM/ERM associates (J. E. Hobbie & Hobbie 2006, E. A. Hobbie & Hobbie 2008). N derived directly from rock breakdown can greatly increase ecosystem C storage in coniferous ECM forests, and in many of the parts of the globe where they grow the relative increase in total N coming from rock breakdown is highest, sometimes over 100%, and of course other nutrients are released at the same time (Morford et al 2011; Houlton et al. 2018) - ECM fungi are "rock-eating fungi" (Jongmans et al. 1997: p. 682). ECM fungi can produce extracellular enzymes that break down organic N (Perez-Moreno & Read 2000), and N may also be transferred to the plant in an organic form, e.g. as glutamine (Alexander 1989a; Newbery et al. 1997; Michelsen et al. 1998; Read & Perez-Moreno 2003; Cairney & Meharg 2003; Lindahl & Taylor 2004; F. Martin & Nehls 2010; Bonfante & Genre 2010; but c.f. Persson & Näsholm 2001: common in plants in vitro). Indeed, ECM fungi can switch betweem a saprotrophic and biotrophic life style (Habib et al. 2013). The fungi may retain N in their mycelium, perhaps especially when the supply of photosynthesate from the plant is high, and the result is soil with very low N concentrations that is unfavourable for the growth of non-ECM/ERM plants (Näsholm et al. 2013; see also Read 1998), although details of the N balance between fungus and host are complex (Garcia et al. 2015; Terrer et al. 2016). In general, ECM aid in the uptake of both P and N, as well as helping the plants deal with metal toxicity in the low pH soils that develop in pine-dominated communities (Read 1998), but it is unlikely that carbon moves from humus to the plant via ECM fungi (e.g. Kohler & Martin 2017). However, a recent study suggested that transfer of N from soil organic matter to the plant via ECM fungi had yet to be conclusively demonstrated (Pellitier & Zak 2017). Interestingly, litter from AM plants has a lower C:N ration than that of ECM plants and decomposes more quickly (Gehring et al. 2017b and references), nevertheless, the mineral-associated soil organic matter produced is stable and the organic N is inaccessible to the AM fungi, perhaps more so than the N in the more slowly-decomposing litter of ECM plants to ECM fungi (Terrer et al. 2016 and references).

ECM fungi are not saprotrophic, i.a. having few enzymes that break down crystalline cellulose, PCWDEs/Carbohydrate-Active enZymes/CAZymes (e.g. Michelsen et al. 1996; Jonasson & Michelsen 1996; Lindahl & Tunlid 2014; Habib et al. 2013; Kohler et al. 2015; F. Martin et al. 2016; Tunlid et al. 2017). The viability of the plant-fungus association may hinge on the ability of the fungus to utilise the humic substances that make up soil organic matter (e.g. Schmidt et al. 2011; Rosling et al. 2015; Shah et al. 2016). Similarly, they cannot break down lignin (e.g. Martin et al. 2016), ECM fungi having only a limited ability to break down lignocelluloses, although some have coopted oxidizing enzymes that, with sugar from the plant, do break down lignocellulosic material, liberating N and P which is taken up by the plant, or they use peroxidases for this purpose (Shah et al. 2016; Cheeke et al. 2016); similarly, they cannot break down lignin (e.g. Martin et al. 2016). These enzymes, normally characteristic of brown and white rot fungi respectively, have evolved functions that differ from those of the same enzyme in fungi in non-ECM associations, for example, rather than obtaining carbon they scavenge nutrients (Doré et al. 2015; Shah et al. 2016). Certainly the saprotrophic activities of ECM fungi are much more constrained than those of ERM fungi (see below: Martino et al. 2018).

In addition to their influence on plant mineral nutrition, ECM can help mitigate some negative effects that dark septate hyphal endophytes (see below) may have on the plant (Reininger & Sieber 2012). The role ECM fungi might play in helping the plant cope with drought stress is unclear, although they may help the plant get water from weathered/weathering bedrock (Lehto & Zwiazek 2011 for a review: root hairs?). Overall, ECM-associated taxa may be more sensitive to drought that AM-associated taxa (Gehring et al. 2017b). In another aspect of ECM-plant associations, wounding Populus x canescens associated with the ECM Laccaria bicolor was found to cause extensive changes to the secondary metabolites the plant produced; the new metabolites deterred the activities of the leaf beetle Chrysomela populi (Kaling et al. 2018). For other aspects of the ecological interactions of ectomycorrhizae, see below.

The association between fungus and plant is close, if not quite so close as in AM associations, although in both there has been loss of PCWDEs, etc. (Perotto et al. 2018). Around a quarter of the genes that are upregulated in ECM interactions are restricted to a particular fungal species (Kohler et al. 2015), and around a fifth of these are mycorrhizal-induced small secreted proteins, which produce effector proteins that move into the nucleus and suppress the defence pathways of the host, so allowing the ECM association to be established (F. Martin et al. 2016).

The basidiomycete Laccaria bicolor has lost the ability to produce invertase and so cannot break down sucrose (see also AM fungi), the glucose it needs being produced by the activities of the host plant invertase (Kohler & Martin 2017), and this may be true of all ECM basidiomycetes, but at least some ascomycete ECM fungi have their own invertase and so can produce glucose (Nehls et al. 2010 and references). When all is said and done, ECM fungi may acquire up to 70% of the N and P that a plant requires, and at the same time they utilize 8-34%, or perhaps even half, of the photosynthetically-fixed C the plant produces, and host allocations to the fungus being particularly high in nutrient-poor soils (Hobbie 2006; Nehls 2008; S. E. Smith & Read 2008; Högberg et al. 2010; Nels et al. 2010; Koide et al. 2013; Phillips et al. 2013; Ekblad et al. 2013; Allen & Kitajima 2014: NPP; Fernandez et al. 2015 and literature. How long EM root tips, fine roots, fungi and fungal rhizomorphs persist varies, but the latter, at least, may persist for months or even years (Agerer 2001; Allen & Kitajima 2014 and references). Indeed, as Read (1998: p. 328) noted of the Pinus/ECM association, "pine roots are simply food bases which nourish an extremely dense mycelial system". Sucrose from the plant may be convereed into trehalose in the fungus; the fungus may also produce glycogen (Nehls 2008). Since the mycelium:root ratio is something like 200,000:1 on a soil volume basis (Read 1998), the high carbon requirements of the fungus are not surprising, and ECM trees increase their photosynthetic capacity in response (Nehls et al. 2010 and references). Estimates of the increase of biomass in the host plant after inoculation with ECM fungi are 80.3% ± 27.1 SE in a meta-analysis carried out by Hoeksema et al. (2019), however, the figures were rather lower in those Fabaceae examined and higher in Myrtaceae. Interestingly, overall the benefits were lower if N was added (c.f. Fabaceae) but were higher with the addition of P. Overall, there was quite a pronounced interaction between fungal and host phylogeny, and basidiomycetes like Scleroderma and in particular Pisolithus (both Boletales) were particularly common ECM associates (Hoeksema et al. 2019).

There is probably an evolutionary sequence, white rot fungi → brown rot fungi → ECM fungi (e.g. Kohler et al. 2015), although too much should not be made of these decomposition "types"; Hibbett et al. (2000) also suggested that white-rot fungi might also be derived from within ECM clades. White rot and brown rot fungi are both wood decomposers that utilize cellulose, but only the former, widely scattered through Agaricomycetes, can mineralize the lignin so removing it from the ecosystem (Eastwood et al. 2011; Floudas et al. 2012; Kohler et al. 2015). Development of ECM associations in Boletales, at least, may be favoured by low N and high organic matter (= mostly lignin), ecological conditions that result from the activities of brown rot and to a certain extent from AM fungi. The soil organic matter accessed by ECM fungi is made up of small fragments of lignin, polysaccharides, peptides, etc. (Schmidt et al. 2011; Shah et al. 2016), however, carbon does not move from this organic matter to the fungus, it all comes from the plant (Tunlid et al. 2017).

The genome size and content of ECM fungi varies dramatically and 7-38% of the genes may be lineage specific and without known functions (Kohler & Martin 2017). Given the extreme polyphyly of the ECM habit, different clades of ECM fungi may vary in details of the nature of their associations with plants (see also Tunlid et al. 2017), but overall physiological similarities between different clades of ECM fungi can be quite extensive (Peter et al. 2016).

Vrålstad (2004; see also Villareal et al. 2004; Brundrett 2004; Imhoff 2009; Tedersoo et al. 2010b) suggested that ECM and ERM fungi form a single ecological guild, one of whose characteristics is the uptake of dissolved organic N by the plant (Lindahl et al. 2002: opposition between decomposer and mycorrhizal fungi; D. L. Jones et al. 2004: ability to take up amino acids universal?; J. E. Hobbie & Hobbie 2008: tundra, two mycorrhizal types difficult to distinguish; Clemmensen et al. 2014: ECM and ERM of different importance in forests of different ages; Talbot et al. 2008: VAM; Inselbacher et al. 2012). Ericaceae-Pyroloideae, -Arbutoideae and -Monotropoideae are associated with ECM Sebacinaceae (see also Selosse et al. 2007), but a number of ascomycetes in particular form distinctive ERM associations with many other Ericaceae (Read 1996; Garbaye 2013; Martino et al. 2018; Perotto et al. 2018; Martino et al. 2018), as do the basidiomycete Sebacinales-Serendipitaceae (Selosse et al. 2007; Imhof 2009 for a summary; Weiß et al. 2016), etc.. Importantly, species of ascomycetes can form ERM with Ericaceae as well as endophytic or ECM associations with Pinaceae growing in the same habitat (e.g. Vrålstad et al. 2000, 2002; Grelet et al. 2009; Perotto et al. 2018; Martino et al. 2018), suggesting that at least at one level the distinction between ERM and ECM fungi is not that great. Indeed, as Grelet et al. (2009: p. 364) observed, "the morphology of the mycorrhizal association (ERM or ECM) is under the control of the host plant".

ERM fungi have far greater saprotrophic abilities than ECM (or AM) fungi, and in this are more in line with plant pathogens and soil/litter saprotrophs, and they have numerous PCWDEs/CAZymes, and in general enzymes involved in secondary metabolism (Martino et al. 2018; Perotto et al. 2018). Note that such differences in saprotrophic activity, etc., occur within those species that form both ECM and ERM associations. ERM associations usually lack a Hartig net and the hyphae may be intracellular - Ericaceae have a variety of morphologically different associations with ERM fungi. Melanin, involved in stress resistance, is commonly synthesized, and it makes fungal biomass recalcitrant, like lignin, being resistant to decay, and is an important component of the sequestered carbon in at least some older boreal forests (Read et al. 2004; Clemmensen et al. 2014; Lindahl & Clemmensen 2017) - although collembola like to eat it (Fernandez et al. 2015) - yet ERM fungi can digest most organic compounds to be found in soil organic matter (Martino et al. 2018; Selosse et al. 2017c). The plant of course pays.

ERM plants are found from the Arctic to temperate areas, usually where soils are acid, and also in tropical montane regions, their distribution partly overlapping that of ECM plants. Thus Leotiomycetes, ascomycetes that are commonly ERM, become more diverse towards the poles and their ericaceous associates are also common there (Wardle & Lindahl 2014; Tedersoo et al. 2014b), indeed, ERM fungi are common and conspicuous in heath vegetation north of the boreal forest zone (Lindahl & Clemmensen 2017). ERM have been dated to ca 90 Ma (van der Heijen et al. 2015 and references), but the fossil (see Paleoenkianthus), on which this age is based, may not even be stem Ericaceae, and in any event Enkianthus itself does not have ERM; the origin of ERM - probably an apomorphy of the [[Cassiopoideae + Ericoideae] [Harrimanelloideae [Epacridoideae + Vaccinioideae]]] clade - can perhaps be dated to around 77-65 Ma (Wagstaff et al. 2010; Z.-Y. Liu et al. 2014), although this age needs confirmation (and see also Schwery et al. 2014; Martino et al. 2018 for unlikely older dates).

Mucoromycotes, Including Fine Root Endophytes. Mucoromycote Endogonales, like other AM fungi, have large genomes (96-230 Mb for four taxa examined), with numerous repeats, and also a reduction in number of PCWDEs, although there were not likely to have been that many in stem members of this clade. Interestingly, there were not that many small secreted proteins other than those specific to single species, and their transposable element content was high (Chang et al. 2018). The level of parallelism in genome size and content between Endogonaceae and the glomeromycote Glomerales and Diversisporales is high (Morin et al. 2019).

Endogonales form associations with a variety of plants, both bryophytes s.l., especially liverworts but not mosses, and a few vascular plants like Lycopodiaceae and monilophytes, and these associations are not nested (Rimington et al. 2019: 36 taxa of fungi involved, in two or perhaps five families); other species form ECM associations (see above). Haplomitriopsida (liverworts) form associations with Endogonales alone, and such associations show complex patterns of gains and losses in other liverworts, while in glomalean fungi in these liverworts, once the AM association is lost, it is not regained (Rimington et al. 2019: ?sampling). The two kinds of fungi quite often co-occur in associations that appear to be stable (c.f. Werner et al. 2018) and have considerable implications for biosphere evolution (see elsewhere); in such associations the fungal networks are nested (Rimington et al. 2019). Interestingly, mucoromycote-liverwort associations are regained only in liverworts that have maintained their AM associations (Rimington et al. 2019), and mucoromycotes are nutritionally more flexible than glomeromycotes and the liverwort associates of the former can transfer organic N to their hosts (Field et al. 2019).

The distinctive nature of the association between fine root endophytes (FREs) and land plants was recognized only quite recently (Orchard et al. 2016). FREs are also mucoromycotes and are the only group apart from AM glomeromycote fungi that also form arbuscules inside cells (Orchard et al. 2017; Hoysted et al. 2018). Note that the morphology of the FRE can vary considerably, both within an individual and between species (Hoysted et al. 2019). Mucoromycote associations have been found in fossils, including some from the Rhynie Chert some 407 Ma (Strullu-Derrien et al. 2014b). They colonize extant plants quite widely, and bryophytes, lycophytes, ferns, gymnosperms, and angiosperms have all been recorded as hosts, in the latter perhaps especially in Fabacae and Poaceae but around fourty or so families in total, and this despite various challenges in identifying and growing them (Orchard et al. 2017).

It has recently been shown that, like AM and ECM, these mucoromycotes (Orchard et al. 2016, 2017; Hoysted et al. 2018 for reviews) also form nutritional mutualisms with land plants. FREs may enhance the uptake of phosphorus in low P conditions, and they tend to be tolerant of some more ecologically extreme conditions than AMF (and they also tolerated high N conditions better) (Orchard et al. 2017). Glomeromycotes and mucoromycotes seem to play similar roles in the physiology of the plant, although the latter may have more plant cell wall degrading enzymes (PCWDEs: Field & Pressel 2018 and references). In the case of liverworts N and also substantial amounts of P may be transmitted to the host (Field et al. 2014, 2015c, 2019), but in Lycopodiella inundata N was the main nutrient involved (Hoysted et al. 2019); C moves from the host to the fungus. Interestingly, mucoromycotes associated with the lycophyte Lycopodiella both took far more C from, and gave far more N to, their host - over two orders of magnitude more - than in mucoromycote associations with hepatics (Field et al. 2015c; Hoysted et al. 2019). In general, fine root endophytes may be important in the N metabolism of their hosts.

On mycorrhizal networks. Since individual mycorrhizal fungi can form simultaneous associations with more than one plant or with more than one species of plant, or such associations may be sequential, or relationships between different members of the one network may change through the season, or the one fungus can form different kinds of mycorrhizal associations, very complex common mycorrhizal networks can form - the only combination that seems to be forbidden is associations between different genets of the one species of fungus (e.g. Bruns et al. 2002; Villareal-Ruiz et al. 2004; Simard & Durall 2004; Selosse et al. 2007; McGuire 2007a; Horton & van der Heijden 2007; van der Heijden & Horton 2009; Kjøller et al. 2010; Kennedy et al. 2012; Simard et al. 2012; Hynson et al. 2013; van der Heijden et al. 2015a; Michaëlla Ebenye et al. 2017; Wipf et al. 2019). For example, complex ECM networks may link the oldest to the youngest individuals in Pseudotsuga menziesii forests (Belier et al. 2010), although this does not necessarily mean that there is nutrient exchange between them (Michaëlla Ebenye et al. 2017), while ECM networks link tall trees of Pinus, Larix and Fagus, and they exchange photosynthesate, maybe 40% of the fine root carbon coming via mycorrhizae (Klein et al. 2016). Serendipitaceae from orchids formed ERM on Calluna vulgaris in vitro (Weiß et al. 2016). Genets of the suilloid Rhizopogon formed a complex web of associations with Douglas fir (Pseudotsuga menziesii, Pinaceae), the youngest and oldest plants being linked (Beiler et al. 2010; for ECM legumes, see Fabaceae). In forests at the Arctic treeline, ECM fungi from shrubs move on to seedlings germinating after fires, although it is unclear if direct mycorrhizal connections between the two parties are established (Hewitt et al. 2017). However, associations between seedlings and mycorrhizal adults of the same or different species can facilitate mycorrhizal establishment in the former, especially in the case of ECM (Horton & van der Heijden 2007 and literature). Generalist ECM species from the same habitat in Oregon associated with Cercocarpus ledifolia (Rosaceae) also form associations with Quercus garryana (Fagaceae) and Arctostaphylos spp. (Ericaceae) (McDonald et al. 2010). The fungal associates of Arbutus menziesii (Ericaceae-Arbutoideae), from Pacific North America and with arbutoid mycorrhizae, are diverse, and are found on other angiosperms growing in the same area, and in particular as ECM on Pseudotsuga (Kennedy et al. 2012), while in central Italy a PinusArbutus unedo shift of ECM associates has been documented, with several extra-zonal fungi that are usually associated with conifers now being found on Arbutus (Di Rita et al. 2020). Although it has been suggested that temperate ECM associations usually develop from fungal propagules in the soils (Hewitt et al. 2017), this may be particularly true of early successional situations where fungi like Laccaria develop from spores in the soil spore bank (Horton & van der Heijden 2007 and literature). There is lability in the associations in which the ascomycete Rhizoscyphus ericae (c.f. also Meliniomyces) is involved; it often grows as an ERM on the hair roots of north temperate Ericaceae and as an ECM on Pinus in the same community (Grelet et al. 2010; see also Villarreal-Ruiz et al. 2004; Martino et al. 2018); the fungus also forms mycorrhizal associations with Jungermanniales-Schistochilaceae, leafy liverworts (Pressel et al. 2006c, 2008). ECM fungi in woodlands and savannas of Africa and Madagascar are relatively uniform, and one species of fungus can form associations with plants in different families and at different successional stages, and this can facilitate ecological succession and perhaps also community shifts (Tedersoo et al. 2008, 2011). Villareal-Ruiz et al. (2004) persuaded a single fungal mycelium to form an ECM with Pinus sylvestris and an ERM with Vaccinium myrtillus at the same time, and the ascomycete Acephala, a dark septate hyphal species, showed similar behaviour, although not simultaneously (Lukesová et al. 2015). In some, at least, ECM fungal networks there may indeed be one- or two-way carbon transfer between the plant parties involved (see Robinson & Fitter 1999 for a review).

Seedlings quite commonly benefit when growing close to mycorrhiza-associated adults (e.g. Horton & van der Heijden 2007; Simard et al. 2012; Gerz et al. 2017 and references), indeed, scattered through these pages are a number of examples where the Janzen-Connell effect seems to be inapplicable to ECM associations in particular, perhaps rather less so when AM fungi are involved (Horton & van der Heijden 2007). Furthermore, stimuli activating defense-related genes may be transmitted to one plant from another via fungal hyphae, at least in some AM associations (Jung et al. 2012 and references). However, common mycorrhizal networks are not some kind of egalitarian system, rather, they may enable asymmetrical distribution of nutrients within the system. In an important early study Simard et al. (1997) found that ca 6.6% of the carbon fixed by Betula papyrifera moved to Pseudotsuga menziesii via their common ECM associate in conditions of deep shade (there was also mutual exchange of carbon), and in another case 5-15% of fixed 15N2 moved from Alnus glutinosa to Pinus contorta (Ekblad & Huss-Danell 1995), although the ecological significance of the latter is unclear. However, N and P movement through other common mycorrhizal networks has been demonstrated (Simard et al. 2012 and references). Vigorously-growing seedlings involved in such networks are sinks for nutrients, etc., and outperform smaller seedlings (Simard et al. 2012 and references), similarly, mycorrhizae of large, well-lit AM plants acquired nutrients from the soil around neighbouring plants, which as a result did not thrive as well (Weremijewicz et al. 2016: experimental set-up).

What goes on between the partners in AM-plant networks is less clear, and Robinson and Fitter (1999) found little evidence for the movement of C between plants linked by AM fungi (Plantago lanceolata with Festuca ovina or Cyonodon dactylon were two studies they examined). Networks involving Thuja plicata (Cupressaceae) and its AM fungi did suggest that these fungi in young trees came from older trees (Gorzelak et al. 2017), but an investigation of nutrient exchange was no part of this study. Liverwort gametophytes can form associations with fungi that are also ECM on the flowering plant on which the liverwort is epiphytic. Examples include Marchantia/mycorrhizal fungus/Podocarpus and the mycoheterotrophic chlorophyll-less liverwort Cryptothallus/the basidiomycete Tulasnella/Pinus-Betula (Read et al. 2000; Bidartondo et al. 2003; Kottke & Nebel 2005 and references).

Field et al. (2012) estimated the amount and efficiency of phosphate uptake by AM plants when simulating mid- to later Palaeozoic CO2 concentrations. In both mucoromycote symbioses with liverworts and glomeromycote symbioses with vascular plants (Osmunda, Plantago were the placeholders) the P gained from the fungus per unit C delivered by the plant was the same or increased as CO2 concentrations declined, and the efficiency of the symbiosis greatly increased, however, glomeromycote-liverwort symbioses were a "direct contrast", the amount of P delivered decreasing (Field et al. 2012; Hoysted et al. 2017: p. 3, 2018). Preissia (mycoheterotroph a glomeromycote) and Marchantia (?mycoheterotroph) were the liverworts, and the amount of P gained and in particular the efficiency of the symbiosis in the former was negatively affected by decreasing CO2 concentrations, so as CO2 concentrations declined in the Palaeozoic, AM associations in liverworts were perhaps likely to have become less efficient, those in vascular plants more efficient - there were of course also changes in the vascular system, stomatal size and density, rooting, etc., during this period (Field et al. 2012). Interestingly, however, the above-ground biomass of fern sporophytes decreased notably with decreasing CO2 concentrations, but liverwort gametophytes and the Plantago sporophyte showed no change, and the C allocation to AM fungal networks decreased in one of the the two vascular plants, Osmunda again, and one of the two hepatics being studied (Field et al. 2012). Integrate with Field et al. 2019....

3C. Endophytic Fungi: General. Aside from mycorrhizal associations, endophytic associations of non-parasitic (at least initially) fungi growing inside plants, are well known. Fungi in the endosphere, i.e. growing inside the plant, are often quite different than those in the rhizosphere, although at the same time some endophytic fungi may end up functioning rather like AM associations (e.g. Almario et al. 2017). Endophytes have been placed in four groups. Class one endophytes are ascomycete clavicipitaceous fungi and occur in Poaceae with which they form very close associations (e.g. Schardl 2010; Card et al. 2014). Transmission is via the seed (vertical transmission). In other endophytes, transmission is usually via fungal spores (horizontal transmission) (Arnold 2008). Class two endophytes pervade all the tissues of the plant; the fungi involved are not particularly speciose. Class three endophytes are restricted to shoots and are very diverse. Class four endophytes, dark septate endophytes, dark because of fungal melanin (see Fernandez et al. 2013), are restricted to roots (Rodriguez et al. 2009); some of these also form ECM and/or ERM and/or saproptrophic associations (Lukesová et al. 2015; Schlegel et al. 2018); Pressel et al. (2016) are sceptical that dark septate endophytes in pteridophytes, at least, are other than saprotrophic... (The "fine endophytes" described as growing in the roots of some Arctic plants [Newsham et al. 2009] have been considered to be AM associations. However, molecular analyses suggest that these fungi, which produce fan-like arbuscules and small vesicles, are mucoromycotes, not Glomus or relatives - Orchard et al. 2016, 2017; see below.)

All seed plants, but perhaps particularly Poaceae and Ericaceae (e.g. Petrini 1986, other papers in Fokkema & van den Heuvel 1986; Saikkonen et al. 2004), harbour endophytic fungi (Rodriguez et al. 2009; Hoffman & Arnold 2010; Friesen et al. 2011), and the number of species of fungi involved is probably very large indeed. Arnold et al. (2001) found 418 morphospecies of class three endophytes in only 83 leaves of two species of tropical trees, Ouratea (Ochnaceae) and Heisteria (Erythropalaceae) while Sato et al. (2015) found 1,245 ascomycetes, probably endophytes, in 442 root samples taken from dipterocarp forests in Sarawak (see also Bills & Polishook 1994; Frohlich & Hyde 1999; Arnold & Lutzoni 2007 and other articles in Ecology 88[3]. 2007; Ferreira et al. 2017). Diversity of fungi associated with single individuals of Dactylis glomerata were not reaching an asymptote (Sánchez-Márquez et al. 2007), while 21 species of trees in New South Wales had on an average ca 140 taxa/sample (Lee et al. 2019). However, Vincent et al. (2015) found rather low beta diversity of the endophytic community in New Guinea rain forest trees - relatively few generalist species, many rare species. Most endophytes are Ascomycota, Xylariaceae and Clavicipitaceae bring notably common (Rogers 2000; Sánchez-Márquez et al. 2007). A recently-described ascomycete order, Phaemoniellales, was initially based on endophytes growing in the sapwood of Peruvian Hevea (K.-H. Chen et al. 2015; see also Gazis et al. 2012). R. Martin et al. (2015) found 118 basidiomycete taxa (12% of all isolates) in leaves and sapwood from 192 collections of two species of Hevea; many appear to be wood decaying white rot fungi (brown rot fungi are at most uncommon as endophytes), and some have very wide distributions. Endophytes are particularly common in Sebacinales-Serendipitaceae (Weiß et al. 2016 and references).

The ascomycete clavicipitaceous Epichloë is a class one endophyte found in Poaceae, q.v., especially in Poöideae, and affects the host in a variety of ways, including suppression of the establishment of AM associations (König et al. 2018 and references). Thus the level of aphid infestation and that of their parasites and parasitoids can be affected by the fungus (Omacini et al. 2001). In Epichloë symbioses, the jasmonic acid pathway, involved in defence against microorganisms that kill plant tissue before getting nutrients from it and also against chewing insects (as are the metabolites synthesized by the fungus), is upregulated, as in AM assocations (Bastias et al. 2017; Harley & Grange 2008). Anatagonism between this pathway and the salicylic acid pathway causes the down-regulation of the latter, perhaps thereby making the plant more vulnerable to attacks by microorganisms that utilize living tissues and by sap-sucking insects (Bastias et al. 2017).

In other than class one endophytes, details of what the fungi are doing, let alone of any advantages accruing to any of the parties involved, are still poorly understood (e.g. see Jumpponen 2001: dark septate endophytes; S. E. Smith & Read 2008: summary; Rodriguez et al. 2009: summary; Chaverri & Samuels 2013: Trichoderma; May 2016; Koudelková et al. 2017: Rhododendron leaves; Schlegel et al. 2018; Christian et al. 2019: endophytes and N uptake in Theobroma). However, dark septate endophytes such as the ascomycete Chaetomium may be involved in making the host more tolerant of heavy metals (Haruma et al. 2017). The basidiomycete Piriformospora [= Serendipita] indica (Sebacinales-Serendipitaceae) can help the plant tolerate root hebivory (Cosme et al. 2016; see also Oberwinkler et al. 2013; Franken 2012), it alleviates salt stress and increases resistance to root and leaf diseases in barley (Baltruschat et al. 2008), and has other effects on the very wide variety of plants it colonizes, and typically it increases vegetative growth by up to ca 50% (see Franken 2012; Ray & Craven 2016; Weiß et al. 2016 for summaries and literature). Helotialean fungi (ascomycetes), common inside the root of non-mycorrhizal Arabis alpina, are involved in the uptake of P in low-P soils and consequent improved growth of the plant, and they seem effectively to replace the AM fungus; CAZymes were notably diverse, perhaps being involved in plant cell wall degradation (Almario et al. 2017). Hiruma et al. (2016) looked at the association of another ascomycete endophyte, Colletotrichum tofieldiae, with Arabidopsis and found a similar relationship; indole glucosinolates were also part of the story. Fungi may affect seed germination and survival (U'Ren et al. 2009), and endophytic associations are probably at least intermittently mutualistic (Carroll 1988, 1995); they may facilitate stress tolerance in the host (Rodriguez & Redman 2008).

The effects can be complex. Metarhizium robertsii (Clavicipitaceae: see also below) can help its host plant alleviate salt-induced oxidative stress (Behie & Bidochka 2014). Van Bael et al. (2009) found that leaf-cutter ants seemed to dislike plants with numerous endophytes. Beetles fed high-endophyte Merremia umbellata (Convolvulaceae) leaves seemed to do fine, but their chances of getting captured by predatory ants increased nine times (Hammer & Van Bael 2015). Endophytic fungi have also been implicated in the protection of the host plant against fungal pathogens (L. L. Martin et al. 2012: in vitro; Raghavendra & Newcombe 2013). Lignases have been detected in a number of ascomycetes endophytic in Uruguayan Myrtaceae (Bettucci & Tiscornia 2013), while metabolites synthesized by endophytic fungi in Carapa guianensis may be antibacterial, antiviral, or trypanocidal (Ferreira et al. 2015). Interestingly, saliva from herbivores may disrupt the endophyte-plant relationship, for instance, by reducing the amount of toxic alkaloids produced by the endophyte (Tanentzap et al. 2014). There are several endophytes in Bromus tectorum, introduced into North America, and one of them, Fusarium c.f. torulosum was the preferred food of the fungivorous nematode Paraphelenchus acontioides, the nematode apparently increasing the abundance of the fungus in the plant, although without any apparent ill effects to the latter (Baynes et al. 2012).

Dark septate endophytes, rich in the refractory melanin, may decrease the growth of conifers with which they are associated, but the effect depends on host, temperature, and association of the plant with ECM, just to begin with (Reininger & Sieber 2012). They may also facilitate the uptake of N (Newsham et al. 2009; Newsham 2011), indeed, Bjorbækmo et al. (2010) and Timling & Taylor (2012) noted the high frequency of melanized fungi and dark septate endophytes (not mutually exclusive categories) in high northern latitudes; dark septate endophytes have been found up to 82oN and 77o S (Newsham et al. 2009). N has been found to move from caterpillars parasitized by the ubiquitous insect pathogen Metarhizium (Clavicipitaceae, related to grass class 1 endophytes) to grasses and Fabaceae-Faboideae, plants in which the fungus is also an endophyte (Behie et al. 2012). Mutualisms can get complex here: Metarhizium can be free-living, endophytic in a wide variety of plants, some species preferring trees or herbaceous plants, at least locally, and/or a parasite of insects, and N may move to the plant from the parasitized insect and photosynthate from the plant to the endophyte (Wyrebek et al. 2011; Behie et al. 2012, 2017). Indeed, Metarhizium and fungi like the related Beauveria are widely distributed, parasitize numerous species of insects, are similarly promiscuous when it comes to being endophytes, and have been shown to benefit the plants with which they are associated; they may be important elements of the global N cycle (Behie & Bidochka 2014). One of the commonest endophyes, globally distributed, is the asmomycete-Helotiales PAC species complex - the Phialocephala fortinii- Acephala applanata species complex. This is both pathogenic, saprophytic on dead roots, has over 20,000 protein-coding genes, including numerous CAZymes and enzymes involved in secondary metabolism including melanin synthesis, etc. (Grünig et al. 2008; Schlegel et al. 2016).

Interestingly, fungal endophytes in a group of Australian trees tended rather strongly to show positive interactions and so were often found together, although there were some negative interactions between fungi that had the same trophic mode (Lee et al. 2019).

How the complex interactions betweeen endophytic fungus and host are set up and maintained has been studied in some detail in Serendipita indica. Within minutes of the initial contact of fungal chlamydospores with the plant there are changes in gene expression in the host, changes that go on for weeks as the fungus establishes itself in the plant (Weiß et al. 2016 and references). Some endophytes affect the rest of the endophytic community. Thus the toxic indolizidine alkaloid, swainsonine, is synthesised by the endophyte Undifilum (= Alternaria), an imperfect stage of Pleosporaceae, Dothidiomycetes, and Harrison et al. (2018) found that general endophyte richness was inversely related to plant size and the presence of Alternaria. Like ECM associations, the biotrophy involved in being an endophyte may be achieved by modifications in saprotrophic ancestors (Weiß et al. 2016). However, the extensive reprogramming of the Poa host genome by Epichloë endophytes, at least, is more like what happens when pathogenic fungi attack the plant, whereas mycorrhizal fungi have a much smaller effect (Dupont et al. 2015).

The life styles of the immediate relatives of fungal endophytes are various. Endolichenic fungi, fungi living inside lichens (but not the lichenising mycobionts whose association with algae constitutes the lichen thallus), and endophytic fungi may be phylogenetically close, while lichenising and endophyte clades tend to be exclusive (Arnold et al. 2009b; U'Ren et al. 2012; K.-H. Chen et al. 2015). Sebacinales (basal basidiomycetes) are notably diverse ecologically (Selosse et al. 2009; Weiß et al. 2012, esp. 2016), and the immediate relatives of the endophytes there may be saprotrophic taxa, although these are little known. Grass endophytes (related to ergot, also ascomycetes) are perhaps derived from insect pathogens, and some species of fungi are both pathogen and endophyte (Spatafora et al. 2007; Sasan & Bidochka 2012;). As mentioned above, Metarhizium and relatives (ascomycetes, Clavicipitaceae) can be both endophytes and insect parasites (Behie & Bidochka 2014). In Colletotrichum the difference between being a pathogen and an avirulent endophyte may be a single gene (Schulz & Boyle 2005). In general, endophytic fungi are often more related to pathogens and other endophytes than to saprotrophic fungi (U'Ren et al. 2009), and they are sometimes derived from necrotrophic fungi, parasitic fungi that first kill the host cell before digesting it (Delaye et al. 2013; García-Guzmán & Heil 2013; see also Chen et al. 2015). Trichoderma (Hypocreales, an ascomycete) has various associations with plants and other fungi, and endophytes may become saprotrophic or mycoparasitic on the death of the host (Chaverri & Samuels 2013; see also Schulz & Boyle 2005 and references; Oberwinkler et al. 2013). Fusarium graminearum behaves more as a vertically-transmitted endophyte on both C3 and C4 North American grasses, with which it may have evolved, and produces only trace amounts of trichothecene toxins, while on introduced wheat and barley it behaves as a pathogen and produces substantial amounts of the toxins (Lofgren et al. 2017). There is more on the plasticity of expression ef endophyte-forming fungi in the paragraph below. Given all these life styles adopted by the relatives of endophytes, even endophytes themselves, and the complexity of the plant-microbe interactions they suggest, it seems only reasonable to talk about "the endophytic continuum" (title of Schulz & Boyle 2005), or, as they say (ibid.: p. 675), "many endophytes seem to be masters of phenotypic plasticity: to infect as a pathogen, to colonize cryptically, and finally to sporulate as a pathogen or saprophyte.".

3D. Mycorrhizae and Endophytes in General. All these associations result in a complex web of interactions between plant and microorganism. Thus a single species of plant or even an individual plant may have a variety of associations with fungi, or one fungus can form different kinds of associations with different species of plants, or the nature of the association changes as the plant ages or soil nutrients change, and this is as true of fungal associations of a liverwort as of a grass - flexibility in the interactions of plant and fungus may be of advantage to both parties (Gerz et al. 2017; Nelson et al. 2018; Martino et al. 2018; Teste & Laliberté 2018, see also above). The distinction between different mycorrhizal or endophyte "types" can be less than clear-cut, as is the very distinction between having or lacking mycorrhizae and being a parasite or a mutualist (e.g. Lekberg et al. 2005; Gao & Yang 2010; esp. Vrålstad 2004; Perotto et al. 2012; Peterson 2012; Maherali et al. 2016; Brundrett 2017b: AM fungi and endophytes; Almario et al. 2017; Toju et al. 2018). It is not uncommon for a single species of plant to have more than one kind of mycorrhizal association, thus Salix may harbour ECM and/or AM fungi, not to mention dark septate endophytes (e.g. Van der Heijden 2001; Becerra et al. 2009; see also Poole & Sylvia 1990; Molina et al. 1992; Bidartondo et al. 2011: thalloid liverworts; Desirò et al. 2013: hornworts; Bennett et al. 2017: Supplement for AM/ECM associations; Tedersoo & Brundrett 2017; Tedersoo 2017b). Dual ECM/AM associations may be particularly common in Australia (Brundrett 2017a), although it has been suggested that in dual associations, AM are most common early in the plant's growth, ECM predominate latterly (Adams et al. 2006; Gerz 2017 and literature). Similarly, the one species of fungus can form more than one kind of association with plants (e.g. Vrålstad et al. 2000, 2002; Grelet et al. 2009; Perotto et al. 2018; Martino et al. 2018), and this is especially true of fungi that can form both ERM and ECM associations. Tuber, normally ECM, may also form endophytic associations, perhaps the ancestral state of the ECM habit, at least here, and also formed ascocarps, contributing to them as the maternal parent (Schneider-Maunoury et al. 2019). As mentioned before, the line between mutualism - or at least prolonged symbiosis - and parasitism is a fine one (Rogers 2000; Schulz et al. 2002; Eaton et al. 2010 and references; Oberwinkler et al. 2013), thus AM fungi in some grasses efectively become parasites at high soil P concentration (Grman 2012 and literature) and some fungal associations may be both harmless endophyte, parasite and saprotroph (e.g. Schlegel et al. 2016). Similarly, Rasmann et al. (2017) discuss the complex interactions between fungi in the roots, plant growth-promoting rhizobacteria, etc., and plant herbivores and their parasites. Some ERM fungi may form endophytic associations, and some endophytic fungi are physiologically quite similar to ERM fungi (Martino et al. 2018). Weiß et al. (2012) found the same nuclear LSU sequences of basidiomycetous Sebacinales in taxonomically unrelated plants growing in different areas, sometimes the association was mycorrhizal, sometimes endophytic, and they suggested that Sebacinales might play an important role in ecosystem integration. As mentioned, Sebacinales are poorly understood, but their members have a variety of relationships with plants (Weiß et al. 2009, 2016; Oberwinkler et al. 2013; Varma et al. 2013). Members of the Phialocephala/Acephala species complex may be dark septate endophytes, ECM or ERM (e.g. Grünig et al. 2008; Lukesová et al. 2015). Chambers et al. (2008) found that dark septate endophytes on a variety of plants in southeast Australian sclerophyll vegetation may also be ERM fungi on Epacridoideae there. Not surprisingly, ascomycetous fungal isolates from ECM and ERM plants and from dark septate endophytes may more or less interdigitate on phylogenetic trees and show little phylogenetic divergence (e.g. Vrålstad et al. 2002; Walker et al. 2011; Perotto et al. 2012). Understanding how a plant that is the host of different groups of symbionts simultaneously - often called tripartite symbioses - reacts to their presence is difficult, especially when aspects of the environment such as water and nutrient status are taken into account (e.g. Villareal-Ruiz et al. 2004; Grelet et al. 2009; Larimer et al. 2010, 2012; Worchel et al. 2013; Augé et al. 2015).

AM presence and various aspects of root architecture such as root thicknesss, root branching, and the development of root hairs are linked (Baylis 1975; Schweiger et al. 1995; B. Liu et al. 2015), although it can be difficult to make connections between root attributes and mycorrhizal status/plant response to mycorrhizal establishment (Maherali 2014). Species with thin roots may tend to forage for nutrients directly via their roots while species with thicker roots forage more indirectly via mycorrhizae, and members of the first group regenerated roots after pruning more than did the latter, although mycorrhizal colonization in the latter was higher (B. Liu et al. 2015). Thick-rooted species may be particularly common at least in the magnoliids and the [Taxaceae + Cupressaceae] clade (e.g. Kong et al. 2015; B. Liu et al. 2015). AM root architecture - complex/branched (finer) versus simple/little branched (thicker) linked with identity of the Glomus AM partner affects infection by the pathogen Fusarium oxysporium (Sikes et al. 2009). Notably fine roots are associated with ERM (modified ECM), while a few measurements of presumably ECM Fagaceae also suggested roots on the thinner side (Liu et al. 2015). In a small sample of temperate trees growing in an environment with patchy distribution of nutrients, AM trees with fine roots exploited these patches by producing more roots, while in ECM trees exploitation was via mycelial growth; the three members of Pinaceae (ECM) and two magnoliids (VAM) included had the thickest rootlets and none exploited the nutrient patches with any precision (W. Chen et al. 2016, 2017). However, a general survey of root thickness, root order, and possible links with root longevity and rate of decay, mycorrhizal type, infection by pathogens, different ways of acquiring nutrients, etc., is in order, although making sense of the correlations obtained can be difficult (Clowes 1951; Silva & Miya 2001; Bardgett et al. 2014; McCormack et al. 2015, 2017; Laliberté et al. 2015; Laliberté 2017 for some connections and the need to standardize measurements; esp. Iversen et al. 2017 - see FRED, the Fine Root Ecology Database at http://roots.ornl.gov; Valverde-Barrantes et al. 2017: strong phylogenetic signal in many root traits, confounding effects major...; Maherali 2017: taxa notably (in)variable in root width). Freschet et al. (2017) provide a preliminary analysis (see also W. Chen et al. 2017: root density, thickness, and mycorrhizal type). Early-evolving plants may have thicker first-order roots and mycorrhizal associations, and there is a connection between the herbaceous habit, thin roots, high specific root lengths (i.e. long roots per unit of biomass), and absence of mycorrhizae (Ma et al. 2018: mycorrhizal type not discussed). Variation in families like Euphorbiaceae is extreme (Kong et al. 2014) and first-order "adventitious" roots in epiphytic Orchidaceae and some woody monocots like Pandanaceae are notably thick (see also Pagès 2016 for descriptions of branching patterns, root thickness, etc.), as thick or thicker than the much higher order "thick" roots (>5 mm diameter) in studies like those of Silva and Miya (2001). Some distinctive root morphologies like proteoid roots, cluster roots of determinate growth, in Proteaceae and other families and bottlebrush-like dauciform roots (these are roots with exceptionally long root hairs) are responses of the plant to nutrient-poor conditions (e.g. Schweiger et al. 1995; Neumann & Martinoia 2002; Shane et al. 2004b, 2006; Playsted et al. 2006; Gao & Yang 2010).

Extant forests made up of AM trees are often more diverse than forests dominated by ECM trees (Malloch et al. 1980; McGuire 2007a; Peay 2016). There are rather few species of angiosperms in ECM-dominated communities in temperate and in particular boreal areas (see below). Laliberté et al. (2013) discuss the diversity of vegetation associated with ECM in terms of the youth of the soils; tropical soils are older, more strongly weathered, and the vegetation more diverse. However, tropical ECM dipterocarp-dominated forests are as diverse as tropical forests anywhere (Read 1991, 1996; Page et al. 2012).

Associations between plant and fungus may become very close. Thus mycoheterotrophic plants lack chlorophyll entirely and are nutritionally dependent on the activities of their associated fungi. Glomeromycotes are the associates in some mycoheterotrophic plants including the gametophytes of lycophytes and ferns and the sporophytes of some monocots, Gentianaceae, etc. (e.g. Bidartondo et al. 2002; Merckx et al. 2012; Jacquemyn & Merckx 2019). Fungi from the Glomus A group are involved (Schußler et al. 2001; Winther & Friedman 2009). Such associations with glomeromycotes may be sensitive to the amount of P in the soil, as was found in a Panamanian study when they disappeared when soil P increased above 2 mg P kg-1 (Sheldrake et al. 2017), rather as increasing soil P is known to negatively affect the abundance and behaviour of AM fungi in general (N. C. Johnson 2009). Overall, up to 90% of the plant's P may come from AM fungi (Sheldrake et al. 2017). Basically, mycoheterotrophic plants are cheaters of AM or EM associations, and tend to grow in humid, shaded, low-fertility conditions conditions where soil pH, nitrate, N:K ratios are low, etc. (Gomes et al. 2019). In mycoheterotrophic (and mixotrophic) Ericaceae and Orchidaceae modified ECM associations are common, as well as are associations with saprotrophic fungi, and these allow the plant to get carbon indirectly from other angiosperms also associated with the fungus or from decaying organic matter (e.g. Perotto et al. 2012; Oberwinkler et al. 2013; Hynson et al. 2013). Mycoheterotrophic plants can seem to be very rare, although since they spend most of their life underground and the plants may be small and inconspicuous even when in flower it is difficult to be sure, and their overall distributions can be quite wide (Gomes et al. 2019); it will be interesting to see what the distributions of the numerous Thismia species currently being described turn out to be! However, their rarity can sometimes be linked to the restricted distributions of their fungal associates (Merckx et al. 2013b). Of course there is a continuum between autotroph and echlorophyllous obligate mycoheterotroph (e.g. Johnson et al. 1997; Johnson 2009; Simard et al. 2012; Jacquemyn & Merckx 2019). Interestingly, in mixotrophic Orchidaceae carbon acquired from the fungus may be more or less restricted to below-ground parts of the orchid (the future mycoheterotroph!), while that fixed via photosynthesis of the plant remains in the above-ground parts of the orchid (see below), and in general, the specificity of the fungal association is correlated with the amount of carbon supplied by the fungus - and o See the papers in Merckx (2013a) for mycoheterotrophy in general.

Maherali et al. (2016) discuss evolutionary changes between ECM and AM associations, and also the non-mycorrhizal condition, in seed plants. There are few transitions between the three conditions, mycorrhizal associations being rather stable (see also Kiers & van der Heijden 2006). Thus Rimington et al. (2019) in their study of mycorrhizal associations in hepatics noted that there was perhaps one origin of glomeromycote associations there, and one or two losses, but no re-development of this association, however, mucoromycote associations with liverworts were notably more labile. However, there have been a number of suggestions that endophytism may be a precursor to the mycorrhizal habit (e.g. Selosse et al 2009; Strullu-Derrien at al. 2018; Martino et al. 2018, etc.). As mentioned, the loss of mycorrhizae is a feature of clades that are usually herbaceous (Proteaceae are an exception) or aquatic, and it may be associated with a higher speciation rate (Maherali et al. 2016 and references). This loss is also associated with thinner roots, etc. (Ma et al. 2018).

3E. Further Complexities. Bacteria, including N-fixing bacteria (see the N-fixing clade, are also common endophytes, part of the endosphere, i.e. the microorganisms growing inside the plant. Thus a diversity of bacteria and fungi have been isolated from the latex of Euphorbia spp., a habitat that one would not have thought to be notably suitable for such organisms (Gunawardana et al. 2015). As with fungi, the line between parasite and pathogen is not sharp, and bacteria may move from e.g. hemipterans associated with plants to the plants themselves (Caspi-Fluger & Zchori-Fein 2010). Overall, the microbiome of the endosphere varies more from species to species and is more linked with host plant phylogeny than is that of the rhizosphere (Fitzpatrick et al. 2018). Under drought conditions, or with abiotic stress in general, the composition of the microbiome in the endosphere may change, although the implications of this are unclear (Fitzpatrick et al. 2018). Carrión et al. (2019) found that bacteria like Chitinophaga and Flavobacterium in the endosphere became enriched upon invasion by a fungal pathogen (Rhizoctonia solani), activities associated with fungal cell wall degradation being stimulated and fungal root disease being suppressed. A great variety of bacteria and other microorganisms flourish in the rhizosphere, the surface of the root and the soil immediately surrounding it (e.g. Hardoim et al. 2008; Lemaire et al. 2011b). It has even been suggested that plants may derive N, etc., from bacteria that enter the root and then are digested by the plant (Paungfoo-Lonhienne et al. 2010); rhizosphere bacteria may facilitate the uptake of organic N (J. F. White et al. 2015); and bacteria have been implicated in fixing N in tuberculate ECM on conifers (Paul et al. 2007). Indeed, bacteria can be very important for ECM (and AM) associations, whether facilitating the establishment of mycorrhizal associations (mycorrhization helpers) or being integral to its subsequent functioning (mycorrhizal helpers), being involved in N fixation, solubilization of nutrients in the soil, protection against root pathogens, and the like (Garbaye 1994; D. L. Jones et al. 2004; Frey-Klett et al. 2007; Müller et al. 2016; Devaux & Labbé 2017: mycorrhiza helpers). However, interactions between plants and rhizosphere microorganisms are poorly understood, although substantial amounts of carbon (maybe 3-5% GPP) may move from the root to the soil (Jones et al. 2004) - and then one has to add the larger amounts of C moving from plant to mycorrhizal fungus (see above) to understand overall C balances. In general, bacteria can affect the growth of the plant, whether by fixing N, deterring pathogens, or the like (Kembel et al. 2014 for references; Pennisi 2015).

Both bacteria and fungi are also found in the phyllosphere, the above-ground surface of the plant (Leveau 2006, Vacher et al. 2016 for reviews). Thus some 7,300 bacterial OTUs were found on 57 species (ca 420 OTUs/tree) from Barro Colorado Island, Panama, and there may be some 11,615 bacterial OTUs in the phyllosphere of plants on that island alone (Kembel et al. 2014: use of current phylogenetic ideas does not affect results - S. W. Kembel pers. comm. xi.2014; Griffin & Carson 2015 for a review). Indeed, it has been estimated that there are some 1026 bacteria in the phyllosphere (see Leveau 2006). Similarly, several hundred species of fungi have been found in the phyllosphere of the temperate Quercus macrocarpa (Jumpponen & Jones 2009). The combined fungal phyllosphere/endophyte community may be stratified in the tree canopy, and there may also be intraspecific geographic variation (Harrison et al. 2016), but little is known about such things - in/on Sequoia sempervirens, for example, basidiomycetes were surprisingly common (Harrison et al. 2016), Amend et al. (2019) found that the diversity of the phyllosphere on Hawaiian Scaevola taccada was nested, rhizosphere > soil > ... > fruit, while Chi et al. (2005) tracked the migration of endophytic rhizobia in rice plants from the roots to the leaves. For a brief review, see Laforest-Lapointe and Whitaker (2019).

Fungi associated with plants may have endosymbionts themselves (Frey-Klett et al. 2007). Mycorrhiza-plant associations often include bacteria as additional partners, whether growing on the surface of the mycelium and synthesising crucial metabolites or living within the hyphae. The bacterium Candidatus Glomeribacter gigasporarum (near Burkholderia) is found in the AM fungus Glomus (Castillo & Pawlowska 2009, 2010; Bonfante & Genre 2010). Such bacteria can affect the growth of the fungi, and they may be vertically transmitted like the fungal genome itself (Bianciotto et al. 2003; Hoffman & Arnold 2009). Numerous bacteria (mostly Proteobacteria) and a diversity of fungi have such relationships, although they sometimes seem to be rather casual (Hoffman & Arnold 2010; Oberwinkler et al. 2013). Viruses in endophytes may affect the ability of the host plant to grow in particular conditions. Thus Márquez et al. (2007) found that only when the endophytic fungus Curvularia protuberata (Pleosporaceae) was infected with a double-stranded RNA virus was Dicanthelium lanuginosum, the host of the fungus, able to grow in volcanically-heated soils in Yellowstone at ca 65o C.

It turns out that several distinctive "plant" metabolites such as indolizidine (swainsonine) and ergoline alkaloids are not synthesized by the plant itself, but by fungal or bacterial associates of the plant; they are toxic to animals and presumably protect the plant (e.g. Popay & Rowan 1994; Tan & Zou 2001; Gunatilaka 2006; Kusari et al. 2012; Markert et al. 2008; Wink 2008; Schardl et al. 2013: Convolvulaceae; Pryor et al. 2009: Fabaceae; Wink 2008), and Celastraceae and especially Poaceae are also distinctive in this regard. Such compounds seem to be ordinary plant metabolites (Zhang et al. 2009; Friesen et al. 2011), and they are also of considerable interest for those working on applied uses of "plant" compounds. Similarly, endophytic bacteria are involved in selenium (Se) uptake by Se-accumulating plants (Lindblom et al. 2013; Sura-de Jong et al. 2015), while "true" plants secondary metabolites like terpenoids and quinolizidine alkaloids are produced more or less exclusively in mitochondria and/or chloroplasts, i.e. in endophytic bacteria whose associations with plants are very ancient (Wink 2008).

In any event, there are numerous important ecological linkages between the aboveground and the belowground (Wardle et al 2004 and references). . Indeed, mycorrhizal relationships, and these mutualistic interactions in general, call into question exactly what an individual "is" and what is the unit of selection. Thus rather than thinking of individual plants and their genomes it may be more useful to think of plants as holobionts with hologenomes, a microcosm composed of plant, bacteria, fungi, etc. (so maybe not so micro) that makes up the holobiome, and the fungal associate may affect aspects of the plant such as floral morphology and hence pollination (Bordenstein & Theis 2015; Vandenkoornhuyse et al. 2015; Gilbert & Tauber 2016; Gundel et al. 2017; Gehring et al. 2017a, b: fungus, pine genotype, water balance; Hawkins & Kranabetter 2017: N uptake and ECM; Tripp et al. 2017a: symbiome; Brody et al. 2019: the effect of host genotype on AM and ErM associationx). Indeed, as Gehring et al. (2017a: p. 11170) noted, "ECM community composition represents a heritable plant trait", while earlier Rayner (1915: p. 128) had noted that many fungus-dependent ericaceous species had "solved the problem of growth upon the poorest and most unpromising soils, but have solved it at a price of their independence", to which Pirozynski and Malloch (1975: p. 162) added "they and other land plants never had any independence, for if they had, they could never have colonized the land".

4. Angiosperm History I - Evolution in stem group angiosperms.

4A. Relationships. The angiosperm stem group probably diverged from other seed plants by the late Palaeozoic (Moldowan et al. 1994; E. L. Taylor et al. 2006; Lutzoni et al. 2018: (395-)365(-338) Ma). So how might the heterosporangiate strobilus with short internodes with ovules enclosed in a carpel that is the angiosperm flower today have evolved from the separate male and female strobili with ovules borne free on ovuliferous scales of most gymnosperms, living and fossil? Is the flower a monoaxial structure, as in most cycad strobili, or is it polyaxial, as in a pine cone, or are the flowers of some groups monoaxial and those of other groups polyaxial? The latter idea might suggest that angiosperms are polyphyletic (see Friis et al. 2011: pp. 141-144 for a summary of such ideas), and X. Wang and Wang (2010) and Z.-J. Liu and Wang (2015) indeed suggest that angiospermy may have arisen more than once and Fu et al. (2017/2018) toy with the idea. Indeed, the envelopment of the seed to produce a fruit-like structure is likely to have happened independently in a number of groups like Bennettitales, Doyleales, and angiosperms (Rothwell & Stockey 2010, 2016 and references; see also Friis et al. 2011; Tomlinson 2012 and angio-ovuly). Pollen tubes in Pinales/Gnetales and in angiosperms also probably evolved independently since at least some glossopterids had multiciliate male gametes (Nishida et al. 2004; c.f. Lee et al. 2011: cilia lost and regained?).

Candidates for relationships along the stem of angiosperms include Corystospermales (Pteruchus, Ktalenia, etc.: Frohlich & Parker 2000), Bennettitales and Caytoniales; note that the first two are known fossil from deposits in Jordan dated as late Permian (Blomenkemper et al. 2018). In the much-discussed Caytonia, poorly known, from the younger Mesozoic, the ovule is borne inside an inverted cupule, the cupule wall being equated with an outer integument - with interesting implications for pollination, although, like most other gymnosperms, the pollen grains end up at the micropyle (e.g. J. A. Doyle 2006, 2008b; Doyle & Donoghue 1986a, b, 1992; Doyle & Endress 2010; Friis et al. 2011). Caytoniales have been linked with Umkomasiales (= Corystospermales, including Doyleales) and Petriellales (and also Ktalenia) as the CUP group (Friis et al. 2019d). Relationships between corystosperms and Ginkgo are quite likely (Shi et al. 2016). Here the cupules, often in the axil of a bract, may almost completely enclose the seed, but with a transverse, slit-like opening, or there are two or more flaps in the outer part of the cupule, the seeds escaping through this valvate structure, and there are one-many seeds per cupule, and such CUP plants are widely known from both hemispheres in the Early Cretaceous, persisting to the Late Cretaceous (leaves of Sagenopteris, = Caytoniales) (Friis et al. 2019d). Pteridosperm groups that have been linked with angiosperms include Pentoxylon and glossopterids, the latter a poorly known group in which ovules are borne on the leaf (Friis et al. 2011 for a summary), and also the diminutive Petriellales, Peltaspermales, and Doyleales (see above), with compound seed cones (Taylor & Taylor 2009; Rothwell & Stockey 2016).

Seed morphology and anatomy in particular, but also pollen morphology, suggest that Bennettitales, Erdmanithecales and Gnetales in particular should be placed together (the BEG group), and Caytonia may also belong here (Friis et al. 2007, 2009a: four new genera in this complex, 2011: see especially chapter 5, 2013, 2019c; Mendes et al. 2010). Note that some Jurassic-Early Cretaceous fossils described as angiosperms seem to have a better home in the gnetalean area (Herendeen et al. 2017 and literature). Chlamydospermous seeds have an ovule with a long integument, a nucellus at least half fused to the integument, and a seed surrounded by a sclerenchymatous envelope developed from a structure outside the integument; method of micropyle closure and leaf insertion are other possible similarities (Friis et al. 2019c). There are some 16 genera and 28 species of plants with such seeds that have been described from the early Cretaceous alone (perhaps 20 genera and 50 species all told), but they soon became extinct (Friis et al. 2019c). One of these, Ephedrispermum, even has ephedroid pollen in the micropyle. Note, however, that Rothwell and Stockey (2013), Pott (2016), and others suggest alternative interpretations of such fossils. Drewria, from the Early Cretacous in Virginia (Crane & Upchurch 1987), and some other chlamydosperms seem to have been quite small plants of open habitats.

Bennettitales were especially common in the Jurassic (although they may have persisted to the Oligocene - McLoughlin et al. 2011), Erdmanithecales persisted into the Late Cretaceous, while Gnetales, also diversifying considerably in the Jurassic-Cretaceous (e.g. Y. Yang et al. 2020), are still extant. The reproductive morphologies of some of the early (Upper Triassic) Bennettitales are rather different from those of later fossils (e.g. Pott et al. 2010), and the interpretation of their complex reproductive structures is not easy (see Crane & Herendeen 2009 for careful analyses) Thus Crane and Kenrick (1997) suggest that the interseminal scales of Bennettitales are sterile [ovule + cupule] complexes. Stockey and Rothwell (2003) noted that in Williamsonia pollination appeared to be siphonogamous, there was no pollen chamber, and a nucellar plug filled the micropylar canal. A more or less close phylogenetic association between Cycadeoids or Bennettitales, so-called "fossil beehives", and angiosperms has long been mooted (see also J. A. Doyle 2006 and Hilton & Bateman 2006 for morphological cladistic analyses and literature; Little et al. 2014). Interestingly, the triterpenoid oleanane, found pretty much throughout angiosperms, also occurs in Bennettitales, but it is also scattered elsewhere (Moldowan et al. 1994; E. L. Taylor et al. 2006; Feild & Arens 2007; see also Banta et al. 2017), so it is not easy to interpret the significance of its presence.

Establishing the relationships of Gnetales has been difficult. In the 1980s and '90s morphological phylogenetic studies suggested that extant seed plants were probably to be placed in five groups: cycads, Ginkgo, conifers, Gnetales (Gnetum, Ephedra and Welwitschia), and angiosperms. Extant gymnosperms were thought to be paraphyletic, the botanical equivalent of reptiles. Plants with a heterosporangiate strobilus, the anthophytes, included flowering plants, Gnetales, and also fossil gymnosperms like Bennettitales; the glossopterid seed ferns were also thought to be fairly close. These groups were embedded in a paraphyletic assemblage made up of conifers, cycads, etc. (e.g. Crane 1985a, b; Doyle & Donoghue 1986a, b; Nixon et al. 1994; Taylor & Hickey 1995; Doyle 1998a, b; Friis et al. 2011 for a good summary); Doyle (in Sanderson et al. 2000: p. 783) noted that this position was "well supported" in bootstrap analyses that were carried out subsequently. As mentioned, various fossil groups, especially Erdmanithecales and Bennettitales, are commonly associated with Gnetales, rather complicating our understanding of relationships and evolution. The evolution of features such as insect pollination and endosperm have then been interpreted in the context of the anthophyte hypothesis (Lloyd & Wells 1992; Friedman 1995), and indeed this phylogenetic context has not lost its appeal (Rudall & Bateman 2019).

Analyses of morphological data, which may include fossil taxa, continue to suggest that extant gymnosperms are para/polyphyletic, the five main groups (Cupressales, Pinales, Gnetales, Ginkgoales and Cycadales) being independently derived from plants of a pteridosperm grade, with Gnetales in particular close to angiosperms and often associated with Bennettitales and their like. Thus they support some kind of anthophyte hypothesis (Ye et al. 1993; Rydin et al. 2002; Doyle 2006; Hilton & Bateman 2006; Rothwell et al. 2009; Schneider et al. 2009; Crepet & Stevenson 2009, esp. 2010; Friis et al. 2007: seed morphology, 2011: summary, 2013a; Zavialova et al. 2009: pollen, walls homogeneous or granular; Rothwell & Stockey 2016). However, bootstrap support for such relationships was low (e.g. Doyle 2006; Hilton & Bateman 2006; Rothwell & Stockey 2016). In one study possible relationships among seed plants even included a paraphyletic Gnetales, with angiosperms sister to [Gnetum + Welwitschia]; [Archaefructus + Ceratophyllum] were sister to all other angiosperms (S. Wang 2010: e.g. Fig. 8.10), although this would seem to be rather unlikely.

Doyle (2006) studied seed plant evolution in the context of a morphological analysis that was constrained by a (molecular) topology in which Gnetales were nested within gymnosperms; he noted that this was almost as parsimonious as if Gnetales were linked with angiosperms. Recent work shows that even morphological analyses now place Gnetales with conifers; careful reconsideration of the morphological characters thought to show relationships with angiosperms in the context of a possible relationship with conifers and use of Bayesian, not just Maximum Parsimony, methods of analysis have driven the shift. Coiro et al. (2017/8) provide a careful analysis of this whole problem, and as they noted of some character reconsiderations, "These changes in character definition do involve a subjective element and were doubtless influenced by knowledge of the molecular evidence for a relationship of Gnetales and conifers,..." (ibid.: p. 21/504) - indeed, knowing where you want to go does help.

In fact, some characters that Gnetales and angiosperms have in common fail to meet one or more of Remane's three criteria of similarity ("homology"), those of position, special properties, and intermediates. Thus the sieve areas in the phloem cells of Gnetales are very like those of other gymnosperms and are unlike those of the sieve tubes of angiosperms (Behnke 1990a). Vessels in Gnetales develop from circular pits and those in flowering plants from scalariform pits (e.g. Rodin 1969; Carlquist 1996), and although Muhammad and Sattler (1982) suggested that in Gnetum, at least, the distinction was not so clear, there is some evidence from whole genome analyses that vessels in Gnetum and in angiosperms have little immediately to do with each other (Wan et al. 2018). The tunica of Gnetales has only a single layer, not two or three as is common in angiosperms (e.g. Donoghue & Doyle 2000a; Doyle 2006). Similarly, what appears to be tension (reaction) wood in Gnetum, normally produced as the branches maintain their orientation against gravity, consists of gelatinous extra-xylary fibres adaxial on the branch; this makes it unique among seed plants and unlike the tension wood of angiosperms (Tomlinson 2001b, 2003; see also Höster & Liese 1966). Indeed, in Ephedra these fibres seem not to function as reaction wood (Montes et al. 2012). Ovule size in angiosperms does not increase between pollination and fertilization, while the ovule in Gnetum increases appreciably in size during this period as in a number of other gymnosperms, but also with some gametophyte development continuing after fertilization (Leslie & Boyce 2012), unlike other gymnosperms (Friedman & Carmichael 1996). Other characters in common between Gnetales and angiosperms such as fast pollen tube growth (Williams 2008) have been deconstructed in the same way. It is not known if details of the loss of sperm cilia and the associated development of a pollen tube growing towards the ovule and the increased venation density of the leaves of Gnetum (Boyce et al. 2009) are similar in the two groups. However, leaf development, particularly the expression of members of the WOX (Wuschel-related homeobox) gene family, does seem to be quite similar in Gnetum and angiosperms, with fewer similarities to other gymnosperms (Nardmann & Werr 2013). Similarly, xylan substitution patterns in Gnetales and angiosperms are quite similar, and there is less similarity with those of other gymnosperms (Busse-Wicher et al. 2016), and Busse-Wicher et al. suggest that the Gnetales pattern could be plesiomorphic for seed plants, or it might be somehow functionally associated with the vessels that are found in both groups. See Coiro et al. (2017/2018) for more details.

Morphological phylogenetic analyses have often suggested a connection between the "flowers" of Bennettitales and those of angiosperms (Rothwell et al. 2008a, 2009; Crepet & Stevenson 2009, esp. 2010: c.f. relationships among angiosperms). However, in Crepet and Stevenson (2009, 2010) the topology is sensitive to change of one character state in one taxon, in some morphological analyses Bennettitales do not group with anthophytes and are associated with cycadofilicalean plants, and extant gymnosperms are not monophyletic, Gnetales being sister to angiosperms. On the other hand, Rothwell et al. (2009) and Rothwell and Stockey (2013) strongly questioned the idea of a close relationship between Bennettitales and Gnetales, noting i.a. that the former had spiral, not decussate, insertion of parts, the nucellus formed a plug in the micropyle, and there was no pollen chamber. Little et al. (2014) suggest that Bennettitales lacked motile sperm, just like Pinales. Interestingly, the triterpenoid oleanane, found in angiosperms, is also found in Bennettitales, but it is also scattered elsewhere (Moldowan et al. 1994; E. L. Taylor et al. 2006; Feild & Arens 2007), the triterpenoid isoarborinol also being synthesized by a marine bacterium (Banta et al. 2017). But wherever Gnetales end up, Bennettitales et al. are not necessarily to be associated with them (Coiro et al. 2017/8).

Gnetales themselves are far from having a flower, but they do have strobili with both types of sporangia, even if only one type normally produces spores. However, many early morphological studies linked Gnetales and angiosperms, the anthophyte hypothesis (see elsewhere), although evidence currently suggests that the former may be best placed sister to Pinales (see below), but if inside Pinales, do these other fossils go with it? (probably not: Coiro et al. 2017/2018), and how do we understand evolution in Pinales? Puttick et al. (2017) reanalysed the data matrix of Hilton and Bateman (2006) using a variety of methods and additional data matrices, both real and simulated, and found that Caytonia consistently came out sister to angiosperms, but broader groupings depended on the methods of analysis used (see also O'Reilley 2016), and with Bayesian analyses, that was all you got. As Puttick et al. (2017: p. 6) noted of a Bayesian analysis they carried out on the Hilton and Bateman data set, "This [low resolution] suggests that the available data are insufficient to discriminate among the competing hypotheses, and this long-standing debate [about the relationships of angiosperms as evident in the fossil record] is largely an artefact of the false resolution of parsimony methods." Although Rudall and Bateman (2019b: p. 619) agreed: " the instability of taxonomically broad morphological phylogenies should not be underestimated", they clearly hoped that careful coding of mophological characters might solve the problem. Overall, the BEG clade would seem to have little immediately to do with angiosperm origins, and if only morphology allows us to infer the morphology of the immediate ancestors/relatives of crown-group angiosperms, and of the ancestors/relatives of those plants, and morphology is currently not helping much. Gernandt et al. (2016), Rothwell et al. (2018a) and Coiro et al. (2017/2018, esp. 2019) discuss fossils, morphology and phylogeny.

Rudall and Bateman (2010) reasonably thought that the morphology of crown-group conifers, being highly derived, might be of little help in thinking about that of the ancestors of angiosperms. This in part turns the problem over to the interpretation of fossil remains and association of particular fossils with the angiosperm stem, and here there has been little progress over the last fifty years or more. Similarities between the ovules of some Magnoliaceae and the cupules of Caytonia (e.g. Umeda et al. 1994) are probably superficial; features like the lobing of the integuments which induced this comparison seem to have little phylogenetic significance (e.g. Endress & Igersheim 2000; Endress 2005c). However, it has been suggested that the two integuments are of quite different origins and ages. The inner integument perhaps represents a modified dichotomising telomic system while the outer integument is leaf-like and derived from a cupule wall (see J. A. Doyle 2006; see also Skinner et al. 2004: expression of a YABBY gene in the o.i.); the inner integument of angiosperms develops first (e.g. Schneitz 1999). Although the ovule-bearing structures of Caytonia and Glossopteris can be linked with the ovules and carpels of extant angiosperms by invoking appropriate morphological gymnastics, it does not make for satisfactory reading.

Baum and Hileman (2006) proposed a developmental genetic model for the evolution of the angiosperm flower which may help in the interpretation of the significance of particular fossils (see also Ruelans et al. 2017: gene duplications involved). Frohlich and Parker (2000) had suggested that the heterosporangiate strobilus had evolved in pteridosperms like Corystospermales (the CUP group) from a male strobilus on which ectopic ovules developed - their "mostly male" theory of the origin of angiosperm flowers. LEAFY/FLORICAULA genes were likely to be associated with male reproductive structures, they suggested, and NEEDLY genes with female. However, work on the expression of LFY/FLO and NLY orthologs suggest that both genes are expressed in early-stage primordia, but the former are then expressed in ovules and microsporangia while the latter are expressed in the ovuliferous scale, aril, microsporophylls, etc.. The expression of both genes in both male and female cones is not consistent with the "mostly male" theory (Vásquez-Lobo et al. 2007 and references; Moyroud et al. 2010; Tavares et al. 2010; see also Bateman et al. 2011b). Mathews and Kramer (2012; c.f. Kelley & Gassner 2009 for a more conventional approach) analyse ovule development in seed plants and floral development in angiosperms to think how these structures might have evolved, and i.a. they suggest that evolution is less the change in form of pre-existing structures than the assemblage of new developmental modules in the context of homeosis, heterotopy, and heterochrony (see also Harrison et al. 2005b; Pires & Dolan 2012; Lovisetto et al. 2012; Almeida et al. 2014). As the Amborella Genome Project (2013) note, most of the genes involved in the development of the flowers of crown-group angiosperms are ancient, it is how they became assembled that made the flower what it is.

To summarise: Ideas of relationships between angiosperms and other seed plants remain in limbo (Scott et al. 1960; Feild & Arens 2005, 2007; Frohlich & Chase 2007; Herendeen et al. 2017; Sauquet & Magallón 2018). As Cronquist (1981: p. 2) noted, "no direct connection of the angiosperms to any other group is known from the fossil record", while a quarter of a century later Feild and Arens (2007: p. 292) observed, "It is not hyperbole to claim ‘all bets are off’ on the question of angiosperm sister groups", and things have not changed. In particular, it is unclear what gymnosperms/seed ferns are to be linked with stem-group angiosperms, regardless of whether or not extant gymnosperms are monophyletic or where Gnetales are to be placed on the tree (e.g. Rudall & Bateman 2010; see also above). In a comprehensive review on the bearing of fossil data on the origin of the flower, J. A. Doyle (2008b) concluded that our understanding of the fossil record was insufficient to help much in understanding angiosperm origins. Taylor and Taylor (2009) reached much the same conclusion, and they emphasized that timing was important. For instance, glossopterids are hardly known after the Permian-Triassic boundary, i.e. some 100Ma before the earliest angiosperms - at least, by some estimates. Cascales-Miñana et al. (2016b) are inclined to think that stem-group angiosperms can be dated to the Triassic, perhaps to the events immmediately following the Permo-Triassic extinction event (for which, see elsewhere, also Blomenkemper et al. 2018). Some of the recently-described Jurassic fossils that are thought to be angiosperms (but see below, also Meyer-Berthaud et al. (2018)) might then be better identified as stem-group angiosperms, the younger Cretaceous angiosperm record representing the diversification of crown-group angiosperms (Cascales-Miñana et al. 2016b: Fig. 1, c.f. topology, extant gymnosperms there polyphyletic). Indeed, as with the ages of crown-group angiosperms (see below), the ages of "phylogenetic fuses" or stem groups have to guide one's thinking. In this case, by just about any estimate, stem-group angiosperms must have been around for a very long time - 84-273 Ma is the spread of possibilities in Barba-Montoya et al. (2018) alone.

For further information on the major seed plant groups, see angiosperms, Cupressales, Cycadales, Ginkgoales, Gnetales and Pinales, and for discussion about their relationships, see also above, conifers in general, and extant seed plants in general.

4B. Pollination & Seed Dispersal. Early seed plants are likely to have been wind pollinated, and the ovules of seed plants in general, including pteridosperms, probably had pollination droplets (Nepi et al. 2017). A number of gymnosperms, both living and extinct, have saccate pollen, the sacci seeming to be wings that aid in the dispersal of the pollen, but these sacci are more like water wings, and help float the pollen, captured in the pollen droplet, on to the micropyle, although they may also reduce the settling velocity of the pollen grains, a low settling velocity being characteristic of wind pollination (Schwendemann et al. 2007; Bolinder et al. 2015; c.f. Hall & Walter 2011 in part). There is a general correlation between saccate pollen, erect cones, inverted or downwards-facing ovules, and the presence of a pollination droplet - although perhaps not in Cordaitales. Saccate pollen has evolved more than once, and it has also been lost - but apparently never regained (e.g. Stützel & Röwekamp 1999b; Leslie 2008, 2010b; especially Leslie et al. 2015a). Pollination droplets may be involved in pollen capture as long ago as in the Middle Pennsylvanian seed fern, Callospermarion pusillum, capturing pollen being dispersed by wind (Rothwell 1977; c.f. Labandeira et al. 2007). They may also become a part of an insect pollination mechanism, as in some extant Gnetales (Rydin & Bolinder 2015; see also Frohlich 2001), in the past Ginkgoales, and in Cycadales, albeit in the latter in a somewhat different way (Nepi et al. 2017). Little et al. (2014) and von Aderkas et al. (2014) discuss the composition of the nectar-like pollination droplet of gymnosperms - even in wind-pollinated taxa it is rich in sugars. Recent work suggests that pollination droplets of wind-pollinated taxa are lower in sugar but higher in total amino acids than those of ambophilous taxa, i.e. taxa in which there was pollination by both wind and insects (Nepi et al. 2017; 13 species examined); there is also a greater proportion of non-protein amino acids, amino acids that can influence pollinators, in the droplets of ambophilous taxa - all in all, the droplets were rather like angiosperm nectar, although the latter had higher sucrose concentrations.

Insects first appeared in the Middle Silurian (Misof et al. 2014: molecular dates) to Late Devonian (Garrouste et al. 2012: fossils). Early insects are implicated in the pollination of some Middle Jurassic to Early Cretaceous gymnosperms, and a diversity of insect groups - beetles, Neuroptera, mecopterids (scorpion flies, Mecoptera, perhaps), Kalligrammatidae (lacewings) and true flies (bee flies may be early Jurassic - see Wiegmann et al. 2011), thrips, etc., a number with quite long proboscides - are thought to have been involved (Labandeira 1998, 2006, 2010; Grimaldi 1999; Labandeira et al. 2007; Ren et al. 2009; Peñalver et al. 2012; Labandeira & Currano 2013; Peris et al. 2017). Crown-group ages of all these clades date to the Carboniferous (Misof et al. 2014), however Tong et al. (2015, see reply by Kjer et al. 2015) thought that a number of ages should be ca 100 Ma older. Crown-group lepidoptera may be around 260 Ma, later Permian (Tong et al. 2015), while Kawahara et al. (2019) suggest that they are Late Carboniferous, proposing a Middle Triassic age of (261.1-)241.4(-218.9) for the origin of nectarivory in the Glossata, the moth clade with proboscides. This makes any link between angiosperm and lepidopteran diversification more complicated, however, Kawahara et al. (2019: p. 22658) note of their ca 241 Ma age for the origin of nectarivory, it "overlapp[ed] with the estimated diversification periods of speciose flowering plant crown groups", but this does seem something of a stretch. Lepidopteran diversification may have begun on Jurassic or even Late Triassic gymnosperms (Labandeira et al. 1997; Wahlberg et al. 2013), although this has been questioned and is perhaps unlikely (Grimaldi 1999; Grimaldi & Engel 2005; Regier et al. 2015: esp. Fig. 10), and stem-group ages in Wiens et al. (2015) also cover quite a spread. Note that the first couple of lepidopteran clades are jawed moths, but most lepidoptera are Glossata. Indeed, some lepidopteran scales in deposits from the late Triassic 212 and 201 Ma that are hollow and may have a serrated apex have been linked to non-ditrysian (basal) Glossata (i.e. they have a single opening for mating and laying eggs), the adult moths perhaps visiting gnetalean-type plants for nectar and pollinating them (van Eldijk et al. 2018). Kawahara et al. (2019) suggest that the first Glossata may have used their proboscides to drink water or sap.

Bennettitales, flourishing from the Triassic to the Cretaceous, had large, rather flower-like reproductive structures, those of Cycadeoidaceae in particular producing both pollen and ovules, but little is known about how they were pollinated (Friis et al. 2011). However, Peñalver et al. (2015) found a ca 105 Ma brachyceran fly in amber, and it had a long proboscis with pollen probably from a bennettitalean plant on it; such flies were widely distributed and are known from deposits 125-100 Ma. Another brachyceran dipteran fly ca 160 Ma with a ca 12 mm long proboscis may have pollinated the bennettitalean Williamsoniella karataviensis (Khramov & Lucanshevich 2019). Scorpion flies are also quite large pollinators with a diversity of proboscis lengths (ca 0.75-14.5 mm long - Liu et al. 2018) and may also have fed on a variety of Mesozoic gymnosperms (Peñalver et al. 2015). Labandeira et al. (2016) suggest that kalligrammatid lacewings (Neuroptera) were also likely to have pollinated bennettitalean plants. These lacewings show notable parallelisms with Lepidoptera, including eyespots and scales on the wings, etc., and their probosces were ca 0.5-18 mm long; they flourished 165-99 Ma (Labandeira et al. 2016; Q. Liu et al. 2018). Diptera are suspected of pollinating Williamsoniaceae in the Late Albian ca 105 Ma (Peris et al. 2017). Features of fossil "ephedroid" pollen frequently fit the insect pollination syndrome, i.a. the pollen tends to clump (Bolinder et al. 2015). Certainly, angiosperms never had a monopoly on insect pollination (see also Erbar 2014).

In extant gymnosperms (but less so in Gnetales) unfertilised ovules are sometimes about as large as seeds, and they keep on growing until fertilization, which may be a long time after pollination; seed reserves are gametophytic tissue. In angiosperms, however, ovules are small, the time to pollination is short, and the seeds are nearly always relatively much larger than the ovules, reserves commonly coming from tissue (endosperm) that is produced after fertilization. Angiosperm ovules can be aborted with little loss to the plant if pollination does not occur, but in gymnosperms the loss is often much more substantial (Haig & Westoby 1989).

Sims (2012) suggested that during the middle Mississippian to Pennsylvanian average seed size increased to about 8 mm3, a value that held largely steady until the evolution of flowering plants, when it decreased; cycads (large) and many Pinales (small) are the two ends of the seed size spectrum in extant gymnosperms. Mesozoic seeds are diverse morphologically (e.g. Anderson & Anderson 2004), and animal dispersal is likely to have been quite common (Friis et al. 2011). The Cretaceous-Aptian (ca 120 My) toothed and long-tailed avialan Jeholornis primus had numerous spherical ovules/seeds/fruits (the form genus Carpolithes) ca 8-10 mm long in its gut (Z. Zhou & Zhang 2002), their size suggesting that they came from a gymnosperm. Lovisetto et al. (2011) discuss the evolution of fleshiness in disseminules of seed plants in general; in general, similar genes are involved, even if fleshiness may develop in very different places/from very different tissues. It has been suggested that the evolution of fleshiness in diaspores was initially as a defence against seed predators rather than an adaptation for seed dispersal - seed dispersers were not around when fleshiness evolved, and fleshy, edible fruits are then more of an exaption than anything else (Mack 2000; Moles 2017). However, Tiffney (2004) suggested that potential seed dispersers have been around for a long time, also cautioning about the perils of uniformitarianism, for instance, noting that ovules of early seed plants might have been very large before pollination (see above), complicating any understanding of what fleshiness was "for".

Plant-Animal Interactions. For plant-arthropod associations, see Labandeira (2006), who noted that medullosan pteridosperms showed a particularly high level of herbivory in the Carboniferous compared to that of other plants of that time. Galls, often thought to have diversified along with early angiosperms, are known on Glossopteridales and Peltaspermales in particular from the Lower Permian (Schachat & Labandeira 2015). Herbivory of angiosperms by llepidopteran larvae has been dated to (276.7-)257.7(-234.5) Ma, the age of the Angiospermivora, and they may initially have been internal plant feeder (Kawahara et al. 2019). See above for more on lepidopteran diversification.

5. Angiosperm History II: Cretaceous (or much earlier?) Origins, subsequent Cretaceous Diversification.

5A. Introduction. Ages of fossil plants that have been called with greater or less degrees of confidence crown-group angiosperms, or at least plants with the features of crown-group angiosperms, vary greatly, as will be discussed immediately below, and molecular estimates show an even wider spread; these latter are summarized above and range from 280-130 Ma or so.

There are number of reports of pre-Cretaceous angiosperm pollen. Some can be dismissed because of incorrect dating or contamination (Hochuli & Feist-Burkhardt 2013 for literature). However, there are several pollen types from the Middle Triassic (ca 243 Ma) of northern Switzerland that are similar to angiosperm pollen, being monosulcate, columellate, semitectate and reticulate, but with a very thin nexine (Hochuli & Feist-Burkhardt 2013). Since such pollen grains are found in a variety of habitats, it is a little difficult to explain the absence of angiosperm megafossils if the pollen really was from angiospermous plants. Hochuli and Feist-Burkhardt (2013) reasonably elect to consider these grains to be pollen of relatives of stem-group angiosperms, which could include a variety of plants. If columellate pollen is ancestral in angiosperms, there may be connections with the Triassic reticular-columellar Crinopolles pollen type (J. A. Doyle 2001; Zavada 2007). Other grains from the Triassic and Jurassic are remarkably like the very distinctive pollen of Acanthaceae-Tricantherinae (see also Tripp & McDade 2014b), but they are not associated with macrofossils. However, other angiosperm-like pollen types are associated with macrofossils, for instance, the late Triassic Sanmiguelia, although whether any are stem-group angiosperms is uncertain (Friis et al. 2011: pp. 158-162; Herendeen et al. 2017; Coiro et al. 2019; c.f. Cornet 1986); overall, such pre-Cretaceous pollen identified as being angiosperm is unlikely to belong there (Coiro et al. 2019).

Literature calling into question the generally accepted understanding of the morphology of angiosperm flowers, if confirmed, will change our search image of stem-group angiosperms. Thus when reporting on possible pre-Cretaceous angiosperms from China, X. Wang (2010a) thought that carpels consisted of an axis (cauline, = placentae) subtended by bracts (foliar, = carpel walls: see also Guo et al. 2103; W. Liu & Ni 2013; W.-Z. Liu et al. 2014; X. Zhang et al. 2017, 2019). The early to mid-Jurassic Schmeissneria, previously placed in Ginkgoales, was considered to be angiospermous, having closed carpels (X. Wang et al. 2007; X. Wang 2009, 2010b; Z.-J. Liu & Wang 2015 for discussion), while another mid-Jurassic fossil, Xingxueanthus, is reported to have had closed carpels as well as a style (Wang & Wang 2010); these were thought to be stem-group angiosperms, but they have angiospermous gynoecia. Euanthus panii, also from the same locality, was described as having 5 sepals and petals, probably 5 tetrasporangiate stamens, a semisuperior ovary with quite a long hairy style, and ovules, although details of ovules and anther contents in particular are hard to make out (Z.-J. Liu & Wang 2015) and the authors were not sure whether these were fossils of stem- or crown-group angiosperms. Han et al. (2016) described the very small - <4cm tall - Jurassic (>164 My) Juraherba which looks faintly like an onion but with long-pedicellate, axillary flowers that perhaps have parietal placentation. Herendeen et al. (2017) and Coiro et al. (2019) in particular have reviewed such fossils and the literature bearing on their identity thoroughly and thought that none was definitely an angiosperm, while some that could be identified seemed rather to be Gnetales or Ginkgoales (see also above), while Euanthus, for example, may well be part of a conifer cone. Similarly, the carpels of Solaranthus (= Euanthus) daohugouensis are probably resin bodies of a male cycad cone (Deng et al. 2014). However, findings of early angiosperm fossils continue to be reported, and include the Early Jurassic Nanjinganthus dendrostyla at least ca 174 Ma (Fu et al. 2017/2018); D. W. Taylor & Li (2018) are inclined to think that it indeed could be a flowering plant. Nanjinganthus has an inferior ovary with ovules on the outer wall, no obvious carpels, and a remarkable branched style; although there are large numbers of fossils, neither angiosperm pollen nor vegetative remains were recovered with the flowers (Fu et al. 2017/2018: are the "calyx" and "corolla" clearly separate?, they look like cone-scales...). As Fu et al. (2017/2018) noted when discussing the morphology of this curious flower, either present angiosperm morphology is no guide to working out what might be ancestral fossils, or angiospermy has evolved more than once. The ca 125 Ma monocot Sinoherba ningchengensis is younger, and it has small flowers that have a 2- or 3-seriate perianth, and the gynoecium, bilocular apically, has a single basal ovule; morphological phylogenetic analyses place it in a clade with Hydatellaceae, Cyclanthaceae and Najadaceae, and the larger clade to which it belongs is sister to all other angiosperms and also includes monocots and Nymphaeales (Z.-J. Liu et al. 2018a). However, this and some other phylogenetic analyses incorporating morphological data from fossils inspire little confidence (see above).

If crown-group angiosperms are 280 to 186 Ma (e.g. Zeng et al. 2014; Foster et al. 2016a), so certainly Jurassic if not mid-Permian (see above; c.f. Beaulieu et al. 2015) in origin, rather than Cretaceous in origin and only some 140-130 Ma, we have a series of problems. There will probably be different sets of gymnosperms/pteridosperms involved in the evolution of the angiosperm flower, and the ecological context for the evolution of angiosperms and of the insects associated with them will differ. Thus Salomo et al. (2017) suggested crown-group angiosperms appeared (342-)284(-226) Ma when conditions in general were rather dry, even if they were probably growing in locally more mesic areas, or, as Feild et al. (2009a: p. 258) put it, angiosperms may have persisted as "a rare, localized line that was highly dependent upon water for several million years". How angiosperms persisted as this presumably not very diverse clade for 50 Ma or much, much longer is yet another question. Furthermore, there are suggestions of an extinction at the Jurassic-Cretaceous boundary ca 145.5 Ma, and this is discussed in some detail by Tennant et al. (2016). There were extinctions (and radiations) over a rather protracted period of some 25 Ma, a period that saw eruptions that produced the huge Ontong Java plateau in the Pacific ca 125 Ma, the arrival of bolides between 145-142 Ma, and so on, although there seems to have been little effect on the land flora (Tennant et al. 2016). Agaricomycote fungi may have been affected, however, increased diversification had begun in the early Jurassic ca 180 Ma and lineage numbers and speciation rates seem not that much affected (Varga et al. 2019: Fig. 2).

In a recent review of the problem of when crown-group angiosperms first appeared, Coiro et al. (2019) examine the fossil data carefully and conclude that it does not support pre-Cretaceous origins, rather, they think the older dates in molecular studies are the result of some inherent biases in the latter. However, in addition to the wide spread of ages suggested for the first crown-group angiosperms, there are very differing narratives for later angiosperm evolution. As is mentioned below, some suggest that angiosperms achieved ecological dominance by the end of the Cretaceous, while others suggest that tropical rain forest as we know it had barely developed then, and that Cretaceous angiosperms were ecologically and physiologically rather unlike many of their Caenozoic successors. There are similar tensions in the literature on the evolution of animals associated with plants.

Angiosperm diversification is often discussed in terms of the consequence of the evolution of flowers (and fruits). To oversimplify: The diversity of flowers and fruits represent adaptations to pollination and fruit dispersal and help ensure reproductive isolation; flowers in particular allowed greater speciation rates (e.g. Hickey & Doyle 1977; Niklas et al. 1983). This may well be true - floral differences are often involved in species barriers - and there are other important changes in the life cycle like increased efficiency in producing seeds and shorter gametophyte phase (see below). However, there are a couple of things to remember here. First, although what pollinated the seed plants immediately ancestral to angiosperms is unclear, insect pollination in extant gymnosperms (e.g. Zamiaceae, Gnetales) is quite common, and the same was probably true of their fossil relatives (e.g. Peris et al. 2017 for a summary; see also above). Insect pollination is by no means unique to angiosperms. Second, herbivory was widespread throughout the Mesozoic. For example, it has been suggested that Jurassic mammals (162.4-)161, 158.5(156.9) Ma - gliding eleutherodonts in this case - were herbivores, perhaps also eating ovules (Meng et al. 2017). Third, there have been vegetative and associated physiological changes that have profoundly affected the ecology and evolution of angiosperms and indeed the climate of the whole earth (see e.g. below). It is not only species numbers and flowers and fruits that matter, but also what clades "do", i.e. their roles in the ecosystem and their effects on the biosphere as a whole, that help us understand angiosperm evolution (e.g. Bengtsson 1998; in particular Augusto et al. 2014: emphasis on the ecophysiological dimensions; Rothman 2001; Minelli 2016: emphasis on morphological disparity).

Our understanding of the eco-physiological dimension of the evolution of extant angiosperms is changing fast (e.g. Feild & Arens 2007; Internat. J. Plant Sci. 173(6). 2012; Shah et al. 2016; etc.). Some of the differences between the leaves of early angiosperms and those of other vascular plants have long been evident (e.g. Hickey and Doyle 1977), and both these and other vegetative changes can be linked with changes in the rate of photosynthesis, nutrient cycling and acquisition, silicate breakdown and rock weathering, and the like (e.g. Knoll & James 1987; Volk 1989: emphasis on deciduous ecosystems; Sack & Scoffoni 2013; Augusto et al. 2014). Angiosperms may have facilitated the spread of the l.t.r.f. habitat in which so much biotic diversity is now to be found, and they may also be implicated in the long-term decline in atmospheric CO2 concentration that characterises the Caenozoic (e.g. Selosse et al. 2015). Interactions of plants with their pollinators and seed dispersers have taken place in the context of this changing biosphere (see also Boyce et al. 2010; Marazzi & Sanderson 2010). Finally, the interactions of plants with their fungal associates, whether ecto- or endomycorrhizal or endophytic, the bacteria associated with them, and again, their effects on the weathering rocks, soil structure, carbon sequestration, and nutrient cycling are all part of the story (e.g. Rothman 2001; Herre et al 2005; Beerling 2005a; L. L. Taylor et al. 2009; Quirk et al. 2012; Bragina et al. 2014), and here the idea that an individual plant can usefully be thought of as a holobiont, a group of more or less closely integrated organisms (Bordenstein & Theis 2015; Vandenkoornhuyse et al. 2015; Gilbert & Tauber 2016) comes into play. The complexity of the interactions between plants, animals and microorganisms almost beggars description. For example, the effect of a caterpillar on the plant can be manipulated by parasitoids of caterpillars, and ultimately by polydnaviruses in the parasitoids - the saliva of the infected caterpillar is less likely to elicit plant defences that that of an uninfected caterpillar (Tan et al. 2018; see also Zhu et al. 2018).

In what follows, I initially focus on angiosperm evolution, and later I attempt to integrate angiosperm diversification with that of some of their more important plant and animal associates. Faute de mieux, I assume an early Lower Cretaceous age for the appearance of crown-group angiosperms, although this is becoming ever more debatable. As already mentioned, estimated ages of crown-group angiosperms are all over the shop, and ages within angiosperms are not much more satisfactory. Thus Fleming and Kress (2013) suggested that crown-group Zingiberales were late Jurassic or early Cretaceous, and so some 150 My; what vertebrates pollinated them (they think vertebrate pollination is the plesiomorphic state for the order) is unclear under this scenario, and as to what might pollinate ancestral angiosperms, which must be substantially older, would be anyone's guess.

5B. Early Cretaceous Evolution - to the end of the Aptian, ca 113 Ma. Background. The climate in the late Upper Jurassic-early Lower Cretaceous was dry - certainly Pangea had a notably dry interior - but continents were drifting apart, and sea levels were rising (e.g. Tennant et al. 2016). Atmospheric CO2 concentrations were about 1,400 p.p.m. around the mid Cretaceous, possibly the highest concentrations since the late Devonian ca 360 Ma, but they declined subsequently, if with the odd hiccup; there may have been a particularly abrupt decrease in the middle of the Cretaceous (Feild et al. 2011b; Barclay et al. 2010; McKenzie et al. 2016). Oxygen concentrations were also high (Shi & Waterhouse 2010; Beerling & Franks 2010; He et al. 2012; Franks et al. 2013).

The most comprehensive reviews of Cretaceous angiosperm history are those of E. M. Friis and collaborators (e.g. Friis et al. 2011, 2019a and references), on which this whole section draws heavily; see also Krassilov (1997), Dilcher (2010), Taylor (2010: genes possibly involved), Cantrill and Poole (2012: Antarctica), and J. A. Doyle and Upchurch (2014). Doyle (2008b), Specht and Bartlett (2009), Endress (2010a), Doyle and Endress (2010), Sauquet et al. (2017: see esp. Supplementary Data 13) and others provide surveys of the floral morphology and biology of extant basal angiosperms. For the floral morphology of the immediate common ancestor of angiosperms, see e.g. Doyle and Endress (2000), Endress (2001a, 2010a), etc., and in particular Sauquet et al. (2017, 2018). This section takes to the story to the end of the Aptian, ca 113 Ma, i.e. a little before the end of the Lower Cretaceous and just after the first appearance of tricolpate pollen, the signature of eudicots, in the fossil record, although as we will see this appeared a little later towards the poles.

The earliest angiosperms are particularly well represented as pollen remains, much less so as flowers, etc. (Coiro et al. 2019). The pollen of early angiosperms is more or less globose, monosulcate, with a continuous tectum, columellate infratectum, and thin endexine (e.g. Doyle 2005; Friis et al. 2011; esp. Coiro et al. 2019; see also Hughes 1994 and references). Doyle (2001) noted the abundance of fossils with ascidiate carpels and exotestal seeds in these floras - and in extant members of the ANA grade and Chloranthaceae. Friis et al. (2019b) note the presence of an endothelium in Early Cretaceous seeds from Portugal associated with Chloranthaceae and think that an endothelium may have been common in early angiosperms in general (an endothelium is also found in some extant Piperales); they summarize the possible functions of such an endothelium. Fossil pollen from the Cretaceous Valanginian-Hauterivian 141-132 Ma has been attributed to angiosperms, and their diversification was well under way by 137 Ma as judged by these pollen remains, but it was 10-30 Ma or more before crown-group diversification really got going (e.g. Feild & Arens 2005). Thus in the Barremian-Aptian ca 125 Ma there are some 140-150 taxa recorded from Portugal alone (e.g. Friis et al. 1999, 2000a, 2010b, 2017c, 2019a). This is a remarkably diverse flora, even if some of the material may be somewhat younger, perhaps Albian and ca 112 Ma (Heimhofer et al. 2005, 2007).

Very few of these fossils can be assigned to extant families, but many of them (85% in some studies) are from magnoliid-type or somewhat monocot-like plants (e.g. Friis et al. 1997a, 1999, 2001, 2010a, 2011; Heimhofer et al. 2007; J. A. Doyle et al. 2008: early putatively monocot fossils; Doyle 2014b: identifications of Early Cretaceous fossils; Friis et al. 2017a). Doyle and Endress (2010: relationships [Chloranthaceae [[magnoliids + monocots] eudicots]], 2014) and Friis et al. (2011) should be consulted for possible phylogenetic placements of a number of mostly magnoliid and ANA-grade Cretaceous fossils. Indeed, many older plant fossils have very odd character combinations (e.g. Feild & Arens 2005; Fries et al. 2011 and references). For example, Archaefructus, probably an aquatic herb from the Barremian-Aptian at least 124 Ma (Sun et al. 2002), has been interpreted as having flowers unlike those of any extant angiosperm - they are perfect, there is no perianth, the receptacle is very elongated, the stamens are paired, and the carpels are conduplicate - or perhaps these "flowers" are inflorescences, the paired stamens representing staminate flowers and the carpels carpelate flowers. Hyrcantha, also more or less aquatic, is from the same Barremian-Aptian deposits in China (Dilcher et al. 2007); it has leaves with sheathing stipules and partly connate carpels with apparent resin bodies at their apices. Z. Zhou et al. (2003), Friis et al. (2003b, 2011), Crepet et al. (2004), Ji et al. (2004), Doyle and Endress (2007) and others offer interpretations of the flowers of this and other early fossil angiosperms.

Such fossils may represent quite distinct but now extinct clades (von Balthazar et al. 2008). Indeed, a variety of strange-looking putative angiosperms have been discovered in deposits from northeastern China, and although the identities of a number are disputed (Sun et al. 2006; Y. Yang & Ferguson 2015: some "angiosperm" fossils are near Ephedra), fossil finds from this area continue to challenge our understanding of angiosperm evolution (e.g. Sun et al. 2011, but see Z. Zhou 2014: eudicots). However, fossils ascribed to Sarraceniaceae (asterids-Ericales) from deposits of about the same age as those in which Archaefructus was found (H. Li 2005a, b) have turned out to be galls of the conifer Liaoningocladus boii (W. Wong et al. 2015), which is something of a relief (see also Friis et al. 2011: pp. 158-162; Herendeen et al. 2017; and discussion above).

It is interesting that a number of early fossils are of aquatic angiosperms. However, even if Archaefructus, ca 124 Ma, is an aquatic and sister to all extant angiosperms (Sun et al. 2001; Crepet et al. 2004), given that we do not know of other plants that can be placed in that part of the tree, it does not necessarily mean that angiosperms had an aquatic ancestry. Although Du et al. (2016) do seem to incline in that direction, they prefer to think that Nymphaeales, Acorales, Alismatales, and Ceratophyllales are extant members of an early radiation of aquatic angiosperms that includes Archaefructus. Recent work suggests that Archaefructus could belong to Nymphaeales (Doyle & Endress 2007, 2010a; Doyle 2008b), and Hyrcantha, also more or less aquatic, is from the same Barremian-Aptian deposits in China (Dilcher et al. 2007). There are fossils of stem-group Nymphaeaceae, Cabombaceae and/or [Nymphaeaceae + Cabombaceae] from several parts of the world in the Lower Cretaceous (e.g. D. W. Taylor et al. 2001, 2008; Friis et al. 2011). Pluricarpellatia, probably to be placed in or near Cabombaceae (Doyle & Endress 2014), is known from the Early Cretaceous (Mohr et al. 2008), while Monetianthus, also Early Cretaceous, was embedded within crown Nymphaeaceae in morphological analyses (Friis et al. 2009b, but c.f. 2011; see also Doyle & Upchurch 2014; Doyle 2016; see also Carpestella, from Virginia - Doyle & Endress 2014), although other analyses placed Monetianthus at the node above Nymphaeales along the spine of the angiosperm tree (Friis et al. 2009b). Ceratophyllales are another ancient clade of aquatics which once may have been quite diverse. The Portugese fossil Montsechia vidalii, 125 Ma or more old, is thought to be close to Ceratophyllum (Gomez et al. 2015), while the distinctive fruits of Ceratophyllum itself are known from the Aptian and Albian onwards (see Dilcher & Wang 2009; Friis et al. 2011 for references), more fossils of both these orders appearing by ca 100 Ma (e.g. Friis et al. 2017b and references; Wang & Dilcher 2018), by which time Nelumbonaceae (Proteales) are also part of the mix. In any event, the development of an angiosperm-wide morphological database to be used in the context of molecular backbones (see Schönenberger et al. 2020) should certainly help in the placement of these early fossils.

Coiffard et al. (2012) give the impression that around 130-125 Ma angiosperms were largely a clade of aquatics, noting that 5/11 genera that grew in the first stage of the rise to dominance of angiosperms "competed with charophytes" (p. 20955). They did not say that the ancestral angiosperm was aquatic, but Goremykin et al. (2012) do think that this was likely and Gomez et al. (2015, but c.f. Herendeen et al. 2017 for Montsechia) also entertained this possibility. Indeed, early herbaceous nymphaealean-type plants are likely to have grown in aquatic or marshy habitats (G. Sun et al. 2008, see also Friis et al. 1999, 2011), and plants in such habitats may well become fossilized easily and so are over-represented in the fossil record (Coiro et al. 2019). However, both very aquatic (and xeromorphic) plants are likely to be derived (Coiro et al. 2019); indeed if the loss of cambial activity in aquatics may be difficult to reverse (Groover 2005; Feild & Arens 2007), then the aquatic habitat may be an evolutionary dead end. However, whether Nymphaeales are sister to Amborellaceae or not may have little effect on ancestral reconstructions of habitat preferences, but, as already mentioned, an important issue is where very old and long-aquatic clades like Ceratophyllales go on the tree, and how they relate to Chloranthales. Are they related, or are Nymphaeales and Ceratophyllum members of old clades of freshwater angiosperms (Du et al. 2016)? Others have suggested that the angiosperm progenitor was a "diminutive, rhizomatous to scrambling herb" (Taylor & Hickey 1992: p. 137, c.f. the palaeoherb hypothesis). In any event, early climatic niche (habitat) evolution is likely to have been slow (S. A. Smith & Beaulieu 2009).

Moore et al. (2007) suggested some time between 148.6-135.5 Ma for a rapid separation of the Chloranthaceae, monocot, magnoliid, eudicot and Ceratophyllum clades (see also Sun et al. 2011). Plants ca 130-120 Ma belong to the ANA grade, Chloranthaceae, and Ceratophyllum areas, while magnoliids, Platanales, and Buxales predominate later. Fossils assignable to Chloranthaceae date from the late Barremian ca 130 Ma onwards, some being very like extant Hedyosmum (e.g. Crepet & Nixon 1996; Friis et al. 2006b, 2011; Doyle & Endress 2014, 2018). Pseudoasterophyllites, vegetatively like Ceratophyllum, has been linked with Tucanopollis, an abundant palynomorph from Africa-South America over 125 Ma, and also with Chloranthaceae (Kvacek et al. 2012). A [Chloranthales + Ceratophyllum] clade would be of considerable interest and importance (see also Doyle et al. 2015; Gomez et al. 2015, c.f. Herendeen et al. 2017), but molecular evidence for it is as yet not compelling (see elsewhere for more discussion). Monocot pollen 120-110 Ma has been identified as Araceae-Pothoideae (Friis et al. 2004; see also J. A. Doyle et al. 2008: Friis et al. 2010; c.f. Hoffmann & Zetter 2010), but overall monocot fossils are not very common, perhaps because monocots are largely herbaceous and may have fossilized less well. Pollen data suggest a monocot/magnoliid split in the early Aptian-mid Albian 125-105 Ma (Heimhofer et al. 2005; Hochuli et al. 2006), while Jud and Wing (2012) thought that monocots and eudicots might have diverged 125-119 Ma (see also Jud & Hickey 2013; W. Wang et al. 2014), initial angiosperm diversification having occurred a mere 5-10Ma before that (see also Wang et al. 2016a, b). Magnoliids diversified somewhat later for the most part (Friis et al. 1997a, 2006b for reviews).

The flowers of early angiosperms appear to be rather generalized and are small to very small, quite often less than 1 mm long and/or across - there are very small fossil waterlilies, very small Hedyosmum-like flowers, etc. (see e.g. Crane et al. 1995; Doyle & Donoghue 1986a; Friis & Crepet 1987; Friis & Endress 1990; Friis et al. 2000, 2006b, 2010b, 2011; Endress 2001a; Weberling 2007; Doyle 2008b; Doyle & Endress 2000, 2010, 2011), although Sauquet et al. (2017) do not mention a size for the ancestral angiosperm flower that they reconstructed. The stamens are often wedge-shaped, with a relatively massive apex, stout filaments and connectives, and anthers opening by laterally-hinged valves (e.g. Crepet & Nixon 1996; Endress 2008c and references; Endress 2011a) - although perhaps not in the earliest angiosperms. Styles were at most short, and dry stigmas and protogyny were probably the common conditions (e.g. Sage et al. 2009; Endress 2010a). There was often only a single ovule per carpel, perhaps the ancestral state for angiosperms (e.g. Doyle 2012), the stamens produced only a few pollen grains, and there is initially no evidence of nectaries (e.g. Crepet et al. 1991; Dilcher 2000; Friis et al. 2006b, 2011). However, quite "derived" features are early apparent. Thus Sinocarpus, from the Barremian-Aptian 139-122 Ma, had carpels that were apparently connate at the base (Leng & Friis 2003), and flowers with inferior ovaries were surprisingly common early on (e.g. Crane et al. 1995; Friis et al. 1999, 2011). Interestingly, Oyston et al. (2016) suggested that angiosperms as a whole have shown remarkably little variation in disparity over time, that is, the amount/extent of morphological variation in a sample of taxa (c.f. gymnosperms), and this is true of individual angiosperm groups like palms; Lupia (1999) had noted that pollen disparity increased early in angiosperm evolution but was flat in the late Cretaceous-Palaeocene.

Little is known about the evolutionary physiology of flowers. However, Roddy et al. (2016) suggest that the flowers of magnoliids and ANA-grade angiosperms have relatively numerous stomata, high transpiration rates, and much-branched vascular bundles connected to stem xylem, overall, carbon investment was high. However, in monocot and eudicot flowers much less water was lost, stomata are largely absent (see also Lipayeva 1989) and vascular connections to the stem less evident, perhaps occurring more via the phloem; water loss and overall carbon costs were lower. Galen et al. (1993) looked at the carbon cost of various parts of the flower of Ranunculus adoneus from opening to fruit ripening and found that the petals were responsible for most of this cost, and although there were very few (two!) stomata per carpel, photosynthesis there appreciably defrayed the carbon cost of fruit and seed production. Of course, sepals in taxa in which they persist may also affect the carbon balance of the fruit, but the costs of producing fruits of extant taxa like Bertholletia must be very appreciable. Sampling is an issue here, and how such findings might relate to the very small flowers and fruits of early angiosperms is unclear.

In some extant magnoliids and ANA grade angiosperms in particular distinguishing between perianth and prophylls and bracts can be difficult, there can be intermediates between perianth and stamens, the numbers of parts and their arrangement vary, pollen morphology and embryo sac development vary, etc. (e.g. Buzgo et al. 2004; M. L. Taylor et al. 2008, 2015; Endress 2008a; J. A. Doyle & Endress 2011; Abercrombie et al. 2011; Wortley et al. 2015; L. Lu et al. 2015). Taxa in which DEF-like and GLO-like proteins can form homodimers predominate in the ANA grade and magnoliids (DEF-like proteins cannot form homodimers above the latter node, and GLO-like proteins cannot form them in eudicots); heterodimers are formed throughout flowering plants, although the situation in gymosperms is less clear (Melzer et al. 2014). Similarly, Hernández-Hernández et al. (2006) and Gloppato and Dornelas (2018) note a duplication in a B-class MADS-box gene in the ancestor of angiosperms producing APETALA3-like (DEF) and PISILLATA-like (GLO) gene lineages. This ability to form both homo- and heterodimers may contribute to the diversity of floral morphologies in these basal clades (Melzer et al. 2014), and perhaps in their extinct relatives. In the fading borders/sliding boundaries model of floral evolution (see also Thiessen & Melzer 2007), genes whose expressions are quite tightly linked to particular floral whorls in eudicots show much less specificity in more basal angiosperms (Chanderbali et al. 2009, esp. 2010 and references: Lauraceae and Nymphaeaceae emphasized, 2016). Along the same lines, Warner et al. (2009) noted that "petalness" and "sepalness" in flowers of taxa from Nymphaeales and Austrobaileyales was environmentally controlled, the exposed areas in buds of Nymphaea, for example, being sepal-like on the outside. Understanding floral development in these clades will clarify floral evolution in angiosperms as a whole (Chanderbali et al. 2016a), for instance, the development of fully whorled phyllotaxis in the perianth, i.e., not the monocot condition where the tepal whorls do not fully encircle the floral apex, and change in the control of "sepalness" would be a route to developing eudicot flowers. A recent reconstruction of the possible ancestral angiosperm flower (Sauquet et al. 2017) which, although apparently very Magnolia-like, differs in important details, may also help us to understand these early fossils since it suggests a different route to the evolution of typical monocot and pentapetalan flowers (see above for further discussion).

As to what pollinated early angiosperms, little directly is known. It has often been suggested that early angiosperms and their insect pollinators diversified together, even if the orgin of angiosperms is as early as the Triassic (e.g. Zeng et al. 2014). Some sort of insect pollination is likely (Hu et al. 2008), although Hu et al. (2012 and references) suggest that pollination may also have been by both insects and wind, ambophily (see also Friis et al. 2011). Labandeira (2000) included Coleoptera, Diptera, Hymenoptera, Lepidoptera and Thysanoptera in his "big five" angiosperm pollinators. How pollinators handled the flowers of early angiosperms, mostly very small, is unknown, although extant members of early-branching clades of Lepidoptera (monotrysian "microlepidoptera"), at least, are also usually very small (Regier et al. 2015 for a phylogeny; Mitter et al. 2016). B. Wang et al. (2013) reviewed the possible role of beetles in the pollination of early angiosperms, noting that potential pollinators such as Scarabaeoidea and Chrysomeloidea had evolved by or at the early Cretaceous. Labandeira (2010 and references) suggested that many pollinating clades of Hymenoptera, Diptera, and Lepidoptera originated around the late Barremian/end Albian some 125-100 Ma, and although Misof et al. (2014: p. 767) note, "we dated the spectacular diversifications within Hymenoptera, Diptera and Lepidoptera to the early Cretaceous, contemporary with the radiation of flowering plants" (see also Grimaldi 1999; Lunau 2004), any connection between insect and plant diversification then is unclear, and anyhow dates in Tong et al. (2015; c.f. Kjer et al. 2015) are sometimes substantially earlier (see also Regier et al. 2015 for caveats over dating). Butterflies, which are moths that became diurnal perhaps because of the existence of an appropriate food source (= angiosperm flowers), so escaping nocturnal predators (bats),and a major group of pollinators, may have diversified perhaps (110.3-)98.3(-86.9) Ma (Kawahara et al. 2019), (129.5-)107.6(-89.5) Ma (Chazot et al. 2019) or (143-119(-91) Ma (Espeland et al. 2020: other estimates there). Bees are perhaps unlikely to have pollinated the very earliest flowers. Thus Cardinal and Danforth (2013; see also Cardinal et al. 2018) suggest that crown-group Anthophila (bees) are some (132-)123(-113) Ma, all families having diverged by the K/P boundary, while Sann et al. (2018) date stem-group Anthophila at (148-)128(-108) Ma. Certainly, their diversity in the earlier part of the Cretaceous was low (e.g. Grimaldi & Engel 2005). Interestingly, Ammoplanidae, the crabronid wasps perhaps sister to bees, catch flower-visiting thrips as food for their larvae, and both these wasps and the earliest fossil bees are tiny, less than 4 mm long (Danforth & Poinar 2011; Sann et al. 2018).

Brachyceran diptera, which can have probosces over 5 mm long "especially suitable for visiting long tubular flowers" (Ren 1998: p. 86) for nectar, might have pollinated early angiosperms. These flies are known from the late Jurassic and early Cretaceous (Ren 1998; c.f. Z. Zhou et al. 2003 for dates), although given both their age and their relatively large size, their role in angiosperm evolution is unclear; there is no evidence that they pollinated early angiosperms (Peñalver et al. 2015), "long tubular flowers" apparently evolving later. Indeed, Labandeira and Currano (2013) and Labandeira et al. (2016 and references) suggest that the mid-Cretaceous decrease in prominence of gymnosperms was associated with the demise of many of their insect associates, including putative pollinators with elongated mouthparts. It has been suggested that an apparent plateau in diversity in the mid-Cretaceous 125-100 Ma, the Aptian-Albian gap, in the middle of the angiosperm radiation, might be due to a lag period as insects became habituated to angiosperms (Labandeira & Sepkowski 1993; Labandeira 2014: complex analysis, families the units, no tree). Peñalver et al. (2015) wonder why, if Bennettitales were pollinated by insects (as they suggest), an apparently adaptively superior mode of pollination, they should have gone extinct. Pieces of the puzzle are missing.

Looking at pollination in extant basal angiosperms, Gottsberger (1970, esp. 2016a; see also Thien et al. 2000) suggested that the basic condition for angiosperms is to be protogynous and self-compatible, and with generalized pollinators, mostly flies and beetles. The flowers are also quite often often unisexual. Thermogenic pollination - beetles are often pollinators in such circumstances, but diptera may be involved - occurs in some extant members of many "basal" lineages, including the ANA grade, some magnoliids (Annonaceae, Aristolochiaceae, Magnoliaceae, Winteraceae), monocots (Araceae), etc., although it is unknown in Amborella, Acorus and Laurales (e.g. Thien et al. 2000; Seymour et al. 2003; Seymour 2010; Gottsberger 2016a). Beetles are attracted to haplomorphic flowers lacking definite symmetry signals (Leppik 1957), and other factors such as scent are also involved. (Note that some beetle-flower interactions may be quite recent, such as pollination by scarabeid beetles, a group that evolved only in the Palaeogene - Gottsberger 2016a.) Interestingly, pollinators have been observed visiting droplets on the stigma, particularly in Winteraceae, apparently as part of the pollination process (e.g. Lloyd & Wells 1992 and references), and these droplets are perhaps analogs of the pollination droplets of conifers (c.f. Frohlich 2001). An extragynoecial compitum might also function in the same way. Of course, unlike early fossils, many extant basal angiosperms have quite large flowers (Gottsberger 2016a), and they are likely to be derived, large size, along with thermal regulation, carcass mimicry, etc., being adaptations to pollination by small flies and beetles (Davis et al. 2008). S. Hu et al. (2008) list early records of pollen, tabulate pollen morphology and suggest possible pollinators of ANA-grade angiosperms, magnoliids, basal eudicots and monocots, and finally optimise pollination mode on the tree.

The small flowers of early angiosperms were probably aggregated into inflorescences to attract pollinators (Friis et al. 2006b, 2011), and pollen was perhaps a likely reward for these early pollinators (Erbar 2014). Although pollen was initially produced in rather low quantities, it is found in insect coprolites and may have been eaten by the pollinator (Friis et al. 1999). The paucity of dispersed pollen morphs and the diversity of pollen morphs associated with plant remains also suggest some kind of insect pollination (Friis et al. 1999). Pollenkitt, produced by tapetal degeneration and rich in plastid-derived lipids that are used by bees, helps pollen grains stick together and to the pollinator, although the pollen of cycads, for example, may clump in the absence of pollenkitt (e.g. Hall & Walter 2011; Bolinder et al. 2015); pollen of early flowers is not often clumped (J. A. Doyle et al. 1975; Hu et al. 2008: examples given). However, there is no signal from fossil pollen that unambiguously suggests particular pollinators, and although pollen size and style length are correlated in a general way, there are no plots of pollen size over time (Roulston et al. 2000).

Despite the estimated age of nectarivory in moths, some 241 Ma (see above), there is no evidence that nectar, whether in the form of stigmatic secretions or from a more organized kind of nectary, was a common reward in early angiosperm flowers (see Erbar 2014: Table 1 for floral nectaries of basal angiosperms; Gottsberger 2016a). However, septal nectaries may be an apomorphy for monocots, being found in some Alismatales although probably not in Acorales; they may date to 120 Ma or earlier. Many Laurales have paired glandular bodies at the bases of the stamens that may provide some reward to the pollinator, and the clade with these nectaries is dated to around 127-89 Ma (see Laurales).

Florivory by insects - extant angiosperms being for the most part notably palatable to insects, although there is little obvious difference in protection of flowers versus their leaves - may be a route by which angiosperm-pollinator relationships became established (Frame 2003), although florivory may well cost the plant (McCall & Fordyce 2010). Indeed, Luo et al. (2018, esp. 2017b) drew attention to the mode of pollination in Kadsura (Austrobaileyales-Schisandraceae). Here the pollinators, nocturnal resin-feeding gall midges, also lay eggs on the flowers, the larvae eating resinous substances produced by the plant in response to midge oviposition, however, since neither pollen nor ovules are eaten, the cost to the plant is low. Thermogenesis occurs in Schisandraceae, but beetle pollination is derived (Luo et al. 2018).

Angiosperm seeds range in size from 10-7 to 104 grams, i.e. from about the size of an embryo sac to massively larger. Seeds of extant gymnosperms, at 10-3 to 103 grams, are rather larger than those of the basal clades of extant angiosperms (Haig & Westoby 1991; Moles et al. 2005a), which are also rather small compared with those of other angiosperms (Tiffney 1986a; Eriksson et al. 2000a, esp. b, also Haig & Westoby 1991). Eriksson et al. (2000b) estimated an average volume of less than 1 mm3 for seeds in the ca 124 Ma Famalicão flora, while Friis et al. (2015b) emphasized the minute size of the fruits, the seeds, and in particular that of the embryos (length relative to that of the seed) in 125-110 Ma floras from east North America and Portugal. Dispersal of disseminules of many early plants is likely to have been by wind (Wing & Boucher 1998). However, in a sample of some 100 taxa from the Barremian-Aptian (132-112 Ma), it was estimated that ca 25% of the disseminules were probably fleshy and animal dispersed (Eriksson et al. 2000b; c.f. Tiffney 1984). The fruits were small and mostly single seeded, although some may have been aggregated (see Eriksson et al. 2000b; Eriksson 2008; Dilcher 2010, in part; Sussman et al. 2013). It is unclear what ate these fruits; Eriksson et al. (2000b) suggest reptiles and multituberculate mammals, and perhaps other mammals and birds. However, the dentition of multituberculates from that period does not suggest herbivory (G. Wilson et al. 2012), while many early birds are likely to have been carnivorous and/or aquatic (Jarvis et al. 2014). However, gut contents with seeds have on occasion been recovered, perhaps most notably in the ca 120 Ma long-tailed bird Jeholornis primus, although the seeds/ovules, at 8-10 mm long and across (Z. Zhou & Zhang 2002), may have been those of gymnosperms.

Vegetative changes were very important in early angiosperm evolution. Angiosperm leaves are very different from those of most other vascular plants (Hickey & Doyle 1977; Sack et al. 2012). Fossils show the development of more regular and hierarchical venation, and teeth and compound leaves were early evident (see Doyle & Hickey 1976 for foliar evolution). How fast nutrient turnover in the litter of these angiosperms might be is debatable (Berendse & Scheffer 2009; Royer et al. 2010; c.f. G. Liu et al. 2014). The ecological consequences of such changes are discussed elsewhere.

Growth rates of early angiosperms may have been high and reproduction rapid compared with gymnosperms (J. A. Doyle & Hickey 1976; Bond 1989; Wing & Boucher 1998; Verdú 2002), although high growth rates in extant angiosperms are most evident in more productive habitats (Lusk et al 2003; Augusto et al. 2014). The reproductive cycle was relatively short, with a short time between pollination and fertilization (see below), seeds were small, the plants were small. However, since the embryos were so tiny, the seeds would probably have needed a period of dormancy before germination, and this would increase generation time (Friis et al. 2015b). A small plant size is perhaps to be expected since plant and seed size are quite strongly linked (e.g. Harper et al. 1980; Eriksson et al. 2000b; Moles et al. 2005b; Kavanaugh & Burns 2014: increase in size in island taxa) and early angiosperms were not large plants. In another study, seed size was estimated to be 2-3 mm3 (but c.f. Eriksson 2016, barely 1/10th this volume ca 80 Ma); seed contents were probably mostly endosperm, and again, given the small to minute size of the embryo with respect to the endosperm, a period of morphological dormancy was likely before a plantlet developed and germination could occur (Forbis et al. 2002; Linkies et al. 2010;; Kadereit et al. 2017). In summary, there was a short pre-reproductive period and short overall generation time (e.g. J. H. Williams 2008, 2009; Crepet & Niklas 2009; Bond & Scott 2010; Abercrombie et al. 2011: pollen tube growth). In extant angiosperms, overall diversification rates are higher in smaller- than larger-seeded groups, perhaps linked to improved colonization potential (Igea et al. 2017), and the rate of seed mass change is also higher when rates of diversification are higher. Overall, early angiosperms may have been rather small plants of open habitats, perhaps on flood plains, rather like some of the gymnosperms with chlamydospermous seeds that were contemporaneous with them (Friis et al. 2019c).

How woody the early angiosperms were is unclear. Seed size in the 132-112 Ma Famalicão flora is similar to that modern, tree-dominated floras, but Eriksson et al. (2000b) thought the plants that produced these seeds were herbaceous or shrubby (see also Hickey & Doyle 1977). Similarly, Jud and Hickey (2013) thought that angiosperms around 130-120 Ma were often herbaceous, but they would have maintained cambial activity; Jud (2015) described Fairlingtonia, an herbaceous angiosperm from 120-112 Ma which was apparently quite common in riparian habitats in the Virginia-Maryland area of the U.S.A.. Unfortunately, no flowers of the plant are known; the "adventitious" roots have a distinctive branching pattern, the one root apparently being made up of successive arch-like units (Jud 2015: Fig. 4a). Philippe et al. (2008) thought that earliest angiosperms might have had cambium, but noted that their wood lacked thick-walled fibres and in general cell walls were thin (see also Amborella). Fossil angiosperm wood is known from deposits of up to about 120 Ma (Aptian), although its assignment to extant clades is not easy (Oakley et al. 2009).

There are other suggestions. Arguing by analogy with extant angiosperms, early angiosperms may have been fairly small (Friis et al. 2010b) tropical trees with decumbent lignotubers and sympodial growth that tolerated shady, humid and disturbed ("dark and disturbed") conditions, perhaps like extant members of the ANA grade and Chloranthaceae, although not the aquatic Nymphaeales (e.g. Feild et al. 2003a, b, 2004, 2009a; Feild 2005; Feild & Arens 2005, 2007; Coiffard et al. 2006; Berendse & Scheffer 2009 for a summary). (Most eudicots have seedlings/young plants that are at least initially erect.) Conditions were certainly not dry ("ancestral xerophobia"), and were probably rather like today's lower montane subtropical forest with prolonged immersion in cloud and minimal drought stress (Feild et al. 2009a). Leaves of plants growing in such conditions are likely to have had relatively low venation density (e.g. Feild et al. 2009a: ca 3.8 mm mm2; de Boer et al. 2012), a vascular system with vessels, etc., is unlikely to have been at a premium, and productivity may have been low (Feild et al. 2011a). The age of crown-group angiosperms is important here. If ca 285 Ma (Salomo et al. 2017), not much in the way of evolution seems to have happened for the next 80 Ma or so, then this drought intolerance and low productivity might suggest that early angiosperms would have limited opportunities at a time when equatorial to mid-latitude areas were predominantly dry (Salomo et al. 2017). There are also suggestions that angiosperms initially grew in semi-arid (e.g. Stebbins 1965, 1974; Hickey & Doyle 1977) or at least seasonally arid (Bond & Scott 2010) conditions, and some early leaf floras from Portugal do have leaves that are small in size and xeromorphic in appearance (Friis et al. 2010b, see also 2011: pp. 46-47). Barba-Montoya et al. (2018) estimated a lengthy cryptic early history of angiosperms, which were Jurassic or much earlier, of some 23-121 Ma, which gives plenty of options for the ecology of these plants.

Move: Our knowledge of the dark reversal of P[phytochrome]fr → Pr may be a bit skimpy, but the general pattern of the ability of the plant to do this seems interesting: monocots are similar to centrosperm eudicots are rather different, Brassicaceae are somewhat intermediate (Kendrick & Hillman 1971; Hennig et al. 1999).

Whatever habitats the first angiosperms prefered, there is a consensus that they were soon components of disturbed, well lit or open and mesic communities (e.g. J. A. Doyle & Hickey 1976; Hickey & Doyle 1977; Tiffney 1984; Eriksson et al. 2000b; Heimhofer et al. 2005; Green & Hickey 2005; Feild & Arens 2007; Bond & Scott 2010; Boyce et al. 2010), and mention of flood-plain, swamp and riparian habitats is common (see also Wing & Boucher 1998; Coiffard et al. 2008, 2012) - Aethophyllum, a small (but hardly herbaceous), fast-growing conifer (?Voltziales), had occupied similar habitats in the lower middle Triassic (Rothwell et al. 2000). In the Late Barremian-Aptian 124-112 Ma of Portugal climate and environment were unstable, which might favour angiosperms adapted to disturbed habitats (Heimhofer et al. 2005). It has been suggested that these ± weedy early angiosperms instituted a novel fire regime, being both productive (in terms of biomass) and flammable, and this would have facilitated their spread (Bond & Scott 2010), although a small experiment suggests that "weedy" angiosperms would be unlikely to ignite when fresh, while "shrubby" angiosperms, ferns and conifers were more flammable (Belcher & Hudspith 2016) - however, the higher atmospheric oxygen then might have encouraged burning.

Angiosperms first appear in the pollen/spore record south of palaeolatitude 30o N and by the Aptian, around 125 Ma, they started to become quite abundant (>20% of the record, but see below) and more widespread, spreading both north and south; the earliest pollen grains are monosulcates (Crane & Ligard 1989). Similarly, fossil ephedroid pollen was common early, peaking between the Barremian and Santonian, i.e. 125-83 Ma, angiosperms and ephedroids apparently preferring similar habitats (Crane & Lidgard 1989). Tricolpate pollen, the signature of eudicots, is reported from the Late Barremian-Early Aptian some 125-120 Ma in deposits from southern Laurasia and particularly northern Gondwana (e.g. Magallón et al. 1999; Sanderson & Doyle 2001; Coiro et al. 2019: Fig. 2). However, if the relationships of Leefructus from early Cretaceous deposits 125.8-122.6 Ma old in China and assigned to stem Ranunculaceae (Sun et al. 2011: no associated pollen) is confirmed, these ages may need to be revised (but c.f. W. Wang et al. 2014, 2016a, b). The functional advantage of tricolpate pollen may be that the grains germinate faster than monoaperturate pollen, even if they remain viable for a shorter time (e.g. Furness & Rudall 2004). In west Portugal and elsewhere tricolpates were initially in low numbers (see also Jud 2015), but by the Early Albian ca 112 Ma angiosperms, including tricolpates, were diversifying rapidly, although they were still not very abundant and apparently absent above 50 o N (Heimhofer et al. 2005; Friis et al. 2006b; Horikx et al. 2016; Coiro et al. 2019). For a critical re-evaluation of the North American Potomac floras, largely Aptian to Albian in age (125-100 My), see Doyle and Upchurch (2014). Tricolpate pollen does not appear in southern Australia until the end of Middle Albian/early Late Albian ca 107 Ma, so this is technically part of the next section, and it first appeared at similary high northern latitudes at about the same time, and the overall palynoflora of the two was similar - perhaps there was global warming then (Korasidis & Wagstaff 2019; see also Playford 1971).

5C. Later Cretaceous Evolution - Albian to Maastrichtian: The Upper Cretaceous began ca 99.6 Ma with the Cenomanian and ended ca 66 Ma with the bolide impact in the Yucatan and the eruptions that produced the Deccan Traps. The sea was initially about one hundred and fifty meters above its present level. A long-term warming trend from the early Aptian culminated in the Cenomanian-Turonian thermal maximum ca 99 Ma (Heimhofer et al. 2005: Coiffard et al. 2007, 2007). However, there is the possibility of some glaciation in the Aptian and Maastrichtian (Ladant & Donnadieu 2016; see also Bowman et al. 2014).

The period from 110-80 Ma encompasses the so-called Cretaceous Terrestrial Revolution (KTR: Lloyd et al. 2008; Benton 2010; Meredith et al. 2011), and so the Albian period (ca 113-100 Ma) is also included here. (W. Wang et al. 2016a suggest that the KTR lasted from 125-80 Ma, while for Barba-Montoya et al. 2018, their idea that the major groups of angiosperms diversified 150-100 Ma is compatible with the idea of a KTR.) During this period there were major changes in the terrestrial vegetation (e.g. Crepet 2008; Coiffard & Gomez 2011; Coiffard et al. 2012), although neither geography, i.e. the fact that continents were drifting around, nor humidity seems to have affected this early dispersal of angiosperms (Morley 2003). Pollen indicates an Albian-Cenomanian rise to dominance by angiosperms, perhaps still weedy, rather before the fossil wood record 83.6-66 Ma showing angiosperm wood rising from ca 32 to 78% of the total (Peralta-Medina & Falcon-Lang 2012; Heimhofer et al. 2005). Angiosperms continued to spread latitudinally from the more tropical areas they initially inhabited (Axelrod 1959; Hickey & Doyle 1977; Wing & Boucher 1998; Hemihofer et al. 2005 for further references). By around 100 Ma, the beginning of the Cenomanian, angiosperms were rapidly becoming dominant in terms of species numbers (Crane 1983; de Boer et al. 2012).

There was replacement of conifers, perhaps especially araucariacean conifers, denizens of more tropical climates, by angiosperms around 126-65 Ma. Indeed, it has been suggested that extinction rates in conifers in general may have increased during the period 110-100 Ma, i.e. near the beginning of the Cretaceous Terrestrial Revolution, and the rate of conifer decline may have increased as temperatures decreased in the Oligocene (Condamine et al. 2020). This decline is perhaps the result of competition between conifers and angiosperms; these extinction rates have remained high while speciation rates appear always to have been rather low; the former may be due to more rapid growth, animal pollination, etc., in angiosperms and the latter in part to long generation time, large genome size, etc., in conifers (Condamine et al. 2020 and references). In gymnosperms in general, rates of genome change are low, as are speciation rates, the two being correlated; Pinaceae have overall the lowest speciation rates among land plants (Puttick et al. 2015) in clades of any size. Note, however, that competition can be considered from other points of view, and then conifers may see to come out not so badly - see elsewhere.

However, in the early Albian in Portugal ca 112 Ma angiosperm pollen was only (2-)5.3(-7)% of the total, and that figure had increased only slightly, to (1-)8(-20)%, by the late Albian-early Cenomanian ca 100 Ma; polyporate pollen, a tropical floral element, also occurs in this mid latitude palaeoflora (Horikx et al. 2016). Mid-latitude floras in particular did change considerably, large trees first appearing in the fossil record in the Middle to Late Albian less than 110 Ma, and fossil woods became notably more common (Philippe et al. 2008; Wheeler & Lehman 2009). The diversity of the woods increased, vessels with scalariform perforation plates decreasing from >50% to <20% and woods with broad rays decreasing from ca 25% to 5% in the Albian-Santonian ca 110-85 Ma (Martínez-Cabrera et al. 2017). Nevertheless, in the late Albian ca 107-100 Ma angiosperms are still described as being herbaceous to shrubby early successional plants of open habitats where they competed with, but did not necessarily displace, ferns (Peppe et al. 2008: Aspen Shale, Wyoming), and ephedroid pollen was also common between 30o N and S (Crane & Lidgard 1989). There was a fair diversity of pollen in Australia by the late Albian ca 102 Ma, and it showed similarities with the palynoflora of southern South America, suggesting that the climate had warmed and eudicots were moving from South America to Australia via Antarctica (Korasidis et al. 2016). By the Albian-Turonian some 100 Ma there was a significant number of eudicots in mid-latitude North America (ca 450 N), and by the end of the Cretaceous they were about 40% of the flora at 80o N and about double that number towards the equator (e.g. Crane & Lidgard 1989, 2000; Lupia et al. 1999). Platanoid fossils were found along river channels ca 110 Ma in the Late Albian of northern Alaska (Spicer & Herman 2010; Pott et al. 2012 and references), and although frequent associates included Ginkgo and Metasequoia, as well as Cercidiphyllum (Royer et al. 2003), angiosperm abundance increased there, although diversity in the high Arctic was low and there was at most little endemism (Wolfe 1987; Spicer & Herman 2010).

Most fossils from the Aptian/Albian ca 112 Ma still have odd assemblages of characters (see also Friis et al. 1995; Horikx et al. 2016: pollen). Thus Friis et al. (2017a) described Kenilanthus - P 5, A extrose, pollen tricolpate. The flowers of Paisia from Portugal, 5-merous, with pantoporate pollen and perianth, stamens, and carpels all opposite each other are from this time - these flowers are less than 1 mm long (Friis et al. 2018a). Slightly younger is Antiquifloris, in amber at least 99 Ma from Myanmar, which has some magnoliid features like broad filaments with three traces and a prolonged connective, but just a single style, and the flower was about 6 mm long (Poinar et al. 2016a), Zygadelphus aetheus, from the same deposits, has a spirally arranged perianth, two styles, and four stamen pairs, the smaller stamen developing on the back of the anther of the larger stamen (Poinar & Chambers 2019b), while the enigmatic Prisca reynoldsii with its slender receptacle 5-7 cm long grew 94-92 Ma in Kansas (Retallack & Dilcher 1981). At ca 100 Ma or slightly younger there are fossils like Cecilanthus, from early Cenomanian Maryland with a floral formula of * P many; A many; G many, with a well-developed receptacle and probably one ovule/carpel, that while perhaps Magnolialean cannot really be assigned to any extant family (Herendeen et al. 2016). Groups like Laurales had become common by around 112 Ma (J. A. Doyle & Upchurch 2014), and as late as the Cenomanian ca 96 Ma many fossils are probably of plants of the ANA-magnoliid grade (e.g. Coiffard et al. 2006; Kvacek & Friis 2010; Friis & Pedersen 2011). Woods from this period like Paraphyllanthoxylon and Icacinoxylon are also difficult to identify, the former, for example, having been compared with woods both of Laurales and of three completely unrelated rosid families (Martínez-Cabrera et al. 2017).

Tricolporate pollen, common in Pentapetalae today, is first known from around 107 Ma in the mid Albian (Friis et al. 2011), and Crepet (1996, 2008) noted the first appearance of many floral characters of the Pentapetalae beginning in the Albian, but especially the Cenomanian/Turonian some 100.5-90 Ma, however, these early pentapetalan fossils are difficult to identify. Indeed, although tricolporate pollen is common in Pentapetalae, it is not restricted to them (e.g. Coiro et al. 2019). A flower, the Rose Creek fossil from the earliest Cenomanian in Nebraska about 99 Ma, is the earliest known pentapetalan fossil flower. It was originally described as having five stamens that are somewhat unexpectedly opposite the petals, fused carpels and short, spreading styles (Basinger & Dilcher 1984), but recent work has suggests that the plant had 10 stamens (the androecium is shown as being obdiplostemonous) and fruits have been associated with it; it has been described as Dakotanthus cordiformis (Manchester et al. 2018a). Caliciflora mauldinensis, of about the same age and from Maryland, is also pentapetalan, but it has a rather different floral morphology, not to mention stellate hairs and a valvate-recurved calyx (Friis et al. 2016). The floral formulae of the two are * K 5; C 5; A 10; N+; [G 5] and * K 5; C 5; A 8 [?10]; N-; G 3 respectively. Tropidogyne, one of the eudicots from amber deposits in Myanmar and also, at ca 99 Ma, of about the same age, has a floral formula of * K 5; C ?0; A 10; G [3]; N+ (Chambers et al. 2010; see also Poinar and Chambers 2017). Other fossils to be taken into account include Eoëpigynia burmensis (Poinar et al. 2007), Lachnociona (Poinar et al. 2008) and Lijinganthus revoluta (Z.-J. Liu et al. 2018b: tricolpate pollen, G [3], style single, bisporangiate anthers, ?calyx), and there are other tantalizing early records (e.g. Poinar 2011), all in amber from Myanmar (as of xii.2019, some 15 genera had been described from these deposits, see Poinar 2018; Poinar & Chambers 2019a). Interestingly, there is tricolpate pollen on a tumbling flower beetle ca 4.25 mm long, Angimordella burmitina (Mordellidae), that was also found in the Myanmar amber deposits (Bao et al. 2019). These Myanmar fossils may be core eudicots, but hardly surprisingly none can be securely placed in an extant family. Furthermore, although it was thought that Tropidogyne and Lachnociona might be Cunoniaceae, their ages conflict with other age estimates for that clade; the Rose Creek flower might be rhamnaceous, but it is notably larger than most extant Rhamnaceae (Jud et al. 2017 for references). Described as Dakotanthus cordiformis, it is thought to show a particularly close association (among extant angiosperms) with the southern South American Quillajaceae (Manchester et al. 2018a). In a formal phylogenetic analysis it was placed in a number of positions (equal maximum parsimony) in the rosids, both Rosales and Sapindales, but also in Fabales including Quillajaceae again, and that despite not being able to include a couple of characters that are particularly characteristic of extant Quillajaceae (Schönenberger et al. 2020).

Of course, some fossils are exceptions. Thus flowers of Kajanthus lusitanicus, from Portugese Cretaceous deposits around 113 Ma, may even be assignable to crown-group Lardizabalaceae (Mendes et al. 2014). Y. Wu et al. (2018, age from Prasad et al. 2011) suggest that basal Poaceae (i.e. non-PACMAD grasses) became widely distributed across both Laurasia and Gondwana 129-125 Ma during the Barremian. Many major gentianid, rosid and monocot clades seem to have radiated by around 90 Ma at the latest (e.g. Sanderson et al. 2004; Jian et al. 2008 and references; Wang et al. 2009). The diversity of floral form in the Turonian (93.9-89.8 Ma) of east North Americas is very considerable, magnoliids, rosids and asterid-Ericales all being represented (e.g. Crepet & Nixon 1996). Thus Crepet et al. (2018) recently described the ca 90 Ma Teuschestanthes, covered by scales and also perhaps Ericales - and with the odd floral formula ↑ K [5]; C 5; A 8 + 8 nectariferous staminodes; G [3]. Rosids in particular were common in the Late Cretaceous (Friis et al. 2010b). Flowers perhaps assignable to Saxifragales (ovary inferior, crowned by a nectary, styles more or less separate, i.e. they look very like the old - polyphyletic - woody Saxifragaceae) were especially common, as were ericalean flowers with a variety of morphologies (Friis et al. 2006b, 2011). Indeed, Saxifragales, although now not very speciose, may represent an ancient and rapid radiation (Fishbein et al. 2001; Fishbein & Soltis 2004; Jian et al. 2008). Flowers from this period commonly have spreading petals and stamens with the anthers distinct from the filaments and short styluli or styles; they are still mostly quite small (Friis et al. 2011), thus the flowers of Caliciflora at about 1.5 mm across represent one end of the size spectrum (see also Crepet et al. 2018). At ca 2.5 cm across, the Rose Creek flower/Dakotanthus is relatively large compared to the flowers of many other angiosperms of its age (ca 100 My), and there are a few other fossils of quite large flowers, mostly terminal in position. Fossils referable to extant angiosperm families begin to appear in east North America around 115-90 Ma, and by some 85 Ma their diversity had increased considerably (Crane & Herendeen 1996; also Lidgard & Crane 1988; Friis & Crepet 1987; Friis & Endress 1990; Crepet et al. 2004, etc.).

Peris et al. (2017) talk about an Aptian—Albian gap in insect pollination, i.e. a period 125-100 Ma in which insect groups that pollinated gymnosperms became less common (or even went extinct) and those pollinating angiosperms, including groups that had not pollinated gymnosperms, became more common. By the mid-Cretaceous pollen had become more abundant and is quite often found in clumps, suggesting that pollenkitt had evolved and that pollinators were becoming more specialized (Hu et al. 2008; Leslie & Boyce 2012). Interestingly, the putative early bee Melittosphex (placed along the stem group of modern bees) collected from the same amber locality as the 99 Ma Caliciflora is a mere 3 mm long (Danforth & Poinar 2011). Nectaries and "food sources" have been reported in flowers collected from Lower Cretaceous amber from Myanmar ca 99 Ma (Santiago-Blay et al. 2005), and angiosperm flowers from the Cenomanian-Turonian 110-90 Ma have a variety of quite specialized zoophilous morphologies, and nectar secretion became common (Crepet 1996, 2008; Hu et al. 2008), while nectaries are conspicuous in floral diagrams drawn for Late Cretaceous flowers (Friis et al. 2011: fig. 16.6, 17.10). With the diversification of Pentapetalae nectar produced by receptacular or gynoecial nectaries is likely to have become a common reward for pollinators (Friis et al. 2006b; Erbar 2014). Of course, nectar is a major pollinator reward in many extant angiosperms, and both receptacular and ovarian nectaries are found in extant Proteales (even some fossils of the now wind-pollinated Platanaceae are described as having nectaries), Trochodendrales and Buxales.

Citerne et al. (2010) thought that 93.5-89 Ma in the Turonian was a period of floral innovation and evolution of pollinators, while Cardinal and Danforth (2013) suggested that there is link between the diversification of the eudicots and that of bees (Anthophila). Cardinal and Danforth (2013; see also Cardinal et al. 2018) suggest that crown-group Anthophila are some (132-)123(-113) Ma, all families having diverged by the K/P boundary, and other crown-group ages are ca 125 Ma (Ronquist et al. 2012) to ca 112 Ma (Grimaldi 1999), while Sann et al. (2018) date stem-group Anthophila at (148-)128(-108) Ma, slightly older than ages in Peters et al. (2017a) who offer ages of (147-)124, 111(-93) Ma, the beginning of the KTR, so there may - perhaps - be a connection between the KTR and bee evolution. Peters et al. (2017a) thought that the crown-group age of corbiculate bees, many eusocial, was around 62 Ma, i.e. early Palaeocene. Interestingly, the plesiomorphic condition for pollination specificity in extant bees seems to be oligolecty, oligolectic bees tending to pollinate rather few species of related plants, yet the flowers they pollinate appear to be unspecialized. Indeed, unspecialised flowers precede specialised flowers, the latter often pollinated by one or a few species of polylectic pollinators that also visit a variety of unrelated flowers (for more on oligolecty versus polylecty, see below). Sympetaly (Actinocalyx bohrii had flowers ca 3 mm long with connate petals) and monosymmetry (evidence for the latter is indirect, seeds and other fossils assignable to Zingiberales: Rodríguez-de la Rosa & Cevallos-Ferriz 1994) appear in the Late Cretaceous, but in general monosymmetric flowers are uncommon in the Cretaceous (Friis 1985; van Bergen & Collinson 1999; Friis et al. 2003a, 2011). Rather surprisingly, ca 50% of end-Cretaceous mesofossil flowers had inferior ovaries, a higher proportion than in the present flora (Crepet & Friis 1987; Friis et al. 2011: fig. 16.8). Of course, some of these Late Cretaceous plants are still hard to place, thus the monocot Viracarpon, some 67 Ma, cannot be identified much more precisely (Matsunaga et al. 2018).

Insect-plant relationships [Add from elsewhere, and/or move to elsewhere...]. Crown-group Coleoptera have been dated to the beginning of the Permian ca 297 Ma (some other estimates 333-253 Ma), and of the really large beetle families (>10,000) species seven of these, eating plants, plant detritus, or wood, diversified in the Cretaceous, while the other two, predominantly carnivores, originated in the early Jurassic (S.-Q. Zhang et al. 2018). Phytophaga, which inludes the three large groups, the weevils, chrysomelids, and cerambycids, began to diversify in the mid-Jurassic ca 175 Ma, and although diversification rates in beetles may have started increasing in the Triassic, they were very high indeed throughout the Cretaceous (S.-Q. Zhang et al. 2018). Ages of obligately symbiotic beetle-fungus associations in fungus farming by weevils (ambrosia beetles: for details of this association, see also below and papers in Vega & Hofstetter 2015) that grow on a variety of angiosperms, their original hosts, although some are on gymnosperms, have been suggested. Thus the origin of fungus farming in platypodine weevils, a major clade of ca 1,400 known species of ambrosia beetles, has been dated to ca 119 Ma (stem age) and 88 Ma (crown age) (Jordal 2015; Vanderpool et al. 2017: Tesserocerini are paraphyletic). It is estimated that the ages of ascomycete ophiostomatalean fungi that are obligate associates of these weevils are ca 101 (stem) and 86 (crown) Ma (something of an underestimate?), with rapid diversification soon after the latter date (Vanderpool et al. 2017). An ophiostomalean ambrosia fungus, Paleoambrosia entomophila, has recently been described from the mycangia of a platypodine beetle found in amber from Myanmar ca 98.8 Ma (Poinar & Vega 2018). To summarize: All the evidence suggests that these first beetle-fungus associations developed during the KTR. Diversification of ambrosia beetle scolytine weevils has been very largely within the last 50 Ma or so (however, fossils identified as the extant genus Microborus, a near basal scolytine, have been found in amber from Myanmar ca 98.8 Ma - Cognato & Grimaldi 2009). Today scolytines are notably diverse in (sub)tropical rain forests, which may say something about the development of such forest, indeed, Platypodinae-Platypodini themselves did not start diversifying until the early Eocene, ca 54 Ma, and there were only five clades of Platypodinae then (Vanderpool et al. 2017). Fungus cultivation in termites and ants is similar in age (Jordal & Cognato 2012). There are some 11,000 species of predominantly leaf-rolling Tortricidae moths, the 10th most diversified clade of phytophagous insects (Fagua et al. 2017). The stem group age of that clade is (140.5-)120.5(-99.3) Ma, and there is a fuse of slightly under 25 Ma before crown-group diversification began (118-)97(-75.9) Ma; according to the model used, there was an increase in their diversification rate around 72 Ma and a fairly gradual decline since (Fagua et al. 2017) - the K/P boundary might as well not exist for these moths. Angiosperms were becoming dominant ca 100 Ma, so it would seem that their rise to dominance enabled the subsequent diversification of Tortricidae, which was initially on the southern continents, as with Lepidoptera as a whole (Wahlberg et al. 2013: ?location). As with bees, tortricid leaf rollers were initially oligophagous, but polyphagy is common in more recently diverging Tortricinae lineages (Fagua et al. 2017 and references). Yponomeutoidea are another large group (ca 1,800 species) of rather small moths that represent an early ditrysian radiation, and they include a variety of mostly oligophagous internal and external feeders that are today especially common on rosids; associations with woody plants are likely to be plesiomorphic here (Sohn et al. 2013). However, van Eldijk et al. (2018) found scales of non-ditrysian Glossata 212 Ma in Germany, and thought that that suggested that caterpillars of these moths initially ate gymnosperms, only later moving on to angiosperms even though angiospermivory might seem to be the ancestral condition for glossatan (with probosces) moths. Indeed, crown-group Angiospermivora, = [Heterobathmoidea + Glossata], have been dated to (276.7-)257.7(-234.5) Ma, and caterpillars of this clade may have initially been leaf miners, perhaps feeding on stem-group angiosperms (Kawahara et al. 2019). At least 27 families and of mites live in acarodomatia that are found on over 2,000 species and 80 families of broad-leaved angiosperms (Maccracken et al. 2019). The oldest domatia in the fossil record date to the Late Cretaceous in North America ca 75.7 Ma on an unidentifiable and much-eaten leaf, and other early records are from the Eocene; mites are of course far older, and Maccracken et al. (2019) suggest that domatia may have first appeared about 100 Ma along with the evolution of the woody habit.

The macro- and mesofossil record is likely to be skewed to large, woody plants, although W. Wang et al. (2016a) discuss the evolution of angiosperm-dominated herbaceous floras in more open vegetation. They suggest that in Ranunculaceae, at least, herbaceousness evolved in plants living in forests in which angiosperms were becoming dominant during the KTR before moving into more open conditions, and they date initial diversification in Ranunculaceae to around 109 Ma, the first branches in the family being represented by forest-dwelling herbs (Wang et al. 2016a: cf. topology and ages, crown-group Ranunculaceae (ca 119-(114.6-)108.8(-101.6)-ca 82) Ma - see also their Fig. 2). They note that a similar scenario may explain the evolution of other herbaceous clades like the grasses and orchids. However, the great expansion of open, herb-dominated vegetation exemplified by today's grasslands and savannas and the massive diversification of a largely epiphytic clade of Orchidaceae are very much later phenonomena, and it is also unclear exactly what kind of angiosperm forest these Ranunculaceae were inhabiting, given that Wang et al. (2016b) thought that Ranunculales represented a very early radiation of the eudicots and that all the other early groups they mentioned are either non-forest plants, magnoliids, or extinct clades of uncertain affinities - Platanaceae are the only exception.

By the Late Cenomanian/Early Turonian ca 93.5 Ma there were four main pollen provinces. 1. The tropical province was dominated by palm-type pollen, although conditions may commonly have been rather dry (Burnham & Johnson 2004). Indeed, fossils of palm stems of Cretaceous age have fibro-vascular bundles with two vessels per bundle as in extant palms that live in places with a dry period (Thomas & Boura 2015 and references). 2. Plants with distinctive pollen assignable to the Normapolles complex and comparable with that of some extant Fagales were both diverse and ecologically prominent in rocks from east North America to western Asia (Friis et al. 2006a); interestingly, fossil flowers with such pollen are often perfect, while flowers of extant Fagales with Normapolles pollen are generally imperfect. 3. Of the other pollen provinces, Aquilapollenites, a pollen type of unknown relationships (Santalales have been suggested), characterised an area that made up the rest of the more temperate northern hemisphere. 4. The southernmost province was characterized by Nothofagites pollen, probably also from plants that would be included in Fagales (e.g. Pacltová 1981 for a review; Kedves & Diniz 1983; Diniz 1988; Friis et al. 2006a, 2010b, 2011; Cantrill & Poole 2012). Extant Normapolles plants and Nothofagaceae are trees, although some are shrubs, and Fagales as a whole are largely ectomycorrhizal (ECM). Pinaceae, also ECM, were also part of the vegetational mix then, so ECM plants may have been an appreciable component of the vegetation, at least locally, and this has implications for carbon sequestration (see also below).

Seeds/diaspores in the later Cretaceous remained small, and abiotic dispersal may have been prevalent (Tiffney 1984, 1986a, b), however, Eriksson et al. (2000a) suggested that biotic dispersal was probably appreciable. For Tiffney, ants, fish, reptiles and archaic mammals were possible dispersers, but not birds, for Eriksson et al. birds and multituberculate mammals (see also Eriksson 2016) were likely candidates. Eriksson et al. (2000a) distinguished between the "disperser" and "recruitment" hypotheses for changes in fruit size; in the former, the evolution of animals that could handle large fruits drove the evolution of plants with large diaspores, while in the latter the increasing size of angiosperms was associated with larger seeds/diaspores that were initially dispersed by rather generalist dispersers, only later did more specialist frugivores evolve. Eriksson et al. (2000a, see also Moles et al. 2005a; c.f. Tiffney 1984) incline to the latter hypothesis, and from what is known about mammal and bird evolution (see below), this seems most likely. Increasing seed size became evident about 85-75 Ma, probably reflecting the increasing size of the plants (e.g. Eriksson et al. 2000a; Moles et al. 2005a, b; Eriksson 2016; Kavanaugh & Burns 2014: increase in size on islands independent of habit), while by 70 Ma around a quarter of the disseminules in European palaeofloras were drupes (thinking of animal dispersal, add berries in Annonaceae, Vitaceae, etc.), a proportion that held quite steady until the early Eocene, at least (Eriksson 2016). Interestingly, fruits and seeds are larger and the former are more likely to be fleshy in woody vegetation of the humid tropics today, or at least in areas that have some periods of heavy rainfall (S.-C. Chen et al. 2016 and references).

Although carnivorous plants are unlikely ever to have been major elements of the vegetation, their distinctive morphology and physiology attracts attention. There are reports of fossils of carnivorous plants from the earlier (Archaeamphora, Sarraceniaceae, ca 124 Ma, see H. Li 2005a, b) and later (Paleoaldrovandra splendens, Droseraceae, ca 90 Ma, Knobloch & Mai 1984) Cretaceous, however, the former is probably a conifer gall (Herendeen et al. 2017 for literature) and the latter an insect egg (Hermanová & Kvacek 2010). However, molecular ages for Sarraceniaceae and Droseraceae for example, do go back to the Cretaceous, those for the latter (or more specifically, the stem-group age of Droseraceae) being anything from 75-67 (Wikström et al. 2001) to ca 99.3 Ma (Magallón et al. 2015; see also Yao et al. 2019).

Around 108-94 Ma (Late Albian), and again at the end of the Cretaceous, the venation density of angiosperms increased and became markedly greater than that of non-flowering plants and ANA-grade angiosperms (Feild et al. 2011b; c.f. Bond & Scott 2010 in part, discussed further below), and l.t.r.f. of sorts may have begun to spread ca 100 Ma (Eiserhardt et al. 2017, see also below). Friis et al. (2006a; see also Heimhofer et al. 2005) note a dramatic increase of phylogenetic diversity and ecological abundance of angiosperms at this time. Towards the end of the Cretaceous there were changes in atmospheric CO2 concentration, etc., and some taxa in the Aquilapollenites flora went extinct, perhaps because of the loss of their pollinators (Spicer & Collinson 2014), although extinction may also have been the result of the bolide impact (Fuentes et al. 2019).

Eudicots replaced free-sporing plants (see also Fiz-Palacios et al. 2011: "continuous replacement"), but initially not conifers and ephedroids (see e.g. Crane & Ligard 1989; Wing & Boucher 1998; Lupia et al. 1999). The decline of cycads and Bennettitales (cycadophytes, an ecological grouping) might be linked with the contemporaneous decline in herbivorous stegosaurian dinosaurs, but there is no indication of any even loose coevolutionary relationships between early angiosperms and dinosaurs (Barrett & Willis 2001; Butler et al. 2009; Barrett 2014 and references, but c.f. Bakker 1978; Y. Wu et al. 2018: dinosaurs eating grass, also below). A variety of insects are associated with these gymnosperms, including scorpion flies (Mecoptera) and the remarkable butterfly-like kalligrammatid lacewings, and although Labandeira et al. (2016) thought that the latter had declined by ca 120 Ma, they are well represented in amber from Myanmar ca 90 Ma, even if the demise of gymnosperms may have been a factor in their own demise (Q. Liu et al. 2018); again, connections are hard to establish.

Areas where conifers remained common seem to have become more restricted, ecological factors such as slow seedling growth, details of leaf construction, narrow stomatal apertures (ca 2 µm: Walker 2005), etc., perhaps explaining this (e.g. Bond 1989). The increase in venation density in angiosperms to 6 mm/mm2 and more around 100 Ma coupled with high stomatal conductance, etc., enabled higher photosynthetic rates at a time when atmospheric CO2 concentrations were declining and probably made angiosperms ecophysiologically more competitive compared with gymnosperms, and also more flexible (McElwain et al. 2015; Simonin & Roddy 2018; Yiotis & McElwain 2019, and references). Furthermore, Sack et al. (2012) note that in large simple leaves with high venation density (an angiosperm innovation, as we shall see), the high venation density allows the leaf to function even when irradiance is high because the evaporating water copiously supplied by dense veins cools the leaf. However, conifers have a very high leaf area index, a measure of leaf area/unit ground surface area, of up to 21 (Maguire et al. 2005), and the leaves can be very long-lived, and such features in extant conifers still make them formidable competitors with angiosperms in well-lit conditions on soils that are other than nutrient-rich (Brodribb et al. 2012; see also Chabot & Hicks 1982: still useful). Conifer pollen and also spores from "bryophytes" and "pteridophytes" decreased in diversity, although gnetalean pollen was quite diverse at rather lower latitudes in the mid-Cretaceous (Crane & Lidgard 1989; see Y. Yang et al. 2017 for references to the rich fossil history of gnetophytes). After 100 Ma there was a moderate decrease in pteridophyte diversity, the beginnings of a decrease in gymnosperm and cycad diversity, and a dramatic increase in angiosperm diversity (Niklas et al. 1983; S. Brown et al. 2012) - although diversity and ecological importance/dominance are not simply connected, as is evident in the current flora where the area dominated:species number ratio is much higher for Pinales than angiosperms...

Fires were relatively common throughout the Cretaceous, and they may have encouraged/been encouraged by a rather shrubby, low stature vegetation with a relatively short life cycle (Bond and Scott 2010; He et al. 2012; Bond & Midgley 2012; Hill & Jordan 2016 and Carpenter et al. 2016: both Australia; c.f. Belcher & Hudspith 2016, in part; Lamont et al. 2018b); Cretaceous mesofossils are often charcoalified (Friis et al. 2011). The increasing prevalence of fires may have been an important element of the KTR. Berner (2003) noted that rocks rich in charcoal derived from plants are particularly prominent in the mid Cretaceous 120-90 Ma to the Palaeocene, and fires may also have been encouraged by the relatively high atmospheric oxygen concentrations of 21-25% then (S. A. E. Brown et al. 2012). Lamont et al. (2018b: fig. 12) suggest that there was a small peak in the prevalence of fires later in the Cretaceous ca 110 Ma, and with considerable effects on the evolution of angiosperms. Moreover, the presence of shrubby angiosperms and ferns in the mid Cretaceous may have increased the prevalence of intense (especially the burning shrubby angiosperms) and rapidly-spreading fires, and conifer forests then may have preferentially burned, being replaced by angiosperms (Belcher & Hudspith 2016). Frequent fires were likely to have been accompanied by increased runoff/erosion and loss of P to the ocean (Brown et al. 2012), while nitrogen is lost by volatilization (Forrestal et al. 2014 and references). Lamont et al. (2018a) suggested that by the early Upper Cretaceous fire-adapted traits had become common in plants from a number of clades, not just Pinus. Thus fire-adapted Proteaceae-dominated heathlands are to be found ca 89-84 or 75-65.5 Ma in Central Australia, numerous species of Proteaceae being involved, and Banksia may have evolved by then (Lamont & He 2012; Carpenter et al. 2015 respectively; Myrtaceae were not involved, spores of Sphagnum were abundant - Carpenter et al. 2015). Indeed, He et al. (2016b) suggest that at this time adaptations associated with the prevalence of fires were evident in members of the Cape flora like Restionaceae. For fires and the evolution of groups like Proteaceae, Fabales, Haemodoraceae, Myrtaceae, etc., see Lamont et al. (2018b). Substantial amounts of inertinite, fossilized charcoal from fires in mire systems, are found through the Cretaceous, but less since, even quite recently, which is difficult to understand (Scott & Glasspool 2006; Bond 2015). It is thought that fires were unable to burn closed angiosperm forests when these finally developed in the Caenozoic (Bond & Midgley 2012).

In the Albian-Cenomanian of Europe ca 100 Ma angiosperms were most evident in backswamp, flood plain, levee, and braided river habitats (Coiffard et al. 2006), and the deciduous habit was relatively common in these habitats, and, more generally with deciduous conifers, north of 65-70o N and through the Eocene (Wolfe 1987). Prisca was a common tree of such habitats in Kansas 94-92 Ma (Retallack & Dilcher 1981). As angiosperms were increasing around this time, gymnosperm groups like Ginkgoales, Bennettitales, Caytoniales, Cycadales, Czekanowskiales and Gnetales were all decreasing, even if conifers, a number of species of which prefer backswamp conditions, did not decline (Royer et al. 2003). Turonian forests of ca 90 Ma were still found primarily in disturbed and/or riparian-type habitats (see also Jud & Wing 2013; Spicer & Herman 2010: Late Cretaceous in northern Alaska; I. Miller 2013: 105 Ma Albian); Platanaceae, found along channel margins in the Cenomanian, had spread onto flood plains in the Turonian. In Australia, angiosperm pollen had increased from a low level in the middle Albian ca 105 Ma to about 35% of the total spores at the end-Cretaceous, pollen of free sporing plants dropping from 80% to 45% over the same period. However, not all fern families behaved the same, and there are differences between Australia and North America (Nagalingum et al. 2002). By ca 80 Ma angiosperms had come to make up ca 40% of both floristic diversity and abundance even at higher latitudes (e.g. Axelrod 1959; Crane & Lidgard 1989; Lupia et al. 1999; see also Nagalingum et al. 2002), with monocots and magnoliids predominating between 50oN and 20oS. Things are made more confusing by the persistent predominace of woods with simple perforations in fossil woods from the tropics in the later Cretaceous, while in more temperate climes fossil woods show Baileyan trends, with features like scalariform perforation plates initially being common (Wheeler ∧ Baas 2019). In the late Cretaceous-earliest Palaeocene in South America, Australia and on the Antarctic Peninsula podocarps, Proteaceae, Myrtaceae, Winteraceae, Araucaria and Nothofagus were major components of the vegetation, with open ericaceous heathlands at higher altitudes (Askin 1989; Dettmann 1989; Bowman et al. 2014).

Tree trunks up to some 1.8 m in diameter and thought to be from a plant somewhat over 50 m tall are known from Turonian deposits ca 92 Ma in Utah (Jud et al. 2018b). Jud et al. (2018b) suggest that the plant, placed in Paraphyllanthoxylon, may have been a member of Lauraceae, a family in which the pollen exine is reduced so the grains do not preserve well; pollen of palms was found fairly close. (Note, however, that P. alabamense, with which Jud et al. 2018b compared their wood, is in a group of species whose woods have been compared with those of Elaeocarpaceae, Burseraceae and Anacardiaceae, while wood of other Paraphyllanthoxylon species have been compared with at least seven more unrelated families - see Herendeen 1991...) In parts of Campanian (83.6-72.1 Ma) North America, angiosperms seem to have lived in rather species-poor and open woodlands (Lehman & Wheeler 2001; Wheeler & Lehman 2001, 2009). The trees, up to 1.3 m diameter, may have produced pollen of the Normapolles type (Lehman & Wheeler 2001). Today such pollen is associated with Fagales, largely an ectomycorrhizal group, so a diverse forest might not be expected - contemporary forest dominated by Fagales are not notably speciose. However, identifications of these woods - and woods like Paraphyllanthoxylon and Icacinoxylon have been placed in a wide variety of families - rarely include Fagales (Wheeler & Lehman 2009). Trees up to 2 m in diameter have been found in Late Cretaceous riparian swamps (Parrott et al. 2013).

When - and where - closed-canopy angiosperm-dominated forest first appeared is of considerable interest. Fleshy fruits reported in monocots 120 Ma may reflect the closing of the canopy (see also Dunn et al. 2007), but they are not associated with any particular vegetation type. Rather, it may have been the Late Cretaceous that was the "dawn of modern angiosperm forests" (Coiffard & Gomez 2011: p. 164; also Coiffard et al. 2012). Until the Mid or even Late Cretaceous angiosperms were mostly small herbs to small trees of the understory growing in dryish conditions (Bond & Scott 2010), perhaps rather weedy plants (Feild et al. 2011b). Analyses using variables like leaf area and vein density, plant height and seed size, suggest angiosperms were mostly not canopy trees, seed size remaining small (Jud & Wing 2013). However, some low latitude floras were dominated by angiosperms in the Cenomanian-Turonian ca 94 Ma (Coiffard et al. 2012 for references). Later in the Cretaceous angiosperm diversity was quite high even close to the Arctic Circle (Hofmann et al. 2011; see Spicer & Herman 2010 for the high Arctic). Strömberg et al. (2013a) suggested that in the Late Cretaceous (ca 73 Ma: Wyoming) angiosperms were not notably abundant compared with other co-occurring vascular plants, but they did vary considerably in niche optima and niche breadth. (Similarly, non-polypodiaceous ferns dominated in Late Campanian [73 Ma] North American fern prairies, but Polypodiales, less dominant, were quite diverse - Wing et al. 2013.) However, angiosperms may have formed a canopy at least locally by the end-Cretaceous (e.g. Upchurch & Wolfe 1987; Crane & Lidgard 1990; Boyce et al. 2010), although there is no evidence of closed-canopy forests in Maastrichtian vegetation from N.W. South America (H. V. Graham et al. 2019). How diverse forests then were is unclear. Afrocasia (Araceae) from over 72 Ma is thought to have lived in terrestrial forest understory (Coiffard & Mohr (2016). There was an end-Cretaceous rise to dominance of angiosperms in Patagonia (Iglesias et al. 2011), with vegetation-induced warming at high latitudes at this time (Otto-Bliesner & Upchurch 1997). Wing and Boucher (1998: p. 379) concluded that diversification of flowering plants then was "the evolution of a highly speciose clade of weeds but not necessarily a major change in global vegetation", while Eriksson et al. (2000a) suggested that Late Cretaceous vegetation was open, rather dry (leaf size was relatively small - Upchurch & Wolf 1987), and disturbed by herbivores such as herds of large dinosaurs (for their speciation then, see Sakamoto et al. 2016).

5D. Venation Density, Stomatal Size, and Vascular Evolution: For the evolution of angiosperm leaves in the mid-Cretaceous, the paper by Doyle and Hickey (1976) is still well worth consulting. Atmospheric CO2 concentration declined from the late Jurassic-early Cretaceous to the later Oligocene ca 40 Ma, bottoming out in the Pleistocene (Rothman 2001; Shi & Waterhouse 2010; Franks et al. 2013). This provides the background for thinking about changes in CO2 uptake and water loss, both associated with increased photosynthetic efficiency. Overall, a combination of features unique to many, but not all, angiosperms - xylem dominated by vessels, and leaf blades with high venation density (Sack et al. 2012), precise and optimal positioning of the veins in the blade, and dense, small stomata - allowed productivity to increase and more carbon to be sequestered (e.g. Boyce & Zwieniecki 2012; Zwieniecki & Boyce 2014; Simonin & Roddy 2018); note, however, in arid habitats veins may be closer than would appear to be optimal (de Boer et al. 2016). Simonin and Roddy (2018) suggest that these linked changes were enabled by decreasing nuclear size, which can probably be considered an apomorphy of angiosperms (see above) even if basal clades do not have the combination of foliar features just mentioned. A productivity increase may have contributed to the declining atmospheric CO2 concentrations, as may the decrease in continental arc vulcanism (McKenzie et al. 2016).

The ecological context for the evolution of venation density and vasculature can perhaps be provided by living members of the ANA grade (Feild 2005; Feild & Arens 2005; Feild et al. 2009a), other than the aquatic Nymphaeales; J. A. Doyle and Upchurch (2014) also noted the similarity between leaves of early angiosperms and those of the ANA grade. Vessels in magnoliids and ANA-grade angiosperms, so-called "basal vessels", are rather different from those in core eudicots, and water conductivity is not high (Hacke et al. 2007; Sperry et al. 2007). Venation density of the leaves is low, and spacing of the veins is suboptimal (Zwieniecki & Boyce 2015), as in non-angiosperm vascular plants. Ancestral angiosperms are likely to have had low drought tolerance (Feild et al. 2009a, 2011a, c); their leaves had large and distant stomata, often lacked any palisade mesophyll tissue, and the abaxial surface of the blade reflected light back inside (Feild & Arens 2007).

If early angiosperms were pioneer plants, they might be able to tolerate high herbivory because they had metabolically cheap, rather thin, rapidly expanding leaves with a low amount of fibre and low concentrations of secondary metabolites like terpenoids, phenols, and tannins; their high quality habitat allowed rapid growth and low defence (see e.g. Bond & Scott 2010), and their whole reproductive cycle was relatively short (e.g. Verdú 2002; J. H. Williams 2008, 2009). Extant angiosperms with the highest leaf venation densities are woody pioneers (Feild et al. 2011b). Royer et al (2010) estimated the SLA of fossil angiosperms 110-105 Ma was low, implying fast nutrient turnover, and the disturbed habitats of early angiosperms are likely to have had elevated levels of nutrients (see also Berendse & Scheffer 2009). However, many extant magnoliids and ANA-grade angiosperms have a "slow" leaf spectrum with a higher SLA, etc. (G. Liu et al. 2014), and their venation density is low; there is no indication of a pioneer strategy for early angiosperms here.

Venation. During the 200+ million years prior to the diversification of flowering plants, the venation density of the leaves of vascular plants held largely constant at below ca 3 mm/mm2, and this despite considerable fluctuations in atmospheric carbon dioxide concentrations (e.g. Boyce et al. 2010; Lee & Boyce 2010; Boyce & Zwieniecki 2012; Boyce & Leslie 2012 for a summary). Venation density in non-flowering plants continued to hold steady through the Cretaceous (Feild et al. 2011b). Extant members of the ANA grade, Chloranthaceae, shade tolerant and succulent plants, etc., as well as fossils from the first ca 30 Ma of the angiosperm record all have similar low venation densities of around 2.4 mm/mm-2 (Feild et al. 2011b: post Hauterivian), and they would also have lower CO2 exchange than most magnoliids and basal eudicots (Feild et al. 2011a).

Most extant angiosperms have distinctive hierarchical-reticulate and very dense venation. Small-diameter minor veins develop during the final expansion phase of the angiosperm leaf, the density of these veins contributing greatly to the dense venation of many angiosperm leaves. Before ca 113 Ma (pre-Albian) about half the fossils still had leaf blades with a low venation density ca 3 mm/mm2 (e.g. Feild et al. 2011b; Boyce & Zwieniecki 2012). However, the venation density of many angiosperms doubled to abobe 6 mm/mm2 some 108-94 million years ago in the late Aptian-early Cenomanian, a change occurring independently in monocots, magnoliids and eudicots (Boyce et al. 2009; Feild et al. 2011b; de Boer et al. 2012), and the veins became much more ordered hierarchical-reticulate. This density increase greatly reduced the main element in the resistance to water flow through the plant by shortening its path through the mesophyll (Sack & Holbrook 2006; Sack & Scoffoni 2013). When venation density surpassed 6 mm/mm2 the path length for water transport inside the leaf (from vein to stomatal pore) equalled and then became shorter than the internal diffusion path of CO2 (from stomatal pore to chloroplast); add high venation density, high stomatal conductance, and the like and assimilation rates increase, even as atmospheric CO2 declined (McElwain et al. 2015).

The venation density of angiosperm leaves increased again around the Campanian-Maastrichtian boundary ca 70 Ma, and plants whose leaf blades had a venation density of ca 3 mm/mm2 were then a mere 4% of the total (Brodribb & Feild 2009; Feild et al. 2011b; Boyce & Zwieniecki 2012). Only after this second increase could forests assume a more "modern" physiology, and only then did trees have a venation density around 10 mm/mm2 or more like to that of plants in the most productive l.t.r.f. today (Brodribb & Feild 2009; Feild et al. 2011a, b). Very high venation densities, >20 mm mm-2, are found only in some rosids, e.g. eucalypts growing in arid environments (Boyce et al. 2010; de Boer et al. 2016). Furthermore, angiosperms alone have leaves in which veins are the same distance from each other as from the lower surface of the leaf, which allows optimal uniform delivery of water to the stomata (Zwieniecki & Boyce 2014; see above). This spacing is found even in some shade-dwelling angiosperms, perhaps because they maximize photosynthesis in sunflecks (c.f. monilophytes). Of course, increased vein density comes at a cost of increased carbon allocation to the veins, but this is partly offset by vein tapering (McKown et al. 2010; Beerling & Franks 2010).

Venation density increase was accompanied by a decrease in stomatal size and an increase in stomatal density, which together increased stomatal gas exchange capacity. For any given stomatal area, smaller stomata allow more water to be lost, but, importantly, more CO2 to be taken up so counteracting falling atmospheric concentrations; pore depth is shallower in small than in large stomata (Franks & Beerling 2009a; de Boer et al. 2012: Fick's and Stefan's laws are relevant here). Overall, carbon assimilation per unit water loss increased (e.g. Franks & Beerling 2009; Haworth et al. 2011; de Boer et al. 2012; Franks et al. 2012). Dense veinlets allowed an easy flow of water into the mesophyll, their proto- and metaxylem having vessels with simple perforation plates (Feild & Brodribb 2013). Water could be supplied to the leaf by the efficient vascular system even if humidity decreased, whether because of drying climates or the emergence of the tree into the canopy (e.g. de Boer et al. 2012; see also Boyce & Zwieniecki 2012). Small stomata also have a faster relative response time than large stomata as do grass stomata compared with those of other vascular plants (Franks & Farquhar 2006; Franks & Beerling 2009a), however, little is known about the stomatal size/response time connection (Raven 2014). Details of guard cell shape are also important (Hetherington & Woodward 2003), and the great range of relative stomatal pore area (pore area/guard cell + pore area) in vascular plants as a whole, but even when comparing Tradescantia with Triticum (Franks & Farquhar 2006) is also pertinent, although there seems to be little information about this. Drake et al. (2013) compared five species of Banksia and found that leaves with smaller stomata indeed had a higher rate of gas exchange, maximum operating stomatal conductance, and overall high productivity. Trees have small, dense stomata when compared with shrubs and herbs (Beaulieu et al. 2008: n = 101). Small stomata may also allow areas of the epidermis to be freed up for other functions (Franks & Beerling 2009a), although this seems somewhat notional.

Overall, with a three-fold increase in venation density, there is a 178% increase in maximum photosynthetic CO2 uptake (e.g. Brodribb et al. 2007; Brodribb & Feild 2010; Feild et al. 2011a; Roth-Nebelsick et al. 2001: vein architecture; McKown et al. 2010: leaf hydraulics). Furthermore, the water potential of angiosperm leaves can decrease 50% before stomatal closure occurs, so maximum leaf hydraulic conductivity can persist in dry conditions, whereas in ferns, for example, closure occurs earlier (Brodribb & Holbrook 2004; see also Haworth et al. 2011, 2013 for stomatal opening in land plants). Increased transpiration resulting from increased stomatal conductance will promote evaporative leaf cooling (Hetherington & Woodward 2003; Boyce & Lee 2010), perhaps particularly important at times like the Palaeocene-Eocene Thermal Maximum (see below) when temperatures globally were very high.

However, other aspects of leaf venation independent of density like elongated vs compact areoles and reconnecting vs branching veinsare also important variables along an altitudinal transect in South America. These variables have less important hydraulic implications, rather, they are connected with such things as investment in defensive compounds and leaf stiffness and strength (Blonder et al. 2018), and there is an extensive literature in this area going back about 40 years (e.g. Givnish 1979, raferences in Blonder et al. 2018).

Areas with ever-wet tropical humid climates seem to have been rather restricted in the Cretaceous (e.g. Boyce et al. 2010; Boyce & Lee 2010; Feild et al. 2009a; maps at the end of Willis & McElwain 2014). However, the increased transpiration from angiosperm leaves may have helped to drive the spread of l.t.r.f. with reliably high rainfall (e.g. Boyce et al. 2008, 2009, 2010), conditions in the Late Cretaceous becoming more humid (Eriksson et al. 2000a). Thus simulations in which the Amazon rain forest was replaced with non-angiosperm vegetation led to a decrease in the extent of ever-wet rain forest there by about 80% (Boyce & Lee 2010; Lee & Boyce 2010; see also Feild et al. 2011b; Boyce & Leslie 2012; Feild & Brodribb 2013); conversely, with human-induced deforestation, predictions are that rainfall will decrease (Zemp et al. 2017). There was less change in the extent of rain forest in other parts of the tropics in such simulations, but this might have been different under conditions earlier in the Caenozoic; large areas of continental Africa had not yet become elevated, continents were in different positions, etc. (Boyce et al. 2010). Models suggest that atmospheric oxygen concentration, for which estimates vary by a factor of three, also affects estimates of precipitation for the Cretaceous-Cenomanian (Poulsen et al. 2015).

Vascular Anatomy. Vessels in ANA-grade plants are short, not very dense, with scalariform perforations, little difference between the pitting of the end and lateral walls, incomplete break-down of the pit membranes, and intertracheidal pit resistance lower than that of intervessel pits, etc. (Hacke et al. 2007; Sperry et al. 2007). The resistance to water flow of the scalariform perforation plates in such xylem is higher than had been estimated (e.g. Christman & Sperry 2010). Even if individual vessels may be more effective in transmitting water than individual tracheids, when comparing xylem cross-sectional area plants with such vessels may have a hydraulic efficiency little different from that of tracheid-bearing gymnosperms (Sperry et al. 2006; Feild & Holbrook 2001; Hudson et al. 2010).

Indeed, tracheids in Pinales in particular may be short and have end walls, yet their overall hydraulic efficiency is higher than might be expected because of the low resistance to water flow of the margo-torus pits. The central torus can block the pit, so localizing air bubbles developing in the cells during embolism, yet the fibrils in the margo are widely spaced compared with those in angiosperm pits, so allowing water to flow quite readily (Pittermann et al. 2005; Sperry et al. 2007; Hacke et al. 2007; Hudson et al. 2010); there are also lipid surfactants in the xylem that i.a. coat nanobubbles that are forming, so preventing the formation of embolisms (Schenk et al. 2017). Some scrambling or climbing seed ferns like Callisophyton, Lyginopteris and particularly Medullosa had long and wide - from 65-237 μm across (the upper part of the range is Medullosa) - tracheids, and their water conductivity was probably on a par with that of some extant angiosperms with vessels (J. P. Wilson & Knoll 2010). Ferns in general have quite wide and long tracheids that can have surprisingly high rates of water transport (Pittermann et al. 2011).

Thus the acquisition of vessels is no simple key innovation, rather, their evolution is likely to have been a rather protracted process (Feild & Arens 2007; Feild & Wilson 2012). They may have been of functional value initially because heteroxylic wood, i.e. wood with vessels and tracheids, allows the specialization of cells in the xylem for support, storage, etc., the heteroxylly [sic] hypothesis (e.g. Sperry et al. 2007; Hudson et al. 2010; J. P. Wilson & Knoll 2010). Indeed, despite lacking vessels (or almost so), as in most gymnosperms, the wood of Amborella has a small amount of parenchyma (Carlquist & Schneider 2001; Feild et al. 2000b; c.f. Carlquist 2012), and axial parenchyma is notably slight to absent in Austrobaileyales, Laurales, and Chloranthales as well (Herendeen et al. 1999). Plavcová and Jansen (2015) discuss the role of xylem parenchyma (in families other than those just mentioned) in the metabolism of non-structural carbohydrates. The avoidance of cavitation may also have driven the early evolution of vessels and of other aspects of wood anatomy (Sperry et al. 2007; Hacke et al. 2007; Philippe et al. 2008; Brodribb et al. 2012: angiosperm:gymnosperm comparisons).

Vessel conductance increases substantially when the peforation plates become simple and the vessels - concatenated vessel elements - become long (see e.g. Christman & Sperry 2010; Hudson et al. 2010; Feild et al. 2011c). Wood with scalariform perforation plates was common in the Cretaceous, particularly before the Santonian ca 85 Ma (Wheeler & Baas 1993, also 1991; Martínez-Cabrera et al. 2017), but as Wheeler and Baas (1993) noted, there is conflict between features of Cretaceous fossil woods and palaeoclimatic indicators, indicators that are based on our understanding of how the wood of extant plants functions (see also Philippe et al. 2008), so functional interpretation of these early woods is not easy. Eventually wide vessel elements with simple perforation plates, the vessels being well over 10 cm long, became an integral part of an efficient water transport system. However, vessel elements with long, oblique, many-barred scalariform perforation plates occur throughout the angiosperm phylogeny, as in the common ancestor of gentianids and in Paracryphiaceae, for example - see Lens et al. (2016) for vessel/tracheid evolution in asterids as a whole. Indeed, it is not easy to understand vessel evolution. As mentioned, from the fossil record it seems woods with simple perforations have always predominated in the tropics, while in temperate and subtropical area, including the southern hemisphere woods, follow a more Baileyan trend, with solitary vessels, apotracheal parenchyma, and scalariform perforations, etc., initially being common (Wheeler & Baas 2019). Along the same lines, both Sambucus, which has vessels and the xylem of which differs greatly in other respects from that of the vessel-less Viburnum, and Viburnum itself have similar climatic niches, at least going on present-day distributions; perhaps these striking differences reflect past events... (Lens et al. 2016)? In general, hydraulic efficiency is not well understood (e.g. Roddy et al. 2019).

Less is known about the functional/ecological significance of variation in the phloem of angiosperms and gymnosperms, although the literature on phloem loading, the movement of sugars produced during photosynthesis into the sieve tubes/cells, is quite extensive (see also Jensen et al. 2012 and references for flow across the sieve plates of seed plants). Even if differences between sieve tubes (angiosperms) and sieve cells (gymnosperms) may be somewhat over-emphasized - the nuclei in both is non-functional, although they differ in how they get to that condition - they do differ in sieve plate morphology, occlusion mechanisms, and ontogenetic/functional associations with neighbouring cells (e.g. see Behnke 1986; Evert 1990b; Schulz 1992).

Passive phloem loading ("open" minor veins) in which sucrose predominates, there are high sugar concentrations in leaf cells, symplastic transport, and numerous plasmodesmata between the sieve tubes and associated cells, are all correlated with the woody habit. This occurs both in Pinales (probably) and many woody angiosperms (Gamalei 1989; Rennie & Turgeon 2009; cautionary comments in Liesche et al. 2010). A variant of Gamalei's type 1 minor veins in which raffinose family oligosaccharides are commonly involved and loading is via specialized intermediary cells has been distinguished, and this is not connected with plant habit, but perhaps rather with a warmer climate (e.g. Rennie & Turgeon 2009: mechanism described; Davidson et al. 2010). Active phloem loading ("closed" minor veins), where a sieve tube element is loaded by the active transport of solute molecules across the cell membrane, seems to be most common in the predominantly herbaceous gentianids, and a variety of selective advantages for active loading can be suggested (Turgeon 2010b; Fu et al. 2011: see also Icacinales - some cautionary comments). Woody plants with active phloem loading are usually asterids (e.g. Buddleja, Catalpa, Ilex, Syringa), while herbs with passive transport are members of the predominantly woody rosids (e.g. Rosaceae, Paeonia, Lythrum). However, Saxifraga, herbaceous, and the woody Cercis and Styrax, none asterids, have active loading (Rennie & Turgeon 2009; Fu et al. 2011). For sugars, etc., other than sucrose in phloem loading, see Rennie and Turgeon (2009), for phloem loading and intermediary cells, see Davidson et al. (2011); see also Batashev et al. (2013) for minor-vein phloem anatomy and physiology.

Other variation in plumbing includes that in the water supply to the flower. The large flowers of at least some magnoliids may obtain their water through the xylem, whereas smaller flowers, as in the core eudicots, may be hydrated primarily via the phloem (Feild et al. 2009a, b, but sampling).

5E. Wood and Litter Decay. Another element of the ecological impact of angiosperms is mediated through litter and wood breakdown, as well as their loss by burning (Cornwell et al. 2009). About 30% of the organic carbon in the biosphere is currently locked up in lignin (Boerjan et al. 2003). Factors like leaf mass per area (MA, the inverse of specific leaf area [SLA], the relation of leaf area to dry mass, cm2 g-1) and primary and secondary venation type have been linked with features like the rate of photosynthesis, plant growth, litter decay, nitrogen content, and nutrient cycling. However, there is much within-community variation in such features in angiosperms and any phylogenetic signal in such correlations is not well understood (e.g. Cornwell et al. 2008; Wieder et al. 2009; Walls 2011). Moreover, it should not be forgotten that most carbon is to be found in the soil, particularly in the subsoil (Schmidt et al. 2011).

Both low MA and high amounts of nutrients in litter are quite common in angiosperms and are implicated in speedy litter breakdown. Thus angiosperm floras in the Cretaceous (Potomac, 110-105 Ma: Royer et al. 2007) and Eocene (49-47 Ma: Royer et al. 2010) had a low leaf MA, under 100g/m2; the three gymnosperms examined in the former flora had a mean of 291 g/m2. Even contemporary tropical non-riparian lowland rain forest may have only a moderate MA, e.g. ca 198 g/m2, as on Barro Colorado Island (Royer et al. 2010), and extant gymnosperms, like the extinct gymnosperms just mentioned, have a higher MA than do angiosperms (Berendse & Scheffer 2009 and references). However, low MA per se is unlikely to be the only factor speeding breakdown of angiosperm remains.

Decay is affected by the composition of plant parts. At around 20%, the lignin content of angiosperms is about half that of lycopsids, and the bark wood ratio has shifted from around 8-20:1 to ca 1:4 (Robinson 1990). Denser gymnosperm woods have more lignin and less nitrogen than angiosperms, and angiosperm woods, although denser than gymnosperm woods, decay faster (Robinson 1990; Weedon et al. 2009). In general, both lignin and polysaccharide content are negatively correlated with the rate of litter breakdown (Cornwell et al. 2008; Martínz et al. 2005: decomposition of lignocellulosic compounds). The syringyl-rich lignins that characterise many angiosperms are more easily decomposed by fungi than the guaiacyl-rich lignins of other seed plants (Ziegler et al. 1985), although overall in angiosperms the rate of litter decomposition, although very variable, seems to increase at the core eudicot node (LeRoy et al. 2019). Brown rot fungi, which do not degrade lignin, are more common in conifer forests than lignin-decaying white rot fungi which are common in angiosperm-dominated forests (Boddy & Watkinson 1995), although it should be noted that the distinction between the two is somewhat artificial (Riley et al. 2014; Floudas et al. 2015). Mean annual precipitation and temperature are positively correlated with litter and wood turnover and so with the release of the nutrients they contain (Yin 1999; Weedon et al. 2009; Wieder et al. 2009). A higher rate of lignin decomposition is correlated with nitrogen and phosphorus release during the process (Wardle et al. 2002). Overall patterns of wood decomposition vary in detail, and differences in decay rates depends on the local decay organisms and tree species, and even on the age of the tree (Weedon et al. 2009). The chemistry of woods is complex (e.g. Kögel-Knaber 2002)!

To summarise: Litter from extant ferns and fern allies and bryophytes is slow to decompose compared to that of gymnosperms and especially that of most angiosperms (Cornwell et al. 2008; Lang et al. 2009), although monocot litter can be be very slow to decompose (Wardle et al. 2002: New Zealand, no Poaceae). Note, however, that data in LeRoy et al. (2019) show no difference between monocots, gymnosperms and basal eudicots in in-stream decomposition rate. However, there is variation within gymnosperms. Thus litter of the arbuscular mycorrhizal juniper has a lower C:N ratio than that of the ectomycorrhizal Pinus edulis and decomposes faster (Gehring et al. 2017b and references). The litter of deciduous angiosperm trees decomposes faster than that of evergreens, angiosperm wood faster than gymnosperm wood, and the litter of angiosperm forbs in particular decomposes faster than that of any other group of land plants (Cornwell et al. 2008; Weedon et al. 2009). However, not only are the properties of wood important when thinking about decay, but manganese ions (Mn+++) may be an important element facilitating the oxidation and breakdown of litter (Keiluweit et al. 2015).

Angiosperm leaves, litter and wood all have more nitrogen and phosphorus (on a %age basis) than do those of gymnosperms (Cornwell et al. 2008; Weedon et al. 2009). Fast decomposition of core eudicot angiosperm litter, particularly associated with the deciduous habit (Knoll & James 1987), speeds up nutrient cycling and plant growth (Cornwell et al. 2008; Berendse & Scheffer 2009). The high photosynthetic rates of most angiosperms allow high growth rates and the nutrients they need are released by the fast decay of their litter; eudicot angiosperms in particular may utilize any flushes of nutrients produced by litter and wood breakdown, they scavenge nutrients effectively (Berendse & Scheffer 2009). The disturbed habitats of early angiosperms are likely to have had elevated levels of nutrients (Berendse & Scheffer 2009).

Graminoid litter, i.e. that of Poaceae and Cyperaceae, decomposes more slowly than that of forbs (Pérez-Harguindeguy et al. 2000; Cornelissen et al. 2001; Cornwell et al. 2008; Lang et al. 2011; LeRoy et al. 2019). Graminoid lignin, with its appreciable component of p-hydroxyphenyl units, is somewhat different in composition from that of other plants, and it is low in nitrogen which is removed before the leaf dies (e.g. Cornelissen et al. 2001; Wedin 1995). Tissues of C4 grasses have a particularly low N content which negatively affects their decomposability (Forrestal et al. 2014; C3 grasses not included in LeRoy et al. 2019). Roots of Poaceae also decompose more slowly than those of other plants (Birouste et al. 2012: sample small, Mediterranean species). There is also a negative correlation between between silicon concentration - especially high in annual grasses - and rate of leaf decomposition (Cook & Leishman 2011b).

G. Liu et al. (2014) pointed out that many ANA-grade angiosperms and magnoliids have a high SLA and their leaves decompose relatively slowly, they are "slow return" leaves, compared with those of other large angiosperms (see also Cornwell et al. 2014). Piperales and Nymphaeales, with more quickly decomposing leaves, are smaller plants (Liu et al. 2014). The litter of ectomycorrhizal (ECM) trees also tends to decompose more slowly than that of AM plants, although not as slowly as that of Sphagnum (Pérez-Harguindeguy et al. 2000; Cornelissen et al. 2001; Cornwell et al. 2008; Lang et al. 2011; Midgley et al. 2015; Gehring et al. 2017b and references); for the remarkable properties of rather slowly-decomposing Sphagnum peat, see e.g. Painter (1991) and Hájek et al. (2011). With ECM trees in particular the rate of decay of the ECM fungal mycelium, whether separate hyphae or rhizomorphs, also has to be taken into account, since the melanin in their their hyphae, perhaps involved in water stress tolerance (Fernandez & Koide 2013), is notably resistant to decay; some dark septate endophytes, also with melanin in their walls, can be either ECM or ERM on occasion (e.g. Butler & Day 1998; Lukesová et al. 2015). Organic matter in the subsoil below 30cm, a major component of global carbon, has very long turnover times, 1,000-10,000 years or more (Schmidt et al. 2011), although it is unclear whether or not there is a link to mycorrhizal type here.

It is not only the amount and rate of decay that matters, but its seasonality. In evergreen plants, whether angiosperm or gymnosperm, nutrient cycling is gradual, nutrients being released throughout the year and tending to be taken up by the plants again. In deciduous species, however, nutrients tend to become available in flushes, and some are lost to the ecosystem, and this will increase weathering (Knoll & James 1987). Fire also removes dead organic matter and affects the availability of plant nutrients (see elsewhere).

Limited data suggest that the diameter of first order roots has decreased in angiosperms, root length per unit biomass, SRL, increasing, so allowing for more efficient exploration of soil space to satisfy the increased water demands of the plant as carbon dioxide concentration decreased and also enabling more efficient nutrient scavenging (Comas et al. 2012; W. Chen et al. 2013: diameter decrease - and branching ratio increase - phylogenetically correlated, breakpoint ca 64 Ma). Many magnoliids and the ANA grade taxa have rather thicker roots, thus magnoliids are prominent in the small group of tropical angiosperms with relatively broad first-order roots compared with the narrower roots of gymnosperms and temperate angiosperms (Chen et al. 2013), and there are links between these root attributes, mycorrhizal status and nutrient foraging status. Species with thin roots may forage for nutrients directly while those with thicker roots forage via their mycorrhizal associates (B. Liu et al. 2015; Eissenstadt et al. 2015). However, Maherali (2014) found no particular correlation between root diameter and response to colonization by AM fungi. Work by W. Chen et al. (2016) suggests the complexity of possible interactions. In a small sample of temperate trees growing in an environment with patchy distribution of nutrients, AM trees with fine roots exploited these nutrient-rich patches by producing more roots, while ECM trees expolited them via mycelial growth, interestingly, the three members of Pinaceae (ECM) and two magnoliids (VAM) in the study had the thickest rootlets and none exploited the nutrient patches with notable precision (Chen et al. 2016). In general, fungi with long-distance exploration-type rhizomorphs are commoner in places where there is low nitrogen availability (Hobbie & Agerer 2010).

At ca 40µm across the so-called hair roots of Ericaceae with ericoid mycorrhizae (ERM) are perhaps the narrowest fine roots, but it is unclear if there are systematic differences in fine root diameter between ERM/ECM and AM plants, either within Pentapetalae or in general (but see preliminary studies in Comas et al. 2012; W. Chen et al. 2013; references in Raven & Edwards 2001). Within monocots the very thick and often little-branched roots of many epiphytic orchids, and particularly those of some Pandanaceae and palms, suggest that different relationships may hold there. Roots in mycoheterotrophic plants also tend to be very thick, to 15 mm across in Voyria rosea (Gentianaceae), and they, too, are often little branched (Imhof 2010; Imhof et al. 2013; Bolin et al. 2016 and references). This whole area would repay more detailed study.

6. Angiosperm History III: Caenozoic Diversification.

Introduction. For a summary of global climate during the Caenozoic, see Zachos et al. (2001). Atmospheric CO2 concentrations briefly spiked at a high of over 1,200 p.p.m. at the Palaeocene/Eocene Thermal Maximum, but they continued to fall throught the Caenozoic (Arakaki et al. 2011). During the recent glaciations CO2 concentrations dropped to 180-190 p.p.m., as low as any time during the whole period of land plant evolution (e.g. Zachos et al. 2008; Gerhart & Ward 2010; Boyce et al. 2010; McKenzie et al. 2016). Similarly, global temperatures were high at the beginning of the Caenozoic, but have since declined, although with pronounced if sometimes short-term increases. Continental drift was active, India and the Deccan plate colliding with Asia, Australia approaching Southeast Asia, and the Atlantic Ocean widening (e.g. Morley 2003), however, the importance of India, separating from Madagascar 96-84 Ma and more or less completely isolated for some 30-50 M years, as a source of distinctive (ex-Gondwanan) taxa when it docked with Asia is unclear (Datta-Roy & Karanth 2008; Jagoutz et al. 2015; see also below). Land connections around the southern end of the world were broken, but fossils from the early Caenozoic in Patagonia are changing our ideas about the distributions of vascular plants in the Southern Hemisphere then. Major mountain ranges such as the Andes and the Qinghai-Tibet plateau were elevated towards the middle-later part of this period, the latter reaching 4,000 m altitude by 35 Ma at the latest, and with its elevation came the development of monsoonal climate in Southeast Asia (Spicer et al. 2003; Favre et al. 2015).

6A. The K/P Boundary Event. At the end of the Cretaceous ca 65.5 Ma a large bolide hit the northern Yucatan region of Mexico, the Chicxulub impact, and on the other side world there were massive volcanic outpourings that now form the Deccan Traps in India a little before and after this time (Schoene et al. 2015; see also Keller 2014; Hull et al. 2020). There is some debate as to exactly when these eruptions happened and how much CO2 was produced (Hull et al. 2020). The first of these eruptions preceded the bolide impact, and Meyer et al. (2019) found a marine temperature spike along with a spike in mercury, presumably from the eruptions, in marine carbonates deposited immediately preceding the impact, although there seems to have been little or no extinction (Hull et al. 2020). Indeed, global climate was unstable in the last ca 1 Ma of the Cretaceous, perhaps because of the production of CO2 and other gases by the Deccan Trap eruptions (Renne et al. 2013), which began a little over 66 Ma. Rock weathering increased, and this increased rate continued through the later Caenozoic (Pälike et al. 2012). The later stages of the eruptions may even be geophysically connected with the bolide impact (Renne et al. 2015). This is no place to discuss the relative importance of these two events, although S. V. Petersen et al. (2016) noted two major temperature spikes in Antarctica, while Schoene et al. (2019) recently affirmed and Sprain et al. (2019) and Hull et al. (2020) questioned the role of the Deccan Traps in major climate change at this time. In any event, very large amounts of sulphate aerosols, dust, etc. were injected into the atmosphere, contributing to the world-wide changes in the biota that were evident at the Cretaceous/Palaeogene (K/P) boundary (K/Pg; Cretaceous/Tertiary, K/T or C/T boundary in other literature). There was an associated 5oC temperature increase that lasted for around 100,000 years (MacLeod et al. 2018: Tunisia).

Almost separate from the question of what actually happened at the K/P boundary, there is the question of what happened to the earth's biota. An estimated 75% of species may have become extinct globally (Vajda & Bercovici 2014), although changes in some groups, including dinosaurs, may have begun a little earlier in the Late Cretaceous (references in Schoene et al. 2015). However, a major question is, how seriously did these events at the K/P boundary actually affect the land flora and fauna?

Land Plants. Maybe as many as 80% of species of plants were lost in some places in North America at the K/P boundary (Upchurch & Wolf 1987), although other estimates are around 25-33% (Nichols & Johnson 2008; c.f. Salas-Leiva et al. 2013), extinction being described as being "moderate" (Knoll 1984). Wilf and Johnson (2004: p. 347) suggested that "sudden ecosystem collapse" occurred at least locally, even some common plants not surviving the K/P boundary, and they estimate that there was 30-57% extinction of the flora (data are from pollen) in southwest North Dakota (see also Vanneste et al. 2014a). Similarly, in an unusually complete sequence from Corral Bluffs, Colorado, 46% of Cretaceous leaf morphospecies did not reappear after the impact as did a number of Cretaceous palynomorphs, including Aquilapollenites spp. (e.g. Fuentes et al. 2019) - but there was also a ca 4.6oC cooling immediately prior to the impact during the last 100,000 years of the Cretaceous (immediately followed by a warming), and a decrease in plant richness was also evident then (Lyson et al. 2019). The familial composition of Early Caenozoic forests in North America differs from that of their Late Cretaceous counterparts (e.g. K. R. Johnson 2002; Wilf & Johnson 2004), interestingly, leaf mass/unit area in the surviving plants decreased, at least in Colorado, suggesting faster growth strategies (Lyson et al. 2019). Insect-pollinated and/or evergreen taxa of seed plants suffered more than wind-pollinated and/or deciduous taxa (Collinson 1990; McElwain & Punyasena 2007).

Looking more globally, it is unclear just how severe the effects of the bolide impact/eruptions were on the flora, certainly, moving away from the Yucatan the effects sometimes seem less marked. In New Zealand the iridium anomaly associated with the bolide impact was followed by a thin layer high in fungal remains, the fungi probably growing on plant material recently killed by K/P boundary events (Vajda & McLoughlin 2004), and changes there seem to be quite pronounced, Araucariaceae in particular declining substantially (Pole 2008; Cantrill & Poole 2012). Indeed, there is no evidence for fungal extinctions at the K/P boundary (Varga et al. 2019), fungal spores and hyphae peaking in the fossil record at this time (Vajda & McLouughlin 2004). In both hemispheres there were fern spikes (and, in the Netherlands, a bryophyte peak) after the impact/eruptions, and immediately following any fungus increase (Saito et al. 1986; Wing 2004; Vajda & McLoughlin 2004, 2007; Nichols & Johnson 2008: esp. Vajda & Bercovici 2014: Fig. 6; Lyson et al. 2019), however, post-impact spikes in ferns reported from New Mexico (Berry 2019) have been questioned - see below. Evidence from Australia is unclear (e.g. Macphail et al. 1994; Hill & Brodribb 2006). In Colombia, there were changes in ecological structure but not extinctions (De la Parra et al. 2007). In general, macrofossils tend to show higher extinction rates than sporomorphs, the macrofossil record being more and the microfossil record less local (e.g. Mander et al. 2010). No major plant group is known to have disappeared at the end of the Cretaceous (Nichols & Johnson 2008). By and large the main pollen genera persisted across the K/P boundary in interior North America, even if species did not (Tschudy & Tschudy 1986; Lupia 1999: no change in general morphological disparity, reduction in pollen species numbers; c.f. Lyson et al. 2019), and there was no major change in the Wodehouseia spinata pollen assemblage in North Dakota during the K/P change (Nichols & Johnson 2008; Vajda & Bercovici 2014). Indeed, the macroflora there seems to have been exceptionally diverse in the million years preceding the K/P boundary, hence perhaps accentuating the subsequent decrease (Vajda & Bercovici 2014: Fig. 4). In a graph of stem-group ages for monocot families [sic: clade origination in general might provide a better estimate?], Givnish et al. (2018b) notice a bimodal distribution, but with the trough just prior to the K/P boundary. More generally, C. Zhang et al. (2020) and others have found that many extant angiosperm families first appeared in the Cretaceous. Green and Hickey (2005) detected no significant changes when various aspects of leaf architecture were plotted across the K/P boundary in North America, so any effect in ecosystem structure was likely to have been minor and/or short-lived (this study not unreasonably was based on a connnection between ecosystem properties and leaf morphology).

Nothing much at all seems to have happened in the high Arctic (Spicer & Herman 2010). In Patagonia in particular and some other places in the Southern Hemisphere, changes at the K/P boundary were rather muted (Wolfe 1987; Iglesias et al 2007; Wilf et al. 2013, 2017a), and in both Patagonia and Antarctica the K/P boundary is visible botanically mainly as a transient fern spike, (Cantrill & Poole 2005a, b) or even less (Bowman et al. 2014), and there were no major extinctions (Cantrill & Poole 2014). The extinct conifer group Cheirolepidaceae increased and Podocarpaceae decreased in Patagonia in the early Palaeocene, but with little extinction, as is true in New Zealand, too (Barreda et al. 2012b). Floras across the end-Cretaceous from northwestern Chubut and vicinity, Argentina, showed "a major floristic shift and decline in richness at the beginning of the Paleocene, but not a great extinction event" (Barreda et al. 2012b: p. 6) and a "relatively muted K–Pg floral extinction" (Comer et al. 2015: p. 544; Iglesias et al. 2007; see also Apesteguía, S. et al. 2014: rhynchocephalians). There seem not to have been widespread fires (Belcher 2010). Plant productivity, at least, in the Deccan Traps shows little change, although the pollen record seems to be poorly known (Cripps et al. 2005).

Things are also confusing when looking at the records of individual taxa. Groups like Annonaceae, Arecaceae, and Araceae seem to have constant diversification rates across the K/P boundary (Couvreur et al. 2011a and references); low diversity forests with ferns and palms would suggest ecological disequilibrium to Lyson et al. (2019). Menispermaceae, today often lianas of the l.t.r.f., showed a burst of diversification close to the K/P boundary (W. Wang et al. 2012, c.f. dates in W. Wang et al. 2016a; Eiserhardt et al. 2017). Several clades of Cretaceous heterosporous water ferns did not survive to the Cainozoic, although others persisted to the present (Collinson et al. 2013). There may have been decreases in the diversification rates of ferns and low diversification and high extinction of gymnosperms, but diversification rates at the level of genus increased and overall net diversity for vascular plants was high (Silvestro et al. 2015: mostly leaf fossils; see also Niklas et al. 1983; Knoll 1984; Wing 2004). It has been suggested that angiosperm-dominated herbaceous floras, of which the largely north temperate Ranunculaceae are an example, may have diversified in the Cretaceous and were relatively unaffected by events at the K/P boundary (W. Wang et al. 2016). Seed ferns survived until well into the Palaeocene in Tasmania (McLoughlin et al. 2008), cheirolepidaceous conifers into the Palaeocene in Patagonia (Barreda et al. 2012b), and Bennettitales perhaps into the Oligocene in Australia (McLoughlin et al. 2011: a ghost lineage of ca 65 My). Some animal groups in the same area show similar patterns of persistence (Iglesias et al. 2007; Barreda et al. 2012b).

Finally, the interpretation of the fossil record continues to change. Thus a reevaluation of the post-bolide record in the Raton Basin, New Mexico, suggests that the fern spike there, probably caused by schizaeaceous ferns, was accompanied by a leaf fossil flora dominated by Lauraceae, indeed, fossils initially identified as Ficus, but now reidentified as Lauraceae, seem to have been little affected by events at the K/P boundary (Berry 2019). Moreover, looking at historical records of vegetation recovery following eruptions at Krakatau, early colonisation by Lauraceae occurred there, too, and it even preceeded an increase of schizaeaceous ferns and hence also any fern spore spike that they might produce - and of course Lauraceae pollen with its thin exine is in general very poorly preserved in the fossil record (Berry 2019). However, DePalma et al. (2019, but see Barras 2019) suggested that deposits from North Dakota that contain fish with ejecta spherules in their gills and dinosaurs represented the effects of a 10 m or more wave generated by seismic waves arriving from the Chicxulub impact perhaps only 13 minutes after the impact...

Animals. Herbivorous insects were decimated, and most immediate post-bolide floras show little diversity of herbivory types, full recovery taking some 9 Ma (Donovan et al. 2016 and references), localities like Mexican Hat, in Montana, with high post-bolide herbivory, definitely being exceptions (Wilf et al. 2006; Donovan et al. 2014). There was a change in the composition of these herbivores, diet-specific insects being seriously affected (Labandeira et al. 2002a, b; Wilf 2008: surveys of leaf damage types). Diversity of insect damage types in Colombia was higher than in North America and there was less of a decrease in damage types across the boundary, but as in Patagonia (Donovan et al. 2015, 2016; see also Jud et al. 2017) and in North America (Wilf et al. 2006; Donovan et al. 2014) there were no obvious boundary-crossing types of leaf mines (De la Parra et al. 2007). However, in Argentina while there was severe extinction of insect herbivores there was also a rapid subsequent recovery, with a diversity of insect mines apparent in immediate post-impact deposits and ± complete recovery by 4 Ma (Donovan et al. 2016). Diversification rates in the large clade of leaf-roller moths, Tortricidae, shows at most a blip at the K/P boundary - or if there were blips, they were just before the boundary (Fagua et al. 2017). Polyphagous beetles (not including whirlygigs or carabids, but including 90% of all beetles) show low extinction rates, and beetles in general seem largely unaffected by the K/P boundary (D. Smith & Marcot 2015), however, patterns of herbivory in North America suggested that "terrestrial ecosystems remained depauperate and ecologically unstable for 10 million years" (Labandeira & Currano 2013: p. 295).

Some animal groups - not only herbivorous dinosaurs (e.g. Samant & Mohabey 2014; Sakamoto et al. 2016) - did suffer more or less severely. Thus there are estimates of about 60% loss of butterfly diversity at the K/P boundary, with a decrease in diversification in Nymphalidae (Wahlberg et al. 2009) but not Riodinidae (Espeland et al. 2015), and a mass extinction of birds, lizards and snakes in western North America, and perhaps elsewhere, with North American snakes and lizards taking perhaps 10 Ma to recover their Late Cretaceous diversity (Longrich et al. 2011, 2012). In Corral Bluffs, Colorado, extinction of mammals, especialy those with larger bodies, was considerable, but a rebound in both species numbers and sizes was evident by as little as 300,000 years post-impact, and both aspects of the recovery continued unabated (Lyson et al. 2019: there is no obvious connection between the evolution of large-fruited Juglandaceae and this rebound - see below). Rather surprisingly, given text-book accounts, placental mammal diversification - but perhaps not metatherian (inc. marsupials) diversification - may have been little affected (Bininda-Emonds et al. 2007; Meredith et al. 2011; G. P. Wilson et al. 2012; Grossnickle & Newham 2016; esp. L. Liu et al. 2017; c.f. O'Leary et al. 2013: dates probably underestimates, see dos Reis et al. 2014), although eutherian mammals may have had much increased evolutionary rates just after the K/P boundary (Halliday et al. 2016; see also Wilson 2014). The diversity of multituberculate mammals was at a peak across the K/P boundary and also seems unaffected by events then (Wilson et al. 2012), while the biota like frogs and turtles entombed in the Indian Intertrappean floras show little evidence of major changes (Cripps et al. 2005), overall, there was little evidence of animal extinctions there (Spicer & Collinson 2014). With 1,000 species, the predominantly leaf-rolling Tortricidae moths is the 10th most diversified clade of phytophagous insects; crown-group diversification here began (118-)97(-75.9) Ma, there was an increase in diversification rate around 72 Ma and a fairly gradual decline since (Fagua et al. 2017), but the K/P boundary might as well not exist for these moths. Janzen (1995) suggested that seed and detritus eaters whose food did not immediately depend on products of active photosynthesis would be less affected than animals with other dietary preferences and might best survive a nuclear winter (see also Berry 2019).

As with some earlier extinction events, there are a number of reports of severe changes in marine habitats (Knope et al. 2020). For marine extinctions and flood basalts, see Clapham and Renne (2019), although there is not much discussion about this particular event there. Marine productivity at the crater itself was high within 30,000 years of the impact, although some more regional effects were still apparent after 300,000 years (Lowery 2018). Henehan et al. (2019) noted rapid oceanic acidification after the impact, and the effects of this event were apparent for about a million years. Ammonites became extinct and plankton with calcareous skeletons were greatly reduced in numbers (Schulte et al. 2010; Ohno et al. 2014), and the relatively few survivors dominated the seas for many thousands of years (Schueth et al. 2015). And even if there were quite low levels of extinction of fish, after the bolide impact ray-finned fish became far more common than sharks and their relatives, a major and permanent change in marine ecosystems (Sibert & Norris 2015). Molluscs other than ammonites were severely affected in the Antarctic (Witts et al. 2016 and references; S. V. Petersen et al. 2016) and marine animals in general were severely impacted (Muscente et al. 2018).

Recovery of algal primary productivity in marine ecosystems may have taken as little as the order of century or perhaps even less (Sepúlveda et al. 2009: Denmark; Aberhan & Kiessling 2015). However, other aspects of marine recovery took considerably longer (literature in Wilf & Johnson 2004), with changes in marine assemblages lasting for a few million years (Aberhan & Kiessling 2015). Interestingly, as with other aspects of recovery, there can be considerable local differences, as when comparing the immediate impact area, where diversity recovered quickly, and the surrounding areas, where diversity recovered more slowly (Fuentes et al. 2019).

Discussion. Not surprisingly, estimates of the time that vegetation took to fully recover range from only a few thousand years in New Zealand (Vajda & McLoughlin 2007) to over a million years years (McElwain & Punyasena 2007) or considerably more. Indeed, it has been estimated that more or less complete recovery in southwest North America was slow, taking some 9 Ma (e.g. Labandeira et al. 2002; Donovan et al. 2014, 2016). In North America initial recolonization may have been by swamp- and mire-loving plants, which apparently survived the impact better (K. R. Johnson 2002; Labandeira et al. 2002b), mire vegetation being least affected by the impact (Nichols & Johnson 2008). In North Dakota species growing along river channels were more adversely affected than those in the flood plain, and fast-growth ecological strategies were favoured: Leaves were thin and often deciduous, and there was a small increase in venation density from ca 3.5 mm/mm2 to ca 4.6 mm/mm2, i.e. ca 1.8 mm/mm2 (Blonder et al. 2014), only a small change given the average difference between nonangiosperms and eudicots, ca 8 mm/mm2, and any long-term effects of these changes are unclear (see also Green & Hickey 2005). A fast-growth strategy perhaps suggested by these thinner leaves might allow the plants to deal better with changing light regimes (see the "impact winter": Ohno et al. 2014 and refs.) and heterogeneities in resource availability (Blonder et al. 2014). In higher-latitude North America predominantly broad-leaved evergreen mesothermal forest was replaced by comparable broad-leaved forest (Wolfe 1987). Tropical rain forest was growing in Colorado ca 1.4 Ma after the K/P boundary (K. R. Johnson & Ellis 2002), although this seems to be atypical (Donovan et al. 2016 and references) and immigration of taxa from less affected areas may have been important here (Renne et al. 2013).

Overall, the Patagonian flora ca 4 Ma after the bolide was notably diverse compared to its northern counterparts, suggesting that this whole area was something of a biodiversity refugium because of its distance from the impact, or that it had higher pre-impact diversity, and/or there was higher immigration (but from where?) or post-impact speciation rates (Iglesias et al. 2007; Barreda et al. 2012b; Donovan et al. 2016). Patterns of leaf herbivory before and after the K/P boundary did differ, so questioning the idea of refugia (Donovan et al. 2016). Wappler et al. (2009) found that the diversity of herbivore damage in the rich Palaeocene flora in Menat, France, was high, again suggesting less effect of the K/P events there than in North America - this flora is dated to 61-60 Ma, ca 5 Ma post-impact.

Clearly, what went on in the southern half of North America should not be extrapolated globally (e.g. Wappler et al. 2009), and as already mentioned the effects of the end-Cretaceous bolide impact/Deccan Traps eruptions on angiosperms and even some animal groups sometimes seem quite muted. However, land plants and animals might almost be expected to differ in how they were affected by such catastrophes (Traverse 1988: "plant evolution dances to a different beat"; Vajda & Bercovici 2014; Cascales-Miñana et al. 2016a; c.f. Blomenkemper et al. 2018 and Nowak et al. 2019: questions about the severity of the Permian-Triassic extinction, Gastaloo 2019 for comment), and bolide impacts and serial volcanic eruptions may have rather different effects on the biosphere. Plants may have been unable to grow for a mere one or two years, even in places with much devastation (Spicer & Collinson 2014), and many plants can tolerate extensive damage (Knoll 1984). Plant propagules in the soil are likely to have survived transient (months, even a few years) atmospheric or other changes at the K/P boundary, while metazoan animals lack resting stages in their life cycles (Cascales-Miñana & Cleal 2013). Overall seed plants show a fair amount of ecological resilience (Nimmo et al. 2015 for resistance/resilience). As Green and Hickey (2005: pp. 998-999) observed, "If all the forests in North America were burned over in a single summer, that would clearly count as a dramatic ecological effect, but it seems intuitively likely that such an effect would have few or no effects that lasted longer than the time taken for the forests to regrow"; 'mass death' would seem a more appropriate descriptor rather than 'mass extinction'. Indeed, no major adverse effects on land plants around the K/P boundary were detected in some recent analyses (Cascales-Miñana & Cleal 2013; Magallón et al. 2015: level of family origination was high then), although such results in part depend on the units of analysis. Wing (2004) also suggested that extinctions in general tended to be ecologically selective, but major clades of plants had diverse life histories and so would be unlikely to go extinct (a shotgun blast) - unlike animals (a chain saw). Nowak et al. (2019 and references) note that plants and animals might have obligate associations, and, if so, then mutual vulnerability to extinction events might result. However, Gradstein and Kerp (2012: p. 235) note "there is no evidence for any worldwide mass extinction event in plants during geological history", which seems a reasonable summary of a complex issue.

Fawcett et al. (2009; see also Vanneste et al. 2014a, b; Landis et al. 2018) dated a series of genome duplications within angiosperms to about 70-57 Ma, around the time of the Deccan Traps/bolide impact, suggesting that allopolyploids forming then, or which had quite recently been produced, were at a selective advantage because of their hybrid vigour, having extra genes/alleles available for selection given the changing environmen, and also being being more tolerant of environmental stress (see also Z. Li et al. 2016; G.-Q. Zhang et al. 2017: age of Orchidaceae ca 81 My). As Lohaus and Van de Peer (2016: p. 64) observed, "the correlations of polyploidization with both plant survival at the K/P boundary and plant invasiveness in general are related, as the plant survivors of the K/Pg mass extinction event turned into plant invaders and recolonizers of the post-cataclysmic, low-plant diversity evironment." (see also Van de Peer et al. 2017; J. Wang et al. 2019b for allopolyploidy and duplication). Disaster here would seem to have been averted by genome duplication, but it is clear that the first thing to do is to understand the extent of the disaster.

6B. Flowering Plants. There were changes in angiosperm ecology and diversity in the early Caenozoic. For angiosperm clades that crossed the K/P boundary, the average seed mass, initially rather low, increased markedly (e.g. Tiffney 1986b; Eriksson et al. 2000a; Collinson & van Bergen 2004; Sims 2010). This trend can be seen within Juglandaceae, which have many winged disseminules (Eriksson et al. 2000a; Friis et al. 2011), as well as in Fagaceae, which largely lack such disseminules (Tiffney 1986a). X.-G. Xiang et al. (2014) thought that relatively open habitats after the K/P boundary may have favoured the diversification of fagalean clades with winged disseminules, although such habitats are unlikely to have persisted. Around the end of the Cretaceous flower size must have increased, flowers even in the later Cretaceous usually being only a few millimetres in size (Friis et al. 2011).

The increase in seed size that began at the end of the Cretaceous and reached a maximum in the early Eocene may be linked to a change in forest type, with closed, more humid forests made up of tall trees becoming more common (e.g. Eriksson et al. 2000a, b; Tiffney 2004; Mack 2000; Moles et al. 2005a, b; Eriksson 2008, 2016; Dilcher 2010). H. V. Graham et al. (2019) found evidence of closed-canopy forests in N.W. South America in the Palaeocene 50-58 Ma (there was no indication of such forests in the Late Cretaceous). Large seeds are common in plants that at least initially grow in shaded habitats, providing reserves for the early growth of the seedling, although they may also be favoured by dry conditions, soils with low mineral nutrients, etc. (Leishman et al. 2000; Bolmgren & Eriksson 2005). Within individual forest communities there is great variation in seed size, in part connected with the successional status of the species, early successional species tending to have smaller seeds (e.g. Westoby et al. 1996; Eriksson et al. 2000b). In Europe rodents shifted from those that ate soft fruit (Middle Eocene) to those that ate hard seeds (Stratiotes: Late Eocene), perhaps connected to the deteriorating climate (Collinson & Hooker 2000), and this shift has been associated with the evolution of large nuts then (see also Eriksson 2008, 2016). Interestingly, the emergent trees, epiphytes and lianas of today's tropical rain forest canopy are quite commonly wind dispersed (Herrera et al. 2014b for references), and the evolution of minute dust seeds, as in epiphytic Bromeliaceae and Orchidaceae, and the development of a closed canopy has been linked (Eriksson & Kainulainen 2011). Recalcitrant seeds, which have relatively high water content and are short-lived, predominate in phanerophytic taxa that grow in humid rainforests today, and such seeds may tend to be derived (Subbiah et al. 2019: but c.f. comparisons between species/seed types); see also Will is et al. (2014b).

Gentianids, many of which are herbaceous or shrubby, tend to have small seeds (Eriksson & Kainulainen 2011), and they diversified greatly in the Cainozoic. Seed volume decreases somewhat from the end-Eocene onwards (e.g. Tiffney 1984; Eriksson et al. 2000a; Friis et al. 2011). Seed mass of extant angiosperms currently drops quite abruptly (seven-fold) at the edge of the tropics (Moles et al. 2007: sampling in the tropics not very good). The reasons for this are unclear, but wind dispersal of smaller seeds in the open habitats that are more common outside the tropics may be involved (Lorts et al. 2008, but see above). The current prevalance of ectomycorrhizal (ECM) forests in regions outside the tropics and the often rather acid, humus-rich, and nutrient-poor soils that they favour may also affect seed size, although temperate mast-fruiting ECM Fagaceae have notably large, often animal-dispersed seeds while ECM Pinaceae and in particular pioneering Salix and Betulaceae-Betuloideae have small, wind-dispersed seeds.

In general angiosperm diversity in the tropics and warm temperate areas seems to have been rather low during the Palaeocene (Wilf 2008). However, a middle Palaeocene (ca 61 My) flora in France was diverse and also supported a diverse assemblage of herbivores, as in a number of sites far distant from the point of impact of the bolide (Wappler 2009 and references). There are no particular changes in diversification rates at around this time (Silvestro et al 2015; Magallón et al. 2015). By around 64.5 Ma the Castle Rock flora in Colorado is described as "an excellent example of early modern tropical rain forest in North America" (Burnham & Johnson 2004: p. 1607). A Late Palaeocene flora from Colombia ca 59 Ma had a familial composition similar to that of current neotropical rain forest, including Arecaceae, Araceae, Fabaceae, Malvaceae, Menispermaceae, Lauraceae and Zingiberales, even if overall both plant (esp. beta diversity) and herbivore diversity were rather low (Jaramillo et al. 2006; A. Graham 2010: vegetational history of Latin America). Thus Palaeocene-Eocene palynomorph diversity in both Venezuela and Colombia was low (Morley 2007). This may reflect a rather belated recovery from the bolide impact and/or that the tropical rain forest ecosystem was just developing (Wing et al. 2009). Laminar venation density in the Colombian forests is very high (Wing et al. 2009; see also Burnham & Johnson 2004), and this is the first fossil evidence of functional equatorial neotropical megathermal rain forest (Feild et al. 2011b; see also Jud & Wing 2013). Note, however, that taxa with wind-dispersed fruits, common in today's canopy trees, lianas and epiphytes, are uncommon in Palaeocene Colombian floras (Herrera et al. 2014b). Epihov et al. (2017) suggest that tropical forests rich in N-fixing legumes, although with a reduced representation of palms, spread in the Palaeocene-Eocene 58-42 Ma; these forests are known from the Americas, Africa and Europe. For woods in Late Middle Eocene deposits ca 39 Ma on the Pacific side of Peru, see Woodcock et al. (2017).

During the short-lived Palaeocene-Eocene thermal maximum (PETM) of about 55 Ma temperatures increased 3-8o or more - estimates vary - to mean annual temperatures (MAT) 31-34o C (e.g. Willis & MacDonald 2011). (Note that the MAT of l.t.r.f. today is ca 27.5oC, and photorespiration predominates over photosynthesis above 35o - Sun et al. 2012, but c.f. Busch et al. 2017.) Over 2,000 gigatons of carbon were released in ca 10,000 years, the whole event lasting a mere 100,000-200,000 years, and during this time there was a 23oC oscillation in deep ocean temperatures (Zachos et al. 2008; Taggart & Cross 2009; McInerney & Wing 2011). The cause may have been the impact of some extra-terrestrial body (Schaller et al. 2016) or, more likely, volcanic eruptions, perhaps in the North Atlantic Igneous Province (Gutjahr et al. 2017; S. M. Jones et al. 2019: thermogenic methane predominated). Humidity, precipitation and so rock weathering also increased during this period (Zachos et al. 2001, 2008). In South America plant diversity and origination rates increased at about the time of the PETM, but there is no evidence of thermal damage to the leaves (Jaramillo et al. 2010) despite temperatures 5-70 above current values (Jaramillo & Cárdenas 2013). In North America (Wyoming area) mesophytic plants, especially conifers, were temporarily replaced by species that could tolerate both increased temperature and decreased precipitation, and plant and herbivore diversity and herbivore activity were high, perhaps correlated with the high temperature (Currano et al. 2008; Jaramillo et al. 2010). Movement of floras and replacement of Cupressaceae and Podocarpaceae occurred elsewhere, too (Wing & Currano 2013). Diversity was also very high - if stable - in Late Palaeocene Gulf Coast floras, pollen diversity increasing ca 15% (Harrington & Jaramillo 2007; Jardine et al. 2018: any changes due to immigration). Cai et al. (2017/19) found that ages of whole genome duplications in Malpighiales clustered around the PETM (19/24 of the events found, mean age ca 53.9 Ma, or (60.4-)56.8(-54.8) Ma), suggesting that such duplications enhanced clade survival at this time. Although Citerne et al. (2010) thought that this was a period of floral innovation, overall increases of diversification were not detected (see also Magallón et al. 2015; Silvestro et al. 2015). Diversification in Loranthaceae (B. Liu et al. 2018) and epiphytic ferns (Schuettpelz & Pryer 2009) may have increased around this time - and of course trees are essential for both.

The PETM may be associated with some marine extinctions, although these were not particularly notable (Clapham & Renne 2019) and there were distributional shifts in both terrestrial plants and animals and reworking of sediments and increased sediment flux (and more clay moving to the ), but overall there seems to have been little terrestrial extinction (Wing 2004; Wing et al. 2005; Willis & MacDonald 2011; McInerney & Wing 2011; Foreman et al. 2012; Wing & Currano 2013; c.f. Mander et al. 2010). However, at the end of the Palaeocene there was a pronounced (ca 20%) decrease in palynological diversity in the then paratropical Gulf Coast floras (Harrington & Jaramillo 2007), which Mander et al. (2010) consider to be a significant decline - sporomorphs are a more reliable indicator of presence than are meso- or macrofossils.

Fires decreased notably in many parts of the world from the mid Palaeocene to the Pliocene (Bond et al. 2005; Bond & Scott 2010; Belcher et al. 2010b; He et al. 2012; Bond & Midgley 2012), Australia perhaps being an exception (e.g. He et al. 2011; Crisp et al. 2011; Crisp & Cook 2013). Perhumid conditions had spread and large angiosperms dominated; litter decayed quickly and there were few shrubs to support fire (c.f. the Cretaceous, fire-susceptible Proteaceae-dominated heath evolved in Australia then, Carpenter et al. 2015). However, in parts of Europe there is evidence for episodic fires in a vegetation dominated by ferns and perhaps Fagales (Collinson et al. 2007), while Lamont et al. (2018b: fig. 12 for new clades that had various fire-associated traits) suggest there was a first massive peak in the incidence of fires 60-40 Ma. There was a small peak of fire activity in the Oligocene.

Although temperatures soon moderated after the PETM, they became gradually warmer again, peaking at similar high values during the Early Eocene Climatic Optimum (EECO) of 52-50 Ma, a much warmer, wetter and more temperate period than now (e.g. Greenwood & Wing 1995; Upchurch et al. 2007; Zachos et al. 2001, 2008; Sluijs et al. 2009; Kroeger & Funnell 2011). Paratropical forest was common in the early Eocene from the eastern part of the Northern Hemisphere (Mayr 2009). Species diversity of both herbivores and plants was probably at a maximum in later Eocene forests, partly because of flatter global temperature gradients and an overall warmer earth (Jaramillo & Cárdenas 2013; Morley 2007; Labandeira & Currano 2013: both hemispheres; see below); in the earlier Eocene temperate and subtropical taxa grew together in western North America (Wilf 1987). Palaeocene and Eocene Patagonian vegetation was mostly more diverse than its North American counterparts, similarly, the diversity of herbivore damage in a fossil flora from the early Eocene in Argentina was appreciably greater than that in comparable North American floras (Wilf et al. 2005; Iglesias et al. 2007; Wilf 2008; Wilf et al. 2011). Early Eocene South American fossil floras were notably diverse, even at 47oS in Patagonia, and included lianas, the diversity there declining only at the end of the Eocene (e.g. Jaramillo et al. 2006, 2010; Herrera et al. 2011; Wilf et al. 2003, 2011). Diversity in western North America seems to have been comparable to that in these southern floras (R. Y. Smith et al. 2012).

In the Eocene, palm trees grew well inside the Arctic circle (Eldrett et al. 2009; Sluijs et al. 2009). Some angiosperms even grew at the then-north pole ca 60 Ma, and Wolfe (1978), Chaloner and Creber (1989) and Daly et al. (2011) discuss how plants could grow at such high latitudes. The diversity in unique mixed deciduous broad-leaved and evergreen and deciduous conifer forests that grew north of 65-70o N, the polar deciduous forest, had increased markedly since the Cretaceous, a number of new genera appearing, and these forests were remarkably speciose considering that it was dark for about a third of the year (Hickey 1984; Wolfe 1987; Collinson 1990; Jahren 2007; Taggart & Cross 2009 for references; see also Spicer & Herman 2010), although Zanne et al. (2018) provide another way of thinking about the ecology of plants growing in areas in which there are sometimes freezing temperatures. There seems to have been some local endemicity in these forests, and from the late Cretaceous to the Eocene both plants and animals that later moved south evolved there (Hickey et al. 1983; Harrington et al. 2011). Rich forests ca 45 Ma (Eocene-Lutetian) have been described from palaeolatitude 78.6o N in Canada (Jahren 2007). This Arctic flora has also been compared with that of the Pacific Northwest, although overall there may be more similarity with eastern Asia floras, especially in precipitation seasonality (Schubert et al. 2012). In the southern hemisphere, too, palm trees grew well inside the Antarctic circle, with paratropical rain forest recorded from off Wilkes Land, eastern Antarctica, in the early Eocene ca 51 Ma (Pross et al. 2012). There was also considerable diversity in Patagonia and the Antarctic Peninsula (Barreda et al. 2012b), where pollen diversity in an Early Eocene site from South Ellesmere Island (76o S) was similar to that of vegetation in the southeastern United States today (Harrington et al. 2011; see also Bowman et al. 2014).

Chrysomelid bruchine beetles, now restricted to palms, have been found in Eocene deposits in Washington and British Colombia in North America in rather mountainous localities and in Primorye, eastern Russia (Archibald et al. 2014). Palms, gingers, tree ferns are found in 64-53 Ma deposits from west central North America, suggesting warm equable climates with little frost (modern lats 39.17-48.5oN: Wing & Greenwood 1995; see also Wolfe 1978). At this time, extratropical climates showed little seasonality, and plants which would seem to have mutually exclusive climatic preferences grew together. Harrington and Jaramillo (2007) noted the mixture of families that are now temperate or tropical in floras of the Gulf Coast in the late Palaeocene. Overall, latitudinal diversity gradients were flatter or even peaked in temperate regions (e.g. Archibald 2013), as they had since the Triassic (see below). Land bridges in the Palaeocene that would allow warmth-requiring plants to migrate around the northern hemisphere include the Beringian (North America-East Asia) and North Atlantic bridges during the Paleocene in particular (e.g. Tiffney 1985b; Brikiatis 2014); Morley (2003) discussed interplate plant movements during the Cretaceous and Palaeogene. Many tropical taxa whose ranges are restricted and/or interrupted had much wider distributions in the Eocene (e.g. Wing 1987; Archibald et al. 2010; Plaziat et al 2001: Nypa; S. Y. Smith et al. 2008: Cyclanthaceae; Herrera et al. 2011: Stephania; Collinson et al. 2012: survey of the middle Eocene Messel flora; Stull et al. 2016 and references: Icacinaceae), and some of these broader distributions persisted as late as the Miocene (e.g. Ferguson et al. 1997; Manchester et al. 2009: East Asian endemics). There have been similar range changes in the Southern Hemisphere, for example in conifers (Wilf 2012), Eucalyptus (Hermsen et al. 2012), etc.. It has been suggested that southern temperate forests and Mediterranean vegetation are perhaps the best modern analogues of this rather aseasonal early vegetation, and both are notably speciose (Archibald et al. 2010).

In South America the tropical flora did not shift south when temperatures were high during the Early Eocene Climatic Optimum, rather, a new kind of vegetation the ecophysiology of which is unknown developed in paratropical areas below 24oS, and the dissimilarity between tropical and extratropical floras increased (Jaramillo & Cárdenas 2013; see also Romero 1993 and references). Aspects of the subsequent evolution of the extra-tropical flora there are rather distinctive (see below). Interestingly, the diversity of the flora of the U.S. Gulf Coastal Plain was also independent of what was going on in Colombia in the early Palaeogene - stable, any changes due to migration (U.S.), versus in situ origination (Colombia) (Jardine et al. 2018).

Over a period of perhaps 800,000 years in the middle Eocene ca 50 Ma Azolla covered vast areas of the Arctic Ocean, this was the Azolla event (Brinkhuis et al. 2006; F.-W. Li et al. 2018). Azolla sequestered large amounts of CO2, drawing down an estimated 55-470 ppm pCO2, and it may well have contributed to the beginning of the greenhouse → icehouse earth transition (Speelman et al. 2009). Indeed, a long-term cooling trend had begun by the end of the Eocene, perhaps also assoociated with such factors as the drawdown of atmospheric CO2 as it was used up in the weathering of the large amounts of mafic and ultramafic rocks exposed by the collision of India and Africa with Eurasia between 80 and 40 Ma (Jagoutz et al. 2016) or changes in circulation patterns in the North Atlantic and North Pacific by blocking of the Arctic-Atlantic gateway (Hutchinson et al. 2019). Cooling in the Antarctic Wilkes Land shelf area and spread of more temperate Nothofagus fusca-type pollen had begun by the middle Eocene ca 45 Ma (Pross et al. 2102). The cooling trend was accentuated at the beginning of the Oligocene (e.g. Wolfe 1978; Millar 2011; Pagani et al. 2005). It has been estimated that there has been a 30o C or more reduction in the MAT in the far north since the end of the Eocene (Jahren 2007), temperatures dropping 8.2±3.1oC in just 400,000 years at the beginning of the Oligocene some 33.5 Ma in central North America, apparently with little change in precipitation (Zanazzi et al. 2007), although in parts of North America woodlands became more open and arid in the Late Eocene/Oligocene (Mayr 2009). Eriksson (2016) saw this opening woodland as favouring frugivores - birds (passerines) and bats - that could readily move between the patches of trees. Evergreen broad-leaved forests, today very diverse especially in mainland South East Asia, seem to have developed by the Oligocene (X.-G. Xiang et al. 2016).

There may in fact have been short periods of glaciation in the Antarctic by 42 Ma (Tripati et al. 2005) or perhaps even earlier (Bowman et al. 2014; Ladant & Donnadieu 2016). The Antarctic ice sheet appeared ca 33.5 Ma at the Eocene-Oligocene boundary and persisted through much of the Oligocene (Zachos et al. 2001; Coxall et al. 2005; Eldrett et al. 2009 and references). It had been thought that Arctic ice started developing around a mere 7 Ma (Zachos et al. 2001), becoming widespread only in the early Pleistocene 2.4-2.2 Ma (Brigham-Grette et al. 2013; Knies et al. 2014), 10 Ma later than the development of major ice sheets in the Southern Hemisphere (e.g. Zachos et al. 2001, 2008; Retallack 2009; Millar 2011; Crisp & Cook 2011). However, recent work suggests that there was ephemeral Middle Eocene to early Oligocene ice on Greenland, and sea ice in the Arctic, too, and at least the latter may have persisted (Tripati & Darby 2018). Importantly, seasonality greatly increased at the Eocene-Oligocene boundary, even if it was to decrease somewhat later in the Caenozoic (Wolfe 1978). This seasonality was evident in extratropical floras at the end of the Eocene or somewhat earlier (e.g. Wing 1987; Eldrett et al. 2009), but marked seasonality in fossil woods is a Neogene (Pliocene and since - the last ca 23 My) phenomenon (Wheeler & Baas 1993); more or less ring porous woods showed a marked increase from the Palaeocene to the Eocene, and again from the Oligocene to the Miocene (Wheeler & Baas 2011). Zanne et al. (2018) look at correlations shown by members of extant floras and cooler temperatures, suggesting that freezing, and the vessel cavitation that might result, is likely to have played an important role in evolution, interacting with habit (deciduous versus evergreen, herb versus woody), precipitation, and vessel diameter. Mountain-building in the northern hemisphere beginning ca 40 Ma, first in the Himalayas/Quinghai-Tibet Plateau area, led to the evolution of a cold-adapted flora that later moved into the Arctic (Hagen et al. 2019).

Overall, tropical floras became less widespread after the Eocene (Morley 2007), extra-tropical floras became less diverse and less cosmopolitan (Archibald et al. 2010), and deciduous plants became more widespread. Diversification in microthermal deciduous forests was mostly within genera (Wolfe 1987). Although few Eocene leaf remains can be assigned to modern genera, by the end-Eocene this can commonly be done (e.g. Wing 1987; Dilcher 2000). Although ecological conditions then may differ somewhat from those of today, any differences are surely less than those between Cretaceous and extant vegetation (Mittelbach et al. 2007). In addition to climatic changes, biota like that of Australia were affected by its continuing movement northwards that facilitated southwards migrations (Oliver & Hugall 2017).

Temperatures in the later Oligocene rebounded slightly and then oscillated through the Miocene (e.g. Wolfe 1978. An essentially modern Malesian flora developed around the Oligocene-Miocene boundary ca 23.6 Ma (Morley 2007). The mid-Miocene ca 16 Ma in particular was quite warm and wet (Londoño et al. 2018 for CO2 concentration then), and the Atlantic and Amazonian rain forest were continuous then, although they are now separated by a band of drier vegetation (Morley 2000, 2007). In the mid-Pliocene some 6-3.6 Ma MATs were 2-3o C warmer than they are now and there were novel vegetation assemblages and increased diversity (Willis & MacDonald 2011). However, this warmer period was followed by a further long-term temperature decline, latterly precipitous, to the Pleistocene.

Two major vegetational changes involving the appearance of new biomes have taken place in the last 10 million years or so. First, the now very extensive boreal forest/taiga, some 30% or more of global forest cover, developed, probably from high altitude forests in the far north rather than from the extensive polar deciduous forest, and this forest has formed within the last 12 Ma as climates have become drier and cooler (Taggart & Cross 2009; Pound et al. 2012). Here the decreasing [CO2] which had given a photosynthetic advantage to angiosperms was offset by the cooling temperatures, which more or less nullified these advantages (Yiotis & McElwain 2019). Ca 60% of the total carbon in forests globally is in these boreal forests (Taggart & Cross 2009). Details of carbon storage suggest considerable dynamism since the last glaciation. Thus in the far north the extensive peatlands where Sphagnum is now common have been dated to within the last 17,000 years, i.e. this side of the last glaciation maximum at ca 21,000 y.a. (Morris et al. 2018; c.f. Treat et al. 2019: peat accumulation in the north up to 130 ka). Increased carbon stored here more than compensates for the reduction in carbon stored in permafrost mineral soils and loess deposits since the last glaciation (Lindgren et al. 2018), although the balance between new peat accumulation and its longer-term sequestration after burial by marine or glacial depositis is complex (Treat et al. 2019; see also Rogers et al. 2019: different system, same principles).

Second, the great ecological importance of grasses, especially those that carry out C4 photosynthesis, has developed only within the last (10-)7-4 Ma (e.g. Jacobs et al. 1999; Keeley & Rundel 2003; Jacoobs 2004; Edwards et al. 2010). Indeed, the expansion of C4 vegetation may be as recent as ca 3.5 Ma, as in NW Australia (Andrea et al. 2018), and not much older in South America (Palazzesi & Barreda 2012). The origin of this trait goes back 20 Ma or more (and Bowes 2010 suggested that C4 photosynthesis first occured in aquatic systems), and by the late Eocene grass was 30-40% of the pollen in the Niger Delta region, and there was evidence of fires (indeed, the hygrophilous African flora became impoverished by stages, at the K/P boundary, at the end of the Eocene and Miocene, and in the Plio-Pleistocene) (Maley 1996 and references). Tropical and subtropical savanna, including the widespread Cerrado vegetation of Brazil (for which, see Simon et al. 2009; Simon & Pennington 2012) and both tropical and temperate grasslands have spread very widely, and although dominated by relatively few species of grasses, they have come to play a major role in the terrestrial biosphere (Lehmann et al. 2019). For Old World savannahs, see Denk et al. (2018) and Fortelius et al. (2019) and discussion.

A considerable increase in the frequency of fires over the last 10 Ma is associated with the spread of grassland and savanna (Bond et al. 2010; Bond & Scott 2010; Belcher et al. 2010b), indeed, Lamont et al. (2018b: fig. 12) suggest that the incidence of fires in the Tertiary showed a second massive peak that has lasted the last 30 Ma. An interaction between CO2 and fires may be important here: Decreasing atmospheric CO2 would have given C4 grasses an advantage (Ehleringer et al 1997), and it would also reduce the growth rate of woody vegetation, so hindering its recovery from fires (Bond et al. 2003a, b) - see also Poaceae. Fires do cause the release of CO2 into the atmosphere, and although inertinite, also produced by fires, is highly resistant to decay and so can be involved in C sequestration, it is largely absent from the Caenozoic geological record (Bond 2015). Fires also affect the nitrogen cycle, volatilizing N (Forrestal et al. 2014 and refs.), although ammonia may become strongly attached to fire-derived organic matter, if with unclear consequences for the N cycle (Hestrin et al. 2019).

6C. Latitudinal Gradients of Diversity. Details of the relationships between groups diversifying in seasonal temperate regions and their tropical relatives and the different numbers of species in temperate and tropical regions have long been a matter of speculation (e.g. Bews 1927). Some questions are, what global patterns of biodiversity can be discerned?, what causes them?, when have they been evident? (e.g. Pianka 1966; Kier et al. 2005; Schemske & Mittelbach 2017). Knowing the numbers of taxa involved, the ages of the clades, and the ecological attributes that can be linked with these clades are all important, and the perspective changes when the focus broadens to incorporate past climate changes.

Much current plant and animal diversity on both land and sea is broadly correlated with climate, with fewer species towards the poles. However, during the Palaeocene-Eocene period climates were much less seasonal than now and the flora was more homogeneous with considerable diversity even at higher latitides, although not at the very highest. Groups like palms that are now tropical were then found far both to the north and south of their current distributions (see above). Such diversity distributions that are not strongly tropicocentric may be the normal condition for the planet, gradients like those associated with the ice age climates of the present being the exception rather than the rule. The climate is currently strongly seasonal in both southern and in particular northern latitudes, and diversity usually declines away from the tropics (Fischer 1960; Hillebrand 2004: comprehensive metaanalysis, very few exceptions; Francis & Currie 2003: families!, see Qian & Ricklefs 2004 for problems with distribution maps; Hawkins et al. 2011 for a reanalysis; Jablonski et al. 2013; Khine et al. 2019: ferns in eastern Asia). Similar latitudinal diversity gradients have been most obvious in cooler/glacial periods in the past (Mannion et al. 2013; c.f. Benton et al. 2010), and more recently they may be a post-Eocene phenomenon (Archibald et al. 2010, 2012; Rose et al. 2011: temperature gradient similar to that of today; Mannion et al. 2012; also Boyero 2014 for literature). Indeed, the development of latitudinal diversity gradients evident in extant North American mammals have been dated to within the last 4 Ma or so (Marcot et al. 2016): see Fraser (2017) for a discussion on the detection of such gradients in mammals in the fossil record.

An ever-growing literature focusses on establishing mechanisms that would cause/explain current global patterns of diversity (e.g. Pianka 1966; Willig et al. 2003; Mittelbach et al. 2007; Schemske et al. 2009; Hurlbert & Stegan 2014; Chu et al. 2018; Fine 2015 and Pontarp et al. 2019 for critical discussion), and the mechanisms have tended to shift from ecological to evolutionary over the years (Schemske & Mittelbach 2017). Differences in rates of speciation or extinction, differences in the amount of incident energy or habitable areas, longer times of climate stability, more/closer interactions between organisms at lower latitudes, and the like have all been invoked in such articles (see Etienne et al. 2019 for a summary of the main classes of mechanisms -variation in diversification rates, carrying capacity/ecological limits,time to accumulate species). Connections between diversity and environmental energy variously estimated (and this links with latitude), species richness and the rate of molecular evolution are likely to be independent (Davies et al. 2004b; Moser et al. 2005; Jaramillo et al. 2006; Dowle et al. 2013). Evapo-transpiration, topographical diversity, and related factors may also be important (Kreft & Jetz 2007). Lamanna et al. (2014) looked at the alpha, beta and gamma components of functional trait space (specific leaf area, seed mass, plant height), noting that trait hypervolume was greater in temperate areas, although it was unclear how species filled this space. A longer growing season in the tropics means that overlap in the timing of reproduction may be less, and so intraspecific competition increases, at least proportionally, because seedlings of the one species compete more strongly, there being fewer heterospecific seedlings around, so abundance of individual species is constrained, so facilitating the increase of overall diversity (Usinowicz et al. 2017: Pasoh sometimes an outlier, but masting there; see also Mittelbach 2017). Attempts to explain diversity patterns along similar lines continue (e.g. J. H. Brown et al. 2004; Condamine et al. 2011; Gillman & Wright 2014; J. H. Brown 2014; Tomasových et al. 2016). However, it is likely that more than one mechanism is involved, and results of studies of such biodiversity patterns have been likened to reports from the blind men examining the elephant (Hurlbert & Stegan 2014). Indeed, the richness of the Gulf pollen flora in the early Palaeogene (stable, any changes due to migration) was independent of that in Colombia (in situ origination) (Jardine et al. 2018), and similar independence is evident in floras in the southern part of South America (Jaramillo & Cárdenas 2013); there appears to be no single driver of diversification then (Jardine et al. 2018). Allen et al. (2002) thought that productive environments could support more individuals (but c.f. Jansson & Davies 2015), therefore ceteris paribus more mutations and evolution. However, Antonelli et al. (2015) found no differences between speciation and extinction rates of tropical and non-tropical angiosperms, although these rates were significantly higher in the Neotropics compared with tropical Asia and Africa and more species moved from the New to the Old Worlds. Etienne et al. (2019) thought that latitudinal changes in the carrying capacity of the environmant best explained these patterns.

Herbivore diversity is often thought to be greater in tropical than in temperate forests, and so defences should be disposed likewise (Adams et al. 2011 and references; see also Agrawal et al. 2012). Novotny et al. (2006) suggested that individual species of temperate and tropical plants (controlled for phylogenetic relationships) supported a similar number of insect species, but since there were many more species of plants in the tropics, there would be many more species of insects there. However, there may be other patterns of association (c.f. Novotny et al. 2007 and Dyer et al. 2007), and an analysis of the food plants of Californian butterflies showed plant and butterfly diversity to be at most weakly correlated, whether the caterpillars had broad or narrow host plant preferences (Hawkins & Porter 2002). Two comprehensive analyses suggest that both herbivory and allocation of resources to plant defences tend to be greater at higher latitudes away from the equator (Moles et al. 2011a, b), while Salazar and Marquis (2012) noted that although the diversity of herbivores on Piper increased towards the equator, the amount of herbivory did not (see also Moles 2013; Richards et al. 2010, 2015). Overall, specialized insect herbivores are more frequent in tropical regions, and this may be connected with the greater lineage diversity of their hosts there; the numbers of generalist herbivores showed no latitudinal correlations (Forister et al. 2015; see above for herbivore specialization). How this might relate to amount of herbivory is unclear. For more on whether or not there are gradients of biotic interactions, see HilleRisLambers et al. (2002) and Schemske et al. (2009).

Wiens and Donoghue (2004; see also Kerkhoff et al. 2014; Qian & Ricklefs 2016) suggest that phylogenetic niche conservatism might contribute to the higher diversity in the tropics. Groups that are tropical in origin adapt with difficulty to seasonal temperate climates (and vice versa), and older plant clades tend to be more tropical/southerly in distribution, younger ones are more cold-tolerant and northern, in line with the ages of temperate and tropical climates. The tropical clades probably evolved in the Cretaceous, i.e. before the current latitudinal climatic patterns were established, the temperate clades are mostly (but not all) post-Eocene in age (Kerkhoff et al. 2014: New World; Hawkins et al. 2014 and references: conifers not included; Richardson et al. 2015). The older a clade, the more time it will have to speciate, and amount of speciation may also be linked with the area of biomes (Fine & Ree 2006), however, clade age and clade diversity seem not to be linked (Rabosky et al. 2012). Similarly, Linder (2008) linked the timing of diversification in particular areas to whether or not the local environment had been climatically and geologically stable during the Caenozoic, while Fine (2015) linked diversity to the greater extent of tropical environments and their climatic stability. In the early study of temperate-tropical family pairs by Judd et al. (1994), the temperate family, younger and tending to be herbaceous, often arose from within a tropical family, older and woody. Laliberté et al. (2013) discussed diversity in terms of the youth of the soils; tropical soils were older and more strongly weathered and supported a diverse vegetation, soils at high latitudes were younger, less weathered, and supported less diversity. (They noted that there were different forms of nitrogen like NH3 and NO2 in the soil, but the mycorrhizal status of the vegetation also needs to be taken into account.) Jansson and Davies (2015) noted that current diversification rates were highest in the tropics, but outside l.t.r.f., and they thought phylogenetic niche conservatism might well play a role in explaining diversity in l.t.r.fs; rapid diversification in the areas outside l.t.r.f. might be quite recent and connected with areas that showed little change despite climatic changes that were dependent on changes in the earth's orbit. In a global survey, LaManna et al. (2017) noted a strong positive correlation between the diversity of plants and negative density dependence relationships between them, the Janzen-Connell effect. An additional wrinkle is that overall species richness in the tropics is greatest in wetter environments, but lineage diversity is highest in areas of intermediate precipitation where clades that have members living in drier condition and those that have members living in wetter conditions can cohabit (Neves et al. 2020: South America). However, separating historical and ecological signals in patterns of plant diversity is not at all straightforward (Ricklefs 2005).

B. T. Smith et al. (2012) refined the niche conservatism hypothesis and proposed that in New World vertebrates, at least, families with southern origins were more likely to show conservatism than those of northern origins. Southern families have not penetrated the highly seasonal Nearctic, not simply because they diversified less/were exposed to less competition in the southern temperate zone, much smaller than the north temperate zone, but perhaps also because the southern temperate zone is on balance more equable, so it is temperate in a way different from the northern temperate zones with their greater greater climatic fluctuations. Indeed, there are some differences in diversity patterns between the two hemispheres. Thus Kerkhoff et al. (2014) found that high latitude southern floras were less absolutely diverse but were older than comparable northern floras, however, the southern flora had an only somewhat lower phylogenetic diversity than that of the tropics, while that in northern latitudes was notably lower. In the equatorial Andes, as minimum temperatures decrease with increasing elevation, the ages of the clades of the woody plants there increases (Qian 2014: focus on families). To Segovia and Armesto (2015), this was because the woody Andean plants were representatives of an old, temperate, Gondwanan flora, for which there is also evidence in Patagonian fossils ca 52.2 Ma (e.g. Wilf & Escapa 2014 and references; see also Cantrill & Poole 2012; Kooyman et al. 2014: phylogenetic biome - Nothofagus forests - conservatism; Segovia & Armesto 2015). Similarly, Leslie et al. (2012) found that most southern hemisphere clades of Pinales are older than northern clades, the latter having been more subject to major climate swings beginning in the Oligocene. Equable climates most similar to those of the early Caenozoic are now to be found mostly in tropical and to a certain extent south temperate areas (Janzen 1967; Platnick 1992; Chown et al. 2004; Ghalambor et al. 2006; Leslie et al. 2012). However, Qian and Ricklefs (2016) suggest that the ages of clades tend to increase with altitude in tropical mountains in general, and there is niche convergence between unrelated clades (see also Culmsee & Leschner 2013). Complicating the interpretation of such patterns, Schluter and Pennel (2017) found that current rates of speciation tend to be highest where species richness is low, and species of birds and mammals, at least, tend to be younger at higher latitudes, although on the other hand Calcagno et al. (2017) noted that diversity spurred diversification, at least in the initial stage of an adaptive radiation...

It should be noted that not all groups are increasingly diverse towards the tropics (Kindlmann et al. 2007 for review). Looking at N-S curves at a global level, Africa-Europe is the odd man out since there is no clear latitudinal trend in species richness there (Mutke & Barthlott 2005), in Europe probably because of the E-W trending mountain ranges and in Africa because of Palaeogene-Neogene drought and extinctions. In China, mosses showed weaker latitudinal diversity gradients than liverworts (and angiosperms), perhaps because the former can handle a wider diversity of environments than liverworts in particular, a number of which are epiphyllous and need humid conditions (S.-B. Chen et al. 2015); more globally, species diversity in mosses shows non-significant increases towards the equator, unlike the increase in vascular plants (Möls et al. 2013). However, a few angiosperm clades like Carex and some other Cyperaceae (Spalink et al. 2016), Polygonaceae and Ranunculaceae are most diverse away from the tropics (e.g. Escudero et al. 2012b; Kostikova et al 2014b). In grasses the diversity gradient - much flatter than might be expected - is affected by the diversification of Poöideae, largely temperate, the specialization of many members of the PACMAD clade to more or less arid conditions, and the occurrence of grasses in topographically heterogeneous mountainous regions (Visser et al. 2013). Gymnosperms like Pinaceae and Cupressaceae are also more diverse away from the tropics, while Cactaceae, especially at the species level, are most diverse at ca 20o N and S (Mutke & Barthlott 2005). The diversity of forest understory herbs also does not increase towards the tropics (Ramos & Skillman 2015). Weiser et al. (2018) found that of families in the North American flora, 32 (13%), including Cyperaceae, Juncaceae and Rosaceae, had a reversed gradient, while Polygonaceae showed no particular pattern. And in another wrinkle, Weiser et al. (2018) noted that 10 families account for more that 70% of the gradient in North American flora, Asteraceae and Fabaceae alone contributing a third, and 53% of the families contributed little or nothing; surprisingly, the number of species per family did not vary with latitude. /p>

Thinking about these latitudinal diversity patterns in the context of the interactions between angiosperms and their fungal associates, interactions which affect plant diversity, soil fertility and carbon content, etc., provides another way of looking at the issue (see below), although overall there are a number of gaps in our knowledge of below-ground diversity (e.g. E. K. Cameron et al. 2018; Phillips et al. 2019). The distribution of ectomycorrhizal(ECM)-dominated communities, particularly pronounced polewards and especially prominent in the northern hemisphere where, for example, all boreal tree species are ECM plants (Smith & Read 2008; Steidinger et al. 2019), may contribute to the current diversity gradients, and bears on the hypothesis that diversity is in some way linked with productivity (e.g. Willig et al. 2003; see also Veresoglou et al. 2019: light). Contemporary ECM-dominated communities are common in more extreme and unproductive environments and they are species-poor, at least when it comes to angiosperms (Gillman & Wright 2006; Cusens et al. 2012; c.f. Adler et al. 2011: focus on herbaceous communities, see also below). ECM plants are abundant in boreal forests in particular, but also in many temperate forests (for ECM plants in Africa and tropical Southeast Asia, see Fabaceae-Detarioideae and Dipterocarpaceae). Although overall diversity of AM and general soil fungi increases towards the equator (less so in Africa), the diversity of ECM (but not AM) fungi increases in mid to high northern latitudes - or at the very least it is flat - and ERM fungi also show a diversity increase towards the poles, if not at the very highest latitudes, patterns that are consistent with the distribution of their seed plant associates (e.g. Bjorbækmo et al. 2010; Timling et al. 2012; Timling & Taylor 2012: D. L. Taylor et al. 2013; high frequency of melanized fungi; Wardle & Lindahl 2014; Tedersoo et al. 2014b: details of distributions of functional types and taxonomic groups; Davison et al. 2015; see also Pärtel et al. 2016; Toju et al. 2018). Thus the diversity of ECM fungi like Amanita may peak in more temperate climates (Sánchez-Ramírez et al. 2015a), while in Russula diversification rates are highest in extratropical lineages/those associated with Pinaceae (Looney et al. 2015). ECM diversity also increases with altitude on Mt Kinabalu (Borneo), being highest in mid-elevation montane forest for most taxa (Geml et al. 2017: fungi in the Tomentella area, very diverse in arctic-alpine habitats, an exception) - indeed, a similar pattern is found in several groups of plants and animals on the mountain (Geml et al. 2017 for references). ECM diversity is highest in northern temperate mid-latitude areas (see also Tedersoo et al. 2012), AM diversity/abundance decreasing with altitude (Geml 2017). However, Bueno et al. (2017) found that mycorrhizal associations in general decreased in frequency with latitude and increased with elevation in Europe, at high latitudes non-mycorrhizal species making up ca 30% of the total, neither ERM nor ECM associations being that frequent at higher latitudes (see also Delavaux et al. 2019). Duffy et al. (2019) noticed a cline of decreasing diversity of orchid mycorrhizal fungi associated with Spiranthes spiralis with increasing latitude. Indirectly connected with this, in a number of cases, both tropical and temperate, seedling ECM plants and their associated fungi may have positive interactions (see above), an interesting gloss on global patterns of increasing negative density dependence towards the tropics (LaManna et al. 2017). A recent study confirms these general patterns: Agaricomycetes have their greatest species diversity as well as net diversification rates in temperate regions (Varga et al. 2019). Hardly surprisingly, AM fungi do not show reversed diversity patterns (Gorzelak et al. 2017: fungi associated with Thuja plicata, western red cedar]. Such species-poor boreal forests are currently the largest single terrestrial biome, and they are also the youngest woody biome, having formed within the last 12 Ma at the expense of evergreen broadleaf and mixed forests as climates became drier and cooler (Taggart & Cross 2009; Pound et al. 2012; Yiotis & McElwain 2019), so this takes the discussion back to thinking about the existence of latitudinal gradients of diversity over time...

Diversity gradients of other members of the soil biota are rather similar. These include soil bacteria, although comparison with fungal gradients introduces some complications. Thus although bacterial diversity increases towards the tropics in the southern hemisphere, in the north it is largely flat across latitudes, and this may be connected with often lower soil carbon content in the southern hemisphere, many bacteria depending on the decomposition of organic matter for energy (Delgado-Baquerizo et al. 2016; see also Lauber et al. 2009); diversity may be flat with increasing altitude (Fierer et al. 2011: E. Peru) or even increase (J. Wang et al. 2011: Yunnan). Interestingly, phylogenetically diverse Mediterranean shrub communities, which tend to be more productive and have more fertile soils, have richer bacterial communities, but only in terms of species numbers, not phylogenetic diversity (Goberna et al. 2016). Furthermore, different bacterial groups predominate in the two hemispheres (Delgado-Baquerizo et al. 2016). However, in Central and North America bacterial diversity was found to correlate with temperature (J. Zhou et al. 2017). Indeed, in a recent study both the taxonomic and functional diversity of bacteria tended to be highest in middle latitudes, the latter still being quite high in the highest latitudes; taxonomic diversity responded more to mean annual precipitation, functional diversity to soil pH (Bahram et al. 2018). On the other hand, overall fungal taxonomic diversity was slightly higher at middle latitudes and was lower than tropical diversity at higher latitudes, but functional diversity was lower at middle latitudes; the connection here was more with a high C:N ratio (Bahram et al. 2018). Overall there was some antagonism between bacteria and fungi. Earthworms, very important in soil dynamics, were both more diverse locally and more abundant at higher latitiudes, although overall tropical diversity was higher (Phillips et al. 2019). Finally, soil nematodes are notably more abundant in the tundra and in boreal and temperate forests when compared with other regions of the globe, and three of the main functional groups are bacterivores, fungivores and herbivores (van den Hoogen et al. 2019: focus is on top 15 cm of soil, not litter). Comparing overall diversity patterns (plants, animals, above and below ground) E. K. Cameron et al. (2019) found that there were mismatches in above and below ground diversity over 27% of the surface of the globe, notably, diversity was high above ground but low below ground in temperate broad-leaved and mixed forests, while boreal areas tended to have intermediate soil diversity but low above-ground diversity.

One might almost expect organisms that are dependent on plants, have at least some host specificity, yet have more or less passive means of finding their host, would be more diverse if their host plant is more abundant, the "common host hypothesis" (Kindlmann et al. 2007). This idea has been invoked to explain the increased diversity of aphids, although not necessarily of myrmecophilous species, away from the tropics (Bristow 1991; Stadler & Dixon 2005). To the extent that this hypothesis is true for ECM plants and their associates, this has to be factored in to the reverse diversity clines. Along the same lines, there may be latitudinal gradients in the specialisation of mutualistic interaction networks. Interestingly, numbers of galling species, mostly cecidomyiids, are greatest 28-38o N and S (they are notably abundant of chenopods and Asteraceae - Dorchin et al. 2019), especially in sclerophyllous vegetation (Price et al. 1998). Paradoxically, it might seem, specialization decreases towards the equator, and in environments with low plant diversity, pollinators (to a lesser extent) and fruit dispersers become more specialized on the plants that are there - and the mutualistic networks are less stable (Schleuning et al. 2012). In the distinctive ecosystem of Sarracenia pitchers the diversity of organisms from invertebrates to bacteria in the pitchers increases with increasing latitude, the reverse of the normal trend, but here it may be because the numbers of predatory Wyomyia mosquito larvae decrease (Buckley et al. 2003; Kindlmann et al. 2007).

Looking at organisms not so immediately connected with land plants, savanna ants in the Neotropics also show a reversed gradient of species richness (Vasconcelos et al. 2018). Much marine diversity is bimodal, lower at the tropics (Chaudhary et al. 2016), thus Wooley et al. (2016) found that brittle stars living below 2,000 m showed diversity greatest at 30-50o from the equator. Fish showed a very strong bimodal pattern in speciation rates, even though current tropical piscine diversity is very high (Rabosky et al. 2018; see also Mooers & Greenberg 2018). Brittle stars (ophiurids) show signals of diversification in colder, deeper, oceanic waters in response to the episodic reduction in tropical waters during the Caenozoic, while shallower temperate-tropical areas have a more conventional diversity pattern; exchanges between different areas also factor into the patterning here (O'Hara et al. 2019).

6D. Gene and Genome Duplication and Genome Size. There is a fast-growing literature on gene and particularly genome duplication and its evolutionary consequences, also the related issue of genome size where the focus tends to be on its physiological consequences. The literature focusses on genome duplications/polyploids, but there are other ways in which genes can be duplicated and duplicated genes lost (Qiao et al. 2019 for a summary).

For a review of polyploidy, which emphasizes how difficult it is to make generalizations about what one would have thought was a much studied subject, see Soltis et al. (2016) and other papers in American J. Bot. 103(7). 2016; also P. Soltis and Soltis (2012), M. S. Barker et al. (2012, 2016b), papers in Ann. Bot. 120(2). 2017, Doyle and Coate (2018), Rice et al. (2019 and references), etc.. Perhaps 15% of angiosperm speciation events are associated with genome duplication/polyploidy (see Otto & Whitton 2000 and Meyers & Levin 2006 for general overviews). Polyploidy has occurred many times and at all levels of the tree from events that led to the recent formation of species like Sporobolus (Spartina) anglica, formed by allopolyploidization in the 1870s, to the common ancestor of small groups of genera deep within Poaceae and Brassicaceae, or much larger groups such as the asterids, core eudicots and all seed plants - and of course it occurs in individual cells of the plant body - endoreduplication. Overall, angiosperm genomes have been duplicated several to many, many times, so that of Brassica, for example, is estimated be multiplied 288 times, i.e., it should have 1,440-2,016 chromosomes given an ancestral number of 5-7 and no subsequent losses (Wendel 2015), yet its diploid chromosome number is only 38 and its genome size is unremarkable (also e.g. Wolfe 2001 for chromosome number reduction), while the genome of Arabidopsis thaliana is one of the smallest known.

Wood et al. (2009) suggested that polyploidy is quite common during speciation, occurring in ca 15% speciation events in angiosperms, 31% in ferns, but the diversification rate of polyploids is not remarkable. Indeed, polyploidy is quite commonly an evolutionary dead end, with recently-formed polyploid plants speciating less and in particular showing higher extinction rates than diploids (Mayrose et al. 2011, 2014). Scarpino et al. (2014) proposed that the prevalence of polyploids was the result of a ratchet mechanism, being irreversible (but see below); diploids speciated more. A distinction between short- and long-term effects may help clarify the discussion between Mayrose et al. (2011, 2014: emphasis on the former) and Soltis et al. (2009, 2014: emphasis on the latter). For the effects of dysploidy, see Escudero et al. (2014).

Isozyme duplications suggest past polyploidization in clades like Magnoliaceae, Aesculus, and Salix/Populus (Soltis & Soltis 1990). The notably small stomata of some fossils when compared with extant members of these clades may also imply subsequent polyploidization, extinct members perhaps having chromosome numbers half of any of those known in extant members (Masterson 1994: Lauraceae, Magnoliaceae [one point], Platanaceae), this thesis being largely based on the assumption that there is a correlation between stomatal size, DNA content and chromosome number. However, the examples just mentioned are from woody groups that are unlikely to have had herbaceous ancestors, so ceteris paribus polyploidy would be less likely in such groups, smaller stomata might be expected given given falling CO2 concentrations, and genome size and stomatal size are not always coupled (see below). S.-C. Chen et al. (2014), commenting on the small genomes and absence of polyploidy in Fagaceae, suggested that there extensive interspecific hybridization substituted for polyploidy, the two providing similar evolutionary benefits.

Genome duplications have been reported in most land plants (see Szövényi et al. 2014; Devos et al. 2016 for mosses), 244 duplications being recorded in the one thousand plant transcriptomes initiative (Z. Li & Barker 2019); they are infrequent in conifers (Scott et al. 2016; c.f. Z. Li et al. 2015), liverworts and hornworts and still more so in more basal streptophytes (O.T.P.T.I. 2019). Genome duplication, probably usually the result of an allopolyploidy event (J. Wang et al. 2019b; c.f. Y. Liu et al. 2017: Salicaceae-Saliceae; J.-P. Wang et al. 2018: Actinidia) is increasingly being invoked to explain the evolution and diversification of angiosperms (e.g. Vision et al. 2000; Bowers et al. 2003; Blanc & Wolfe 2004a; Schlueter et al. 2004; Adams & Wendel 2005; Maere et al. 2005; de Bodt et al. 2005; de Martins et al. 2006; Chapman et al. 2006; Cui et al. 2006; Jaillon, Eury et al. 2007; Soltis et al. 2009; Van de Peer et al. 2009b, 2017; Duarte et al. 2010; Barker et al. 2010; Jiao et al. 2011, but c.f. Ruprecht et al. 2017; Mühlhausen & Kollmar 2013: myosin motor proteins; Guo et al. 2013; Vanneste et al. 2014a, b; Tank et al. 2015; P. Soltis & Soltis 2016; Gao et al 2018: the VOZ - Vascular plant One Zinc-finger - gene family; R. Ren et al. 2018). By no means all these duplications are included in the order pages, although I hope the more important are... Such genome duplications can be recognised i.a. by the occurrence of numerous gene duplications at a particular node. However, detecting duplications and deciding where they are to be placed on the tree is by no means straightforward, not only because small-scale gene duplications that are independent of genome duplications can be quite common - c.f. e.g. Leebens-Mack et al. (2019) and Larson et al. (2019), also Ren et al. (2018) and Zwaenepoel et al. (2018), see also Zwaenepoel and Van de Peer (2019).

It is often suggested that gene/genome duplications (rather different beasts, e.g. X. Wang et al. 2009; Jiao & Paterson 2013, see above) may facilitate subsequent morphological evolution by allowing the subfunctionalisation and neofunctionalisation of genes (e.g. Ohno 1970; Freeling 2009; Edger & Pires 2009; Renny-Byfield et al. 2014; also Jiao & Paterson 2013; Rensing 2014 and Conant et al. 2014: reviews). One of the gene copies may be lost (for a nice example, see de Martino et al. 2006), perhaps particularly in important housekeeping genes, which thus revert to being single copy genes (de Smet et al. 2013; Wendel 2015 and references). Sankoff et al. (2010) noted that which particular member of a paralog pair was reduced was random, however, contiguous pairs could be resistant to reduction, and there tended to be runs of single copy genes on a single chromosome. In a comparative study of core genes found in all the angiosperms examined, Z. Li et al. (2016) note that multicopy genes tend to represent ancient duplications and are involved in signaling, transport, development and metabolism (but not in mosses like Physcomitrella - Devos et al. 2016 for literature), while single-copy genes are younger and are often involved in the maintenance of genome stability/integrity and organelle function. Novel regulatory pathways may evolve, and there are other changes (e.g. Veron et al. 2007: retention and evolution of MIKC-type networks after genome duplications; Guo et al. 2013; Murat et al. 2013; Conant 2014: Saccharomyces; Vanneste et al. 2015). As Jiang et al. (2013) noted, features of the gene/genome such as complexity of the gene, GC content at the third position, etc., also affect what happens after duplication. Such changes may happen faster when chromosomes or the whole genome are duplicated, and change may be slower if individual genes have duplicated and lie in tandem, perhaps because of co-regulation (Lan & Pritchard 2016: animals; see also Jiang et al. 2013). In biased fractionation/genome dominance, an heritable feature associated with allopolyploidy, genes in one of the genomes may be preferentially retained, while autopolyploidy goes along with the absence of genome dominance (Garsmeur et al. 2013; Woodhouse et al. 2014; Wendel 2015; Steige & Slotte 2016 and references). Dosage-based selection on retained homoeologous genes can persist for millions of years, indeed, still reflecting pre-polyploidy selection regimes (Hao et al. 2018). Conant et al. (2014) summarize various models of duplicate genome evolution. P. Wang et al. (2018) found that gene families in Solanaceae involved in things like fruit ripening and secondary metabolism were very variable, tending to differ between species, and such genes had evolved by tandem duplications, however, housekeeping genes were less variable, and genome duplication was likely to be involved in their evolution. On the other hand, Levy (2019) suggests that in some cases new genes may arise from non-coding sections of the genome. For genome duplication and number of 35S rDNA loci - not always linked - see Hidalgo et al. (2017a). Along similar lines, Garcia et al. (2016) noted disparities in genome size and the number of 35S rDNA loci, finding that in some Brassicaceae the genome was ca 0.5 pg and there were up to eight 35S loci, while in some Liliaceae the corresponding figures were ca 125 pg and two loci; in terms of picogram:locus ratios, this is a difference of around 500[!]. For well worked-out examples of genome evolution after duplication, see e.g. Y. Wang et al. (2016: what happened after the core eudicot duplication(s)) and S. Liu et al. (2013: ditto the Brassica oleracea duplication); as is evident both from a cursory examination of chromosome number and genome size in angiosperms, diploidization after genome duplication is the rule (e.g. Sankoff et al. 2010).

One sometimes gets the impression that the literature on the importance of genome duplications for the clade involved can be summarized by saying duplications can be implicated in diversification when they can, but not when they cannot, but that generally they are/can be made to be implicated... Numerous duplications are thought to have occurred in angiosperms (e.g. Landis et al. 2018), and most seem to be allopolyploidization events (J. Wang et al. 2019b); it has been suggested that they cluster around events like the K/P boundary when allopolyploids were at a selective advantage because of their hybrid vigour, having extra genes/alleles available for selection given the changing environment (Fawcett et al. 2009; see also Vanneste et al. 2014a, b; Lohaus & Van de Peer 2016; Landis et al. 2018; Van de Peer et al. 2017; see also Wallis and Jorge 2018). Species-rich clades and genome duplications have been specifically linked, for example, Soltis et al. (2009: p. 336) associating genome duplications with "a dramatic increase in species richness" in Poaceae, Fabaceae, Brassicaceae and Solanaceae (for Poaceae, see also Salse 2016), and P. Soltis and Soltis (2016) tentatively link the origins of a number of key innovations in angiosperms to ancient polyploidy events while R. Ren et al. (2018) suggest that duplications may be associated with increased speciation, although in the latter case the sister-group comparisons may be suspect. However, since duplication and subsequent diversification may be separated by tens of millions of years - hence the genome duplication/radiation lag-time model (see Van de Peer et al. 2009a; Schranz et al. 2012) - establishing direct connections between the two can be difficult, putting it mildly (see also J. W. Clark & Donoghue 2017). Small-scale events include a decrease in genome size after polyploidy in Veronica (Plantaginaceae) that was linked with increased diversification rates (Meudt et al. 2015b), again with a lag between polyploidy events and subsequent diversification. Estimates of timing of genome duplications and subsequent diversification in angiosperms were summarized by Tank et al. (2015), and the lag-time ranges from 0 to almost 50 Ma. This lag period may reflect the time that it takes for the genome to become diploidized (Dodsworth et al. 2016) and for the other processes mentioned above to occur. At 50 Ma, the lag period suggested by Clark and Donoghue (2017) between the angiosperm ε duplication (319-297 Ma) and crown-group angiosperm age is at the upper limits of Tank et al.'s estimates, however, if Clark and Donoghue (2017) are right in their estimate of the age of the duplication as 319-297 Ma and authors like Magallón et al. (2015) are right in their estimate of the crown-group age of angiosperms as ca 139 Ma, the lag period is almost 170 Ma.

Genome duplications, especially if reflecting allopolyploidy events, may reduce the probability of extinction by e.g. increasing genetic variation and environmental tolerance (Crow & Wagner 2006 and references; see also Van de Peer 2009a; Franzke et al. 2011; Kagale et al. 2014; Z. Li et al. 2016, etc.), and individual mutations are less likely to have an immediate effect. Along these lines, Fawcett et al. (2009; see also Vanneste et al. 2014a, b; Lohaus & Van de Peer 2016; Landis et al. 2018) dated a series of genome duplications within angiosperms to about 70-57 Ma, around the K/P boundary, the time of the Deccan Traps eruptions/bolide impact, suggesting that these polyploids were at a selective advantage because of their hybrid vigour, also having extra genes/alleles available for selection given the changing environment, etc. (see also Visser & Molofsky 2014 for possible advantages of polyploids). For a critical review of polyploidy/genome duplication and its effect of diversification, see Vamosi et al. (2018); Kellogg (2016a) was sceptical about claims of connections between polyploidy and subsequent diversification.

The relationship between genome duplication and genome size is another area where there has been much recent work. Leitch et al. (2005) summarise what is known about genome size for seed plants, for genome sizes in vascular plants, see Lomax et al. (2013), and for genome size in land plants as a whole, see Puttick et al. (2015) and Pellicer et al. (2018); Leitch and Leitch (2013) also discuss the evolution of genome size in land plants. Doyle and Coate (2018) review the complex interactions of autopolyploidy, genome size, and evolution, and polyploidy and genome size are also involved in comparisons of the rate of pollen tube growth in seed plants (Reese & Williams 2018). Nuclear genome size varies some 2400-fold within angiosperms alone (Greilhuber et al. 2006; Garcia et al. 2014), and the maximum in any organism is about 150 Gb (1C value), a figure approached by both some ferns and monocots (Hidalgo et al. 2017c; see also Characters).

Most angiosperms have rather small genomes, and those of broad-leaved angiosperms are notably smaller than those of monocots (e.g. Grime & Mowforth 1982; Soltis et al. 2003c; Pellicer et al. 2018), and looking at the genome sizes of the major groups of land plants, it appears that there has been a reduction over time (e.g. Puttick et al. 2015). Genomes tend to be smaller in island taxa (Hidalgo 2017a and references) and, interestingly, in true mangroves, in the latter case because of fewer long terminal repeat-retrotransposons (Lyu et al. 2017). Devos et al. (2002) noted that duplication of Long Terminal Repeat (LTR) retrotransposons was responsible for changes in genome size, but in Arabidopsis illegitimate recombination between two LTRs in the one chromosome could lead to the loss of one LTR and other genomic material - thus there are mechanisms for both increase and decrease of genome size. Many changes in genome size are due to retrotransposon amplification and removal (Bennetzen et al. 2005). Bennetzen and Kellogg (1997) had early floated the idea that increase in genome size might be irreversible, which could be true of some gymnosperms (e.g. Nystedt et al. 2013; Ickert-Bond et al. 2020), but not of angiosperms.

Stebbins (1966) had noticed a correlation between increase in latitude and increase in chloroplast size, and this was confirmed (and linked to increase in genome size) in a survey of angiosperms as a whole, and also monocot and dicots, by Levin and Funderberg (1979), and similar trends in genome size can be seen in individual clades like Poaceae (Levin & Funderberg 1979) and the Caesalpinia Group (Souza et al. 2019). However, more global trends may be driven by replacement of clades with latitude, the clades involved differing in genome size (Levin & Funderberg 1979). Grime and Mowforth (1982) had early noted links between large genomes in the taxa they examined (focus on the British flora), the geophytic habit, and fast growth by cell expansion of large cells (link to large genomes) under cooler conditions in the spring; smaller genomes and cells were associated with cacti and tropical plants, and growth there was by cell division and occurred under warm, moist conditions, this is discussed elsewhere. Connections between the amount of available nitrogen and genome size have been suggested, thus both nitrogen availability and genome size tended to be high in parasites and low in carnivorous plants (Vesely et al. 2013). Polyploidy/large genomes negatively affected plants' ability to compete successfully in low N/P conditions (e.g. Smarda et al. 2013; Pellicer et al. 2018). Bennett (1972) noticed that both mitosis/cell cycle and meiosis were shorter in annual herbs and this was associated with a lower nuclear DNA content; annuals/short-lived plants in general tend to have smaller genomes than longer-lived plants (e.g. Hlousková et al. 2019; Kobrlová & Hrones 2019). Knight et al. (2005) suggested that plant lineages with large genomes were likely to diversify more slowly, have a lower photosynthetic rate, and be under-represented in extreme environments - and ultimately be more likely to go extinct. Such associations would also link to increased speciation rates, etc., of annuals/herbs - see below. For genome size (and other aspects of the genome) and plant invasiveness, see Suda et al. (2014).

There are possible connections between genome and stomate size. Small genomes have been linked to high stomatal and vein densities and associated high photosynthetic rates and the success of angiosperms (Simonin & Roddy 2018). Genome size can be correlated positively with cell size, while Franks et al. (2012) found a correlation between guard cell length and nuclear and genome size among north temperate herbs (see also Lomax et al. 2014; Simonin & Roddy 2018). Stomatal size is negatively correlated with stomatal density; trees, with rather small genomes, have the highest stomatal density (Beaulieu et al. 2008, see also Bainard et al. 2012, c.f. Rupp et al. 2010 for Polystachya [Orchidaceae]). Stomatal size, and hence genome size, may have been inversely correlated with atmospheric CO2 concentration over the last 300 Ma or so (Franks et al. 2012: Fig. 3 - Fig. 4 has problems with ancestral state reconstructions), so both increasing and decreasing in size; Lomax et al. (2013: c.f. fig. 2A and 2B) thought that maximum genome size (derived from maximum guard cell length) may have been steadily increasing from 360 Ma (the Mississippian), but a time-binned average shows a decrease over the last 250 Ma, there was no signal of increased genome size in the early Caenozoic, and grasses, for example, have notably small stomata (Franks & Beerling 2009). Franks et al. (2012) do not suggest any mechanism that facilitated changes in genome size. Indeed, Hodgson et al. (2010) thought that changes in stomatal size could have driven changes in genome size, and they emphasize the complexity of the relationship between stomatal amd genome size. Surprisingly, of the species they examined, stomatal length was highest in vernal geophytes, not species in shade. However, Jordan et al. (2014) found that in Proteaceae, at least, changes in genome size drove changes in stomatal size, the two often being correlated, but some species showed substantial changes in stomatal size without change in genome size, and this was related to the ecology of the plants involved. Note that stomata in early vascular plants may have been involved in drying out the sporangium and so in spore discharge, and there the relationships between genome and stomatal size and atmospheric CO2 concentrations are likely to have been different - none (Renzaglia et al. 2017). Although Furness et al. (2015) found a correlation between genome size and pollen size in Liliales, broader studies suggest that there is little general correlation here (Knight et al. 2010).

There is often little correlation between genome size and chromosome/genome block number and in particular ploidy level (e.g. Leitch & Bennett 2004; Chase et al. 2005; Weiss-Schneeweiss et al. 2005; Bennett & Leitch 2005; Lysak et al. 2007, 2009, 2016; Schnable et al. 2009; Peruzzi et al. 2009; Bliss & Suzuki 2012; Vaio et al. 2013; Gorelick et al. 2014; Jordan et al. 2014; Fleischmann et al. 2014; Gunn et al. 2015; Hohmann et al. 2015; Barrett et al. 2019a: Arecaceae; c.f. in part Jakob et al. 2005; Hlousková et al. 2019), and in Hippeastrum chromosome morphology (bimodal) is largely unaffected by substantial changes in genome size/genome copy after polyploid events (Poggio et al. 2014). There is no correlation between chromosome number and genome size in Cycadales, and chromosome number changes there are probably caused by fissions or fusions (Gorelick et al. 2014). Similarly, when chromosomes are holocentric, genome size is little affected by chromosome number change, as in Cyperaceae (Chung et al. 2012; Escudero et al. 2012a; Lipnerová et al. 2013; López et al. 2017). However, genome size and chromosome number are positively correlated in monilophytes (J. Clark et al. 2016; Pellicer et al. 2018).

Genome downsizing after autopolyploid events was analysed by Zenil-Ferguson et al. (2016), who found interactions between monoploid number, ploidy level and genome size, many and duplicated chromosomes perhaps negatively affecting the meiotic process. Frajman et al. (2015 for references) discuss genome downsizing after polyploidization; since genome duplications have been common in seed plants and there are no comparable changes in genome size or chromosome number, these must be decoupled (e.g. Bennett & Leitch 2005; Leitch et al. 2005; Hodgson et al. 2010; Schneider et al. 2015; Lazarevic et al. 2015; Wendel 2015); for biased fractionation, see above. Systematic signal in genome size may be more apparent at lower taxonomic levels, i.e., when polyploidization has been more recent (Frajman et al. 2015 for references), and there may be a connection between genome size and chromosome number is some ferns (Vanneste et al. 2015 and references), chromosomes being lost less easily there. Note that changes in the amount of repetitive DNA can have major effects on C-values (e.g. Jakob et al. 2004) independent of chromosome number. A recent study suggests that genome size per se does not drive evolution, rather, it is the rate of change in genome size, especially high in many angiosperms (but perhaps not in the ANA grade, or even magnoliids), that is correlated with speciation rate (Puttick et al. 2015).

The rate of root meristem growth seems to be negatively correlated with genome size, and since holoparasites in particular have little need for roots (Gruner et al. 2010), they may have much enlarged genomes (e.g. Piednoël et al. 2012). For the GC content of genomes, which shows interesting correlations with genome size, karyotype morphology (esp. holocentric chromosomes) and some aspects of ecology, see Smarda et al. (2014). Genome size, but not ploidy level, increased the length of the cell cycle, particularly in perennial monocots (Francis et al. 2008). In extant plants, seed size is more or less correlated with genome size (Beaulieu et al. 2007a; Linkies et al. 2010) and more so with plant habit. Seed size and plant height are correlated, but for a possible negative correlation of genome size and plant size, see Beaulieu et al. (2007b). Recent work on Liliaceae suggests that species with large genomes have larger cells, the plants are larger and have larger flowers; Liliaceae are geophytes, and there fast growth/large genomes (= large cells) may be an advantage, although in more extreme conditions where the growth period is short genomes may be smaller and polyploidy is common (Carta & Peruzzi 2016). Obviously, extending this approach to other groups in which geophytes predominate is the next step.

Plant genomes, and perhaps particularly those of flowering plants, are very dynamic (Cavalier-Smith 2005 for general discussion), and the importance of genome duplication is in the opportunities it provides for gene and genome, and hence plant, evolution. However, the significance of much of the variation in genome size is unclear, and strong correlations between genome size and other features remain hard to come by (Garcia et al. 2010) - but see above for latitude. Moreover, different major groups of plants behave differently. Ferns, for instance, tend to have have very high chromosome numbers (n = 57 on overage), and here genome duplication may have been accompanied by gene silencing but not chromosome loss (Haufler 1987; Barker 2013 and references), indeed, in general, heterospory is accompanied by relatively low chromosome numbers (Barker 2013). Gymnosperms show relatively little variation in chromosome number and genome size (Murray 2013; Scott et al. 2016), the latter is, however, on average the largest of the main plant groups (Leitch & Leitch 2013; Pellicer et al. 2019).

6E. Diversification in other Plant and Animal Groups associated with Flowering Plants.

Other Embryophytes. What about the diversification of embryophytes other than angiosperms? It is clear that there has not been a simple replacement of gymnosperms, other vascular plants and "bryophytes" by angiosperms during the Cretaceous-Caenozoic. Mosses and liverworts for the most part seem to have undergone bouts of rapid diversification earlier, but in both there has also been extensive Caenozoic diversification (Cooper et al. 2012; Feldberg et al. 2014; Laenen et al. 2014). Diversification in the speciose pleurocarpous mosses, about 40% of all mosses, seems to have been early-Cretaceous and rapid, with subsequent semi-stasis. Many mosses, especially members of Hypnales, are epiphytic (Shaw et al. 2003b; Newton et al. 2006, 2007; see also Kürschner & Parolly 1999), and their initial radiation is at about the same time as the early rise of the angiosperms during the KTR (Laenen et al. 2014). Porellales, largely leaf-epiphytic liverworts, diverged from the terrestrial Jungermanniales somewhere between the Late Carboniferous to Triassic, but they, too, diversified in the Cretaceous and early Caenozoic (Heinrichs et al. 2007; Feldberg et al. 2014; see also Ahonen et al. 2003; Forrest & Crandall-Stotler 2004; Cooper et al. 2012), and much divergence within liverwort families has been Caenozoic (Cooper et al. 2012). However, Lejeuneaceae initially diversified in the Cretaceous and neither here nor in Cephaloziineae were there rate changes in the Caenozoic (R. Wilson et al. 2007a, b; Feldberg et al. 2013). Laenen et al. (2014) suggest that, as in gymnosperms, there may also have been massive extinction events. Overall, the fossil record of bryophytes is sparse, and many early records are from amber (Tomescu et al. 2018; Heinrichs et al. 2018; Bippus et al. 2019 and references).

Ferns found the low-light environment created by the dominant angiosperms to their liking, as the title of one study, "Ferns diversified in the shadow of the angiosperms" (Schneider et al. 2004) suggests. About one third of all leptosporangiate ferns are epiphytic, and these ca 3,000 species make up about 10% of all epiphytes and the great majority of non-angiosperm vascular epiphytes. Epiphytic ferns commonly grow on angiosperms and prefer humid conditions (see Watkins et al. 2007a, b; Watkins & Cardelús 2012 for their adaptations), and they are major components of the epiphytic vegetation, particularly in the rain forests of the Antipodes and Oceania (Dubuisson et al. 2009). They are commonly crown epiphytes and grow high up in the tree and on branches (see also Lehnert et al. 2017; Lehnert & Krug 2019), and the adoption of the epiphytic habitat was perhaps facilitated by the evolution of a distinctive new photosystem that allowed them to grow in shady conditions, although that may have been acquired considerably earlier around 179 Ma (Kawai et al. 2003; F.-W. Li et al. 2014). Most epiphytic ferns are Polypodiales, and their initial diversification (it happened in several clades) began in the Palaeocene, perhaps around the PETM (Schneider et al. 2004a, b; Schuettpelz 2007; Dubuisson et al. 2009; esp. Schuettpelz & Pryer 2009: Supplemental Tables 2, 3; Watkins et al. 2010), although Watkins and Cardelús (2012: p. 701) talk about a "Cretaceous Pteridophytic stampede into the canopy" and Testo and Sundue (2016) did not find particularly rapid diversification among epiphytic ferns. However, Testo and Sundue (2016) emphasized that although many fern clades are quite old - Cretaceous or even older (Testo & Sundue 2016: Fig. 2, S1, etc.) - their diversification, both on the forest floor and in trees, was mostly a Caenozoic phenomenon.

The epiphytic habit did evolve quite early in a few clades of vascular plants. Trichomanes and relatives (Hymenophyllaceae) diversified in the early Cretaceous, but they are commonly epiphytic on tree ferns, a relatively old clade (Schuettpelz 2007; Hennequin et al. 2008; see also Schuettpelz & Pryer 2009), furthermore, epiphytes in Hymenophyllaceae are often low epiphytes, growing on the trunks and lower branches, perhaps an older habitat that that of crown epiphytes (Lehnert et al. 2017; Lehnert & Krug 2019). About half - 190/380 species - of clubmosses, Lycopodiaceae, are also epiphytic, and their diversification may have begun in the Late Cretaceous (Wikström & Kenrick 1997, 2001; Wikström 2001), however, Testo et al. (2018b) suggest that Phlegmariurus was ancestrally epiphytic (see also Field et al. 2016), perhaps by the middle Jurassic. Botryopteris (Ophioglossaceae) may have been another early epiphyte growing on the extinct marattialean tree fern Psaronius (Rothwell 1991). Bippus et al. (2019: Early Eocene) discuss early "epiphytic" communities which are known largely from osmundaceous rhizomes (see also McLoughlin ∧ Bomfleur 2016: Early Jurassic), while earlier a variety of plants have been found growing/climbing on Psaronius, marattialean tree ferns, from the mid-Carboniferous to Permian (Rößler 2000).

The complex of changes that occurred in angiosperm leaves did not occur in gymnosperm leaves (e.g. de Boer et al. 2012), and gymnosperms may have been at least locally disadvantaged by the temperature changes happening at around the K/C boundary (Blonder et al. 2014). Podocarps with flattened foliage units - i.e., including cladodes - are often shade tolerant and their diversification may have occurred somewhat after the venation density of angiosperm leaves increased, (94-)64(-38) versus 109-60 Ma (Biffin et al. 2011a; Brodribb & Feild 2009; Biffin & Lowe 2011). Diversification within extant genera of both Cycadales and Pinales is quite recent, mid to later Caenozoic (e.g. Oberprieler 2004; Nagalingum et al. 2011; Crisp & Cook 2011; Davis & Schaefer 2011; Leslie et al. 2012; c.f. Salas-Leiva et al. 2013, in part). Extinction may have been higher in gymnosperms than in angiosperms, hence contributing to lower diversity in the former, certainly, gymnosperm clades have long stems and shallow crowns (Crisp & Cook 2011). Brodribb (2011, see also Brodribb et al. 2012) and others have emphasized that conifers, and Pinaceae in particular, an ECM clade, are very successful in high light but other than high-nutrient conditions, some can tolerate extreme cold, and a few species dominate a considerable area of the earth's surface (see below, also Augusto et al. 2014).

Animals. Here I focus on some animal groups that are common in tropical rain forests today.

Ants make up only ca 2% of known insect species, with about 13,000 species described. However, they make up one third of insect biomass, overwhelmingly dominate in samples collected when l.t.r.f. canopies are fogged (86% of arthropod biomass, to 94% of arthropod individuals), their 15% of total animal biomass is greater than that of terrestrial vertebrates, and they are the major consumers of plant resources in the canopy (e.g. Davidson et al. 2003; Rico-Gray & Oliveira 2007; Pie & Tschá 2009; Ward 2014; Barden & Grimaldi 2016). For ants and plants, see Chomicki and Renner (2017b) and papers in Proc. Royal Soc. B 284(1850). 2017.

Ants are are sister to the Apoidea, spheciform wasps + bees (e.g. B. R. Johnson et al. 2013; Ward 2014; Peters et al. 2017a; Branstetter et al. 2017a). Crown-group diversification of ants may have begun 176.4-132.6 Ma (Moreau et al. 2006: depends on calibration used), 143.2-108.6 Ma (Brady et al. 2006), or 158-139 Ma (Moreau & Bell 2013, 2014); ants diverged from Apoidea ca 163 Ma to the end Jurassic ca 145 Ma (Peters et al. 2017a; Branstetter et al. 2017a). One scenario suggests that ants and plants have coexisted for at least 120 Ma (Chomicki & Renner 2015), between 100 and 60 Ma the rise of angiosperm-dominated forests being tracked by ant diversification (Moreau et al. 2006). The ants initially evolved in pretropical forests, moving in to tropical rain forest as it developed (Moreau & Bell 2013; see also E. O. Wilson & Hölldobler 2005). On the other hand, Grimaldi and Engel (2005) date stem ants to only some 120 Ma, the oldest fossil stem-group ants being from the Middle Albian some 105 Ma, and crown-group ants are estimated to be ca 95Ma by LaPolla et al. (2013).

Although the evolution of ants and angiosperms may be connected, the relationships between the two are complicated. Ants may obtain sugars, etc., from extrafloral nectaries, but they are not directly herbivorous, and they also indirectly obtain nutrients from the plant via aphids and scale insects that feed on plants, or via leaf-decomposing fungi in the case of leaf-cutter ants (Peeters et al. 2017). The subfamilies Dolichoderinae, Formicinae, Myrmicinae and Ponerinae include ca 90% of all ants (Ward et al. 2015), and of these Formicinae and Dolichoderinae are the major groups of canopy-dwelling ants in the l.t.r.f. today, while they and Myrmicinae are all much involved in close associations with plants, hemipterans and fungi (Chomicki & Renner 2017b). All authors suggest that extant ant subfamilies and several of the tribes are Cretaceous in age (Moreau & Bell 2013; Ward 2014; Blaimer et al. 2015; Ward et al. 2015). The crown-group ages of Formicinae and Dolichoderinae have been estimated to be late Cretaceous-early Palaeocene in age, 80-70 and 70-60 Ma or (90-)80, 75(-65) and (69-)66.5, 53(-?) respectively (Ward 2014; Moreau & Bell 2013). However, Blaimer at al. (2015) date crown-group Formicinae to (137-)117.5, 104(-88) Ma in the Aptian/Albian, late Lower Cretaceous, rather older than many other estimates, and although there was little divergence for the next 20 Ma, four of the major tribes had diverged by the beginning of the Palaeocene ca 66 Ma (crown-group Camponotini and Melophorini are both early Eocene) and there had been a substantial amount of divergence by the beginning of the Eocene ca 56 Ma. Crown-group Pseudomyrmecinae, a small clade of arboreal ants more or less closely associated with its plant hosts has been estimated to be (78.7-)71.7(-65.7) Ma, but the crown-group age of all Pseudomyrmecinae minus two small clades (= ca 230 species) is only ca 49 Ma (Chomicki et al. 2015), however, the pseudomyrmecine crown-group age in Moreau and Bell (2013) is only ca 40 Ma. Ward et al. (2015) dated crown-group Myrmicinae to (109.5-)98.5(-88) Ma in the early Upper Cretaceous, with the six tribes diverging in the late Cretaceous (Maastrichtian) to early Eocene 71-52.3 Ma, however, the first three of these to diverge have stems of 25-45 Ma, and no fossil myrmicines are known from the Cretaceous.

Recent work is clarifying ant evolution. The extinct Sphecomyrminae, perhaps sister to all other ants, were quite diverse in Kachin amber from Myanmar ca 99 Ma (Barden & Grimaldi 2014). Eusocial ants are known from this amber, although they are all members of extinct basal clades (Barden & Grimaldi 2016). Interestingly, highly specialized staphylinid social parasites of ants and termites have also been found in the same amber (Yamamoto et al. 2016). Brownimecia, known from fossils 94-90 Ma from New Jersey (Engel & Grimaldi 2005 and references), is sister to extant ants (Barden and Grimaldi 2016). Fossil evidence suggests that the ecological dominance of ants began only in the Eocene, 50-35 Ma, modern ants first becoming common in the Eocene fossil amber record (e.g. Grimaldi & Engel 2005; Stadler & Dixon 2005; Dunn et al. 2007; Rust et al. 2010; LaPolla et al. 2013; Ward 2014; c.f. in part Wilson & Holldöbler 2005; Moreau et al. 2006; Rico-Gray & Oliveira 2007), perhaps around the PETM ca 55.8 Ma (Ward 2014). Ants make up less than 1% of the insects in Cretaceous deposits (Zherihin & Eskov 1999), 5-12% in early to mid-Eocene deposits 52-42 Ma, but around 20% in Miocene Dominican amber (Barden & Grimaldi (2016), and to over 40% by the end of the Eocene (Grimaldi & Agosti 2000). However, ants recently found in ca 72.1 Ma amber from Tilin, Myanmar, were not only very numerous, ca 20% of all insects in the amber (n = 34), but they included crown-group ants - three genera of Dolichoderinae and one probable ponerine, while no sphecomyrmines were found (Zheng et al. 2018). This suggests a somewhat earlier diversification of ants...

Extant members of several basal ant clades live largely underground (Rabeling et al. 2008), although details of the relationships of these ants need clarifying (Ward 2014). Many ground-dwelling ants are carnivores, some basal clades eating termites (Bourguignon et al. 2014). The tree-loving Formicinae and Dolichoderinae eat plant materials (Rico-Gray & Oliveira 2007). Ants, fungi, hemipterans and tropical rain forest plants are associated in a variety of ways (e.g. Oliver et al. 2008). For general information about ant-hemipteran-plant interactions, see Huxley and Cutler (1991) and Rico-Gray and Oliveira (2007). Rhizobiales (and other bacteria, including γ-proteobacteria) are endosymbionts of herbivorous ants, overall, bacteria play an important role in the nitrogen metabolism of the insect, for instance, glutamate is converted into essential amino acids (Davidson et al. 2003; Russell et al. 2009). Members of Chaetothyriales, the black yeasts, a poorly-known group of often melanized ascomycetes, are commonly associated with ants, both carton-builders - the fungus is integral to the construction of the carton - and domatium-formers (see below), as well as being commonly found on the cuticle of leaf-cutter attine ants (Vasse et al. 2017). Clades of chaetothyrialean fungi tend to be either carton-formers or domatium dwellers, but they show little geographical signal (Vasse et al. 2017).

There are a variety of important associations between ants and plants, and the main associations are mentioned below; for a still useful general survey, see Davidson and McKey (1993). Plants that are myrmecophytes will be found in groups 3 and 4, although 1 and 2 can be elements of myrmecophytic associations. Note that myrmecophytes are commoner, and the clades they form are larger, in tropical America than in Africa, but Indo-Malesia has the largest number of myrmecophytic genera (Davidson & McKey 1993). Alkthough associations between plants and ants may be close, strict coevolution of the two seems to be at most limited. Thus Davidson and McKey (1993: p. 44, App. 1) noted that Allomerus, all eight species of which are plant-ant specialists, was to be found on seven genera of plants in five families, and this suggested quite extensive host switching; most, but not all, of the examples they gave suggested similar behaviour. Furthermore, several instances were cited in which one ant may have replaced another within a single association, which thus may be quite labile, ants such as Pseudomyrmecinae predictably often being displaced, Crematogaster and Azteca doing the replacing, and these replacements often occurred in resource-rich (for the ant) light-demanding and fast-growing host plants (Davidson & McKey 1993: esp. pp. 45-56).

1. Ants and extrafloral nectaries. Ants can obtain sugar directly from extra-floral nectaries, and the ants protect the plant against herbivores. Metaanalyses suggest that the plant's reproductive (seed) output is increased in such associations (Trager et al. 2017). Thousands of species of plants have such nectaries and they are common in some Pentapetalae, Dioscorea, etc., but not in early angiosperms, gymnosperms or magnoliids, and it is estimated that they have evolved 457 times (Weber & Keeler 2013); extrafloral nectaries are relatively uncommon in annuals (Trager et al. 2017). Clades of plants with extrafloral nectaries have a rate of diversification that is double that of their sister clades that lack them (Marazzi & Sanderson 2010; Weber & Agrawal 2014; c.f. Käfer & Mousset 2014). About a third of woody broad-leaved angiosperms and vines in Panamanian forests have extrafloral nectaries or other ant attractants (Schupp & Feener 1991); extra-floral nectaries provide a rather generalized resource for ants, rarely being monopolized by a single species of ant (e.g. Blüthgen et al. 2004). The earliest such nectaries in the fossil record are found on the lamina of ca 49.5 Ma Prunus fossils from western North America (DeVore & Pigg 2007), while Marazzi and Sanderson (2010) suggest a crown-group age for a clade of Senna with extrafloral nectaries of some 40.8-30.6 Ma. Interestingly, ants, largely the same species as those visiting extrafloral nectaries, were notably common visitors to sap exuding from wounds on Fagaceae in subtropical S.E. China (Staab et al. 2017: planted forests). The composition of sap exuding from such wounds can be quite like that of extrafloral nectaries, and such wound-visiting behaviour may have been a precursory stage in the evolution of full-blown extrafloral nectaries (Staab et al. 2017) - indeed, secretion at regular extrafloral nectaries is stimulated by wounding (Heil 2015). Soil-dwelling ants that visit plants tend to nest close to the plants on which they forage, and nutrients from the ant nests may be taken up by the plants, so providing the latter with additional benefits (Wagner & Nicklen 2010).

2. Ants, hemipterans, and honeydew. Ca 41% of ant genera obtain sugars indirectly from plants by way of honeydew produced by their hemipteran associates that tap the host's phloem, an indirect mutualistic association between ant and plant. Myrmecophilous aphids are commonest in the tropics, although they also serve as "temperate-zone extrafloral nectaries" (Bristow 1991: p. 116; see also Stadler & Dixon 2005). This is a very largely mutualistic association - trophobiosis, the aphids are the trophobionts, and scale insects, Hemiptera-Coccoidea, may also be involved. Honeydew is an important food/energy source for many arboreal ants, and the species of ants involved may dominate their local communities, forming huge colonies, even if the ants themselves tend to be quite small, and they may also occupy disturbed habitats (e.g. E. O. Wilson & Holldöbler 2005; Blüthgen et al. 2004; Oliver et al. 2008). Associations of plants with honeydew-eating ants are more specialised than those with ants that visit extra-floral nectaries, and a single species of ant often dominates (Blüthgen et al. 2004; Rico-Gray & Oliveira 2007). Plants benefit from such associations in about 3/4 of the cases studied, and the differences in herbivore communities in plants with and without these associations can be considerable (Styrsky & Eubanks 2006). The larger ant families commonly involved in hemipteran-honeydew associations are Myrmicinae, Formicinae, and Dolichoderinae, and they are common in other subfamilies in the immediate clade formed by these three groups (Oliver et al. 2008; Styrsky & Eubanks 2006; Chomicki & Renner 2017). For general discussion on the economics of plant-ant-scale insect interactions,including domatium-dwelling scales, see Davidson and McKey (1993). It is interesting that none of the fossil aphid groups known from the Early Cretaceous survived until the Caenozoic (Weigerek et al. 2017).

3. Ants and domatia. Ants, particularly Dolichoderinae (Oliver et al. 2008), along with scale insects, etc., live together in plant domatia the morphology of which varies from quite elaborate to little more than dense, long, erect hairs between which small ants can move easily (Davidson & McKey 1993). These associations are young, dating to the Miocene and later, i.e. within the last (19-)16 Ma, and have evolved ca 160 times in plants (Chomicki & Renner 2015) and in ants ca 40 times (Davidson & McKee 1993). Both the plant and ant clades involved tend to be quite small (Cecropia and Hydnophytum and relatives are among the larger) and the plant clades in particular are scattered phylogenetically; all told around 680(-?1140) species of plants are involved (Chomicki & Renner 2015; see also Ueda et al. 2008: Macaranga). In the New World, some species of Pseudomyrmex have close associations with plants, as in the swollen-thorn acacias, Vachellia spp.; these associations date to (15-)13(-11) Ma, although the crown age for Pseudomyrmex as a whole, most species of which have more generalized associations with plants, is around 36 Ma (Chomicki et al. 2015). Ants living in domatia tend to be quite closely related to honeydew-collecting clades (Oliver et al. 2008) and to clades that have close associations with fungi (Chomicki & Renner 2017b). Davidson & McKey (1993) suggested that successional plants and plants growing in light gaps were associated with more aggressive ants than herbs or shrubs in the understory.

4. Ant gardens. Ant gardens are known from both the America and the Australian-Southeast Asian region. They consist of a mass of plant material and soil all assembled in trees by ants who put the material in the carton of their nests. The ants bring along seeds of epiphytes, including myrmecophilous plants, that grow in these gardens (Davidson 1988: America; Rico-Gray & Oliveira 2007; Orivel & Leroy 2011). Some of these myrmecophilous plants have domatia in which the ants live - examples include Psychotria s.l. (Rubiaceae), Dischidia (Apocynaceae) and Lecanopteris (Polypodiaceae), all Old World (Chomicki et al. 2017a); seven families of plants are recorded from the nests of Amazonian carton-ants (Davison 1988 and references; see also Ule 1902, 1905). Around 13 origins of domatium-bearing ant-garden plants have been documented in the Australo-Malesian region and they have arisen within about the last (9-)6 Ma, although one of the ants involved, Philidris, is estimated to have begun diversifying (20.2-)13.2(-6.2) Ma (Chomicki et al. 2017a). The seeds of non-domatium-bearing ant-garden plants, which include some species of orchids, attract ants which carry them to the gardens, although some may lack obvious elaiosomes (Davidson 1988; Dunn et al. 2007; Morales-Linares et al. 2018), and the ants may later harvest nectar, pearl bodies, etc., from the plants (e.g. Davidson 1988).

5. Ant faeces and ant-associated plants in general. Here there are no obvious features of plant or insect that suggest an association between the two. However, nitrogen from the amino acid and urea-containing faeces of ants foraging on plants (the subjects: Oecophylla smaragdina, coffee) deposited at random on the leaves may get taken up by the plant and translocated within it, and although there was little effect on overall growth, the shoot:root ratio of the plants increased (Pinkalski et al. 2017). Pinkalski et al. (2017: see references) note that the effects of ants in reducing herbivory on plants may be insufficent to explain the improvement in plant performance of plants associated with ants, but this mechanism of nitrogen uptake might close the gap. Potentially, such nitrogen uptake could occur from any ant wandering on the plant, and clearly the generality of the phenomenon needs to be established - it could be very important in tropical nutrient cycling.

6. Ants and seeds. Ants disperse the seeds or fruits (= diaspores below) of many angiosperms other than plants growing in ant gardens. They are attracted by elaiosomes, quite common on small diaspores, seeds in particular (Beattie 1985; Rico-Gray & Oliveira 2007) which vary considerably in their morphological nature and chemistry but are often a source of important nutrients for the ants (e.g. Bresinsky 1963; Kubitzki et al. 2011; Turner & Frederickson 2013). These elaiosomes are eaten by omnivorous or even carnivorous ants that do not eat the diaspores themselves, but discard them on their rubbish piles (c.f. granivorous ants, which eat seeds). The fatty acids in the elaiosomes that attract the ants may mimic those in their animal prey, the elaiosomes being "dead insect analogue[s]" (Carroll & Janzen 1973: p. 235; Hughes et al. 1994). Oleic acid, a very common fatty acid in many insects, is a particularly potent attractant and may be found in seeds that although dispersed by ants in fact have no nutritional value for them, so plants with such seeds are effectively cheaters in this context, dispersal being the result of deceit (Pfeiffer et al. 2010; Turner & Frederickson 2013). Myrmecochory is particularly common in herbs of the ground flora of the E. North American and European forests, shrubs in Brazilian Caatinga, some 1,500 species of shrubs and small trees in Australia that often grow in infertile soils outside the rain forest, and a number of South African species (Sernander 1906; Berg 1975; Giladi 2006; Orians & Milewski 2007; Milewski & Bond 1982; Bond et al. 1991; Lengyel et al. 2009; Leal et al. 2017, and references). Diaspore dispersal over 2 m is uncommon, but distances up to 77 m have been recorded (Gómez & Espadaler 1998). (Gastropods may also be involved in the distribution of diaspores with elaiosomes - Türke et al. 2011.) Overall, myrmecochory is commonest in smaller plants, shrubs/trees less than 5 m tall and perennials (Giladi 2006; Leal et al. 2015). Germination and plant establishment may be facilitated in diaspores deposited in ant rubbish heaps, and if the diaspores are buried they may be protected from seed predators; Beattie (1985), Giladi (2006) and others discuss these and other possible benefits of myrmecochory to the plant.

A very conservative estimate is that some 11,500 species of angiosperms in 77 families are myrmecochorous and the trait has evolved 100, or even 140 or more times (Lengyel et al. 2009, 2010). Thus caruncles have evolved ca 13 times in Euphorbia alone (Horn et al. 2012). Myrmecochorous clades have about twice as many species as their non-myrmecochorous sister clades (Lengyel et al. 2009), and adoption of myrmecochory in clades like Polygalaceae-Polygaleae seems to be linked to their diversification (Rico-Gray & Oliveira 2007; Forest et al. 2007b; Lengyel et al. 2009, 2010, see also Fokuhl 2008). Myrmecochory may have first evolved when ants themselves became abundant, measured as the time when they make a substantial component of the insect fauna trapped in amber, that is, in the Eocene (Dunn et al. 2007 and references); one estimated early date for the evolution of myrmecochory (in Proteaceae) is (58-)44.5(-32 Ma (Dunn et al. 2007; Lamont & He 2012). Interestingly, perhaps half the species of stick insects (Phasmatodea) lay eggs that mimic the seeds of myrmecochorous plants (seed capitula mimic elaiosomes), and such eggs are first known from the Late Eocene (Hughes & Westoby 1992; Sellick 1997).

7. Leafcutter ants and fungi. The ecologically very important leaf-cutter attine ants (Atta, Acromyrmex: ca 48 species) are restricted to the New World tropics. The leaves they harvest are the substrate for the fungus Leucocoprinus gongylophorus (Agaricaceae: for its biogeography, see Mueller et al. 2017) that they cultivate. The fungus no longer produces spores, and its swollen modified hyphal tips, gongylidia, are fed to the ant larvae (but see Abril 2011 for an alternative interpretation); when the ants eat the fungus, they do not digest many of the enzymes, rather, they are in faecal fluid the ants deposit on the chewed-up leaves the ants are adding to the fungus garden (De Fine Licht et al. 2014). Van Bael et al. (2009) found that leaf-cutter ants seemed to dislike plants with numerous endophytes, indeed, this whole system is very complex, N-fixing bacteria, esp. Klebsiella (Pinto-Tómas et al. 2009), and the necrotrophic ascomycete Escovopsis weberi (Hypocreales) also being integral parts of the whole; in at least some cases, the ant associates seem to be unable to break down cellulose (?or lignin) which is removed by the ants as refuse and on which Leucocoprinus grows (Abril 2011). Escovopsis parasitises Leucocoprinus (see also de Man et al. 2016), but it is kept in check by antibiotics produced by actinobacteria growing on the cuticle of the ants (e.g. Beattie & Highes 2002). There have been extensive genomic changes in leaf-cutter ants, and they have lost various genes involved in nutrient acquisition, including the arginine biosynthesis pathway (Nygaard et al. 2016), while E. weberi has lost genes involved in plant degradation (de Man et al. 2016), and so on; there have also been extensive changes in the abdominal microbiota, the main increase in abundance being linked with the movement of attine ants to the Central/North American continent ca 20 Ma where the higher attine ants then evolved - this was after the domestication of the gongylidia (Sapountzis et al. 2018). Fungus-farming attines initially formed associations with coral fungi, Pterulaceae, and in particular Lepiotaceae, although L. gongylophorus is also cultivated by an isolated lineage of attine ants in a clade that otherwise forms associations with other Lepiotaceae or coral fungi - but like them, it grows on arthropod frass (Schultz et al. 2015). Stem and crown-group ages of leaf-cutter farmers are estimated to be only (16-)13, 9(-7) Ma and (14-)11, 8(-6) Ma respectively (Schultz & Brady 2008), the latter age is ca 18 Ma in Branstetter et al. (2017b), but the beginning of the association between agaricalean fungi and ants, with the fungus initially growing on detritus or arthropod frass, has been dated to ca 61 (stem) to 57 (crown) Ma (Branstetter et al. 2017b) or ca 55 Ma (Schultz et al. 2015; see also Jordal & Cognato 2012; Nygaard et al. 2016; de Man et al. 2016: Trichoderma sister), and obligate crop status of Leucocoprinus has been dated to 25-20 Ma (De Fine Licht et al. 2014). For chaetothyrialean fungi and attine ants, relationships between which are poorly understood, see Vasse et al. (2017).

Close ant-plant relationships are largely phenomena of the lowland tropics, and are less evident at higher altitudes and latitudes. Thus in Central America, ants maintain their abundance until about 1,000 m altitude, the numbers of species peaking at lower altitudes, while above 2,500 m there were very few species of ants at all, although those species that were there might have quite large numbers of individuals (Longino et al. 2014). In line with this altitudinal zonation of ants, extrafloral nectaries become less frequent with increasing altitude and latitude (Rico-Gray & Oliveira 2007), thus extrafloral nectaries in Andean centropogonid lobelioids, found on the outside of the inferior ovary, occur mostly in species growing at lower altitudes where ants are to be found (Stein 1992). In the New World those species of Cecropia growing above 1,500 m altitude are not associated with ants (Gutiérez-Valencia et al. 2017). In the Malesia-W. Pacific area the association between ants and myrmecophilous species of Psychotria s.l. (= Hydnophytinae) breaks down at higher altitudes (Chomicki & Renner 2017a) while in Papua New Guinea the ant-plant associations became more generalized and interconnected, and there seemed to be less in the way of benefits for the plant, at higher altitudes (Plowman et al. 2017).

Some other important insect-plant associations involve ants more indirectly. Lycaenidae and Riodinidae, sister taxa that split around 110-70 Ma - there are some 5,200 lycaenids, about a quarter of all butterflies, and ca 1,500 riodinids - are important here. Their caterpillars have single-celled pore cupola organs, perhaps involved in reducing ant aggression (de Vries et al. 1986; Pierce et al. 2002). Many caterpillars of lycaenids (blues, hairstreaks: Lepidoptera) are herbivorous but they also produce amino acids and sugars prized by ants, indeed, around 75% of lycaenids are associated with ants, as are 20% of the riodinids. 40% of host plant records of these caterpillars are from Fabaceae, and over 90% of these are myrmecophilous; about 5% of lycaenids are more or less carnivorous, eating larvae of the ants with which they are associated (see e.g. Cottrell 1984; Fiedler 1991, 1995, 2001, 2006; Lohman & Samarita 2009; Kaliszewska et al. 2015), or homoptera or homopteran honeydew, for example, the ca 190 species of Miletinae. Ant associations evolved at the base of the lycaenids ca 78 Ma, and twice more recently in the riodinids (Espeland et al. 2018). The age of the clade of lycaenids whose members are commonly associated with ants has also been dated to around 71.7 Ma (Wahlberg et al. 2013), the [Lycaenidae + Riodinidae] clade being somewhat over 95 or (110-)90(-70) Ma (Espeland et al. 2015, 2018). Lycaenid larvae are often found on plants that have notably high foliar nitrogen, especially high amino acid concentrations, and so are often found on plants like (

Fabaceae) that fix nitrogen, but they are also common on Santalales, especially Loranthaceae, and limited evidence suggests that some Santalales have high nitrogen concentrations. About 18% of ant genera have some species that are associated with lycaenids, but these are particularly common in just some genera - thus 178 lycaeinid species associate with Crematogaster, a myrmecine, 95 of the associations being obligate, and the corresponding figures for Campanotus, a formicine, are 90/15, and for Anonychomyrma, a dolichoderine, are 18/16. Interestingly, although both Lasius and Formica are quite commonly associated with plants in the North Temperate region, only 5 of the 86 associations recorded are obligate (Fiedler 2001).

Hemipterans include Auchenorrhyncha, various kinds of leaf hoppers (spittle bugs are the larvae of some), and Sternorrhyncha, in turn including scale insects and mealy bugs (Coccoidea: >8,000 spp.), whiteflies (Aleyrodoidea: >1,500 spp.) and aphids (Aphidoidea: >4,400 spp.), all told around 14,000 species of sap-sucking insects (Iluz 2011). Hemiptera in general have associations with γ-proteobacteria, thus almost all aphids harbour Buchnera aphidicola, an obligate bacterial endosymbiont involved i.a. in synthesizing amino acids for the aphid that it cannot get from the plant (Iluz 2011; Meseguer et al. 2017 and Manzano-Marín et al. 2018: also other obligate bacterial associations in hemipterans; Gomez-Polo et al. 2017). Interactions can become very complex and changes in the bacterial symbiont may be quite common, however, these do not seem to affect the ecology of the aphid (e.g. Meseguer et al. 2017). Soft scales (Coccidae ca 1,140 species) have recently been found to be associated with ascomycetes, members of two families of Hypocreales that otherwise parasitize insects (Gomez-Polo et al. 2017).

Termites are derived from cockroaches, relationships being [Cryptocercis [a cockroach] [Mastotermes [other termites]]] (Evangelista et al. 2019: Zootermopsis switched with Mastotermes in some analyses). Crown-group termites are perhaps around 149 Ma (Bourguignon et al. 2014: mitochondrial genomes), although ages in Evangelista et al. (2019) are rather younger - [Cryptocercis + termites] (=Tutricablattae, termites with child care) are 134±13.5 MA, Neoisoptera 53±13 Ma. Termites are very important globally in plant decomposition, basal clades having protozoa, spirochaetes, etc., in their guts that can break down lignins (Sugimoto et al. 2000; König & Dröge 2011; Bignell et al. 2011; Ni & Tokuda 2013 and references). The speciose and primitively soil-eating Termitidae (crown-group age ca 54 Ma, but c.f. above) become common about the same time as ants, common predators of termites, in the Caenozoic (Bourguignon et al. 2014). Old World Macrotermitinae (Termitidae) are often described as "cultivating" lignin-decomposing fungi, i.e. the agaricalean Termitomyces, but details of the obligate relationship between the two are complex and there is no simple coevolution, furthermore, bacteria that can decompose cellulose and xylose have also been found in the guts of these termites (e.g. Aanen et al. 2007; Nobre & Aanen 2012). Fungus-growing termites are estimated to be 34-24 Ma or thereabouts, with a recently described fossil 26-24 Ma (Roberts et al. 2016; see also Jordal & Cognato 2012). Note that termites with social castes are known from amber ca 99 Ma from Myanmar (Engel et al. 2016), and there are tiny aleocharine staphylinid beetles of about this age that are likely to have been social parasites either on termites or on ants (Yamamoto et al. 2016).

Beetles.McKenna et al. (2015) found that most beetle families started diverging in the Jurassic, and of the later-diverging groups associated with angiosperms, the Chrysomeloidea and Scarabaeoidea began diversifying in the very early Cretaceous. Agaric fungi, also associated with angiosperms, show high diversification rates in the Cretaceous, although these has begun to increase in the Jurassic (Varga et al. 2019). Phytophaga, the phytophagous beetle sister taxa [weevils (Curculionoidea) + leaf beetles (Chrysomeloidea)] include well over 125,000 species, about half of all herbivorous insects. It has been suggested that they may have diversified largely in parallel with angiosperms (Farrell 1998), although initially diversifying on gymnosperms in the Jurassic (e.g. Labandeira et al. 1994; Farrell 1998; McKenna et al. 2009; c.f. Rainford & Mayhew 2015). Ages for Phytophaga are (169.5)161.5(-155.5) Ma, of Chrysomeloidea (159.5-)145(-124.5) Ma, and of Curculionoidea (160.5-)149.5(-138.5) Ma (McKenna et al. 2015) - Hunt et al. (2007) suggested that the latter were ca 22 Ma older. Chrysomelidae, leaf beetles s. str., may diversify (86-)79–73(-63) Ma in the Late Cretaceous-Eocene, especially in the early Caenozoic (Gómez-Zurita et al. 2007; also Winkler & Mitter 2008). However, an association between phytophagy, whether of gymnosperms or angiosperms, and diversification has been questioned (Hunt et al. 2007), although here, as elsewhere, dating is critical - in general the ages in Hunt et al. (2007) are (7-)22-46(-59) Ma older than those in McKenna et al. (2015) and with no or little overlap. Thus Smith and Marcot (2015) suggest that an uptick in family numbers of polyphaga in the mid-Cretaceous, contemporaneous with angiosperm diversification, may rather be due to the first appearance of beetle-bearing amber from that period, while in a careful analysis that utilized twice the number of sister-group comparisons in Mitter et al. (1988), Rainford and Mayhew (2015) found no evidence that plant-feeding clades were more species-rich than their sister taxa with different diet preferences. There may have been a major diversification of herbivorous beetles in particular and insects in general around the PETM (Farrell 1998; Wilf & Labandeira 1999; Wilf et al. 2001; Lopez-Vaamonde et al. 2006).

Within Scarabaeoidea the Pleurosticti and Glaphyridae are the main plant-eating groups, both including clades that eat flowers/pollen in particular (Cetoniinae in the former), and Pleurosticti are dated to (142-)128-109(-96) Myb.p. (there is a 40-50 Ma stem) and Glaphyridae to ca 141 or 101 Myb.p., with the flower-eating clades originating 79-62 Ma (Ahrens et al. 2014). Under this scenario, there was diversification of the plant-eating groups ca 100 Myb.p., although more in the Palaeogene; feeding on dinosaur dung by scarabs seems unlikely. However, Gunter et al. (2016) suggested that early scarabs did eat dinosaur dung, and when dinosaurs started to eat more angiosperms, this indirectly helped facilitate the switch of scarabs to eating angiosperms. Under this scenario, pleurosticts are much older, some (177.5-)158-142.5(-101.5, 91) Ma, and there was an increase in diversification ca 100 Myb.p., and perhaps some decrease (it depends on the group) across the K/P boundary (Gunter et al. 2016).

Citerne et al. (2010) thought that 93.5-89 Ma in the Turonian was a period of floral innovation and evolution of pollinators, while Cardinal and Danforth (2013) suggested that there is link between the diversification of the eudicots and that of bees (Anthophila). Peters et al. (2017a) date the stem group age of bees to (147-)124, 111(-93) Ma, the beginning of the KTR, while Sann et al. (2018) date stem-group Anthophila at (148-)128(-108) Ma. Certainly, bee diversity in the earlier part of the Cretaceous was low (e.g. Grimaldi & Engel 2005). Estimates of when bees began to diversify range from (132-)123(-113) Ma (Cardinal & Danforth 2013), ca 125 Ma (Ronquist et al. 2012), and to ca 112 Ma (Grimaldi 1999), with families diverging by the beginning of the Caenozoic; most diversification occurred within the last 100 Ma (see also Engel 2000; Grimaldi & Engel 2005). The crown-group age of corbiculate bees, many eusocial, is around 62 Ma, i.e. early Palaeocene (Peters et al. 2017a). The plesiomorphic condition for pollination specificity in bees seems to be oligolecty, that is, bees initially pollinated one or a few species of plants, all more or less related (Sipes & Tepedino 2005; Danforth et al. 2006; Sipes et al. 2006; Larkin et al. 2008; Praz et al. 2008; Michez et al. 2008; Litman et al. 2011 and references: note early ages for bee diversification; also Sedivy et al. 2013; c.f. e.g. Moldenke 1979). Several species of oligolectic bees may pollinate a single plant species, and the floral morphology of the latter is likely to be rather unspecialized (see also below). Floral specialization has increased over evolutionary time, and unspecialised flowers, initially probably pollinated by pollinators other than bees, precede specialised flowers, probably pollinated by one or a few species of polylectic pollinators, pollinators that visit a variety of unrelated flowers. As mentioned above, there is no signal of pollinator type in pollen protein content, etc. (Roulston et al. 2000). Buzz pollination (sonication) and collection of oils are two very distinctive bee behaviours. The former has evolved ca 45 times (and been lost even more) in bees, probably within the last (55-)39.2(-25) Ma, although there is weak support for the common ancestor of Anthophila as a whole being a buzz pollinator, and around 22,000 or more species of plants are pollinated in this way (Cardinal et al. 2018). In a number of cases bumble bees may obtain pollen from the one species, even from the same flower, by both buzzing/sonication and scrabbling pollen directly from the anthers, what happens in any particular case depending on how much pollen is readily accessible; this behaviour may represent an early stage of the evolution of more obligate sonication bee-plant relationships (Raine & Chittka 2007b; Russell et al. 2017: see Table 1).

Lepidoptera, butterflies and moths, with ca 160,000 described and perhaps 500,000 total species, are the biggest insect clade almost totally dependent on plants both as adults and as larvae (Powell et al. 1998; Mutanen et al. 2010). Adults of species in the basalmost clades (non-ditrysian lepidoptera) are for the most part small (Hepialidae are an exception), and although members of the two most basal clades, both small, have jaws, probosces are otherwise the norm (Regier et al. 2015; Mitter et al. 2016). Caterpillars of these extant basal lepidopteran clades are plesiomorphically mostly internal feeders (e.g., leaf miners, stem borers), and apart from most members of the basalmost Micropterigidae and Agathiphagidae (Mitter et al. 2016), they eat mostly woody superrosid angiosperms - certainly magnoliids, members of the ANA grade and monocots do not figure prominently in their diet, while fungivory and detritivory is scattered (Regier et al. 2015: Fig. 10). Relationships between some of the main clades are poorly supported, perhaps reflecting very fast initial evolution (Mutanen et al. 2010). The basal Ditrysia are also microlepidoptera, and within this clade Tineoidea s.l. have recently been found to be paraphyletic; all five of the basal ditrysian clades are fungivores and detritivores, although one is also phytophagous (Regier et al. 2014; Mitter et al. 2016). Most other Ditrysia are herbivores, clades like Gracillariidae being mostly internal feeders (Kawahara et al. 2016), perhaps a derived condition (Regier et al. 2013). Diversification of the non-ditrysian lepidoptera may have begun in the Jurassic and that of the ditrysians in the Cretaceous (Labandeira et al. 1997; Wahlberg et al. 2013), or both may be dated to the Cretaceous (Grimaldi 1999; Grimaldi & Engel 2005); stem-group ages of Lepidoptera range from under 200 to about 300 Ma (Wiens et al. 2015). Wahlberg et al. (2013) estimated the start of diversification of the speciose Macroheterocera group of moths, with ca 100,000 spp., at about 90 Ma. However, Regier et al (2015) thought that the very different estimates for both lepidopteran and angiosperm diversification made it difficult to think about mutual connections in their evolution.

Papilionoidea (butterflies) are well embedded in monotrysian lepidoptera (e.g. Mitter et al. 2016). Heikkilä et al. (2011) suggested that the main clades (= families) diverged quickly in the early Cretaceous, Wahlberg et al. (2009) thought that these clades were largely of late Cretaceous origin while Wahlberg et al (2013) dated initial butterfly diversification to ca 104 Ma - eudicots radiated ca 100 Ma and lepidoptera in general shortly after ca 90 Ma. Estimates for diversification within clades representing extant subfamilies are after (e.g. Vane-Wright 2004; Wahlberg et al. 2009; Wheat et al. 2007; Heikkilä et al. 2011), more or less at (Simonsen et al. 2011), or before (Michel et al. 2008; Pohl et al. 2009: 113-84 Ma, gene duplications) the K/P boundary. Overall, much diversification of of Nymphalidae-Nymphalinae and -Papilioninae seems to have occurred 65-33 Ma (Wahlberg 2006; Zakharov et al. 2004). Although diversification of Pieridae may have begun in the Late Cretaceous (112-)95(-82) Ma (Braby et al. 2006), much speciation was Caenozoic (Simonsen et al. 2011 gives a range of divergence times). Caterpillars of these groups tend to show rather high food-plant specificity. Some butterfly (and other herbivore) clades that survived the K/P boundary may initially have eaten several different food plants, but subsequently they diversified on a more restricted set of plants and/or they shifted their food preferences (Janz et al. 2006; Nylin & Wahlberg 2008; Fordyce 2010; Nylin et al. 2014; but c.f. in part Hamm & Fordyce 2015).

Placental mammals have a substantial fossil history before the Cretaceous, and their crown-group age is around 77.8-76.5 Ma (e.g. dos Reis et al. 2014), with notable diversification through the end-Cretaceous and the early Caenozoic, and these early mammals were probably insectivorous (Bininda-Emonds et al. 2007; see also Stadler 2011a, esp. L. Liu et al. 2017), generalists persisted through the K/P boundary, although there may have been a shift to frugivory soon after (Grossnickle & Newham 2016). A radiation of multituberculate mammals, now extinct, had begun ca 85 Ma, and their adoption of a more herbivorous diet may be associated with the increasing prominence of angiosperms (G. P. Wilson et al. 2012). Primates initially may have been arboreal omnivores eating plants and their associated insects, but even in the Palaeocene ca 46% were probably frugivores, the figure rising to ca 76% in the Holocene, and the relationship between plants and primates has been described as a "facultative mutualism" (Goméz & Verdú 2012: p. ; Sussman et al. 2013). Although radiation of extant primates began in the late Cretaceous, e.g. 71-63 Ma (Springer et al. 2012), (89.8-)84.8, 78.8(-64.9) Ma (Fabre et al. 2008) or (98.6-)87.2(-75.9) Ma (Perelman et al. 2011), nearly all diversification is Caenozoic. "Old" estimates date the split within Anthropoidea/Simiiformes to the later Eocene ca 43.4 Ma, crown-group ages of Old World and New World monkeys (Catarrhini and Platyrrhini) being around 31.6 and 24.8 Ma respectively (Perelman et al. 2011). Recent discoveries in Amazonian Peru of fossil teeth ca 36 Ma similar to those of African anthropoids (Bond et al. 2015) are consistent with these dates and with the movement of (ancestral) platyrrhines from the Old to the New World.

In bats, both frugivory and nectarivory have arisen in parallel, even within New World bats, and some combination of insectivory with these modes of nutrition is common (Datzmann et al. 2010; Rojas et al. 2011). Crown-group diversification of phyllostomid bats occurred between 43.1 and 33.4 Ma, probably in the late Eocene, with diversification of the more specialized fruit-, pollen- and nectar-eating bats dated to around 26-16 Ma in the late Oligocene to mid-Miocene (Datzmann et al. 2010; Rojas et al. 2011; K. E. Jones et al. 2005; Teeling et al. 2005). However, only a few species of species of bats are involved, for instance, there are only ca 15 species of nectar-eating Old World pteropodids (Datzmann et al. 2010: see also below).

Birds. Jetz et al. (2012) and Jarvis et al. (2014: skeleton tree) provide phylogenies for all birds; crown-group Aves are estimated to be 115.9-94.8 Ma (Lee et al. 2014). By 62.5-62.2 Ma, 4 million years after the K/P extinction event, several major neoavian clades had diverged, and a fossil found then, a stem mousebird (Coliiformes) from New Mexico, probably ate fruits or seeds (Ksepa et al. 2017: crown mousebirds are African!). Radiation of important seed-dispersing birds such as Columbiformes (pigeons) began some (63.6-)54.4(-46.1) Ma (95% CI) in the earlier Caenozoic (e.g. Tiffney 1986b; Pereira et al. 2007; Jarvis et al. 2014), but extant Columbidae are a mere 33.4 Ma (Jetz et al. 2012). Bird pollination is also likely to be a Caenozoic phenomenon, three groups, Trochilidae (hummingbirds), Nectariniidae (sunbirds, etc.) and Meliphagidae (honey eaters) being most important (Cronk & Ojeda 2008). Hummingbirds diversified only in the Pliocene (Bleiweiss 1998a; McGuire et al. 2007, 2014), while that part of Passeriformes, the songbirds, that includes sunbirds and honey eaters did not begin to diversify until the Oligocene around 30 Ma (Jarvis et al. 2014; c.f. estimate in Friis et al. 2011, up to 65 Ma) or a little later (Moyle et al. 2016: Pardalotus diverged ca 21 Ma). Meliphagids are basal oscines, and Selvatti et al. (2015) suggest that they diverged from the rest ca 35 Ma and speciated/radiated ca 27 Ma (Early and Late Oligocene respectively), but in Moyle et al. (2016) the crown-group age of the group is a mere 11 Ma or so. Passerida initially diversified 26-20 Ma, Zosteropidae and [Dicaeidae + Nectarinidae] - Promerops is in this part of the tree - are all less than (31-)27.6, 27.1(-23.1) Ma (Selvatti et al. 2015).

Parrots (Psittaciformes) are important in both pollination and seed dispersal, but very differing ages have been suggested for their diversification. Wright et al. (2008) offer two sets of ages based on different geological calibrations. Older crown-group ages are around 82 Ma, with the split of the cockatoos from the rest of the family bar the few Strigopoidea being dated to around (82.9-)74.1, 70(-63.9) Ma; corresponding younger ages are around 50 Ma and (51.4-)45, 42.7(-38) Ma respectively (Wright et al. 2008). The younger ages are consistent with the fossil record and the age for Psittaciformes suggested by Jarvis et al. (2014); no fossil parrots from the European Palaeogene, with rather rich records, can be placed in extant crown groups (e.g. Mayr 2002, 2009). Loriinae, mostly Australasian, are nectarivorous, nectarivory having evolved at least three times, the major nectarivorous clade, ca 53 species of lories and lorikeets, evolved within (15-)13 Ma (Schweizer et al. 2014, 2015).

6F. Discussion. We can now return to the issue of when l.t.r.f. as we know it appeared. The question is, to what extent can "the ecological opportunities provided by humid megathermal forests" (Feldberg et al. 2014: p. 1) that seem to have driven the diversification of epiphytic ferns, etc., in the Late Cretaceous translate to the existence of l.t.r.f., particularly a l.t.r.f. that might look something like that of today and have a broadly similar complexity of ecological relationships? What were the interactions between animals and plants in tropical forests in the Late Cretaceous and Palaeogene?; what birds, butterflies and bees were involved in pollination?; what vertebrates were involved in fruit dispersal?; how important were ants? Is there any direct connection between the diversification of animal groups that are now more or less directly dependent on plants for food, etc., and of the plants themselves? Dating plant and animal clades is critical here, so the problems with dating that have frequently been mentioned should be borne in mind.

Tropical rain forest today is characterized by having an annual rainfall of at least 1800 mm, at most three months having less than 100 mm per month. There is little variation in the temperature, the mean for the coldest month being around 18o, and there is (almost) no frost (e.g. Morley 2000; Burnham & Johnson 2004). Much woody angiosperm diversification has been in such forests, and epiphytes and lianas (but see below) are common there. Estimates of the numbers of tree species (10< cm d.b.h.) in tropical forests in general, i.e., including drier forests, are between around 40,000 and 53,000, both the Neotropics and the area from India to the Pacific having at least 19,000-25,000 species, but Africa has fewer than a quarter of these (Slik et al. 2015); for the Amazon alone, there are an estimated 16,000 species (ter Steege et al. 2016, but c.f. Cardoso et al. 2017: ca 6,727 a more likely number), even if at the more local level diversities in all three areas can be similar (Ricklefs & He 2016). The most diverse families of woody plants in neotropical rain forests are Fabaceae (by far), Moraceae, Annonaceae, Euphorbiaceae, Malvaceae, Lauraceae, Sapotaceae and Myristicaceae (Burnham & Johnson 2004: from Fig. 2); for a global look at tree diversity, see Beech et al. (2017: note definition of tree).

Saxifragales, although now with very few species in the tropics, may represent an ancient and rapid radiation (Fishbein et al. 2001; Fishbein & Soltis 2004; Jian et al. 2008). Molecular estimates of the diversification times of the major rosid clades are around (114-)108-91(-85) Ma, and of Fabidae and Malvidae very soon after, (113-)107-83(-76) Ma (H. Wang et al. 2009). The origins of several clades within Malpighiales and Ericales, major components of today's lowland tropical rain forest, are to be pegged to the Mid Cretaceous or slightly later. Initial malpighialean diversification was rapid, and relationships have been hard to disentangle (Wurdack & Davis 2009; but c.f. Xi et al. 2012), many clades originated some time in the late Aptian/Albian, (119.4-)113.8(-110.7)/(105.9-)101.6(-101.1) Ma (Davis et al. 2005a: high and low estimates). Palms diversified 102-98 Ma (see Eiserhardt et al. 2017 - spread rather greater), 124-101 Ma are some ages for crown-group Menispermaceae (Jacques et al. 2011; W. Wang et al. 2016b), and all this might suggest a mid-Cretaceous origin of l.t.r.f. (Eiserhardt et al. 2017 and references).

Areas of current very high diversity like western Amazonia are wet, less seasonal, relatively nutrient-rich, and have a mosaic of ecological conditions - and they are young, less than 20 Ma, indeed, many neotropical clades characteristic of l.t.r.f. are relatively young (Burnham & Johnson 2004; Hoorn et al. 2010).

Epiphytes. See Zotz (2016) for an overview. Around 24,750 species of flowering plants, 9% of the total, are epiphytes (Feild et al. 2009a; Boyce et al. 2009, 2010; Boyce & Lee 2010; Boyce & Leslie 2012; Zotz 2013), and they are common in l.t.r.f. (see Wagner et al. 2015 for possible host specificity). About 70% are Orchidaceae-Epidendroideae (Ramírez et al. 2007; Gustafsson et al. 2010; Conran et al. 2009), around 7% are Bromeliaceae (Givnish et al. 2008a, 2014a), and add Gesneriaceae (570-700 spp.) and Ericaceae (630 spp.), which together account for ca 5%, and the great majority of epiphytes are included. Bromeliaceae are all New World and Oligocene or even Pliocene in age, while highly speciose and commonly epiphytic-CAM crown Epidendroideae include around 19,560 species (figures from Pridgeon et al. 2005, 2009, 2014), i.e., about two thirds of the species in the whole family, and they may have diverged from each other only 37.9-30.8 Mya (Givnish et al. 2015: c.f. stem age of subfamily!). Epiphytic angiosperms commonly have small seeds (of the families just mentioned, Ericaceae have the largest), and in Orchidaceae, at least, association with a fungus is needed for germination (Eriksson & Kainulainen 2011). Ca 3,000 species of ferns are another major element of the epiphytic flora (Schuettpelz & Pryer 2009). Knowledge of epiphytic assemblages is rather rudimentary (Mendieta-Leiva & Zotz 2015).

CAM-type photosynthesis is particularly prevalent in clades that either grow in arid terrestrial environments or are epiphytes; succulent epiphytes are also quite often CAM-type plants. All told, some 17,000-18,000 or more species especially in Crassulaceae, Bromeliaceae, Cactaceae and Orchidaceae-Epidendroideae, all succulents of one sort or another, have CAM or its variants (Winter & Smith 1996b; Sayed 2001). The origin of CAM clades is largely contemporaneous with that of C4 clades, being Miocene and younger, indeed, there seems to have been a "global surge" of succulent CAM plant diversification within the last 10 Ma like that of C4 grasslands (Edwards & Ogburn 2012: p. 726), and overall, diversification rates are moderate to high in CAM clades (Ferrer et al. 2014). Succulence of some form, whether of root, stem or leaf, occurs in some 690 genera and 12,500 species (Nyffeler & Eggli 2010b; see also von Willert et al. 1990; Eggli & Nyffeler 2009) and is associated with CAM photosynthesis. Succulents include species which either avoid drought, although they are rarely found in the driest conditions, or are salt tolerant - usually mutually exclusive strategies (Ogburn & Edwards 2010). They include many C4 chenopods, and also CAM Cactaceae, Crassulaceae, and epiphytes, perhaps particularly orchids (10,000+ species). The venation density of the leaves of succulent plants tends to be low (Sack & Scoffoni 2013). Overall, both CAM-type photosynthesis and succulence are in part connected with tropical rain forest,

Lianas. For general information, see Schenck (1892), Putz and Mooney (1991), Rowe et al. (2004) and Schnitzer et al. (2015) and references. Lianas may make up some 25% (10-44%) of both stem density and species richness of woody plants in tropical forests, and they are especially prominent in drier and disturbed forests (e.g. Schnitzer 2005, 2015a) and in forests over limestone soils (Manzané-Pinzon et al. 2018). Climbing Arecaceae, mostly Calamoideae, are a feature of wetter forests (Couvreur et al. 2015) and Menispermaceae are also quite diverse in l.t.r.f. (R. Ortiz-Gentry pers. comm.). Stem-twining vines seem to be commoner in old forests, tendril vines in younger forests (Schnitzer & Bongers 2002). Lianas have an important effect on carbon cycling (increased), storage (reduced) and sequestration (reduced) since liana stems are thinner and the vessels are closely packed, overall wood density being low, and the carbon the plant sequesters goes mainly into leaves, and these break down fast (van der Heijden et al. 2013, 2015; Durán & Sánchez-Azofeifa 2015; Isnard & Feild 2015; Ewers et al. 2015). Indeed, although the biomass of liana stems is often less than 10% of the total, that of liana leaves may be up to 40% (Putz 1984; Isnard & Feild 2015). Lianas have often rather negative effects on the diversity, reproduction and regeneration of l.t.r.f., particularly of canopy trees, where they may be relatively abundant (Campbell et al. 2018), less on palms and understorey plants, an effect exacerbated in the New World by their increasing abundance, perhaps caused by climate change (Schnitzer et al. 2005; Schnitzer 2015b; García León et al. 2017: lianas themselves as food resource for animals not taken into account), although this increasing abundance is less evident in perhumid forests (J. R. Smith et al. 2017). Indeed, they can be thought of almost as ruderals doing best in early successional vegetation with their thin, metabolically inexpensive stems and throw-away leaves (Ewers et al. 2015: lianes as structural parasites; Campbell et al. 2018). There may be some connection between lianas and the phylogeny of the plants on which they grow, probably mediated by such things as bark characteristics - but in communities like savannas these features have evolved in parallel (Zulqarnian et al. 2016: lianas and mature forests?).

Lianas are important components of seasonally dry tropical forests today, where their higher investment in hydraulic efficiency and lower investment in support when compared with trees pays off (Dias et al. 2019). Liana stems are poorly insulated and water in the very wide and long vessels can freeze easily, with disastrous consequences for the plant - some of the reasons that liana abundance decreases rapidly north of the Tropic of Capricorn, ca 23.5o N (Schnitzer 2005: exceptions interesting).

Burnham (2009, 2015) noted that there were few lianas through most of the Mesozoic, but a number in the Palaeozoic, and they reappeared in some numbers in the fossil record only with the evolution of angiosperms. Today, ca 8,700 species of scandent plants of one sort or another are recorded from the New World alone (Gentry 1991), for a total of perhaps double that number, and they are found in 162 families of seed plants (Scnitzer et al. 2015 and references). Liane woods from the Cretaceous-Palaeogene were dominated by Menispermaceae (S. Y. Smith et al. 2013a); crown-group Menispermaceae are around 109.1-106.3 Ma and there was a burst of diversification close to the K/P boundary (W. Wang et al. 2012). About 400 species of Menispermaceae are vines/lianas, and they are common in l.t.r.f. today. There are around 800 species of lianas in Vitaceae-Vitoideae, the crown-group age of which is ca 91 Ma (Smith et al. 2013; Wen et al. 2013). There are ca 400 species of Bignoniaceae, especially Bignonieae and largely New World (ca 50 My), ca 535 species of Arecaceae, most in Calamoideae-Calaminae and largely Old World (maximum age 48-40 Ma, or early Eocene to Miocene), Sapindaceae-Sapindoideae-Paullinieae (?age), ca 360 species of Celastraceae (?Cretaceous, especially Africa: Bacon et al 2015), and perhaps 400 species of Malpighiaceae lianas (Gentry 1991), largely New World (Late Cretaceous, ca 75 Ma, or younger). Dioscoreaceae include ca 630 species, most being lianas or vines; diversification there may have begun up to 80 Ma, although it may have begun in crown-group Dioscorea only some 48.3 Ma (Viruel et al. 2015), while about 625 species of Passifloraceae-Passifloreae are vines/lianas of one sort or another (estimate derived from Feuillet & McDougal 2007; see also references in Schnitzer et al. 2015). There are ca 190 species of vines/lianas in the [Smilacaceae [Philesiaceae + Rhipogonaceae]] clade, and diversification in Smilax, which includes most of these plants, began at the end-Eocene ca 40 Ma (Qi et al. 2012). In South East Asia-Malesia around 600 species of Old World Piper are climbers/lianas (age very uncertain, see Piperaceae), there are ca 500 species of Annonaceae in the palaeotropics that are lianas (Couvreur et al. 2015), and Combretaceae and Loganiaceae are also important components of the liana community there (Addo-Fordjour & Rahmad 2015; also other papers in Parthasaranthy 2015). Although nearly all the 975 species of Cucurbitaceae are lianas or vines, many are relatively small and herbaceous and grow outside l.t.r.f..

Parasites and Mycoheterotrophs. Clades of parasitic and mycoheterotrophic angiosperms that currently live in l.t.r.f. or similar conditions are of particular interest. Stem-group Rafflesiaceae, holoparasites now largely restricted to l.t.r.f., are estimated to have diverged from other Malpighiales some 95 Ma, divergence within the family beginning (95.9-)81.7(-69.5) Ma (Bendiksby et al. 2010). Naumann et al. (2013, q.v. for discussion) estimate the stem age of Rafflesiaceae to be ca 65.3 Ma, around the K/C boundary, although stem-group Tetrastigma (Vitaceae), on which Rafflesiaceae are parasitic, is estimated to be only some (68-)57, 51(-36.4) Ma (P. Chen et al. 2011b; Lu et al. 2013). Fruits of crown-group Vitaceae are known from the Deccan Traps at around or a little before the K/C boundary ca 66 Ma (Manchester et al. 2013). Mycoheterotrophic clades of Dioscoreales, Burmannia, etc., Thismia, often grow in similar habitats, and some are estimated to have diverged (118-)109-79(-68) Ma, the mycoheterotrophic habit being established well before the beginning of the Palaeocene (Merckx et al. 2008a, 2010). The stem age of the mycoheterotrophic Gentianaceae-Voyrieae is (65.2-)54.0, 46.8(-40.1) Ma (Merckx et al. 2013). Mycoheterotrophy has probably evolved over fourty times (Jacquemyn & Merckx 2019), but most origins are younger. Eupolypod ferns diversified in the late Cretaceous-early Caenozoic (H.-M. Liu et al. 2014 and references).

Characteristic of today's l.t.r.f. is the diversity of various animal and fungal groups directly dependent on plants that live there, whether mycorrhizae, endophytes, herbivores, gallers, frugivores, decomposers, or pollinators. The connection between animal and plant can also be indirect: animals eat other animals. That the diversification of orb-weaving spiders, insect-eating bats (vespertilionids), stinging wasps, etc., was more or less contemporaneous with that of angiosperms is because they were eating insects, many of which were eating plants (see also Hawkins & Porter 2003; Penney 2004; J. S. Wilson et al. 2012b). Thus Pompilidae, wasps which now almost exclusively eat spiders, radiated in the Caenozoic after a Mid/Late Cretaceous origin (J. S. Wilson et al. 2012b), the speciose carnivorous Carabidae-Harpalinae (19,800 or so spp.) showing a similar pattern after originating in the Cretaceous Aptian/Albian ca 115 Ma and then doing nothing for ca 32 Ma (Ober & Heider 2010). However, the spider-eating archaeid spiders began diversifying around 200 Ma (Wood et al. 2015). Diversification of parasitic and hyperparasitic wasps such as the hyperdiverse Chalcidoidea (the chalcid wasps), perhaps with half a million species (only ca 22,500 described) and most ultimately dependent on flowering plants, is most intensive in the Caenozoic (Heraty et al. 2013), although they began diversifying back around the Early Cretaceous (208-)129(-89) Ma (Peters et al. 2017b: fuse (65-)45(-25) My). Similarly, the fleshy fruit "niche" was exploited by particular groups of flies, particularly by Drosophilinae, and the relationships between particular fruits and flies may be very close (Ashburner 1998: alcohol dehydrogenase in flies; Harry et al. 1996, 1998: fig-breeding Lissocephala). The sugar-rich fruits of angiosperms may have provided a habitat for budding yeasts such as Saccharomyces cerevisiae; the latter have a genome duplication ca 100 Ma that is perhaps connected with their ability to exploit this habitat (Wolfe & Shields 1997; Conant & Wolfe 2007), much evolution here is also Caenozoic (Guzmán et al. 2013).

In the following paragraphs emphasis is placed on direct relationships. Plant-feeding insects make up at least one quarter of all described species, and maybe 26% coleoptera (Weins et al. 2015; c.f. Janz et al. 2006; Farrell 1998; Hunt et al. 2007), although quite how many insect species, or arthropods in general, there are is unclear. Ølgaard (2000; see also Basset et al. 2012: insects; Hamilton et al. 2013; Stork et al. 2015) estimated that there were around (2.75-)4.9-6.7(-10) million species, down from the some 30,000,000 estimate based on the early tree-fogging experiments.

There are well over 100,000 species of extant phytophagous beetles in some five clades, particularly the chrysomelids and curculionids, that eat angiosperms. Initial diversification has been dated to the Jurassic, with over 100 extant beetle lineages diverging before the beginning of the Cretaceous 140 Ma (Farrell 1998, but c.f. dates; see also Mayhew 2007; Hunt et al. 2007). Beetle and angiosperm evolution seem not to be tightly linked, and there is no strong association between diversification and the adoption of herbivory, or shifts from gymnosperms to angiosperms as a food source (Hunt et al. 2007; Weins et al. 2015; see also Rainford & Mayhew 2015). Diversification may have begun first on monocots and then moved on to broad-leaved angiosperms (Reid 2000). About two thirds of herbivorous beetles eat only one or a few species of angiosperms, i.e. they are are mono- or oligophagous. Herbivorous beetles and herbivory in particular and insects in general increased with the warming trend of late Palaeocene-Eocene (Farrell 1998; Wilf & Labandeira 1999; Labandeira et al. 2002b; Wilf et al. 2001; Lopez-Vaamonde et al. 2006, Wilf 2008).

There is no simple connection between the diversification of plants and the insects associated with them. Kergoat et al. (2005b) suggested that diversification of bruchids and Fabaceae may have occurred more or less together. More commonly, bouts of insect diversification have occurred (well) after the appropriate angiosperm host clades originated, particularly in herbivores (implicit in Futuyma 1983; see Funk et al. 1995; Berenbaum 2001; Percy et al. 2004; Lopez-Vaamonde et al. 2006: Phyllonorycter, leaf-mining Gracillaridae; Winkler & Mitter 2008; McKenna et al. 2009; Janz 2011; Cruaud et al. 2012b; Kergoat et al. 2015). Diversification and overall diversity of phytophagous insect groups may at least initially increase after they adopt new hosts (Janz et al. 2006; c.f. Hamm & Fordyce 2015). However, close co-evolution/cospeciation seems to be the exception rather than the rule, and is most evident in shallow rather than deep clades (Berenbaum & Passoa 1999 for references; c.f. Farrell & Mitter 1998); a looser relationship may be more common (see Futuyma & Mitter 1996). Patterns of relationships between plants and herbivores/gallers may differ from those between plants and pollinators. The complexity of such relationships is evident in fig wasps, which are both gallers and pollinators. Crown group Ficus may be (101.9-)74.9(-60.0) Ma and the age of the fig wasps, at (94.9-)75.1(-56.2) Ma, is similar (Cruaud et al. 2012b), although at most of the nodes within these groups the wasps seemed to be older than the figs. Around 30% of figs are pollinated by more than one species of wasp, and Yang et al. (2015) found that in dioecious species, where the figure may be over 40%, the wasps were sister species, and there cospeciation was possible, however, in monoecious figs around two thirds of the co-pollinators were not sister species. Indeed, Satler et al. (2019) found that in neotropical strangler figs host switching predominated, cospeciation being no more than expected by chance, and that there was "[s]trong evidence for host specificity in ecological time combined with strong evidence for host switching in evolutionary time" (ibid., p. 2305). Plant clades with "specialized" pollinators may diversify despite there being little if any associated evolution of their major pollinators, which are themselves not notably diverse in the first place (see below).

Seed-dispersing animals and the plants they dispersed may have diversified roughly in parallel (see below: e.g. Tiffney 1984, 2004; Wing & Tiffney 1987; Collinson & Hooker 1991; Dilcher 2000; c.f. Herrera 1989; Eriksson et al. 2000a; Moles et al. 2005a). Vertebrates are important in the dispersal of the rather large propagules of rain forest trees and the pollination of the rather widely dispersed individuals that produce them (e.g. Regal 1977); mammals in particular are also herbivorous. However, the acquisition of fleshy fruits is not linked to notable increases in diversification of clades with them (Bolmgren & Eriksson 2010 and literature; c.f. Eriksson & Bremer 1991).

Returning to the issue of the evolution of angiosperm-dominated l.t.r.f.. The "museum" hypothesis is that current centers of diversity are centers of origin, they are areas in which there has been little disturbance over the last 50-100 Ma (Stebbins 1974). Initial rapid diversification may have slowed down ("ancient cradle"), or diversification rates may have increased towards the present, the "recent cradle" theory (see Couvreur et al. 2011c for references), recent diversification patterns in palms fitting this model (see Eiserhardt et al. 2017 for a critque of such ideas).

There have of course been forests in the tropics earlier, as in Carboniferous era, but they were species poor with around 120 species in areas up to 105 km2 and they lacked much in the way of epiphytes, even if lianas were quite common, and so are not comparable with modern l.t.r.f. (Cleal et al. 2012). Morley (2007) plots the extent of tropical forests through the Caenozoic. A biota from the late Early Eocene from Messel, Germany and about 48 Ma is notable both for its floristic richness and the diversity of interactions betweren plants and what appear to be quite specialized herbivores (Dunne et al. 2014). Kapgate (2013) suggests that there was rain forest in central India in the Palaeocene-Eocene - >2,000 mm rainfall/yr, prolonged rainy season - and large trunks from the Mandla-Dindori area have been identified to extant species, similar vegetation perhaps being as old as the Upper Cretaceous (Chandra & Bande 1992 in Kapgate 2013: Fig. 3).

Some clades now prominent in l.t.r.f. seem to have been little affected by events at the K/P boundary. Thus the diversification rate for palms was more or less constant for some 65 Ma after their origin ca 100 Ma (Couvreur et al. 2011b, esp. c). Annonaceae (Couvreur et al. 2011a; see also Erkens et al. 2012) and Araceae (Nauheimer et al. 2012), and the liverworts Lejeunaceae (R. Wilson et al. 2007) and Cephaloziineae (Felberg et al. 2012) show similar constant rates of diversification (but c.f. Marshall 2017: rather unlikely?). Many major clades in Malpighiales, conspicuous small trees of today's l.t.r.f. are Cretaceous in age (Davis et al. 2005a; see also Xi et al. 2012b). The presence of palms has been used to suggest that there were rain forests in mid-Cretaceous Laurasia ca 100 Ma, although this suggestion was tempered by noting that rain forest assembly had probably been a gradual process (Couvreur et al. 2011c).

Palms in particular are almost iconic plants of l.t.r.f. today, so either 1) the rain forest of 100 Ma was rather different from that of today (see e.g. Feild et al. 2011b), 2) l.t.r.f. of "modern" aspect remained very restricted in extent for millions of years, 3) not all palms have the same ecological preferences, and/or 4) there are methodological problems with the analyses (see e.g. Quental & Marshall 2010). Coiffard and Gomez (2009) thought that early palms may have been swamp plants like living basal Arecales (their examples were Calamus, Nypa, and Mauritia). Most palm fossils in the New World may have been from plants growing in seasonally dry tropical forest (Burnham & Johnson 2004). Indeed, no palm fossils with a single wide metaxylem element in their stem vascular bundles are known from the Cretaceous, and there is an association between extant palms with two-vessel bundles and climates with a dry period; Arecoideae have a single vessel, and they are conspicuously l.t.r.f. plants (Thomas & Boura 2015). Couvreur et al. (2011c) did acknowledge the absence of Cretaceous fossils that might support their hypothesis, and they also noted that Coryphoideae, more frequent outside l.t.r.f., were Palaeocene in age.

Late Cretaceous megathermal forests from western Africa ca 72 Ma are supposed to provide "[t]he first evidence for typical closed, multi-stratal forest synusiae" (Morley 2007: p. 6). However, little recent work has been carried out on this flora, from which relatively very large seeds/fruits/infructescences 1.2 (Annonaceae seeds)-10 cm long have been reported (Chesters 1955; Monteillet & Lappartient 1981). Large seeds imply large plants (Morley 2007), but there I have seen no estimates either of the size of the plants in these forests or of the venation densities of their leaves.

Both Kapgate (2013) and Srivastava (2011) suggest that there were quite widespread humid tropical forests on India at the K/P boundary when the Deccan Traps were being laid down - India was ca 15o S at the time. Woods of genera like Eucalyptus, Ailanthus, Canarium, Aglaia, Garcinia and Sterculia are reported from intertrappean deposits, also bamboos and seeds of Artabotrys, but these last records are over 150 years old (Bande et al. 1988; Bande 1992 and references). Wheeler et al. (2017) recently reevaluated as much of this fossil record that was available, and they found that the woods are characterized i.a. by the prevalence of simple perforation plates, rarity of exclusively solitary vessels, and absence of growth rings, suggesting overall warm and rather dry conditions and the absence of a monsoon. India had started separating from Gondwana ca 120 Ma, its movement north speeding up ca 80 Ma - it had separated from Madagascar ca 85 Ma (some estimates to 96 Ma) and remained isolated for the next 30-50 Ma - and it collided with Eurasia ca 40 Ma (Datta-Roy & Karanth 2008; Jagoutz et al. 2015: the Kohistan-Ladakh Arc had collided with India ca 10 Ma before). The flora from the Traps contains some supposed Gondwanan elements like the African Turraeanthus (Meliaceae, = ? - Wheeler et al. 2017 ) as well as Clusiocarpus, Eucalyptus (= Myrtoideae - Wheeler et al. 2017) and Cyclanthodendron (perhaps a palm: Srivastava 2011); overall palms were common, but there were no dipterocarps (Srivastava 2011). Interestingly, a 52-50 Ma amber fauna from India showed little evidence of Indian insularity (Rust et al. 2010), while vertebrate taxa, for example, were pretty cosmopolitan and showed little endemism, and distinctive plants like Viracarpon are more widely distributed than had been thought (Matsunaga et al. 2018).

There is no evidence for l.t.r.f. in the New World Cretaceous (Burnham & Johnson 2004, but c.f. Eiserhardt et al. 2017 and above for references suggesting a mid-Cretaceous origin). There may have been l.t.r.f.-like forests in the increasingly moist climates that were developing then (see also Schönenberger 2005; Wing et al. 2012; etc.), but wind-dispered fruits, conspicuous in the canopy, etc., of current neotropical rain forests, were at best very uncommon (Herrera et al. 2014b). However, the Castle Rock flora, from the early Palaeocene in Montana and ca 64.5 Ma, is described as "an excellent example of early modern tropical rain forest in North America" (Burnham & Johnson 2004: p. 1607), and l.t.r.f. may have first developed around the K/P boundary (Eiserhardt et al. 2017 for references). A later Palaeocene flora from Colombia ca 59 Ma had a familial composition similar to that of current neotropical rain forest, including Arecaceae, Araceae, Fabaceae, Malvaceae, Menispermaceae, Lauraceae and Zingiberales, even if overall both plant (especially beta diversity) and herbivore diversity were rather low (Jaramillo et al. 2006; Graham 2010: general vegetational history of Latin America). The venation density in this flora (Wing et al. 2009; see also Burnham & Johnson 2004) is very high, and the flora is the first fossil evidence of functional equatorial neotropical megathermal rain forest, i.e. l.t.r.f. (Feild et al. 2011b; see also Jud & Wing 2013; Ricklefs & Renner 2012). Dunne et al. (2014) found a "modern" trophic structure in rich forests in Europe in the early Eocene, but exactly when this first appeared and what it implies in the context of animal-plant relationships is unclear. The first fossil record of l.t.r.f. from Asia is from the early Eocene 52-50 Ma of western India (Rust et al. 2010).

During the early Caenozoic, angiosperms with their often dense veinlet reticulum, high leaf specific conductivity, and small and dense stomata (Watkins et al. 2010; Feild et al. 2011b; de Boer et al. 2012; Feild & Brodribb 2013), in turn associated with high rates of photosynthesis and transpiration, may have facilitated the rise to dominance and spread of widespread closed tropical forest with reliably high rainfall (e.g. Boyce et al. 2008, 2009, 2010; Boyce & Lee 2010). Features like vessel length and perforation plate morphology (scalariform perforation plates became less common) that would enhance xylem conductivity, and wood parenchyma, but not ray morphology, changed across the K/P boundary (Wheeler & Baas 1991). Such changes may be connected with changes in venation density and water conductance needs, but their interpretation is not easy (see Wheeler & Baas 1993; Philippe et al. 2008; Lens et al. 2016: c.f. Sambucus and Adoxa). Fast decomposition of angiosperm litter, particularly associated with the deciduous habit and with core eudicots (Knoll & James 1987; LeRoy et al. 2019), may also have speeded nutrient cycling and plant growth (Cornwell et al. 2008; Berendse & Scheffer 2009; Liu et al. 2014). Jaramillo et al. (2010 and references) note that high levels of CO2 and high levels of soil moisture improve the plant performance performance when temperatures are high, as earlier in the Caenozoic, perhaps also another element in angiosperm success.

Meredith et al. (2011: p. 523; c.f. O'Leary et al. 2013) observed that the KTR of 125-80 Ma, during which angiosperms increased from 0 to 80% of the vegetation in many parts of the world, was "a key event in the diversification of mammals and birds".

If "modern" or even "early modern" l.t.r.f. is dated to around the middle part or end of the Palaeocene some 60 Ma, one expectation might be that the distinctive fauna that currently inhabits it is of similar age. However, in the ca 58 Ma Cerrejón flora, although insect damage was quite high, most was from external feeding, and specialized feeding types like leaf mines and galls were uncommon (Wing et al. 2009). As we will see, our understanding of ecological inter-relationships in early l.t.r.f. is not very good.

In extant l.t.r.f. there are often numbers of closely related species (= species in the same genus!) growing together (c.f. Turner et al. 2013), although both the extent of this phenomenon and its history are problematic. Tropical forests as we think of them, the "modern archetypal tropical rain forest" of Burnham and Johnson (2004: Table 1), with multi-layered vegetation, lianas, epiphytic ferns, bromeliads (New World) and orchids, the trees in particular having large seeds and relatively large leaf blades with dense venation and entire margins, seem to be a Caenozoic (Palaeocene-Eocene) phenomenon (Upchurch & Wolf 1987; Wing 1987; Eriksson et al. 2000a; Morley 2000: general account; Pennington et al. 2006a; Schuettpelz 2006; Boyce et al. 2009: Schuettpelz & Pryer 2009; Crane & Carvell 2007: the early Caenozoic fossil record; Burnham 2009: climbers; Watkins et al. 2010).

Many angiosperm groups have diversified since the Cretaceous, including speciose clades that often include many herbaceous and/or shrubby taxa (e.g. Tiffney 1985a, b????). Magallón et al. (1999) noted that major core eudicot clades like Fabaceae (19,000+ species: e.g. Bruneau et al. 2008b; Bello et al. 2009) and (most of) Lamiales that together represent about 45% of core eudicot diversity appear only in the upper Cretaceous (Maastrichtian) and Caenozoic. Woody Fabaceae are prominent in today's tropical forests, while core Lamiales with monosymmetric flowers include many species pollinated by recently-evolved generalist pollinators like birds and corbiculate bees. Diversification of Asteraceae (23,000+ species: K.-J. Kim et al. 2005; Funk et al. 2009c for a summary) is also Caenozoic in age (but c.f. Barreda et al. 2015)

Diversity in families that one might thinks of as being characteristic of l.t.r.f. may be quite recent. Thus the modern Malesian flora has been dated to ca 23.6 Ma (Morley 2007), while. Diversification in a largely epiphytic clade in Orchidaceae-Epidendroideae that contains about 19,560 species, about 2/3 of the whole family, can be dated to within the last 37.9-30.8 Ma (Givnish et al. 2015). Even in old clades like Myristicaceae, crown-group diversification is also Caenozoic at some 21-15 Ma, and there is little molecular divergence between its extant members - all very recent indeed given possible stem ages for the family that are over 100 Ma, and this extensive stem-group history is largely unknown (Sauquet et al. 2003; J. A. Doyle et al. 2004, 2008; Scharaschkin & Doyle 2005; Richardson et al. 2004; but c.f. in part Couvreur et al. 2011a, c). Meliaceae are characteristic of extant l.t.r.f., but their ancestors may have been a deciduous tree of seasonal or montane habitats. Crown-group ages of the speciose rain forest clades in the family are a mere 23 Ma (Late Oligocene/Early Miocene), although their stem-group ages are Eocene, and Keunen et al. (2015) suggested that they may have been quite species-rich in pre-Late Oligocene times, but extinction then obscured this early history. Indeed, a number of plants (and animals) have very different past and current distributions are (e.g. Nypa (Arecaceae, Roridula (Roridulaceae), Leea (Vitaceae and Phytocreneae (Icacinaceae), so the composition of past l.t.r.f. will be hard to predict.

Next few paragraphs little more than notes.... Around 53.5% of vascular plants are herbs, which would suggest that there are about 132,000 species of herbaceous flowering plants (Fitzjohn et al. 2014; Ølgaard 2000: 65,000 species). The high assimilation rates facilitated by high venation density may, in conjunction with a shortening of the life cycle, particularly the pre-fertilization gametophytic stage (e.g. J. H. Williams 2012b), etc., have enabled the evolution of annual herbs (Boyce & Leslie 2012). This life form is exceedingly uncommon in other vascular plants, fossil or extant (Boyce & Zwieniecki 2012). Maybe --- species of flowering plants are annuals.

Palaeocene-Eocene diversification in l.t.r.f. took place in a warmer world with flatter temperature gradients and with the continents in rather different places than they are now. Boreotropical migration and extensive long distance dispersal were important in shaping the assembly of tropical forest biomes (Thomas et al. 2015 and references; Moncrieff et al. 2016 for biomes and history). The role of India in "seeding" tropical diversity is unclear. Early Eocene dipterocarp resins have been found there, but its biota seems not to indicate particular isolation (Rust 2010; see also above). The beginning of diversification in Piper subgenus Ottonia, a small clade with species in both the Amazonian and Atlantic rain forests in South America, may be dated to some 55.3 Ma (Molina-Henao et al. 2016), around about the PETM, although fossil records of the genus from Maastrichtian Colombia that could be associated with the Schilleria clade of Piper led to estimates of the stem age of New World Piper of (106-)93, 69(-67) Ma or even more (Martínez et al. 2012, esp. 2014: note topology). Much diversification in the genus is Oligocene in age or younger, thus the Macrostachys and Radula clades, which between then include around half the 1,300 neotropical species of the genus, began diversifying around ca 45 Ma (Martínez et al. 2014).

Diversity may have been high in the early Eocene l.t.r.f., although what plant-animal associations looked like then is unclear. Major diversification of birds, both pollinators (hummingbirds, various passeriform groups) and fruit-dispersers (pigeons and parrots) (e.g. et al. ; Jarvis et al. 2014; see also Wright et al. 2008 for parrots), the successfull generalist orchid and bumble bees, fruit-eating mammals, especially fruit-eating and nectarivorous bats and fruit eating primates, and arboreal and seed-eating and -dispersing ants (Ward 2014), was in or after the Oligocene ca 40 Ma. Conspicuous elements of today's vegetation like grasslands and savanna are still younger, most being less than (15-)10 Ma old.

Increased plant productivity and diversity would allow animals that ate, pollinated (e.g. birds, butterflies, bees) or dispersed (e.g. ants, mammals, birds) angiosperms to diversify (see also Boyce et al. 2010). Thus Novotny et al. (2010) noted that congeneric rather than confamilial, etc., plant species tended to share herbivore species, so as phylogenetic diversity increased, so would the number of herbivore species.

Exactly how, when and where plants and animals that now show "mutualistic" relationships evolve? Our understanding of the ecological-evolutionary connections between animals, in particular insects, and plants remains unclear (e.g. Futuyma 1983; Janz 2011); there is no simple underlying theory to explain the variety of the interactions. We know little about the details of long-term evolutionary-ecological interactions of plants and the organisms associated with them (see above; Fine et al. 2004 for habitat specialization and herbivore activity in the Amazon, also Janzen 1974a). Furthermore, the sheer complexity of the matrix of defensive compounds inside the plant make simple explanations of the evolutionary dynamics of plant-insect herbivore relationships in particular difficult. As Berenbaum and Zangerl (2008: p. 806) note, idiosyncracy is central to the nature of chemical co-evolution, a problem that is only exacerbated by the difficulty of understanding ecological relationships over time. Yet, as Grimaldi and Engel (2005: p. 625) note, "Despite the fact that the mechanism is obscure as to how insects diversified with angiosperms, the overall patterns are extremely clear that the angiosperm radiations had a profound impact on insects, and vice versa."

Paragraph out of place. Light conditions in the closed forest habitat pose a particular challenge. The PHYA/C gene pair duplicated before the origin of crown-group angiosperms, and PHYA in particular may have been very important in angiosperm evolution. It is intimately involved in germination and in etiolation responses of the seedling, especially in shady conditions such as occur on the forest floor, for example, preventing seedling etiolation in response the the far-red light that dominates the spectrum there. Furthermore, it is involved in the germination response of hydrated seeds to very brief pulses of light (the very low fluence response - VLFR) such as sunflecks (Mathews et al. 2003). Unfortunately, how ANA grade angiosperms might respond to manipulation of their PHY genes is unknown.

7. Flowers and Pollination.

Here I discuss some aspects of angiosperm reproduction in a rather conventional way, i.e. as flowers (and fruits) being a key to understanding angiosperm diversification. I outline the general distribution of particular flower "types" with respect to phylogeny. For details about the interaction of particular groups of pollinators with the flowers they pollinate, see below.

7A. Flowers, Pollination and Fertilization. Most narratives of angiosperm evolution focus on flowers and fruits and their influence on speciation. The flower is considered a key innovation - or, better, a group of innovations - and the success of angiosperms has been attributed in considerable part to the evolution of flowers (e.g. Eriksson & Bremer 1992; Dilcher 2000). Burger (1981) saw insect pollination as a key to the diversification of angiosperms, insects being able to find isolated plants in small populations, and angiosperms were able to subdivide the environment effectively. Frame (2003) emphasized flexibility in construction of the flowers, the speed of the reproductive cycle, the closure of the carpels, and the fact that flowers are edible hence attracting potential pollinators as contributing to the sucess of angiosperms. Monosymmetry in particular has often been thought of as a key innovation (e.g. Donoghue et al. 1998; Neal et al. 1998; Endress 2001; Sargent 2004: see below). Pollination, especially by insects, but also bats and other mammals as well as birds, and seed dispersal, especially by mammals and birds, may interact; both can increase outcrossing, gene flow in general, and, given a heterogeneous environment, speciation (e.g. Kay et al. 2006; Kay & Sargent 2009).

Adoption of syncarpy was an important evolutionary event (e.g. Armbruster et al. 2002 and references; Friis et al. 2006b). It has evolved seventeen or more times, and a compitum, which allows pollen tubes from one stigma to fertilize the ovules in more than one carpel, evolved in three quarters of these cases (Armbruster et al. 2002). A style/compitum allows competition among male gametophytes (Mulcahy 1979; Endress 1980a); note that there are some cases, perhaps most notably in Loranthaceae, of competition between female gametophytes where the embryo sac becomes as tubular and invasive as a pollen tube (Bachelier & Firedman 2011 and references. Several/many pollen grains can potentially fertilize a single ovule, and since over 60% of the genes expressed in the sporophyte are also expressed in the gametophyte (Erbar 2003 and references), gametophytic competition can directly benefit the sporophyte. Syncarpy also allows "excess" pollen landing on one stigms to "switch" and pollinate ovules normally served by pollen landing on another stigma (Carr & Carr 1961; Armbruster et al. 2002 and references), incidentally reducing pollen tube competition. Interestingly, many basal angiosperms, often apocarpous, have rather few ovules per carpel, but the advantages of syncarpy are more evident if there are numerous ovules per carpel, however, sometimes several pollen grains per stigma are needed for the pollen to germinate in the first case, and the amount of pollen arriving is marginal for ensuring full seed set, a compitum of some sort might be generally useful (Armbruster et al. 2002). Indeed, many members of the ANA grade, for example, do have an extragynoecial compitum (e.g. E. G. Williams et al. 1993; Lyew et al. 2007; J. H. Williams 2009 - see also Igersheim & Endress 1997; Endress & Igersheim 2000; X.-F. Wang et al. 2011). An extragynoecial compitum allows pollen tubes to switch between carpels, as do compita developed at the tip of otherwise free styles, as in Apocynaceae-Asclepiadoideae, some Malvaceae, etc. (Armbruster et al. 2002). For pollen tube competition, whatever the morphology of the gynoecium, see also Erbar (2003).

Closed carpels also protect the ovules and often become much elaborated as the seeds mature, so promoting dispersal (Armbruster et al. 2002 and references). The evolution of the carpel may have allowed the parental sporophyte to control both fertilization and the allocation of resources to the seed (Lord & Westoby 2012 and references), although complexities are introduced by the varying proportions of maternal and paternal contributions in different endosperm types. For discussion about the rate of pollen tube growth in angiosperms and those gymnnosperms that have pollen tubes, and also the importance of tissues like endosperm and perisperm in seed development in angiosperms, see above.

7B. Major Clades With Monosymmetric Flowers. Successful animal pollination entails that the pollinator follows a more or less complex and specific set of cues. Early angiosperms seem to have had what would be described as rather generalized flowers (but see below). A number of large clades (2,000+ species each) can be characterised by specialized floral features that seem likely to affect diversification/speciation. Five of these major clades, Orchidaceae, Zingiberales, core Lamiales (= [Gesneriaceae + The Rest]), Fabaceae, and Asteraceae, have a preponderance of members with monosymmetric flowers, although reversion to polysymmetric flowers has also occurred within these clades, most notably in Fabaceae-the mimosoid clade. Together with some rather smaller clades, e.g. Campanulaceae-Lobelioideae, Caprifoliaceae, some Iridaceae, they include almost one third of all angiosperms. Diversification in many clades with monosymmetric flowers seems to be greater than that in their sister taxa with polysymmetric flowers (Sargent 2004; Kay & Sargent 2009; c.f. Kay et al. 2006), perhaps because pollinator fidelity is increased. However, using current ideas of sister taxa, putting examples into a broader phylogenetic context (as in Sanderson & Wojciechowski 1996), and correcting for what might be called "apomorphy lag", the fact that the evolution of the apomorphy of interest does not necessarily happen when the clade of interest diverges from its sister taxon (Käfer & Mousset 2014), such stories may well become more ambiguous.

Diversification as a possible result of the acquisition of monosymmetry can be studied at much finer evolutionary scales. For instance, Stebbins (1974) suggested that monosymmetry had evolved more than 25 times within angiosperm families, Westerkamp and Claßen-Bockhoff (2007) noted that it was found in 38 families. In fact, monosymmetry has evolved hundreds of times (see also Endress & Matthews 2006a; Endress 2008a; Jabbour et al. 2009, esp. 2014), the precise number of origins depending critically on definitions/character circumscriptions and implicit or explicit optimizations (e.g. Endress 1997a; Donoghue et al. 1998; Reeves et al. 2003; Cubas 2004; Jabbour et al. 2008; J. Zhang et al. 2017); for the evolution of monosymmetry in asterids, see also Ree and Donoghue (1999) and Donoghue and Ree (2000). Reyes et al. (2015, 2016) found over 200 changes in symmetry in angiosperms, of which up to perhaps 1/3 were reversals from mono- to polysymmetry; 148 origins was the figure in Reyes et al. (2018). Thus there have been 10-18 or more origins of monosymmetry in Proteaceae alone, with 4-12 reversals (Reyes et al. 2016; Citerne et al. 2016); the Cape species Mimetes cucullatus has monosymmetric groups of flowers. Monosymmetry in Fabaceae is almost protean in its expression, monosymmetric papilionoid flowers in particular having arisen five times or so (e.g. Cardooso et al. 2012a; Legume Phylogeny Working Group 2017; Bukhari et al. 2017).

Bukhari et al. (2017) found that monosymmetry in Pentapetalae with normal floral orientation was sometimes associated with the odd petal being abaxial, but in an almost equal number of examples the odd petal was adaxial, the flower being obliquely oriented (from the point of view of the pollinator it is inverted), while in the much less common situation when the flower was inverted early in development, the plane of monosymmetry of the monosymmetric flower at maturity remained inverted. In most monocots and eudicots examined CYC or CYC-like genes are involved, and in eudicots they are expressed mostly on the adaxial side of the flower although sometimes affecting the abaxial stamen pair (e.g. Song et al. 2009; Hileman & Cubas 2009; Hileman 2014b: Fig. 4, c.f. definition of monosymmetry and Fig 2a). Connected with monosymmetry is the horizontal orientation of the flower, and horizontally-held flowers increase the precision of pollen deposition on the pollinator even when the flowers are polysymmetric (Robertson 1888a-c; Fenster et al. 2009); polysymmetric peloric flowers tend to be held verically. In gloxinia (Sinningia speciosa) at least, both monosymmetry and floral orientation are under pleiotropic control of a single gene, SsCYC, and so polysymmetric, vertically-held and monosymmetric, horizontally-held flowers switch (Dong et al. 2018). There is extreme parallelism, and even within Asteraceae different CYC2c paralogs are involved in the expression of the monosymmetric (ray flower) phenotype (Chapman et al. 2012). However, in the magnoliid Aristolochia CYC genes seem not to be so involved in the development of monosymmetry (Horn et al. 2014), while recent work suggests that CYC genes are also little involved in the development of monosymmetry in flowers of Orchidaceae (Madrigal et al. 2017).

From a structural point of view, many flowers are monosymmetric at some stage of their development (Endress 2008a, also 1999, 2001b; see also Characters), and monosymmetry, as in pentapetalous angiosperms, may have very different developmental pathways (Bukhari et al. 2016, 2017). Melastomataceae are particularly difficult to categorise, some species having strongly monosymmetric flowers, others polysymmetric, while yet others are intermediate, and this is why they are not included in the list below; Solanaceae (e.g. J. Zhang et al. 2017) are not much better. Many highly reduced flowers are monosymmetric, not only in Poaceae (see below), but also in the speciose Piperaceae, etc.. On the other hand, many inflorescences are functionally the equivalent of polysymmetric or haplomorphic flowers, the often strongly monosymmetric peripheral flowers in Asteraceae, some Brassicaceae (e.g. Busch & Zachgo 2007), Adoxaceae and Apiaceae being the visual equivalent of petals to the pollinator. Inflorescences may also be the functional equivalent of a single monosymmetric flower, as in Pedilanthus (= Euphorbia subg. Euphorbia sect. Crepidaria), etc., while the polysymmetric flowers of Iridaceae may be the functional equivalent of a small inflorescence with three monosymmetric flowers (e.g. Guo 2015b). The bottom line is that as more becomes known about floral morphology and development, a clear definition of monosymmetry becomes elusive, and, like disseminule morphology, the definition will in part depend on the question being asked.

Taxa. There are five particularly large clades with monosymmetric flowers. 1. Orchidaceae (ca 27,800 species) are ground-dwelling or epiphytic and in part mycoheterotrophic herbs of small, sometimes moderate size producing as many as millions of tiny, wind-dispersed seeds per flower. The tepals are more or less free and the flowers, with their distinctive and complex morphologies centred on the gynostemium (stamens and stigma-style all congenitally fused) and labellum (median tepal of the inner whorl), are often inverted. Deceit pollination is particularly common, but pollinators also visit for fragrance and nectar rewards. Although the diversity of floral form in Orchidaceae is great, it is attained by variation on a rather limited basic theme. 2. Zingiberales (2,100 species) are quite large tropical plants, mostly herbs, and they have large flowers. Their fruits usually have only a moderate number to few seeds that are animal dispersed. Flowers in Zingiberales vary considerably in orientation, the parts that are petal-like, number of stamens, etc., and from this point of view are more variable than Orchidaceae. (Note that many taxa in Commelinales, sister to Zingiberales, also have monosymmetric flowers, and it is possible the the common ancestor of the two had such flowers.) 3. Core Lamiales (23,360 species) are often more or less herbaceous plants perhaps particularly abundant oustide the tropics, although there are many tropical members; woodiness is more common in the tropics. Individual flowers are moderate to quite large in size, each usually producing quite many and small seeds; in general seed dispersal is by wind. Although clades like Lamiaceae have only four seeds per flower, they are still quite small. Most of the monosymmetric 4. Fabaceae (19,400 species, of which 3,300 are in the polysymmetric mimosoid clade) have inverted keel flowers with more or less free petals (Stirton 1981). The plants are either trees, especially common in neotropical forests, or herbs. Dispersal is either autochorous (ballistic) or animal-mediated; there are rather few seeds per fruit. Many Fabaceae are nitrogen-fixers, and the family is noted for the diversity of secondary metabolites that it produces, sometimes in association with endophytic fungi.

Of the smaller but still quite large clades in which monosymmetry predominates, 5. Campanulaceae-Lobelioideae include some 1,200 species of laticiferous herbs or shrubs often with slit-monosymmetric flowers that have plunger-type secondary pollen presentation devices. Their dehiscent fruits have many small seeds, although some clades have fleshy fruits. 6. Caprifoliaceae comprise some 850 species; the flowers are often rather weakly monosymmetric and the often indehiscent and dry fruits have at most few seeds. 7. Some 200 species of Lecythidaceae-Lecythidoideae (Ericales) are trees that are a prominent element of neotropical forests. The polystaminate androecium alone is monosymmetric, the fruit is large, and the seeds are few and large. In 8. Iridaceae, monosymmetry of the flowers of the speciose Gladiolus is obvious. However, from the point of view of the pollinator the flowers of Iris, Moraea, etc., are also monosymmetric; a single Iris flower consists of three strongly monosymmetric meranthia or part-flowers (see also Westerkamp & Claßen-Bockhoff 2007). All told some 750+ species have monosymmetric flowers of one sort or another. The fruits have a moderate number of seeds. 9. Polygalaceae, with some 1,050 species most of which have monosymmetric flowers, are phylogenetically close to Fabaceae, although the exact relationships of the two are unclear (see Bello et al. 2009). They have keel flowers that are superficially like those of many Fabaceae, although different parts are involved (Westerkamp & Weber 1999). The fruits usually have few seeds and myrmecochory is common (Lengyel et al. 2009).

The final clade to be mentioned is 10. Asteraceae (25,100 species). These are mostly herbaceous to shrubby plants with small flowers that are aggregated into heads or capitulae; in most species at least some flowers of each capitulum are monosymmetric and secondary pollen presentation is widespread. Only a single usually quite small seed per flower is produced, and dispersal is often by wind. However, if the capitulum can be considered to be functionally equivalent to a flower, then each capitulum is a single polysymmetric or haplomorphic flower which produces quite numerous seeds. Asteraceae are also noteworthy for the diversity of secondary metabolites they contain. Exactly where acquisition of monosymmetry is to be fixed within Asterales is unclear.

What pollinates such flowers? Different pollinator groups have different floral preferences. Adult Lepidoptera prefer to visit radiate flowers, that is, polysymmetric flowers with a definite number signal (e.g. three-merous, five-merous flowers) and that have enclosed nectar (e.g. Leppik 1957). In Apidae, honey bees (Apini) frequent polysymmetric (radiate) flowers with relatively accessible nectar (for a review of the cues used, see Horridge 2009). Halictidae-Rophitinae (sister to all other halictids) prefer to visit flowers of asterids, especially those of the lamiids, although other plants are also visited (Patiny et al. 2008). Oligolectic (specialist) bees prefer to visit shallow often polysymmetric flowers with easily accessible rewards (Wcislo & Cane 1996); Asteraceae are the group above that are most notably commonly pollinated by such bees (e.g. Larkin et al. 2008), but a variety of insects visit flowers with generally accessible rewards (see also Benton 2017). However, derived polylectic (generalist) euglossine and bumble bees (see below) in particular pollinate flowers with complex, monosymmetric corollas that the insect has to learn to work before visits are effective (see Westerkamp & Claßen-Bockhoff 2007 for monosymmetry and bee pollination). Thus oligolectic basal megachilids pollinate polysymmetric and sometimes rather large flowers, while the more polylectic derived megachilids commonly pollinate monosymmetric Fabaceae and Lamiaceae (Westerkamp 1997), or polysymmetric Boraginaceae (Litman et al. 2011; Sedivy et al. 2013). Bumble bees in particular appear to have an innate preference for monosymmetric flowers (Leppik 1957; Kalisz et al. 2006; Rodríguez et al. 2004), but can pollinate a variety of floral morphologies using a combination of learning and general behavioural flexibility (e.g. Gegear et al. 1995; Raine & Chittka 2007b; Russell et al. 2015, 2017). Birds also visit monosymmetric flowers, and in the gullet-type ornithophilous syndrome stamens in the mouth of the flowers often deposit pollen on the head of the pollinator.

CYC genes have independently been coopted in the development of monosymmetric flowers (Reyes et al. 2016 and references). However, in Zingiberales and Commelinales CYC-like genes are expressed abaxially in the flower, while in eudicots CYC genes are expressed adaxially (Reyes et al. 2016); this difference may be connected with the inverse orientations of the flowers in the two groups. Similarly, obliquely symmetrical flowers may appear inverted from the pollinator's perspective. In angiosperms in general with 3- or 5-merous flowers the perianth part that is differentiated is the odd member (and sometimes also adjacent members) of the inner whorl (see e.g. Reyes et al. 2016: fig. 4), Pelargonium being a notable exception. (A survey of the position of the differentiated petal in the context of floral orientation would be interesting....)

The evolution of monosymmetric flowers may be connected with the evolution of dichogamy (separation of the time of pollen dispersal and stigma receptivity), in particular, to that of protandry (Kalisz et al. 2006; for dichogamy, see e.g. Bertin & Newman 1993; Routley et al. 2004), but dichogamy is advantageous in other situations, too. Monosymmetry, self sterility, apomixis, dioecy and other features tend to be common in families with high diversification rates (Ferrer et al. 2014). Monosymmetry, presence of a corolla, and reduced number of stamens relative to perianth parts have been found to act together, clades with them having a doubled diversification rate compared with those that lack this combination perhaps because more precise interactions with pollinators improve the efficacy of pre-zygotic isolation barriers (O'Meara et al. 2016). In asterids and monocots in particular, monosymmetry and polyandry tend to be mutually exclusive (Jabbour et al. 2008; Reyes et al. 2016, q.v. for possible molecular reasons for this), but some Neotropical Lecythidaceae-Lecythidoideae (Ericales) are spectacular exceptions to this.

Most of the plant clades just mentioned are more or less herbaceous, and Qian and Zhang (2014) discuss the evolution of life forms (woody vs herbaceous) in angiosperms. Among extant seed plants, only angiosperms are herbaceous, and the annual habit, known from ca spp of angiosperms, is not known from other vascular plants. There are long-standing suggestions of a correlation between the rate of molecular evolution and plant habit: Molecular evolution is faster in herbs/annuals (e.g. M. A. Wilson et al. 1990; Bousquet et al. 1992: esp. chloroplast genes; Gaut et al. 1992: chloroplast rbcL gene, grasses evolve notably faster even than other monocot herbs, 1996, 2011: summary; Barraclough et al. 1996: rbcL gene; Andreasen & Baldwin 2001; Soria-Hernanz et al. 2008: ITS, correlation not very strong; Rydin et al. 2009b; Barker et al. 2009; Korall et al. 2010 [ferns]; Müller & Albach 2010; Yue et al. 2010; Frajman & Schönswetter 2011; Comer et al. 2015; c.f. Whittle & Johnson 2003: comparisons of branch lengths of species pairs, ?sampling), and as discussed elsewhere this also links to increased speciation rates, etc.. Both chloroplast and nuclear genes show an increased rate of molecular evolution (Gaut et al. 2011), but not all genes are equally affected (Yue et al. 2010). The rate of molecular evolution in woody angiosperms is faster than that of those conifers studied, although the rate of non-synonymous substitutions is lower (Buschiazzo et al. 2012). In a series of extensive analyses of both monocots and eudicots, S. A. Smith and Donoghue (2008) confirmed that there are usually substantial increases in the rate of molecular evolution in herbs as compared to trees, shrubs, or simply to plants with long life cycles. Thus within the commelinids, Arecaceae, Bromeliaceae and Rapateaceae, all three with long life cycles, have a low rate of molecular evolution, and it could be argued that none is very diverse, although this will depend on how comparisons are made/the methods that are used. Y. Yang et al. (2015: focus on Caryophyllales) also demonstrated this correlation, taking into account synonymous and non-synonymous sites in protein-coding genes, and found substitution rates in herbaceous lineages that were up to three times those in their woody relatives (see also Luo et al. 2015: rosids, also within Rosaceae).

This correlation may be connected with mutation rate or population size and thence to speciation rate (Gaut et al. 2011 and references). There is a general correlation between rates of molecular evolution (substitution rates) and the rate of speciation, more molecular evolution occurring in speciose clades (Webster et al. 2003, see also Barraclough & Savolainen 2001), although this is not always the case (Müller & Albach 2010: Veronica). Herbs also show an increased rate of climatic niche evolution (S. A. Smith & Beaulieu 2009). The evolution of herbs from trees has been linked with higher speciation rates in the former (Dodd et al. 1999), although Verdú (2002) suggested that is was not so much the herbaceous habit per se that was important, but the associated condition, length of time to maturity (see also Baker et al. 2014: Amazonian trees). However, in the fire-prone Mediterranean ecosystem neither diversification nor molecular evolution differs between seeders and resprouters, two ways in which plants survive fires there, yet seeders, having shorter generation times, should perhaps have diversified more (Verdú et al. 2007). In the related "rate of mitosis" hypothesis, mitosis rates in the apical meristem may slow when the plant is near its maximum height, hence leading to reduced plastid substitution rates in trees (Lanfear et al. 2013; Barrett et al. 2015; Bromham et al. 2015a). Overall, trees may have a distinctive evolutionary rhythm, speciating rather slowly; any one species may have quite large numbers of individuals, and although they may be rather dispersed they are long-lived, the species themselves also being rather long-lived (Petit & Hampe 2006; see also de la Torre et al. 2017). However, Nürk et al. 2019) found that diversification rates in plants of (sky) islands that had secondarily become woody and lengthened their life cycles nevertheless showed an increase when compared with that of their herbaceous relatives.

To summarize: Much angiosperm diversity, but not necessarily biomass production or net primary productivity (see below), is concentrated in groups that are annuals or herbaceous or shrubby perennials and that have animal-pollinated flowers (see also Ferrer et al. 2014); disseminules are small, rarely fleshy (Eriksson & Bremer 1991, 1992), any animal dispersal often being by hooks and the like. Several of the large groups with monosymmetric flowers mentioned above (core Lamiales, Asteraceae) are largely made up of such plants. Overall diversification rates/species numbers are high in these clades, particularly in Asterales and Lamiales (Magallón & Sanderson 2001; Magallón & Castillo 2009; O'Meara et al. 2016), although these rates are more properly associated with particular clades within those orders (see the euasterids). Indeed, O'Meara et al. (2016) noted that species with monosymmetric flowers are perhaps less abundant than expected, in part because the apparently unitary "monosymmetric flower" had to be assembled from separate and independent characters like a monosymmetric corolla and reduced stamen number relative to perianth parts, and it took some time, over 60 Ma, for these features to come together.

7C. Major Clades With Wind-Pollinated Flowers. Monoecy and dioecy are associated with features such as woodiness, biotically-dispersed fruits, and the wind (or, more generally, abiotic) pollination syndrome (for which, see e.g. Whitehead 1983; Timerman & Barrett 2018: frequency of filament oscillation and pollen release connected; Galati et al. 2019: orbicule micromorphology). There are ca 16,160 dioecious species, ca 6% of angiosperms (Käfer et al. 2014), and dioecy has evolved hundreds of times (Renner 2014; Käfer et al. 2017), and diversification rates can be high (Käfer et al. 2014). Clades in which wind pollination predominates are usually not notably speciose, the adoption of abiotic pollination often being associated with a decrease in speciation rate (e.g. Dodd et al. 1999). A clear exception is 1. Poaceae, 10,050 or so species of frequently monoecious largely herbaceous wind-pollinated plants with single-seeded fruits (in most species the flowers can be categorized as being reduced-monosymmetric). 2. Cyperaceae-Juncaceae, also with more or less reduced flowers and often with single-seeded fruits, contain about 4,800 species. 3. Fagales include about 1055 species, nearly all monoecious trees with much reduced flowers, the staminate flowers being borne in catkins and the pistillate flowers usually having an inferior ovary; here the single-seeded fruits are often quite large.

In wind-pollinated angiosperms there has been selection for small flowers, monoecy is common, and the pollen is smallish, smooth and often some kind of porate. Most wind-pollinated taxa have few to a single ovule per flower, and this is coupled with a high pollen:ovule ratio, pollen loads on a single female flower being quite high; the fruits very often have a single seed (e.g. Linder 1998; Culley et al. 2002; esp. Friedman & Barrett 2008, 2009 [useful table], 2011). In Fagales and Leucadendron (Proteaceae) monospermous fruits seem to have evolved before wind pollination (D. W. Taylor et al. 2012; Welsford et al. 2016). However, fossil Normapolles plants (Fagales) may have perfect flowers even with nectaries (e.g. Friis et al. 2011; see also fossil Platanaceae). Interestingly, Fagales are sister to Cucurbitales with around 3,000 species nearly all of which are monoecious. 1-seeded fruits may be a synapomorphy for the combined clade, however, nearly all Begoniaceae and Cucurbitaceae, which make up the vast majority oif Cucurbitales, are insect-pollinated and have many-seeded fruits.

Kay and Sargent (2009) noted that Poaceae and Cyperaceae/Juncaceae were exceptions to the rule that it is animal pollination that leads to an increase in speciation rate, the two clades being about seven times more diverse than their animal-pollinated sister clades. Clades immediately below Poaceae are small, and their flowers are small, but are probably pollinated by insects; the relationships of the [Thurniaceae [Cyperaceae + Juncaceae]] clade are unclear. The mostly animal-pollinated Cucurbitales, probably sister to the overwhelmingly wind-pollinated Fagales, have ca 2,300 species, i.e. about twice as many species as in Fagales.

For the evolution of dioecy, see Renner and Ricklefs (1995), Vamosi et al. (2003), Dufay et al. (2014), Renner (2014), and Käfer et al. (2014, 2017) and references. Correlates of dioecy are tropical or island distributions, woody habit, abiotic pollination, inconspicuous flowers and inflorescences, and fleshy fruits, and these may also be correlated among themselves, e.g. tropical, woody, fleshy fruits, although dioecy seems not to evolve notably frequently within clades with these features (Vamosi et al. 2003). Clades in which dioecy predominates are not notably speciose, and it has been suggested that dioecious clades diversify less (Heilbuth 2000; Vamosi & Otto 2002; Kay et al. 2006, etc.; c.f. in part Leslie et al. 2013). However, recent analyses suggest that the evolution of dioecy seems to have little effect on diversification rates (Sabath et al. 2015), in general, rates of changes of sexual system type being similar (Goldberg et al. 2017). Indeed, correcting for the lag in evolution of dioecy in dioecious clades (it is a derived character, thus it may not have evolved spot on at the clade divergence), even an increase, dioecy being associated with families with moderate to high diversification rates, but reversals to monoecy are also common (Ferrer et al. 2014; Käfer & Mousset 2014; Käfer et al. 2014; Q. Zhang et al. 2018). Dioecy tends to be associated with pollination by wind (Sabath et al. 2015) and if insect pollinated, the floral displays tend to be dimorphic, those of the staminate plants being showier and so more visited, and it has been suggested that for this and other reasons extinction is thus perhaps quite likely (Käfer et al. 2017 for literature). Dioecy may be relatively uncommon because reversions back to monoecy may be quite frequent. After all, in dioceous taxa the ability to express genes allowing the development of both carpels and stamens must persist in both staminate and carpelate individuals, so Dollo's Law is not really applicable (see Käfer et al. 2017; Goldberg et al. 2017).

7D. A Cautionary Note. But what is really known about the relationships of such features just mentioned to the diversification of the clades that have them? Using Poaceae as an example, we can see how complex and difficult a question like "Are Poaceae diverse, and why?" can be. A series of points:

1. The first three clades of Poaceae that are successively sister to the remainder contain some 26 species out of the 11,000+, and these three "basal" clades are forest plants (e.g. Givnish et al. 2010b).

2. Ca 1,300 species of bamboos are woody and have a distinctive, synchronized monocarpic flowering habit; they are arguably ecologically distinct from the rest of the family.

3. Poaceae-Poöideae (ca 3,850 spp) are noted for their association with fungal endophytes, an association that could be ca 40 Ma old (Schardl et al. 2004). The presence of these endophytes affects the palatability of foliage to herbivorous mammals and of seeds to granivorous birds, and animals eating the infected material may not thrive. The level of aphid infestation and that of their parasites and parasitoids, and even the pattern and rate of decomposition of dead grass, are also affected (e.g. Madej & Clay 1991 - birds; Omacini et al. 2001 - aphids; Lemmons et al. 2005 - decomposition). A variety of alkaloids, including loliine (pyrrolizidine) and ergot alkaloids, are produced by the fungi; the distinctive loliine alkaloid is primarily active against insects (Schardl et al. 2007). Various aspects of root growth may also be affected (Sasan & Bidochka 2012). Poöideae are largely a cold-tolerant group. Poöideae are also noteworthy in that they are by far the largest temperate clade in a largely tropical family.

4. Grasses are well known for the diversity of silica bodies in their leaves, and these play a role in protection against herbivory (but probably not against that by mammalian grazers), while silicon concentration itself is correlated with the rate of breakdown of plant tissues, and so with nutrient cycling (see silica).

5. About 75% of the PACMAD clade, some 4,500 species, have C4 photosynthesis (Sage et al. 1999; Grass Phylogeny Working Group II 2011, and are ecologically very distinctive (see below). C4 plants tend to be less attractive to herbivorous animals because of their lower nitrogen concentration and greater amount of fibrous tissue (Caswell et al. 1973).

8. Clade Asymmetries.

When thinking of overall patterns of seed plant diversity and evolution, there are a number of striking examples of what may be called clade asymmetries involving quite small clades of animals and plants. There are two rather different kinds of these asymmetries. One includes small (in terms of species numbers) clades of animals involved in the pollination and seed dispersal of relatively very large numbers of plants. As we will see, these are quite well known, especially on a fairly local scale, but here the scale is global. The other kind of asymmetry is that in more physiological-ecological relationships, where relatively small clades of plants have major effects on biome functioning globally, particularly through their effects on carbon cycling. Both kinds of asymmetries have major implications for species persistence, the ecological structuring of communities and ecosystems, and the way one thinks about diversity and evolution in general.

8A. Plant-Animal interactions.
8B. Carbon Sequestration.


As Ollerton et al. (2011: p. 321) noted, "if a policy-maker or conservation planner were to ask an ecologist the straightforward question, 'How many species of flowering plants are pollinated by animals?' the answer would be: 'We do not know'.". They did provide an estimate - around 308,000. However, for questions like "How many species of plants are dispersed by [such and such a group of] bats?", and "How many species are pollinated by [such and such a group of] of bees?", it is much more difficult to obtain reliable estimates.

Estimates of the numbers of species of particular groups of bees, birds or bats may be fairly accurate, but the same certainly cannot be said of the numbers of species of plants that they service. Many observations in the literature do not allow one to distinguish between a visit of an animal to a plant that is casual or one that results in pollination, and assigning a pollinator to a plant is not simple. Thus of the seven birds visiting the flowers of Butea monosperma, only one was an effective pollinator - and so was a squirrel (Tandon et al. 2003), and similar examples are common. Furthermore, studies of such plant-animal relationships still often focus on only one of the partners.

Pollinating bees, for instance, were initially categorized as such based largely on observations on plants visited for pollen, but pollination also occurs when bees are nectaring, and bees, even oligolectic bees, tend to be more florally promiscuous when they are nectaring, and they may pollinate in such circumstances (e.g. Waser et al. 1996; Sipes & Tepedino 2005; Michener 2007). Add variation in time - from season to season, within a season, as well as time of day that the stigma is receptive - and space, and characterizing pollination relationships becomes difficult (e.g. Fishbein & Venable 1996; Waser et al. 1996; Kandori 2002; Ollerton et al. 2003, 2007; Thompson 2009; Crone 2013). Furthermore, relationships between plant and pollinator are by no means constant (e.g. Aizen et al. 2012; Natalis & Wesselingh 2013), and even if they are constant locally, this may not be true across the range of the species (Newman et al. 2014; van der Niet et al. 2014 and references: pollinator ecotypes). Rosas-Guerrero et al. (2014) suggested that secondary pollinators might often be ancestral pollinators. Some insects visiting flowers may not be effective pollinators at all, but scavenge pollen remaining after pollination had actually occurred, or they are otherwise irrelevant to the pollination process, the plant and insect having some kind of commensal relationship (Linsley 1958; Linsley et al. 1973; Michener 1979; Roubik 1989; Roulston et al. 2000: p. 618).

The estimates below are based on very scanty observations sometimes extrapolated to flowers in the same immediate clade with a similar floral morphology; floral syndromes have been used (see Rosas-Guerrero et al. 2014 for justification), despite the reservations mentioned elsewhere. However, the connection between red colour and bird pollination is not straightforward. Thus red colour may make the flowers less conspicuous to bees, and such less bee-conspicuous flowers tended to have longer tubes (see also Raven 1972; Chittka & Waser 1997; Rodríguez-Gironés & Santamaría 2004; Coimbra et al. 2020). Grant (1966) noted that flowers with the ornithophilous syndrome - large, red, often tubular flowers, whether mono- or polysymmetrical, no scent (even the pollen lacks scent - Dobson & Bergström 2000) - were particularly prominent in North America where the hummingbirds that pollinated them were migratory. However, Waser et al. (2018) noted a number of instances where hummingbirds in western North America visited "non-ornithophilous" flowers, although the benefit to the plant was less evident than the benefit to the bird. In tropical America Grant (1966) thought that such syndromes were less evident, but in Mexico, at least, Martin González et al. (2018) noted that it was the migrant humminbirds that were flexible in the flowers that they visited and the residents preferred to visit more typical ornithophilous flower. Overall, it would seem that non-migratory birds in the tropics might have a better chance to learn the local flora while for migrating birds consistent signals in different places might desirable.

Of course, nectarivorous birds may visit a variety of flowers usually pollinated by other animals (e.g. Muruyama et al. 2013), while other animals may on occasion pollinate flowers visited by hummingbirds (e.g. Snow & Snow 1980; E. D. Brown & Hopkins 1995; Fleming et al. 2005). Thus only 2-3% of the species in Neotropical cerrado vegetation have ornithophilous flowers, but hummingbirds take nectar from a similar number of species that do not have ornithophilous flowers, and some of these are pollinated (Muruyama et al. 2013); birds may visit flowers with "inappropriate" morphologies if preferred flowers are not present or there is competition for those that are there (Temeles et al. 2002, 2009). The pollination behaviour of birds on islands may differ from that on the mainland, the birds being more promiscuous in the plants they visit, a phenomenon called interaction release (Traveset et al. 2015; see also Dalsgaard et al. 2018), and plants may also become more promiscuous in their pollinators (Serrano-Serrano et al. 2017, see also below). Of course, many birds also eat insects they find on flowers, although this may be irrelevant when thinking about plant/pollinator relationships.

All this emphasizes the amount of salt needed when reading the discussion below: Estimates of numbers of pollinators rely in part on floral syndromes, anecdotal evidence, and other suspect data.


Plant groups with species pollinated by euglossines include Araceae, e.g. the very speciose Anthurium and Spathiphyllum, visited for fragrances (e.g. Hentrich et al. 2010b and references), Bignoniaceae, Gesneriaceae-Gesnerioideae (ca 300 spp., both male and female bees involved - Wiehler 1978), Lecythidaceae-Lecythidoideae, 900 to perhaps 2,000 species of Orchidaceae-Epidendroideae (for the latter figure, most Stanhopeinae, Zygopetalinae and Cataseteinae may be visited by male bees for fragrances: N. H. Williams 1982; numbers from Pridgeon et al. 2009), and Zingiberales, especially Costaceae (ca 19 species of Costus - Salzman et al. 2015) and Marantaceae, also Apocynaceae, Convolvulaceae, Euphorbiaceae (perhaps 70 species in Dalechampia alone - Armbruster 1993; Armbruster et al. 2009b), Fabaceae (including the nectarless Swartzia), Solanaceae, and Rubiaceae (see Dressler 1968; K. M. Cameron 2004; Roubik & Hanson 2004; Ramírez et al. 2011; Schiestl 2012, and references).

In any one community there may be up to 50 species of bees, and bee populations are often notably stable (Roubik 1989; Roubik & Hanson 2004; Zimmermann et al. 2009). Hentrich recorded ca 23 species of bees visiting ca 48 species of plants in Nouragues, French Guiana, while Ramírez (2009) noted that one species of Euglossa might visit 74 species of plants from 41 families - and that at a single locality.

Overall, these 200 or so species of euglossine bees are likely to be the major pollinators of well over 4,000 species of Neotropical plants (Wiehler 1976; N. H. Williams 1982; Ramírez et al. 2002 for a summary of the literature; Ramírez pers. comm.). However, very little is known about the pollination of most orchids, thus Nunes et al. (2017) found that the Zygopetalum species they examined were pollinated by bumble bees and Dichaea by weevils, although both genera (members of Zygopetalinae) had been thought (tentatively) to be pollinated by orchid bees (Pridgeon et al. 2009).

Age: Crown group euglossine diversification probably began only 42-27 Ma, montane clades diverging only in the last 8-4 Ma (Ramírez et al. 2010); another estimate of crown-group age is slightly younger, some (35-)28, 26(-17) Ma (Cardinal & Danforth 2011; Martins et al. 2014). For relationships in Euglossini, see Bossert et al. (2018).

Flowers with a diversity of morphologies in temperate and Arctic-Alpine floras in particular are pollinated by bumble bees (see above for bumble bees and learning skills). Ericaceae, including Rhododendron and Vaccinium, are common in Arctic communities, but practically all conspicuous flowers in these habitats are visited by the bees (e.g. Heinrich 1979; Ranta & Lunberg 1981; Tomono & Sota 1997; Kudo et al. 2011 and references). I have not found estimates for the number of species pollinated there, but it is likely to be appreciable. Bumble bees are also prominent in alpine environments, where hundreds of species in large genera like Gentiana, Rhododendron (Corlett 2004 and references) and Pedicularis (perhaps 600 species here alone - Macior 1994; P. Williams et al. 2009; Eaton et al. 2012) largely depend on them for pollination. Lamiaceae and many other Lamiales, Fabaceae, particularly Faboideae, and Impatiens (Balsaminaceae) are also often visited (Williams et al. 2009). Most of the 600-700 species of Ranunculaceae-Delphinieae are bumble bee-pollinated (Jabbour & Renner 2012b), this having long been recognised in Aconitum in particular (see map in Kronfeld 1890). In South America bumble bees pollinate genera like Rubus, Scutellaria, Lathyrus and Lupinus, all of which have diversified substantially there (Asmussen & Liston 1998; Hines 2008).

Bumble bees in Europe and elsewhere visit monosymmetric flowers (Raine & Chittka 2007a, b; Goulson & Darvill 2004), and they can also handle polysymmetric flowers quite easily (Laverty 1994; Sedivy et al. 2013). Many bees are more specialized in pollen collection than in nectar foraging (for nectaring, see Raine et al. 2006); learning to handle pollen flowers can quite difficult even for bumble bees (Strickler 1979; Goulson & Darvill 2004; Benton 2006; Raine & Chittka 2007b; Goulson 2010), although they also show considerable behavioural flexibility (e.g. Russell et al. 2017). Bumble bees are effective buzz pollinators (Goulson 2010). In at least some Ericaceae, bumble bees (and Andrena) both buzz the flower and also take nectar (e.g. Moquet et al. 2017a, b). Furthermore, buzzing/sonication and scrabbling, bumble bees picking up pollen using their legs (female bees alone are of course involved), are in part alternative ways of collecting pollen when there is plenty vs little obviously available and can be used on the same flower (Russell et al. 2017). ?Other bees doing this - ?Andrena.

Local diversity of bumble bees can be quite high, with 4-12 species occurring in a single community (Hines 2008 and references), and one fifth or more of the world's bumble bee species (40<) are found in the Sichuan-Chongqing region of China alone (P. Williams et al. 2009). In the Front Range of the Colorado Rockies ca 18 species of bumble bees pollinate ca 43 species of plants (Macior 1974).

Estimates of the numbers of species of plants pollinated by bumble bees are hard to make, but upwards of 3,000 is a plausible number.

Age: Bumble bees diversified over a similar time frame as did the euglossine bees, i.e. about 47-25 Ma, although their stem group (they split from meliponines) age may be considerably more, 100-80 Ma (Hines 2008; see also Martins et al. 2014); Cardinal and Danforth (2011) suggest a somewhat more recent age for crown-group bumble bees of (31-)21(-12) Ma. The Eocene-Oligocene boundary of ca 34 Ma was a time of sharp cooling and increase of seasonality, and bumble bees flourish in cooler climates, being facultatively endothermic (Hines 2008 and references). The bees moved into South America about 8-6 Ma, perhaps along with the plant genera of northern origins that they now pollinate (Asmussen & Liston 1998; Hines 2008).

For Apis mellifera and pollination - it is a major pollinator - see Hung et al. (2017).

Age: Stem-group Apis is late Eocene/earliest Oligocene, with the diversification of extant species beginning a mere ca 13.5 Ma; it is unclear if diversification began in Europe (fossils) or Asia (Kotthoff et al. 2013: they favour the first hypothesis). Other estimates for the age of the group are (30-)22(-16) Ma (Cardinal & Danforth 2011; Martins et al. 2014).

The evolution of flowers which have oils as their primary reward may have begun in the Eocene (Renner & Schaefer 2010). However, Cardinal and Danforth (2013) estimated that Centradini and Tetrapedia, which take oil from Malpighiaceae in particular, evolved in the Late Cretaceous 87-52 and 92-66 Ma respectively; see also Neff and Simpson (1981) for the bees. The malpig Eoglandulosa warmanensis known from the Eocene Claiborne Formation in Tennessee, U.S.A., in deposits ca 34 Ma has distinctive paired calyx glands and may have been pollinated by oil-collecting bees (Taylor & Crepet 1987; Friis et al. 2011). Martins et al. (2014a) found Centradini to be paraphyletic (but c.f. Bossert et al. 2018), Epicharis diverging (102-)91(-79) Ma and Centris (95-)84(-72) Ma, about contemporaneous with the stem-group age of Malpighiaceae, estimated at (100-)86(-73) Ma (Xi et al. 2012b: other estimates are 75-60(-32) Ma, see Wikström et al. 2001; Davis & Anderson 2010; Renner & Schaefer 2010). This is broadly consistent with some kind of co-evolutionary story, Martins et al. (2014) even suggesting that oil collection in the bee clade evolved in the common ancestor of Epicharis and other bees, corbiculate bees later losing the ability to collect oils. However, crown ages of Epicharis are (39-)28(-18) Ma and of Centris (58-)44(-36) Ma, all (much) younger, while Michez et al. (2007) described the ca 53 Ma Paleomacropis eocenicus from early Eocene amber in France which might have pollinated Lysimachia. All in all, details of the evolution of the association between the bees and malpigs are not clear, and the origins and timing of oil collecting in bees need clarification (Neff & Simpson 2017).

Some 1500-2,500 species of oil flowers in 11 families are pollinated by females of 365-447 species of bees (Alves-dos Santos et al. 2007; Martins et al. 2013, 2014; Possobom & Machado 2017a and references), the ca 26 species of Rediviva pollinating around 140 species (Neff & Simpson 2017)

There are four groups of these flies. The Moegistorhynchus longirostris group includes two species of flies that pollinate 21 species of plants in two families and are suspected of pollinating 7 more species (and one more family). In the Moegistorhynchus-Philoliche group 6 species of flies visit 36 species of plants in 3 families, pollinating perhaps another 45 species (and two more families), while in the Prosoeca ganglbaueri group 3 species of flies pollinate 39 species of plants in 7 families, with 39 more species (three more families) as possible nectar sources. The fourth group, which includes just Stenobasipteron wiedemanni, pollinates 19 species in six families; in perhaps 9 species the relationship is obligate, and the fly may pollinate another 12 species (Manning & Goldblatt 1995; Goldblatt & Manning 2000, 2006; Potgieter & Edwards 2005; Johnson 2010; Newman et al. 2014).

All told, some 12 species of flies pollinate around 108 species of plants, although the latter figure is likely to be a considerable underestimate. Iridaceae figure particularly prominently, with about 34 species being pollinated by Nemestrinidae - but 100 more are thought to be pollinated by the flies (Manning & Goldblatt 1996, 1997; Goldblatt & Manning 2000)!

Age: Nemestrinids are an old group known from as far back as the Jurassic, but

Some 22,000 or more species of angiosperms, mostly core eudicots, are buzz pollinated (e.g. de Luca & Vallejo-Marín 2013; Cardinal et al. 2018). However, the sonication technique used in buzz pollination is also used in other situations, in particular it may be used in flowers that have little pollen remaining, although in the same flowers that have more pollen the bees (a bumble bee was being observed) may collect pollen by scraping it up (Russell et al. 2017: compilation of species that do not have poricidal anthers but are nevertheless sonicated).


In the estimates below of numbers of plants visited I have focussed on larger clades of plants and on literature covering flower visitation in particular regions. As a result taxa like Brachychiton are probably included (in the general figures for bird visitation in Australia), as are the 40 or more Old World bird-pollinated species of Erythrina (Bruneau et al. 1997), while species like Holmskiodia sanguinea are probably not (see also Porsch 1936 and references for early literature). I also focus on the major groups of birds that pollinate plants. Not mentioned, for example, is the role played by New World orioles (Icterus), tanagers, and other more or less casual - perhaps more from their point of view - flower visitors (e.g. Stiles 1981; Rocca & Sazima 2010) in pollination, although they are needed for pollination in a number of taxa; orioles are particularly important pollinators in drier forests (Stiles 1985). Note that New World flowers pollinated by birds other than hummingbirds tend to be longer-lived, the nectar is more copious and dilute and somtimes coloured, there may be various kinds of food bodies, perches for the birds, etc. (Rocca & Sazima 2010: Table II). For recent summaries of various aspects of bird pollination, see Fleming and Kress (2013) and Zanata et al. (2017).

Hummingbirds may be trap-liners (especially hermits), generalists, or territorial, while "parasitic" hummingbirds visit flowers normally visited by other species of hummingbirds (e.g. Feinsinger & Colwell 1978). There is considerable diversity in morphology and behaviour of hummingbirds, especially in the most species-rich assemblages (Abrahamczyk & Kessler 2014), and male and female birds can have bills of different lengths, curvature and degree of serration of the margin and/or the birds may differ in aggressiveness, etc., and so may pollinate quite different plants (e.g. Bleiweiss 1999; Temeles et al. 2009, 2013; Rico-Guevara et al. 2018). Depending on the birds and plants in the community, different foraging types of hummingbirds will visit different species of plants (Feinsinger & Colwell 1978); trap-lining birds tend to visit fewer species of plants than do territorial birds (Snow & Snow 1980). As is well known, hummingbirds often visit herbaceous plants, in part because they do not need perches when feeding (e.g. Stiles 1981; Kress & Beach 1994; Fleming & Muchhala 2008). Self incompatability was commonest is woody plants pollinated by other than trapliners and also less marked in herbaceous taxa generally (Wolowski et al. 2013: Itatiaia).

Trap-lining hermits, which one would think tended to disperse pollen over longer distances (e.g. Wolowski et al. 2013) tend to be commoner at lower altitudes in the Andes, and several clades have independently moved into high-altitude habitats (Bleiwiss 1998b; McGuire et al. 2007 and references). Andean hummingbirds have notably small mean ranges, a quarter the size of other hummingbirds, and hummingbirds as a whole have just over a third of the mean range when compared with other birds (McGuire et al. 2014). Nevertheless, there are still usually several species of hummingbirds in any one place, up to 25-30 being recorded from a single local assemblage (Graham et al. 2009, 2012), and the common species at least visit several species of plants. Figures in Fleming et al. (2005) are 3-28 hummingbird species per site, pollinating 14-51 species of flowers, mostly herbs, whether epiphytic or not, to trees. In an Andean rain forest at around 2000 m, 79 species of flowering plants (in twelve families and 29 genera) were visited by 26 species of hummingbirds, of which the eight commonest visited 74% of the plants (Dziedzioch et al. 2003), while in the Monteverde forest, Costa Rica, 23 species of hummingbirds (excluding uncommon and rare species) visited 181 species of plants in 60 genera and 28 families, 8.8% of the total flora (Murray et al. 2000; see also Maglianesi et al. 2014a). Interestingly, in an analysis along an altitudinal gradient in southern Ecuador curved-billed specialists (Phaethorninae) tended to be overrepresented at lower altitudes, long-billed specialists (several origins) in the high Andes, while at medium altitudes generalist species with small- to medium-sized bills predominated (Sonne et al. 2019). In southeastern Brazil, home to relatively few hummingbirds, four species of birds visited 23 species of plant belonging to 21 genera and 14 families, individual species of birds visiting between three and eighteen species of plants (I. Sazima et al. 1996: species visited only once excluded), while Buzato et al. (2000) recorded 86 species of plants (23 families, 44 genera) being visited by 15 species of hummingbirds, although at any one site there were 30-41 species of plants in 13-16 families mostly visited by 3-4 species of birds - only Bromeliaceae were abundant, with 20 species in 9 genera, of which Vriesia had seven species, at the lowland site. Buzato et al. (2000) suggested that despite the relatively few species of hummingbirds in the Atlantic Forest area, as many species of plants were pollinated there as in Colombia or Ecuador. Gentry (1982) outlines the general diversity of bird-pollinated taxa of Gondwanan origin in tropical and premontane parts of the northern Andes and Ferreira et al. (2016) discuss hummingbird pollination in the Brazilian Cerrado.

Specialization in plant-hummingbird networks is most evident in places where quaternary climate change, as measured by metres needed to move each year to stay in the same climate, is low and the species richness of plants high (Dalsgaard et al. 2011; see also Stiles 1978), hence it would be expected to be less in areas in North America with few (and migratory) species of hummingbirds. This runs somewhat counter to earlier suggestions by Grant (1966), however, it is supported by recent work in Mexico and western North America (Martin González et al. 2018; Waser et al 2018) where migratory hummingbirds visit flowers of various morphologies in addition to more typical ornithophilous flowers.

Some 1,000 or more species of Ericaceae-Vaccinioideae-Vaccinieae and Gesneriaceae-Gesnerioideae (around 600 species of the latter alone - Wiehler 1978; see also Perret et al. 2007; Rodriguez et al. 2010; Clark et al. 2015; Serrano-Serrano et al. 2017), also with their centres of diversity in the Colombian-Ecuadorean region, may be pollinated by hummingbirds (Luteyn 2002; Weber 2010). To these plants can be aded some 250 species of Salvia (Wester & Claßen-Bockhoff 2011), 225 species of Heliconia (Pedersen & Kress 1999), 500-600 species of Acanthaceae (E. A. Tripp & L. McDade, pers. comm.; Tripp & Manos 2006; Tripp & Tsai 2017), about 365 species of the Centropogon alliance of the Campanulaceae-Lobelioideae (L. Lagomarsino, pers. comm. 9.iii.2017; see Stein 1992), around 505 species in 39 largely unrelated genera of Rubiaceae, about 12% of the neotropical species (C. M. Taylor, pers. comm. 22.viii.2015), some 55 species of Erythrina (Bruneau 1997), over 40 species of Penstemon (P. Wilson et al. 2006, 2007; Wessinger et al. 2016), hundreds (perhaps 1,060-1,300) of species of Bromeliaceae, mostly at higher altitudes in the Andes, not in drier habitats or in terrestrial lowland forest habitats (e.g. Benzing et al. 2000a; Kessler & Krömer 2000; Givnish et al. 2014), about 125 species of Loranthaceae, perhaps 100 species of Fuchsia (e.g. Berry 1989; Wagner et al. 2007 and references), 138/555 species of Passiflora (K. Porter-Utley & J. MacDougal, pers. comm.), and so on. Just about all sizeable sympetalous families in the hummingbirds' ranges are visited, even including ca 45 species of Asteraceae (Vogel 2016), as are polypetalous Rosaceae, Melastomataceae and Symplocaceae, ca 27 species of Costus (Costaceae: Salzman et al. 2015), etc. (see e.g. Snow & Snow 1980; I. Sazima et al. 1996 for other examples). Some 390 species of Cactaceae have reddish flowers, ca 20% of these flowers are often notably tubular, and all told there are perhaps 120 species with flowers that might be considered bird-pollinated (Gorostiague & Ortega-Baes 2015; 187 species is the estimate in Mutke et al. 2015); however, pollinator specificity does not seem to be high and I have not included cacti in the total. I have also not tried to take into account flowers of various morphologies that are visited by migratory hummingbirds (in addition to more typical ornithophilous flowers) in Mexico and western North America (Martin González et al. 2018; Waser et al 2018).

The overall imbalance of species numbers of hummingbirds/plants pollinated is probably similar to the euglossine bees just mentioned, with some 338 species of birds pollinating over 5,000 species of plants (over 4,500 species above, but this is a considerable underestimate); Abrahamczyk and Kessler (2014) thought that the number was ca 7,000 species, but even they, too, thought that this was an underestimate). In any event, the ratio is almost 12:1 to over 20:1. At regional scales, in North America north of 24oN, where the hummingbirds are migratory for the most part, and in South America south of 27oS about eighteen and six species of hummingbirds visit some 184 and 56 species of plants respectively (Abrahamczyk & Renner 2015; see also Cantrill & Poole 2012 and references for hummingbird pollination in southern Valdivian forests).

There are other, more general, estimates. Thus Muruyama et al. (2013; see also Rocca & Sazima 2010) estimated that perhaps 20% of the species in Amazonian rain forest were pollinated by hummingbirds, compared to 2-3% in Cerrado vegetation; if there are around 60,000 species of flowering plants in Amazonia, that would suggest 12,000 species were pollinated by the birds there alone. Kress and Beach (1994) estimated that hummingbirds pollinated 14.9% of the 276 species of plants they examined at La Selva, Costa Rica, and when extrapolated to Amazonia this would yield a broadly comparable number of 8,940 species. However, these 274 species are not a random sample of the flora; Stiles (1985) estimated a total of 56 bird-pollinated species at La Selva (Kress & Beach 1994 recorded 41 species), only 3.8% of a total flora of 1650 species (Hartshorn & Hammel 1994 estimate 1,280 species of flowering plants). The numbers offered by Stiles (1985) extrapolated to the whole of Amazonia give a figure of 2,625 hummingbird-pollinated species there.

Focussing on the numbers of clades of hummingbird-pollinated plants, we find that at the northern and southern ends of the overall range of hummingbirds - i.e. southern South Ameria, much of North America - the clades are small (Abrahamczyk & Renner 2015), as are hummingbird pollinated clades of Rubiaceae (C. M. Taylor pers. comm. 22.viii.2015), Acanthaceae-Ruellia (Tripp & Tsai 2017) and Gesneriaceae (Perret et al. 2003, 2007; Roalson et al. 2007; Martín Rodriguez et al. 2009; Clark et al. 2015) throughout their ranges (but c.f. Serrano-Serrano et al. 2017 in part). There are numerous origins of bird-pollination in Ruellia (Acanthaceae), and speciation in such clades is increased, but reversals to other pollination modes are also frequent (Tripp & Tsai 2017). But such stories get more complex. Although there is a large clade of Andean centropogonid lobelioids that is likely to be plesiomorphically pollinated by straight-billed hummingbirds, there have also been (10-)13(-23) other origins of the straight-billed pollination syndrome when reversals from the sickle-billed and bat-pollinated floral syndromes are taken into account (Lagomarsino et al. 2017). See also Iles et al. (2016) for dates.

Age: Hummingbirds and swifts are sister clades that had probably diverged by the Eocene (58-)53.5, 42.1(-36.9) Ma (McGuire et al. 2014; Jarvis et al. 2014; Prum et al. 2015; Mayr 2016). Rather surprisingly, Eurotrochilus, assignable to stem-group hummingbirds although quite similar to Trochilinae and apparently a nectar-eater, is known from Oligocene Europe in deposits ca 34.3 Ma (Mayr 2004, 2009, 2014; Louchart et al. 2008), there are also fossils from the Late Eocene of the Caucasus (Louchart et al. 2008), and there are stem-group hummingbirds that appear not to have been nectar feeders from deposits ca 48 Ma old. Still more surprisingly, a clade of proctophyllodid feather mites, Rhamphocaulini, known only from hummingbirds, has been variously dated to (106.9-)71.7, 57.2(-47.6) Ma, the younger ages being from BEAST analyses of host phylogeography, so the hummingbird clade may be substantially older than thought; Pterodectini, sister to Rhamphocaulini, are mostly on passerines, and the basal members are Old World (Klimov et al. 2017). Of course, extant crown-group hummingbirds are restricted to the New World, and it is estimated that they moved there between 40 and 22 Ma, diversification beginning in South America (24.7-)22.4(-20.3) Ma (McGuire et al. 2014). Note that Tripp and McDade (2014a) estimated crown-group diversification to have begun in the mid-Oligocene (29.9-)28.8(-28.4) Ma while ca 29 Ma is an estimate from Claramunt and Cracraft (2015: Phae. inc.). However, around 67 Ma is the age suggested by Fleming and Kress (2013; see also van Tuinen & Hedges 2001), more compatible with the estimates of Klimov et al. (2017). The trap-lining hermits, Phaethornithinae (Bleiwiss 1998a) - topazes are in the same clade (McGuire et al. 2014) - are sister to other hummingbirds. Much speciation occurred about 13-12 Ma along with the uplift of the Andes where some 140 species are to be found, hummingbirds moving to Central-North Americas ca 12 Ma (Bleiweiss 1998a; McGuire et al. 2007, 2014; Pacheco et al. 2011; Abrahamczyk & Renner 2015; Prum et al. 2015: Archilochus + The Rest; see for connections between North and South America), the Coquettes and Brilliants, with almost a third of hummingbird species, being very largely Andean. However, Patagona gigas, also an Andean species, is the only extant member of a clade over 14 Ma.

There are four, perhaps five, cases where hummingbird-plant interactions may be quite old and where clades of hummingbird-pollinated plants may be quite large. The first case is Gesneriaceae-Gesnerioideae. There Serrano-Serrano et al. (2017) noted that diversification increased (25.5-)18.5(-5) Ma and that hummingbirds may have arrived in South America 25-20.3 Ma at about this time, the birds spurring this diversification. They estimated that there were (41.5-)31.5(-21.5) shifts to hummingbird from insect (bee) pollination, the latter probably being the original condition for Gesnerioideae, and these shifts were often near the base of large clades that subsequently had very high subsequent diversification rates. Overall ca 60% of Gesnerioideae they studied (351/590) were bird-pollinated (Serrano-Serrano et al. 2017; see also Roalson & Roberts 2016: hummingbird-pollinated clades in Columneinae ca 22.4 My), although they discuss other factors, including the adoption of the epiphytic habit, that may also have increased speciation. Roalson and Roberts (2016) had dated three major clades dominated by hummingbird pollination to 22.4[Columneinae]-15.2 Ma, rather similar ages. The second case are Heliconiaceae-hummingbird relationship. Most Heliconia are from the New World, and their predominant pollinators are hummingbirds; Heliconiaceae make up one of the single most important hummingbird-pollinated clades. Estimates of the age of crown-group/possibly hummingbird-pollinated Heliconia are 40-30 Ma - 32-28 Ma (Kress & Specht 2006), ca 32 Ma (McKenna & Farrell 2006), or (47-)39(-32) Ma, diversification being most evident from a little over 30 Ma onwards (Iles et al. 2016). Interestingly, sicklebills, Eutoxeres, visit Heliconia at lower elevations and Centropogon at higher elevations, perhaps moving from the former to the latter group which diversified only 3-2 Ma (Abrahamczyk et al. 2017a; Lagomarsino et al. 2014, 2016). The third case is in Bromeliaceae, where bird pollination in the [Pitcairnioideae [Puyoideae + Bromelioideae]] clade can be dated to around 14 Ma (Givnish et al. 2014a, see also 2004a, 2011a). Within Lamiaceae, in Nepetoideae-Salviinae, a burst of diversification within the New World Salvia subgenus Calosphace - at 550 species it includes over half the genus, and ca 300 are pollinated by birds (as also are some Menthinae) - occurred in Mexico 17.8-14.1 Ma (Kriebel et al. 2019). Stem and crown group ages for this clade are ca 22 and 20 Ma respectively, and these ages are similar to the crown age of humminbirds, ca 22 Ma (Kriebel et al. 2019; McGuire et al. 2014). Finally, some clades of bird-pollinated Ericaceae-Vaccinieae may also be quite large, although little is known about ages of phylogenetic relationships in this group.

Perhaps hummingbirds that early evolved in association with these groups of plants became the templates, as it were, for a variety of younger plant clades as they adopted bird pollination (Mayr 2005, 2009). Thus it has been suggested that the evolution of hummingbird-pollinated plants was "facilitated by this pre-existing relationship" in Heliconiaceae (Iles et al. 2016: p. 161; see also Abrahamczyk et al. 2017a). Some younger plant clades pollinated by hummingbirds, although quite young, may be quite large. Thus the long-billed hummingbird Ensifera ensifera visits some 37 species of Andean Passiflora supersection Tacsonia alone as well as a few other species (Abrahamczyk et al. 2014, 2017a: 62-64 spp.), while some 50 species of a clade in Centropogon are pollinated by the sickle-bill hummingbirds Eutoxeres condaminii (Lagomarsino et al. 2014, 2016; Abrahamczyk et al. 2017a).

Plants visited by Old World nectarivorous birds are mostly woody and the birds usually perch on the twigs when feeding (Stiles 1981; Fleming & Muchhala 2008). Compared with hummingbirds, the relationships between bird and flower sometimes seem rather indiscriminate (see also Stiles 1981). From Africa to Australia, the birds often fly in mixed flocks (c.f. in part E. D. Brown & Hopkins 1995), with up to nine species of lories alone occurring together in New Guinea (Schweizer et al. 2015). Any one species of bird may visit many species of plants, and any one species of plant may be visited by many species of birds, although bird-plant relationships are not totally promiscuous (Gill & Wolf 1975; Ford et al. 1979; Rebelo 1987; Brown & Hopkins 1995; Franklin & Noske 2000). Here I discuss bird pollination in Australia, South East Asia-Malesia, New Zealand, New Caledonia, Africa and Hawai'i separately.

In Australia there are about 75 species of honeyeaters (E. T. Miller et al. 2016), a number of lorikeets, and a few species of other bird groups, for a total of about 111 species that pollinate flowers; Stiles (1981) estimated that there were ca 310 nectar feeding birds in the whole Australasian region (this includes New Guinea), a figure that he thought was very definitely an upper estimate. Estimates in Fleming & Muchhala (2008) are 56 genera and 242 species, of which over 60% are honeyeaters, but perhaps only about half the honey eaters, ca 80 species, are nectarivorous, individual species varying considerably in their diet (see also Recher 1981; Higgins et al. 2008).

Keighery (1980, 1982) recorded about 21 species of birds visiting about 750 species of flowers in Western Australia alone (estimates lower in E. M. Brown et al. 1997, but pollinators listed); Western Australia may be the epicentre of bird pollination in that region. In monsoonal northwestern Australia, some 24 species of birds visited 116 species of plants in twenty eight families (Franklin & Noske 2000). Ford et al. (1979) thought that in Australia some 300 species of both Proteaceae and Myrtaceae were pollinated by about 100 species of birds, the brush flowers of Myrtaceae in particular being pollinated by lorikeets, and these may eat pollen (Stiles 1981 and references); all told, Ford et al. (1979) estimated that around there were around 800 species pollinated by birds in Australia. There are about 1,100 Australian species of Proteaceae, and many species, including members of the large genera Grevillea and Banksia, are likely to be pollinated by birds (Maynard 1995), and a moderately conservative estimate is ca 350 species are pollinated by birds (E. T. Miller pers. comm. 12.ii.2017). In addition, there are about 70 species of bird-pollinated Loranthaceae in Australia (Barlow 1984), while Toon et al. (2014) discuss the evolution of bird pollination of ca 37 species of Fabaceae-Mirbelieae/Bossiaeeae.

A (guess)timate is that around 120 species of birds pollinate 1,000< species of plants in Australia. (Cheke and Mann [2008] had suggested that honey eaters alone visited about 450 species of flowering plants from 100 families.) The proportion of ornithophilous flowers may be relatively high on that continent since the copious nectar produced incurs little cost to the plant, an advantage given the nutrient-poor soils so common there; there is less ornithophily in the more nutrient-rich east coast forests (Orians & Milewski 2007).

It is more difficult to estimate numbers of bird-pollinated plants outside Australia, although in the general area China and India to New Zealand the same groups of birds are involved. Here I do little more than list groups of plants where bird pollination may be expected to preponderate.

In the area from South East Asia to Malesia, there are a few groups of plants that are likely to be visited by birds. These include Ericaceae like Rhododendron, where 80 species from the island of Papua may be bird-pollinated (Stevens 1976; see also Corlett 2004), Paphia and Dimorphanthera, also largely Papuan and with around 75 species with red, tubular flowers, and the old genus Agapetes, with about 95 species in its centre of diversity in southwest China and adjacent Myanmar and India that also often have red, tubular flowers (see also below). In addition, species of the widespread Aeschynanthus (Gesneriaceae) typically have red, gullet-type flowers; there are about 185 species in the genus. There may be some 36 species of bird-pollinated Loranthaceae in China (Qiu & Gilbert 2003) and 165 in Malesia (Barlow 1998). The ca 14 species of the largely New Guinean Tapeinochilus (Costaceae), are likely to be pollinated by sunbirds (O. Gideon, in Salzman et al. 2015). However, E. D. Brown and Hopkins (1995) note the apparently unspecialised morphologies of many of the flowers visited by birds in a site they studied in southeastern Papua New Guinea. They describe "knob" flowers from Schefflera and "fluffy cups" from Syzygium, etc.; apart from one species of Loranthaceae, none of the 17 species of flowers they list as being pollinated by birds are in groups mentioned above, although they do include both Myrtaceae and Proteaceae, important nectar sources in Australia. The 17 species of plants pollinated by three nectarinids in a locality in Sarawak included Malvaceae-Bombacoideae, Musaceae, Sapotaceae, Myrsinaceae and Zingiberaceae (Momose et al. 1998); two spiderhunters pollinated 8 species of Zingiberaceae in four genera (Sakai et al. 1999b) as well as three species of Loranthaceae (Yumoto et al. 1997) at localities in Borneo.

New Zealand has a mere 12 species of pollinating birds, of which only three - two meliphagids and the white eye - were recorded as making almost 90% of the visits to flowers (Kelly et al. 2010). They are probably major visitors to 29 species of plants, and all told 85 species, or perhaps double that number, may be visited by the birds (Kelly et al. 2010; Lee et al. 2013). However, understanding plant/flower interactions in the islands is particularly difficult. One of the common plant visitors, the silver- or white-eye Zosterops lateralis, arrived from Australia in 1832 and 1856 (it is quite often a nectar robber - Anderson et al. 2011), three other species have suffered serious recent declines, and the stitchbird, Notiomystis cincta, is in a monotypic family, Notiomystidae, unrelated to any of the other passeriforms so far mentioned (Kelly et al. 2010). Furthermore, on some smaller islands in particular flowers with morphologies very unlike those of bird pollinated plants on the main islands are visited and apparently effectively pollinated by the two meliphagids and the stitchbird (Castro & Robertson 1997: Kapit Island, pollination release?), although this has also been noticed elsewhere in New Zealand (Kelly et al. 2010). Finally, understanding plant-pollinator relationships is difficult because of the extensive recent human-caused changes to the New Zealand avifauna, and seed set in a number of plants appears to be pollinator limited (Kelly et al. 2010 and references).

In New Caledonia some 17 species of Cunoniaceae, mostly Geissois, have red, brush-type inflorescences that are thought to be bird-pollinated (Hopkins et al. 2014), and there are ca 20 species of Metrosideros, also with brush-type inflorescences, on the island (Wright et al. 2000a; Pillon et al. 2015).

Africa, especially The Cape. There are about 200 species of probably bird-pollinated Loranthaceae in all of Africa (Polhill & Wiens 1998), while there are 64 bird-pollinated species of Iridaceae (Goldblatt & Manning 2006) and 13 species of Cyrtanthus (Amaryllidaceae) in southern Africa alone (Snijman & Meerow 2010). There are probably over 400 species of Aloe, nearly all in Africa, and many in southern Africa are bird pollinated (Rebelo 1997) although bees also pollinate a number (Symes et al. 2009; Hargreaves et al. 2012). Perhaps another 287 species of bird-pollinated plants can be added from southern Africa (Rebelo 1987: 424 species, from which Proteaceae, etc., have been removed). At a more local scale, Rebelo (1987) estimated that perhaps 318 species of plants were pollinated by six species of birds (including the sugarbird, Promerops caffra) in the Cape region alone of South Africa, while Rebelo et al. (1984) suggested that ca 86 species of Proteaceae in the South African Fynbos were pollinated by Promerops caffra, although nectariniid sunbirds also visited them (Rebelo 1987). The figure in Johnson (2010) are somewhat different, although high, ca 20 species (unspecified) of angiosperms are visited by Promerops but ca 44 species are visited by Nectarinia famosa and 66 species by Anthobaphes violaceae, the orange-breasted sunbird, both nectariniids.

Perhaps 950 African species are pollinated by birds, which are mostly Nectariniidae and a few Zosteropidae; the sugar bird is placed in Promeropidae. The relatively high frequency of ornithophily in the Cape flora may be connected with the prevalence of nutrient-poor soils, as in Australia (Rebelo 1987; Orians & Milewski 2007).

Age: Recent phylogenies are clarifying avian evolution. Old World pollinators belong to two major clades. 1. Psittaciformes include parrots and cockatoos, and Loriinae, mostly Australasian, are nectarivorous parrots, nectarivory having evolved at least three times (Schweizer et al. 2014). Their age has been estimated as the early Eocene ca 59 Ma (Fleming & Kress 2013; see also Fleming & Muchhala 2008), and late Palaeocene may indeed be the stem age of the clade (Prum et al. 2015). However, the crown-group age of parrots is late Oligocene, and the some 53 species of lories and lorikeets, by far the biggest clade of nectarivorous parrots, are a mere (14.8-)ca 10(-4.8) Ma (crown-group age) or (15-)13 Ma (stem group) (Schweizer et al. 2014, 2015).

2. The other clade is Passeriformes, a large clade including over half the species of extant birds which began diversifying towards the middle of the Eocene around 50 Ma (Prum et al. 2015: Acanthisitta, the rifleman, diverging from the remainder; ca 4 Ma older, Claramunt & Cracraft 2015) or, a rather different estimate, around 82 Ma (e.g. Barker et al. 2004, see also Ericson et al. 2014; c.f. Mayr 2013; Ksepa & Phillips 2015). However, Jarvis et al. (2014) estimated that the part of Passeriformes that includes all the Passeri or oscines below may not have begun diversifying until the Oligocene ca 30 Ma, with the nectarivorous taxa being embedded in clades that are Miocene in age (Prum et al. 2015); general estimates in Moyle et al. (2016) are in the same ball-park. Within the oscine Meliphagida, which started off in Australia, Moyle et al. (2016) estimate the time of divergence of Acanthorhynchus from other meliphagids at less than 12 Ma (see also Driskell and Christidis 2004; c.f. Marki et al. 2016). However, Joseph et al. (2014) and Marki et al. (2016) found that Myza, from Wallacea, was sister to other meliphagids, diverging 26.6-15.9 or ca 24 Ma respectively, and that the rest of the group, including Acanthorhynchus, did not radiate until ca 18 or 20 My; there was substantial diversification of the group accompanying the Miocenne aridification of Australia (Joseph et al. 2014). Within the Passerida, Yuhina, Timaliidae, is paraphyletic to white-eyes s. str., and the whole clade (= Zosteropidae/Zosteropinae) began diversifying 8.1-6.3 Ma, Zosteropidae s. str. diversified 5.6-4.5 Ma, and the speciose Zosterops itself a mere ca 1.8 Ma (Moyle et al. 2009). Within the finches, sunbirds and spider hunters are both members of Nectariniidae-Nectariniini while flower peckers are Nectariniidae-Dicaeini, and they began diversifying in the middle Eocene around 45 Ma; the stem group is perhaps Palaeocene in origin (Barker et al. 2004). On the other hand, comparable estimates in Fleming and Muchhala (2008) are 30 and 35 Ma respectively (see also Friis et al. 2011). Promerops has been mentioned on occasion; its stem age has been dated to ca 39 Ma (Fleming & Kress 2013) or (39.5-)33.4(-28.3) Ma (Beresford et al. 2005: note calibration).

Hawaii is noted for the remarkable radiation of Drepanididae/Drepanidinae, the Hawaaian honeycreepers, which were/are (about 60% the species have become extinct since human arrival on the islands ca A.D. 1250 - Wilmshurst et al. 2010; Ricklefs 2017) both nectarivorous and insectivorous. All told, about 7 extant species of Drepanidinae seem to be nectarivorous, to which can be added the 5 extinct species of the unrelated Mohoidae (look-alikes of Meliphagidae) that were endemic to the islands. These few species of birds are/were probably pollinators of some 178 species of plants, of which around 125 species alone are Campanulaceae-Lobelioideae (Brighamia and Cyanea), although the main nectar source is probably Metrosideros, especially the protean M. polymorpha (Carlquist 1970; Lammers & Freeman 1986; Givnish et al. 1995; T. J. Givnish pers. comm. x.2013).

Age: Hawaiian Drepanidinae are oscine cardueline finches, their ancestor being something like a rose finch. It is thought that much diversification of clades representing extant drepanid species took place 5.8-2.4 Ma, especially after the formation of Oahu ca 4 Ma (Lerner et al. 2011). However, Givnish et al. (1995) estimated that the age of the common ancestor of Cyanea (Campanulaceae), one of the main nectar sources for the birds, was of the order of 17.4-8.7 Ma, about three times estimates of the age of diversification of Drepanidinae, their pollinators. Pender et al. (2013) date the diversification of both birds and plants to within the last 17 Ma, while the arrival of Metrosideros, another important nectar source, on the islands has been dated to (6.3-)3.9 Ma, probably initially on Kaua'i (Percy et al. 2008).

To summarize some differences between Old and New World nectarivorous birds and the plants that they pollinate: Hummingbirds usually hover and tend to pollinate herbs and epiphytic plants, while the Old World nectar-feeding birds perch when feeding and more frequently pollinate larger trees (e.g. Stiles 1981; Kress & Beach 1994; Fleming & Muchhala 2008). Hummingbirds, all New World, are thought to be more specialized than both other New World flower pollinators like orioles and also their Old World ecological counterparts, and in the Old World, although African sunbirds show at least a moderate degree of specialization, honeyeaters and lories are rather generalized (Schweizer et al. 2014). Similarly, bird-pollinated flowers, at least, in the New World seem to be more specialized (Stiles 1981; Fleming & Muchhala 2008) than their Old World counterparts, and Papuan plant-pollinator relationships may be less specialised that those in Australia (E. D. Brown & Hopkins 1995). Johnson and Nicholson (2008; see also ) compared hummingbirds with other New World passerine pollinators and sunbirds with other African pollinators - thus neither Meliphagidae, honeyeaters, nor lories and lorikeets were involved in the comparison. Flowers visited by hummingbirds and by sunbirds tended to produce medium amounts of nectar with high concentrations of sugars, of which sucrose was a major component (ca 1/2), while flowers visited by other birds in both regions tended to produce larger amounts of more dilute nectar, with sucrose being at most 0.5% (Johnson & Nicholson 2008), indeed, it was found that the nectar of about a third of Old World species pollinated by birds had low overall sugar concentration, and that nectar was extremely sucrose poor, to the extent that Abrahamczyk et al. (2017b) toyed with the idea that pollinators visitng such flowers were being deceived. Assemblages of hummingbirds in any one site in the New World are more diverse than pollinator assemblages in the Old World (Fleming & Muchhala 2008), although lories can be quite diverse locally in New Guinea (Schweizer et al. 2015). Overall, the diversity of plants pollinated by birds is lower in the Oriental region than in the New World or the Australo-Papuan region (e.g. Corlett 2004).


For good introductions to bat pollination, see Dobat and Peikert-Holle (1985) and Fleming and Kress (2013).

Phyllostomid bats are relatively small and often hover when they feed, and the plants they pollinate are trees as well as vines and epiphytes; they are active in the l.t.r.f. as well as in deserts (Fleming et al. 2005, 2009). In any one site 1-6 species of bats may pollinate 4-19 species of flowering plants (Fleming et al. 2005). In the Monteverde forest, Costa Rica, 7 species of bats visited 33 species of plants, 1.6% of the total flora (Murray et al. 2000). Phyllostomines are important pollinators of columnar cacti, Agave, and Malvaceae-Bombacoideae, with 16 species recorded as pollinating 90 species of plants including some of these cacti (Arizmendi et al. 2002). Clairmont et al. (2014) found pollen of 11-14 species of plants (2.4-3.9/night) in the excreta of three phyllostomines in Cuba, although the figure was 17-21 species (3.5-5.3 species/night) in two species of fruit-eating bats; the study was carried out in the wet season. M. Sazima et al. (1999) found that Anoura caudifera pollinated flowers belonging to 3 families and 10 species (3 bromeliads) in a lowland site and 5 families and 7 species (2 bromeliads) at a montane site in the Brazilian Atlantic Forest.

Estimates are that in the New World at least 500 species of plants in 27 families (Vogel 1969) or 590 species of plants in 43 families (Dobat & Peikert-Holle 1985) are pollinated by bats. Although other estimates are substantially lower, e.g. 364 species in 44 families (Fleming et al. 2009), such figures need to be revised substantially upwards. Thus Fleming et al. (2009) list 20 bat-pollinated species in the Centropogon alliance (Campanulaceae-Lobelioideae), although L. Lagomarsino (pers. comm. 9.iii.2017) estimates that about 167 species of that group may be pollinated by bats and 110 species is the estimate in Dobat and Peikert-Holle (1985). Similarly, 5 bat-pollinated species of Passiflora are listed by Fleming et al. (2009), while Jørgensen et al. (2012) estimate the number to be 17, 8 known to be bat-pollinated, 9 likely to be so.

Age. The bats have diversified within the last 30-12 Ma (Datzmann et al. 2010), crown-group ages of the two major clades involved, Glossophaginae and Lonchophyllidinae, being around (20.8-)17.5(-14.2) and (11.4-)11.1(-10.9) Ma, with stem group ages of ca 21.6 and 20.6 Ma respectively (Baker et al. 2012). Comparable ages in Rojas et al. (2011) are 20.1-12.9 and 23.5-22 Ma respectively. Similarly, Fleming et al. (2009) date the evolution of bat-pollinated flowers to the Miocene ca 20 Mya (see also Fleming & Kress 2013).

IIIB. Pteropodidae. Pteropodid bats, about 173 species, are predominantly frugivores or nectarivores. Many pteropodids visit flowers and eat fruit, and on Samoa 2 species of bats visited 78 species of plants in 37 families, many of them for both flowers and fruits, and some also for leaves (Banack 1988). In the Old World there are only 5-6 genera and 12-15 species of nectar-feeding bats (again, other species visit flowers on a more opportunistic basis) in the macroglossine Pteropodidae (e.g. Marshall 1983; Fleming et al. 2009). The bats are found from Africa to Australasia and the Pacific (Fleming & Muchhala 2008). They are on average larger than phyllostomid bats, they tend to hold on to the plant when feeding, and the flowers they pollinate are robust and are usually borne on trees (Fleming et al. 2009).

Bats pollinate fewer than 200 species of flowers in the Old World (Dobat & Peikert-Holle 1985); 57 genera in 22 families is the estimate in Marshall (1983: records from only 19/44 genera of bats), 168 species in 41 families is the figure in Fleming et al. (2009).

Age: The crown-group pteropodid clade may be around 56 Ma (Fleming & Kress 2013), older than the phyllostomids, 38-24 Ma (Almeida et al. 2009) or 28-18 Ma (Teeling et al. 2005).

The New World phyllostomid bats are more efficient pollinators than hummingbirds, transporting larger amounts of pollen, and bat-pollinated flowers evolved from bird-pollinated ancestors (Muchhala & Thomson 2010 and references) - note, however, that pollen transport in trap-lining hummingbirds may be remarkably effective with multiple paternities in the plants they visit because the birds do not clean themselves very much and are aggressive (Krauss et al. 2017). The bats, all New World, are thought to be more specialized than their Old World ecological counterparts. Assemblages of phyllostomid bats in any one site in the New World are more diverse than in the Old World (Fleming & Muchhala 2008). Overall, the general diversity of plants pollinated by bats is lower in the Oriental region than in the New World or the Australo-Papuan region (e.g. Corlett 2004).

DISCUSSION. [This is still choppy...]

Introduction. Ollerton et al. (2011) estimated that the number of animal-pollinated plants as 308,006, 87.5% of some 352,000 species of flowering plants (the latter number from Paton et al. 2008). Van der Niet and Johnson (2012) estimated that one quarter of plant speciation events might involve pollinator differences. However, details of con- and heterospecific pollen movement and related topics like the establishment and efficacy of barriers to crossing between species are not the main issue here (see e.g. Armbruster 2014 for a recent review). Rather, the focus is on numbers of species of plants and animals involved in particular sets of pollination relationships, since this is one way of allowing us to think about ideas of co-evolution (the term will be defined precisely when used), mutualism, and the like.

Indeed, floral variation and plant-pollinator interactions have long been of central interest to biologists. Darwin (1876: p. 371) observed that the "beaks of humming-birds are specially adapted to the various kinds of flowers they visit", although one might now add the width of the tube as another important variable (Temeles 1996; Temeles et al. 2002, 2009). Stebbins (1970: p. 308) noted that in animals major groups tended to be rather invariant, differing in characters related to survival, and in plants reproductive features showed comparable invariance: The "flower must become a highly integrated stucture, with all of its parts precisely adjusted to each other" for cross pollination by animals with specialised habits to be successful. Fifty years or so ago plant-pollinator relationships were often thought of in terms of what might be thought of as mutual co-evolution, with almost lock and key-type relations between particular flowers and their pollinators, the two evolving together with the implication that the relationships between the two were almost one to one (e.g. Grant & Grant 1965), as the often-published pictures of birds with long, curved bills next to flowers with a long, curved corolla tube suggest. Stebbins (1970), although cautious, conveyed the same general idea, as when he noted that euglossine bees obtained fragrances from orchids, extensive speciation in both being the result, and also when he discussed intermediates between different pollination syndromes.

Unfortunately, the terminology here is complex. Ollerton et al. (2007) distinguished between phenotypic generalization and specialization, thinking about the plant, and ecological and functional specialization, thinking about the pollinator, ecological generalization emphasizing the numbers of pollinators, and functional generalization their (difficult to quantify) diversity. Thus a species pollinated by 100 species of flies might be ecologically generalized yet functionally specialized. As Ollerton et al. (2007: p. 725) noted, functionally and phenotypically specialised South American hummingbird flowers might be visited by more than one species of bird (also functionally and phenotypically specialized), and so were "to some degree ecological generalists", the birds likewise being ecological generalists. Similarly, Rebelo (1987) suggested that both plants and pollinating birds in the South African fynbos were generalists, while K. M. Cameron (2004) thought that most species of orchids attracted many species of euglossine bees, and most species of bees were attracted to several different species of plants. Pauw and Stanway (2014) found that pollinators in some South African communities were often specialists, 154/217 species visiting only one species of plant, but the plants were generalists, 38/62 species being visited by more than two species of pollinator (the former figures might change if other communities had been studied). And in Ollerton's earlier work on South African stapeliads, he noted that specialised flowers were specialized on common, ubiquitous insects, and flowers pollinated by the most specialized insects had easily accessible nectar that attracted other visitors, too (Ollerton et al. 2003).

As one tries to get one's brain around the terms, one also has to remember that there is much debate over related issues such as (1) the nature and extent of flower-pollinator co-evolution, discussed more below, and (2) the existence of wide-ranging pollination syndromes versus sometimes quite local guilds facilitating the success of both plant and pollinator (S. D. Johnson 2010; see also F. Zhang et al. 2012), (3), the suggestion that the very idea of pollination guilds and pollination syndromes is overly simplistic (e.g. Waser et al. 1996, 2018; S. D. Johnson & Steiner 2000; Pellmyr 2002; Waser & Ollerton 2006; Morales & Aiizen 2006; Olesen et al. 2007; Raguso 2008; S. D. Smith et al. 2009; Ollerton et al. 2009a for references; but c.f. in part Fenster et al. 2004; Willmer 2011; Quintero et al. 2016; Wilson et al. 2017; Lagomarsino et al. 2017; Serrano-Serrano et al. 2017; Guzmán et al. 2017; S. D. Johnson & Westra 2017; Dellinger et al. 2018; Vendelook et al. 2019: nectar composition; esp. Rosas-Guerrero et al. 2014), and (4) what exactly pollinators might see and respond to (Macior 1971; Waser et al. 1996; Chittka et al. 1999; Fenster et al. 2004; Waser & Ollerton 2006; Raguso 2008; Ollerton et al. 2009a; Schaefer & Ruxton 2009, 2010; Schiestl et al. 2010; Caves et al. 2018: categorical perception of colour in some birds); see also the papers in Ann. Bot. 113(2). 2014. Thus Rodríguez et al. (2004) and Horridge (2009) discuss the bee's point of view, the former, thinking about monosymmetry in particular, the latter more generally. Function(al) groups/types or guilds are groups of animals with similar mouthparts, for example, so generating similar selection pressures on the plants they pollinate (Armbruster 2014 for literature). Note, however, that a group of meliphagids with rather similar morphologies living in arid areas of Australia occupy ecological pollination space as fully as mesic meliphagids with different morphologies (E. T. Miller et al. 2016).

Oligolecty vs Polylecty, Generalization vs Specialization. Thinking about pollinator:plant relationships leads to a series of apparent paradoxes (see also Johnson & Steiner 2000). Snow and Snow (1980) noted that although a particular species of plant might have but a single hummingbird pollinator, i.e., the plant was a specialist, the same bird, specialized though it might be, might pollinate several species of plants, i.e. it was a generalist. Similarly, specialized fruit-eating phyllostomid bats are generalists when looking at bat:plant networks, a single species of bats may eat fruits of several species of plants (Mello et al. 2011a). Thus the flowers visited by polylectic pollinators like hummingbirds, orchid and bumble bees, and long-tongued flies are specialized, often being monosymmetric, with concealed nectar, and so on.

The plesiomorphic condition for pollination specificity in bees is likely to be oligolecty (Danforth et al. 2006; Sipes et al. 2006; Michez et al. 2008); polylectic behaviour in bees is often derived (e.g. Müller 1996; Sipes & Tepedino 2005; Danforth et al. 2006; Sipes et al. 2006; Larkin et al. 2008; Praz et al. 2008; Michez et al. 2008; Sedivy et al. 2008, 2013; Litman et al. 2011 and references: note early ages for bee diversification; also Danforth et al. 2013; c.f. e.g. Moldenke 1979 and references in Larkin et al. 2008). Bees initially pollinated one or a few species of plants, all more or less related, but clades of bees that visited a variety of unrelated plants evolved. Of course, there are relatively young, very speciose clades of plants that are commonly pollinated by oligolectic bees, Asteraceae being a prime example (q.v. for details). Around half of all bees, which total around 17,500 species, are oligolectic (Michener 2007; Larkin et al. 2008). Several species of oligolectic bees may pollinate a single plant species, and the floral morphology of the latter is likely to be rather unspecialized (see below), indeed, plants are not generally dependant on particular oligolectic pollinators for their pollination (Michener 2007). Floral specialization has increased over evolutionary time, and unspecialised flowers, probably pollinated by several species of oligolectic bees or other pollinators, precede specialised flowers, probably pollinated by one or a few species of polylectic pollinators. Monosymmetric flowers in which precise interactions between plant and pollinator are needed for effective pollination are largely a Caenozoic phenomenon, and many of them are likely to have polylectic pollinators, indeed, relationships with particular species of generalist pollinator can be quite precise. Note that in general bees are much less specific when nectaring (e.g. Michener 2007). In any event, one should not get too excited over this and related typologies (e.g. Benton 2017 for problems, etc.).

Oligolectic bees visit relatively few species of flowers, and/or a group of rather closely related plants, i.e. within a single family, or a tribe within that family, so they are specialized from that point of view (oligolecty often refers to pollen-collecting behaviour, not nectaring, where bees tend to be more catholic - Waser et al. 1996). The bees may have few obvious morphological adaptations for pollination (Michez et al. 2012), and the flowers they visit are often what would be described as unspecialized, with radial symmetry, poorly concealed nectar, etc., although oligolectic bees may also pollinate monosymmetric flowers (e.g. Bawa 1990; Sedivy et al. 2008; Benton 2017). The bees, which are often solitary, are in general most diverse in arid and often extratropical regions, they tend to be short-lived, and the flowering times of the plants they pollinate are also often rather short (Linsley 1958; Michener 1979). Thus in arid and semi-arid areas like deserts, the Great Basin, and parts of California and Chile there are many species of oligolectic/specialist bees (e.g. Waser et al. 1996) that at least sometimes compete for the same resource (see in part Moldenke 1976, 1979a, b; Petanidou & Ellis 1996; Lindberg & Olesen 2001; Stang et al. 2007). The flowers that they visit are accessible to a wide variety of pollinators. For example, the creosote bush, Larrea tridentata (Zygophyllaceae), common in the American southwest and with an open and "unspecialized" floral morphology, is the focus of visits by 22 species of oligolectic bees, as well as being regularly visited by another 22 species of polylectic bees, not to mention still other more transient visitors (Hurd & Linsley 1975). Similarly, 12 species of oligolectic Andrena bees are major visitors to Camissonia campestris (Onagraceae: Linsley et al. 1973: see also Cruden 1971: Nemophila; Ehrenfeld 1979: Euphorbia; Waser et al. 1996: Ranunculaceae). So rather paradoxically, individual specialist pollinator species may be less effective in and necessary for pollination than are generalists (e.g. Ehrenfeld 1979; Olesen 1997 and references); that a pollinator specializes says nothing about its success in pollinating.

In many situations one can think of both plant and pollinator as being specialists - and at any particular time and place, this may well be true. Certainly, from the point of view of the plant promiscuity of its pollinator may be more apparent than real. For example, even if a single polylectic (generalist) bee species, or colony, or even an individual bee, may visit many species of plants, on any one trip a particular bee may be much more selective (e.g. Heinrich 1976; Chittka & Thomson 1997; Heard 1999; Hagbery & Nieh 2012 for general pollen/nectar constancy), so being functionally mono- or oligolectic (specialist). Similarly, pollinators like hummingbirds may be widely distributed and pollinate many unrelated species of plants, but at any one place the plant-pollinator ratio is rather lower than the figures from, say, the whole of Costa Rica, might suggest (e.g. Rebelo 1987); situations where male and female birds have bills of different lengths will also lower the ratio. Interestingly, hummingbirds with smaller ranges show greater community-level, i.e. local, specialisation on flowers with different morphologies (Sonne et al. 2016). Specialization may also increase as resources decrease (Tinoco et al. 2016) and in species-rich hummingbird communities there are higher levels of specialization and modularity, as in communities where there is competition between closely-related hummingbid species (Martín González et al. 2015). (Note that on islands hummingbirds tend more to be generalists (Traveset et al. 2015), and the modularity of hummingbird communities is less (Martín González et al. 2015), and the plants are pollinated by a greater variety of pollinators (e.g. Serrano-Serrano et al. 2017)). In southern Africa Goldblatt and Manning (2006; see also Johnson 2010) estimated that there were often about six species pollinated by a particular long-tongued fly at any one locality - quite a number, but still substantially fewer than the 20-30 species that typically depended on that pollinator - and the pollen from different plants might be deposited on different places on the pollinator. Similarly, a single species of orchid bee may visit several species of orchid locally, but the pollinaria may be deposited on different places on the bee (N. N. Williams 1982).

In the well-known Diascia/Rediviva association, the extreme long-legged morph in Redivia is likely to have evolved about five times independently, but leg length was not correlated with environmental/geographical/biological variables examined, in particular, leg length did not seem to restrict the spectrum of oil host usage (Kahut et al. 2017). Indeed, Borrell (2005) had suggested euglossine bees with longer feeding appendages had access to a wider range of food resources, and from a short-tubed plant's point of view such bees visiting them would be nectar thieves.

Plant-Pollinator Interactions and Species Number Imbalances. There are marked asymmetries in the numbers of players in the plant-pollinator relationships discussed here (e.g. Bronstein 1994: question 2; S. D. Johnson & Steiner 2000; Johnson 2010: Table 1; Mello et al. 2012), and plant-pollinator ratios of 10:1 are at the low end of the spectrum. Wiens et al. (1983) expected to find similar asymmetrical relationships in plants with wind or water pollination, also in those pollinated by social bees, passerine birds, most flies, and perhaps beetles; they themselves looked at pollination of South African Protea species by small mammals.

These asymmetries are evident at all scales. Sekercioglu (2006) suggested that some 600 species of birds visited plants for nectar, 350 more being casual visitors, and that they visited 500 (3.7%) of a total of 13,500 plant genera (his estimate). By extrapolation, and using the same figure of 352,000 species of flowering plants and assuming that there is no variation in pollen syndrome within a genus(!), some 13,000 species may be bird-pollinated. Overall, one species of bird would pollinate around 21 species of plants. In the relatively well-studied Costa Rican flora, Stiles (1981) estimated that 55 species of hummingbirds pollinated mostly or exclusively around 300 species of plants. Bees are quite a diverse group, with some 17,500 species (Michener 2007), but it is particular groups of bees, not notably speciose, that play a disproportionately important role in current plant:bee interactions (see above; c.f. in part Cappellari et al. 2013: Fig 2B, C). In New Caledonia 43 species of native bees are the major pollinators of a flora of over 3,050 species of flowering plants, a plant:bee ratio of 71:1, and although this must be a considerable overestimate, it bears clarifying, while in New Zealand the ratio was estimated to be 57:1 (Donovan et al. 2013: but recent disruptions of plant-pollinator interactions caused by humans).

Smaller-scale examples show similar imbalances. Long-tongued dipteran Nemestrinidae in southern Africa have plant-pollinator ratios anywhere from 6:1 (highly conservative) to much in excess of 30:1; thus the nemestrinid Prosoeca ganglbaueri alone pollinates 20 or more species (van der Niet & Johnson 2012) and P. gangelbaueri and two tabanids between them are estimated to pollinate about 200 species of plants (Goldblatt & Manning 2000; Pellmyr 2002). Also in southern Africa, the satyrid mountain pride butterfly Meneris/Aeropetes tulbaghia pollinates 19 species of plants belonging to 8 genera and 4 families (Johnson & Bond 1994). In the Cape region of South Africa, Rebelo (1997) estimated that perhaps 318 species of plants were pollinated by five species of sunbird and the Cape sugarbird, Promerops caffra, although he thought that the 15:1 plant:bird ratio was particularly high; figures for a diversity of pollinators given by Johnson (2010) are 420:24. Thus Promerops is particularly attracted to 80 or more species of Cape Proteaceae (Rebelo 1997; Johnson 2010: 20 spp.), while perhaps 37 or 42 species (Geerts & Pauw 2009 and Johnson 2010 respectively), mostly other than Proteaceae, depend for their pollination on the malachite sunbird, Nectarinia famosa. The convolvulus hawk moth (25 spp.), tabanid flies (3 species pollinating 48 species of plants) (Johnson 2010), spider hunting wasps, Hemipepsis (Johnson 2010; Shuttleworth & Johnson 2012) which pollinate especially Orchidaceae, Apocynaceae:Asclepiadoideae and Asparagaceae:Scilloideae, and so on. In Western Australia, the New Holland honeyeater, Phylidonyris novaehollandiae, visits flowers of 142 species of plants in 32 genera and ten families (E. M. Brown et al. 1997). The long-billed hummingbird Ensifera ensifera pollinates 37 species of the Andean Passiflora supersection Tacsonia as well as other plants (Abrahamczyk et al. 2014), so there is some phylogenetic clumping of pollinated plants there. Interestingly, many of these examples are from the southern hemisphere (Johnson & Steiner 2000), and Johnson (2010) suggested that in Apocynaceae and Orchidaceae pollinator specificity was greater in southern Africa than in the northern hemisphere.

Species of the polylectic Apidae visit over twice as many families of flowering plants as do species of Halictidae and over five times those of Colletidae (Waser et al. 1996), and in the Mediterranean, Bombus and Apis were the most generalized visitors, visiting the most plant species (Olesen et al. 2007b). In the humid tropics polylectic bees, which often live longer, are proportionally more common than oligolectic bees (Michener 1979). Although in Venezuelan forests, at least, fruit set in specialized (monosymmetric, gullet types) flowers may be less than that in generalized flowers (Ramírez 2003), this cannot be linked to the particular pollinators involved.

In such asymmetric relationships the idea of pollinators and frugivores being keystone species or keystone clades that show phylogenetic niche conservatism (e.g. Fleming et al. 2005; S. D. Johnson 2010: see below) readily spring to mind. Many species of plants have specialized flowers and many species of bees, for example, are oligolectic, visiting a relatively few species of plants. Pollinator-plant relationships are not nested in any simple fashion: Specialist flowers interact more with polylectic pollinators, while generalist flowers often interact both with oligolectic and polylectic pollinators (e.g. James et al. 2012; c.f. Bascompte et al. 2003; Pawar 2014). For a polylectic pollinator that serves as a hub (when diagramming out plant-pollinator relationships) the effect of the extinction of a single species of plant may be slight, but the extinction of the pollinator may affect some of the plants it pollinates more seriously. For oligolectic pollinators, the relationship will tend to be the reverse; plant specialists or species with low numbers of interactions are more likely to go extinct than species with more diverse sets of interactions (Aizen et al. 2012; James et al 2012). The consequences of such changes will depend on the overall patterns of modularity and connectedness of the plant-animal relationships, connectance being notably higher in polylectic bumble bees or hummingbirds, less in agaonid wasps and oligolectic bees - and in the examples given by Jordano (1987) also euglossine bees (see also Waser et al. 1996; Lindberg & Olesen 2001; Rezende et al. 2007; Stang et al. 2007; Olesen at al. 2007a; Vamosi & Vamosi 2012; also Vásquez & Simberloff 2002: disturbance and pollination; Colles et al. 2009; Armbruster 2012). Thus the removal of a single species of bumble bee from a subalpine community in the Rockies perturbed general pollination relationships, even though there were other polylectic pollinators there (Brosi & Briggs 2013).

The high degree of asymmetry in pollinator/plant interactions is compatible with recent work demonstrating the apparent inevitability of the development of such asymmetries. In an experimental study, communities in which generalist species, more highly nested, developed as the population of the whole system increased, even as the whole became less resilient to perturbations, unstructured systems being more resilient (Fontaine 2013; esp. Suweis et al. 2013), although the number of species in simulations was held fixed in the latter study. [elaborate or delete: Rezende et al. (2007) suggested that extinctions might have some phylogenetic signal, but this is certainly not always true (Ramírez et al. 2011).

Co-evolution? Plants involved in the pollination asymmetries discussed here represent only ca 3% (11,380) of all animal-pollinated plants (308,006: Ollerton et al. 2011 above), although this figure is likely to be a considerable underestimate. A question is, how did the morphological features in the animal partner that we now see as the features of a supergeneralist/"specialized" pollinator arise? That there are pervasive interactions between plants and their pollinators is incontestable, but how did the morphologies of the two evolve (Thompson 2009; Guimarães et al. 2011)?

We often think that relationships between animals and plant directly reflects the past, with plants and their pollinators, for example, being involved in some kind of evolutionary pas de deux that results in the apparently co-adapted morphologies of the two that we see today. Individual plant-pollinator interactions may be very precise, witness the deposition of pollinaria by the orchid Catasetum on a visiting euglossine bee, and the complex morphologies of the staminate and carpelate flowers of this orchid that ensure the correct placement of the pollinaria (for which, see e.g. Darwin 1862a). Plants and hummingbirds appear be mutually adapted, there having been "complementary trait evolution" (Maglianesi et al. 2014a: p. 3325). But that characters are complementary in plant and pollinator does not mean that they have been equally labile evolutionarily. The imbalances between the numbers of species of plants pollinated and the numbers of species of their pollinators almost by itself demonstrates that strict co-evolution has not occurred: "Since most pollinator species visit a broad range of host plant species ... morphological traits in pollinators are only rarely thought to have evolved in direct response to a specific plant morphology" (Kahnt et al. 2017: p. 95). What are the relationships that might be considered co-evolutionary, and what groups of plants and animals show them (In the following, I do not discuss fitness trade-offs, for which, see Armbruster 2014 and references)?

1. The majority of examples given in the seminal paper by Ehrlich and Raven (1964) had to do with insects and the food plants that their larvae ate. This particular form of coevolution has been described as "escape and radiate coevolution" (Thompson 1994, but see Marquis et al. 2016 who link Ehrlich and Raven's ideas to scenarios more like 2 below). A plant develops defences against an insect, reducing herbivory, but the insect then evolves counter-measures to the defences, neutralizing them, and then speciates on the plant clade - which may develop further defences, and so on. The ability to eat the plant can be thought of as a key innovation, subsequent diversification may involve conventional geographical isolation and divergence (Althoff et al. 2014). There are many cases where an insect previously unassociated with a plant becomes able to develop on it, and a good example of this kind of coevolution is pierid butterflies that have diversified particularly on Brassicales, especially Brassicaceae (see also Pichersky & Raguso 2018).

2. In simple coevolution, that is, co-cladogenesis or parallel cladogenesis (Janzen 1980), both plant and pollinator/frugivore diversify together, there being some kind of reciprocal evolution. However, there is little evidence for such a mutual 1:1 co-evolution in plant-pollinator interactions (Jordano 1987; Waser et al. 1996; I. Sazima et al. 1996; Chittka et al. 1999; Fenster et al. 2004; Waser & Ollerton 2006; Raguso 2008; Placentini & Varassin 2007; Winkler & Mitter 2008; Ollerton et al. 2009a; Fleming & Kress 2012: pp. 182-188). Coevolutionary diversification is reciprocal natural selection that leads to an increased diversification of both parties, but, as Althoff et al. (2014: esp. p. 88) showed, this is very difficult to demonstrate (for a good example of co-evolution in Heliconius butterflies that are Müllerian mimicks, see Hoyal Cuthill et al. 2019). Plant-pollinator relationships do not seem to involve reciprocal bursts of speciation (Hembry et al. 2014).

3. Sequential evolution/radiation, plants diversifying earlier that their bird or bat pollinators, but with an overall similar pattern of cladogenesis, has also been invoked to explain both plant-pollinator relationships (Fleming & Muchhala 2008: plant family ages used). Fleming and Kress (2013: esp. table 5:3) list a number of examples of pollination and frugivory where they think particular plant groups become ""colonized" by a group of animals which then radiates in parallel with the plants" (ibid., p. 183). Their examples include phyllostomid bats with Agavaceae and columnar cacti, Hawaiian Campanulaceae with honeycreepers, Asian Loranthaceae with Dicaeidae, frugivorous phyllostomid bats with Piper, Araceae, Cyclanthaceae, and Ficus, and Southeast Asian hornbills/birds of paradise with Meliaceae/Myristicaceae. (Similar delayed sequential speciation - the delay is 65-170 Ma - has also been invoked in the relationship between the mycoheterotroph Thismia and its associated glomeromycote fungal symbiont - Merckx & Bidartondo 2008.) However, it is unclear how sequential cladogenesis might work (see Fleming & Kress 2013: Fig. 5.4b; c.f. Winkler & Mitter 2008).

4. In a more diffuse coevolution one or both partners is "represented by an array of populations that generate a selective pressure as a group" (Janzen 1980: p. 611). Complex geographical patterns of infraspecific variation in pollinators and the plants they pollinate are consistent with such an idea (e.g. Cotton 1998: hummingbirds; Steiner & Whitehead 1990, 1991: oil-collecting bees; Johnson 2010). Species of both plants and animals converge and specialize on a core set of mutualistic traits, rather than directly on each (Thompson 2005: p. 289; Fleming & Kress 2013), and there is no suggestion that a single clade of either plant or animal is involved. Perhaps a variant of this is where a group of plants are pollinated by the one pollinator: "Reciprocal selection drives the evolution of the community" of pollinators, the set of associations in which the insect is involved subsequently enlarging as "additional species become attached to the network of interacting mutualists by convergence" - the insect here was a long-tongued fly, Moegistorhynchus longirostris (Pauw et al. 2009: p. 268).

5. A plant may also evolve to "fit" the morphology and/or senses of the pollinator, advergent evolution (Johnson et al. 2003; Anderson & Johnson 2009; Pauw et al. 2009: the converging species above; S. D. Johnson 2010; Newman et al. 2015). Many of the asymmetries discussed here can be interpreted as representing groups of plants that have independently converged on the morphologies of what are now super-generalist pollinators; the plants may evolve, but their pollinators are less likely to (Guimarães et al. 2007, 2011, p. 881: "non co-evolutionary cascading events"; Thompson 2009). Wiens et al. (1983) included such relationships in their idea of unilateral evolution, in which plants evolve profoundly altered floral features to attract non-specialised (from one point of view, but see above) pollinators, while Stiles (1981) contrasted the behavioural flexibility of the pollinator (hummingbirds in this case) with the evolutionary flexibility of the plant. Long-tongued euglossine bees that pollinate a variety of Zingiberales, for example, have less specialized on these flowers than the plants have specialized on long-tongued pollinators (Borrell 2005; see also S. D. Johnson 2010), and relationships between orchids and their pollinators, although often quite precise, are unlikely to signify reciprocal co-evolution (e.g. N. H. Williams 1982; Ackerman 1983; T. Jermy in Szentesi 2002; Jersáková et al. 2006; Ramírez et al. 2011). As Lagomarsino et al. (2017: p. 1970) note when discussing the pollination of Andean centropogonid lobelioids, "floral morphological diversity is extremely labile, likely resulting from selection imposed by pollinators". Earlier, Whittall and Hodges (2007) showed that there had been directional evolution in spur length within the 25 North American species of Aquilegia, with floral morphology evolving to fit the morphologies of their pollinators, which included hummingbirds and sphingid moths. In Cirrhaea and Stanhopeinae, Pansarin et al. (2018) found that one species of orchid could be pollinated by two or more species of bee, and one species of bee visited two or more species of orchid; as they noted (ibid. p. 436) "[these] orchids depend unilaterally on bees". Abrahamczyk et al. (2016: p. 112) reached somewhat different conclusions, noting that in asterids nectar sucrose proportion and pollinator groups were correlated, "in a manner dictated by pollinator behaviour" and dietary requirements, but they also thought that nectar sucrose proportion was an adaptation of the plant to the pollinator group.

In pollination niches, positive interactions occur in the plants involved as they adapt to the pollinator (Johnson 2010), and this is associated with the idea of pollination guilds, "large numbers of species (typically about 20-30) depending on a single pollinator species or small set of pollinators of the same functional type" (Johnson 2010: p. 504). Trait-tracking by rare or non-rewarding species in a diffuse co-evolutionary setting (Johnson 2010) would also fit here.

There may be variation in the pollinator that has a geographical component, but there is no speciation, although there may be infraspecific variation in corolla tube length in the plant when different pollinators become involved (Newman et al. 2014). Thus ages of patterns of infraspecific variation in the bee-fly Megapalpus capensis show quite extensive covariation with biome ages, but there is just the one bee-fly species, and it pollinates 12 species or so (7 separate clades - de Jager & Ellis 2016 and references; Struck 1997, see Röschenbleck et al. 2014) in various biomes.

Knowing the relative timing of plant and pollinator diversification is critical in determining the evolution and nature of plant-pollinator relationships - unfortunately, given the state of dating phylogenies. One example that may be quite old is the Heliconia-hummingbird relationship. The predominant pollinators of New World Heliconia are hummingbirds and most Heliconia are from the New World; Heliconiaceae make up one of the single most important hummingbird-pollinated clades. Most estimates of the age of crown-group Heliconiaceae are 40-30 Ma - 32-28 Ma (Kress & Specht 2006), ca 32 Ma (McKenna & Farrell 2006), or (47-)39(-32) Ma (Iles et al. 2016). Diversification in New World Heliconia is most evident from a little over 30 Ma onwards (Iles et al. 2016). Diversification of crown-group hummingbirds has been dated to as late as the early Miocene (24.7-)22.4(-20.3) Ma, much speciation occurring about 13-12 Ma along with the uplift of the Andes (Bleiweiss 1998a; McGuire et al. 2007, 2014; Abrahamczyk & Renner 2015; Prum 2015), however, Tripp and McDade (2014a) estimated crown-group diversification of hummingbirds to have begun (29.9-)28.8(-28.4) Ma. Iles et al. (2016) suggest that there was diffuse coevolution (sensu Tripp & McDade 2014a; see also Ehrlich & Raven 1964 - ecological interactions between two parties driving adaptations in both) between hummingbirds and Heliconia, the two becoming associated early in the history of the former in the New World. Some changes in bill length and flower curvature may have involved mutual changes, as perhaps in situations like the pollination of Heliconia by Eulampis (= Anthracothorax) (e.g. Temeles & Kress 2003), and there certainly seems to have been divergence of body mass within the latter genus on the islands (Dalsgaard et al. 2018). However, much of the early fossil history of stem-group hummingbirds is from Oligocene Europe in deposits ca 34.3 Ma (Mayr 2004, 2009, 2016; Louchart et al. 2008) and from the Late Eocene of the Caucasus (Louchart et al. 2008). Furthermore, Heliconia itself may have a very long stem history - ca 45 Ma (Iles et al. 2016), about which precisely nothing is known.

Be all this as it may, hummingbirds may have been the templates, as it were, for a variety of younger plant clades as they adopted bird pollination - hummingbird pollination was "facilitated by this pre-existing relationship" (Iles et al. 2016: p. [12]), "evolutionary fitting" (Janzen 1985) of the plant to the bird being likely Abrahamczyk et al. (2017a) noted that the ages of plants in the five mutualisms they examined (Chile to California) were all more or less substantially younger than that of their pollinating hummingbirds. Indeed, asynchrony in timing of the evolution of the two partners in relationships that may now seem to represent an obligate association for at least one of the two (e.g. Fleming 2004; Ramírez et al. 2011; Schiestl & Dötterl 2012; Dalsgaard et al. 2018) suggests that there was no reciprocal co-evolution. If plants diversified before their pollinators, for example, this would suggest that their current pollinators are "exploiting" a mutualism that had evolved with other pollinators, and if the pollinators diversified first it is the plants that are doing the exploiting. In a summary of studies examining timing of parallel insect-plant phylogenies - here the insects are all herbivores of one sort or another - Winkler and Mitter (2008) found that equivalent ages of the two partners had only rarely (four studies only) been demonstrated, in seven cases the insect radiation was younger than that of the host plant, and there were no examples of the opposite relationship.

Members even of some of the classic examples of apparently strict co-evolution show disparities in diversification times, and mutual reciprocal evolution between plant and pollinator, cospeciation, seems to be the exception rather than the rule. Figs (q.v.) may be a partial exception (e.g. Cruaud et al. 2012a, b and literature), although more than one species of wasp pollinates a single species of Ficus in about 30% or more cases, indeed, host switching rather than cospeciation may predominate as a speciatiom mechanism here (Stadler et al. 2019). Much of the divergence in Yucca seems to have occurred before that of its main pollinator, Tegeticula, but only a mere 6-4 Ma, indeed, given the vagility of the moth, it is difficult to imagine how strict co-evolution might work. Initial diversification in Yucca may have been in association with Parategeticula, a poor flier and now rather uncommon (Althoff et al. 2012). Similarly, crown-group euglossine bees can be dated to 42-27 or 38-34 Ma with especially rapid diversification 20-15 Ma (Ramírez et al. 2010) or (35-)28(-21) Ma (Cardinal & Danforth 2011). Ages of three immediately unrelated clades of orchids that these bees pollinate suggest that they speciated up to 12 Ma later, (31-)27-18(-14) Ma (Ramírez et al. 2011). In the apparently less tight association of aroids with scarab beetles, the beetles are thought to have diversified well before the plants (Schiestl & Dötterl 2012).

In several of these quite small but very successful pollinator groups that pollinate many species of plants it can be difficult to maintain that there is much evolution of the pollinator going on at all (see esp. Thompson 2009; Guimarães et al. 2011; Suchan & Alvarez 2015). As Kay and Sargent (2009: p. 638) noted "a plant's reproductive success is tied to the attraction and manipulation of its floral visitors". Plants exploit pre-existing perceptual/sensory biases of the pollinator, effectively manipulating their behaviour, resulting in widespread parallelism in pollination syndromes (= floral convergence), and both plant and pollinator are benefited (e.g. Ackerman 1983; Chittka 1996; Schaefer & Ruxton 2009, 2010; Ramírez et al. 2011; Schiestl 2010; Schiestl et al. 2010; Johnson & Jurgens 2010; Ramírez et al. 2011; Jürgens et al. 2013; Schiestl & Johnson 2013; Shrestha et al. 2013; Hembry et al 2014; Borghi & Ferni 2017; Reyes et al. 2018: focus on five floral characters; c.f. Strong et al. 1984), and such plant responses can happen very rapidly (Gervasi & Schiestl 2017: bumble bees and Brassica rapa). Thus Dyer et al. (2012) found that flower reflectance curves in Australia matched those in the northern hemisphere, and had probably evolved to match the colour discrimination abilities of their hymenopteran pollinators, which did not differ between the hemispheres; the evolution of the hymenopteran visual system predates angiosperm evolution. Similarly, colour perception is not an apomorphy for bees (Chittka 1996) while the sucrose-rich nectars that characterise hummingbird-pollinated flowers are associated with the ability of hummingbirds, but not starlings, thrushes, swifts, mocking birds, etc., to digest sucrose (Baker et al. 1999) and with the very ability of the birds to taste sweetness (Baldwin et al. 2014). Darwin (1876) is turned upside down; the corollas of various species of plants are specifically adapted to the beaks of the particular hummingbirds that visit them, not vice versa, indeed, the morphology of the pollinator may be more a witness to past than to present plant:animal interactions - although this simply pushes one aspect of the problem back in time. Sensory exploitation also occurs in floral mimicry, advergent evolution, pollination by deception of one sort or another (e.g. Schaefer & Ruxton 2009; Moré et al. 2013; Whitehead et al. 2018: c.f. Batesian mimicry), and this includes cases where plants produce odours that attract saprophagous, necrophagous and coprophagous insects (Schiestl & Dötterl 2012; Schiestl & Johnson 2013; Jürgens et al. 2013). Such mimicry is widespread, with about 10,000 species of orchids alone being examples (Ackerman 1986). The plants evolve, their models do not, and benefits here are very largely one-sided.

If the effects on the pollinator in such interactions were independent of some of the other interactions in which it is involved, there could be evolutionary specificity in the response of the animal (see in part Bawa 1990). However, this seems to be uncommon. In southern Africa a number of species of plants belonging to unrelated families depend on the services of a small group of nemestrinid pollinators (S. D. Johnson 2010; see also Huang & Shi 2013 and above), and Pauw et al. (2009) described rather diffuse co-evolution between individual species of these flies and and a group of species with long-tubed flowers. Interactions here take place in a complex geographical mosaic, with some of the flies having considerable infraspecific variation in proboscis length and a number of pollination ecotypes (e.g. Pauw et al. 2008; Newman et al. 2014; van der Niet et al. 2014; Anderson et al. 2014; see also Bascompte & Jordano 2007). This relates to the idea of functional guilds, groups of animals with similar mouthparts, for example, that generate similar selection pressures as a group on the plants that they pollinate (Armbruster 2014 for literature).

Suchan and Alvarez (2015 and references) suggest that any mutual coevolution that would lead to the pollinator developing particular adaptations might happen over a relatively short period, a brief co-evolutionary fling, as it were. Moreover, today's pollinators could have ecologically "taken over" older clades in which there were already established animal-plant relationships (suggested by Janzen 1980), with current plant-pollinator relationships as it were superposed on an earlier set of relationships. For example, the earliest evidence of birds visiting flowers comes from the Middle Eocene of Germany ca 47 Ma where eudicot (tricolpate) pollen was found in the gut of Pumiliornis tesselatus. Pumiliornis, although of uncertain relationships, is not close to any of today's major clades that pollinate flowers, such as parrots, sunbirds and hummingbirds (Mayr & Wilde 2014). Indeed, whether any extant clade of bird-pollinated plants dates back to 47 Ma is debatable, but if they do they would have had to switch their partners (Mayr & Wilde 2014). Indeed, stem-group, but quite highly derived, hummingbirds are first known from Europe in the Oligocene only some 34.3 Ma (e.g. Mayr 2004; Louchard et al. 2008; Mayr & de Pietri 2014). It has been suggested that plants like Ericaceae-Vaccinioideae-Agapetes and Bignoniaceae-Tecomaria capensis have evolved in the Oligocene in the Old World and have flowers that were originally pollinated by hummingbirds, but now by sunbirds, Nectariniidae, while hummingbirds moved to the New World where they and their flowers are now found (Mayr 2005, 2009, 2016). Although bird-pollination is common in the [Styphelioideae + Vaccinioideae] clade, there is no evidence that this was the original condition for the clade, which may be only ca 37.8 Ma (Wagstaff et al. 2010). Similarly, although diversification of Nectariniidae has been dated to ca 45 Ma in the Eocene (Barker et al. 2004), Jarvis et al. (2014) suggest that that the part of the very speciose Passeriformes to which Nectariniidae belong may not have begun diversifying until the Oligocene ca 30 Ma, while estimates from Prum et al. (2015) are still younger, less than 20 Ma. Bird pollination in the Australian bacon-and-egg peas, [Mirbelieae + Bossiaeeae], has evolved many times, and diversification of the meliphagid pollinators seems to be contemporaneous with that of the peas, crown-group ages for origins of bird pollination in the latter being (23.2-)16.8(-10.4) Ma (cpDNA data) or (19.2-)13.4(-7.6) Ma (ITS) (Toon et al. 2014), while the crown meliphagids are dated to 29.4-15.9 Ma (Joseph et al. 2014).

There is no evidence that hummingbirds and plants they pollinated co-evolved in Europe and then moved to the New World. Although the early Eocene avian faunas of Europe and North America are rather similar, neither crown nor stem hummingbirds are known in North America that early (Mayr 2009). Old - ca 30 Ma - clades of hummingbird-pollinated New World plants would be consistent with Eocene bird pollination. Crown-group diversification in hummingbirds is estimated at (24.7-)22.4(-20.3) Ma, much speciation occurring about 13-12 Ma along with the uplift of the Andes (Bleiweiss 1998a; McGuire et al. 2007, 2014; Abrahamczyk & Renner 2015; Prum 2015), however, other estimates of crown-group diversification are somewhat older, (29.9-)28.8(-28.4) Ma (Tripp & McDade 2014a). Turning to the plants they pollinate, most estimates of the age of crown-group Heliconiaceae are (47-)39, 32(-28) Ma (Kress & Specht 2006; McKenna & Farrell 2006; Iles et al. 2016) while hummingbird pollination in Gesneriaceae has been dated to (25.5-)18.5(-5) Ma (Serrano-Serrano et al. 2017). Diversification in two groups of bird-pollinated New World Acanthaceae may have been more recent than that of their hummingbird pollinators, e.g. crown-group neotropical Ruellia, in which bird pollination is common, is ca 9 Ma (Tripp et al. 2013c; Tripp & McDade 2014; McGuire et al. 2014; Tripp & Tsai 2017). Of the five examples studied by Abrahamczyk et al. (2017a), in three (Loranthaceae-Tristerix, Ericaceae-Arbutus, Centropogon-Campanulaceae) the hummingbirds are thought to be notably older than the plants that they pollinate, in two (Grossulariaceae-Ribes, Asteraceae-Chuquiraga) the ages were about the same. A mismatch is less clear in the sword-billed hummingbird, which split from other hummingbirds ca 11.6 Ma (McGuire et al. 2014), while the stem age of Passiflora section Tacsonia whose species it largely pollinates is ca 10.7 Ma and the crown age for the bird is ca 7.1 Ma (Abrahamczyk et al. 2014: many support values rather low). The fifteen species of hummingbirds in the West Indies pollinate some 101 species (90 endemic) of plants in associations dated 9-5 Ma, and bird and plant diversifications are roughly contemporaneous (Abrahamczyk et al. 2015). However, Dalsgaard et al. (2018) noted that much of the evolution in bill length had occurred before the likely arrival of ihe birds in the Antilles, specialization of the birds on the plants there depending on factors such as topographic complexity, high preciptation and numbers of both hummingbird and plant species - for the most part ecological fitting (Janzen 1985) rather than mutual coevolution.

Recent work on hummingbird pollination in North America north of 24oN and southern South America south of 27oS (Abrahamczyk & Renner 2015) can perhaps be understood in the same way. Seven species of migratory hummingbirds pollinate ca 130 species of plants in the western United States alone, particularly at high elevations (Grant 1994; Grant & Grant 1968; J. H. Brown & Kodric-Brown 1979). These associations are dated to 9-5 Ma or perhaps even younger, there being independent evolution of migratory behaviour in clades of bee hummingbirds ca 6.8 (stem) and 5.6 (crown) Ma (Abrahamczyk & Renner in Abrahamczyk et al. 2015; Abrahamczyk & Renner 2015; Licona-Vera & Ornelas 2017). The ages of plant and associated hummingbird diversifications in both North America north of 24oN and South America south of 27oS are thought to be largely contemporaneous, although much older in S. South America (Patagona and Sephanoides both ca 15 My) than in North America (Abrahamczyk & Renner 2015). Bird pollination in neither area led to massive diversification within particular clades of plants, but there was repeated evolution of bird pollination in unrelated clades. Thus individual clades of plants pollinated by hummingbirds were usually small, and only 8/ca 70 hummingbird-pollinated plant clades in North America and 0/35 clades in South America include five or more species (Castilleja is an example of moderate diversification). Only 18 (8 are Bees) and six species respectively of birds are involved, yet P. Wilson et al. (2006, 2007; see also Wessinger et al. 2016) estimated that up to 21 shifts from bee to bird pollination occurred in Penstemon s.l. alone, where over 40 species are pollinated by hummingbirds; Abrahamczyk and Renner (2015) estimated ten shifts, but either way the plant clades involved are small. Speciation within these little clades may primarily be geographic, despite the extensive gene flow probably mediated by the birds (Abrahamczyk & Renner 2015). One-on-one coevolution is unlikely here since no single species of birds uses just a single species of plant as a nectar source (Abrahamczyk & Renner 2015), and although such dependencies seemed to occur, albeit uncommonly, in some other hummingbird-plant interactions, this was for only part of the year (Abrahamczyk et al. 2017a). Although migratory birds sometimes/quite often visit plants the flowers of which lack the bird pollination syndrome (Waser et al. 2018; Martín González et al. 2018), and this may complicate bird-plant interrelationships, however, it does not affect the overall story here.

diffuse coevolution on plant traits by interaction networks of animals, animal traits change at the same time (Poelman & Kessler 2016: plant-insect interactions)

The diversity in bill length and pollinating behaviour of extant hummingbirds is unlikely to be the result of interactions between particular extant plant and bird clades. There cannot be simple one-on-one co-evolution since many more independent clades of plants than birds are involved (e.g. Abrahamczyk & Renner 2015). Indeed, it is unlikely that there is even loose coevolution - "a long shared evolutionary history" - between guilds of hummingbirds and guilds of the plants they pollinate (Cotton 1998: p. 639), even here Cotton noting that "the floral traits of many flowers have undoubtedly evolved to exploit hummingbirds as pollinators" (ibid.: p. 645; see also Borrell 2005). Abrahamczyk et al. (2017a: p. 1847) note "the apparently rapid 'addition' of further plant species" to the Centropogon/Eutoxeres association, described as a co-evolutionary mutualism in which the partners seem to be perfect fits for each other, but they note that the bird is much older than the plant, ca 21.5 vs 3.6 Ma old.

These are largely temperate bird-plant associations in a group that is predominantly lowland and montane tropical, and the plants in North America are serviced largely by hummingbirds that migrate over long distances, while elsewhere in the range of the family any movement of the bird is usually over shorter distances, so the representative nature of such associations may perhaps be questioned, but, given the discussion above, this seems unlikely. Thus most of the neotropical hummingbird pollinated genera in Rubiaceae are unrelated, with 39 or more cases of adoption of bird pollination, and the clades of plants involved are small (C. M. Taylor, per. comm. 2015). Although Roalson et al. (2007), Perret et al. (2003, 2007), Martín Rodriquez et al. (2009), and Clark et al. (2015) note that clades of Neotropical Gesneriaceae tend to be small, this is not always the case (see ). There may be larger clades in Heliconia (see Iles et al. 2016), Centropogon (e.g., a clade of 34 spp. studied by Abrahamczyk et al. 2017a), Ruellia, the Aphelandra pulcherrima group, and other Acanthaceae (Tripp & McDade 2014a and references), montane neotropical Vaccinieae, and Erythrina that are pollinated by hummingbirds, and in Parkia, Stenocereus, Agave subgenus Agave and Burmeistera clades of plants are pollinated by bats.

It has been suggested that oligolecty may facilitate the evolution of large numbers of sympatric species of plants because each is pollinated by different species of pollinators (e.g. Linsley & MacSwain 1958), however, the behaviour of many polylectic pollinators could have the same effect, since at any one time or place they may visit only one or a few species of plants. Schemske and Bradshaw (1999) in a classic paper discussed possible links between pollinator behaviour and pollination preferences of hummingbirds and bumble bees as drivers of speciation (see also Gegear & Burns 2007, floral features considered more or less separately).

9A2. Seed Dispersal.

Bats in which fruits are a major component of the diet are phyllostomids (New World) and pteropodids (Old World), and together they account for about 350 species, about 30% of all bat species (Baker et al. 2012).

Major diversification of fruit-eating phyllostomid bats, a clade of 68 species and twenty genera (Carolliinae, Glyphonectarinae, Rhinophyllinae, Stenodermatinae), began (27-)22(-18) Ma in the late Oligocene to mid-Miocene (Datzmann et al. 2010; see also Rojas et al. 2011) or (18.6-)17.0(-15.4) Ma (Baker et al. 2012), the stem-group age being around 17.0 Ma (Baker et al. 2012). The fruit-eating habit may have evolved three times or so. Of the more obligate frugivores Carollia has a stem-group age of about 19-17.5 Ma, while that of the [Stenoderma + Ariteus] clade, 8 species in eight genera, is ca 10.3 Ma, with a crown-group age of (6.2-)5.4(-4.6.) Ma (Baker et al. 2012, q.v. for other estimates, all broadly similar; slightly younger in Rojas et al. 2011). They eat soft fruits of secondary species as well as harder, more fibre-rich fruits of plants of primary growth and in the canopy (Muscarelli & Fleming 2007; Rojas et al. 2011). Phyllostomids may take nectar from 500 genera in 27 families (). The bats can be very abundant, and species of New World phyllostomids are wide-ranging compared to Old World pteropodids (Muscarella & Fleming 2008).

Old World fruit-eating bats (pteropodids) tend to disperse seeds of later-successional trees in families like Sapotaceae, Meliaceae, Arecaceae and Rubiaceae (Muscarella & Fleming 2008), and these tend to be canopy trees with larger fruits (least so for Rubiaceae). Pteropodids tend to be larger (less so in New Guinea) and they more commonly fly long distances (Cristoffer & Peres 2003). They take fruits from 144 genera of plants in 55 families (Marshall 1983: records from 19/44 genera of bats). On Pacific islands where they occur they may be represented by but a one or two species that are the major dispersers of 17 species of plant or more (Samoa: Banaack 1988; Vava'u, Tonga: McConkey & Drake 2015). Crown dates for pteropodids are around (29-)24(-20) Ma (Teeling et al. 2005) or (30.3-)25.1(-22.7) Ma (Bininda-Emons et al. 2007).

The ecological importance of these bats in terms of the services they provide plants is considerable (Freeman 2000; Muscarella & Fleming 2008).

Figs are commonly dispersed by bats (Muscarella & Fleming 2008) and birds (Snow 1981). New and Old World bats search for food in different ways and have selected figs with different qualities; overall fig morphology is quite diverse. In the New World some 21 species of Artibeus bats (phyllostomids) are the predominant ficivores. The bats are slow feeders and spit out larger seeds, fibre, etc., but they commonly disperse the tiny fig achenes. Like other bat-dispersed taxa in the New World, including Cecropia (Urticaceae) and Trema (Cannabaceae), particularly in Mexico (Lobova et al. 2009), species of Ficus can be found in early successional communities (Muscarella & Fleming 2008), and the altitudinal ranges of the bats and figs are similar (Fleming 1986). Bat-dispersed fruits in the New World often tend to be greenish or dull-coloured and odoriferous (Compton 1996 [a whole series of papers]; Korine et al. 2000; Shanahan et al. 2001; Harrison 2005). In the Old World fruits dispersed by pteropodid fruit bats are often yellow, etc., and are sometimes quite large (Kalko et al. 1996; Harrison & Shanahan 2005; Lomáscolo et al. 2008, 2010); bat-dispersed figs in Malesia, also dull-coloured, may not smell (Hodgkinson et al. 2003).

Although it has been debated to what extent figs really are keystone species, they are certainly eaten by a great variety of vertebrates (Shanahan et al. 2001). Kattan and Valenzuela (2013) summarize the information pro and con the importance of New World Ficus as a food resource, noting that in some places the local figs, although at times producing large amounts of food for frugivores, nevertheless did not do so throughout the year. However, "[a]ll accounts agree that figs are among the most important foods of specialized frugivores [birds] in Africa, southeast Asia, and Australia..." (Snow 1981: p. 9), but in the New World specialised fig-eating frugivores were less prominent. Leighton and Leighton (1983) suggested that the loss of figs in the Bornean forests that they studied would deplete frugivore populations in general, which would impact the dispersal of other plants. Similarly, Terborgh (1986: p. 339) observed, "[s]ubtract figs from the ecosystem and one could expect to see it collapse"; in Cocha Cashu, Peru, fig species were less than 1% of the total (12 out of ca 2,000), but they sustained nearly the entire frugivore community for three months of the year. There is no comparative study of the nutritional content of Old and New World figs (Snow 1981).

New World phyllostomid fruit bats have networks that are more highly nested and have higher connectance than fruit-bird networks, but were less robust to extinctions (see also Guimarães et al. 2011; Suweis et al. 2013 for this phenomenon); at the same time, there was low complementary specialization within a network (Mello et al. 2011a, b).

Fruit-eating birds: frugivory is common in the toucan-barbet clade, where the clade [Psilopogon + The Rest] had been dated to the late Oligocene (Prum et al. 2015). Just four species of toucans were found to eat fruits of 34 families of angiosperms - the largest genus had four species - in a study in the Brazilian Atlantic Forest (Galetti 2000).

Fruits in some families dispersed by specialist frugivores may have evolved before the particular plant:frugivore relationships that are evident today (Snow 1981). New World Piper, whose seeds are dispersed by Carollia, a phyllostomid bat, presents problems. Major diversification of fruit-eating phyllostomid bats began (27-)22(-18) Ma in the late Oligocene to mid-Miocene (Datzmann et al. 2010; see also Rojas et al. 2011, perhaps 3 Ma later), but this is (a very long time) after the beginning of diversification of New World Piper that has been variously estimated at 111-34 Ma (e.g. Martínez et al. 2014), so how this apparently close relationship (Fleming 2004) developed is unclear.

8B. Carbon Sequestration.


Bengtsson (1998), Augusto et al. (2014), Cernansky (2017) and many others have emphasized that species numbers are only one metric of evolutionary importance or success, however, numbers are quite easy to come by and they are usually the metric of choice. Estimates of biomass, productivity, even area occupied, whether for clades or for individual species, are much harder to obtain, but are other possible metrics, and from the point of view of global ecosystem functioning are of great importance. Here the focus is on particular clades of seed plants, sometimes quite small in size, that have a disproportionately large effect on communities, ecosystems, or even the global environment by fixing and/or sequestering large amounts of carbon; Avolio et al. (2019) discuss some of the different uses of "dominant", and although their prefered definition emphasizes similar aspects of the environment, their focus is more on individual species.

It has been noted in a rather general way that in nearly all communities there are common species; one can almost expect 25% of the species to contribute around 50% of the biomass or over 80% of the individuals, and the slope of biomass accumulation is less than that of species number accumulation (Gaston 2011; see also Gaston 2010; M. D. Smith & Knapp 2003). The monodominance discussed here is more extreme, a single species representing 60% or more of the canopy individuals and/or basal area (Connell & Lowman 1989; see also Peh et al. 2011b), although it is sometimes extended to include members of the one clade (quite restricted) that are growing together. Species that are dominant only in early successional phases (Type II transitional dominance, including Shorea albida!: Connell & Lowman 1989) are usually not considered further, although Newbery et al. (2013) described the long-lived Microberlinia bisulcata (see below) in the Cameroons as a "transient dominant". Note that even if they are not strictly successional, some of the vegetation types mentioned below may have been affected by human activities, as is discussed later.

Pitman et al. (2001, 2013) noted a surprisingly large number of common (>1 individual/ha) and widespread species of trees at least 10 cm across in the terra firme forests of western Amazonia. Some 83 of these 150 "oligarchic" species (from several families) occurred per hectare in Peru, and although they represented over 50% of the individuals, no species was close to being a monodominant, even by the somewhat relaxed definition sometimes adopted below. Building on this work, ter Steege et al. (2013) found that half the individuals with stems at least 10 cm across in Amazonian rain forests were accounted for by only 227 species in 41 families in 11 orders; they called these species "hyperdominants". In any one plot 32 of these species represented about 40.7% of the individuals (medians; ranges 0-78 species, 0-93% individuals: ter Steege et al. 2013). However, a few of these common species may be dominants as defined here (see Arecaceae, Fabaceae below). Fauset et al. (2015), also looking at Amazonian forests, found that about 1% of the trees (ca 160 species) were responsible for around 50% of the above-ground carbon storage/biomass production, but again, single species or even groups of species from the one family had relatively small effects. Thus the maximum percentage of biomass represented by a single species was a mere 1.93% (VAM/ECM Eperua falcata: Fabaceae) and the maximum total biomass of any one family in the top 20 biomass accumulators was 4.96% (Fabaceae, 1/4 the species). On the other hand, Bastin et al. (2015) found that only 18 hyperdominant species (1194 species in total recorded) produced 50% of the above-ground biomass in Central African rain forest, but 17 of these species produced 3.6% or less of this biomass production - interestingly, at 20% Gilbertiodendron deevrei (Fabaceae-Detarioideae-Amherstieae) was an exception (see also below). However, some of these Amazonian species may be locally abundant/hyperdominant as a result of encouragement/domestication by human activities in pre-Columbian times... (Levis et al. 2017; Junqueira et al. 2017; c.f. McMichael et al. 2017b; see also Watling et al. 2017a; Maezumi et al. 2018). In any event, most of the dominance mentioned in such literature is very different from that under discussion here (see also Arellano et al. 2016 and references).

The estimates of area, biomass, carbon accumulation, primary productivity, etc., given below should be taken with more than a grain of salt, and they can vary very widely from author to author (e.g. L. Brown & Lugo 1984; Botkin & Simpson 1990; Dixon et al. 1994; Chmura 2011). Different countries may have different definitions of "woodland", "grassland", "peat soils", and the like, and the distinctions between communities such as sea-grass, marine salt-marsh and mangrove are not always clear. It is also easy to forget that there are three different kinds of tons, and which is being used is not always mentioned. Ideally, biomass estimates for both above and below ground are needed, and the latter in particular should include dead plant biomass; residence times ("stored carbon") of dead biomass can be (much) longer than that of living biomass (for estimates of living woody biomass, see Galbraith et al. 2013). Frequently estimates of above-ground biomass only are given, and any estimation of the one from figures of the other should be undertaken with caution (e.g. Cairns et al. 1997; Pan et al. 2013); here the focus is mostly on below-ground carbon accumulation. Indeed, soils accumulate over three quarters of the carbon (as soil organic matter) in ecosystems, carbon above-grpound biomass being less than a quarter, and it has also been estimated that half the global carbon is in subsoil more than 30cm deep, not always sampled (Schmidt et al. 2011). Furthermore, mycorrhizal biomass, particularly that of ectomycorrhizae, may have to be incorporated into such figures, given that a significant proportion of tree productivity may be diverted to their upkeep (see below).

Major Players.

C4 photosynthesis, Grasses and Grasslands.

C4 photosynthesis, grass, and grasslands together play a major role in global ecology. Overall only a little over 2% of angiosperms - perhaps some 7,500 species - are C4 plants (R. Sage et al. 2012). They can be divided into three main groups: grasses, sedges, about which rather little is known ecologically, and core eudicots; in none is the origin of C4 photosynthesis monophyletic. The C4 photosynthetic syndrome has evolved 22-24 times in grasses, and 66-68 times in angiosperms as a whole (R. Sage et al. 1999, 2011, 2012; Ludwig 2011b). It is found in ca 1,500 species of Cyperaceae, 4,500 species of Poaceae, 500 species of chenopod-style Amaranthaceae as well as in other core Caryophyllales, ca 340 species of Euphorbiaceae (Euphorbia subg. Chamaesyce section Anisophyllum), and so on (Arakaki et al. 2011; R. Sage et al. 2012). Perhaps the immediate drivers of the evolution of this distinctive syndrome are the notable decreases of CO2 in the atmosphere at the beginning and again towards the end of the Oligocene along with briefly increasing temperatures, which together would lead to an increase in photorespiration, and temperature, high light levels, etc., are all important, with C4 photosynthesis favoured above ca 23oC (Still et al. 2003: fig. 1) In grasses, C4 photosynthesis is associated with arid climates, wheras in rosids it is more immediately associated with succulence (Lauterbach et al. 2016 and references). However, we still understand rather little about details of the origin and spread of C4 photosynthesis (e.g. Edwards et al. 2010; Kellogg 2013a; Cowling 2013; Heckmann et al. 2013; B. P. Williams et al. 2013; Christin & Osborne 2014; Y. Wang et al. 2014; etc.), although parallelisms at the morphological molecular level have been noted several times (e.g. N. J. Brown et al. 2011; Külahoglu et al. 2014; Aubry et al. 2014).

The global distribution of C4 vegetation is ca 18.8 x 106 km2 occupying somewhat over 15% of the total land area (Still et al. 2003: see map). All told C4 photosynthesis accounts for about 23-28% of terrestrial gross primary productivity (35.3 Pg C yr-1, vs 114.7 Pg C yr-1), although the biomass of C4 plants is less than 5% of the global total, 18.6 vs 407.9 Pg C yr-1 (figures from Still et al. 2003: see also Lloyd & Farquhar 1994; Ehleringer et al. 1997; Gillon & Yakir 2001; Retallack 2001; R. Sage et al. 2012). Most of the difference is in the woody biomass, that of C3 plants being 352.7 PgC and that of C4 plants zero (obviously Caryophyllales not factored in); in both cases root and leaf biomass was estimated to be about equal, that of C3 plants being about twice as much as C4 plants, 36.6 vs 18.6 Pg C (Still et al. 2003). Other estimates of total biomass are similar: 15.6 vs 488.5 Pg C (Ito & Oikawa 2004). Such figures are in line with recent estimates of the global biomass of plants, around 450 Gt C, of which <1 Gt was represented by marine plants, ca 315 Gt by stems and tree trunks, and ca 139 Gt by roots (Bar-On et al. 2018).


Productivity estimates for grasslands in particular, including both C3 and C4 species - are that they currently account for 11-19% of net primary productivity on land (slightly higher estimates in Lehmann et al. 2019) and 10-30% of soil C storage (Hall et al. 2000); soil C storage estimates in Averill et al. (2014) are (12.3-)14.5(-17.7) kg C m-2, NPP figures being (477-)576(-675) kg C m-2 yr-1. Gibson (2009) suggests that grasslands store 650-810 GtC, ca 33% of the global total, and 55-95% of that is stored underground, the higher values being in higher-latitude (probably = Arctic) areas. More general figures for tropical savannas and grasslands together - the grasses are likely to be C4 grasses - in Carvalhais et al. (2014: Tables S1 + S2) are ca 338 Pg total C, a carbon density of around 17.7 kgC m-2, and a mean turnover time of (12.2-)16(-22.1) years; the last set of figures is only slightly higher than those for l.t.r.f.. Comparable figures for temperate grasslands and shrublands are (145-)187(-249) Pg total C, a carbon density (13-)16.7(-22.2) kgC m-2, and a mean turnover time of (32.8-)41.3(-54.6) years; the residence time for carbon is much longer here, largely because of the cooler temperatures. Overall, grasslands are a long-term carbon sink and contribute to long-term global cooling (Volk 1989; Retallack 2001). As Retallack (2009: p. 100) noted, "grasslands did not merely adapt to climate change, but were a biological force for global change".

Grasslands often have distinctive soils. Root systems in mature grasslands are dense, and the soils are up to 1 m deep with good crumb structure and much organic matter (mollisols: Retallack 2001, 2009). The total carbon sequestration in grasslands is greater than that of the forests they replaced, and in particular the proportion of the biomass sequestered in the soil increases. Grassland soils are notably moister than corresponding woodland soils because woodlands have a lower albedo, so they reflect incoming radiation less, they are warmer and so transpire more, and so their soil is drier. Somewhat paradoxically grasslands support a cooler, drier climate, yet one that allows increased weathering, which consumes carbon; erosion from grasslands leads to a loss of organic carbon in sediment that is an order of magnitude larger than the corresponding loss from forests (Retallack 2009). Evidence from palaeosols suggests that grasses replaced woodland (Retallack 2001, 2013a). Savanna trees allocate relatively more carbon to below-ground biomass than do forest trees, while in both Cerrado and African savanna there has been the recent evolution of woody plants with massive stem and root systems underground, but with little permanent above-ground biomass (Scheiter et al. 2012; Pennington & Hughes 2014; Bond 2016a). Nutrients are also rapidly mobilized and when lost in run-off they ultimately support ocean productivity (Volk 1989). Nitrogen is volatilized in fires, but C4 grasses have low nitrogen requirements.

Associated with the spread of grasslands is a pronounced increase in fires ca 10 Ma (e.g. Bond & Midgley 2000; Keeley & Rundel 2005; Bond & Scott 2010; Lehmann et al. 2019). Only some 600 species of grass dominate ecologically worldwide, and most of these are C4 photosynthesizers (Edwards et al. 2010). Andropogoneae, with some 1,200 species and 90 genera, are notable in responding positively to annual burning (Forrestel et al. 2014; Visser et al. 2015), and may dominate in fire-prone grasslands. In African, Australian and North American grasslands and savannas in particular members of the C4 ASH clade (Andropogon, Schizachyrum, Hyparrhenia), with some 244 species, are prominent among the dominants (Estep et al. 2014). Lehmann et al. (2019) carefully summarized the whole issue of grass dominance.

Age: C4 photosynthesis may have originated in the Oligocene ca 33 Ma, but C4 grasses became diverse - and made a corresponding major contribution to overall vegetation biomass - only in the late Miocene 9-8 Ma, the process being complete as recently as the late Pliocene 3-2 Ma (e.g. Edwards et al. 2010; Strömberg & McInerney 2011; McInerney et al. 2011; Strömberg et al. 2011; Arakaki et al. 2011; R. Sage et al. 2012; Bouchenak-Khelladi et al. 2014; etc.); for further details, see Poaceae. C4 grass-rich and fire-prone savannas in Africa and the Cerrado in South America developed at about the same time (Simon et al. 2009; Simon & Pennington 2012; Maurin et al. 2014; Pennington & Hughes 2014).

Ectomycorrhizal Plants.

Introduction. In the following discussion, AM = arbuscular mycorrrhizae, ERM = ericoid mycorrhizae, and ECM = ectomycorrhizae, C = carbon, N = nitrogen, P = phosphorus.

The ECM habit predominates in only a few clades of seed plants, ca 8,500 species in 335 genera in 30 clades being a recent estimate, ca 2% of plant species (Brundrett & Tedersoo 2018). These numbers of ECM taxa do not include species with ericoid (ERM) or orchidaceous mycorrhizae; orchids and their mycorrhizae are not mentioned further here, but ERM are included in the discussion (see below). The major groups of ECM plants include most Fagales (some 1000 species), Pinaceae (210 spp.), Dipterocarpaceae (680 spp.) and relatives (together 915 spp), and Fabaceae-Detarioideae (250 spp. - Brundrett 2009; 450 spp. - B. Mackinder pers. comm. viii.2012), while some Salicaceae (Salix, Populus, 485 spp.) and a number of other small clades are also ECM (see also Tedersoo & Brundrett 2017; Tedersoo 2017; etc.).

Despite including only 2% of seed plant species, ECM plants occupy about half the forested area of the globe (L. L. Taylor et al. 2011) and make up about 60% of all tree stems - over 80% outside the tropics (Steidinger et al. 2019), especially in cooler conditions in the northern hemisphere, but there are also quite extensive areas of ECM plants in the tropics (see Corrales et al. 2018 for a critical review of tropical ECM). The map shows very approximately areas where Ericaceae (olive: ericoid mycorrhizae, ERM, see below), Pinaceae (red), and Fabaceae-Detarioidieae communities (blue), the last two ECM plants, predominate (from end-papers of Specht 1979a; White 1983; White et al. 2000; Matthews et al. 2000; Andersson 2005 - see also Read 1991). The area shaded red and olive green above 50o N in the map - tundra and much of the boreal forest - is mostly in the permafrost region. In the maps showing the distributions of different mycorrhizal types in Brundrett and Tedersoo (2018) Russia and Australia show particularly high percentages of ECM plants in their vegetation; the latter does not figure prominently in the narrative below because surprisingly little is known about the ecology of ECM in Myrtaceae, Eucalyptus in particular, and Acacia.

There may not be that many species of ECM seed plants, but there are 7,750 described species - but probably many more, perhaps up to 25,000 species - of ECM fungi (Blackwell 2011; esp. Rinaldi et al. 2008; F. Martin et al. 2016; Weiß et al. 2016; see also Kottke & Kovács 2013; Pickles & Pither 2013). Even single-site fungal diversity can be very impressive (Gardes & Dahlberg 1996; Horton & Bruns 2001: examples mostly Pinaceae-dominated forests; Kennedy et al. 2012; Walker et al. 2011; Timling & Taylor 2012). Indeed, Sánchez-García and Matheny (2017) suggest that in Tricholomatineae, at least, the ecological success of ECM plants may have positively affected fungal diversification as much as the ECM habit itself being an (ecological) key innovation. Overall, there are about as many (or considerably more) species of ECM fungi as ECM plants, although there may be relatively fewer species of fungi in Ericaceae (see below) and Orchidaceae (Rinaldi 2008, but c.f. van der Heijden et al. 2015a). (There are relatively far fewer species of AM fungi than their hosts, although the size of the imbalance depends on the number of species of glomalean fungi, currently unclear - e.g. Kottke & Kovács 2013.) ECM associates of taxa like Alnus show considerable host specificity (Bruns et al. 2002; Walker et al. 2013; Wicaksono et al. 2017), although on the other hand species in a small clade of boletes, [Strobliomyces + Afroboletus], ca 50 species, are associated with Fagales, Pinaceae, Pabaceae, Fabaceae-Detarioideae and Dipterocarpaceae (Sato et al. 2016: host shifts associated with changes in diversification rates?).

The almost 4,000 species of Ericaceae with ectendomycorrhizal ericoid mycorrhizae (ERM) are not included in these totals. Nevertheless, Vrålstad (2004), Villareal et al. (2004), Brundrett (2004), Imhoff (2009), Tedersoo et al. (2010b) and others (see also above) have suggested that ECM and ERM form a single ecological guild, and together they are found in ca 3% of flowering plants (Brundrett 2009). One of the characteristics of the guild is that the fungi are intermediaries in the uptake of N from soil organic matter by the plant (e.g. Read 1991, 1996; Peay 2016; c.f. Persson & Näsholm 2001), although ERM fungi are more or less saprotrophic, unlike ECM fungi (Kohler et al. 2015; Peay 2016; Selosse et al. 2017c). Another similarity is the architecture of the plant-fungal networks (Toju et al. 2016 and references). ERM fungi include ascomycetes, the basidiomycete Sebacinales-Serendipitaceae, etc., but the number of species of fungi involved is unclear (Weiß et al. 2016). ERM are included in the discussion below, but orchid mycorrhizae (see Orchidaceae), also often considered to be modified ECM and with similar plant-fungal networks, are not discussed further (see also Imhof 2009). It should also be remembered that some dark septate endophytes can also form ERM and ECM associations, indeed, sometimes a single species of fungus forms both kinds of associations (Lukesová et al. 2015; see also below).

ECM/ERM plants are especially common in subarctic to (cool) temperate habitats - all boreal tree species are ECM plants (S. E. Smith & Read 2008) and shrubs are ECM/ERM plants - and in tropical montane vegetation, basically, in areas with predominantly organic-bound nutrients (Comas et al. 2012), the fungi being able to access soil organic matter, with N ultimately moving to the plant (Shah et al. 2016; Peay 2016: see below). However, ECM plants are also common in lower altitude tropical West Malesian dipterocarp forests, African Miombo (ca 2.6 x 106 km2 alone - see Iimberlake et al. 2010) and Sudanian woodlands, and also considerable areas of the Guineo-Congolian coastal and Ituri rain forests (e.g. Malloch et al. 1980; White 1983; Safer 1987; Connell & Lowman 1989; Hart et al. 1989; Read 1991; Sanford & Cuevas 1996; Torti et al. 2001; Peh et al. 2011b); Sudanian and Miombo woodlands are biogeographically close (Linder et al. 2012). Australian Eucalyptus woodlands are also ectomycorrhizal. ECM plants are not often common in Amazonian forest, although they are known from acidic white sand vegetation and other less fertile soils (Roy et al. 2016; Peay 2016). Dark septate endophytes also occur far N (82o) and S (77o) (Newsham et al. 2009).

ECM/ERM plants are generally woody, although there are some herbaceous ECM taxa, perhaps particularly in Arctic-Alpine environments (e.g. Newsham et al. 2009). ECM/ERM plants tend to dominate the communities in which they grow, although these communities are often not notably diverse, in part perhaps because of the dearth of readily available nutrients (L. L. Taylor et al. 2009; Reich 2014 for literature; see below). Many ECM plants are large individuals, each representing a substantial amount of standing biomass, and, very importantly, the soil in which they grow is often acid, litter commonly accumulates, and peat formation is common (e.g. Yu et al. 2010; Page et al. 2011). A very approximate estimate of the number of ecologically dominant plant ECM species is 1,000. This includes ca 160/388 species of Dipterocarpaceae, 11/13 Nothofagaceae, and 50/165 Fagaceae from Malesia alone (data from Ashton 1981; Soepadmo 1972), however, P. S. Ashton (pers. comm. vii.2012) noted that only ca 13 species of dipterocarps were major dominants. Ca 20 species of Ericaceae are widespread in boreal forest and tundra habitats, a few species of Vaccinium are very abundant locally in montane forests in Malesia, and species of Rhododendron similarly dominate locally in the Himalaya-Yunnan region.

Vegetation dominated by particular ECM species often also includes other ECM/ERM plants. Thus in the Mediterranean Maquis vegetation, Cistaceae (close to Dipterocarpaceae), Fagaceae and Pinaceae, all ECM plants, are important components of different aspects of the vegetation. There are extensive oak-pine ECM forests in the eastern United States and Mexico, in western North America the oak-pine forests include a substantial element of Arbutus menziesii (Waddell & Barrett 2005), an ericaceous tree with arbutoid mycorrhizae, and in boreal forests ECM Salicaceae and Pinaceae grow together (Hewitt et al. 2017 and references). Forests with ECM Fabaceae, Dipterocarpaceae and Phyllanthaceae are common in tropical Africa. Boreal forests are dominated by ECM Pinaceae with some Betulaceae and Salicaceae, also ECM, and the understory often includes ERM Ericaceae (Read 1993).

Evidence suggests that the establishment of conspecific seedlings in ECM forests is often less affected than might be expected given the density of the parental trees, i.e. ECM species are less subject to the Janzen-Connell effect (Terborgh 2012). Thus seedlings of ECM trees can germinate close to their parents (e.g. Horton & van der Heijden 2007 and literature; Simard 2009; McGuire 2007b; Michaëlla Ebenye et al. 2017; Gerz et al. 2017) while seedlings of AM trees do so less readily (van der Heijen & Horton 2009; Laliberté et al. 2015; Bennett et al. 2017: Fig 2a). In part, at least, this may be because the ECM investing the root helps protects the roots against pathogen attack (e.g. Koele et al. 2012; Laliberté et al. 2015), and this would facilitate the development of the dominance shown by many ECM trees. L. Chen et al. (2019) noted that in the subtropical forest they examined taxa of all mycorrhizal types, seedlings and adults, that accumulated pathogenic fungi more quickly suffered more from conspecific negative density dependence (CNDD) effects, species with lower CNDD accumulating ECM fungi more quickly (ECM species lest affected?!). Dicymbe seedlings quickly tap in to the ECM network in Dicymbe forests in Guyana (McGuire 2007a, b). In the Arctic seedlings germinating after fire acquire ECM like those in the resprouting shrubs immediately surrounding them (Hewitt et al. 2017). Interestingly, Corrales et al. (2016) suggested that despite strong negative plant-soil feedback to seedlings, monodominant stands of Oreomunnea mexicana formed because the ECM fungi associated with the plant could obtain N directly from organic matter. In temperate forests the regeneration of the more abundant species (not necessarily ECM) may show weaker CNDD that that of the less common species (D. J. Johnson et al. 2012a, b; c.f. Dickie et al. 2012; Liu et al. 2015: Fusarium oxysporum a specialist on Ormosia glaberrima), perhaps enhancing their dominance, although this is not a general latitudinal effect (Comita et al. 2014). In more species-rich areas of South American forests there are stronger CNDD effects, fungal pathogens and insect herbivores driving up diversity (Mangan et al. 2010: Panama; Terborgh 2012: review, esp. Peru; Johnson et al. 2012a, b: Bagchi et al. 2014; Comita et al. 2014; see also Kulmatiski et al. 2008 and Schnitzer et al. 2011: grasslands).

Many ECM plants have a distinctive ecological syndrome: They are trees, often locally dominant and casting deep shade; they are mast fruiters, i.e., they fruit more or less simultaneously yet intermittently over large areas and they have large seeds; and the soils they grow on are poor in inorganic nutrients, often with deep leaf litter and accumulating humus, yet the organic N is indirectly accessible to them (e.g. Connell & Dawson 1989; Alexander 1989b; Richards 1996; Torti et al. 2001; Newbery 2005; Peh et al. 2011b; Terrer et al. 2016: see also Koenig & Knops 2000, 2005; Norden et al. 2007; Veller et al. 2015; Pearse et al. 2016 for mast fruiting). However, although Fabaceae-Detarioideae have most of these features, not all are mast fruiters, and some other Fabaceae that also have most of them are AM plants (see also Torti et al. 1997, 2001). Similarly, Nascimento and Proctor (1997) noted that there was little difference between the soils on which the ECM Peltogyne dominated and those in adjacent more diverse forests. Furthermore, ECM/ERM plants like Pinaceae, Salicaceae, and some Betulaceae, and perhaps ERM Ericaceae, from cooler areas have a somewhat different ecological syndrome. Although they are trees or shrubs, they have much smaller seeds and often do not show mast fruiting, but like other ECM plants, they can dominate the communities in which they are found, and they grow on acidic and peaty soils - and they can tolerate soils with toxic metals (e.g. Read & Perez-Moreno 2003; Nara et al. 2003; Cairney & Meharg 2003). Indeed, Koele et al. (2012) found no particular correlations between foliar traits and ECM status. Note that mast fruiting has been described from a number of other plants, e.g. trees and lianas from tropical French Guiana in South America, but these are not known to be ECM plants nor to be particularly dominant in their communities - see Norden et al. 2007.)

Before going further, a digression on lignin decomposition is in order. White-rot basidiomycete fungi are the most important decomposers of lignin. The capability to degrade lignin may have evolved in the ancestor of the Agaricomycetes clade (372-)290(-222) Ma around the beginning of the Permian, in particular, in the [Auriculariales + The Rest] clade which have fungal class II peroxidases and other components of the lignin decay system, and this is consistent with the fossil record, the earliest fossils of white-rot (lignin-decomposing) fungi being around 260 Ma (Floudas et al. 2012: also other similar dates; Kohler et al. 2015: ca 294 My; Nagy et al. 2015). The white-rot fungi show extensive expansion of the gene families involved in the decay process - they can break down both lignin and cellulose - and they also have many gene families of as yet unknown function (Nagy et al. 2015, 2016). Eastwood et al. (2011) suggested an age of around 219.6 Ma for the [Russulales + Agaricales] clade, while a larger clade in which the basal pectinations are white rot fungi can be dated to 250-234 Ma (Kohler et al. 2015). Dacrymycetes are sister to Agaricomycetes and are brown rot fungi that lack lignin-decomposing enzymes, in particular, the dacrymycete Calcera cornea is a brown-rot fungus with a soft-rot, not white-rot, ancestor (Kohler et al. 2015; Nagy et al. 2015, 2016). The basal agaricomycete Sebacinales and Cantharellales also have other life styles.

Peroxidases of white rot fungi play a central role in lignin degradation by mineralizing the lignin, at the same time producing CO2, etc., and white rot fungi also have several enzymes that can degrade crystalline cellulose. Brown rot fungi can access much of the crystalline cellulose in the cell wall quite readily, but they do not destroy the lignin; they may have evolved from white-rot fungi (e.g. Floudas et al. 2012; Lundell et al. 2014; Kohler et al. 2015; Nagy et al. 2016), although Dacrymcetes are brown rot fungi without any obvious white rot ancestry (Nagy et al. 2015). Indeed, the distinction between the two groups of fungi is somewhat artificial (Riley et al. 2014; Floudas et al. 2015), and they are better thought of representing two ends of a spectrum of fungi that have various combinations of lignin- and cellulose-degrading enzymes, furthermore, agaricomycetes and also ascomycetes that cannot break down lignin may have enzymes that are involved in the breakdown of cellulose (see also Soudzilovskaia et al. 2015; Nagy et al. 2015). Some ascomycetes, soft rot fungi, can also break down lignin (Shary et al. 2007 and references), and many xylariaceous fungi can break down lignin and cellulose, "the most efficent of them rival[ling] basidiomycetes in substrate degradation" (Rogers 2000: p. 1414). The ECM ascomycete Oidiodendron maius is saprotrophic, breaking down Sphagnum peat, and it has both cellulose and some lignin-decomposing enzymes (Kohler et al. 2015), while UV light also decomposes lignin (Austin & Ballaré 2010), although I do not know how globally important these processes are/were ecologically.

ECM fungi are not saprotrophic, i.e. they have few enzymes involved in cellulose and lignin breakdown, indeed, these are genes that may get lost as ECM symbioses develop (see Hess et al. 2018) and they do not obtain their metabolic C from dead organic matter, rather, this comes from the plant. However, ERM fungi may be rather different, retaining a number of saprotrophic genes from their ancestors and being able to break down lignin, to a cerain extent switching between saprotrophic and biotrophic life styles (e.g. Michelsen et al. 1996; Jonasson & Michelsen 1996; Hashimoto et al. 2012; Vohník et al. 2012; Habib et al. 2013; Lindahl & Tunlid 2014; Kohler et al. 2015; Martino et al. 2018). Given the extreme polyphyly of the ECM habit, different clades may vary in details of the nature of their associations with plants, but in at least some cases physiological similarities between different clades of ECM fungi are quite extensive (Peter et al. 2016), and if the genetic tool kit for ECM symbioses is to be found in free-living fungi (Hess et al. 2018) the extensive polyphyly of the ECM habit may have a partial explanation. The hyphae of ECM fungi form a complete sheath investing the fine roots, the Hartig net, and hyphae penetrate between the exodermal/cortical cells; this sheath, not the root/root hair system, forms the interface between the plant and the soil (e.g. L. L. Taylor et al. 2009), and they also form rhizomorphs of various morphologies that extend sometimes many metres from the plant (this would constitue the hyphosphere of the plant), the nature of the rhizomorph depending on the particular association which is in turn linked to overall N availability (Agerer 2001; Hobbie & Agerer 2010; Águeda et al. 2014 and references) and which do a lot of the foraging for nutrients (see below). Usually the Hartig net involves only the outermost layer of root cells, but sometimes, as in Pinaceae, the hyphae penetrate more deeply (Brundrett 2004). The fungal hyphae are septate and are not intracellular except in ERM, which usually lack a Hartig net (Ericaceae have quite a variety of associations with fungi: Weiß et al. 2016 and references). The ecological effect of these plant-fungus associations hinges on the ability of the fungus to utilise the humic substances that make up soil organic matter (e.g. Schmidt et al. 2011; Rosling et al. 2015; Shah et al. 2016). Although ECM fungi have only a limited ability to break down lignocelluloses, some have coopted oxidizing enzymes that, with sugar from the plant, do break down lignocellulosic material, liberating N and P which is taken up by the plant, or they use peroxidases for this purpose (Shah et al. 2016). These enzymes are normally characteristic of brown and white rot fungi respectively, and have evolved functions that differ from those of the same enzyme in non-ECM associations, e.g. rather than obtaining C they scavenge nutrients (Doré et al. 2015; Shah et al. 2016).

ECM plants differ from AM plants in their N metabolism, the former probably obtaining N from the persistent litter, while soils with AM plants accumulate less litter, N produced by the activities of soil organisms and inorganic N being the source of N for the plant (e.g. Read 1991; D. L. Jones et al. 2004; J. E. Hobbie & Hobbie 2006; Philips et al. 2013; Bardgett et al. 2014; Lin et al. 2016). ECM hyphae are most abundant in the litter layer and in mineral soil layers immediately below the litter (e.g. Lehto & Zwiazek 2011). Even when ECM and AM forests have similar litter inputs, there is more litter accumulation in the former (Lin et al. 2016: largely temperate plantation comparisons; see also Baskaran et al. 2016: modelling). Trees from higher-latitude habitats in particular often have both long-lived leaves and fine roots, the latter being well-defended (as by ECM: Laliberté et al. 2015 and references). Furthermore, ECM taxa like oaks and pines are efficient at removing N and P from their own leaves before they die, and the result is nutrient-poor humus, unsuitable for AM plants which are often faster-growing and need N (Phillips et al. 2013). Indeed, the result of the activity of ECM growing with Pinus is a low pH soil, and the litter still around after N and P has been removed by the fungi is of low quality and slow to break down (Read 1998). ECM and ERM plants can utilize N as complex organic compounds (see above), and this is affected by pH and litter polyphenols, which also vary in tandem at the infraspecific level (Northup et al. 1995). Indeed, Averill et al. (2014; also Soudzilovskaia et al. 2015: top/organic soil layer) noted that there was about 70% more C per unit N in the soils of ecosystems dominated by ECM plants; the N remained accessible to the fungi, but not to their microbial competitors. However, Averill et al. (2014) thought that litter with a higher C:N ratio would lead to less overall C storage in the soil because microbes would respire relatively more C to obtain N. Low rates of litter decay (often accompanied by high MA - leaf mass per area - values) are features of both gymnosperms and angiosperms that are able to grow in stressful, nutrient-poor environments (Berendse & Scheffer 2009).

The litter of some ECM trees tends to decompose more slowly than that of other vascular plants, whether deciduous or evergreen (Pérez-Harguindeguy et al. 2000; Cornelissen et al. 2001; Alexander & Lee 2005; Cornwell et al. 2008; Lang et al. 2011; Phillips et al. 2103) and conifer litter tends to decompose more slowly than that of core eudicot angiosperms (e.g. Wardle et al. 2008; Cornwell et al. 2008b; Weedon et al. 2009; LeRoy et al. 2019), and within Pinales litter of the ECM Pinus edulis decomposes slower than that of AM Juniperus having a higher C:N ratio (Gehring et al. 2017b and references). Litter accumulation is also notable in tropical ECM communities (Torti et al. 2001). For example, the litter of 4 ERM species (Fabaceae-Detarioideae) in Cameroonian forests decomposed faster that that of three AM taxa, although the former were richer in P and N (Chuyong et al. 2002). However, the generality of correlations between ECM plants and their litter composition is perhaps questionable (esp. Koele et al. 2012; see also Liu et al. 2014: Fagaceae; Corrales et al. 2016: Oreomunnea). Litter from ERM plants may be the most recalcitrant (Read 1991; Cornelissen et al. 2001; Alexander & Lee 2005). Interestingly, although litter from AM plants decomposes faster than that of ECM plants, the mineral-associated soil organic matter produced is stable and the organic N is inaccessible to AM fungi (Terrer et al. 2016 and references).

In Rhododendron, at least, N in stable protein-tannin complexes formed by the plant is more easily accessed by its own ERM associates than by ECM or AM roots (Wurzburger & Hendrick 2009). N mobilization is also affected by the polyphenols in the litter (e.g. Northup et al. 1995). Enzymes, etc., produced by the ERM plant/fungus association contribute to the formation of acidic mor humus that ERM plants like and AM plants do not (e.g. Read 1991). Brown rot fungi are associated with conifer forests in particular, but they are unable to destroy lignin, which accumulates in large amounts and binds N and cations (Eastwood 2011 and references). ECM forests may have few bacteria and litter-decomposing fungi, and, as noted below, ECM fungi themselves also affect litter decomposition (Ekblad et al. 2013). Although Näsholm et al. (2013) interpret ECM fungal activities in conifer forests somewhat differently, with N in some circumstances being retained in fungal mycelium, the consequences are similar; non-ECM plants will be at a disadvantage in the N-poor conditions that result. Strullu-Derrien et al. (2018 and references) note that ECM fungi may have lost PCWDEs, but they do have enzymes that allow them to decompose organic material and scavenge organic N and P. imilarly, Newbery et al. (1997) found that in some forests on poor soil in Cameroon the phosphorus in soil and litter was preferentially accessed by the dominants, ECM Dialioideae (Fabaceae), again, ECM plants can exploit the rather extreme soil conditions they themselves make. At the same time, it should be noted that some ECM communities grow in quite N-rich habitats, as in the Pacific Northwest, but the N is taken up as NH4+, not NO3-, ions (Kranabetter et al. 2015) as is a feature of the organic nutrient ecomonomy of ECM communities (Phillips et al. 2013). ECM, ERM, and their seed-plant associates together break down soil organic matter to get at the nutrients it contains, N in particular, the fungi, supported as they are by host C, competing successfully with microbial decomposers for N (e.g. E. A. Hobbie & Hobbie 2008). ECM, ERM, and their seed-plants together form the soil/humus conditions that they all prefer (see also Read 1993; Phillips et al. 2013; Shah et al. 2016). Indeed, interactions between the N metabolism of the plant and their ECM/ERM associates is a recurring theme in this whole literature (e.g. Koele et al. 2012; Hawkins & Kranabetter 2017). However, Pellitier and Zak (2017) suggested that despite the interest and importance of these interactions between humus, fungus and the plant, transfer of N from soil organic matter to the plant via ECM fungus had yet to be conclusively demonstrated...

Read (1998: p. 328) noted of the Pinus/ECM association, "pine roots are simply food bases which nourish an extremely dense mycelial system". Substantial amounts of mycelium may be produced every year, whether as individual hyphae, as in ascomycete ECM, or in more massive rhizomorphs, as in many basidiomycete ECM (Visser 1999; Agerer 2001; Hobbie & Agerer 2010; Koide et al. 2013; Águeda et al. 2014), although the actual contribution of ECM fungi to soil organic matter is unclear (Ekblad et al. 2013: few systems studied; Soudzilovskaia et al. 2015). Hyphal biomass ultimately comes largely from the host plant, the ECM fungi utilizing (1-)7-34(-ca 50%) of the host's photosynthetic NPP (J. E. Hobbie & Hobbie 2006, E. A. Hobbie & Hobbie 2008; S. E. Smith & Read 2008; Högberg et al. 2010; Koide et al. 2013; Phillips et al. 2013; Ekblad et al. 2013; Allen & Kitajima 2014; Fernandez et al. 2015 and literature). Other estimates are higher, Litton et al. (2007) noting that the total below-ground C flux in forests was 25-63% of their gross primary productivity and Nehls et al. (2010) estimating that up to half the C produced by the plant might be committed to the fungus. C flux in ERM appears to be similar, although figures specifically for them are hard to come by. These figures are broadly similar to the amount of productivity devoted to fine root production and turnover, and so the total figure of plant-soil biomass changes considerably if both are incorporated into the community C budget, as needs to be done (E. A. Hobbie & Hobbie 2008; D. D. Cameron et al. 2008a; Philips et al. 2013; Bargett et al. 2014; McCormack et al. 2015; Laliberté 2016; Field & Pressel 2018), indeed it has been estimated that the mycelium:root ratio is something like 200,000:1 on a soil volume basis (Read 1998), so the high C requirements of the fungus are not surprising. As Schmidt et al. (2011) noted in a review that did not focus on ECM, soils contain at least three times as much C in organic matter as there is in the above-ground biomass. Fungi with long distance rhizomorphs may contribute around 15 times more biomass to the soil than do those with short-distance rhizomorphs (Koide et al. 2013), indeed, the rhizomorphs may live longer than the roots with which they were initially associated (references in Ekblad et al. 2013).

Contributing to the persistance of fungal remains in the soil is the fact that ECM and ERM fungal hyphae in particular, but also hyphae of dark septate endophytes (of course, one species of fungus can be placed in two or all three of these categories depending on the plant with which it is associated), commonly have melanin, a substance particularly resistant to degradation, in their hyphae (e.g. Butler & Day 1998: white rot fungi may be able to decompose it; Read et al. 2004; Bardgett et al. 2014; Clemmensen et al. 2014; Peter et al. 2016; Lindahl & Clemmensen 2017; Martino et al. 2018). The result is that C in these ECM hyphae turns over more slowly than that in AM hyphae (Phillips et al. 2013; Fernandez et al. 2013, 2015; Fernandez & Kennedy 2018). Fernandez and Koide (2014) found that amounts of melanin, along with those of N, determined the rate of hyphal breakdown, suggesting that from this point of view melanin was an analogue of lignin, being notably decay-resistant (see also Fernandez et al. 2015); fungal rhizomorphs were likely to decay more slowly than individual hyphae (Koide et al. 2013 and references). Bjorbækmo et al. (2010) and Timling & Taylor (2012) noted the high frequency of melanized fungi and dark septate endophytes in high northern latitudes. Indeed, it was found that hyphae of the common and widespread ECM (it can also be a saprophyte - Meyer 1964) ascomycete Cenococcum humile, which have much melanin, persisted up to ten times longer than the hyphae and rhizomorphs of other ECM fungi (Fernandez et al. 2013; see also references in Allen & Kitajima 2014: rhizomorphs may persist for months to years). (Amusingly, somewhat carbonized sclerotia of Cenococcum or similar fungi were mistaken for carbonaceous spherules produced by intense fires following a supposed bolide impact that triggered the Younger Dryas period - Scott et al. 2010.) The hyphae of Cenococcum are notably common in dry and nutrient-poor mor humus in boreal and temperate forests, thus mor in Fagus sylvatica forests has both a very high density of beech rootlets and substantial amounts of mycorrhizae, about half of which is Cenococcum (Meyer 1964; see also Pellitier & Zak 2017). In Swedish boreal forests, much root biomass was in the soil to about 20 cm down, the depth of soil in younger forests, but older soils were deeper and in the deeper portions ECM and ERM were major contributors to the biomass. In younger soils, although the fungal biomass might be considerable, it decayed faster than that in the older, deeper soils (Clemmensen et al. 2013). Dark septate endophytes also have melanin, and these may on occasion form ECM or ERM associations (Lukesová et al. 2015). All in all, dead and rather decay-resistant fungal mycelium makes up a substantial component of the total soil C in ECM forests, and the source of this mycelial C is largely the ECM trees (Nehls et al. 2010; Clemmensen et al. 2013; Koide et al. 2013). Note, however, that the rhizomorphs of the large genus Cortinarius, an agaric, decay quickly (Zak et al. 2019).

The combined result of plant decay and fungal activity in ECM plants is often soil with acid conditions and persistent litter with a high C:N ratio; this will tend to constrain mineralization by other than ECM plants, as will the seasonal/cold and sometimes dry climates that many ECM/ERM plants favour (Read 1991; Augusto et al. 2014). Indeed, ECM/ERM communities are often found on rather extreme soils, including serpentines (Branco & Ree 2010), that are either poor in nutrients (e.g. Michelsen et al. 1998), and/or rich in organic materials and/or without much other vegetation (e.g. Read 1993). In such communities soil pH is, or becomes, low, and sometimes massive amounts of mor humus commonly accumulates, especially in cooler climates; podzolization may also occur (van Schöll et al. 2008). In general, the acid, nutrient-poor conditions and high water tables that develop are not conducive to the activity of many potential decomposers of humus - a self-reinforcing cycle - and there is CO2 sequestration.

There is one more aspect of ECM activities which affects global C balances. ECM fungi in particular - AM fungi less so, although they, too, are active - facilitate subsurface weathering of rocks, especially when basaltic, so sequestering additional CO2 in the process (e.g. Landeweert et al. 2001; van Schöll et al. 2008: Al moves from the rock to the humus layer?; Taylor et al. 2009, 2011, 2012; Comas et al. 2012; Quirk et al. 2012, 2014, c.f. Porada et al. 2016 for remarkably high estimates of Ordovician bryophyte and lichen rock weathering; Strullu-Derrien et al. 2018). Siderophores and low molecular weight organic chelators like oxalic acid produced both by ECM fungi and their bacterial associates all increase the breakdown of silicate minerals and weathering of rocks, both basaltic and granitic (e.g. Knoll & James 1987; Frey-Klett et al. 2007; L. L. Taylor et al. 2009, 2011, 2012; Comas et al. 2012). These low molecular weight organic acids can mobilize cations such as Ca++ and Mg++, increase phosphorus availability, etc. (Taylor et al. 2012); siderophores chelate iron and oxalate forms complexes with aluminium ions, detoxifying the aluminium but also increasing the weathering of aluminium-containing minerals in rocks (Landeweert et al. 2001; Hoffland et al. 2001; van Schöll et al. 2008). CO2 produced by the respiration of the fungus-plant association is used up in this weathering as it reacts with water, CaCO3 and silicate minerals, and C is ultimately carried out to sea (e.g. Berner 1997; Beerling 2005a; L. L. Taylor et al. 2009). Quirk et al. (2014) showed that plants allocated more C to ECM than to VAM, and that ECM caused correspondingly higher rates of calcium silicate dissolution from basalt, a rate proportionally reduced when atmospheric CO2 concentrations were low. It has been suggested that the evolution of the ECM habit "represents the most profound alteration in root functioning to occur in plant history..." (Taylor et al. 2011: p. 369; c.f. in part Boyce & Lee 2011), and it is not for nothing that ECM fungi in particular have been dubbed "rock-eating fungi" (Jongmans et al. 1997: p. 682; Lehto & Zwiazek 2011). Indeed, N derived directly from rock can greatly increase ecosystem C storage in coniferous forests in particular (Morford et al 2011; Houlton et al. 2018).

Thinking about ERM and ECM emphasizes the below-ground dimension of plants, i.e. the large proportion of the plant's metabolic capabilities that go to support both its roots and their microbial associates and the important part that roots and mycorrhizae together play in C sequestration and the like. It also emphasizes how difficult it is to understand what is going on below ground, and how little is really known about biomass accumulation there and its longevity (e.g. D. L. Jones et al. 2004; Bardgett et al. 2014; Laliberté 2016). Indeed, biomass itself is a rather vague term - the mass of living organisms in a particular area, possibly including microroganisms associated with the organisms of interest, measured in various ways, as living tissues, dried or not, as net primary productivity, dead tissues (strictly speaking necromass: see Fernandez et al. 2015), particularly important in the discussion here, and so on.

Age: So what are the major fungal clades involved and when and where did ECM-plant interactions develop?

ECM associations may have initially appeared in the tropics (e.g. Matheny et al. 2009; Strullu-Derrien et al. 2018 and literature), even if ECM/ERM plants today are especially common in some tempereate to Arctic communities. The ability to form ECM associations has evolved perhaps 82-86 times in fungi, especially in ascomycetes and basiodiomycetes, but also in Zygomycota, and more origins are likely to be discovered especially in tropical and south temperate areas (F. Martin et al. 2010; Tedersoo & Smith 2013, 2017; see Wurzburger et al. 2016 for clades). Ages of the origins of these fungal ECM clades, including those associated with Pinaceae, are split about equally between Late Cretaceous (e.g. Amanita) and Cenozoic (e.g. Hebelomateae, Tricholomatineae) (Ryberg & Matheny 2012; Tedersoo et al. 2014a and references; Sánchez-García & Matheny 2017: movement on to Pinus). Bonito et al. (2013) looked at the evolution of truffles (ascomycetes), and they suggested that the age of the clade that included Helvellaceae and Tuberaceae, all ECM fungi, was (184.7-)160.8(-137.4) Ma. In another estimate. the stem age of Cantharelles and Sebacinales, a proxy for the evolution of the ECM habit, was given as ca 229 m.y. and somewhat after the early diversification of gymnosperms if well before the diversification of Pinaceae at (233.4-)183.3(-150) Ma (Lutzoni et al. 2018). The m.r.c.a. of Tuberaceae was dated to some (179.1-)156.9(-134.5) Ma, and Bonito et al. (2013) thought that its host was likely to have been an angiosperm. Kohler et al. (2015) suggest that ECM associations with plants developed in the last ca 175 Ma, and a clade of ECM, brown rot, and one white rot fungus is dated to ca 115 Ma. Augusto et al. (2014) dated confirmed ECM symbioses in both angiosperms and gymnosperms as being mid-Cretaceous, some 115 Ma, probable ECM symbioses in gymnosperms might be over 200 Ma in the Late Triassic and possible ECM symbioses over 250 Ma, as early as the Permian, although the authors warn about extrapolating from the ecophysiological proclivities of modern gymnosperms to those of early gymnosperms. Clade size varies greatly. Thus the ascomycete Cenococcum geophilum, perhaps the commonest ECM fungus, is the only dothideomycetan fungus with such a life style (Peter et al. 2016), although this fungus in particular is unlikely to be able to do much to soil organic matter (Pellitier & Zak 2017).

Additional dates: Sclerodermatinae (Boletales) diversified (115.5-)82.5, 80.5(-54.5) Ma, core Sclerodermatinae (90.5-)66, 58(-34.5) Ma ancestral hosts perhaps being Pinaceae, Fabaceae, Fagaceae or Betulaceae (A. W. Wilson et al. 2012). The large, cosmopolitan and plesiomorphically ECM Inocybaceae are estimated to have begun diverging at the very beginning of the Cretaceous (191-)143(-99) Ma, early hosts being Fagales, Myrtaceae and Phyllanthaceae, and a number of major clades separated in the Cretaceous, however, movement on to Pinaceae was later, no earlier than 65 Ma, both relatively recent and derived as in other groups (Matheny et al. 2009: ages of preferred calibration more compatible with the evolution of angiosperms). Still younger are fungal associations with Dipterocarpaceae (49 Ma or less) and Nothofagaceae (33-13 Ma) (Matheny et al. 2009). Strobilomyces and its sister Afroboletus diverged (70-)57(-44) Ma, diversification within the latter genus beginning (65-)50(-35) Ma, probably in Africa, and with hosts of the early diverging clades (S. echinatus sister to the rest) including Fabaceae-Detarioideae, Dipterocarpaceae-Monotoideae and Phyllanthaceae (Han et al. 2018). Basal clades of the ECM false truffle group Hysterangiales are found predominantly on Australian Myrtaceae (Mesopelliaceae grow on Eucalyptus), other clades being more widely distributed and with a diversity of hosts, and Hosaka et al. (2008) think that the group is post continental drift in age. On the other hand F. Martin et al. (2002) thought that the ECM Pisolithus was originally Australasian - a generalist as regards its hosts, now found on Acacia and eucalypts - and radiated as Pangaea broke up in the Triassic.

The Plants. ECM plants are clumped phylogenetically (e.g. Alexander & Lee 2005; L. L. Taylor et al. 2009, 2011; Maherali et al. 2016), and the ecologically most important ECM species are found in four main clades, Pinaceae: 1 origin, Fagales: 1 origin, Dipterocarpaceae/Sarcolaenaceae/Cistaceae: ?1 origin, and Fabaceae-Detarioideae: ?origins; note that the last three clades are in the fabid or N-fixing clade. All told, members of 30-40 families of seed plants are involved (Hibbett et al. 2000; Hibbett & Matheny 2009; Bruns & Schefferson 2004, B. Wang & Qiu 2006; S. E. Smith & Read 2008; Tedersoo et al. 2010b, 2014a; Koele et al. 2012 in part), with perhaps 6,000-7,000 species in 250-300 genera (Tedersoo & Brundrett 2017) or as many as around 29,300 species (Maherali et al. 2016: inc. gymnosperms; angiosperm total numbers from Paton et al. 2008) or 16,100 species (starting off with figures from Brundrett 2002). In a number of taxa there are records of both ECM and AM associations (Wurzburger et al. 2016; Brundrett 2017a; Tedersoo & Brundrett 2017; Tedersoo 2017b). For the evolution of ECM associations from vesicular-arbuscular mycorrhizae, see Maherali et al. (2016); reversals from the ECM condition are uncommon and are unknown in Pinaceae, mycorrhizal associations in general being rather stable (see also Kiers & van der Heijden 2006). ERM are restricted to Ericaceae. See below under individual groups for suggestions of dates.

some ECM/ERM forestsPinaceae (red in map) are the major component of the vast boreal/subarctic/taiga forests; there are also Salicaceae and Betulaceae, also ECM (see below), while ERM Ericaceae may be common in the understory (e.g. Read 1991; Villareal et al. 2004; Vrålstad et al. 2002; Vrålstad 2004; Kranabetter & MacKenzie 2010; Gauthier et al. 2015). Estimates of the area occupied by these forests range from 12 x 106 km2, or ca 17% of the land surface of the earth (Moore 1996; Lindahl et al. 2002), to 9.2 x 106 km2, 73% of the conifer forests of the world (Kuusela 1992), or 30% or more of the world's forest (Taggart & Cross 2009; Gauthier et al. 2015). However, () suggested that the core area of boreal forests in Eurasia alone was 12 x 106 km2, so the global total may be as much as 17.1 x 106 km2 (see also Melillo et al. 1993: estimates for boreal woodland + forest are 18.5 x 106 km2). In addition to these boreal forests being the largest single terrestrial biome, they are also the youngest woody biome, having formed within the last 12 Ma at the expense of evergreen broadleaf and mixed forests as climates became drier and cooler (Taggart & Cross 2009; Pound et al. 2012).

Botkin and Simpson (1990) estimated above-ground biomass and C for boreal forests in North America to be 4.2±1.0 kg/m2 and 1.9±0.4 kg/m2 respectively which, when extrapolated to a total forest area of 5,172,427 km2 gave biomass and C figures of 22.5±5 x 1015 g and 9.7±2 x 1015 g respectively. Moore (1996) estimated living above-ground biomass ("phytomass") in North American boreal forests as 12 x 1015 g, to which could be added 76 x 1015 g in soils (including dead and fallen trees) and 135 x 1015 g in peatlands. Extrapolating to a total area of 10.6 x 106 km2 (the average of the first two figures in the preceding paragraph), this would then give a figure of some 420 x 1015 g of C in soils and peatland combined. C burial figures are estimated at 49.3 Tg C y-1 (Chmura et al. 2011: area 13.7 x 106 km2), while soil C storage estimates in Averill et al. (2014) are (49.7-)61.4(-73.1) kg C m-2, NPP figures in the latter being (292-)319(-346) kg C m-2 yr-1. Estimates of the current C pool for forests in Russia, Canada, and Alaska, roughly boreal forest, are 88 x 1015 g (vegetation) plus 471 x 1015 g (soils), the area under forest cover being 13.7 x 106 km2 (Dixon et al. 1994). Figures for the total C in boreal forests in Taggart and Cross (2009: Table 1) is 703 Pg, almost 60% of the total forest C, while those in Carvalhais et al. (2014: tables S1 and S2) are some 505 PgC, ca 34.2 kgC m-2, and a mean turnover time of (45.4-)53.3(-73.4) years. Another way of putting it is that boreal forests contain at least 32% of the global C stock, sequestering around 20% of total forest C (Pan et al. 2011; Gauthier et al. 2015).

Buffam et al. (2014) noted the importance of both peat-containing wetlands and lakes in long-term C storage in rather mixed Wisconsin-Michigan forests of the Northern Highlands Lake District; at 33% of the area, they represented over 80% of the fixed C storage, and so were major players in long-term C sequestration there. Clemmensen et al. (2013), working on Swedish conifer forests, emphasized that in older, less disturbed forests much C came from roots and in particular from their fungal associates, and the latter also contributed substantially to total respiration, which affects rock weathering (see also Högberg et al. 2010; Tedersoo et al. 2012, also below). In younger forests cord-forming basidiomycete ECM were commoner, mycelial biomass turnover was fairly rapid, and little C was stored, while in older forests ERM, often with decay-resistent melanized hyphae, were associated with increased C sequestration (Clemmensen et al. 2014; see also Fernandez et al. 2013; Martino et al. 2018). Indeed, the factors affecting C storage are complex. For example, Pinaceae in boreal forests differ in the kinds of litter that they produce. There is massive accumulation of non-flammable litter under genera like Tsuga and Picea while the litter of Pinus is much more flammable and accumulates much less readily (Cornwell et al. 2015 and references). Furthermore, brown-rot fungi are commonly associated with Pinaceae, and these decompose wood less thoroughly than white-rot fungi, being unable to break down lignin (Riley et al. 2014; Floudas et al. 2015 and references), but there is no sharp distinction between these two types of fungi. Interestingly, bacteria related to Rhizobium dominate in coniferous forests where N-fixing plants are vanishingly uncommon (VanInseberghe et al. (2015).

Peat has figured prominently in the last few paragraphs, and it will continue to be important in this discussion as being a major element globally of short(ish) term C storage; J. Xu et al. (2018) provide a map of global peat deposits (see https://doi.org/10.5518/252). Most extant peat deposits in temperate-subpolar regions seem to have formed within the last 17,000 years or so after the last glacial maximum (Kleinen et al. 2016; Morris et al. 2018), however, as Treat et al. (2019) note, the balance between peat formation and sequestration is complex. Over the last 130,000 years northern peatlands in general have higher peat production in interglacial periods, however, peatlands near the coast may be buried by sediment as sea levels rise (and so the carbon is sequestered), but on the other hand during colder periods mineral deposition, as by glaciers, may result in the burial and sequestration of recently formed peat, at the same time the fall in sea levels exposing new areas suitable for peat formation. There are comparable dynamics for tidal marshes in particular (temperate, some subtropical) and tropical peatlands, which include mangroves and coastal dipterocarp forests (Treat et al. 2019; Rogers et al. 2019; FitzGerald & Hughes 2019).

Age: The age of ECM Pinaceae is uncertain. They may be some 350-200 Ma old (see Eckert & Hall 2006), although the earliest fossils identified as Pinaceae are from Upper Jurassic deposits ca 155 Ma (Rothwell et al. 2012). Other estimates of the age of crown-group Pinaceae range from (271-)153(-136) Ma (Gernandt et al. 2008; Magallón et al. 2013) to late Cretaceous or even younger (Willyard et al. 2007; Crisp & Cook 2011). Crown and stem ages of around 100 and 263 Ma respectively were suggested by Quirk et al. (2012), of course, establishment of ECM associations can be any time between these estimates. Fossils of crown-group Pinus have been dated to 140-133 Ma (Ryberg et al. 2012; Falcon-Lang et al. 2016a, b, c.f. Hilton 2016), although the genus may be as old as 237 Ma (He et al. 2012). Pseudolarix was widely distributed in the northern hemisphere at latitudes above 400 in the Early Cretaceous (Barremian, 115 Ma). However, most crown-group diversification in Pinaceae is much more recent, largely Palaeogene or younger, and the extensive boreal forest/taiga biome, where Pinaceae now dominate, is young, having developed only within the last 12 Ma (Taggart & Cross 2009; Pound et al. 2012).

Fagales, particularly Nothofagaceae and Fagaceae, are common in forests of temperate areas in eastern and western Eurasia and North America, in southern temperate regions, and on hills and mountains in Central America and Malesia; see below for the fossil history of this clade, which is quite well documented. They often grow in association with ECM Pinaceae, as in eastern North America (e.g. Abrams 1996). Indeed, in most of these forests ECM and AM trees are overall about equal, although ECM trees predominate in the north and also in the southeast (Phillips et al. 2013).

Fagaceae frequently dominate north temperate vegetation. White oak (Quercus alba) alone represents (12-)19-26(-49)% of witness trees, i.e. trees that were probably present before Europeans arrived, in the oak-dominated forests of eastern North America (81% in some southern Illinois forests). White oak, which grows with up to three more ECM species, two of which are usually other Fagaceae, makes up anything from (36-)50-80(-± 100)% of all trees (Abrams 2003). Six of the 30 species of Quercus growing in those forests are notable dominants (Abrams 1996). Fagaceae, again mostly Quercus, are abundant in western North America, and in California the black oak, Quercus kelloggii, is particularly widespread and has the greatest timber volume of any oak (Waddell & Barrett 2005). Oak trunks may get buried in sediment in flood plains, and the mean age of C storage in such conditions is ca 1,960 years, individual trunks being up to 14,000 years old (Guyette et al. 2008), while in mixed ECM temperate forests the half-life of conifer wood was about 20 years, but buried wood persisted for up to 1,400 years (Hyatt & Namian 2001).

The ECM American chestnut, Castanea dentata, was previously the dominant large tree in some 800,000 km2 of forest in eastern North America, but it now persists largely as suckers after its devastation by chestnut blight in the first half of last century (Thompson 2012). However, Faison and Foster (2014) qualify earlier literature reports, i.a. noting that some of the dominance of chestnut may be quite recent, the result of coppicing after being cut down by early European colonists. In any event, it has been replaced by mixed oak or oak-hickory forests (Abrams 1996; see e.g. van der Gevel et al. 2012 for the future), so the forests remain dominated by ECM trees.

Age: The development of ECM associations is likely to be an apomorphy for Fagales. Sauquet et al. (2011) give a range of crown-group ages for the clade, (124.8-)120.2-67.3(-48.9) Ma, and these include the ranges of dates suggested in eight other studies. Stem-group ages for Fagales, i.e. the earliest that the ECM association could have developed here, are 125-84(-73) Ma (e.g. H. Wang et al. 2009; X.-G. Xiang et al. 2014; Tank et al. 2015) Palynological evidence suggests that Fagaceae were diverse 42-40 Ma in western Greenland (Grímsson et al. 2015). Judging by pollen and wood fossils in particular, Nothofagaceae have been very important in southern latitudes (Australia, Patagonia, Antarctica) since the mid-Campanian ca 78 Ma (Cantrill & Poole 2012). Nothofagus is thought to have dominated in mid-Eocene forest on Wilkes Land, the eastern Antarctic, ca 66o S (Contreras et al. 2013) and it made up a major component of Antarctic vegetation and biomass through the Palaeogene, perhaps persisting there in tundra-like vegetation until the Pliocene (Cantrill & Poole 2012).

Plants with distinctive pollen assignable to the Normapolles complex, comparable with pollen of a subset of Fagales, but not Nothofagaceae or Fagaceae, were both diverse and ecologically prominent in the northern hemisphere from the Turonian-Campanian 94-80 Ma (but c.f. Batten 1981, 1989, and Polette & Batten 2017 for cautionary comments, e.g. on pollen identification). Molecular estimates for the age of this clade range from (50-)41, 37(-28) Ma (Bell et al. 2010) to (96.9-)93.4(-88.2) Ma (X.-G. Xiang et al. 2014: c.f. topology). Normapolles pollen has been found in much of the area 20-45oN Cretaceous palaeolatitudes from eastern North America to west central Asia (Kedves 1989; Vakhrameev 1991; Sims et al. 1999; Friis et al. 2003a, esp. 2006a, 2010b and references). Elsewhere in the Northern hemisphere Aquilapollenites and Wodehouseia pollen, of uncertain affinities (Farabee 1993: the former variously linked with Santalales, Apiaceae and Caprifoliaceae-Morinoideae) predominated, and in tropical Gondwanan areas pollen of Arecaceae was common. If Normapolles pollen is correctly identified, it would suggest that Normapolles communities were ECM communities. A temperate Gondwanan pollen province was characterized by Nothofagites pollen, probably from the fagalean Nothofagus, another ECM plant (e.g. Pacltová 1981 for a review; Kedves & Diniz 1983; Friis et al. 2006b, 2010b; Nichols & Johnson 2008). However, there has been little discussion about any ECM-associated activities of these plants.

Dipterocarpaceae-Dipterocarpoideae Dipterocarpaceae (map: Pakairaimaea, close to Cistaceae - blue; Monotoideae - green; Dipterocarpoideae - red: from Trop. Afr. Fl. Pl. Ecol. Distr. 1. 2003; Gottwald & Parameswaran 1966; Ashton 1982) dominate large areas of Southeast Asian l.t.r.f.. Common in Malesia, and in particular Borneo, they form extensive areas of peatlands but are also abundant away from peatlands, but in both cases the soil is rather humus-rich. In Lambir forest, Sarawak, dipterocarps make up only 7.4% of the species but 41.6% of the basal area (918.41 m2); the figures for Shorea alone are 4.7%, 21%, and 467.8 m2, Dryobalanops aromatica and Dipterocarpus globosus between them accounted for 13.2% of the basal area, and seven dipterocarps (out of the ten most dominant species) accounted for 23.1% (Davies et al. 2005). Dipterocarps are also important elements in drier and more open woodlands. Shorea robusta (sal) is a gregarious tree that grows in monsoon areas from Pakistan to China, especially in the India-Assam-Myanmar area. Sal forests occupy 115,000 to 120,000 km2 (11.5 x 106 ha) and make up ca 15% of Indian forests (Tewari 1995). In Africa Monotes is a significant component of the rather dry forests and woodland otherwise dominated by ECM Fabaceae-Detarioideae (see below).

Dipterocarpaceae occupy about 56% of all tropical peatland areas, close to 250,000 km2, about 6.2% of peatlands globally, and they dominate on huge peat lenses (Page et al. 2011, 2012; Richards 1996; Rieley " Page 2016); note, these figures are not corrected to take account of the recent realization that there are ca 145,500 km2 of peat with ca 30.6 petagrams/30.6 X 1015 g. C in the Cuvette Centrale in the Congo (Dargie et al. 2017), with Uapaca paludosa, alone of the four common species mentioned, probably an ECM plant. This peat contains an estimated 68.5 Gt C, some 77% of the tropical and 11-19% of the global totals for peatlands (Page et al. 2011: above-ground biomass not included). These figures are over twice the total C storage in all other forests in Malaysia and Indonesia (S. Brown et al. 1993; Page et al. 2012), which also include a substantial component of other ECM trees. (Other estimates are 84 Gt C in tropical peat - Rydin & Jeglum 2013 and references; also Immirizi & Maltby 1992; Rieley et al. 1996, etc., including estimates of pre-human peatland areas). C in waters draining from disturbed dipterocarp peat swamps may be as much as ca 4,180 years old (Moore et al. 2013), indicating that C storage there can be quite long term. Indeed, in a single near-surface Bornean peat examined carbohydrate concentrations were lower and aromatic (lignins, tannins, etc.) concentrations were higher than comparable more temperate (northern hemisphere) peats examined, hence likely contributing to the relative recalcitrance of this tropical peat (Hodgkins et al. 2018: c.f. deeper temperate peats, see below). Peat deposits started to form in the late Pleistocene 40,000 y.a., and now the peat may be 25 m deep (Page et al. 2004, 2012; see Raes et al. 2014 for dipterocarps on the Sunda Shelf during glacial maxima).

Somewhat unusually for ECM forests, Malesian dipterocarp forests are very diverse on a global scale (e.g. Lee et al. 2002) and from this point of view represent one extreme of the ecological spectrum of ECM plants. They have high above-ground wood productivity, even when compared with west Amazonian forests largely similar in soil, precipitation, etc., the dipterocarps being taller trees gaining in diameter faster, and there is also higher solar radiation in Borneo (Banin et al. 2014). For Malesian dipterocarps, see also papers in Osaki and Tsuji 2016). Amazonian peatlands sequester perhaps 9.7 Gt C, but on a per area basis their C accumulation is only a little over half that in Malesia; unfortunately, the mycorrhizal status of the plants in these peatlands seems to be largely unknown (Lähteenoja 2011; Lähteenoja et al. 2011; Rieley & Page 2016).

Of other close relatives of Dipterocarpaceae that are ECM plants, Cistus in particular dominates in the shrubby Maquis vegetation in the Mediterranean region, while species of Sarcolaeanaceae are often reported to be common in Madagscar (Cavaco 1952).

Age: A substantial clade in Malvales, [[Pakaraimaea + Cistaceae] [Sarcolaenaceae + Dipterocarpaceae]], most/all members of which are ECM, can be dated to as little as (25-)23(-21) Ma (Wikström et al. 2001) or over 88 Ma, the split of Dipterocarpaceae and Sarcolaenaceae (Ducousso et al. 2004). Suggestions that ECM associations in Dipterocarpaceae developed before the break-up of Gondwana over 130 Ma (Henkel et al. 2002; Ducousso et al. 2004; Moyersoen 2006; Alexander 2006; see also Sato et al. 2016) are overly optimistic; later Cretaceous or early Caenozoic ages are more likely. There are massive amounts of dipterocarp resin from India in the Early Eocene around 52-50 Ma (Rust et al. 2010), although the botanical provenance of some fossil resins has been questioned (Kooyman et al. 2019).

Fabaceae are often not thought of as being ECM plants, but Old World Detarioideae, which dominate millions of square kilometres in Africa (blue, two maps above), are ECM plants (e.g. Read 1991; Onguene & Kuyper 2001). Species in some 36 of the ca 82 genera included in Detarioideae are reported to be at least locally dominant (e.g. Letouzey 1968; Mackinder 2005), and 11 of these dominants are in a rather small clade (Macrolobieae/the Berlinia clade) with 16 genera, of which 10 are known to be ECM plants (see also Wieringa & Gervais 2003). A few species of Detarioideae dominate ca 3.27-3.75 x 106 km2 of Miombo forests in the Zambezian region (estimates from White 1983; see also Chuyong et al. 2002: Korup, Cameroon; Newbery et al. 2006). The ECM detarioid Isoberlinia is a major component of Sudanian Woodland (White 1983) which forms an interrupted band south of the Sahara from Mali to Uganda (White 1983; upper band of blue in the map above). This forest is biogeographically closest to Miombo woodlands among other African vegetation (Linder et al. 2012).

Detarioideae represent 20-90% of the trees in Miombo forests, 30-96% of the basal area, and with biomass estimates in the range of 35-97 Mg ha-1 (Högberg & Piearce 1986; Frost 1996). Figures for the C dynamics of tropical savannas and grasslands together in Carvalhais et al. (2014: Tables S1 + S2) are around 328 Pg total C, a C density of ca 17.7 kgC m-2, and a mean turnover time of (12.2-)16(-22.1) years. Detarioideae are ecologically important elsewhere in Africa. Gilbertiodendron dominates large areas of the eastern Congo Ituri rain forest (Torti et al. 2001; Makana et al. 2011), one study finding that it made up 20% of the above-ground biomass there (Bastin et al. 2015). Microberlinia dominates Guineo-Congolian forests in Cameroon; see Chuyong et al. (2002) for the slow breakdown of ECM litter in that forest. Other Detarioideae dominate parts of the forest that grows inland from the coast from Sierra Leone to western Gabon, and again in the periphery of the Zaire basin (White 1983). A caesalpinioid Biafran forest subtype has been recognised that includes this sub-coastal forest, and of the 34 important genera recorded from it, 28 are Detarieae, and 11 of these are described as being characteristically gregarious (Letouzey 1968). Other ECM plants in the woodlands and savannas of Africa and Madagascar include Monotes (Dipterocarpaceae), Uapaca (Phyllanthaceae), Asteropeiaceae and Sarcolaeanaceae (Tedersoo et al. 2011 and references).

Detarioideae like Cynometra are arbuscular mycorrhizal (AM, endomycorrhizal) plants, and Cynometra, too, can dominate in tropical African rain forests (Eggeling 1947; Makana et al. 2011). Interestingly, litter type and amount, soil nitrate, etc., of the caesalpinioid AM Mora excelsa which can dominate forests in Trinidad (see Brookshire & Thomas 2013: it also has endophytes), are like those of several ECM-dominated communities. A few other Fabaceae, both ECM and AM, may also dominate locally (see also Peh et al. 2011b - check).

In the New World, Aldina (Faboideae) and the coppicing Dicymbe (Detarioideae), both ECM plants, dominate forests in the Pakaraima Mountains in the central Guiana Shield region (McGuire 2007b; M. E. Smith et al. 2011). Dicymbe has a remarkably high basal area of 38.4-52.5 m2 per hectare, around 25(-40) m2 being more normal figures (Henkel 2003).

None of the 42 common Amazonian species mentioned by Pitman et al. (2001) is known to be an ECM plant. However, of the 20 most abundant Amazonian trees, plants 10 cm d.b.h. or more, listed by ter Steege et al. (2013) as being "hyperdominants", the ECM Eperua falcata (see Peh et al. 2011b), along with E. leucantha (mycorrhizal status?), Detarioideae, are notable as being 50% more abundant (usually far more) than any other non-palm on the list. 5 of the 20 species with most above-ground woody biomass are Fabaceae (2 known to be ECM, 4.96% of the total biomass) as are 6/top 20 species ranked by productivity (Fauset et al. 2015); most of the others are probably AM plants (Béreau & Garbaye 1994). Peltogyne (Detarioideae: ?mycorrhizal status) is a rare Amazonian monodominant, and it occupies ca 53% of the basal area of trees 10 cm or more in d.b.h. on Maraca Island, Roraima, a figure that increases the larger the trees (Nascimento et al. 1997). For further discussion about Amazonian "hyperdominance", see above.

Age: There are suggestions that ECM associations in Fabaceae-Amherstieae (= Detarioideae) developed before the break-up of Gondwana over 130 Ma (Henkel et al. 2002; Moyersoen 2006), but an early Caenozoic date is more likely. Indeed, the crown age of this clade has been estimated to be ca 29.2 Ma (Lavin et al. 2005), ca 53.8 Ma, or as little as ca 17.3. Ma (Bruneau et al. 2008a). However, there are reports of several extant genera including Brachystegia (ECM) and Cynometra (AM) in Africa fossil in the Eocene 46-34 Ma and they seem to be dominants even then (Epihov et al. 2017 and references).

ERM Ericaceae, a clade including the old Epacridaceae, Prionotaceae, Empetraceae and Vacciniaceae, are characteristic and often very common in heathlands world-wide (green in map above, inc. tundra: end-papers in Specht 1979a; Read 1996: also Grubbiaceae and Diapensiaceae, the latter close to Ericaceae), including those in alpine and arctic tundra (Specht 1979b; Chapin & Körner 1995; Jonasson & Michelsen 1996; Michelsen et al. 1998), in montane shrubberies especially in the northern Andes (especially Vaccinieae), in parts of the eastern Himalayan-Yunnan region and Malesia (Vaccinieae, Rhodoreae), and in heathlands of southern Africa (Ericeae) and Australia (Styphelioideae). Tundra vegetation occupies some 8% of the global land surface (Read 1991; Chapin & Körner 1995; Gardes & Dahlberg 1996; Camill et al. 2001; Kranabetter & MacKenzie 2010), or around 9.7 x 106 km2 (Melillo et al. 1993: alpine tundra included).

Age: ERM associations in Ericaceae have been dated to ca 90 Ma (van der Heijen et al. 2015 and references), but the fossil Paleoenkianthus on which this age is based may not even be stem Ericaceae, and in any event Enkianthus itself does not have ERM; the origin of ERM here can perhaps be dated to around 77-65 Ma (Wagstaff et al. 2010) or around 77 Ma (Z.-Y. Liu et al. 2014), with a stem group age of around 91 Ma (Liu et al. 2014). However, Martino et al. (2018) dated the age of the common ancestor of the four ascomycete ERM fungi (Leotiomycetes) that they sequenced at ca 118 Ma, agreeing with the age of the family - ca 117 Ma - in Schwery et al. (2014), but the age of origin of ERM association from that study would have to be younger (the AM Enkianthus is sister to the rest of the family and the mostly ECM [Arbutoideae [Monotropoideae + Pyroloideae]] is the next branch up), not to mention that the 117 Ma estimate is itself remarkably high.

Tundra, Boreal Forests and Permafrost. ECM/ERM plants are abundant in boreal forests and tundra in particular, indeed, all boreal tree species have ECM associations (S. E. Smith & Read 2008), and most tundra ECM fungi are found also in the adjacent boreal forest (Hewittt et al. 2017). Tundra-type habitats are often dominated by Ericaceae, Vaccinium and Empetrum being two of the seven prominent biomass accumulators there (Chapin & Körner 1995; Gardes & Dahlberg 1996; Kranabetter & MacKenzie 2010), and Read (1993) characterized the tundra by the prevalence of ERM. Ericaceae represent 30-87% of the above-ground biomass and 40-83% of the net annual above ground primary productivity in tundra (figures from Bliss 1979), and ECM and ERM plants together made up more than 95% of the vascular plant biomass in some heath tundra sites (Michelsen et al. 1998; see also Sistla et al. 2013: 6/10 species listed in Table 2). Three of the other major biomass accumulators, Salix, Betula and Dryas, are also ECM plants, while two Cyperaceae (Eriophorum, Carex: see below) are the others (see also Timling & Taylor 2012 for mycorrhizal diversity). ECM Polygonum (Bistorta) viviparum is a perennial herbaceous tundra plant that is both a pioneer and a prominent component of established vegatation (e.g. Gardes & Dahlberg 1996; Michelsen et al. 1998; Brevik et al. 2010). The mycorrhizal status of Diapensiaceae (Ericales) needs clarification, although it may have an ericaceous-like ectendomycorrhiza (Brundrett & Tedersoo 2018); the family is not immediately related to Ericaceae, and only a single species, the circumpolar Diapensia lapponica, is to be found in the tundra.

Interestingly, Betula and Salix (and Alnus) may be prominent in the forests developing at subarctic tundra treelines, and below-ground root turnover in such forests is fast, carbon not accumulating, the result being that soil organic carbon content is lower than in adjacent heaths - in this case with Empetrum and Vaccinium (Parker et al. 2015: canopy height 19 cm?). Permafrost

The northern polar permafrost encompasses the tundra region and also much of the boreal conifer forest above 60o N if areas of patchy permafrost are included, although permafrost is widespread only above 70o N in western Asia and even further north in much of Europe. The permafrost area includes extensive bogs dominated by Sphagnum and Cyperaceae (e.g. Camill et al. 2001). The map opposite is based on that by J. Brown et al. (2001, q.v. for detail; c.f. Tarnocai et al. 2009) - note that quite extensive areas with some, but less than 50%, permafrost are not included. Tarnocai et al. (2009) estimated permafrost to occupy about 18 x 106 km2, about 16% of the global soil area.

Very large amounts of C are in long-term storage in the extensive peat deposits (this is essentially necromass) to be found in tundra, boreal forests, and heathlands, mostly quite near the poles. Thus there is around 98.2 gT of SOC (soil organic C) stored in the 3.04 x 106 km2 of treeless North American Arctic soils (Ping et al. 2008: to 1 m deep). Estimates in MacDonald et al. (2006) are that northern peatlands stored 188-455 Pg C; Z. Yu et al. (2010) thought that there were around 547 gigatons of C there, with an additional 15 gt in Patagonian peatlands; or there may be some 16 Gt C in southern peats and as much as 621 Gt C in northern peats (Rydin & Jeglum 2013 and references; see also Gorham 1991; Immirizi & Maltby 1992). Figures for C storage in the tundra alone are about 158 PgC (Carvalhais et al. 2014: tables S1 and S2) and for that of the permafrost areas as a whole 1,035±150 PgC, but add to that another ca 272 Pg stored below 3 m in the Yedoma region, Russia, and in deltaic deposits, with around 822 Pg of this being perpetually frozen (esp. Hugelius et al. 2014; Schuur et al. 2015; Olefeldt et al. 2016), 1035 pG being about one third the global total (Jobbágy & Jackson 2000: to 3 m depth).

Interestingly, although near-surface peats (dominant plants various - Sphagnum, Cyperaceae, shrubs) may have relatively high carbohydrate and low aromatic (lignins, tannins, etc.) concentrations, at greater depths the proportions reverse, so making the deeper peat relatively more recalcitrant (Hodgkins et al. 2018). Although Sphagnum peat is relatively rich in carbohydrates and low in lignin, breakdown of moss peat in general, and Sphagnum peat in particular, tends to be slow (Turetsky et al. 2008). In Sphagnum cell wall pectin-like polysaccharides and glycuronoglycones have antimicrobial activity and sequester nutrients like N efficiently because of their high cation exchange capacity, so making them unavailable to microorganisms, hence the resistance to decay of Sphagnum litter which has a low respiration rate - microorganisms do not thrive there (Painter 1991; Hájek et al. 2011); this is also discussed elsewhere.

Soils contain much of the C in tundra ecosystems (e.g. Gorham 1991; Ping et al. 2008: ca 60% in non-permafrost conditions), and although the net primary productivity of Ericaceae there may be high, they are only one the major sequesterers of C in peat soil, thus mosses, especially Sphagnum, are very important contributors to peat (Turetsky et al. 2008). Indeed, C cycling in these environments needs more study. Comparing ECM and AM plants in subarctic alpine Sweden, Soudzilovskaia et al. (2015) found much variation, although there was substantial soil C from fungal hyphae of ECM plants, and soils in areas dominated by ECM plants showed evidence of slower C cycling than AM-dominated areas. Of course most of the C in the thick, C-rich sediments in the Yedoma region, made up of deltaic deposits, etc., is currently immobile, while some 20% of the circumpolar permafrost has thermokarst areas with a distinctive topology that are susceptible to rapid C release on further warming (see Tesi et al. 2016 for rapid C release in northern Siberia with warming at the end of the Younger Dryas ca 11,600 y.a.), and these areas store 30%, or perhaps 50% if Yedoma lake thermokarst is included, of the global C storage (Oldefeldt et al. 2016). Interestingly, N derived directly from rock can greatly increase ecosystem C storage in coniferous forests, and such forests grow in parts of the globe where relative increase in total N coming from rock breakdown is highest, sometimes over 100%; this rock breakdown may be facilitated by ECM fungi (Morford et al 2011; Houlton et al. 2018; Perakis & Pett-Ridge 2019: Alnus). In another wrinkle, C uptake by peatlands is fairly constant, while marine CO2 production during CaCO3 preciptation increases as sea level rises, decreases as it falls (Kleinen et al. 2016 and references).

Mosses are very important components of the tundra and boreal forests and represent a substantial proportion of the biomass (Gorham 1991; Chapin & Körner 1995). Particularly in more boggy areas Spagnum is the main moss, and it decomposes more slowly than the other bryophytes and vascular plants there (Verhoeven & Liefveld 1997: secondary metabolites; Lang et al. 2009; Lindo et al. 2013; Sistla et al. 2013; Rydin & Jeglum 2013), although mosses in general tend to decompose slowly (Turetsky et al. 2008). Sphagnum-dominated poor fens in northern Alberta may not be very productive, but respiration tends to be low, the plants start photosynthesizing early in the year, etc., so net C production may be higher than in, for example, Carex-dominated rich fens with their shorter growing seasons (Flanagan 2014; see also Ragoebarsing et al. 2005; Larmola et al. 2010; Hájek 2014; Bragina et al. 2104). Interestingly, the ERM ascomycete Oidiodendron maius is saprotrophic and can break down Sphagnum peat (Kohler et al. 2015).

The ecological role of Cyperaceae is poorly understood (Barrett 2013). Members of the family are often particularly common in rich fens in wet tundra habitats in the Arctic (ca 8% of the land surface), and include Eriophorum and Carex, two of the seven major contributors to the biomass there (Chapin & Körner 1995; see also Flanagan 2014 and above); Carex itself is the most speciose genus in the Arctic (Elven et al. 2011. Roots of cyperaceous plants may penetrate into the mineral soil below the shallow layer of soil dominated by roots of ECM plants (Read 1993). Cyperaceae-dominated communities were notably extensive during the last glacial maximum north of 550 N (Bigelow et al. 2003). Even today about 16% of all species growing in Quebec and Labrador north of 54o N are Cyperaceae, and 13% are Carex (Poaceae are next at 11%), and they can be major components of the plant cover, especially in wetter habitats (Cayouette 2008; Escudero et al. 2012). Moist sedge tundra is quite productive, and C cycles faster there than in adjacent dry tndra (Kade et al. 2012; Parker et al. 2015).

Habitats in alpine and other extreme conditions may be dominated by single species of Cyperaceae, and these are ECM plants. Thus there are some 450,000 (or 1.5 x 106) km2 of meadows between 3,000 and 5960 m altitude on the Tibetan plateau that are dominated by the ECM Kobresia pygmaea (= Carex parvula), and there is around 18.1 x 1016 g of C in the soil (Miehe et al. 2008, 2014; Qiu 2016, see also Zhou 2001). ECM Kobresia, now in the Core Unispicate Clade of Carex (Global Carex Group 2015), are widespread and sometimes dominant in other alpine, Arctic and tundra habitats (e.g. Gardes & Dahlberg 1996; Muhlmann & Peintner 2008; Newsham et al. 2009; Gao & Yang 2010).

Age: For the most part the clades discussed above are small and young. Of the seven major angiosperm accumulators of biomass in the tundra, Dryas (Rosaceae) diverged from other Dryadoideae (83.2-)63.1(-44.7) Ma (Chin et al. 2014) or ca 38 Ma (Y. Xiang et al. 2016), but ages for the other six, Vaccinium, Empetrum (Ericaceae), Eriophorum, Carex (Cyperaceae), almost 70 species of Salix (Salicaceae), and Rubus (Rosaceae) (Chapin & Körner 1995) are unknown, the species involved for the most part being members of large genera, although they are likely to be younger. Dating the most prominent bryophyte in today's Arctic, Sphagnum, is problematic. Sphagnum-like fossils are known from Ordovician rocks ca 455 Ma (Cardona-Correa et al. 2016), Carpenter et al. (2015) found spores of Stereisporites, linked with Sphagnum, to be very common in fire-prone heathlands in Central Australia 75-65.5 Ma, and Daly et al. (2011) suggest that Sphagnum-type mosses were components of peats produced by mire vegetation in northern Alaska ca 60 Ma that later were converted to coal. However, the clade to which extant species of peat-forming Sphagnum belong has been dated to as late as the mid-Miocene ca 14 Ma (Shaw et al. 2010a; Shaw & Devos 2014).

Another way of thinking about the age of tundra - and boreal - vegetation is to think about how long the climate that these vegetation types prefer has been around. The cold climate that characterises today's Arctic and Antarctic communities started to develop perhaps as much as 47 Ma in the Eocene, cooling being accentuated at the beginning of the Oligocene (e.g. Wolfe 1978; Millar 2011; Pagani et al. 2005; ). Estimates are of a 30o C or more reduction in the mean annual temperature in the far north since the end of the Eocene (Jahren 2007), temperatures dropping 8.2±3.1oC in just 400,000 years at the beginning of the Oligocene some 33.5 Ma in central North America (Zanazzi et al. 2007). Arctic ice started developing ca 7 Ma (Zachos et al. 2001), becoming widespread only in the early Pleistocene 2.4-2.2 Ma (Brigham-Grette et al. 2013; Knies et al. 2014), 10 Ma later than the development of major ice sheets in the Southern Hemisphere (e.g. Zachos et al. 2001, 2008; Retallack 2009; Millar 2011; Crisp & Cook 2011). The Antarctic ice sheet appeared ca 33.5 Ma at the Eocene-Oligocene boundary and persisted through much of the Oligocene (Zachos et al. 2001; Coxall et al. 2005; Eldrett et al. 2009 and references), although cooling in the Wilkes Land shelf area had begun by the middle Eocene (Pross et al. 2102) and there may have been short-lived periods of glaciation even ca 42 Ma (Tripati et al. 2005) or perhaps still earlier (Bowman et al. 2014; Ladant & Donnadieu 2016). Along with cooling, the climate also became seasonal, as is evident in extratropical floras by the end of the Eocene or somewhat earlier (e.g. Wing 1987; Eldrett et al. 2009), and of course climatic oscillations during the Quaternary have been extreme. But none of this seems to have anything to do with the subpolar-temperate peatlands mentioned above which seem to have developed within the last 17,000 years, largely as temperatures warmed after the last glacial maximum or, in the case of peats in the West Siberian lowlands, the triggering factor seems to have been increased precipitation (Morris et al. 2018). The long-term history of peatlands seems to be poorly known.

Seagrasses, Mangroves, and Tidal Saltmarshes.

Sea-grasses, mangroves and tidal salt marshes all have C burial rates well over 100 g C M2 y-1, considerably more than twenty times that in tropical, boreal or temperate forests, which are usually substantially less than 10 g C M2 y-1 (Mcleod et al. 2011; Chmura 2011). All three systems are accretionary, in that they also capture much sediment in which the C they produce, but also allochthonous C, is stored, and sediments can reach 10 m or more thick (Chapman 1974; McKee et al. 2007; Mcleod et al. 2011; Chmura 2011; Fourqueran et al. 2012; Marani et al. 2013; FitzGerald & Hughes 2019). C storage can be for thousands of years, well over ten times as long as that in l.t.r.f. (Chambers et al. 2001; Mcleod et al. 2011).

Seagrasses Sea-grasses are the only fully marine angiosperms, aside from some true grasses growing in estuaries (see below). There are about 55 species, all members of Alismatales, so they are unrelated to true grasses; a few are Hydrocharitaceae and the rest belong to the [[Posidoniaceae [Ruppiaceae + Cymodoceaceae]] [Zosteraceae + Potamogetonaceae]] clade. Sea-grasses often form monodominant stands, individual clones of some species being very long-lived. The map here - perhaps too optimistic - comes partly from Green and Short (2003), although there are substantial differences with individual family maps, particularly those for Zosteraceae and Ruppiaceae. This is partly because there is some disagreement over what a sea-grass is, but partly because Green and Short (2003) show very little data both for South America in general and places like the Solomon Islands in particular.

Estimates of the area occupied by sea-grass communities range from 22.8 x 106 (Waycott et al. 2009) to 30 x 106 (Duarte et al. 2005) to 60 x 106 ha (); although the last is an old estimate, it is relevant here where the emphasis is on conditions immediately before human activities became transformative. Although the amount of C in an individual sea-grass plant (= ramet) itself is small, that stored in the "soil", which can form mats up to 11 m thick, as in the Mediterranean sea, is very great (Fourqueran et al. 2012), larger than that of most forests and comparable with mangrove storage, and C can be sequestered for 4,000 years or more in the anoxic sea-grass beds (Orem et al. 1999; Serrano et al. 2011, 2013).

The gross primary productivity of sea-grasses has been estimated at 1903 g C m2y-1, rather like that of mangroves, their global primary productivity is 628 Tg C y-1, while their net ecosystem production (1211 g C m2 y-1 and globally 400 Tg C y-1) is substantially higher than that of mangroves because of their relatively low respiration rates. Sea-grasses are responsible for about 1.13% of all marine primary productivity, yet they bury as much as an estimated 27-44 Tg C y-1, some 12% of the total C storage in the marine ecosystem (Duarte et al. 2005: area 30 x 106 ha; Duarte 2011), although they occupy less than 0.2% of the area of the oceans. Indeed, this burial estimate may be only half the actual amount (Fourqueran et al. 2012). When thinking about sea-grass communities as C sinks, an estimate of 169-186 g C m-2 yr-1 seems reasonable - net community production of ca 120 g plus 41-66 g of allochthonous C (Kennedy et al. 2010: highest areal estimate above). Estimates in Mcleod et al. (2011) vary - they suggest a C burial rate of (100-)138(-176) g C m-2 y-1 (range 45-190), total C burial of 48-112 Tg C y-1, for a sea-grass area of 17.7-60.0 x 106 ha. Other estimates of global C storage by sea-grasses range from 4.2-8.4 or 9.8-19.8 Pg C, depending on the assumptions made, somewhat over 0.5% of the global total (Fourqueran et al. 2012; see Charpy-Roubaud & Soumia 1990 for estimates of benthic algal productivity). A substantial amount of sea-grass C moves into other marine ecosystems, including the deep sea (Suchanek et al. 1985). Bar-On et al. (2018) estimated that the biomass of marine plants, green algae and seagrasses together, was less than 1 Gt C, the total biomass of plants being around 450 Gt C.

To summarize. The sea-grass ecosystem is very productive, supports a considerable amount of diversity, does not suffer from much herbivory, captures much sediment, and stores much C, both autochthonous and allochthonous (Orth et al. 2006; Kennedy et al. 2010 for summaries).

Age: Of the two main seagrass clades, the stem age of [Thalassia + Enhalus + Halophila] in Hydrocharitaceae has been estimated to be 47.8-38 Ma (Iles et al. 2015) - other estimates are dramatically older. Within the [[Posidoniaceae [Ruppiaceae + Cymodoceaceae]] [Zosteraceae + Potamogetonaceae]] clade, the first split has been dated to ca 73 Ma and its stem age is dated to less than 82 Ma (Janssen & Bremer 2004).

Fossils of Thalssocharis bosquetii ca 72 Ma from the early Maastrichtian of western Europe have been identified as those of a seagrass, although to what clade they should be assigned is unclear - they seem to completely lack intravaginal squamules (van der Ham et al. 2017). Fossils from the late Middle Eocene in Florida have been identified as Hydrocharitaceae (Thalassia) and Cymodoceaceae (Thalassodendron, Cymodocea), indeed, some have been referred to extant species, and records from the Eocene in the Old World suggest a considerable age for the seagrass community (Lumbert et al. 1984; Ivany et al. 1990). Sea grasses may have originated in the eastern Tethys in the Late Cretaceous, and some taxa recorded from the New World by the Eocene are now known only in the Old World (Ivany et al. 1990). It is unclear how many times adaptation to the marine habitat has evolved (three times or so), since individual species in sea-grass families and other Alismatales can tolerate a range of salinities (e.g. Barbour 1970).


Mangroves can be divided into two groups, the much more speciose eastern group, from east Africa to the western Pacific, which includes ca 40 species, ca 14 of which are Rhizophoraceae, and the western group, from west Africa to the Americas, with only eight species, three of which are Rhizophoraceae (for their evolution, see Ricklefs et al. 2006). Plaziat et al. (2001) suggested that the separation of the two groups occurred ca 20 Ma. Depending on how species limits are drawn, no dominant mangrove species is common in both areas (Tomlinson 1986, 2016). For salt and water balance in mangroves, see Reef and Lovelock (2015) and other papers in Ann. Bot. 115(3). 2015.

The mangrove ecosystem is very productive and also has high C flux rates (see Feller et al. 2010 for a good summary). Mangroves occupy 13.7-15.2 million hectares, and they store 4-20 PgC globally (Bouillon et al. 2008; Donato et al. 2011 and references; 16.7 m ha in Valiela et al. 2001). Other estimates are that they bury 17.0-23.6 Tg C y-1, their gross primary productivity is 2087 g C m2y-1, global primary productivity is 417 Tg C y-1, but with a rather lower net ecosystem production (221 g C m2 y-1, globally 44 Tg C y-1) because of a relatively high respiration rate, at least when compared with the sea grass community (Duarte et al. 2005: area 0.2 x 1012 m2). Estimates in Mcleod et al. (2011) are a C burial rate of (187-)226(-265) g C m-2 y-1 (range 20-949), total C burial of 25.7-40.3 Tg C y-1, area 13.8-15.2 x 106 ha. Spalding et al. (2010) estimated mangrove net primary productivity to be 140-168 tg y-1, of which 10(-30)% was incorporated into sediments where it made up 15% of the organic C accumulating in marine sediments globally, and this 10% of refractory organic C in marine sediments that is mangrove in origin equals the amount of C in atmospheric CO2. 10% of terrestrially-derived dissolved organic C in the oceans comes from mangroves (Dittmar et al. 2006). Mangrove peat can become very thick, and C storage in it is long term, C in Caribbean peat having been dated to around 7,500 years old (McKee et al. 2007). Although Rhizophoreae, with their stilt roots, and Avicennia, with their numerous pneumatophores, grow in very different ways, both can produce large amounts of peat (Ezcurra et al. 2016: also estimates of productivity, peat age, etc.).

Age: There have been a number of independent adaptations to the mangrove habitat (Tomlinson 1986, 2016; Spalding et al. 2010), and Rhizophoraceae-Rhizophoreae and Arecaceae-Nypoideae-Nypa are particularly important mangrove plants. By the Eocene, ca 50 Ma, many mangrove genera are known from the fossil record, and several are known from both the Old and the New World (Plaziat et al. 2001; Ricklefs et al. 2006 for some dates; but see Martínez-Millán 2010 for Pelliciera). Thus Nypa, today found only in the Indo-Malesian area, is known from the Upper Cretaceous ca 70 Ma and by the early Palaeocene ca 55 Ma was found in both the Old and New Worlds (see Arecaceae). Fossil hypocotyls identified as Ceriops and preserved with good anatomical detail have been found in the Lower Eocene London Clay (Wilkinson 1981; but c.f. Collinson & van Bergen 2004). Rhizophora is known from the Caribbean in the late Eocene (Graham 2006) and Rhizophoreae from the Early Eocene 55-48.5 Ma in western Tasmania, Australia (Pole 2007).

Estuarine productivity is difficult to estimate. Estimates of C burial in salt marshes (2.2-40 x 106 ha in extent) are given by Mcleod et al. (2011): the rate of burial is (194-)218(-242) g C m-2 y-1 (range 18-1713) and total C burial is 4.3-96.8 Tg C y-1. Duarte et al. (2005) estimated that salt marshes occupied ca 40 x 106 ha with a gross primary productivity of some 3595 g C m2y-1 and global primary productivity of 1438 Tg C y-1, while their net ecosystem production was 1585 g C m2 y-1 and globally 634 Tg C y-1, substantially higher than either mangrove or seagrasses. They estimated C burial to be 60.4-70.0 Tg C y-1. FitzGerald and Hughes (2019: emphasis on accretion and erosion) suggest that saltmarshes contain 0.1% of terrestrial C

Age: Pseudoasterophyllites, ca 97 Ma from the European Cenomanian, is possibly the earliest halophyte, and was described as growing in supratidal salt marshes; since morphologically, it tends to link Chloranthaceae and Ceratophyllum (Kvacek et al. 2016) it is not immediately related to any other extant halophytic group. Extant angiosperms that live in estauries include a variety of more or less salt tolerant members of Alismatales, Poaceae such as Spartina (= Sporobolus), some species of Juncus, Limonium, Atriplex, a few Asteraceae, etc. - overall, quite a number of species, but no major clades - so it is difficult to provide useful estimates of the age of estaurine vegetation.

Attempt at a Synthesis.


The ecosystem functions emphasized here are C sequestration and to a lesser extent net primary productivity, but the two are not necessarily linked (Lähteenoja 2011), thus, as mentioned above, production of peat and the sequestration of the carbon it contains are separate operations (Treat et al. 2019). C estimates can be of above-ground biomass, or in soils and peats, and in peats in particular below-ground C can represent over half the total forest C pool (Dixon et al. 1994), with sequestration times being relatively long term. On the other hand, in many speciose tropical lowland rain forests productivity is high, standing C biomass is high, but below ground biomass is relatively low, C sequestration times are short, and biomass turnover is relatively fast (e.g. Dixon et al. 1994; Peh et al. 2011a). There are suggestions that the total standing biomass of the trees in forests is invariant with respect to species number and composition or latitude (Enquist & Niklas 2001; Enquist et al. 2007; but c.f. e.g. Dixon et al. 1994), but above- and below-ground biomass, nutrient cycling and productivity do all vary considerably.

Communities in which ECM clades dominate often grow under relatively extreme environmental conditions, whether of substrate or climate (Read 1991; Augusto et al. 2014). ECM Pinaceae (also other Pinales, which are endomycorrhizal) successfully compete with angiosperms, but not in the most productive environments (Brodribb et al. 2012). The marine and estuarine environments inhabited by mangroves and sea-grasses are physiologically extreme for angiosperms. Ecosystem turnover times for C are much longer in cooler climates in which woody ECM plants become so prominent: Times for tropical, temperate and boreal forests, and tundra are ca 14.2, 23.5, 53.3, and 65.2 years respectively, and above 75o N the mean turnover time is ca 255 years (Carvelhais et al. 2014), all in all a land-efficient and low cost (when it comes to amount of N needed) way of storing C (Leifeld & Menichetti 2018). Note also that C residence time can also be very long in both dipterocarp and mangrove peats, radiocarbon ages ranging from 1,000-20,000 years or more (Page et al. 2004; Schmidt et al. 2011; Ezcurra et al. 2016).

It should be noted that ecophysiological boundaries between different mycorrhizal "types" are not that clearcut. Thus in some cases ECM and ERM are produced by the same fungus on different species of plant (e.g. Martijena 1998), and the effect of ECM and ERM on the environment may not be that different (see also above). On the other hand, related taxa may differ in their mycorrhizal associates, and the ecological attributes under discussion may be associated with plants with different types of fungal associations. Thus not all monodominant legumes are ECM (Torti & Coley 1999; Torti et al. 2001), an example being the AM Mora excelsa, but, like ECM legumes, it has low foliar and litter nitrogen contents are low, there is foliar resorbtion of nitrogen, soil nitrate concentrations are low, litter decomposes slowly and accumulates (Brookshire & Thomas 2013). Two other species of AM Mora may also be monodominants, as is the AM Prioria copaifera, Pentaclethra macroloba (also clonal), Cynometra alexandri (Torti et al. 1997; Henkel 2003; Makana et al. 2011; Peh et al. 2011b; Menge & Chazdon 2015, etc.). Again, communities dominated by these legumes may have some of the features of those dominated by ECM plants, and this is also true of communities dominated by the AM Piranhea mexicanum, in Malpighiales-Picrodendraceae (Martijena 1998: as Celaenodendron).

Answering questions like, "How many times did ECM associations/seagrass communities develop?, When did they evolve?, When did they become common?", is central to our understanding of the long-term effects on the biosphere of the various associations under discussion (see also Eastwood et al. 2011; Maherali et al. 2016: discussion on evolution of different types of mycorrhizal associations). In these associations, one or a few clades largely dominate important aspects of community/ecosystem functioning, sometimes over very large area - relatively few clades of ECM plants occupy perhaps 50% of the earth's forested areas (L. L. Taylor et al. 2011). The dominance of a relatively few groups of plants in often rather unproductive terrestrial environments and particularly marked as one proceeds polewards, or in marine environments, may reflect the relative rarity of successful adaptations to more extreme conditions. Although the physiological/ecological traits under discussion have evolved several times, all have a strong phylogenetic signal. The clusters of origins of C4 photosynthesis in the PACMAD clade of Poaceae, and again in Cyperaceae and in Amaranthaceae (e.g. Kadereit et al. 2012), and the separate origins of adaptations to life growing completely submerged in the sea in Alismatales, suggest further complexities underlying the evolution of these traits. Similarly, the ECM habit has originated several times in the N-fixing clade.

However, understanding ecophysiological relationships earlier in the Caenozoic, let alone in the Mesozoic, presents major challenges. Three immediate issues arise when thinking of the possible changes of C sequestration over time.

1. I have already noted that the definitions of vegetation types are imprecise. Thus the area of grassland mentioned above depends on how "grassland" is defined, similarly, there is no consensus over the definition and hence the extent of vegetation types in the forest/savanna transition (Parr et al. 2014; Timberlake et al. 2010; Torello-Raventos et al. 2013; Bond 2016a, c.f. DeWitt et al. 2016 and Bond 2016b; Denk et al. 2018 and Fortelius et al. 2019; J. Xu et al. 2018: "peat"). Similarly, although I have separated mangrove- and sea grass-dominated communities, members of both are halophytes, that is, plants tolerating at least 200mM salt, and both intergrade with estuarine and inland halophytic vegetation, the former including abundant Poales, especially Poaceae, the latter often dominated by Caryophyllales (see also Veldkornet et al. 2015). In both these other vegetation types C4 plants are common (Flowers et al. 2010), although this is not a feature of plants of the marine habitats.

2. Estimates of the amount of above- and below-ground C and similar measures for current ecosystems are estimates for ecosystems that have often been more or less profoundly modified by the activities of humans. The effects of climate change on productivity and C sequestration that are mediated by changes in temperature, precipitation, atmospheric CO2 concentration, etc., are of course ubiquitous (e.g. H. Chen & Luo 2015 and references). The focus here is on pre-agricultural vegetation, yet as soon as humans started using fire, they began to cause major vegetational changes - and when fires are prevented, there are also major changes in the vegetation, although fires are likely to have been central to the existence of various grass-dominated communities for the last 10 Ma or so (e.g. Retallack 2001; Bond 2016a). Indeed, the Late Quaternary megafaunal extinctions, largely caused by human hunting, perhaps exacerbated by climatic fluctuations, have occurred both in land and marine communities (Janzen & Martin 1982; Valentine & Duffy 2006; Lorenzen et al. 2013; Cooper et al. 2015; Saltré et al. 2015) and have had substantial effects on community composition, nutrient cycling, biome limits, and perhaps the very existence of communities like the dry- and cold-adapted Arctic grasslands (Gill 2013; Hoffmann et al. 2017 and references). Overall, the areas occupied by different vegetation types have changed greatly over the last 10,000 years, hence, in part, the differences in some of the area estimates (Dixon et al. 1994). Longleaf pine savanna has decreased from ca 90 million to less than 2 million acres, and the area occupied by mangroves has also decreased greatly because of cutting (e.g. Spalding et al. 2010). Stands of thermophilous Abies alba in the Mediterranean have disappeared because of human activities (Tinner et al. 2013). Even in "primary" forests, human activities can have an impact on biomass estimates. Thus there was ca 1/3 loss in biomass in primary - but obviously not untouched - forests in Peninsula Malaysia over a single decade late last century (Kerridge et al. 1987; Dixon et al. 1994 and references) and the above-ground C in forest throughout the whole Indo-East Malesian area is substantially below its potential value because of human activities (S. Brown et al. 1993). Forest plots in Amazonia that provide the data used to understand Amazonian forests are notably common in areas that are likely to have been affected by the activities of pre-contact humans (McMichael et al. 2017a; see also Watling et al. 2017a, b; Palace et al. 2017; c.f. Piperno et al. 2017), moreover, the "hyperdominance" of some tree species may reflect their response to pre-Colombian human activity in which plants with fruit, etc., of value to humans were encouraged/domesticated starting some 4,500 years ago (Levis et al. 2017; Junqueira et al 2017; Maezumi et al. 2018; c.f. McMichael et al. 2017b).

Estuarine and salt marsh vegetation have been very considerably affected by human activity. Their floristic composition has changed and their extent reduced because of agricultural runoff, drainage, grazing and reduced sedimentation (e.g. Chapman 1984; Brush & Hilgartner 2000; Kirwan et al. 2011; Mariotti & Fagherazzi 2013). However, land clearance and the resultant increase in sediment in rivers can also facilitate the development of estuarine salt marshes (e.g. Kirwan et al. 2011; Chmura 2011), while agricultural nutrients in runoff negatively affect subaquatic estuarine vegetation (e.g. Brush & Hilgartner 2000). Mangroves have also been severely impacted by humans, and towards the beginning of the last century there may have been 22.0-25.5 x 106 hectares (estimated, from Valiela et al. 2001, current figures corrected by those multiyear records that exist, also with the lower current estimate of Spalding et al. 2010), but now there is at most 3/5 of this area. However, loss of mangrove vegetation since the beginning of the century may be rather less than supposed, even if what will happen in the immediate future remains unclear (Richards & Friess 2015).

However, depending on the nature and intensity of human activities, not all ECM plants, for instance, will be negatively affected. ECM oak has increased in eastern North American forests since Europeans arrived there (Abrams 1996), while grazing pressure associated with the spread of the Tibetan empire in the seventh century CE facilitated the development of the some 450,000 km2 of ECM Kobresia pygmaea (= Carex parvula)-dominated community of Tibetan plateau (Miehe et al. 2008, see also Zhou 2001). In other cases the basic vegetation type may be unaffected, despite human activities. If ECM ecosystems include members of different clades of ECM plants there may be ecological complementarity and stasis (c.f. Cadotte et al. 2012); Salicaceae, Betulaceae and Pinaceae, all ECM plants (or mostly so), may interact in this way in Boreal forests. Castanea dentata, an ECM plant and previously dominant in much of the eastern U.S.A., has suffered ecological death there, but perhaps with little overall effect, since it has been replaced by ECM oak-hickory forests (Abrams 1996). Although the species of trees growing in eastern deciduous forests in the eastern North America have changed considerably - and continue to change - in response to logging pressure, changing fire regimes, etc., since the advent of Europeans, the dominant species have remained ECM plants, even if their relative abundance has changed (Abrams 1996, 2003). However, with the suppression of fires in the last three hundred years or so replacement of oak-pine forests by largely non-ECM species does seem to be under way (Abrams 1996, 2003). The extent to which the ECM-dominated Mediterrananean Maquis vegetation reflects human activity is unclear, but again, the major components of the different successional stages are all ECM plants (Comandini et al. 2006). ECM trees remain important in Californian forests, even if tree size has changed and the ECM species are different, with oaks replacing pines (McIntyre et al. 2015; Gehring et al. 2017b and references).

3. {The next three paragraphs rather heterogeneous - to be reorganized.} The last issue is that adding a temporal component to thinking about ecosphere effects of the vegetation makes life very difficult. How have ecosystems with relatively small groups of species that have a disproportionately great influence on current global ecology behaved over time? Boreal forests, grasslands, peat swamps, mangrove vegetation and the like are not fixed and invariant elements of the biosphere; their extents, and the roles that individual species play in them, have changed over even quite short periods (e.g. Zobel et al. 2018), and as we think about the longer term, change has been ubiquitous - and with climate change, it will continue (J. W. Williams et al. 2007; Maguire et al. 2015; McDowell et al. 2016; Reich et al. 2018). Subpolar-temperate peatlands in both hemispheres seem to have developed within the last 17,000 years, largely as temperatures warmed after the last glacial maximum (Morris et al. 2018); what were their distributions earlier in the Quaternary and in the Neogene? Grasslands, C4 grasslands in particular, and boreal forests/taiga (Taggart & Cross 2009; Pound et al. 2012), both very extensive and ecologically/climatically very important today, are also both novel and rather young biomes that have developed within the last 10 million years or so. What about other communities that in the past had the same or similar species or the same genera as the communities of today? Has their behaviour been fixed over time?

Becklin et al. (2016) discuss how the ecophysiology of individual species may change in response to Pleistocene and Holocene climate changes. The relative recency of the development of latitudinal gradients of diversity is just one aspect of this problem (see above), while our understanding of vegetational, etc., change in response to changing climate, e.g. grasses in grassland, may be depend in an unexpected way from results coming in from longer-term experiments (J. W. Williams et al. 2007; Maguire et al. 2015; McDowell et al. 2016; Reich et al. 2018). Jackson and Williams (2004: see also J. W. Williams & Jackson 2007; Taggart & Cross 2009: late Cretaceous to late Eocene polar deciduous forest; Veloz et al. 2012; Donoghue & Edwards 2014; Maguire et al. 2015; Sluiter et al. 2016: S.E. Australia peat swamps) discuss "no-analog communities" of the past that have species combinations unlike those of any current communities, often associated with "no-analog climates" and "no-analog physiology", indeed, even if one can link extinct and extant taxa phylogenetically, the ecophysiology of the two may be quite different (J. P. Wilson et al. 2017). In the Late Cretaceous-Miocene Ginkgo was common in disturbed streamside habitats, growing along relief/abandoned channels and unstable crevasse splays, but not in backswamps (see also Jordan 2011; Jordan et al. 2019); the morphology and habitat of extant G. biloba would hardly lead one to guess that Ginkgo grew in such habitats then (Royer et al. 2003: frequent associates included Platanus, some species of which are still to be found in such places, Metasequoia and Cercidiphyllum). Thus, to say that the accuracy of the use of nearest living relatives as a guide to the past "decreases with an increasing age of a palaeoflora" (Uhl 2006: p. 95; see also Wing 1987, etc.) rather understates the problem; indeed, the Ginkgo paradox refers to situations in which there may be a tight correlation say between the stomatal index of extant plants and the atmospheric CO2 concentration, but this correlation breaks down when you extrapolate to fossils ( (Jordan 2011). Little et al. (2010) noted that there was no evidence that such iconic temperature indicators as leaf teeth in fact indicated temperature in the past - latitude, perhaps, but temperature, no. However, Zohner et al. (2019, see also Royer & Wilf 2006) suggested that teeth indicate leaves that are preformed and can photosynthesize immediately they emerge from the bud, and are indeed correlated with climate. Little et al. (2010) did find that many "climatic indicators" were in fact more or less linked with phylogeny, and conversely, that phylogeny could at times obscure the relationships of particular traits with climate. Even general patterns of associations in communities that had been stable for around 300 million years seem to have changed since the beginning of the Holocene (Lyons et al. 2016). Similar changes may have occurred in the relationships between plants and the insects that eat them, thus plant-sawfly associations may have become reorganised around the KTR (Schneider 2016). Stomatal number and size as predictors of atmospheric CO2 concentrations may be less informative than one would like (Jordan et al. 2010).

It can be difficult to assess the initial distribution and ecology of a clade that has a long stem (Leslie et al. 2018). What happened morphologically and ecologically along the ca 100 Ma stem of Picea (Leslie et al. 2018), for example, is unclear, while still more perplexing is the longer stem of crown-group angiosperms. And fossils may aggravate the situation. From these pages, it will be clear there are ever-increasing numbers of fossils assigned to extant genera, etc., that were found in parts of the world ways away from when the genus, for example, is now to be found. What was it doing ecologically then, and how does this relate to the ecology of the taxon today? Mediterranean-type habitats have been dated to Ma, but the ages of many genera now characteristic of such vegetation can be considerasbly older ().


Not all ECM/ERM communities are notably productive, but those in the boreal zone in particular (Dixon et al. 1994) sequester considerable amounts of C in their soils, while the accumulation of raw humus on the forest floor in tropical ECM forests has long been noted (e.g. Alexander 1989a). The rate of nutrient turnover in mono- or oligo-dominant ECM vegetation types varies. It can be very high, particularly in the tropics (Torti et al. 2001), and ECM-dominated woody vegetation there is not always species-poor, as White (1983) noted for Miombo vegetation and Beard (1946) for the Mora-dominated forests of Trinidad. The dipterocarp-dominated forests at Bukit Lambir, Sarawak, are among the most species-rich tropical forests anywhere (Lee et al. 2002). Interestingly, Mora in particular behaves like some conifers (Enright & Ogden 1995; Aiba et al. 2007) and is almost an add-on to the vegetation, communities with and without Mora being otherwise similar; rather similarly, emergent dipterocarps may form a separate stratum above the rest of the forest (see Ashton & Hall 1992). Bornean Dipterocarps show very high levels of above-ground wood production when compared with other species in the same community, and also when compared with forests in the western Amazon (Ecuador, Peru) that are similar in soils, precipitation, etc.; there dipterocarps were more productive by as half as much again than their non-dipterocarp counterparts (Banin et al. 2014). This high productivity was despite a much lower amount of phosphorus in the soil in the dipterocarp forests and a C:N ratio about 50% higher (Banin et al. 2014). Litle is known about underground C storage and ECM activities in the Amazonian forests.

The importance of ECM fungi for angiosperm evolution is not just because they facilitate the nutrient and water supply of their associates and make life difficult for non-ECM plants and for free-living microbes, but also because of their direct and indirect effects on the biosphere - on soil, on weathering, on C sequestration, and hence on the earth's climate. ECM plants increase mineral weathering, and rainfall, derived in part from transpiration, also allows more silicate weathering, and this weathering is a principal sink for atmospheric C dioxide (Boyce et al. 2010; Berner 1997). An increase in atmospheric CO2 removed by the rock weathering has been linked to the decrease in atmospheric CO2 concentration during the Caenozoic (Pagani et al. 2009; L. L. Taylor et al. 2009, 2011; Quirk et al. 2014). In drier years, there may even be competition between ECM and lignin-decomposing fungi for water, leading to a reduction in the rate of wood decomposition (Koide & Wu 2003). Finally, C is buried in sediments - at least medium-term sequestration - much more easily than in AM forests, particularly those in the tropics where C turnover is very fast (Tedersoo et al. 2012; see Schmidt et al. 2011 for the very long C turnover times in subsoils, mycorrhizae not mentioned). All these biogeochemical effects of ECM plants are as much caused by the activity of fungi and bacteria associated with the plant as by any activities of the plant itself, all three forming a functional whole (e.g. Landeweert et al. 2001; L. L. Taylor et al. 2009; Bonfante & Anca 2009).

General diversity and community/ecosystem stability in the face of environmental change are connected (e.g. Hautier et al. 2015 and references). How this happens is a matter of discussion. Petchey and Gaston (2002a, b) suggested that if functional diversity/functional traits in the community are to be conserved, a large proportion of species in that community will also have to be preserved; there is little redundancy in functional diversity. Even if dominant species can maintain ecosystem functioning in the face of the loss of rare species, at least for a time (e.g. M. D. Smith & Knapp 2003), phylogenetic diversity may still improve ecosystem functioning, although this may also depend on rainfall, temperature, levels of soil nutrients and CO2, etc. (see e.g. Chapin et al. 1997; Zavaleta et al. 2003; Maestre et al. 2012; Cadotte et al. 2012). Experiments measuring biomass production find that productivity and overall diversity become more closely linked over time (Reich et al. 2012); Isbell et al. (2011) found that 84% of the grassland species studied promoted ecosystem functioning at least under some conditions even in the limited periods during which their experiments were carried out. As conditions change, different species may assume importance - and conditions have never been fixed for long throughout the whole Caenozoic period. Some species may even be quite flexible in the ecological roles they play (Aizen et al. 2012), although other studies suggest more conservatism (Maherali & Klironomos 2007; Stouffer et al. 2012). However, relatively little of the work on community/ecosystem functioning has emphasized the kinds of communities that are the focus here.

Even since the beginning of the Pleistocene ca 2.6 Ma there have been great changes in community composition and location, many plant communities being quite novel; the present is rindeed at best an imperfect guide to the past (e.g. Meseguer et al. 2014b), even over the short term. Prior to 3.3 Ma, boreal forests with ECM taxa like pine, spruce, larch and birch grew in the east Siberian-North American-Greenland area from 60-80o N, although especially since 2.7 Ma the conifers, etc., have been replaced by tundra (Brigham-Grette et al. 2013). The association of Picea, Betula and Alnus, characteristic of Recent boreal forest, was first recorded in North America a mere 7,000 years ago (J. W. Williams et al. 2004). Mapping of post-glaciation forest changes in North America shows that some species have been fairly constant in abundance, if not in location, but they are mixed with other species that as it were appear from nowhere and come to be abundant over wide areas (e.g. Webb 1988; Williams et al. 2004; see also Jahn 1991). Many forest communities found in North America are Holocene in age, communities coming and going even in recent times (Curtis 1959; Williams et al. 2004 and references), and the same is true elsewhere in the world (Torres et al. 2013). Boreal peatlands are post-Pleistocene in age (MacDonald et al. 2006), even if some tropical peatlands are somewhat older (Page et al. 2004). The composition of tundra vegetation changed considerably from glacial to interglacial periods, and more C accumulated in the latter (Brubaker et al. 1995). Changes were complex (Lindo et al. 2013), and mosses were in places replaced by angiosperms with their rather more labile leaf litter, plant biomass showing an overall increase (Sistla et al. 2013). The changing relative proportions of forbs and graminoids in Arctic tundra and steppe over the last 50,000 years are detailed in Willerslev et al. (2014). As permafrost thaws, peat accumulation, especially by Sphagnum, but also by spruce, etc., may increase along with above-ground primary productivity (Camill et al. 2001).

The floristic composition of vegetation seems to have become much more "modern" in the latter part of the Caenozoic, and in North America forest composition may be little changed over the last 15 million years or so (A. Graham 1999; Hawkins et al. 2014). However, although biomes may have been fairly stable over the medium term, this is partly the result of how they are delimited, and individual species and their abundance may vary substantially within the one biome (J. W. Williams et al. 2004); communities, and biomes themselves, have certainly not been stable over this period (Moncrieff et al. 2016). The grassland and savanna biomes that are now such a prominent feature of vegetation can be dated to the Pliocene, within the last 10 Ma or so, and especially within the last 3 Ma (e.g. R. Sage et al. 2012; Pennington et al. 2006b; Simon et al. 2009; Simon & Pennington 2012). The boreal biome, now dominated by ECM taxa, is dated to 10-4 Ma (references in Fine & Ree 2006), while diversification of Sphagnum, now a prominent and ecologically important component of boreal forest and tundra vegetation, is dated to the middle Miocene ca 14 Ma (Shaw et al. 2010a; Shaw & Devos 2014; M. G. Johnson et al. 2014). Betula, now conspicuous in northern forests, has probably diversified within the last 10 Ma (Xing et al. 2014) - yet fossils that are remarkably like Sphagnum are known from Ordovician rocks ca 455 Ma (Cardona-Correa et al. 2016; see also L. E. Graham et al. 2017), while the palynomorph Stereisporites, linked to Sphagnum, was very common in fire-prone heathlands in Central Australia 75-65.5 Ma (Carpenter et al. 2015). What was the ecological role of these plants? Sea grass communities may be quite old, mid Eocene or older, the sea grasses living then being assigned to extant genera and even species. However, sirenian grazers of New World seagrasses, very diverse through the Miocene, declined in the Pliocene (Domning 2001), and Plio-Pleistocene New World sea grasses may have grown in waters to 20-30 m deep, compared with ca 10 m today (Budd et al. 1996). Again, aspects of the ecology of early sea grass communities may be rather different from apparently similar communities today.

In the Palaeocene and Eocene in particular species that today have different climatic preferences grew together, and in the late Eocene mixed deciduous broad-leaved and evergreen and deciduous conifer forests grew within both the Arctic and Antarctic circles (e.g. Collinson 1990; Jahren 2007; Harrington et al. 2011; Collinson et al. 2012; Pross et al. 2012; Cantrill & Poole 2012). A unique biome in which angiosperm and AM gymnosperm trees were mixed developed in Eocene South America below 24oS (Jaramillo & Cárdenas 2013), while much more recently during the warmest part of the Miocene 17-15 Ma in western North America there were associations of species unlike any extant (Millar 2011). The problem of no-analog communities can only increase with age (see also Cantrill & Poole 2012; Moncrieff et al. 2016). However, Bouchal et al. (2014) suggest that vegetation similar to that of the modern chaparral and nemoral conifer forest of the Coastal Ranges was to be found in the Late Eocene Front Range in west North America.

The current latitudinal diversity curves, with diversity commonly - but by no means always - peaking near the Equator, are a phenomenon of a post-Eocene globe with strong and seasonal N-S temperature gradients (see above). Even today, estimates of both living C biomass and dead and below-ground biomass are highest, not in l.t.r.f., but in some temperate and warm temperate forests, as well as mangroves and peat swamps (Keith et al. 2009; Pan et al. 2013: living biomass; c.f. in part Carvelhais et al. 2014). In the Mesozoic and early Caenozoic these curves were almost flat or they even peaked in more temperate areas (e.g. Wolfe 1987; Mannion et al. 2012, 2013; Archibald et al. 2012; see above). As temperatures dropped, particularly in the early Oligocene, vegetation with more local facies developed, seasonality became evident, and latitudinal diversity curved developed. Interestingly, tall trees (80+ m tall: Tng et al. 2012 for records) tend to be found in thermally equable climates, so the distribution of such trees in the early Caenozoic is an interesting unknown (Larjavaara 2013); a single tall tree can sequester a considerable amount of C.

Pinus seems to have been a mid-latitude (30-50o N) plant in the Cretaceous, but Pinaceae in general may have been negatively affected by fires fuelled by angiosperm shrubs and ferns that Belcher and Hudspith (2016) suggested had a major effect on the vegetation from the middle Cretaceous onwards. In the warm Palaeocene and Eocene Pinus retreated to high latitudes (Millar 1998; Daly et al. 2011: also Cupressaceae there, conflict between evidence from pollen and macrofossils; see also Augusto et al. 2014), although it also grew near the equator. In high latitude (65-80o N) Eocene floras Pinaceae could be quite common, and with other ECM plants they made up about 2/5ths of the species on Canadian islands 75-80o N (McIver & Basinger 1999). With the climatic deterioration of the Late Eocene-Oligocene, Pinus moved back to mid latitudes while persisting at higher latitudes (C. I. Miller 1993, 1998), and Late Miocene-Pliocene mountain building would favour Pinaceae (LePage 2003). In the Oligocene Pinus moved into western Malesia (e.g. Muller 1972).

OUT OF PLACE The proportion of wind-pollinated trees and shrubs are higher in humid areas away from the tropics (Regal 1982; Ollerton et al. 2011); temperate wind-pollinated trees and shrubs tend to be ECM plants, as in about half the examples mentioned by Regal (1982).

The decline of atmospheric CO2 over the last 120 Ma may in part be connected with the origin of clades of ECM plants (L. L. Taylor 2009, 2011; Quirk et al. 2012, 2014), although other contributing factors may include reduced volcanism from subduction along continental margins (McKenzie et al. 2016). If the present and past are connected, Normapolles and Nothofagites plants, along with most Fagales, were ECM; fossil remains of these plants are abundant in Late Cretaceous and early Cainozoic rocks (Friis et al. 2011 for a summary), and they may have had a transformative effect on the environment. Given the ages of Fagales and Pinaceae, ECM seed plants "may have [been found] over a larger area and for a much longer time period in northern temperate zones than in the tropics" (Tedersoo et al. 2012: p. 4167). ECM plants may even have progressively supplanted AM plants at weathering hotspots from some time in the Cretaceous (Taylor et al. 2011), indeed, ECM associations may be "the most profound alteration in root functioning to occur in plant evolutionary history" (ibid., p. 369). However, age uncertainties make life particularly difficult here; from the ages given above, the fungal ECM habit is at least sometimes likely to have originated rather later than the ECM seed plant clades on which the fungi are now found (see also Ryberg & Matheny 2012; Bruns et al. 1998; Horton & Bruns 2001).

The ideas of keystone species (e.g. Leighton & Leighton 1983; Terborgh 1986; Watson 2001; Watson & Herring 2012; Mouquet et al. 2012b)), species that directly or indirectly disproportionately control the resources needed by other organisms, and ecosystem engineers (C. G. Jones et al. 1994; Wright & Jones 2006; Sultan 2015), species - beavers, prairie dogs, leaf-roll caterpillars - whose activities substantially affect the environment, may be helpful here. The activities of members of the clades we are talking about have a disproportionate effect on the community, ecosystem or even biosphere relative to their species numbers (see Power et al. 1996). The clades under discussion have major effects at ecological scales from the local community up to the globe and are associated with major biomes or ecosystems (Pennington et al. 2004; see other papers in Proc. Royal Soc. B 359(1450). 2004), and it is at this level that the ecological interactions play out. Thus Brodribb et al. (2012; see also Coomes & Bellingham 2011) thought of conifers in general as being ecosystem engineers because of their major effect on the environnment. However, whether keystone clades or ecosystem engineers, they are not sharply distinguishable from all other clades in terms of their effects on the environment; for demolition of the simple idea that there are keystone species - species "important for something", see Hurlbert (1997).

It is increasingly a matter of comment that a number of clades seem to be more or less restricted to biomes (e.g. Schrire et al. 2004; Pennington et al. 2009; Dick & Pennington 2011; de Nova et al. 2012; Moncrieff et al. 2016), and in such cases ideas of phylogenetic, biome or niche conservatism are invoked: Clades retain niche-related traits or, more generally, have conserved ecological roles (e.g. Wiens & Donoghue 2004; Crisp et al. 2009; Crisp & Cook 2012; B. T. Smith et al. 2012; Kooyman et al. 2014). The ages of the ECM clades mentioned above are usually much more than 10 Ma, and so their evident ecological conservatism is relatively ancient (Tedersoo et al. 2014a). The term "phylogenetic conservatism" has been used in various ways in the literature on phylogenetic community ecology (Mouquet et al. 2012a), while niche conservatism seems to be little more than the recognition that some ecological features are associated with clades more than might have been expected, but this is true of subsets of many groups of characters.

Estimates are that tropical ecosystems store 47% terrestrial C and have 59% terrestrial primary productivity (10% of both in the Amazon Basin alone: Tian et al. 2010). Similarly, of global terrestrial net primary production, that in l.t.r.f. is around 36% (19.1/53.2 x 1015 gC) of the total, that of grassland + savanna ca 19% (9.7/53.2 x 1015 gC, of which over half comes from tropical savanna), contributions of boreal forest, temperate coniferous forest, etc., being less than 6% each (Melillo et al. 1993). On a per area basis the inequalities between biomes are clearer, thus soil C storage, at ca 11.7 kg C m-2 and NPP, at ca 956 kg C m-2 yr1 in tropical forests, compares with figures of 14.5 and 576 respectively for grassland and ca 61.4 and ca 319 for boreal forests (Averill et al. 2014).

Overall, the implications of these asymmetries in relationships between animals, plants, and the environment, are complex, but species numbers are clearly just one way of thinking about seed plant evolution. By focussing on the construction and maintenance of the ecological scaffolding of community structure over evolutionary time and in a phylogenetic context, angiosperms with dense venation, C4 grasses and ectomycorrhizal plants represent pillars, and ants, bumble bees, fruit bats and the like, arches and spandrels. These groups appear to have had a major role in constructing and maintaining the environment, while the bulk of the tens of thousands of gentianid species make up the paintings in the spandrels (apologies to Gould & Lewontin 1979). These paintings are forever changing as individual species go extinct, for instance because of the breakdown of plant/pollinator relationships, while other relationships are evolving. Plant communities come and go, and the relation between present, past and future is unclear (e.g. Torres et al. 2012). Over time the whole biosphere has changed as groups of plants with different eco-physiological capabilities assume prominence, and such changes help shape the background ecological context for the diversification of seed plants and of the animals associated with them.

10. In Conclusion.

Innovations in reproductive biology are thought to characterise the evolution of new plant groups, allowing increases in diversity in part by greater subdivision of the environment (e.g. Niklas et al. 1983). Gorelick (2001) summarized some twenty hypotheses that have been advanced to explain diversification/success of the angiosperms (see also Crepet & Niklas 2009; Onstein 2019), many having to do with flowers, and all told some 120 or more hypotheses have been advanced to explain the patterns of plant species richness that are such a distinctive feature of the environment (Palmer 1994).

A focus has been on understanding speciation within individual very speciose clades (e.g. Davies et al. 2004c), and much literature emphasizes the acquisition of "key innovations", apomorphic features whose advantages - sometimes more or less assumed - allowed a subsequent increase in the overall speciation/diversification rate of the clade in which they arose (e.g. Marazzi & Sanderson 2010), and there are related ideas like "ecological opportunity" (Yoder et al. 2010b) that have much the same flavour. Thus clades with latex (Farrell et al. 1991; see also Powell et al. 1999; Agrawal & Konno 2009: survey of laticiferous plants and latex; Konno 2011: chemistry), nectar spurs (Hodges & Arnold 1995; Hodges 1997; Kay et al. 2006), monosymmetric flowers (Donoghue et al. 1998; Neal et al. 1998; Endress 2001; Sargent 2004; Kay & Sargent 2009, see above; c.f. in part Kay et al. 2006), hummingbird pollination (Schmidt-Lebuhn et al. 2007), animal pollination (Eriksson & Bremer 1992; Kay et al. 2006b), self sterility (Ferrer & Good 2012), or the climbing habit (Gianoli 2004), are often more diverse in terms of extant species than their sister clades lacking these distinctive features (see also Ferrer et al. 2014 for vegetative and reproductive features associated with high diversification rates). However, Foisy et al. (2019) found no evidence for a connection between the evolution of latex and resin canals and Ehrlich and Raven's "escape and radiate" idea of co-evolution. Interestingly, Malpighiales and Ericales, disproportionately common among the small trees of the understory of tropical rain forests (Davis et al. 2005a), include taxa with many kinds of flowers and fruits. Neither clade can be well characterised either morphologically or chemically and they seem to lack innovations, whether key nor not, although the latter are sometimes evident in smaller clades within these two major groups.

It is a challenge to think about the evolution of the morphological and other novelties that are the focus here. Key innovations that cause the more or less immediate diversification of the clade in which they arise may be individually less important than we might like to think, and identifying key innovations is far more than simply linking a feature to a node on the tree (e.g. Sims & McConway 2003; Davies et al. 2004a; Donoghue 2005; Erkens 2007; Crepet & Niklas 2009; Marazzi & Sanderson 2010; c.f. Endress 2011a). Overall, determining that an innovation might be a key innovation is a difficult process, and speciation and the evolution of a key innovation are not necessarily to be linked (e.g. Cracraft 1990; Sanderson 1998; Galis 2001; Ree 2005b; Maddison et al. 2007). Indeed, Hedges et al. (2015) suggest that speciation and diversification are largely dominated by random events, while adaptive changes are something else again. Understanding the not-so-simple idea of persistence is also important (Leslie et al. 2013).

Key innovations are rarely simple features, rather, they may involve a complex suite of changes that occur over a protracted period, as with angiosperm flowers and vessels (e.g. Horn et al. 2012: Euphorbia subgenus Chamaesyce; Schranz et al. 2012; Donoghue 2005; Donoghue & Sanderson 2015; see also Stebbins 1951). Furthermore, the importance of some changes may be less in the changes themselves, but subsequent changes that they make possible and/or their importance in ecological conditions that may develop long after their origin. Thus Donoghue and Sanderson (2015) and Bouchenak-Khelladi et al. (2015) look at diversification and its relation to where characters that might be involved in that diversification actually change on the tree. Similarly, Edwards and Donoghue (2006) suggest that several key elements of the cactus ecological niche were established before the evolution of the cactus life form and subsequent diversification of Cactaceae (Ogburn 2007; Ogburn & Edwards 2009; Nyffeler & Eggli 2010). The evolution of C4 photosynthesis (e.g. Strömberg 2011) and the effects of genome duplications (e.g. Tank et al. 2015) both fit this model.

The increase in speciation that results from the acquisition of a key innovation has to be distinguished from simple radiation of a clade when it moves into in a new area, thus much speciation in Guatteria may have occurred only subsequent to its entry into South America (Erkens et al. 2007). CYC-like genes are widely involved in symmetry changes, especially in core eudicots (X. Yang et al. 2012; Preston & Hileman 2012), but direct links between gene and diversification remain to be established. Howarth and Donoghue (2004, esp. 2005) note possible connections between changes in such genes and changes in floral form in Dipsacales. Here crown-group Caprifoliaceae-Valerianoideae (315 spp.) are estimated to be 60-55 Ma (Bell & Donoghue 2005a), but diversification in the Andean paramo, which has resulted in ca 1/7th of the species currently recognized in the clade, happened only after their arrival in South America less than 5 Ma (Bell & Donoghue 2005b; Moore & Donoghue 2007, see also Viburnum) and is not obviously associated with the evolution of particular floral (or other) key innovations (see also Richardson et al. 2001). Similarly, rapid diversification of Andean species of Lupinus - where most species of the genus are now found - began only some 1.76-1.19 Ma and was probably driven by the ecological opportunities available in the high altitude habitats there (Hughes & Eastwood 2006; Drummond 2008; Drummond et al. 2012); bumble bees, also recent immigrants to South America, may have been an important factor in this diversification (Hines 2008). Finally, in Halenia (Gentianaceae) with its "key innovation" of five nectar spurs, diversification and acquisition of these spurs are not simply linked (von Hagen & Kadereit 2003, see also Gentianella, etc.). Of course, factor(s) that enable a clade to move into new ecological space within which much diversification subsequently occurs may be part of the definition of a key innovation (Sargent 2004; Marazzi & Sanderson 2010). Finally, deciding that a clade is notably speciose is not a simple task (e.g. Sanderson & Wojciechoswki 1996; Donoghue & Sanderson 2015).

In many very speciose clades, including angiosperms as a whole, patterns of clade numbers do not suggest immediate diversification after the acquisition of putative key innovations (see also above). Diversification rates of early diverging clades of angiosperms are low, clades with higher rates came later after what might be thought of as "typical" angiosperm flowers, i.e. those of Pentapetalae and of monocots, slowly assembled (see e.g. Sanderson & Donoghue 1994; Magallón & Sanderson 2001). Friis et al. (2006b) emphasized that such early clades have long fossil records yet include only a few extant species, and they also differ from other angiosperms in ecophysiological features. Thus ANA grade angiosperms have low veinlet densities rather like those of gymnosperms and ferns, hence transpiration rates and photosynthetic capacity are rather low (Brodribb et al. 2007; Boyce et al. 2009; Feild et al. 2009a; Brodribb & Feild 2010); their xylem is functionally not very different from that of vessel-less Pinales (e.g. Sperry et al. 2007; Feild & Thomas 2012; Plavcová & Jansen 2015). Similarly, distinctions between the nature and arrangement of floral parts that are obvious in say, Pentapetalae are less evident in members of the ANA grade, endosperm formation is variable, etc. (e.g. Buzgo et al. 2004; Endress 2005c; M. L. Taylor et al. 2008; Friedman 2008b). Flowers may become important in facilitating diversification only with the evolution of bees (Cappellari et al. 2013). Characters that seem to facilitate diversification but that evolve well before the diversification they are supposed to facilitate may be best thought of as exaptions (Gould & Vrba 1982; de Queiroz 2002, also 2015), and this may be appropriate for several of the characters considered to be key innovations of angiosperms or of major clades within it.

Parallelism and convergence, homoplasy, are everywhere one looks, even in early land plant evolution (e.g. Boyce 2010; Endress & Matthews 2012). Nearly all angiosperm characters are highly homoplastic, both arising in parallel and being lost many times and characterising both large and small clades (Reyes et al. 2018 and references: 148 origins of monosymmetry alone!). As is increasingly being found, there are parallelisms even at the amino acid level, an example being C4 photosynthesis (e.g. Bläsing et al. 2000; Christin et al. 2007b, 2008b, 2009a; N. J. Brown et al. 2011). Particular features are rarely consistently key innovations. The evolution of extra-floral nectaries may be deemed a key innovation in some Senna (Fabaceae) but the loss of such nectaries may equally be a key innovation in related taxa (Marazzi & Sanderson 2010). Similarly, Weber and Agrawal (2014) found that the acquisition of defensive metabolites increased diversification in 4/6 of the clades on which they focussed, although overall families with extra-floral nectaries showed a doubling of diversification rates. Some wind-pollinated clades are speciose, although most are not. Understanding developmental/regulatory pathways is important. The frequent reaquisition of woodiness in clades that have become herbaceous may be because elements of the cambial regulatory program remain untouched (Groover 2005; see also Blein et al. 2010: vegetative development). There are elements of common developmental mechanisms involved in independent acquisitions of monosymmetry (e.g. Feng et al. 2006: Fabaceae and Plantaginaceae; Zhang et al. 2010, Malpighiaceae), duplication of CYC genes being involved (see also Damerval & Manuel 2003; Rosin & Kramer 2009; Preston et al. 2011b). Irish (2009) suggests that petals may have evolved several times because of the independent cooption of underlying gene regulatory networks, and, as has been mentioned, the acquisition of CAM and C4 photosynthesis, and nitrogen fixation, and the development of ectomycorrhizal associations are similar.

Heterochrony (e.g. the male gametophyte of flowering plants - see Takhtajan 1976), heterotopy (e.g. Baum & Donoghue 2002), and homeosis (e.g. Mathews & Kramer 2012) are all part of this mix. As Preston et al. (2011b) put it as they summarized aspects of the developmental evolution of angiosperm flowers, "reduce, reuse, and recycle" has been the order of the day, and it seems that old dogs can indeed be taught new tricks (Rosin & Kraemer 2009; Mathews & Kramer 2012). We have tended to think of evolution as the modification of pre-existing form: "Are petals in x really modified stamens?". Now we have the tools to think more about the evolution of novelty, elements of developmental pathways that merge and form new combinations; Mathews and Kramer (2012) review floral and in particular ovule development across seed plants from this point of view, and this is true of vegetative features, too, as in the distinctive leaves of the Inverted Repeat Loss Clade of Fabaceae.

The often rather sporadic distributions of secondary metabolites has long been difficult to understand. But as with cambia, the ability to synthesise a particular secondary metabolite having been acquired, it may be switched off easily, but not lost, and so the metabolite can be "reacquired" (e.g. Grayer et al. 1999; Wink 2003, 2008, 2013; Liscombe et al. 2005; Albach et al. 2005c; Agrawal et al. 2012). However, in other situations pathways may degenerate, and change becomes irreversible (Zufall & Rauscher 2004). Associations between plant and fungus/microbe in both mycorrhizal and endophytic associations may also go some way towards understanding the rather unpredictable pattern of distribution of many secondary metabolites (Wink 2008; Lamit et al. 2009); endophytes may synthesize metabolites normally ascribed to the plant. Genes can be transferred via grafts in host-parasite connections, chloroplasts can move via grafts between organisms - and perhaps genes may be transferred from live pollen that lands on the stigma, germinates, but does little else (Christin et al. 2012: confirmation needed!).

The idea of evolutionary "tendencies" persists (e.g. Endress & Matthews 2012), similar discussions having recurred periodically in the phylogenetic literature (e.g. Cantino 1985; Sanderson 1991). Indeed, some phenotypes may be the result of parallel mutations that occur only because of a previous change in the larger clade (see Shubin et al. 2009 on deep homology), and Marazzi et al. (2012) attempt to locate such an evolutionary precursor - in this case, extrafloral nectaries in Fabaceae - on the tree. The ability of a plant to form an association with nitrogen-fixing bacteria is a good example (see Fabales: e.g. Soltis et al. 1995b), the clustered origins of C4 photosynthesis in grasses, sedges and core Caryophyllales invite a similar explanation (see e.g. McKown et al. 2004; Christin et al. 2011a, 2013; Grass Phylogeny Working Group II 2011), as do the origins of various symmetry types in angiosperm flowers (Irish 2009; Preston et al. 2009). Recent work on the independent acquisition of the ability of the plant to synthesizie caffeins i.a. shows how difficult it is to distinguish between convergence and parallelism (Huang et al. 2016).

The rise to dominance of the angiosperms and the diversification of particular angiosperm clades also involves other organisms - plants, animals, fungi, bacteria - as well as changes in the environment itself in which angiosperms are also involved, and it is a thoroughly ecological process (e.g. Thompson 1998; Harmon et al. 2009). Lavin et al. (2004) and Schrire et al. (2005) suggest that it is more profitable to think of diversification and distribution of Fabaceae in terms of vicariance of biomes rather than of the classic geographical areas. The area that a clade inhabits, especially if it is non-contiguous, may affect diversification (Vamosi & Vamosi 2010, 2011; see also Marazzi & Sanderson 2010 above). Under such circumstances diversity may be limited by ecological factors (e.g. Vamosi & Vamosi 2010), although if the assumption that clades increase steadily in diversity with time is unreasonable (Rabosky 2009; Vamosi & Vamosi 2010), so is any implicit assumption that the environment does not change.

The evolution of the first angiosperms took place under ecological conditions rather different from those under which they later prospered; then continents were drifting apart, there were high atmospheric carbon dioxide concentrations and rising sea levels, and ever-wet tropical humid climates were rather restricted. Even if "basal" clades that are now species poor were more diverse in their early history (Magallón & Castillo 2009; Friis et al. 2011 and references), conditions then were unlike those of today. Diverse angiosperm-dominated vegetation is largely of Caenozoic age, so the early Caenozoic environment - warm, not seasonal, few fires, the beginning of the diversification of some pollinators, herbivores and frugivores - provide the context for thinking about its evolution. Subsequent diversification occurred as temperatures and atmospheric CO2 concentrations were dropping, seasonality was increasing, and, especially within the last 10 m.y, fires were increasing. Climate has been been changing throughout the history of crown-group angiosperms, spurred in part by angiosperms themselves and their association with ECM fungi (e.g. Knoll & James 1987; Volk 1989; Boyce et al. 2010).

Thus establishing an immediate connection between acquisition of an apomorphy or group of apomorphies and diversification is difficult (e.g. Galis 2001). The evolution of flowers, vascular systems, and just about all aspects of plants seems a complex, protracted, and clade-dependent process involving both intrinsic and extrinsic factors, and often with a time lag between origin and effect. As Feild and Arens (2007: 21) noted, "Among basal angiosperms, the initial transitions to higher-light environments are characterized by a high degree of lineage-dependent, functional experimentation, in which fine-tuned performances were assembled piece-by-piece."

Diversification may well depend "on the fortuitous combinations of a large repertoire of traits" (Feild & Arens 2005: p. 402) rather than on any particular key innovation (see also Crepet & Niklas 2009; Magallón & Castillo 2009; Onstein 2019). Overall angiosperm success is in considerable part the result of diversification of individual angiosperm clades with particular combinations of characters that are responding to various ecological/environmental contexts, especially in a number of asterids and monocots (e.g. Magallón & Sanderson 2001; Sims & McConway 2003; Crepet & Niklas 2009). Thus two thirds (16,360 spp.) of Asteraceae are members of the chemically very distinct Asteroideae. Over four out of five Orchidaceae are Epidendroideae (21,600 spp.) in which the epiphytic habitat predominates, and much diversification occurred in the "higher epidendroids" some (64-)59-42(-36)/(49-)39-34(-22) Ma (Ramírez et al. 2007; Gustafsson et al. 2010). Exactly where monosymmetry is an apomorphy in Asterales as a whole is unclear - and so on. Phylogenetic niche conservatism - another way of saying ecological traits that do not reverse or reverse only little - or adaptations to "major ecological niches" mean that some groups will follow these niches when there is an opportunity (Donoghue 2008; especially Lavin et al. 2004; Schrire et al 2005; Marazzi & Sanderson 2010); adaptation to such niches may not occur very frequently. Extinction, although difficult to document, plays an important role. Thus as recently as ca 30 Ma there were stem-group hummingbirds in Europe (Mayr 2004, 2009) and Cyclanthaceae are known from European Eocene deposits (S. Y. Smith et al. 2008). Both groups are now iconically New World, and hummingbirds are involved in the pollination of (well) over 5,000 species of flowering plants there (see also Abrahamczyk & Kessler 2014).

I finish by returning to a couple of points that are made frequently in these pages. Species number is only one estimate of success in evolution, and there may even be a weak negative correlation between diversity and biomass produced (Wing & Boucher 1998). Biomass production, primary productivity, etc., provide ecological estimates of success, and are to be seen in the context of the eco-physiological evolution of angiosperms and the environmental changes that resulted. Venation density, vascular evolution and other ecophysiologically important features like the evolution of ectomycorrhizal associations and C4 photosynthesis have helped shape the evolution of biomes within which angiosperm diversification has occurred. Here and elsewhere we have to grapple with the evolutionary implications of clade size:ecological importance asymmetries. And mention of mycorrhizae emphasizes that plants are composite organisms. It is not only that several distinctive "plant" metabolites are synthesized by fungal or bacterial associates of the plant and that mitochondria and chloroplasts are bacteria whose associations with organisms are very ancient. But exactly what an individual "is" and what is the unit of selection start becoming problematical, indeed, rather than thinking of individual plants and their genomes it may be more useful to think of plants as holobionts with hologenomes, a microcosm, although maybe not so micro, the holobiome (e.g. Bordenstein & Theis 2015; Gundel et al. 2017; Hawkins & Kranabetter 2017; Tripp et al. 2017a). As Gehring et al. (2017a: p. 11170) note, "ECM [ectomycorrhizal] community composition represents a heritable plant trait" (see also McDowell et al. 2016; Eck et al. 2019). However, much earlier Rayner (1915: p. 128) had noted that many ericaceous plants, all obligately mycorrhizal, had "solved the problem of growth upon the poorest and most unpromising soils, but have solved it at a price of their independence", to which Pirozynski and Malloch (1975: p. 162) added "they and other land plants never had any independence, for if they had, they could never have colonized the land". This broader perspective may help us change how we think about evolution and diversification.

AMBORELLALES Melikian, A. V. Bobrov & Zaytzeva - Main Tree.

Just the one family, 1 genus, 1 species.

Note: In all node characterizations, boldface denotes a possible apomorphy, (....) denotes a feature the exact status of which in the clade is uncertain, [....] includes explanatory material; other text lists features found pretty much throughout the clade. Note that the precise node to which many characters, particularly the more cryptic ones, should be assigned is unclear. This is partly because homoplasy is very common, in addition, basic information for all too many characters is very incomplete, frequently coming from taxa well embedded in the clade of interest and so making the position of any putative apomorphy uncertain. Then there are the not-so-trivial issues of how character states are delimited and ancestral states are reconstructed (see above). Of course, putting apomorphies here in particular is a distinctly dubious proposition, given both the position of the clade and the black hole of ignorance immediately basal to angiosperms.

Includes Amborellaceae.

Synonymy: Amborellineae Shipunov

AMBORELLACEAE Pichon, nom. cons. Back to Amborellales


Shrub or small tree; alkaloids?; phloem loading via intermediary cells [specialized companion cells with numerous plasmodesmata; raffinose, etc., involved]; stem cork cambium?; axial parenchyma apotracheal-diffuse, (some pits in tracheary elements lacking membranes); pericycle with hippocrepiform sclereids alone; mucilage cells 0; nodes 1:1; petiole bundle arcuate; (stomata anomocytic), mesoperigenous; ?tooth morphology; plant dioecious; inflorescence cymose; flowers small [<7 mm across]; hypanthium +; P spiral, 5-10(-15), basally slightly connate, with a single trace; staminate flowers: A spiral, subsessile, 6-25, outer adnate to the base of P, anthers triangular, vascular bundle branched near thecae; middle layer of anther wall from both secondary parietal cells [ wall type]; pollen anaulcerate [pore-like, operculum endexinous, margin poorly defined], ektexine cupulate [distinctive undulate, columella-less exine]; pistillode 0; carpelate flowers: staminodes 1-2; G 3-7, spiral, stigma with uniseriate multicellular papillae; ovule 1/carpel, apotropous, ± median, pendulous, hemianatropous, subsessile, micropyle endostomal, outer integument annular; nucellar cap 0; embryo sac bipolar, 9-nucleate, with three synergids, antipodal cells die very early, polar nuclei in chalazal region; fruit a drupelet, P persistent, stone largely mesocarpial in origin, surface pocked; seed coat tanniniferous, endotesta lignified, collapsed, exotegmen lignified; endosperm triploid, first cell wall oblique, develops mostly from chalazal cell, embryo suspensor triangular, >2-seriate; n = 13; nuclear genome [1C] ca 870 Mb; horizontal transfer of atp1 gene; germination epigeal, phanerocotylar, growth of seedlings/young plants sympodial.

1[list]/1: Amborella trichopoda. New Caledonia. Photo: Leaves, Flower].

Evolution: Divergence & Distribution. Assuming that New Caledonia finally became emergent only some 37 Ma (Grandcolas et al. 2008; Keppel et al. 2009; Cluzel et al. 2012; de Queiroz 2014; Swenson et al. 2014, 2015; Grandcolas 2017; Nattier et al. 2017 for references), proto-Amborella must have been hanging out somewhere else for a very long time. However, Condamine et al. (2016) suggest that metapopulations of the clade may have persisted on ephemeral islands in the region (see also Heads 2018a, b), while Wallis and Jorge (2018; see also He et al. 2016a) suggest that New Caledonia may have remained emergent, and Giribet and Baker (2019), emphasizing the old ages of some clades endemic to the island, also incline in that direction.

Ecology & Physiology. Feild et al. (2003b, see also 2003a) looked at how Amborella coped with the shady conditions common where it grows. Flores Tornero et al. (2020 and references) look at the expression of genes in the egg apparatus and pollen tube.

Pollination Biology. For dioecy in Amborella, see Anger et al. (2017); perfect flowers that set fruit occur uncommonly. Both insects and wind are effective pollinators, so the plants are ambophilous (Thien et al. 2003; see Culley et al. 2002 for ambophily; Gottsberger 2016a). Stigmatic exudates may join all the stigmas of a single flower together, so pollen landing on a stigma can pollinate ovules in more than one carpel, i.e. there is an extragynoecial compitum (J. H. Williams 2009).

Genes & Genomes. For the genome of Amborella, see the Amborella Genome Project (2013). There are ancient transposons, but no evidence of recent transpson insertyions (see also Poncet et al. 2019).is no evidence of

The mitochondrial genome of Amborella is remarkable in that it is huge, containing genes from a number of land plants, including at least three different mosses, and such "foreign" genes may also migrate to the nucleus (Bergthorsson et al. 2004; c.f. Goremykin et al. 2009: methodological problems?). At ca 3.9 Mb, it is about seven times normal size and has acquired about three genomes worth of green algal DNA, two genomes worth of moss DNA, and one genome worth of angiosperm DNA (Rice et al. 2013). The algal DNA is similar to that of the trebouxiophyte Coccomyza, a component of lichens, perhaps suggesting that wounding of Amborella plants somehow facilitated the uptake of mitochondrial DNA from associated epiphytes. The angiosperm DNA came from Fagales, Oxalidales, Santalales, Ricinus (Malpighiales-Euphorbiaceae) and Bambusa (Poales-Poaceae), the two latter genera not currently known from New Caledonia (Rice et al. 2013; Poncet et al. 2019). Mitochondrial genomes like that of Amborella are as yet unknown from other angiosperms, although sampling is still poor, and one wonders what is distinctive about Amborella that might cause it alone to be such "a graveyard of foreign genes" (Rice et al. 2013: p. 70).

Chemistry, Morphology, etc.. For the absence of aluminium accumulation, see Thien et al. (2003). Amborella lacks reaction wood, its stems tending to sprawl, especially when young (but see Roussel & Clair 2015); in terms of architectural models (Hallé et al. 1978) the plant conforms to Troll's model. Some pits of the tracheary elements lack membranes, so technically they are vessels; open conduits are made up of only two such cells (Feild et al. 2000b), but Carlquist (2012a: p. 107; see also 2012c) thought that these were artefacts, noting that "intact porose pits in end walls of Amborella can be found". Stomatal morphology is quite variable, although the brachyparacytic configuration is common (Carlquist & Schneider 2001). The leaves are described as being spiral at first (Cronquist 1981; Takhtajan 1997), but c.f. Posluszny and Tomlinson (2003).

The perianth is spiral and undifferentiated. There is little agreement about pollen morphology: Sampson (2000) and Hesse (2001) suggest that the pollen is not really tectate (see also J. A. Doyle 2000, 2001, esp. 2009; Doyle and Endress 2000), and the aperture is difficult to categorise, as well as not always being present. J. H. Williams (2008, 2009) describes pollen tube development and fertilization. The ovule has been described as being orthotropous (straight), anatropous, or intermediate (Tobe et al. 2000); for the distinctive embryo sac, see above. The drupe of Amborella differs from a drupe in the strict sense in that the bulk of the woody layer is mesocarpial in origin, unlike the drupes of Laurales, etc., where the woody layer is often endocarpial (Bobrov et al. 2005; Romanov et al. 2018). The nature of the "resinous" cavities in the mesocarp is unclear; although not observed by Bobrov et al. (2005), they were conspicuous in material that I saw and are unlikely to be an artefact caused by re-expansion of dried fruits prior to study. The seed coat appears to have thin, unlignified walls, as might be expected in such a fruit, although some lignification has been reported (Tobe et al. 2000).

Additional information is taken from Bailey and Swamy (1948), Money et al. (1950) and Philipson (1993), all general, Metcalfe (1987: anatomy), Rudall and Knowles (2013) and Rudall and Bateman (2019a: stomata), Sampson (1993: pollen), Yamada et al. (2001a: ovules), and for endosperm, see above. The chemistry seems to be poorly known.

Previous Relationships. Amborellaceae were included in Laurales by Cronquist (1981) and Takhtajan (1997).