q 2005 The Paleontological Society. All rights reserved. 0094-8373/05/3102-0006/$1.00

Paleobiology, 31(2), 2005, pp. 77–93

Key innovations, convergence, and success: macroevolutionarylessons from plant phylogeny

Michael J. Donoghue

Abstract.—Improvements in our understanding of green plant phylogeny are casting new light onthe connection between character evolution and diversification. The repeated discovery of para-phyly has helped disentangle what once appeared to be phylogenetically coincident characterchanges, but this has also highlighted the existence of sequences of character change, no one ele-ment of which can cleanly be identified as the ‘‘key innovation’’ responsible for shifting diversifi-cation rate. In effect, the cause becomes distributed across a nested series of nodes in the tree. Manyof the most conspicuous plant ‘‘innovations’’ (such as macrophyllous leaves) are underlain by ear-lier, more subtle shifts in development (such as overtopping growth), which appear to have enabledthe exploration of a greater range of morphological designs. Often it appears that these underlyingchanges have been brought about at the level of cell interactions within meristems, highlightingthe need for developmental models and experiments focused at this level. The standard practice ofattempting to identify correlations between recurrent character change (such as the tree growthhabit) and clade diversity is complicated by the observation that the ‘‘same’’ trait may be con-structed quite differently in different lineages (e.g., different forms of cambial activity), with somesolutions imposing more architectural limitations than others. These thoughts highlight the needfor a more nuanced view, which has implications for comparative methods. They also bear on issuescentral to Stephen Jay Gould’s vision of macroevolution, including exaptation and evolutionary re-currence in relation to constraint and the repeatability of evolution.

Michael J. Donoghue. Department of Ecology and Evolutionary Biology and Peabody Museum of NaturalHistory, Yale University, New Haven, Connecticut 06520. E-mail: michael.donoghue@yale.edu

Accepted: 7 September 2004

Introduction

Much of Stephen Jay Gould’s work was con-cerned, directly or indirectly, with patterns ofcharacter evolution, patterns of clade diversi-fication, and the causal link between thesetwo. Although Gould did not take an explic-itly phylogenetic approach to these problems,others have in recent years. In any case, ourknowledge of the Tree of Life has expandedenormously (Cracraft and Donoghue 2004)and it is worth considering how phylogeneticinsights may be influencing our views on mac-roevolution and especially the link betweencharacter evolution and diversification. In thisessay I provide the perspective of someoneworking on plant evolution, together with afew concrete plant examples. Gould was not,of course, especially interested in plants, buthis ideas were clearly intended to apply to or-ganisms of all sorts.

Specifically, I begin by briefly characterizingwhat we have learned recently about the fun-damental structure of green plant phylogeny,drawing a few generalizations about the na-

ture of that progress. Then I consider how thisprogress has been, or at least should be, af-fecting our understanding of the connectionbetween character evolution and diversifica-tion. My basic argument is that recent phylo-genetic findings are making it increasinglydifficult to sustain the traditional view of keyinnovations and also to maintain standardcomparative approaches to detecting the ef-fects of character change on diversity. Theserealizations suggest several new methodolog-ical needs and research strategies. In closing Ibriefly consider how these ideas relate to someof Gould’s views on macroevolution.

Progress in Understanding Plant Phylogeny

Figure 1 provides an overview of our pres-ent knowledge of phylogenetic relationshipsamong the major lineages of green plants.This is simplified, of course, and consciouslyrendered pectinate to serve my purposes (seeO’Hara 1992, on the representation of trees).Readers are referred to other recent reviews(Bateman et al. 1998; Chapman et al. 1998;

78 MICHAEL J. DONOGHUE

FIGURE 1. An overview of green plant phylogeny, illustrating the recent discovery of major clades (shaded groups);the monophyly of some traditionally recognized groups (shown at nodes with open circles) has been upheld, where-as others are now seen to be paraphyletic (names in quotation marks). † marks denote extinct groups. See text forreferences and discussion.

Doyle 1998; Kenrick 2000; Donoghue 2002,2004; Judd et al. 2002; Delwiche et al. 2004;Pryer et al. 2004; Soltis et al. 2004) for refer-ences to the primary literature underpinningFigure 1, and for levels of support, commen-tary on remaining controversies, and a widevariety of evolutionary implications nottouched upon here.

Several familiar and long-recognized taxa

are strongly supported as monophyletic.These include the entire green plant clade (theviridophytes), land plants (embryophytes),vascular plants (tracheophytes), seed plants(spermatophytes), flowering plants (angio-sperms), and monocotyledons (monocots).Conveniently, these clades are marked bycharacters that relate to their names: greenplants by chlorophyll b, land plants by a rest-

79LESSONS FROM PLANTS

ing embryo stage in the life cycle (hence em-bryophytes), vascular plants by vascular tis-sue with specialized cells for the transport ofwater (tracheids), seed plants by seeds (inte-gumented megasporangia), flowering plantsby one or more carpels in the shortened re-productive axes that we call flowers, andmonocots by embryos with just a single seedleaf (cotyledon).

Phylogenetic analyses conducted over thelast two decades have also shown that severalother traditionally recognized major groupsare not monophyletic, but instead representgrades of organization. Specifically, tradition-al ‘‘green algae,’’ ‘‘bryophytes,’’ ‘‘pterido-phytes’’ (seedless vascular plants), ‘‘gymno-sperms’’ (naked-seed plants), and ‘‘dicotyle-dons’’ appear to be paraphyletic. These hadeach been diagnosed on the basis of what wenow recognize to be ancestral traits. For ex-ample, green algae are green plants that lackthe specialized characteristics of the landplant clade (they live in the water, lack a rest-ing embryo, etc.). In bryophytes the sporo-phyte phase is unbranched and lacks vasculartissues of the sort found in tracheophytes. Asthe names implies, ‘‘seedless vascular plants’’are vascular plants that lack seeds, ‘‘gymno-sperms’’ are seed plants that lack carpels, andso forth.

Recognition that these traditional groupsare paraphyletic has, of course, resulted fromthe discovery of new major clades that uniteone or more of the lineages traditionally as-signed to the grade group directly with an in-cluded clade. For instance, the dismantling ofthe traditional green algae came aboutthrough the recognition that some groups for-merly treated as green algae are actually moreclosely related to land plants than they are toother green algal lineages. Specifically, it wasdiscovered (initially on the basis of ultrastruc-tural features, but now with much molecularsupport; [e.g., Karol et al. 2001]) that the Char-ales and several other lineages (e.g., Klebsor-midiales, Zygnematales, Coleochaetales) aremore closely related to lands plants than theyare to Chlorophyceae, Trebuxiophyceae, andUlvophyceae (the latter three making up theChlorophyte clade in the strict sense). Thename ‘‘streptophytes’’ has now been widely

applied to this newly discovered clade (Del-wiche et al. [2004], prefer the name ‘‘charo-phytes;’’ see Donoghue 2004).

Similarly, in the first phylogenetic analysesof land plants, hornworts and mosses werefound to be more closely related to vascularplants than to liverworts, the other major lin-eage of ‘‘bryophytes’’ (Mishler and Churchill1985). The term ‘‘stomatophytes’’ was coinedfor this clade, reflecting the presence of sto-mates in hornworts, mosses, and vascularplants. In recent years, however, several alter-native hypotheses have surfaced, especiallythe idea that the first split was between horn-worts and a clade containing the other threeclades (e.g., Nickrent et al. 2000; Renzaglia etal. 2000). In any case, phylogenetic analysesthat have sampled a sufficient number of rep-resentatives of these groups have supportedthe view that bryophytes do not form a cladebut rather represent a grade of organizationwithin land plants.

The name euphyllophytes has recently beenapplied to the clade including horsetails,whisk-ferns (psilophytes), various fern line-ages, and seed plants (e.g., Kenrick and Crane1997). These are more closely related to oneanother than to the other extant lineage ofseedless vascular plants, the lycophytes. Thename ‘‘anthophytes’’ was applied to the hy-pothesized clade including the ‘‘gymno-sperm’’ group Gnetales along with the flow-ering plants, to the exclusion of cycads, gink-gos, and conifers (Doyle and Donoghue 1986).As in the bryophyte case, many recent analy-ses (reviewed by Donoghue and Doyle 2000)do not support this anthophyte clade (Gne-tales instead being allied with conifers). Inany case, however, ‘‘gymnosperms’’ remainparaphyletic relative to angiosperms whenfossil groups (e.g., Paleozoic and Mesozoic‘‘seed ferns’’) are considered (Donoghue andDoyle 2000; Pryer et al. 2004); that is, the firstseed plants clearly lacked carpels. Finally,within flowering plants, the recently discov-ered eudicot clade (containing more than160,000 species) and a re-circumscribed mag-noliid clade (containing magnolias, black pep-pers, avocados, etc.) are found to be moreclosely related to monocots than they are tosome other lineages of ‘‘dicotyledons,’’ such

80 MICHAEL J. DONOGHUE

FIGURE 2. The nature of progress in resolving plantphylogeny. The upper tree shows the standard view asof the 1970s; the lower tree depicts current understand-ing. Major clades supported as monophyletic aremarked by open circles at the nodes; newly discoveredclades are marked by black circles.

FIGURE 3. An example of the impact of new phyloge-netic knowledge (discovery of the paraphyly of ‘‘greenalgae’’ and ‘‘bryophytes’’) on our understanding ofcharacter evolution. What once appeared to be clumpedchanges at key nodes (upper tree) can now be sortedinto a sequence of character changes (lower tree) thatclarify the transition to land and the origin of vascularplants.

as Amborella and the water-lilies (Zanis et al.2002; Soltis et al. 2004).

Figure 2 presents a cartoon summary ofthese results to highlight the nature of the pro-gress that has been made since the 1980s. Ingeneral, our advances have entailed confir-mation of the monophyly of some long-rec-ognized major clades, along with the recog-nition of a number of paraphyletic taxathrough the discovery of new major clades.Names such as ‘‘green algae,’’ ‘‘bryophytes,’’and ‘‘dicots’’ are now either being dropped al-together or being used only to refer to partic-ular life styles or grades of organization.Meanwhile, names such as euphyllophytesand eudicots are finding their way into intro-ductory textbooks (e.g., Judd et al. 2002) andare beginning to orient the way we think

about plant diversity and conduct research.The discovery and abandonment of paraphy-letic groups is, in general, what progress is allabout in phylogenetic systematics (Donoghue2004).

Character Sequences andDevelopmental Enablers

How have these advances changed our un-derstanding of plant evolution? The most ob-vious impact has been on our ability to dissectthe evolutionary sequence of events surround-ing the greatest transformations in green planthistory. For example, consider the transitionfrom life in the water to life on land (see Gra-ham 1993). When green algae and bryophyteswere both viewed as clades, this transition ap-peared to entail a very large number of stepsthat could not be placed in any particular tem-poral order (Fig. 3, top). This implied either alarge number of extinctions of intermediarytaxa and, consequently, major gaps in our

81LESSONS FROM PLANTS

knowledge, or a wholesale correlated trans-formation from one life form to another. Underthese circumstances several alternative theo-ries remained viable to explain the evolutionof features such as the land plant life cycle, en-tailing the alternation of multicellular haploid(gametophyte) and diploid (sporophyte)phases. Was a multicellular haploid phase ora multicellular diploid phase added to an an-cestral non-alternating life cycle? Or, perhapsthe ancestor of land plants belonged to a lin-eage within which alternation of generationshad already evolved. Did the precursors ofland plants live in salt water, fresh water, oreven on land (several ‘‘green algal’’ lineagesindependently made the transition to land)?What was the basic body plan from whichland plants evolved? After all, ‘‘green algae’’present an impressive number of alternatives,from unicells, to colonies, to filaments, topseudo-parenchymatous forms, with or with-out cell walls separating the nuclei. With noway to sort out the sequence of events, thetransition to land largely remained a mystery.

Knowing now that both the traditionalgreen algae and bryophytes are paraphyletic,and having succeeded in identifying the clos-est living relatives of land plants (Charalesand Coleochaetales [Karol et al. 2001]), we canstart to establish the sequence of events fromthe origin of the first green plants throughtheir movement onto land (Fig. 3, bottom). Onthis basis, we can be quite certain that landplants arose within a lineage of ‘‘green algae’’living in fresh water, probably quite near theshore. Their ancestors probably had rathercomplex parenchymatous construction, withgametes (and then zygotes) borne on the par-ent plant in specialized containers. Perhapsmost importantly, we can infer that the landplant life cycle originated through the inter-calation of a multicellular diploid phase (bydelaying the onset of meiosis) into a life cycleresembling that retained in Coleochaetalesand Charales (wherein the diploid zygote un-dergoes meiosis directly to form haploidspores). Likewise, we can infer that the firstland plants had a bryophyte-like life cycle inwhich the gametophyte was the dominantphase and the sporophyte was smaller andparasitic on the gametophyte.

Moving within land plants, the discovery ofthe polysporangiophyte clade (Kenrick andCrane 1997; see Pryer et al. 2004) implies thatenlargement and branching of the sporophytepreceded the acquisition of tracheids (Fig. 3).Moreover, fossil reconstructions of the game-tophytes of the first polysporangiophytes(Remy et al. 1993) suggest that the transitionto sporophyte dominance moved through astage in which gametophyte and sporophytephases were more or less similar in structure(so-called isomorphic alternation of genera-tions [Kenrick and Crane 1997]).

I provide this level of detail to draw atten-tion to the great significance of recent phylo-genetic advances, which have basically settledmany major questions about plant evolution.But the main point I want to make here is thatrecent phylogenetic discoveries don’t just helpus to choose among existing hypotheses, butalso shed genuinely new light on such prob-lems. Many of the newly discovered greenplant clades serve to focus our attention onseemingly minor—but in retrospect appar-ently quite profound—shifts in the nature ofplant development. Prime examples concernmeristem structure and function in relation tobranching. The polysporangiophyte clade ismarked by the ability of the sporophyte plantto branch dichotomously, as compared to theancestral unbranched condition retained to-day in the bryophytic lineages (Fig. 4). Di-chotomous branching made it possible for agiven sporophyte to produce more sporangiaand more spores per fertilization event, andperhaps generally to become larger (Mishlerand Churchill 1985; Knoll et al. 1986). Thisability of the apical meristem to branch ap-parently set the stage for a series of changesthat now mark the tracheophyte clade, notablythe evolution of differentiated vascular tissuesfor the flow of water and nutrients through alarger upright plant body. In retrospect, di-chotomous branching may have establishedthe conditions for—or enabled—the evolutionof increased size, of vascular tissue, and ofmany other downstream character changes.

The same line of reasoning applies to theevolution in euphyllophytes of the differenti-ation between a main axis, or trunk portion ofstem, and lateral branches (Fig. 4)—so-called

82 MICHAEL J. DONOGHUE

FIGURE 4. A comparison of sporophyte branchingamong early-branching lineages of land plants. In thebryophytic lineages (left) the sporophyte is unbranched;dichotomous branching evolved at the base of the po-lysporangiophytes (center); overtopping (or pseudo-monopodial growth) evolved at the base of the euphyl-lophytes (right). Insets at the top represent these differ-ences in branching at the level of the apical meristem.(Drawings at the bottom are from Stewart and Rothwell1993.)

overtopping or pseudomonopodial growth(Zimmermann 1965). This seemingly minorshift at the level of the shoot apical meristemappears to have enabled the evolution (mostlikely independently in several lineages [e.g.,Boyce and Knoll 2002; but see Schneider et al.2002; Pryer et al. 2004]) of the determinate lat-eral organs that we call leaves (or, more spe-cifically, megaphyllous leaves, as distinct fromthe so-called microphyllous leaves of lyco-phytes), and, in turn, the evolution of seedsand flowers. These derived traits (e.g., leaves,seeds) are often viewed as the key innovationsresponsible for the evolutionary success (usu-ally measured in terms of the number of spe-cies) of their respective lineages. Recent phy-logenetic discoveries have the effect of high-lighting subtle, but crucial, underlying devel-opmental shifts at the level of the apicalmeristem that made possible the evolution ofthe more obvious characters.

These observations have an important bear-ing on the identification of ‘‘key innovations.’’In two obvious ways the identification of keyinnovations becomes easier. First, as alreadynoted, recent progress has distributed in-ferred character changes across a series ofbranches as opposed to having them piled upat particular nodes (Fig. 3). The problem withhaving character changes concentrated at anode is that it is unclear which one (or whichcombination) of the changes might have trig-gered a shift in diversification rate. Decom-posing such a set of characters can help singleout the character(s) associated most directlywith shifts in diversification. Second, decom-posing paraphyletic groups reduces the num-ber of species in the sister group of the focalclade, thereby increasing the magnitude of thediversity contrast. For example, Charales andColeochaetales contain many fewer speciesthan did the traditionally circumscribed‘‘green algae’’ (with probably more than35,000 species). The discovery that Charales(with approximately 500 species) are sister toland plants (with over 300,000 species), and inturn that the Coleochaetales (with about 30species) are sister to the clade containing thesetwo, greatly accentuates the contrast in diver-sity between land plants and the several lin-eages to which they are most closely related.In general, this sort of change makes it easierto locate a significant shift in diversificationrate (Moore et al. 2004) and therefore increasesthe inclination to explain it with reference toa key character change.

However, a third impact of phylogeneticdiscoveries challenges the very notion of keyinnovations. The existence in our classifica-tions of major groups such as tracheophytes,spermatophytes, and angiosperms has drawnour attention to the obvious traits of theseclades—vascular tissue, seeds, and flowers—as potential drivers of diversification. The dis-covery of a set of new major clades, includingpolysporangiophytes, euphyllophytes, ligno-phytes, etc., likewise focuses our attention ontheir somewhat more subtle features—dichot-omous branching, pseudomonopodial growth,bifacial cambium, etc. The intercalation ofthese new clades between the traditionalgroups, I predict, will bring about a subtle but

83LESSONS FROM PLANTS

fundamental shift in how we view the link be-tween character evolution and success. De-spite the increased ease (just noted) withwhich we may be able to associate particularcharacter changes with shifts in diversifica-tion, I suspect that we will become increasing-ly less comfortable about phylogenetically lo-calizing ‘‘key’’ innovations. Instead, becausethere are often causal links between charactersthat evolved earlier and later in a sequence, itwill seem increasingly natural to think fromthe outset about a series of changes culminat-ing in a combination of traits that togetherserved to increase diversification. Appreciat-ing the interdependencies and the combinedeffects of character changes doesn’t just relo-cate the cause to another node in the tree, butinstead distributes the causation across a se-ries of nodes. As we become increasinglyaware of the ways in which apparently minordevelopmental changes early in a chain ren-dered new morphological designs accessible,we might even be tempted to view early stepsas actually necessitating later ones. But thecausal links will generally be much more sub-tle. Overtopping growth did not, we presume,necessitate the evolution of macrophyllousleaves. Instead, it enabled the exploration of anew set of morphological designs, whicheventually set the stage for the evolution ofleaves.

This refinement in outlook will, I suspect,reveal some important new evolutionary gen-eralities. For example, in the several cases wehave been considering (dichotomous branch-ing, overtopping) the enabling changes ap-pear to have been developmental shifts at thelevel of apical meristems, which presumablyinvolved shifts in gene expression and the lo-calization of signals at the level of cells andcell layers within the meristem. These under-lying changes appear now to be highly con-served, in the sense of showing little homo-plasy, which perhaps implies that the derivedstate was somewhat difficult to achieve in thefirst place and/or that the derived conditionrather quickly became burdened by the evo-lution of dependent traits. Paradoxically, de-spite the current entrenchment of such traits,they may initially have conferred greater flex-ibility, opening up new design possibilities

and consequently the exploitation (or ‘‘crea-tion’’) of new environments.

So far, the basic apical meristem featureshighlighted here (Fig. 4) have attracted ratherlittle attention from molecular developmentalbiologists. These characters are, after all, deep-ly embedded within the phylogeny of greenplants, a very great distance from the popularmodel organisms, and relevant mutationshave rarely been recorded. At this stage, eventhe formulation of credible developmentalmodels, and perhaps the identification of can-didate genes and appropriate study organ-isms, would be quite useful. Along these lines,Geeta (2003) has recently sketched such amodel for the origin of dichotomous branch-ing. This entails a duplication in the locationof the normal activity of the shoot apical mer-istem (regulated in part by the KNOX genepathway), possibly brought about by the pe-riodic expression of so-called MYB genes inthe center of the meristem (specifically theARP genes AS1/rs2/phan). My hope is thatspeculation of this type will encourage morecareful comparisons and experimental workin the relevant organisms (e.g., the apex of themoss sporophyte, branching in lycophytes).

Convergence and Equivalence

New phylogenetic results will also, I believe,bring about a shift in how we interpret the sig-nificance of the recurrence of similar characterstates in different lineages. We have rightlyviewed such cases as providing opportunitiesto test the effect of the evolution of a trait ofinterest on the evolution of other traits or ondiversification rate. A repeated association be-tween the evolution of a trait and elevated di-versification rates suggests a causal connec-tion. This seems reasonable, so long as we alsoappreciate that the effect might be somewhatindirect, or a function of the accumulation ofcharacters, as discussed above. However, rath-er little attention has been paid to negative re-sults—for example, where a trait is associatedwith increased diversification in one or a fewlineages, but not in other lineages, and thecorrelation ends up looking weak with respectto a predicted consistent effect. One has thesense that such mixed results are the norm, al-though this is difficult to assess because such

84 MICHAEL J. DONOGHUE

FIGURE 5. Differences between the bifacial cambium inthe lignophyte lineage (including seed plants) and theunifacial cambium found in extinct tree lycophytes (e.g.,Lepidodendron). The bifacial cambium produces both sec-ondary xylem and secondary phloem, and the cambialinitials are able to divide both periclinally (producingcells that differentiate in secondary tissues) and anti-clinally (producing new cambial initials). The unifacialcambium produced only secondary xylem and the cam-bial initials divided only periclinally, limiting expansionof the cambial cylinder and the production of wood.These seemingly minor differences translated into ma-jor differences in evolutionary flexibility and ‘‘success’’(see text).

‘‘insignificant’’ results tend not to be pub-lished.

What are we to make of such cases? One in-terpretation has recently been discussed by deQueiroz (2002), namely that the influence of aparticular sort of character change is contin-gent on other factors. That is, for example, theorigin of a particular state in a particular en-vironment (say, the herbaceous habit in a ter-restrial setting) may have a positive effect ondiversification, whereas the evolution of the‘‘same’’ trait in a different environment (say,herbaceousness in an aquatic habitat) mighthave little influence on diversification, or may-be even a negative effect. Feild et al. (2004) em-phasized the critical role of the environmentalcontext in understanding the function and theeffect on diversification of such ‘‘key’’ angio-sperm characters as vessels and closed car-pels.

This is an excellent point, but another inter-pretation also comes to mind. Maybe someways of making a trait are really somehow‘‘better’’ than others. After all, traits thatevolved independently in separate, distantlyrelated lineages are apt to be truly convergent(as opposed to parallel) in the sense of havingbeen constructed from different startingpoints, and possibly in quite different ways.Those differences might ultimately be of greatsignificance in terms of both the subsequentevolutionary changes that they enable and lin-eage ‘‘success.’’ Some ways of ‘‘solving’’ aproblem might ultimately be better than oth-ers in the sense of allowing greater evolution-ary flexibility.

Several cases from plant evolution come tomind, especially related to changes in organ-ism size and longevity. The tree growth habit(tall plants, with a thickened single trunk,branching well above ground level) evolvedmany times independently—in lycophytes,equisetophytes, and lignophytes (Fig. 1), toname a few prominent cases (others cases arediscussed briefly below). These cases all in-volved the same basic mechanism, namely theproduction in the stem of a cylinder of cam-bium—a secondary meristematic tissue thatproduces new cells to the inside and/or theoutside of the stem, thereby increasing thegirth of the stem. This, in concert with the evo-

lution of a variety of mechanical support sys-tems, allowed the evolution of large trees(Niklas 1997). Importantly, however, pains-taking paleobotanical studies have shown thatthe cambium functioned differently in thesedifferent clades. In lignophytes (including the‘‘progymnosperms,’’ such as Archaeopteris,and seed plants) we find the familiar situa-tion, in which the cambium is ‘‘bifacial’’—pro-ducing secondary xylem tissue toward thecenter of the stem and secondary phloem tis-sue toward the outside (Fig. 5, top). By con-trast, extinct tree lycophytes and equiseto-

85LESSONS FROM PLANTS

phytes are reconstructed as having had a un-ifacial cambium (Eggert 1961, 1962; Cichanand Taylor 1990). They produced cells only tothe inside, which differentiated as secondaryxylem tissue, but not toward the outside to bedifferentiated as secondary phloem (Fig. 5,bottom). Secondary phloem appears to havebeen entirely lacking in these plants (Eggert1972; Eggert and Kanemoto 1977).

Both types of construction allowed the evo-lution of large trees, but the differences in de-tail appear to have had profound consequenc-es. The familiar seed plant cambium originat-ed in the Devonian at the base of the ligno-phyte clade. Rather shortly thereafter, by theend of the Devonian, the major lineages ofseed plants (aside from the angiosperms) hadcome into existence, including a variety of‘‘seed fern’’ groups (looking rather like mod-ern tree ferns), cycad-like plants, conifer-likeplants, etc. This radiation spawned highlysuccessful lineages of woody plants from thestandpoint of their longevity, structural diver-sity, and species numbers.

By contrast, today there are only perhaps1200 species of lycophytes, the vast majority ofwhich (e.g., Selaginella, with approximately700 species) represent lineages that retainedthe ancestral herbaceous habit and never in-cluded trees. Large lycophyte trees evolved(perhaps several times) within a clade char-acterized by heterospory (i.e., the productionof two kinds of spores) and a flaplike ‘‘ligule’’associated with each leaf, and they diversifiedand became widespread especially during theCarboniferous. Isoetes (so-called quillworts),containing perhaps 150 species of small ro-sette plants, is the only living descendant ofthe lycophyte line in which secondary growthevolved—these plants have retained a cambi-um and ‘‘rootlets’’ that resemble those of theextinct trees (Gifford and Foster 1989). How-ever, Isoetes probably originated within the so-called cormos line of lycophytes (includingChaloneria and Pleuromeia), which never at-tained the great size of Lepidodendron and theother very large lycophyte trees. There are noliving representatives of this ‘‘rhizomorphic’’lineage. Similarly, the equisetophyte lineage(horsetails and relatives), which was diverseand produced large trees in the Carbonifer-

ous, is represented today by just 15 species ofEquisetum (Des Marais et al. 2003), all of themrelatively small plants lacking wood.

Although the down-sizing of the lycophyteand equisetophyte clades (in both plant stat-ure and species number) may not be directlyor entirely attributable to the structure of theirwood, the unifacial cambium does seem tohave placed significant functional constraintson the evolution of these plants—constraintsthat are reflected in a variety of other charac-teristics. First, relative to lignophytes with thebifacial cambium, the unifacial plants pro-duced rather little wood. This was not a func-tion of the unifacial cambium per se, but rath-er of the apparent inability of these plants toexpand greatly the circumference of the cam-bial cylinder (Cichan and Taylor 1990). Cam-bial cells in lignophytes can undergo both per-iclinal and anticlinal cell divisions, the peri-clinal ones adding xylem and phloem and theanticlinal ones adding extra cells to the ring ofcambium (Fig. 5). By contrast, cambial cells inunifacial plants apparently did not divide an-ticlinally. Consequently, any increases in thecambial ring were brought about by thegrowth of cambial initials in length, spreadingapart the cambial initials situated just aboveand below them in the cambial cylinder. Thismechanism can produce only very limited cir-cumferential increases, and the girth of theseplants may have resulted largely from some-thing analogous to the primary thickeningmeristem found today in palm trees (see be-low; Bateman et al. 1992; Bateman 1994). Thepaucity of wood formed by these plants ap-parently had several other consequences. Forone thing it meant that the wood that was pro-duced had to be especially efficient, andachieving this entailed structural changes inthe vascular tissue and the tracheary elementsthemselves (Cichan 1986). Also, because thewood of these plants could provide only min-imal mechanical support (as compared withlignophyte trees), in the lycophyte line a pe-culiar barklike ‘‘periderm’’ tissue (situated inthe outer cortex, beneath the persistent leafbases) was ‘‘invented’’ to stiffen the trunk.

Other peculiar attributes of these plants re-flect the lack of secondary phloem. In ligno-phytes, secondary phloem makes possible the

86 MICHAEL J. DONOGHUE

FIGURE 6. A sample of growth forms in extinct lycophytes. Two drawings on the left (from Phillips and DiMichele1992) show early stages in the life cycle—establishment of the stigmarian ‘‘root’’ system with possibly photosyn-thetic ‘‘rootlets’’ prior to rapid stem elongation. Three drawings on the right (from Stewart and Rothwell 1993)show reconstructed forms of the determinate stems (not drawn to the same scale); from left to right: Sigillaria,Pleuromeia, and Lepidodendron.

transport of carbohydrates from sites of pho-tosynthesis (typically leaves) to distant partsof the plant, such as the roots. In the absenceof secondary phloem, such long-distancetransport would be severely limited, whichwould necessitate the maintenance of photo-synthesizing structures in the vicinity of tis-sues that needed to stay alive in order to func-tion. Consequently, in the unifacial lineageswe find several highly unusual strategies. Fo-cusing now on lycophyte trees, we see themaintenance of photosynthesizing leaf basesall over the stems (and, consequently, the ab-sence of normal bark as in seed plant trees).These plants probably also provisioned theirmassive so-called stigmarian ‘‘root’’ systemsby producing photosynthesizing ‘‘rootlets’’(probably leaf homologs), some of which ap-pear to have been deployed above ground orinto shallow water in the swamps that most ofthese plants occupied (Fig. 6) (Phillips andDiMichele 1992).

The truly weird life cycles inferred for these

plants (Andrews and Murdy 1958; Eggert1961; DiMichele and Phillips 1985; Phillipsand DiMichele 1992) are also probably relatedto the lack of secondary phloem. It appearsthat the large lycophyte trees grew very littlein height for many years, instead remainingstumplike while the stigmarian system be-came well established underground (Fig. 6).Then they bolted up to great heights, quicklyproduced their spores (repeatedly, or onlyonce), and then died. In effect, the enormousabove-ground stems behaved like the inflores-cences of other plants. This highly unusual de-terminate growth mode (also found in equi-setophyte trees) may have been, in part, ameans of avoiding the long-term maintenanceof dispersed functioning tissues without sec-ondary phloem and the production of costlymechanical support tissues.

As noted above, the tree habit evolved inother lineages as well (Fig. 7). For example, ar-borescent forms are found among extinct mar-attialean (Psaronius) and filicalean (Tempskya)

87LESSONS FROM PLANTS

FIGURE 7. Diversity of form among extinct treelike plants from the Devonian and Carboniferous (not drawn to thesame scale). From left to right: Archaeopteris (an early lignophyte); Calamites (an equisetophyte); Psaronius (a mar-attialean ‘‘fern’’), in which the trunk was formed by a mantle of adventitious roots; Tempskya (a filicalean ‘‘fern’’),in which the trunk was formed by numerous smaller stems embedded in a tangle of adventitious roots.

ferns, and in modern tree ferns (Cyatheaceae).Trees were also re-evolved several times with-in the ancestrally herbaceous monocotyledon-ous flowering plants, with palm trees provid-ing a prime example. In each of these cases thetree habit was achieved in a distinctly differ-ent way, and again in each case with obviousdownstream consequences (Niklas 1997). Inthe upper Paleozoic Psaronius (Fig. 7) and inextant tree ferns, a cambium is lacking, andincreased girth and mechanical support areprovided by a mantle of adventitious inter-twining roots. Cross-sections of the trunk ofTempskya reveal yet another way to make atree—its ‘‘false’’ stems were made of manysmaller ramifying stems (each lacking second-ary growth) packed in among a dense thicketof adventitious roots (Andrews 1948). Similarconstruction is found today in the osmunda-ceous fern Todea. Palms also lack a vascularcambium, and increases in diameter are large-ly due to what is called a primary thickeningmeristem, situated in a zone where the youngleaves attach to the stem (Rudall 1991; Tomlin-

son 1995). Their mechanical strength is pro-vided by a combination of a greater density ofvascular bundles in the outer cortical tissueand thickening of the cell walls in that region(Niklas 1997). Some other monocots, within avariety of separate lineages living mostly inarid regions (e.g., Agavaceae, Convalariaceae,Iridaceae, Xanthorrhoeaceae [Chase et al.2000]), have also become trees (Tomlinson andZimmermann 1969; Tomlinson 1995). As inthe palms, the presence of vascular bundlesthat appear ‘‘scattered’’ in the stem (an atac-tostele) and of individual bundles that are‘‘closed’’ to further growth (both conditionsassociated with the origin of the monocots) ef-fectively precluded the re-evolution of a ‘‘nor-mal’’ ring of cambium. Instead, these plantsinvented a novel form of unifacial cambium(the ‘‘etagen’’ cambium), situated near the pe-riphery of the stem, which yields derivativesthat differentiate as additional ground tissueand into whole new vascular bundles contain-ing both xylem and phloem (Rudall 1991;Tomlinson 1995).

88 MICHAEL J. DONOGHUE

The point of giving these details about treeconstruction is to illustrate that different waysof attaining a given condition, evolved con-vergently in different lineages, can be con-strained by prior circumstances (e.g., the re-invention of a cambium in monocots withscattered bundles) and, most importantly forpresent purposes, can sometimes have signif-icant consequences for subsequent evolutionin those lineages (e.g., the bizarre structuresand life cycles of lycophytes). Such differencesamong lineages might translate into ‘‘nega-tive’’ or only weakly positive results in stan-dard phylogenetic comparative tests for char-acter correlations or for correlations with di-versity. As explained in the next section, I donot intend this as an argument against at-tempting to identify common evolutionary re-sponses to convergent characters across line-ages. Instead, I hope to highlight the potentialpower of negative results in such tests in help-ing to pinpoint consequential differences in‘‘the same’’ structure, thereby refining the ini-tial causal hypothesis.

Some Methodological Implications

Phylogenetic discoveries have been affect-ing macroevolutionary studies in a variety ofcompletely obvious ways. In general, in tryingto make sense of the tempo and mode of mac-roevolution it helps to know how species arerelated to one another. The point of my paperis that some much less obvious, but ultimatelymore fundamental, effects are on the horizon.Presently, we tend to want to pin the cause ofthe ‘‘success’’ of a clade on a ‘‘key innova-tion’’—used here to refer to a trait responsiblefor increasing the rate of diversification (seeGivnish 1997; Sanderson 1998; and Hunter1998; for alternative views on ‘‘key innova-tion’’ and ‘‘adaptive radiation’’). Tests of evo-lutionary character relationships and key in-novation hypotheses hinge on phylogeneticcorrelations. Does the character of interest re-ally correlate with a shift in diversification?Do we see repeated instances, in differentclades, of such a correlation?

I made the case above that key innovationsmay not happen at a point in a tree, but overa region. Likewise, shifts in diversificationmay ratchet upward (or downward) not at a

single spot in a tree, but over a series of nodes.New comparative methods need to be de-signed with this image in mind. We need teststhat attempt to identify particular sequencesof change that may have impacted diversifi-cation, as well as clusters of positive, but per-haps individually less than significant, shiftsin diversification rate (see Moore et al. 2004 forsome methodological developments alongthese lines).

Likewise, in testing for repeated evolution-ary correlations, more attention needs to bepaid to potentially significant character dif-ferences in different clades. In the case of trulyconvergent characters, as illustrated by theevolution of the tree habit, differences in con-structional details can have profound effectson subsequent evolution and, ultimately, onwhat we judge to be clade success. Failure toidentify a significant correlation in a phylo-genetic test could reflect such underlying dif-ferences and might help to refine the compar-ison. Ultimately, of course, it is critical to spec-ify a particular mechanistic connection be-tween the evolution of a trait and the evolutionof other traits and/or diversification rate. For-mulating the causal hypothesis as precisely aspossible will more clearly circumscribe whichinstances of ‘‘the same’’ character are relevantin performing a test (see Coddington 1994).

In many cases I imagine that an initial phy-logenetic test will narrow the set of compari-sons to characters with more specific similar-ities, perhaps often to cases of the parallel evo-lution of states in the strict sense (involvingthe same structural modifications and pre-sumably the same genes, and therefore per-haps in more closely related organisms; seediscussion in the next section). But, this is notto say that phylogenetic correlation tests areproperly applied only to parallel changes. In-stead, because the outcome, whether one useswildly convergent or only strictly parallelchanges, is potentially of interest, I am sug-gesting a nested data exploration strategy, be-ginning perhaps with obviously convergenttraits and narrowing down the comparisondepending on the results. For example, itseems well worth testing whether the treehabit, regardless of how it was actually at-tained, had a significant effect on the evolu-

89LESSONS FROM PLANTS

tion of other traits or on patterns of diversifi-cation. Likewise, to mention another popularcase in the plant literature (e.g., Donoghue1989; Heilbuth 2000; Vamosi et al. 2003), it isworthwhile testing whether dioecy and fleshy,bird-dispersed propagules are correlated, orwhether either one has influenced diversifi-cation, regardless of major structural differ-ences (e.g., whether the actual fleshy structureis the wall of the seed, the wall of the fruit, orsome accessory structure). But, where verydifferent structures are involved, we shouldnot be surprised or disappointed by negativeor ambiguous results. Instead, we shouldlearn from such experiences that structuraldetails might make a difference with respectto the presumed mechanistic hypothesis, andthen design more refined comparisons. Suchrefinements should take account of differentorganismal and environmental contexts (deQueiroz 2002), but they also should take moreseriously the distinction between convergenceand parallelism, which is often glossed over insuch work.

Some Connections to Gould

These observations connect to Steve Gould’sthoughts in a variety of ways. Gould presum-ably would have appreciated the idea of de-velopmental enablers—changes early in a se-quence that opened up new design options.But, exactly how such traits relate to Gould’sconcepts and terminology is a bit complicated.In my examples the underlying changes thatset the stage for later, more obvious changesare themselves, I presume, adaptations. Theyare what once would have been labeled ‘‘pre-adaptations,’’ a term that Gould rejected onthe grounds of its being ’’ ‘prepackaged’ forinevitable trouble and misunderstanding’’(Gould 2002: p. 1232). Gould and Vrba (1982)introduced the term ‘‘exaptation’’ to cover anyinstance of co-optation, whether from a pre-vious adaptation or from a nonaptation, butthey emphasized that ‘‘exaptations that beganas nonaptations represent the missing con-cept’’ (Gould and Vrba 1982: p. 12). Unfortu-nately, they left this more specific concept un-named. Gould (2002: p. 1278), therefore, re-cently distinguished between what he called‘‘franklins’’ (‘‘alternative potential functions

of objects now being used in another way’’)and ‘‘miltons’’ (‘‘currently unused materialorgans and attributes’’) as the basic elementsof the ‘‘exaptive pool.’’ As he pointed out,‘‘franklin’’ captures the concept behind theterm preadaptation and ‘‘milton’’ captures thenotion of nonaptations available for co-opting.Where do my plant examples fall in this ex-panded terminology? If I’m forced to useGould’s terms (which I must admit I have ahard time taking seriously), then my examplesare very likely ‘‘franklins.’’ That is, the under-lying traits that I have described as develop-mental enablers (e.g., dichotomous branching,overtopping) were probably adaptations intheir own right, but they also clearly providedinherent potential for future exaptive changes(e.g., to pseudomonopodial growth, leaves).

Having claimed that these cases are frank-lins, I hasten to note that I think there are alsoimportant miltons in plants, which have alsobeen brought to light in phylogenetic analy-ses. For example, in recent studies of the an-giosperm clade Dipsacales (a group of around1100 species of Asteridae), we have discussedthe evolution of a specialized structure calledan ‘‘epicalyx’’ (Donoghue et al. 2003). It ap-pears that the epicalyx evolved (possiblytwice, in Dipsacaceae and in Morinaceae)through modification of several sets of sub-tending ‘‘supernumerary’’ bracts, which weinterpret as having been ‘‘left over’’ from theearlier loss of flowers in the inflorescence(Donoghue et al. 2003). If so, the supernumer-ary bracts are miltons that were co-opted toform the epicalyx. It is difficult to quantify atthis stage, but in view of the nature of plantmorphology, and especially the evolutionaryuse and reuse of ‘‘leaves’’ for a very wide va-riety of purposes, I suspect that the co-opta-tion of miltons has been quite common inplant evolution. As for Gould’s distinctionsbetween ‘‘spandrels,’’ ‘‘manumissions,’’ and‘‘insinuations’’ (Gould 2002: p. 1278), I won’tattempt to further categorize the epicalyx. Inthis case, and in the other real examples thatcome to mind, these categories do not seemmutually exclusive enough to warrant the for-mality.

My discussion of convergence and successintersects another area of relevance to Gould’s

90 MICHAEL J. DONOGHUE

thought, namely the distinction between par-allelism and convergence, which he portrayedas critical for properly understanding the no-tion and the extent of ‘‘historical constraint.’’Parallelisms, he argued, reveal historical con-straints in the evolving system—the samecondition originates again and again within alineage owing to something about the struc-ture and development of the shared ancestor.Convergences, on the other hand, demonstratethe power of natural selection to fashion sim-ilar forms from very different starting points.Here again there are terminological issues.Gould (2002) provided a fine analysis of theconvergence-parallelism distinction but set-tled on a terminology that I think may not beideal. As he stressed, E. Ray Lankester, whocoined the term ‘‘homoplasy’’ in 1870, viewedit (ironically) as a form of homology (equiva-lent to Owen’s ‘‘general homology’’). Specifi-cally, Lankester meant to apply it to ‘‘inde-pendently evolved, but historically con-strained, similarities—what we would nowcall parallelisms’’ (Gould 2002: p. 1073). Nev-ertheless, Gould chose to follow standardpractice in applying ‘‘homoplasy’’ verybroadly to all sorts of non-homology, includ-ing both parallelism and convergence.

My own preference is to use ‘‘analogy’’ forall non-homologous similarities (e.g., as Os-born did in 1905), and to use ‘‘homoplasy’’ inthe more restricted (and original) sense to re-fer to parallelisms. ‘‘Homoplasy’’ would thenrefer precisely to the sorts of recurrent simi-larities detected in phylogenetic analyses.That is, it would refer to recurrences in thestates of characters that are actually includedin phylogenetic analyses on the working as-sumption that they are truly homologous be-cause they pass Remane’s positional, structur-al, and developmental tests of homology (Pat-terson 1982; Donoghue 1992). By contrast,convergences fail such tests and are excludedat the outset (as individual characters) fromphylogenetic analyses. Applying the terms inthis way would serve to connect these abstractdiscussions directly to work on levels of ho-moplasy in the phylogenetic literature (e.g.,Sanderson and Donoghue 1989, 1996). Metricsof the extent of parallelism (e.g., the consisten-

cy index) could then help to quantify the im-portance of historical constraint.

But, leaving aside these terminological is-sues, I quite agree with Gould that the paral-lelism-convergence distinction is importantfrom the standpoint of what it implies aboutthe mechanisms underlying character change.However, the point of my examples is differ-ent, namely that the distinction is also impor-tant because parallelisms and convergencesmay have rather different long-term evolu-tionary consequences. The tree habit as man-ifested by lycophytes had very different con-sequences (in terms of the evolution of othercharacters, and long-term success) than didthe tree habit as it evolved in lignophytes.Making (or failing to make) the convergence-parallelism distinction can have importantconsequences for comparative tests, and Ihave suggested a strategy of nested tests be-ginning with clear instances of convergenceand working toward parallelisms.

This last statement implies the existence ofa continuum between parallelism and conver-gence, which Gould also clearly appreciatedand used to his advantage. His basic argu-ment was that (1) parallelisms are importantbecause they reveal constraints due to deeperhomology, and (2) recent developmental stud-ies have revealed that many instances of sup-posed convergence are actually at least in partcases of parallelism. Therefore, (3) constrainthas been even more pervasive than we mighthave supposed. The critical link in his argu-ment is the contention that many real casesshow signs of both convergence and parallel-ism, a point he illustrated with examples suchas the role of the Pax-6 gene in the evolutionof eyes in different animal lineages. So, whatbegins as a plea for paying more attention tothe convergence-parallelism distinction endsup stressing that the distinction is a blurry oneat best. My guess is that this blurriness is evenmore pervasive than Gould imagined. Inplants, at least, with their modular, open de-velopmental systems, David Baum and I(Baum and Donoghue 2002) have argued thatcases of mixed or partial homology (see Sattler1984, 1991) may be common owing to ‘‘trans-ference of function,’’ especially between ad-jacent organs, brought about by shifts in the

91LESSONS FROM PLANTS

location where genes are expressed (what wetermed ‘‘homeoheterotopy’’). The epicalyx inDipsacales, mentioned above, may provide aconcrete example. That is, the calyx-like ap-pearance and function of the epicalyx mightreflect the activation of calyx identity genes ina newly formed structure adjacent to the calyx(Donoghue et al. 2003). In the end, it may bedifficult to sustain the notion of pure conver-gence, a thought that I suppose Gould wouldhave enjoyed.

Finally, these thoughts about recurrencealso bear on the issue of the role of conver-gence vis-a-vis the repeatability of evolution(Gould 1989; Conway Morris 1998, 2003; Con-way Morris and Gould 1998). Convergences,parallelisms, and mixtures of the two surelywill occur in any evolving systems, and atleast for parallelisms we can make concretepredictions about the frequency of occurrence(depending on the number of branchingevents, the number of character states, andrates of character evolution [Donoghue andRee 2000]). But the mere fact of recurrence, Iwould argue, does little to guarantee convinc-ing repeat performances in running the tapeof Life over again. The idea of convergence isthat structures are put together in differentways from different starting points in differ-ent lineages. If my argument is correct thatdifferences in construction (even seeminglyminor ones) can have major effects on down-stream evolutionary changes and patterns ofdiversification, then convergence on the samebasic form in different iterations might yieldwildly different outcomes. Large size, for ex-ample, may be selected again and again, butdepending on the details of how large size isactually attained, we might end up with verydifferent sorts of organisms. In one iterationwe might get familiar-looking lignophyte-liketrees (e.g., imagine a pine tree, or a maple), butin the next iteration we might see giant club-mosses or horsetails, and in a third go-aroundthe world might fill up with palm trees. Al-though there may be commonalities in what isselected for, different mechanisms underlyingthe response could translate into enormousdifferences in structure, life cycles, patterns of‘‘success,’’ ecological communities, and so on.So, approaching the problem from a different

angle, I end up squarely on Gould’s side of thisparticular argument.

Acknowledgments

I am very grateful to Elizabeth Vrba andNiles Eldredge for inviting me to contribute tothis volume, to J. Cracraft, S. Smith, R. Geeta,D. Stevenson, C. Delwiche, J. Doyle, L. Hickey,and W. DiMichele for helpful discussions, andto T. Feild and an anonymous reviewer fortheir comments on the manuscript. B. Mooredeserves special thanks for critically readingthe manuscript, and for his insights basedpartly on having reached similar conclusionsabout trait interdependence in relation to keyinnovations in his master’s work at the Uni-versity of Toronto. He also kindly producedthe figures. Although I would argue that SteveGould never fully embraced phylogeneticthinking, phylogenetic biologists (even botan-ical ones) are deeply indebted to him for de-veloping macroevolutionary ideas that willkeep us occupied for many years to come.

Literature CitedAndrews, H. N. 1948. Fossil tree ferns of Idaho. Archaeology 1:

190–195.Andrews, H. N., and W. H. Murdy. 1958. Lepidophloios and on-

togeny in arborescent lycopods. American Journal of Botany45:552–560.

Bateman, R. M. 1994. Evolutionary-developmental change in thegrowth architecture of fossil rhizomorphic lycopsids: scenar-ios constructed on cladistic foundations. Biological Reviewsof the Cambridge Philosophical Society 69:527–597.

Bateman, R. M., W. A. DiMichele, and D. A. Willard. 1992. Ex-perimental cladistic analyses of anatomically preserved ar-borescent lycopsids from the Carboniferous of Euramerica: anessay in paleobotanical phylogenetics. Annals of the MissouriBotanical Garden 79:500–559.

Bateman, R. M., P. R. Crane, W. A. DiMichele, P. R. Kenrick, N.P. Rowe, T. Speck, and W. E. Stein. 1998. Early evolution ofland plants: phylogeny, physiology and ecology of the pri-mary terrestrial radiation. Annual Review of Ecology andSystematics 29:263–292.

Baum, D. A., and M. J. Donoghue. 2002. Transference of func-tion, heterotopy, and the evolution of plant development. Pp.52–69 in Q. Cronk, R. Bateman, and J. Hawkins, eds. Devel-opmental genetics and plant evolution. Taylor and Francis,London.

Boyce, C. K., and A. H. Knoll. 2002. Evolution of developmentalpotential and the multiple independent origins of leaves inPaleozoic vascular plants. Paleobiology 28:70–100.

Chapman, R. L., M. L. Buchheim, C. F. Delwiche, T. Friedl, V. A.R. Huss, K. G. Karol, L. A. Lewis, J. Manhart, R. M. McCourt,J. L. Olsen, and D. A. Waters. 1998. Molecular systematics ofgreen algae. Pp. 508–540 in D. Soltis, P. Soltis, and J. Doyle,eds. Systematics of plants II. Kluwer Academic, New York.

Chase, M. W., D. E. Soltis, P. S. Soltis, P. J. Rudall, M. F. Fay, W.J. Hahn, S. Sullivan, J. Joseph, M. Molvray, P. J. Kores, T. J. Giv-

92 MICHAEL J. DONOGHUE

nish, K. J. Sytsma, and J. C. Pires. 2000. Higher-level system-atics of the monocotyledons: an assessment of current knowl-edge and a new classification. Pp. 3–16 in K. Wilson and D.Morrison, eds. Monocots: systematics and evolution. CSIROPublishing, Collingwood, Victoria, Australia.

Cichan, M. A. 1986. Conductance of the wood of selected Car-boniferous plants. Paleobiology 12:302–310.

Cichan, M. A., and T. N. Taylor. 1990. Evolution of cambium ingeological time—a reappraisal. Pp. 213–228 in M. Iqbal, ed.The vascular cambium. Wiley, Somerset, U.K.

Coddington, J. A. 1994. The role of homology and convergencein studies of adaptation. Pp. 53–78 in P. Eggleton and R. Vane-Wright, eds. Phylogenetics and ecology. Academic Press, Lon-don.

Conway Morris, S. 1998. The crucible of creation. Oxford Uni-versity Press, Oxford.

———. 2003. Life’s solution: inevitable humans in a lonely uni-verse. Cambridge University Press, Cambridge.

Conway Morris, S., and S. J. Gould. 1998. Showdown on the Bur-gess shale. [The Challenge by Conway Morris and the Replyby S. J. Gould.] Natural History 107:48–55.

Cracraft, J., and M. J. Donoghue, eds. 2004. Assembling the treeof life. Oxford University Press, New York.

Delwiche, C. F., R. A. Anderson, D. Bhattacharya, B. D. Mishler,and R. M. McCourt. 2004. Algal evolution and the early ra-diation of green plants. Pp. 121–137 in Cracraft and Dono-ghue, 2004.

de Queiroz, A. 2002. Contingent predictability in evolution: keytraits and diversification. Systematic Biology 51:917–929.

Des Marais, D. L., A. R. Smith, D. M. Britton, and K. M. Pryer.2003. Phylogenetic relationships and evolution of extanthorsetails, Equisetum, based on chloroplast DNA sequencedata (rbcL and trnL-F). International Journal of Plant Sciences164:737–751.

DiMichele, W. A., and T. L. Phillips. 1985. Arborescent lycopodreproduction and paleoecology in a coal-swamp environmentof Late Middle Pennsylvanian age (Herrin Coal, Illinois,USA). Review of Paleobotany and Palynology 44:1–26.

Donoghue, M. J. 1989. Phylogenies and the analysis of evolu-tionary sequences, with examples from seed plants. Evolution43:1137–1156.

———. 1992. Homology. Pp. 170–179 in E. Keller and E. Lloyd,eds. Keywords in evolutionary biology. Harvard UniversityPress, Cambridge.

———. 2002. Plants. Pp. 911–918 in M. Pagel, ed. Encyclopediaof evolution, Vol. 2. Oxford University Press, Oxford.

———. 2004. Immeasurable progress on the Tree of Life. Pp.548–552 in Cracraft and Donoghue, 2004.

Donoghue, M. J., and J. A. Doyle. 2000. Demise of the antho-phyte hypothesis? Current Biology 10:R106–R109.

Donoghue, M. J., and R. H. Ree. 2000. Homoplasy and devel-opmental constraint: a model and an example from plants.American Zoologist 40:759–769.

Donoghue, M. J., C. D. Bell, and R. C. Winkworth. 2003. The evo-lution of reproductive characters in Dipsacales. InternationalJournal of Plant Sciences 164:S453–S464.

Doyle, J. A. 1998. Phylogeny of the vascular plants. Annual Re-view of Ecology and Systematics 29:567–599.

Doyle, J. A., and M. J. Donoghue. 1986. Seed plant phylogenyand the origin of angiosperms: an experimental cladistic ap-proach. Botanical Review 52:321–431.

Eggert, D. A. 1961. The ontogeny of Carboniferous arborescentLycopsida. Palaeontographica Abteilung B 108B:43–92.

———. 1962. The ontogeny of Carboniferous arborescent Sphen-opsida. Palaeontographica 110B:99–127.

———. 1972. Petrified Stigmaria of sigillarian origin from NorthAmerica. Review of Paleobotany and Palynology 14:85–99.

Eggert, D. A., and N. Y. Kanemoto. 1977. Stem phloem of MiddlePennsylvanian Lepidodendron. Botanical Gazette 138:102–111.

Feild, T. S., N. C. Arens, J. A. Doyle, T. E. Dawson, and M. J. Don-oghue. 2004. Dark and disturbed: a new image of early an-giosperm ecology. Paleobiology 30:82–107.

Geeta, R. 2003. Variation and diversification in plant evo-devo[book review]. American Journal of Botany 90:1257–1261.

Gifford, E. M., and A. S. Foster. 1989. Morphology and evolutionof vascular plants, 3d ed. W. H. Freeman, New York.

Givnish, T. J. 1997. Adaptive radiation and molecular system-atics: issues and approaches. Pp. 1–54 in T. Givnish and K.Sytsma, eds. 1997. Molecular evolution and adaptive radia-tion. Cambridge University Press, Cambridge.

Gould, S. J. 1989. Wonderful life: the Burgess shale and the na-ture of history. Norton, New York.

———. 2002. The structure of evolutionary theory. Harvard Uni-versity Press, Cambridge.

Gould, S. J., and E. S. Vrba. 1982. Exaptation—a missing term inthe science of form. Paleobiology 8:4–15.

Graham, L. E. 1993. Origin of the land plants. Wiley, New York.Heilbuth, J. C. 2000. Lower species richness in dioecious clades.

American Naturalist 156:221–241.Hunter, J. P. 1998. Key innovations and the ecology of macro-

evolution. Trends in Ecology and Evolution 13:31–36.Judd, W. S., C. S. Campbell, E. A. Kellogg, P. F. Stevens, and M.

J. Donoghue. 2002. Plant systematics: a phylogenetic ap-proach, 2d ed. Sinauer, Sunderland, Mass.

Karol, K. G., R. M. McCourt, M. T. Cimino, and C. F. Delwiche.2001. The closest living relatives of land plants. Science 294:2351–2353.

Kenrick, P. 2000. The relationships of vascular plants. Philo-sophical Transactions of the Royal Society of London B 355:847–855.

Kenrick, P, and P. R. Crane. 1997. The origin and early diversi-fication of land plants: a cladistic study. Smithsonian Insti-tution Press, Washington, D.C.

Knoll, A. H., S. W. F. Grant, and J. W. Tsao. 1986. The early evo-lution of land plants. In T. Broadhead, ed. Land plants: notesfor a short course. Studies in Geology 15:45–63. Departmentof Geology,University of Tennessee, Knoxville.

Lankester, E. R. 1870. On the use of the term homology in mod-ern zoology, and the distinction between homogenetic andhomoplastic agreements. Annals of the Magazine of NaturalHistory, 4th series, 6:34–43.

Mishler, B. D., and S. P. Churchill. 1985. Transition to a land flo-ra: phylogenetic relationships of the green algae and bryo-phytes. Cladistics 1:305–328.

Moore, B. R., K. M. A. Chan, and M. J. Donoghue. 2004. Detect-ing diversification rate variation in supertrees. Pp. 487–533 inO. Bininda-Emonds, ed. Phylogenetic supertrees: combininginformation to reveal the tree of life. Kluwer Academic, NewYork.

Nickrent, D., C. L. Parkinson, J. D. Palmer, and R. J. Duff. 2000.Multigene phylogeny of land plants with special reference tobryophytes and the earliest land plants. Molecular Biologyand Evolution 17:1885–1895.

Niklas, K. J. 1997. The evolutionary biology of plants. Universityof Chicago Press, Chicago.

O’Hara, R. J. 1992. Telling the tree: narrative representation andthe study of evolutionary history. Biology and Philosophy 7:135–160.

Osborn, H. F. 1905. The ideas and terms of modern philosoph-ical anatomy. Science 21:959–961.

Patterson, C. 1982. Morphological characters and homology. Pp.21–74 in K. Joysey and A. Friday, eds. Problems of phyloge-netic reconstruction. Academic Press, London.

Phillips, T. L., and W. A. DiMichele. 1992. Comparative ecologyand life-history biology of arborescent lycopsids in Late Car-

93LESSONS FROM PLANTS

boniferous swamps of Euramerica. Annals of the Missouri Bo-tanical Garden 79:560–588.

Pryer, K. M., H. Schneider, and S. Magallon. 2004. The radiationof vascular plants. Pp. 138–153 in Cracraft and Donoghue,2004.

Remy, W., P. G. Gensel, and H. Hass. 1993. The gametophytegeneration of some early Devonian land plants. InternationalJournal of Plant Sciences 154:35–58.

Renzaglia, K. S., R. J. Duff, D. L. Nickrent, and D. J. Garbary.2000. Vegetative and reproductive innovations of early landplants: implications for a unified phylogeny. PhilosophicalTransactions of the Royal Society of London B 355:768–793.

Rudall, P. J. 1991. Lateral meristems and stem thickening growthin monocotyledons. Botanical Review 57:150–161.

Sanderson, M. J. 1998. Reappraising adaptive radiation. Amer-ican Journal of Botany 85:1650–1655.

Sanderson, M. J., and M. J. Donoghue. 1989. Patterns of variationin levels of homoplasy. Evolution 43:1781–1795.

———. 1996. The relationship between homoplasy and confi-dence in a phylogenetic tree. Pp. 67–89 in M. Sanderson andL. Hufford, eds. Homoplasy and the evolutionary process. Ac-ademic Press, San Diego.

Sattler, R. 1984. Homology—a continuing challenge. SystematicBotany 9:382–394.

———. 1991. Process homology: structural dynamics in devel-opment and evolution. Canadian Journal of Botany 70:708–714.

Schneider, H., K. M. Pryer, R. Cranfill, A. R. Smith, and P. G.Wolf. 2002. Evolution of vascular plant body plans: a phylo-genetic perspective. Pp. 330–363 in Q. Cronk, R. Bateman, andJ. Hawkins, eds. Developmental genetics and plant evolution.Taylor and Francis, London.

Soltis, P. S., D. E. Soltis, M. W. Chase, P. K. Endress, and P. R.Crane. 2004. The diversification of flowering plants. Pp. 154–167 in Cracraft and Donoghue, 2004.

Stewart, W. N., and G. W. Rothwell. 1993. Paleobotany and theevolution of plants, 2d ed. Cambridge University Press, NewYork.

Tomlinson, P. B. 1995. Non-homology of vascular organizationin monocotyledons and dicotyledons. Pp. 589–622 in P. Ru-dall, P. Cribb, D. Cutler, and C. Humphries, eds. Monocoty-ledons: systematics and evolution, Vol. II. Royal Botanic Gar-dens, Kew, U.K.

Tomlinson, P. B., and M. H. Zimmermann. 1969. Vascular anat-omy of monocotyledons with secondary growth—an intro-duction. Journal of the Arnold Arboretum 50:159–179.

Vamosi, J. C., S. P. Otto, and S. C. H. Barrett. 2003. Phylogeneticanalysis of the ecological correlates of dioecy in angiosperms.Journal of Evolutionary Biology 16:1006–1018.

Zanis, M. J., D. E. Soltis, P. S. Solits, S. Mathews, and M. J. Don-oghue. 2002. The root of the angiosperms revisited. Proceed-ings of the National Academy of Sciences USA 99:6848–6853.

Zimmermann, W. 1965. Die Telomtheorie. Fischer, Stuttgart.