Smith & Turner 2005

166 SYSTEMATIC BIOLOGY VOL. 54

Syst. Biol. 54(1):166–173, 2005Copyright c© Society of Systematic BiologistsISSN: 1063-5157 print / 1076-836X onlineDOI: 10.1080/10635150590906000

Morphology’s Role in Phylogeny Reconstruction: Perspectives from Paleontology

NATHAN D. SMITH1 AND ALAN H. TURNER1,2

1Department of Geoscience, University of Iowa, Iowa City, IA 52242, USA; E-mail: [email protected], [email protected] address: Division of Paleontology, American Museum of Natural History, 79th Street at Central Park West, New York, NY 10024, USA

A recent article by Scotland et al. (2003; hereafterreferred to as SEA) purporting to examine the valueof morphological data in phylogeny reconstruction hasbeen received critically by several systematists (Jenner,2004; Wiens, 2004). As paleontologists—and in an area ofsystematics restricted solely to morphological data—wetake exception to many of the arguments put forward bySEA and feel we may provide a unique perspective inthe debate.

In their paper, SEA argued for a redefined rolefor morphology in phylogeny reconstruction, onein which “rigorous and critical anatomical studiesof fewer morphological characters, in the context ofmolecular phylogenies, is a more fruitful approach tointegrating the strengths of morphological data withthose of sequence data” (p. 539). Such a statement isbold and therefore warrants a critical analysis, as itwould effectively neuter the ability of morphologicaldata to generate novel phylogenies. Though issues suchas accuracy, support, character coding, and characterconceptualization were discussed by SEA, in all casesthese discussions resorted to a “too few characters”argument. As the authors characterized it, the “mainconstraint of morphology-based phylogenetic inferenceconcerns the limited number of unambiguous charactersavailable for analysis in a transformational framework”(p. 539). Additionally, the merits of increased taxonsampling in the context of morphological data werediscussed by SEA, and found to be lacking.

Though several of the views presented by SEA are notnovel (see Hedges and Sibley, 1994, and Hedges andMaxson, 1996), they provide the most detailed recent dis-cussion of this position. We agree with many of the pointsmade by SEA, especially the call for more critical and rig-orous analysis of morphology; however, we draw differ-ent conclusions from the data. It is our goal to reexaminesome of the arguments put forward by SEA, in order toillustrate that a much more optimistic conclusion existsregarding the current and future role of morphology inphylogeny reconstruction.

PHYLOGENETIC ACCURACY AND DATA SET SUPPORT

SEA (p. 539) describe the relationship between phy-logenetic accuracy and number of characters with theirFigure 1a (reproduced here as Fig. 1a), adapted from sim-ulation studies of Hillis (1996, 1998). Simulation studieshave generally demonstrated that increasing the num-ber of characters in an analysis will lead to an increase

in accuracy (Nei et al., 1983; Kim and Burgman, 1988;Rohlf and Wooten, 1988; Huelsenbeck and Hillis, 1993;Charleston et al., 1994; Hillis et al., 1994a, 1994b; Hillis,1996, 1998; Givnish and Sytsma, 1997; Rosenberg andKumar, 2001). SEA (p. 541) refer to this relationship asevidence that the number of characters utilized in mor-phological analyses is less than the number that is “typ-ically required to accurately reconstruct phylogenies insimulation studies.”

We take issue with SEA’s (p. 541) interpretation ofHillis’ (1996, 1998) results. The studies of Hillis (1996,1998) were explicitly examining the number of charactersthat were required to accurately reconstruct a 228 taxontree (Fig. 2), and were in no way intended to equate spe-cific numbers of characters with specific levels of accu-racy for any phylogenetic analysis, though they wereintended to examine the possibility of accurately re-constructing large phylogenies. More importantly, themain point of Hillis’ (1996, 1998) simulation studies wasthat relatively few characters were required to accuratelyreconstruct a 228 taxon tree, given dense taxonomicsampling.

An equally critical point to bear in mind is thatthe “true” 228 taxon tree was originally inferred from18S RNA by parsimony. This phylogeny and a specificmodel of nucleotide substitution (Kimura 2-parameter,alpha/beta = 2, gamma distribution of rate heterogene-ity with shape parameter = 0.5; Hillis, 1998) were thenused to generate artificial data sets, and the ability ofparsimony (and neighbor-joining) to accurately estimatethe phylogeny from these data sets was then measured.Given that this particular model of evolution is designedfor molecular, not morphological, characters, and giventhat morphological characters (and more than likelymolecular characters as well) probably do not evolve ac-cording to this specific model, it seems unwarranted tojudge the quality of morphological data sets (in terms ofnumbers of characters needed to accurately reconstructthe true phylogeny) through a comparison with the re-sults of Hillis’ (1996, 1998) simulation studies.

There is simply no fixed relationship between char-acter sampling and accuracy. Furthermore, most simu-lation studies have not examined number of charactersand character/taxon ratios independently; i.e., these pa-rameters tend to covary in most studies. The fact thatcharacter/taxon ratios may not accurately capture theratio between number of taxa and possible variation isoften overlooked as well. A more informative metric inthis regard may be a character state/taxon ratio. This

2005 POINTS OF VIEW 167

FIGURE 1. Possible relationship between an increase in the number of characters and features of the phylogeny. (a) Accuracy (Hillis, 1996,1998). (b) Bootstrap support (bremer et al., 1999). (c) Ease of homology assessment in morphological studies. (d) Character coding in morphologicalstudies. Text and figure reproduced as in Scotland et al. (2003: fig. 1).

would better deal with biases in data type (i.e., con-straints on number of character states present in discretemorphological, nucleotide, amino acid, and quantitativemorphological data).

CHARACTER CODING AND AMBIGUITY

Part of SEA’s (p. 541) argument focuses on the claimthat the “number of unambiguously coded morphologi-

FIGURE 2. Performance of parsimony in estimating a 228-taxon treeplotted against sequence length. “Percent of tree correct” is based onthe partition metric (Robinson and Foulds, 1981; Penny and Hendy,1985). Text and figure are reproduced from Hillis (1998: fig. 2).

cal characters for any study is finite (Fig. 1d) and less thanthe number typically required to accurately reconstructphylogenies in simulation studies (Fig. 1a).” The authorsdefend this claim with a comparison of a figure depict-ing the relationship between accuracy and number ofcharacters (Fig. 1a; Hillis, 1996, 1998), and another figuredepicting the relationship between character coding andnumber of characters (Fig. 1d). Figure 1a appears misap-plied. Though simulation studies may provide an ideaof how many characters are required to reach an upperplateau of accuracy given specific evolutionary models(Hillis, 1996, 1998), no study has attempted to determine,or even examine, the “spectrum” of character ambiguityto which SEA refer. Figure 1d is uninformative, as it isbased on no empirical data.

SEA state that there is no ambiguity in assigning char-acter states to sequence data once the sequences havebeen aligned. Although true, this overlooks the potentialfor error in sequence alignment (Morrison and Ellis, 1997;Goldman, 1998). Additionally, Hillis and Wiens (2000:12)cite multiple sources of homology problems unique tomolecular data, including gene duplication, horizontaltransfer, and exon shuffling. SEA (p. 541) acknowledgebut dismiss these as serious problems.

Though sources of ambiguity in morphological andmolecular analyses are different, both can act to con-found phylogenetic analysis (Hillis and Wiens, 2000).Unless it can be demonstrated that problems unique tothe analysis of morphological data adversely affect morecharacters than do problems unique to the analysis ofmolecular data, claims such as those made by SEA re-duce to assertions that morphological data sets typically


have fewer numbers of characters than DNA sequencedata sets. Such arguments can by no means be used asjustification for excluding morphological data from play-ing an independent role in phylogeny reconstruction.

CHARACTER CONCEPTUALIZATION AND HOMOLOGYASSESSMENT

Character delineation and primary homology assess-ment of morphological data have traditionally been citedas inherently problematic due to their oftentimes sub-jective nature. In recent years, however, much progresshas been made in terms of recognizing and amelioratingthese biases and establishing a more rigorous frameworkfor morphological character conceptualization (Wiens,2000; Rieppel and Kearney, 2002; Wagner, 2001). Further-more, many problems related to homology assessmentare by no means unique to morphological data.

SEA (p. 541) refer to their graph (Fig. 1c) of the “rela-tionship between homology assessment and number ofmorphological characters” to make three claims:

1. “There are few [morphological] characters that seemto be uncontroversial in relation to homology assess-ment” (p. 541).

2. These characters are typically “the first [morphologi-cal] characters to be included in a phylogenetic dataset” (p. 541).

3. “Increasing the number of [morphological] charactersincreases the level of ambiguous or problematic char-acters” (p. 541).

We feel these claims oversimplify the problem of ho-mology assessment in morphological data. As in othertypes of data, these problems are not distributed uni-formly across phylogenetic studies and are more preva-lent at particular hierarchical levels (i.e., levels wherehomologous structures have diverged greatly, or barelyat all, in form from each other). Interestingly, these areoften the hierarchical levels where homology assessmentbecomes more controversial for molecular data as well,albeit for different reasons. We believe that some of themost interesting questions in evolutionary biology existat these particular levels, and (due in part to the specificproblems they cause for various types of phylogeneticdata) that the most fruitful approach for exploring thesequestions involves utilizing a variety of different phylo-genetic data.

Regarding their second point, we suggest that thefirst characters included in any phylogenetic analysisare typically those that are easily accessed and that theresearcher feels will give him/her the best estimate ofthe phylogeny. For a morphological analysis this maymean, as SEA (p. 541) suggest, characters that are rel-atively uncontroversial in regards to homology assess-ment. For an analysis of DNA sequence data this maymean characters from an appropriate gene, or charactersfrom a region where alignment is relatively uncontro-versial (i.e., a region that is uncontroversial in regards tohomology assessment).

ARE MORPHOLOGICAL CHARACTERS MOREHOMOPLASTIC?

SEA (p. 542) argue that additional characters incor-porated into a morphological analysis will generally beof limited value, whereas additional characters incor-porated into an analysis of DNA sequence data willtypically be of equal value. This assertion presumes afixed, low number of characters that are actually phy-logenetically informative in a morphological analysis,a claim that is prevalent throughout SEA and that isnowhere supported by actual evidence. Additional char-acters incorporated into a phylogenetic analysis, be theymorphological or molecular, are just as likely to be homo-plastic as the characters initially utilized (Sanderson andDonoghue, 1989; though see Hedges and Maxson, 1996;Sanderson and Donoghue, 1996; Givnish and Sytsma,1997; and Baker et al., 1998, for claims that morphologi-cal or molecular data are more homoplastic in general),and any claim to the contrary must be backed up by em-pirical data.

One test of SEA’s assertion would be to identify recentphylogenetic analyses of morphological data for whichreferences (author and year) are given for each characterutilized (i.e., each analysis would effectively be a com-pilation of earlier analyses). Retention indices (R.I.) forthe newer characters (from more recent studies) couldbe compared against the R.I.’s from older characters(included in the earliest studies). SEA’s hypothesiswould predict that the R.I.’s of the newer characterswould be significantly lower than those of the older char-acters. To our knowledge, no such comparison has beencompleted, and we suspect it would find no significantdifference (see Wiens, 2004, for an empirical exampledemonstrating the absence of correlation between num-ber of characters and levels of homoplasy).

There is a practical reason why new morphologicalcharacters might not have inferior homoplasy statistics,such as lower R.I.’s (C. Brochu, personal communica-tion). In the case of many data sets incorporating newtaxa, new characters are introduced to express a close re-lationship between the new form and another taxon orclade already in the data matrix. In cases such as this,most new characters would be expected to have higherR.I.’s. The distinction rests between new taxa that requirea wholesale revisit of group relationships and those thatrepresent another twig on the tree.

Hillis and Wiens (2000) discussed some of the prob-lems caused by homoplastic characters in phylogeneticstudies and aptly pointed out that homoplasy in theform of convergent evolution of characters is a problemfaced equally by morphological data and molecular data(2000:11). The authors supported these claims by point-ing to empirical studies (Cunningham et al., 1997; Bullet al., 1997) which conclusively demonstrated that selec-tive convergence in molecular sequence data can con-found phylogenetic analysis. We agree with Hillis andWiens (2000:12) that phylogenetic error caused by con-vergence is probably “of little consequence (relative tosampling error, for instance), regardless of data type”


but feel it is relevant to the discussion to point out thatphylogenetic analyses of both molecular and morpholog-ical data can suffer from these problems, and methodsand practices exist for ameliorating these problems inboth types of analyses. The argument that morpholog-ical data suffer from excessive homoplasy in the formof convergence is simply a straw man (Wiens et al.,2003).

A CRITICAL ROLE OF MORPHOLOGY (AND FOSSILS)REVISITED

Phylogeny Reconstruction of Extinct Taxa

Molecular data cannot reconstruct the phylogenetic re-lationships of extinct taxa, except for rare cases involvingrecently extinct forms (Cooper et al., 2001). Morphologi-cal data provide the only means to understand an extincttaxon’s evolutionary history. This is no trivial matter con-sidering that nearly every species that has ever evolvedis now extinct (Novacek and Wheeler, 1992). Fossil taxa,therefore, comprise the vast majority of the branches andtwigs on the Tree of Life. The importance and relevanceof phylogenetic hypotheses for these fossil taxa cannot beunderstated because our knowledge of the rates of mor-phological evolution and the acquisition of novel bau-plans, the frequency of mass extinctions and the impactof global biodiversity crises, the calibration of molecu-lar divergence estimates, as well as the origination ofmodern clades of taxa, all rely on data from these fossilphylogenies.

Considering the importance of fossil data, it is neces-sary to ask the question “Does SEA’s proposed role ofmorphology in systematics and phylogenetics allow forphylogeny reconstruction of fossil taxa?” The answer tothis is “no.” With ambiguity and relevance of morpholog-ical characters determined from congruence with molec-ular phylogenies, phylogenetic hypotheses for groupswith no living representatives (e.g., trilobites, blastoidechinoderms, pterosaurs, etc.) would of necessity remainunexplored because no molecular phylogeny can existfrom which “unambiguous” morphological characterscould be obtained.

Even the phylogenies of fossil taxa with extant rela-tives would remain unresolved because not all taxa cansimply be inserted into a molecular phylogeny. If thesefossil taxa do not belong within the crown clade properthan they would optimize, given the available unam-biguous morphological data, as a basal polytomy withall other extant members of that crown group. At thispoint, there would be no objective means of resolvingtheir position among other stem members of the ex-tant clade. Moreover, there is no reason to expect theextant members of a group would provide morpholog-ical homologies relevant to elucidating the phylogeniesof their fossil relatives. Resolving the phylogenetic re-lationships within sauropod dinosaurs would be aidedlittle by the discovery of morphological characters ofunambiguous homology assessment in extant aviandinosaurs.

Morphology Bridges Extinct and Extant Taxa

Do cases exist where analysis of morphological datacould be especially useful in elucidating the phylogenyof a group? As described by previous authors (Donoghueet al., 1989; Doyle, 1998; Smith, 1998; Wills et al., 1998),the most likely situation where morphological analysesmay play an important role in phylogeny reconstruc-tion would be when dealing with clades characterized bydeep divergences and rapid radiations. Doyle (1998:456)states that this is “a situation in which molecular data areleast likely to be reliable, because of the small numberof molecular synapomorphies between nodes, the possi-bility that these were erased by later changes (multiplehits), and the related problem of long-branch attraction.”Several cases in which these problems may arise are inanalyses of the higher-level relationships of amniotes(Gauthier et al., 1988), seed plants (Doyle, 1998), echi-noids and ophiuroids (Smith, 1998), arthropods (Willset al., 1998), and metazoans (Rokas et al., 2003).

Rokas et al. (2003: fig. 6) clearly illustrated why theeffects of homoplasy and subsitutional saturation areparticularly problematic for molecular analyses dealingwith taxa characterized by deep divergences and rapidradiations. They related this problem to higher-levelmetazoan relationships by pointing to work conductedon 18S rDNA by Phillipe et al. (1994), which demon-strates that splits in the metazoan tree separated by lessthan 40 million years cannot be confidently resolved. Re-cent simulation studies (Levinton et al., 2004) have con-firmed the relationship between the ratio of radiationtime/post-radiation time and accuracy of phylogeneticinference. Rokas et al. (2003) also aptly characterizedmore fundamental and philosophical problems relatingto these situations (2003:355–356):

Phylogenetic reconstruction of deeply diverged taxa, such as theearly branching metazoans, rests on the presence and availabilityof highly conserved genes that can be accurately identified as or-thologs. Such strong conservation of amino acid sequences over longtime periods can only be explained by invoking the action of strongpurifying selection; in the absence of selection, genetic drift is ex-pected to lead to wide divergence between orthologs, making themindiscernible after a few million generations (Gillespie 1991). Despitethis realization, the assumption behind every phylogenetic analysisis that variation between sequences is neutral. Indeed, models ofsequence evolution and phylogenetic reconstruction algorithms donot specifically take natural selection into account (Swofford et al.1996).

The primary reason that morphological analyses maybe capable of resolving the relationships of such groupsis that morphological analyses can incorporate data fromfossil taxa. Fossil taxa can preserve critical combinationsof synapomorphy and plesiomorphy that can overturnhypotheses misled by homoplasy (Gauthier et al., 1988;Donoghue et al., 1989).

Although increased taxon sampling of extant organ-isms can also increase phylogenetic accuracy via thesetwo means (in cases involving either the analysis ofmolecular or morphological data), fossil taxa are likely tobe much more beneficial as divergence time of the groupin question increases, and the time period within which


the group radiated decreases. Indeed, Kim (1996:363)concluded that “if taxa are added to counter inconsis-tency problems, the added taxa should have low ratesof change and be close to the common ancestor of theclade” (see also Huelsenbeck, 1991; Poe and Swofford,1999; and Poe, 2003).

Taxon Sampling and the Importance of Fossils

Aside from reconstructing relationships of fossil taxa,utilizing morphological data for phylogenetic analysesoffers a considerable advantage in its ability to increasetaxon sampling. This advantage is often best employedthrough the addition of fossil taxa, which may possessunique combinations of characters not present in extanttaxa (Gauthier et al., 1988; Donoghue et al., 1989). Severalempirical studies of widely disparate taxa have shownthat the addition of fossil taxa can impact tree topology(Gauthier et al., 1988; Donoghue et al., 1989; Smith, 1998;Wills et al., 1998). SEA comment on these possible ad-vantages, stating that: “morphological data from fossiltaxa can increase taxon sampling in a way not possiblefor sequence data,” but dismiss them on relatively un-substantiated grounds (p. 543):

Although it can be demonstrated that adding taxa with unique com-binations of characters can alter a topology (Doyle and Donoghue,1987; Smith, 1994, 1998) and sometimes give slightly increased levelsof support (Lecointre et al., 1993; Baker et al., 1998; Smith, 1998), thisis not the same as increasing the accuracy of a given estimate. It is un-clear whether breaking up long branches by dense taxon sampling(Gauthier et al., 1988; Graybeal, 1998) using morphological data onthe basis of reduced cost or specimen accessibility (Hillis and Wiens,2000) will lead to a more accurate assessment of phylogeny . . . giventhat these data will suffer from problems discussed above (Figs. 1c,1d) plus the additional problem of large amounts of missing data.

The first part of this argument presumes an objectivemeans of assessing phylogenetic accuracy (not support)for empirical data. This assumes knowledge of evolu-tionary history unattainable by researchers. Throughouttheir paper, SEA appear to proxy true phylogenetic accu-racy through comparison with currently accepted molec-ular analyses. We caution against such a practice underthese circumstances for the simple reason that most phy-logeny estimates are typically accepted as being accu-rate because of their comprehensiveness and high levelsof support. It would seem logically flawed to utilize levelsof support to consider one estimate of phylogeny as “ac-curate,” but then divorce support and accuracy in orderto dismiss another estimate of phylogeny, as SEA appearto do in the above statement.

SEA (p. 543) question “the beneficial effect of densesampling in the context of fossils,” considering that “inthe context of recent molecular analyses (Mathews andDonoghue, 1999; Qiu et al., 1999; Soltis et al., 1999;Barkman et al., 2000; Bowe et al., 2000; Chaw et al.,2000; Graham and Olmstead, 2000), earlier studies basedon morphological data and dense sampling of fossils(Crane, 1985; Doyle and Donoghue, 1986, 1987, 1992;Loconte and Stevenson, 1990; Nixon et al., 1994) havenow been recognized as being inaccurate estimates ofphylogeny.” SEA do not point out that earlier stud-

ies of amniote relationships based on morphologicaldata and dense sampling of fossil taxa (Gauthier et al.,1988; Donoghue et al., 1989) were initially contradicted(Hedges et al., 1990), but ultimately validated (Hedges,1994), by phylogenetic analysis of molecular data. Fur-thermore, early molecular estimates of seed plant phy-logeny, although consistent in rejecting the “anthophytehypothesis,” still differed significantly in topology fromeach other (Doyle, 1998).

As far as accurate phylogeny reconstruction is con-cerned, the value of including morphological data (in theform of fossil or extant taxa) is indeed difficult, if not im-possible, to determine a priori. We are in agreement withSEA that more studies, utilizing both simulation and em-pirical data, should be conducted to specifically addressand examine these conditions. However, the current lackof these studies is not sufficient grounds for utterly dis-missing the possibility that the addition of fossil taxa canresult in more accurate phylogeny reconstruction. Also,given that increased taxon sampling within a mono-phyletic group increases the accuracy of reconstructingthat clade’s phylogeny (Rannala et al., 1998), we sug-gest fossil taxa may be critical for accurate phylogenyreconstruction of crown groups with few extant mem-bers (Smith, 1998). We recommend the advice of Gauthieret al. (1988:105): “While fossils may be important in phy-logenetic inference only under certain conditions, thereis no compelling reason to prejudge their contribution.We urge systematists to evaluate fairly all of the availableevidence.”

SEA (p. 543) also claim one drawback of increasedtaxon sampling to be “the potential decreased numberof unambiguous characters as more taxa are added to astudy,” citing a decrease in the number of unambiguousmorphological characters observed as more taxa wereadded to a phylogenetic analysis of Strobilanthes. Specif-ically, the original number of 32 discrete morphologicalcharacters coded for 66 taxa in an analysis by Carineand Scotland (2002) was reduced to 12 characters uponthe addition of 22 taxa by Moylan et al. (unpublished).This example does not, as SEA (p. 543) suggest, demon-strate that adding taxa to a study may reduce the numberof unambiguous characters. The 20 ambiguous characterswere already present in the initial analysis, but were notrecognized as such. This example instead demonstrateshow increased taxon sampling can improve a data set byimproving character conceptualization and knowledgeof the range of variation. Moreover, the 20 characterswere not ambiguous due to poor homology assessment,but because they were no longer discrete (SEA, p. 543).Though we do not presume to judge the character se-lection methodology employed by Carine and Scotland(2002) and Moylan et al. (unpublished), it is worth notethat a growing body of work has been directed at utiliz-ing quantitative and continuous data in phylogeny re-construction (Swiderski et al., 1998; Polly, 2001; Wiens,2001; MacLeod and Forey, 2002; Polly, 2003a, 2003b). Apriori dismissal of these characters ignores their potentialas phylogenetically informative data and artificially re-duces the scope of variation that morphological data can


offer to systematics. This example furthermore serves asa caution to all researchers to be careful in character selec-tion and conjectures of primary homology (Rieppel andKearney, 2002), as well as taxon sampling (Hillis, 1998).

CLOSING POINTS

But morphology is a much more complex subject than it at firstappears (Darwin, 1872:584)

Charles Darwin wrote these words in reference toa paper by the eminent zoologist E. Ray Lankester, inwhich several important terms in evolutionary biologywere coined, including the still-used homoplasy and ho-moplastic (Lankester, 1870:39). Lankester distinguishedhomoplastic structures from homogenous structures bydefining homogeny in terms of reference to a commonancestor (Lankester, 1870:36), whereas homoplasywould comprise “all cases of close resemblance ofform which are not traceable to homogeny, all detailsof agreement not homogenous, in structures which arebroadly homogenous, as well as in structures having nogenetic affinity” (Lankester, 1870:41). It was in this waythat Lankester parsed out homoplasy from a broaderset of “cases which have all been equally ranked bynaturalists as homologous” (Darwin, 1872:584) andmade substantially clearer the primacy of commonancestry in characterizing homology.

In the years since Darwin’s and Lankester’s statementseven more progress has been made to this end, anda methodologically rigorous framework now exists insystematics that allows homology and homoplasy to bedistinguished from each other in a relatively clear andscientifically testable manner. We view Darwin’s lamentof morphology’s complexity not as a cry of despair, butas a source of elation, an intrinsic benefit of morphologyand a spur to action and continued research. It is pre-cisely the complexity of morphology that allows us todissect underlying patterns of relationship, which is inpart why we take particular issue with SEA’s (p. 545)final claim that: “One reason why morphology is be-ing superseded by DNA data for phylogenetic studiesis because much of the useful morphological diversityhas already been scrutinized.” Even a cursory reviewof many current morphological- and phylogenetic-basedresearch programs and journals would reveal that a greatdeal of the useful morphological diversity of the earth’sbiota, present and past, has not been “scrutinized.” Todeny this, and instead make the blanket statement im-plied by SEA that we are “all out of morphology,” wouldbe to deny the impact of new specimens (recent and fos-sil), new methods and technology (CT scanning, electronmicroscopy, histological methods, staining/preparationmethods), and, perhaps most importantly, the impact ofnew researchers.

In their paper, SEA attempted to “explore the paradoxof why morphological data are currently utilized less forphylogeny reconstruction than are DNA sequence data,whereas most of what we know about phylogeny stemsfrom classifications founded on morphological data” (p.539). SEA argue that the current prevalence of molecu-

lar studies is due to inherent problems in the analysisof morphological data. In contrast, we view this shift asa logical step towards utilizing and exploring the his-torically untapped phylogenetic information content ofmolecular data. However, it is worrisome if this shift isdone at the expense and dismissal of further morpholog-ical scrutiny and revision (Hillis and Wiens, 2000).

We agree with SEA (pp. 545, 542) that a goal of cur-rent phylogenetic analyses of morphological data shouldbe more “rigorous and critical anatomical studies” ofmorphological characters, as well as “explicit criteria forcharacter selection” (see also Rieppel and Kearney, 2002;Poe and Wiens, 2000). However, we do not find relegat-ing morphology to a subservient position is the meansto achieve this. We maintain that the majority of phylo-genetically informative morphological diversity has notbeen scrutinized, in addition to pointing out that a greatdeal of morphological diversity needs to be rescrutinized.Many thorough, critical, and novel phylogenetic stud-ies of morphological data await and will be crucial inreconstructing the phylogeny of the earth’s biota.

ACKNOWLEDGEMENTS

We gratefully acknowledge members of the University of IowaPaleontology Discussion Group for helpful conversation and com-ments. We also thank John Wiens for discussion, support, and advice,as well as for providing an advance copy of his manuscript. Several dis-cussions with Chris Brochu and Jonathan Adrain were also critical tothe development of this article. We thank the patrons and staff of Joe’sPlace, Iowa City, for tolerating said discussions. This article greatlybenefited from reviews and comments by Norm MacLeod, Rod Page,David Polly, and Andrew Smith. Any mistakes or misinterpretationsremain our own. NDS was supported by the Wilkins Fund and NSFgrant OPP-0229698 to W. R. Hammer, and AHT was supported by theLittlefield Fund and the Evolving Earth Foundation during this work.

REFERENCES

Baker, R. H., Y. Xiaobo, and R. DeSalle. 1998. Assessing the relativecontribution of molecular and morphological characters in simulta-neous analysis trees. Mol. Phylogenet. Evol. 9:427–436.

Barkman, T. J., G. Chenery, J. R. McNeal, J. Lyons-Weiler, W. J. Ellisens,G. Moore, A. D. Wolfe, and C. W. Depamphilis. 2000. Independentand combined analyses of sequences from all three genomic com-partments converge on the root of flowering plant phylogeny. Proc.Natl. Acad. Sci. USA 97:13166–13171.

Baum, D. A., K. J. Sytsma, and P. C. Hoch. 1994. A phylogenetic anal-ysis of Epilobium (Onagraceae) based on nuclear ribosomal DNAsequences. Syst. Bot. 19:363–388.

Bowe, L. M., G. Coat, and C. W. Depamphilis. 2000. Phylogenyof seed plants based on all three genomic compartments: Extantgymnosperms are monophyletic and Gnetales’ closest relatives areconifers. Proc. Natl. Acad. Sci. USA 97:4092–4097.

Bremer, B., R. K. Jansen, B. Oxelman, M. Backland, H. Lantz, and K.-J.Kim. 1999. More characters or more taxa for a robust phylogeny—Case study from the coffee family (Rubiaceae). Syst. Biol. 48:413–435.

Bremer, K. 1988. The limits of amino-acid sequence data in angiospermphylogenetic reconstruction. Evolution 42:795–803.

Bull, J. J., M. R. Badgett, H. A. Wichman, J. P. Huelsenbeck, D. M. Hillis,A. Gulati, C. Ho, and I. J. Molineux. 1997. Exceptional convergentevolution in a virus. Genetics 147:1497–1507.

Carine, M. A., and R. W. Scotland. 2002. Classification of the Strobi-lanthinae (Acanthaceae): Trying to classify the unclassifiable? Taxon51:259–279.

Charleston, M. A., M. D. Hendy, and D. Penny. 1994. The effects of se-quence length, tree toplogy, and number of taxa on the performanceof phylogenetic methods. J. Comp. Biol. 1:133–151.


Chaw, S.-M., C. L. Parkinson, Y. Cheng, T. M. Vincent, and J. D. Palmer.2000. Seed plant phylogeny inferred from all three genomes: Mono-phyly of extant gymnosperms and origin of Gnetales and conifers.Proc. Natl. Acad. Sci. USA:4086–4091.

Cooper, A., C. Lalueza-Fox, S. Anderson, A. Rambaut, J. Austin, andR. Ward. 2001. Complete mitochondrial genome sequences of twoextinct moas clarify ratite evolution. Nature 409:704–707.

Crane, P. R. 1985. Phylogenetic analysis of seed plants and the originof angiosperms. Ann. Mo. Bot. Gard. 72:716–793.

Cunningham, C. W., K. Jeng, J. Husti, M. Badgett, I. J. Molineux, D. M.Hillis, and J. J. Bull. 1997. Parallel molecular evolution of deletionsand nonsense mutations in bacteriophage T7. Mol. Biol. Evol. 14:113–116.

Darwin, C. 1872. On the origin of species by means of natural selection;or the preservation of favoured races in the struggle for existence,6th edition. Random House Inc., New York (1998 printing).

Donoghue, M. J., J. A. Doyle, J. Gauthier, A. G. Kluge, and T. Rowe.1989. The importance of fossils in phylogeny reconstruction. Ann.Rev. Ecol. Syst. 20:431–460.

Doyle, J. A. 1996. Seed plant phylogeny and the relationships of Gne-tales. Int. J. Plant Sci. 157(suppl.):S3–S39.

Doyle, J. A. 1998. Molecules, morphology, fossils, and the relationshipof angiosperms and Gnetales. Mol. Phylogenet. Evol. 9:448–462.

Doyle, J. A., and M. J. Donoghue. 1986. Seed plant phylogeny and theorigin of angiosperms: An experimental cladistic approach. Bot. Rev.52:321–431.

Doyle, J. A., and M. J. Donoghue. 1987. The importance of fossils in elu-cidating seed plant phylogeny and macroevolution. Rev. Paleobot.Palynol. 50:63–95.

Doyle, J. A., and M. J. Donoghue. 1992. Fossils and seed plant phy-logeny reanalyzed. Brittonia 44:89–106.

Gauthier, J., A. G. Kluge, and T. Rowe. 1988. Amniote phylogeny andthe importance of fossils. Cladistics 4:105–209.

Gillespie, J. H. 1991. The causes of molecular evolution. Oxford Uni-versity Press, Oxford.

Givnish, T. J., and K. J. Sytsma. 1997. Consistency, characters, and thelikelihood of correct phylogenetic inference. Mol. Phylogenet. Evol.7:320–330.

Goldman, N. 1998. Effects of sequence alignment procedures on esti-mates of phylogeny. BioEssays 20:287–290.

Graham, S. W., and R. G. Olmstead. 2000. Utility of 17 chloroplastgenes for inferring the phylogeny of basal angiosperms. Am. J. Bot.87:1712–1730.

Graybeal, A. 1998. Is it better to add taxa or characters to a difficultphylogenetic problem? Syst. Biol. 48:9–17.

Hedges, S. B. 1994. Molecular evidence for the origin of birds. Proc.Natl. Acad. Sci. USA 91:2621–2624.

Hedges, S. B., and Maxson, L. R. 1996. Re: Molecules and morphologyin amniote phylogeny. Mol. Phylogenet. Evol. 6:312–314.

Hedges, S. B., K. D. Moberg, and L. R. Maxson. 1990. Tetrapod phy-logeny inferred from 18S and 28S ribosomal RNA sequences and areview of the evidence for amniote relationships. Mol. Biol. Evol.7:607–633.

Hedges, S. B., Sibley, C. G. 1994. Molecules vs. morphology in avianevolution: The case of the “pelecaniform” birds. Proc. Natl. Acad.Sci. USA 91:9861–9865.

Hillis, D. M. 1996. Inferring complex phylogenies. Nature 383:140–141.Hillis, D. M. 1998. Taxonomic sampling, phylogenetic accuracy, and

investigator bias. Syst. Biol. 47:3–8.Hillis, D. M., J. P. Huelsenbeck, and C. W. Cunningham. 1994a. Appli-

cation and accuracy of molecular phylogenies. Science 264:671–677.Hillis, D. M., J. P. Huelsenbeck, and D. L. Swofford. 1994b. Hobgoblin

of phylogenetics? Nature 369:363–364.Hillis, D. M., and J. J. Wiens. 2000. Molecules versus morphology in

systematics. Pages 1–19 in Phylogenetic analysis of morphologicaldata (J. J. Wiens, ed.). Smithsonian Institution Press, Washington,DC.

Huelsenbeck, J. P. 1991. When are fossils better than extant taxa inphylogenetic analysis? Syst. Zool. 40:458–469.

Huelsenbeck, J. P., and D. M. Hillis. 1993. Success of phylogenetic meth-ods in the four-taxon case. Syst. Biol. 42:247–264.

Jenner, R. A. 2004. Accepting partnership by submission? Morpholog-ical phylogenetics in a molecular millennium. Syst. Biol. 53:333–342.

Kim, J., and M. A. Burgman. Accuracy of phylogenetic estima-tion methods under unequal evolutionary rates. Evolution 42:596–602.

Lankester, E. R. 1870. On the use of the term homology in modernzoology, and the distinction between homogenetic and homoplasticagreements. Ann. Mag. Nat. Hist. 6:34–43.

Lecointre, G., H. Phillipe, H. L. V. Le, and H. Le Guyader. 1993. Speciessampling has a major impact on phylogenetic inference. Mol. Phylo-genet. Evol. 2:205–224.

Levinton, J, L. Dubb, and G. A. Wray. 2004. Simulations of evolutionaryradiations and their application to understanding the probability ofa Cambrian explosion. J. Paleontol. 78:31–38.

Loconte, H., and D. W. Stevenson. 1990. Cladistics of the spermato-phyte. Brittonia 42:197–211.

MacLeod, N., and P. L. Forey. 2002. Morphology, shape, and phylogeny.Taylor & Francis Group, London.

Mathews, S., and M. J. Donoghue. 1999. The root of angiosperm phy-logeny inferred from duplicate phytochrome genes. Science 282:947–950.

Mindell, D. P. 1991. Aligning DNA sequences: Homology and phy-logenetic weighting. Pages 73–89 in Phylogenetic analysis of DNAsequences (M. M. Miyamoto and J. Cracraft, eds.). Oxford UniversityPress, New York.

Morrison, D. A., and J. T. Ellis. 1997. Effects of nucleotide sequencealignment on phylogeny estimation: A case study of 18S rDNAs ofApicomplexa. Mol. Biol. Evol. 14:428–441.

Nei, M., F. Tajima, and Y. Tanteno. 1983. Accuracy of estimated phy-logenetic trees from molecular data. II. Gene frequency data. J. Mol.Evol. 19:153–170.

Nixon, K. C., W. L. Crepet, D. Stevenson, and E. M. Friis. 1994. A re-evaluation of seed plant phylogeny. Ann. Mo. Bot. Gard. 81:484–533.

Novacek, M. J., and Q. D. Wheeler. 1992. Extinct taxa—accounting for99.9990. . . % of the earth’s biota. Pages 1—16 in Extinction and phy-logeny (M. J. Novacek, and Q. D. Wheeler, eds.), Columbia UniversityPress, New York.

Penny, D., and M. D. Hendy. 1985. The use of tree comparison metrics.Syst. Zool. 34:75–82.

Phillipe, H., A. Chenuil, and A. Adoutte. 1994. Can the Cambrian ex-plosion be inferred through molecular phylogeny? Development 120(Suppl):15–25.

Poe, S. 1998. Sensitivity of phylogeny estimation to taxonomic sam-pling. Syst. Biol. 47:18–31.

Poe, S. 2003. Evaluation of the strategy of long-branch subdivision toimprove the accuracy of phylogenetic methods. Syst. Biol. 52:423–428.

Poe, S., and D. L. Swofford. 1999. Taxon sampling revisited. Nature398:300–301.

Poe, S., and J. J. Wiens. 2000. Character selection and the methodology ofmorphological phylogenetics. Pages 20–36 in Phylogenetic analysisof morphological data (J. J. Wiens, ed.). Smithsonian Institution Press,Washington, DC.

Polly, P. D. 2001. On morphological clocks and paleophylogeography:Towards a timescale for Sorex hybrid zones. Genetica 112/113:339–357.

Polly, P. D. 2003a. Paleophylogeography of Sorex araneus: Molar shapeas a morphological marker for fossil shrews. Mammalia 68:233–243.

Polly, P. D. 2003b. Paleophylogeography: The tempo of geographic dif-ferentiation in marmots (Marmota). J. Mammal. 84:369–384.

Qui, Y.-L., J. Lee, F. Bernasconi-Quadroni, D. E. Soltis, P. S Soltis, M.Zaner, E. A. Zimmer, Z. Chen, V. Savolainen, and M. W. Chase. 1999.The earlierst angiosperms. Nature 402:404–407.

Rannala, B., J. P. Huelsenbeck, Z. Yang, and R. Nielsen. 1998. Taxonsampling and the accuracy of large phylogenies. Syst. Biol. 47:702–710.

Rieppel, O., and Kearney, M. 2002. Similarity. Biol. J. Linn. Soc. 75:59–82.Robinson, D. F., and L. R. Foulds. 1981. Comparison of phylogenetic

trees. Math. Biosci. 53:131–147.Rohlf, F. J., and M. C. Wooten. 1988. Evaluation of the re-

stricted maximum-likelihood method for estimating phylogenetictrees using simulated allele frequency data. Evolution 42:581–595.


Rokas, A., N. King, J. Finnerty, and S. B. Carroll. 2003. Conflicting phy-logenetic signals at the base of the metazoan tree. Evol. Dev. 5:346–359.

Rosenberg, M. S., and S. Kumar. 2001. Incomplete taxon sampling isnot a problem for phylogenetic inference. Proc. Natl. Acad. Sci. USA98:10751–10756.

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

Sanderson, M. J., and M. J. Donoghue. 1996. The relationship betweenhomoplasy and confidence in a phylogenetic tree. Pages 67–89 in Ho-moplasy: The recurrence of similarity in evolution (M. J. Sandersonand L. Hufford, eds.). Academic Press, San Diego, California.

Scotland, R. W., R. G. Olmstead, and J. R. Bennett. 2003. Phylogenyreconstruction: The role of morphology. Syst. Biol. 52:539–548.

Shaffer, H. B., P. Meylan, and M. L. McKnight. 1997. Tests of turtle phy-logeny: Molecular, morphological, and paleontological approaches.Syst. Biol. 46:235–268.

Simmons, M. P., and H. Ochoterena. 2000. Gaps as characters insequence-based phylogenetic analyses. Syst. Biol. 49:369–381.

Smith, A. B. 1994. Systematics and the fossil record: Documenting evo-lutionary patterns. Blackwell Scientific, Oxford, U.K.

Smith, A. B. 1998. What does paleontology contribute to systematics ina molecular world? Mol. Phylogenet. Evol. 9:437–447.

Soltis, P. S., D. E. Soltis, and M. W. Chase. 1999. Angiosperm phylogeny.Nature 402:402–404.

Swiderski, D. L., M. L. Zelditch, and W. L. Fink. 1998. Why morphomet-rics is not special: Coding quantitative data for phylogenetic analysis.Syst. Biol. 47:508–519.

Swofford, D. L., G. J. Olsen, P. J. Waddell, and D. M. Hillis. 1996. Phy-logenetic inference. Pages 407–514 in Molecular systematics, 2ndedition (D. M. Hillis, C. Moritz, and B. K. Mable, eds.). Sinauer,Sunderland, Massachusetts.

Wagner, G. P. 2001. The character concept in evolutionary biology. Aca-demic Press, San Diego.

Wiens, J. J. 2000. Phylogenetic analysis of morphological data. Smith-sonian Institution Press, Washington, DC.

Wiens, J. J. 2004. The role of morphological data in phylogeny recon-struction: A reply to Scotland et al. (2003). Syst. Biol. 53:653–661.

Wiens, J. J., P. T. Chippindale, and D. M. Hillis. 2003. When are phylo-genetic analyses misled by convergence? A case study in Texas cavesalamanders. Syst. Biol. 52:501–514.

Wills, M. A., D. E. G. Briggs, R. A. Fortey, M. Wilkinson, and P. H. A.Sneath. 1998. An Arthropod phylogeny based on fossil and re-cent taxa. Pages 33–106 in Arthropod fossils and phylogeny (G. D.Edgecombe ed.). Columbia University Press, New York.

First submitted 17 March 2004; reviews returned 14 May 2004;final acceptance 8 June 2004

Associate Editor: Norm Macleod

Documents

Smith & Turner 2005