Aliso: A Journal of Systematic and Floristic Botany Aliso: A Journal of Systematic and Floristic Botany

We used ndhF sequence variation to reconstruct relationships across 282 taxa representing 78 mono cot families and all 12 orders. The resulting tree is highly resolved and places commelinids sister to Asparagales, with both sister to Liliales-Pandanales in the strict consensus; Pandanales are sister to Dioscoreales in the bootstrap majority-rule tree, just above Petrosaviales. Acorales are sister to all other monocots, with Alismatales sister to all but Acorales. Relationships among the four major clades of commelinids remain unresolved. Relationships within orders are consistent with those based on rbcL, alone or in combination with atpB and 18S nrDNA, and generally better supported: ndhF contributes more than twice as many informative characters as rbcL, and nearly as many as rbcL, atpB, and 18S nrDNA combined. Based on functional arguments, we hypothesized that net venation and fleshy fruits should both evolve-and thus undergo concerted convergence-in shaded habitats, and revert to parallel venation and dry, passively dispersed fruits in open, sunny habitats. Our data show that net venation arose at least 26 times and disappeared 9 times, whereas fleshy fruits arose 22 times and disappeared 11 times. Both traits arose together at least 15 times and disappeared together 5 times. They thus show a highly significant pattern of concerted convergence (P < I0-9) and are each even more strongly associated with shaded habitats (P < I0-10 to I0-23); net venation is also associated, as predicted, with broad-leaved aquatic plants. Exceptions to this pattern illustrate the importance of other selective constraints and phylogenetic inertia.


INTRODUCTION
Monocotyledons-with roughly 60,000 species, 92 families, and 12 orders-are the most diverse, morphologically varied, and ecologically successful of the early-divergent angiosperms. Over the past ten years, molecular systematics has revolutionized our understanding of higher-level relationships within the monocots and made them among the best understood in the angiosperms (Chase et al. 1993(Chase et al. , 1995aGivnish et al. 1999;Bremer 2000, Present addresses: 15  2002; Kress et al. 2001;Caddick et al. 2002a, b;Hahn 2002;Patterson and Givnish 2002;Pires and Sytsma 2002;Michelangeli et al. 2003;Zanis et al. 2003;Graham et al. 2006;McPherson et al. submitted). Such studies have laid the groundwork for rigorous studies of adaptive radiation, geographic diversification, and the evolution of development, independent of phenotypic convergence among distantly related groups or divergence among close relatives.
Based on a cladistic analysis of more than 500 rbcL sequences, Chase et al. (1995a, b) identified six major clades of monocotyledons: commelinids (including Poales, Commelinales, Zingiberales, Dasypogonales, and Arecales), Asparagales, Liliales, Pandanales, Dioscoreales, and Alismatales, with Acarus L. sister to all other monocots. However, even when Chase et al. (2000) complemented these data with sequences of atpB plastid DNA and 18S nrDNA for a subset VOLUME 22 Monocots: Concerted Convergence 29 of 140 species-more than tripling the number of nucleotides scored per taxon-relationships among many major clades remained unresolved or weakly supported, and evolutionary ties among several groups of commelinids and asparagoids remained unclear.
To contribute to efforts to resolve these higher-level relationships and investigate the possibility of widespread concerted convergence and plesiomorphy in ecologically significant traits across the monocots, we decided to produce a well-resolved, highly inclusive monocot phylogenetic analysis based on sequences of the plastid gene ndhR This gene provides abundant data for phylogenetic reconstruction: it is more than 50% larger than rbcL (ca. 2200 base pairs [bp] vs. ca. 142S bp) and has substantially more variable positions (Gaut et al. 1997;Patterson and Givnish 2002). Our ndhF tree is also ideally suited for analyzing patterns of repeated convergence and divergence among the monocots: it entails many more characters, better resolution, and higher levels of support for individual clades than phylogenetic trees based on rbcL alone (albeit for fewer taxa), while incorporating many more taxa than the existing three-gene tree  based on rbcL, atpB, and ISS nuclear ribosomal DNA (nrDNA), or the 7-and 17-gene trees now in preparation Graham et al. 2006).
Concerted convergence (Givnish and Sytsma 1997a, b;Givnish and Patterson 2000;Patterson and Givnish 2002) is the independent rise in different lineages under similar ecological conditions of two or more traits that are genetically, developmentally, and functionally unrelated; concerted plesiomorphy involves the retention of the same suite of traits in different lineages under similar conditions (Patterson and Givnish 2002). These phenomena might result from adaptations of unrelated traits to the same environmental conditions, or (possibly more likely) to different components of the shared set of conditions. They should be challenging to detect and study using analyses based on phenotypic data, given that multiple (and seemingly independent) characters would carry the same, misleading "signal" regarding evolutionary relationships. Patterson and Givnish (2002) demonstrated that concerted convergence and plesiomorphy occur among the monocots in the order Liliales. Phylogenetic reconstruction demonstrated that (1) visually showy flowers, capsular fruits, winddispersed seeds, narrow leaves, parallel venation, and bulbs arose upon invasion of open seasonal habitats, and (2) visually inconspicuous flowers, fleshy fruits, animal-dispersed seeds, broad thin leaves, net venation, and rhizomes persisted in lineages inhabiting ancestral forest understories. For each trait, the observed variation in phenotype with environment across lineages appeared to be functionally adaptive (Givnish and Patterson 2000;Patterson and Givnish 2002).
Two of these patterns of concerted convergence and plesiomorphy may hold throughout the monocots. Specifically, we predict that net venation and vertebrate-dispersed fleshy fruits should frequently evolve and be retained with each other under shady conditions in forest understories, and that parallel venation and nonfleshy fruits (dispersed by wind, water, or gravity) should frequently evolve and be retained with each other in open habitats. These predictions are based on the biomechanical economy of branched vs. unbranched support networks in thin leaves adapted to shady conditions, and on the efficiency of dispersal via vertebrates vs. more passive means in less windy forest understories.
Shady conditions favor thin, broad leaf laminas, which cannot support themselves mechanically (especially after small losses of turgor pressure), and therefore require longitudinal and lateral reinforcement from primary and secondary veins (Givnish 1979, 19S7). The cost per unit length of such veins scales like their diameter squared, whereas their strength scales like diameter cubed, favoring the coalescence of nearby, subparallel veins into one or few branching ribs of lower cost (Givnish 1979(Givnish , 1995. Thus, the broader and thinner a lamina or its divisions, the greater should be the advantage of net venation and the greater the advantage of a single midrib. Givnish et al. (submitted) argue that soft, thin, broad leaves are also favored in fast-growing, emergent aquatic plants with access to abundant moisture and nutrients (e.g., Sagittaria L.), and in filmy-leaved submersed species adapted for photosynthesis underwater (e.g., Aponogeton L. f.). Net venation should thus also be selectively favored in such plants. In addition, whereas wind dispersal of seeds is likely to be effective in open, windy habitats, animal dispersal of fleshy fruits should be more effective below closed habitats (Croat 197S;Givnish 199S). In Neotropical rain forests, up to 95% of the woody understory species (mostly dicots) bear fleshy fruits dispersed by birds, bats, or nonvolant mammals (Gentry19S2).
In this paper, we evaluate these hypotheses by deriving a well-resolved monocot phylogenetic tree based on ndhF sequence variation. We compare the resulting clades with those previously resolved based on rbcL, atpB and ISS nrDNA sequence variation. Finally, we use the ndhF tree to test whether fleshy fruits, net venation, and occurrence in shady forest understories show significant patterns of concerted convergence under shady conditions.

Phylogenetic Analyses
We included 2S2 monocot species in our analysis, representing as broad and representative a group of taxa as possible, including members of 7S of 92 families and all 12 orders (Table 1). Families and orders follow APG II (2003), except that we recognize Dasypogonales as equaling Dasypogonaceae (see Givnish et al. 1999;Doweld 2001;Reveal and Pires 2002), and Petrosaviales (Cameron et al. 2003). Most of the families unsampled are small, and several are nonphotosynthetic (e.g., Corsiaceae) or occur in wet or submerged habitats (e.g., Anarthriaceae, Posidoniaceae). The families not represented comprise only 1.2% of all monocot species. Only four (Burmanniaceae, Cyclanthaceae, Hydrocharitaceae, Potamogetonaceae) involve substantial numbers of taxa (100-225 species per family). We used Ceratophyllum L. as the outgroup, given its position sister to the monocots in several recent analyses (Soltis et al. 1997Graham and Olmstead 2000;Zanis et al. 2002;Borsch et al. 2003). Total DNAs were extracted from fresh, deep-frozen, or silica gel-dried leaf material. We amplified and sequenced ndhF for most taxa ourselves following standard techniques (see Patterson and Givnish 2002), obtaining both forward and reverse strands in most cases. New sequences were uploaded to GenBank and accession numbers obtained; se- Table I. Classification. GenBank accession numbers, vouchers, and authors for the 283 ndhF sequences included in this study. Nomenclature follows Bremer et a!. (2002) for orders and families (including "bracketed" taxa), and the International Plant Name Index (2004) for generic names, specific epithets, and taxonomic authorities. Specimens sampled include herbarium vouchers, accessions of living plants provided by various botanical gardens, and, in a few instances, initial citations. Material from curated living collections is designated by institution, followed by accession number. Abbreviations are as follows: ADBG (Adelaide Botanical Garden), ADU (University of Adelaide), NYBG (New York Botanical Garden), SEL (Marie Selby Botanical Garden), SIRG (Smithsonian Institution Research Greenhouses), and UCBG (University of California-Berkeley Botanical Garden).  ALISO quences from previous studies were downloaded from GenBank to complete the data matrix (Table 1). Sequences were visually aligned using MacClade vers. 4 (Maddison and Maddison 2002). Almost all of the 54 indels detected were in-frame and straightforward (albeit laborious) to align, given their general restriction to single species or small sets of close relatives. The aligned data matrix (including 2518 aligned bases) is available upon request from the three senior authors.
Phylogenetic analyses based on maximum parsimony (MP) were conducted using PAUP* vers. 4.0b8 (Swofford 2002). One hundred replicate searches were conducted using tree-bisection-reconnection (TBR) and random stepwise-addition to maximize the chances of detecting multiple islands of trees if they exist. Bootstrap percentages were obtained via TBR searches on 500 random resamplings of the nucleotide data, saving up to 50 trees per replicate. For comparative purposes, an additional MP search was conducted including both nucleotide and indel data; individual indels were treated as equally weighted characters and scored to minimize the number of additional evolutionary events following Baum et al. (1994). We merged our nucleotide data with those analyzed by Chase et al. (2000) to conduct an MP search involving 88 monocot genera for which sequence data are available for 18S nrDNA, rbcL, atpB, and ndhF, using Acarus as an outgroup. Based on this analysis, the numbers of informative and variable characters contributed by each of these sequences were calculated.

Calibration of Molecular Phylogenetic Trees Against Time
As previously shown for rbcL (Gaut et al. 1992(Gaut et al. , 1997, ndhF displays substantial variation in rates of nucleotide evolution across different groups of monocots, precluding the use of simple molecular clocks to place phylogenetic events and character-state changes on a time line. We therefore used the computer program r8s to transform one of the most-parsimonious ndhF trees into ultrametric form-with equal branch lengths from the root after discarding the outgroup Ceratophyllum-using cross-verified penalized likelihood (Sanderson 2002). We calibrated this tree against absolute time by fixing the age of the divergence of Acorales from other monocots at 134 million years ago (Mya) (Bremer 2000), while setting the minimum ages of the stem groups of six clades (Poaceae-Joinvilleaceae-Flagellariaceae-Restionaceae, Typhaceae-Sparganiaceae, Zingiberales, Arecales, Araceae, and Tofieldiaceae) equal to 69.5 Mya, 69.5 Mya, 83 Mya, 89.5 Mya, 69.5, and 83 Mya, respectively, based on the estimated ages of the oldest known Cretaceous fossils for these groups (Bremer 2000).

Tests of Concerted Convergence
We used selected ndhF trees to test whether fleshy fruits, net venation, and occurrence in shaded understories show correlated evolution employing DISCRETE (Pagel 1994(Pagel , 1999. DISCRETE uses a continuous Markov model to analyze the evolution of binary characters, incorporating branch lengths and weighting gains and losses equally. We executed separate tests of correlated evolution between (1) fleshy fruits and life in shady habitats; (2) net venation and shady habitats; (3) fleshy fruits and net venation; (4) net venation and shady habitats, emergent broad-leaved aquatics, or submersed broad-leaved aquatics; and (5) fleshy fruits and net venation, excluding emergent and submersed broadleaved aquatics. DISCRETE produces a likelihood ratio for which the distribution converges on that of x 2 with 4 degrees of freedom. We tested for correlated evolution-and hence, concerted convergence-by comparing the observed likelihood ratio against critical values of x 2 , a conservative approach (Pagel 1999). We conducted each test on four fully resolved trees, chosen randomly from among the maximumparsimony trees to represent each of the four resolutions of the major polytomy at the base of the commelinids. The other unresolved nodes are unlikely to have any substantial effect on inferences regarding the correlated evolution of fleshy fruits, net venation, or life in shaded understories. We ran each test using branch lengths (inferred number of substitutions) as measures of the amount of molecular evolution down each branch, reflective of time discounted by the rate of molecular evolution inherent to different lineages. Rates of phenotypic transitions were fitted to a gamma distribution, based on median rates in quartiles across monocots. We conducted each test five times independently because DIS-CRETE can fit slightly different likelihood models to the data from each random starting point, as a result of nearly fiat response surfaces and/or large numbers of species. In a few instances, the first step of a DISCRETE run-which evaluates a model assuming no correlated evolution between the given pair of traits-returned a log-likelihood substantially below that of other runs. We discarded such cases a priori because they represent a much worse fit of the independent model than is possible and would bias the dependent test toward higher significance of correlated evolution. This procedure would, if anything, create a bias against acceptance of a significant pattern of correlated evolution.
We considered "net venation" to include branching support structures within leaves, including cases of reticulate venation, simple leaves in which the veins diverge from a massive central rib regardless of whether they branch anatomically (e.g., Musa), and compound leaves with a branching rachis (palms). "Fleshy fruits" include berries, drupes, and seeds with showy, massive, nutritional arils dispersed by vertebrates. Proiphys Herb. (Amaryllidaceae) has brightly colored capsules that seem to mimic fleshy fruits (Meerow and Snijman 1998) and were scored as such. Seeds dispersed by ants, bearing small arils (elaiosomes), occur in forest and nonforest habitats and can serve as adaptations for purposes not directly related to dispersal (e.g., placement in nutrientrich ant nests, shelter from frequent fire) (Beattie and Culver 1983; Beattie 1985; Hughes and Westoby 1992;Boyd 2001). The fruits of Acarus are anatomically berries, but are minute, have a relatively thick, dry coat, and lack the sweet or oily composition usually associated with adaptation for ingestion and dispersal by vertebrates. Other features of its morphology and (especially) its geographic pattern of genetic variation suggest that Acarus is water-dispersed (Liao and Hsiao 1998). Thus, we did not score either ant-dispersed seeds or the dry berries of Acarus as fleshy fruits. Species were classified as occurring primarily in open, sunny habitats (e.g., tundra, chaparral, desert) or closed, shady habitats (forest understories). For species growing in seasonally deciduous forests, the timing of leaf activity and fruit production rela-tive to canopy closure was used to categorize the habitats occupied as sunny or shady (see Patterson and Givnish 2002). Assigning species to these two classes was occasionally problematic: light regimes occupied by different species can vary continuously (e.g., see Leach and Givnish 1999;Givnish et al. 2004b) and most published accounts of ecological distributions are qualitative. However, no matter how one slices the light availability gradient, taxa like Schizachyrium Nees and Strelitzia Aiton occur in brightly lit sites, while Trillium L. and Cyanastrum Cass. occur in shaded understories.
For illustrative purposes, we overlaid net venation, fleshy fruits, and life in shady habitats on an ultrametric tree using MacClade. Accelerated transformation was employed to minimize the number of apparently independent origins of each trait. A complete matrix of venation, fruit, and ecological character states is available upon request from the first author.

Phylogenetic Relationships
Maximum parsimony produced one island of 880 trees, each 16,489 steps in length based on 1727 variable characters, of which 1408 are potentially phylogenetically informative ( Fig. lA-D). Across monocots, ndhF strongly supports (85-100% bootstrap) the monophyly of nine of the twelve orders identified by previous molecular studies (Chase et al. 1995a; support is only modest for Commelinales (52%) and Asparagales (65%), and Petrosaviales are represented by only a single taxon (Japonolirion Nakai). Several nodes previously unresolved or weakly supported are resolved in the ndhF phylogenetic tree. Our analysis demonstrates that (1) Asparagales are sister to the commelinids; (2) both of these groups are sister to Liliales plus Pandanales; (3) Japonolirion (Petrosaviales) and/or Dioscoreales are sister to all preceding groups; ( 4) Alismatales are strongly supported (99% bootstrap) as sister to the preceding orders; and (5) Acarus (Acorales) is sister to all other monocots.
Bootstrap support for individual clades is often substantially higher than that based on rbcL, considered alone or in combination with atpB and 18S nrDNA (see Chase et al. l995a, b, 2000). Even so, support values are still only modest at several points along the backbones of the asparagoid portion of the tree and the monocot tree as a whole (Fig.  lA-D). In a four-gene analysis, ndhF contributes 2.2 times as many informative characters as rbcL, and 87% as many as rbcL, atpB, and 18S nrDNA combined. Only nine nodes are unresolved in the ndhF strict consensus tree. Of these, only two-involving a four-way polytomy at the base of the commelinids, and a trichotomy involving four families of Zingiberales-involve substantial numbers of taxa. The commelinid polytomy involves unresolved relationships among Poales (P), Zingiberales plus Commelinales (ZC), Dasypogonales (D), and Arecales (A). Each of these clades is strongly supported individually (79-100% ), as are the commelinids as a whole (85%). Among the most-parsimonious trees based on ndhF sequence variation, we found four different patterns of relationship among the major comme-linid clades: ((P,A),(ZC,D)); (P,(ZC,A,D)); ((P,(ZC,D)),A); and ((P,D),A),ZC).
Within Poales, ndhF places Bromeliaceae sister to Typhaceae-Sparganiaceae at the base of the order, with this overall group sister to an unresolved trichotomy involving (1) Rapateaceae, (2) the sedge alliance-Cyperaceae-Juncaceae-Thurniaceae, Eriocaulaceae-Xyridaceae, and Mayacaceae, and (3) the grass alliance-Poaceae, Joinvilleaceae, Ecdeiocoleaceae, Flagellariaceae, and Restionaceae (Fig. lA-B). Rapateaceae are sister to the grass and sedge alliances in the bootstrap consensus (54% support), and are sister to these groups in the strict consensus tree as well if Ecdeiocolea F. Muell. is excluded or if nucleotide characters are sequentially reweighted based on their consistency index. Poaceae and Poaceae-Joinvilleaceae-Ecdeiocoleaceae have 100% bootstrap support, with Elegia L. (Restionaceae) and/or Flagellariaceae sister to these other elements of the grass alliance. Thurniaceae are sister to Cyperaceae-Juncaceae (93% bootstrap) at the core of the sedge alliance. Eriocaulaceae and Xyridaceae are monophyletic (100% and 84% bootstrap support, respectively) and each other's closest relatives at the base of the sedge alliance minus Mayacaceae (Fig. lA). The four major subclades of Poales-the grass alliance, sedge alliance, Rapateaceae, and Bromeliaceae-show as much sequence divergence from each other as that seen among the remaining orders of monocots. Members of the grass and sedge alliances show the highest rates of ndhF evolution among monocots; bromeliads display unusually low rates, and rapateads are intermediate in this respect (Fig. lA-D). Resolution of relationships within the latter two families by highly informative ndhF indicates that both require new internal classifications, including five new subfamilies and recircumscription of an additional two (Givnish et al. 2004a, in press).
Commelinales and Zingiberales are both resolved as sister clades (79% bootstrap support). In Zingiberales, ndhF resolves three pairs of sister families-Zingiberaceae-Costaceae (61% ), Marantaceae-Cannaceae (80% ), and Strelitziaceae-Lowiaceae (89% ). Within Commelinales, Pontederiaceae and Haemodoraceae both have 100% bootstrap support as sister taxa. Philydraceae are sister to the rest of Commelinales, but this position is weakly supported (Fig.  lB). One indel supports their placement with Pontederiaceae and Haemodoraceae, and a combined analysis of ndhF indels and nucleotides (not shown) places Philydrum Banks ex Gaertn. in an unresolved trichotomy with Pontederiaceae-Haemodoraceae and Commelinaceae-Hanguanaceae. Hanguana is strongly supported (90%) as being sister to Commelinaceae; Cartonema R. Br. is sister to all other members of the latter. Our analysis places the climbing rattan Calamus Auct. ex L. sister to the rest of Arecales (98% bootstrap), with the mangrove palm Nypa Steck next-divergent. Dasypogon R. Br. and Calectasia R. Br. are resolved as forming the monophyletic order Dasypogonales (100% bootstrap).
An important finding of this study is that ndhF places Asparagales sister to the commelinids rather than Liliales in the strict consensus tree (Fig. lC). Asparagales are composed of a ladder of eight clades, with Orchidaceae sister to the rest (Fig. lC). The sequence of families is broadly similar to that seen in recent studies (see Discussion). Hyacinthaceae are sister to Agavaceae rather than Themidaceae, with which they share bulbs and a similar habit. Agapanthaceae are sister to Amaryllidaceae-Ailiaceae. Orchidaceae are strongly supported as monophyletic ( I 00% bootstrap) , and are placed siste r to all other Asparagales with moderate support (Fig. 1 C).
Pandanales are s ister to Liliales in th e ndhF stri ct consen-sus tree (Fig. lD), and to Dioscoreales in the bootstrap majority-ru le tree. Among the famiJjes sam pled, Velloziaceae are sister to Pandanaceae-Stemonaceae in Pandanales, and Nartheciaceae are sister to Di oscoreaceae of Dioscoreales. Japonolirion of Petrosaviales is part of an unreso lved trichotomy in volving itself, Dioscoreales, and corruneHnids-Aspa ragales-L ili a les-Pa nd ana les; together, these groups fo rm a strong ly supported clade (100% bootstrap) consisting of all monocots except Ali smatales and Acorales (Fig. lD). Araceae (100% bootstrap) are siste r to Tofieldiaceae and the remaining Alismatales. The latter fo rm a c lade w ith 100% bootstrap suppo rt and two well-marked su bclades, inc luding Alis mataceae-Limnocharitaceae-Butomaceae (95 % ), and Aponogetonaceae-J un caginaceae-Sche uchzeriaceae-Cy modoceaceae-Zosteraceae. Juncag inaceae and Scheuchzeriaceae are reso lved as sister groups based on ndh.F seque nce vmiation. Tofieldiaceae are weakly suppo rted (69 %) as sister to the fami lies of the former Najadales (Dahlgren et al. 1985). Finally, ndhF provides 100% bootstrap support for the positio n of Acarus sister to all other monocots (Fig. lD).

Concerted Conve rgence
Based on o ur ndhF data, fl eshy fruits appear to have arisen at least 2 1 tim es and been lost 11 times, whereas net venation has arisen at least 26 times and been lost 9 times (Table 2; F ig. 2). As predicted, these traits have undergone concerted convergence. They have done so in highl y signi ficant fas hi on ( P < I0-9 , log-likelihood test), wi th both trai ts arising together (at the same or adj acent nodes) 15 times and di sappearing together 5 times (Table 2; Fig. 2). Fleshy fruits  Table 2. Inferred evolutionary ongms of net venation. fleshy fruits, and life in shady habitats, and of parallel venation, passively dispersed fruits and/or seeds, and life in sunny habitats. Most instances of the evolution of the former character states represent initial transitions from the latter, while most instances of the origin of the latter represent reversals from the former. Transitions on the same line occurred at the same node or (in a few cases) adjacent nodes. Instances where all three character states underwent transition at the same or adjacent nodes-involving concerted convergenceare underlined. All calls are based on overlaying characters on a single most-parsimonious tree using accelerated transformation in MacClade (Maddison and Maddison 2002 Nolina Michx., with the loss of fleshy fruits in the last lagging that of net venation by one node. Both fleshy fruits and (especially) net venation show even stronger patterns of correlated evolution with shady conditions than with each other. In almost every case, the evolution of net venation and fleshy fruits is associated with life in forest understories, whereas their loss is associated with open habitats. Specifically, 19 of 21 gains of fleshy fruits are associated with invasion of-or life in-shady sites, whereas 7 of 11 losses are associated with the invasion of sunny conditions. For net venation, 22 of 26 gains are associated with shady conditions, whereas 8 of 9 losses are associated with sunny conditions. These patterns of origin and maintenance are highly significant (P < 10-10 to lQ-30 ) when tested in DISCRETE, using branch lengths that are equal to the inferred total amounts of molecular evolution down each lineage, a function of time plus plant characteristics such as generation time; Table 3). These results support our hypotheses about adaptation and establish the existence of a highly significant pattern of concerted convergence across the monocots.
Net venation shows an even more marked association with shade if we factor out the four lineages (Alismataceae, Aponogetonaceae, Philydraceae, Zosteraceae) in which it arose in broad-leaved aquatic plants, mostly near the base of the monocots in Alismatales (Tables 2, 3). All origins of net venation are associated with either shady conditions (85%) or broad leaves in aquatic emergents or submersed species (15% ). Fleshy fruits also show a stronger association with net venation if we exclude aquatic plants with broad leaves and net venation, in which we have no a priori reason to expect the evolution of fleshy fruits.
The numerous origins of fleshy fruits and net venation are distributed rather evenly across lineages and time (Fig. 2). Both traits arose nearly 90 Mya ago in Araceae and Arecaceae. The former family is dominated by herbs, vines, and epiphytes of tropical rain-forest understories, together with some temperate forest herbs (e.g., Arisaema Mart., Arum L.) , and broad-leaved aquati c habit (blue boxes). Note that almost all transiti ons to net venati on and fl es hy fruits occur upo n in vasio n of shaded habitats, and th at almost a ll reversals to paralle l venati on and dry, pass ive ly di spersed seeds or fruits occur upon re-invas io n o f ope n, sunny habitats. The tree shown is ultrametri c and has been calibrated again st the age of six C retaceo us fossil s using pe nalized like lihood, so that the te mpo and taxono mi c distributi o n of phenoty pi c tran itions can be visua li zed. Both net venatio n and fl eshy fruit show somewhat constant rates of ecological evo luti on over the past 90 milli on years, with an inc rease in the absolute number of origins toward the present and a decrea e in the numbe r o f ori gins per c lade present. Table 3. Log likelihood ratios (LR) and significance levels (P) resulting from five different tests for correlated evolution across monocots in net venation, fleshy fruits, and life under shaded conditions, conducted on four representative trees using DISCRETE (Pagel 1994(Pagel , 1999 and broad-leaved submersed aquatics (e.g., Anubias Schott, Cryptocoryne Fisch. ex Wydler). The palms contain many rain-forest lineages, but have also invaded open subtropical savannas and scrub. The most recent instances of concerted convergence in fleshy fruits (or mimics thereof) and net venation occurred within the last 5 to 10 Mya, in Gri.ffinia Ker Gawl. and Proiphys-Scadoxus Raf. of Amaryllidaceae and Curculigo of Hypoxidaceae. Fleshy fruits arose at least three times in Poales, twice in Commelinales, once in Zingiberales, once in Arecales, eight times in Asparagales, three times in Liliales, once in Pandanales, once in Dioscoreales, and once in Alismatales. Net venation arose at least three times in Poales, once in Commelinales, once in Zingiberales, once in Arecales, eleven times in Asparagales, four times in Liliales, once in Pandanales, once in Dioscoreales, and four times in Alismatales. During 10-Mya intervals, an average of 2.9 :!: 0.5 lineages evolved net venation, whereas an average of 2.4 :!: 0.4 lineages evolved fleshy fruits, implying a rather clocklike rate of adaptive evolution in both these traits across the monocots. It is important to note that many of the inferred reversals to parallel venation or passively dispersed, dry fruits appear to have occurred quite recently, with the exception of the reversal at the base of subfamily Lilioideae of Liliaceae (Fig. 2). As might be expected given the relative numbers of origins of net venation and fleshy fruits, there are a number of groups of understory plants in which only net venation, not fleshy fruits, evolved. The net-venation-only syndrome characterizes the bambusoids, early-divergent grasses, Costaceae, Hosta Tratt., Cyanastrum Oliv., and Stemonaceae. Cardiocrinum and Tricyrtis of temperate deciduous forests both have net veins only, but are photosynthetically active under shady conditions in summer while releasing seeds af- ter the canopy re-opens in autumn. Net veins also occur in the absence of fleshy fruits in four lineages of broad-leaved aquatics, including Alismataceae, Aponogetonaceae, and Zosteraceae of Alismatales and Philydraceae of Commelinales. Fleshy fruits arose without net venation under shady conditions in bromelioid bromeliads, Amischotolype Hassk., Asteliaceae and relatives, and the apostasioid orchid Neuwiedia Blume.

Phylogenetic Relationships
Cladistic analysis of ndhF sequence vanatwn yields a highly resolved, well-supported phylogenetic tree for the monocots (Fig. 1). Relationships among orders are unclear in only two cases, involving the commelinids and the position of Dioscoreales and Pandanales close to the base of the monocots. The lack of resolution among the four major commelinid clades-Poales, Zingiberales plus Commelinales, Dasypogonales, and Arecales-may simply reflect a rapid initial diversification among the commelinids. Analyses based on seven genes (but many fewer taxa) resolve this polytomy by placing Dasypogonales sister to Poales, and Arecales sister to Zingiberales-Commelinales, but the bootstrap support for both relationships is weak (::;51%) ). Analyses based on 17 genes flip these relationships, placing Arecales as sister to Poales and Dasypogonales sister to Zingiberales-Commelinales, and bootstrap support for these relationships is also weak (<50%) (Graham et a!. 2006). Here Pandanales are sister to Liliales in the ndhF strict consensus, but to Dioscoreales in the bootstrap consensus. The latter position is consistent with that obtained from an analysis based on 7 and 17 genes (Chase et a!. 2006; Graham et a!. 2006). Alismatales are sister to a strongly supported clade (100% bootstrap) consisting of all other monocots except Acorales (Fig. 1). Petrosaviales or Dioscoreales are, in tum, sister to all other elements of this large clade; 7-and 17-gene analyses place Petrosaviales sister to all monocots except Alismatales and Acorales, and Dioscoreales sister to Pandanales (Chase eta!. 2006;Graham eta!. 2006).
Within commelinids, ndhF supports many relationships identified previously based on other sequence data (Givnish et a!. 1999;Graham et a!. 2003;Michelangeli et a!. 2003), and resolves others for the first time. Brome1iaceae and Typhaceae-Sparganiaceae are sister to each other and earliest divergent within Poales, with Rapateaceae being next divergent in the bootstrap consensus and sequentially weighted analyses (see Fig. 1 and Results). Our findings for Poales differ somewhat from those of Michelangeli et al. (2003) based on morphology and sequence variation in rbcL and atpA. Those authors placed Rapateaceae sister to paraphyletic family Xyridaceae, including Eriocaulaceae and Mayacaceae, at the base of Poales; identified Bromeliaceae, then Typhaceae-Sparganiaceae as sister lineages to the remaining members of the order; and positioned Flagellaria L. as sister to two terminal clades, consisting of ( 1) Anarthria R. Br., Aphelia R. Br., and Restionaceae, and (2) Joinvillea Gaudich., Ecdeiocolea, and Poaceae. The nodes at which our results and those of Michelangeli et al. (2003) differ, however, are weakly supported (<50% bootstrap) in their analysis. These include (1) the positions of Bromeliaceae, Rapateaceae, and Typhaceae-Sparganiaceae relative to each other and to Eriocaulaceae, Mayacaceae, and Xyridaceae; (2) the supposed paraphyly of Xyridaceae; and (3) the position of Flagellaria, not Restionaceae, as sister to the remainder of the grass alliance. Our ndhF analysis resolves both Xyridaceae (including Orectanthe) and Eriocaulaceae as being monophyletic with 84-100% bootstrap, and identifies these two families as each other's closest relative (75% bootstrap). We resolve Bromeliaceae as sister to Typhaceae-Sparganiaceae with 64% bootstrap support, and place Rapateaceae as the next-divergent element with 79% support in the bootstrap majority-rule tree, consistent with its strongly supported position in the 7-gene tree . These relationships are similar to those derived by Bremer (2000) based on rbcL, but differ in the placement of the three earliest-divergent clades consisting of Bromeliaceae, Rapateaceae, and Typhaceae-Sparganiaceae. The placement of the last just inside Bromeliaceae-Rapateaceae by Bremer (2000) involves a very short branch, however. We were unable to amplify and sequence ndhF for DNAs of Aphelia (Centrolepidaceae) and Trithuria Hook. f. (Hydatellaceae) kindly provided by J. Davis and D. Stevenson, and so were unable to confirm their strongly supported finding that Aphelia is sister to Restionaceae or the more weakly supported association of Trithuria with Xyridaceae.
The strongly supported placement of Thurnia Hook. f.-Prionium E. Mey. sister to Cyperaceae-Juncaceae by ndhF is consistent with that of several recent molecular studies (Givnish et al. 1999;Bremer 2000;Chase et al. 2000;Michelangeli et al. 2003). The position of Mayaca Aubl. sister to all other elements of the sedge alliance, however, is more weakly supported and problematic. An earlier ndhF se-quence of this taxon had placed it sister to Bromeliaceae (Givnish et a!. 1999), but a new, higher quality sequence from Venezuelan material places it sister to the remainder of the sedge alliance, near Xyridaceae and Eriocaulaceae, which seems more plausible based on morphology and atpA and rbcL sequence data (Michelangeli et al. 2003); rbcL places Mayaca immediately sister to Xyridaceae and Eriocaulaceae (Bremer 2000). The possibility that Mayaca could act as a "wild card" much like Ecdeiocolea (see above) or Aphyllanthes Tourn. ex L. in Asparagales (see Fay et al. 2000), should not be overlooked.
The extensive divergence of the grass alliance, sedge alliance, rapateads, and bromeliads from each other is comparable to that among other groups of monocots already recognized at the ordinal level (Fig. 1). The remarkable isolation of both Bromeliaceae and Rapateaceae from other monocots in both morphology and sequence variation appears to reflect 15 to 40 million years between the origins of each group and when present-day lineages began to diverge from each other (Givnish et al. 2004a, in press). If support for the four major clades of Poales grows in future multigene analyses, and the position of Eriocaulaceae, Xyridaceae, and (especially) Mayacaceae becomes solidified, it would be prudent to revisit the issue of recognizing the four major clades in Poales-representing 31% of all monocot species-as orders in their own right.
The resolution of Commelinales and Zingiberales as sister taxa is consistent with previous molecular analyses (Givnish et a!. 1999;Chase et al. 2000;Evans et al. 2003;Graham et al. 2003). Relationships among families within Zingiberales are largely consistent with a detailed analysis based on morphology and several rapidly evolving stretches of DNA (Kress et al. 2001). Our analysis, however, places Musaceae, Heliconiaceae, and Strelitziaceae-Lowiaceae in an unresolved trichotomy sister to the remaining "ginger" families, rather than in a ladder with Musaceae earliest-divergent as seen in Kress et al. (200 1 ). Our ndhF tree identifies Haemodoraceae and Pontederiaceae as each other's closest relatives (Fig. lB). They fail, however, to provide positive evidence that their immediate sister is Philydraceae, as have other molecular studies (Graham and Barrett 1995;Graham et al. 1998;Chase et al. 1995aChase et al. , 2000Givnish et al. 1999). However, inclusion of indels places Philydrum in a polytomy consistent with a tie to Haemodoraceae and Pontederiaceae (see Results), and thus consistent with previous studies based on morphology (Dahlgren et a!. 1985) and molecular variation. Fleshy-fruited, net-veined Hanguana is strongly supported as being sister to Commelinaceae, consistent with previous molecular analyses (Givnish et al. 1999; but not with morphology, which tends to place this genus of southeast Asian rain forests with Zingiberales instead (Rudall et a!. 1999).
The placement of Calamus as sister to the rest of Arecales, followed by Nypa, is consistent with relationships obtained using 5-7 kilo bases (kb) of coding and noncoding plastid DNA (Asmussen and Chase 2001;Hahn 2002). Although bootstrap support for relationships within the rest of Arecales based on ndhF alone are low (35-95% ), the fact that they are fully resolved based on a single gene is promising, given that many similar relationships are unresolved using rbcL alone (Uhl et al. 1995). Members of Arecales, Bro-meliaceae, and Zingiberales display unusually slow rates of plastid DNA evolution (Gaut eta!. 1992;Givnish eta!. 1999;Fig. lA-D). It should thus not be surprising that relationships within these groups are much better resolved by ndhF than by rbcL, with or without atpB and 18S nrDNA (see Chase et a!. 1995a. Although molecular data (ndhF; rbcL, atpB, 18S nrDNA; and 7-and 17-gene trees in development) do not resolve the relationships among the four major clades of commelinids, certain anatomical and chemical characteristics tend to link Commelinales-Zingiberales to Poales (Stevens 2003).
The placement of Asparagales sister to the commelinids rather than Liliales by ndhF (Fig. 1 C) runs counter to the previous view that Asparagales and Liliales are sister to each other (Dahlgren et a!. 1985). The arrangement of families within Asparagales generally supports that obtained in other recent studies McPherson et a!. submitted). Relationships within and near Amaryllidaceae are largely consistent with those obtained by Meerow et a!. ( 1999) based on rbcL and the tmL-trnF region, except that ndhF places Agapanthaceae sister to Amaryllidaceae-Alliaceae, rather than Amaryllidaceae alone.
Relationships within Liliales are mostly consistent with those obtained by Vinnersten and Bremer (2001) based on rbcL, and by Patterson and Givnish (2002) based on rbcL and ndhR Our results differ slightly from the rbcL tree, in which Alstroemeriaceae and Colchicaceae are sister to the rest of the order, and in which Liliaceae sensu Tamura ( 1998b ), Calochortus, and Scoliopus-Streptopus-Tricyrtis form a trichotomy at the base of Liliaceae sensu Tamura (1998a). Analysis of the combined rbcL and ndhF data produces a tree identical to that based on ndhF alone (Patterson and Givnish 2002). The placement of Prosartes in Liliaceae and Disporum in Colchicaceae supports previous analyses based on rbcL (Shinwari et a!. 1994a, b;Chase et a!. 1995a, b), despite the striking morphological convergence in habit, net venation, and fleshy fruits in these two groups, formerly lumped in Disporum.
The position of Pandanales sister to Dioscoreales in the ndhF bootstrap majority-rule tree is consistent with analyses of placeholders involving 7 and 17 genes, as is the position of Japonolirion (or Dioscoreales) sister to monocots other than Alismatales and Acorales in the strict consensus tree Graham et a!. 2006). Attempts to sequence ndhF for representatives of Burmanniaceae (Burmannia L., Thismia Griff.) failed despite repeated attempts, preventing us from determining where this family belongs. Tacca-Dioscorea L. is sister to Trichopus Gaertn. within the dioscorealean taxa sequences surveyed, consistent with the findings of Caddick et a!. (2002a, b) based on rbcL, atpB, and 18S nrDNA. The isolated position of Japonolirion supports the decision to recognize this genus (and achlorophyllous Petrosavia Becc.) as constituting Petrosaviales, one of the 12 monocot orders (Cameron eta!. 2003).
Relationships among the families of Alismatales based on ndhF are broadly similar to those implied by rbcL (Les et a!. 1997) but differ in detail. Mostly, the divergences between the two trees are not strongly supported in either case. The identification of Juncaginaceae and Scheuchzeriaceae as sister groups by ndhF, however, is probably significant, given that these morphologically similar families share a unique cyanogenic glucoside (triglochinin) known in no other angiosperm family (see Haynes eta!. 1998). Repeated attempts to amplify and sequence ndhF from several of the smallest, aquatic families of Alismatales failed, preventing as detailed an analysis of relationships in this group as desired (D. Les and S. W. Graham pers. comm.). Our results support the important conclusion of Les eta!. (1997) that the "aquatic" families of Alismatales fall into two clades, one including Alismataceae, Limnocharitaceae, and Butomaceae, and the other a series of three independently evolved families of seagrasses, with the Madagascar lace-plant family (Aponogetonaceae) closely related to the latter clade.

Concerted Convergence
The independent origin of net venation at least 26 times in the monocots, always in association with invasion of shady conditions (85%) or life as a broad-leaved aquatic plants-as well as the independent origin of fleshy fruits at least 21 times, 19 in association with shaded forest understories-is one of the most remarkable, widespread, and highly significant (P < 10-30 to I0-30 ) cases of convergent evolution ever documented. The joint evolution of fleshy fruits and net venation 15 times across the monocots, and their joint loss five times, is also-by far-the most striking case of concerted convergence and plesiomorphy thus far demonstrated. These patterns are not only highly significant, they have high explanatory value as well. Phylogenetically unstructured correlation coefficients (r) range from 0.54 for the coincidence of net venation and fleshy fruits, to 0.64 and 0.73 for the coincidence of fleshy fruits and net venation with shaded habitats, to 0.77 for the coincidence of net venation with shaded habitats or a broad-leaved aquatic habit, when all traits are scored as binary characters.
In many ways, the contrast between Trillium and its closest relatives in Melanthiaceae (represented in this and all other surveys by Xerophyllum Michx.) epitomizes the pattern of concerted convergence discussed in this paper. Trillium grows in the shaded understories of temperate mesic forests, has broad, thin, soft leaves, net venation, and fleshy fruits, while Xerophyllum grows in more open habitats (meadows, fireswept pine glades) and possesses narrow, thick, hard leaves with parallel venation and tiny, wind-dispersed seeds released from dry capsules. It would be difficult, based on gross morphology, to infer that these taxa are actually very close relatives; the demonstration that they are is one of the triumphs of plant molecular systematics. The contrast between Trillium and Xerophyllum is paralleled by several other cases, most notably involving the contrast between Hypoxis (mostly grass-leaved, capsule-fruited herbs of meadows, prairies, and glades, occasionally found in woodlands) and Curculigo (broad-leaved, net veined, fleshy-fruited herbs of tropical forest understories). A few Hypoxis occur in tropical forest understories or have broad leaves, and a few Curculigo have rather narrow leaves. Fleshy-fruited, net-veined, forest-dwelling Geitonoplesium also contrasts sharply with all of its dry-fruited, parallel-veined relatives of open habitats (see Conran 1999). Cyanastrum of shady African rain forests and woodlands has broad, cordate leaves with net venation, while confamilial Cyanella L. of open South African fynbos and Tecophilaea Bert. ex Colla of the Chilean high Andes have narrow, rather fleshy, grasslike foliage that lack cross veins. The difference between forestdwelling Hosta (with thin, broad, net-veined leaves) and Agave L., Yucca L., and other elements of Agavaceae (mostly with thick, succulent, parallel-veined leaves) to which Hosta is sister (Bogler and Simpson 1996) could hardly be more striking, although it does not entail the evolution of different fruit types.
In addition to such cases of divergence among close relatives, striking convergence among distant relatives also supports our case. Asian Disporum of Colchicaceae and North American Prosartes of Liliaceae both grow in the understories of temperate mesic forests and share net venation and fleshy fruits, as well as many other features of growth form (e.g., arching stems) and floral morphology. They are so similar that both were placed in the same genus, until Shinwari et al. (1994a, b) used molecular data to demonstrate that the North American taxa were closely related to Streptopus, while the Asian taxa were closely related to Uvularia. Both of these genera, in turn, are remarkably similar in many ways to Polygonatum Miller, Disporopsis Hance, and Smilacina Desf. (also native to temperate forest understories) of Ruscaceae in order Asparagales; indeed, these genera were grouped with Disporum and Prosartes in the asparagoid tribe Polygonatae by Krause (1930), Therman (1956, Hutchinson (1959), and La Frankie (1986)!
The joint evolution of fleshy fruits and net venation is not lock-step: by no means is every invasion of forest understories associated with a gain of both traits, nor is every invasion of open sites associated with a loss of both traits. Nevertheless, this pattern is highly significant and some apparent exceptions are illuminating. Bromelioid bromeliads evolved fleshy fruits, but not net venation-which may be understandable, given that they also possess CAM photosynthesis and thus have thick, succulent leaves in which net venation would not be adaptive. CAM photosynthesis seems obviously adaptive in the open, dry habitats (Winter and Smith 1996) in which bromelioids evolved (Givnish et al. in press), but is also advantageous under the constantly damp, rainforest-interior conditions where most other bromelioids grow because it allows C0 2 recycling when the leaf surfaces are occluded with raindrops (Pierce et al. 2002). Vanilloid orchids (not included in our survey) evolved net venation but not fleshy fruits, except Vanilla Plum. ex Mill. itself (Cameron and Chase 1998)-which may also be understandable, given that mycotrophy in general appears to favor tiny, ex-ceedingly numerous seeds that are independently dispersable, presumably to maximize the chances of contacting a suitable fungal partner. Finally, the retention of net venation in several species of palms (Arecaceae) and yams (Dioscoreaceae) that have invaded open tropical and subtropical habitats speaks for the importance of phylogeny and genetic/ developmental heritage, not ecology, in helping maintain this trait. It is true that even yams of open, hot savannas often have thin, soft-textured leaves; presumably this is related to their short leaf lifespans, the ephemeral period of abundant moisture in their savanna habitats, and the widespread trend for specific leaf mass (g m-2 ) to decline with leaf longevity across biomes and (mainly dicot) lineages (Reich et al. 1997;Ackerly and Reich 1999). However, palms of open savannas and oases often have tough, coriaceous foliage and a relatively compact, palmate form-and yet retain a branching support structure within leaves, strongly supporting a role of phylogenetic morphological conservatism.
There are a few additional cases involving the concerted convergence of net venation and fleshy fruits beyond the monocot taxa we included in our survey. Examples include Palisota Rchb. ex Endl. (Commelinaceae), Vanilla and Selenipedium Rchb. f. (Orchidaceae), Eucharis Planch. & Linden (Amaryllidaceae), and Cyclanthaceae of tropical rainforest understories. Perhaps the most striking evidence that selection strongly favors both net venation and fleshy fruits under shaded conditions, however, is provided by Gnetum L. This genus of tropical vines and trees is characterized by fleshy fruits and broad, net-veined leaves that strongly resemble those of Coffea L. and other understory angiosperms-and yet Gnetum is a gymnosperm, closely related to the xeric-adapted Ephedra L. and Welwitschia Reichb. (Bowe et al. 2000;Chaw et al. 2000). The strong resemblance of Gnetum to certain angiosperms helped inspire the hypothesis that the angiosperms were derived from gymnosperms via Gnetales (Doyle and Donoghue 1986;Donoghue 1994). Molecular data do not support that hypothesis, however, indicating that the gymnosperms as a whole are sister to the angiosperms and that Gnetales arose from within the conifers (Bowe et al. 2000;Chaw et al. 2000;Soltis et al. 2002; but see Rydin et al. 2002). Won and Renner (2003) have recently discovered a horizontal transfer of a group II intron (a self-splicing RNA and putative spliceosomal ancestor) and adjacent exons of mitochondrial nadl from asterid angiosperms to a few Asian species of Gnetum. Although this might seem to open the possibility of a horizontal transfer of genes coding for net venation into Gnetum, such a scenario seems highly unlikely, given that the Asian species involved are nested well within Gnetum, all of whose species are characterized by net venation.
Phylogenetic analyses indicate that fleshy fruits have evolved repeatedly in association with forest understories in Lobeliaceae (Givnish 1998), Gesneriaceae (Smith 2001), and urticoid Rosales  among the dicots. Givnish ( 1979) observed that net venation occurs in several monocot groups with thin, broad leaves in forest understories, including Arisaema, Smilax L., Trillium, and various tropical gingers and their relatives. Conover ( 1983) and Chase et al. (1995a) independently noted similar, qualitative associations of net venation with broad-leaved forest vines; Cameron and Dickison (1998) noted a similar association of net venation with achlorophyllous vanilloid orchids. The association of net venation with the climbing habit among monocots is well marked. We believe it arises for three reasons: (1) most vines are, perforce, growing in microsites shaded by the hosts they are climbing; (2) species growing directly on tree boles are likely to experience especially dense shade, given that the boles occlude half the sky (Givnish and Vermeij 1976); and (3) the vine habit, by its nature, entails low allocation to support tissue, resulting in more rapid rates of upward growth and self-shading of lower leaves than in self-supporting plants with the same photosynthetic rate, and favoring shorter leaf lifespans and thus thinner, softer leaves with lower specific leaf mass . Indeed, a survey of 52 European woody species grown in a common garden showed that climbers/scramblers (6 species) had the lowest specific leaf mass of the species surveyed (Castro-Diez et a!. 2000). The association of net veins with the achlorophyllous vanilloid orchids most likely reflects initial adaptation of chlorophyllous ancestors to shady conditions, including the evolution of net venation (seen today in other shade-adapted orchid genera, such as Goodyera R. Br. and /sotria Raf.). Subsequently, evolution may have favored abandonment of the photosynthetic habit under such unproductive conditions and a focus on carbon input via mycotrophy, with further reduction in leaf size and thickness.
Given that both fleshy fruits and net venation each arose more than 20 times in the monocots, the question immediately arises as to whether the same developmental pathways and underlying genes were involved in each case, or whether these adaptations arose in different ways in different groups (as has occurred in different populations of rock pocket mice that have independently evolved dark pelage on dark-colored soils [Nachman et a!. 2003]). The fact that both fleshy fruits and net venation have arisen without the other in some cases demonstrates that they are unlikely to be the pleiotropic effects of a single gene or supergene. Furthermore, given that several groups show obvious differences in the fine details of their pattern of net venation (e.g., see Shinwari et a!. 1994a, b), and that "fleshy fruits" involve the elaboration of different tissues in different groups (e.g., arils vs. capsule walls), it seems unlikely that all of the multiple origins of net venation and fleshy fruits have each depended on the same genes and developmental pathways for each trait. Determining whether or not this has been the case should be a goal of new studies at the interface of ecology, evolution, and development ("eco-evo-devo"; Givnish 2003).
When Patterson and Givnish (2002) demonstrated that net venation, fleshy fruits, inconspicuous flowers, and rhizomes undergo concerted convergence under shady conditions in Liliales, they also showed that these patterns distorted phylogenetic inference based solely on morphology. When morphology was analyzed cladistically, two clades--characterized by the alternative suites of traits undergoing concerted convergence-emerged; when these traits were excluded from analysis, the relationships inferred were nearly identical to those deduced from DNA sequence variation. It would now be worthwhile to see if the same holds true for monocots as a whole: if both fruit and venation type are excluded, does an analysis of relationships across monocots based on morphology more closely approach that based on molecular data? Chase et a!. (1995a) have already noted that several of the groups placed at the base of the monocots by morphology alone (Stevenson and Loconte 1995) share reticulate venation. It would also be interesting to evaluate whetheras in Liliales-large, visually conspicuous flowers are mainly found in open habitats with strong illumination by broadspectrum light, and if small, visually inconspicuous flowers are mainly found in shaded sites with low illumination by narrow-spectrum, greenish light. Many forest herbs in eastern North America accord with this prediction (Givnish and Patterson 2000). Across the angiosperms as a whole, this hypothesis may account for the striking increase with rainfall in the fraction of woody species with inconspicuous flowers in tropical forests documented by Gentry ( 1982), given that most of the tree diversity added in wetter forests are understory species (Givnish 1999a). Repeated shifts to visually inconspicuous flowers in shade to visually conspicuous flowers under bright, full-spectrum light may be analogous to the increased sexual selection for bright coloration in African rift-lake cichlids in clearer waters (Seehausen et a!. 1997) and the likely role that an analogous process has played in the evolution of bright coloration in tropical coral-reef fish (Givnish 1999b).