Sumida & Modesto, 2001

586

AMER. ZOOL., 41:586–597 (2001)

A Phylogenetic Perspective on Locomotory Strategies in Early Amniotes1

STUART S. SUMIDA2,* AND SEAN MODESTO3,†*Department of Biology, California State University San Bernardino, 5500 University Parkway,

San Bernardino, California 92407†Bernard Price Institute for Palaeontological Research,

University of the Witwatersrand, Johannesburg, PO Wits, 2050, South Africa

SYNOPSIS. Past approaches to understanding the evolution of locomotory strate-gies among Paleozoic amniotes (‘‘primitive reptiles’’ of previous parlance) havebeen influenced by preservational bias: early occurrences of some amniote taxawere used to polarize the acquisition or development of locomotory structuresamong the earliest amniotes. Using a phylogeny representing the current consensusin the literature, we investigate the major locomotory strategies that have beenposited for Paleozoic amniotes (basal synapsids on one hand and early reptiles onthe other) by optimizing the major locomotory styles identified for these taxa ontothe consensus tree, in order to present an overview of the pattern of evolution oflocomotory strategies inherited and adopted by various amniote lineages.

INTRODUCTION

Understanding of the earliest amnioteshas been a central pursuit of vertebrate pa-leontologists over the past century (e.g.,Case, 1911; Olson, 1947; Parrington, 1958;Vaughn, 1962; Carroll, 1969a, b, 1995;Gauthier et al., 1988; Laurin and Reisz,1997, 1999). Until Reisz (1980, 1986)pointed out that Synapsida should be con-sidered as a basal amniote lineage distinctfrom reptiles, ‘‘origin of amniotes’’ wassynonymous with ‘‘origin of reptiles.’’ Themost influential of these earlier studies werethose of Carroll (1969a, b, 1970) whereinhe proposed the morphotype of a hypothet-ical reptilian ancestor, which he foundamongst gephyrostegid anthracosaurian am-phibians. Notably, these studies wereamongst the last to address the question ofamniote origins without a cladistic context.This approach was characterized by a num-ber of factors. First, the earliest known rep-tiles were assumed to be the most primitive

1 From the symposium Beyond Reconstruction: Us-ing Phylogenies to Test Hypotheses About VertebrateEvolution presented at the Annual Meeting of the So-ciety for Integrative and Comparative Biology, 4–8January 2000, at Atlanta, Georgia.

2 E-mail: [email protected] Current address: Department of Paleobiology,

Royal Ontario Museum, 100 Queen’s Park, Toronto,Ontario M5S 2C6, Canada;E-mail: [email protected]

reptiles. As protorothyridid reptiles (‘‘pro-torothyrid reptiles’’ of Carroll, 1969a, b,1970) were known from middle Pennsyl-vanian sediments (Carroll and Baird, 1972),they were taken as the model of the prim-itive reptilian morphotype. Second, taxathat departed from the small, gracile, insec-tivorous bauplan exhibited by protorothyr-idids were generally considered to be morederived (such as in the case of captorhinidreptiles and many taxa now assigned to Par-areptilia) or aberrant (such as in the case ofgroups lumped into ‘‘Cotylosauria’’). Last,the ‘‘search for a reptilian ancestor’’ wasexactly that, a search for an ancestor. Assuch, it endeavored to define successivelymore primitive taxa appropriate to provid-ing a structural origin for reptiles. Althoughthis provided a taxon-to-taxon path withoutmorphological reversals, it also resulted inan ancestral taxon that shared only primi-tive features with any basal amniote group.

Recent phylogenetic work

By the mid-1970s, the advent of cladisticmethods of phylogenetic analysis began togain popularity in vertebrate paleontologi-cal studies. However it was not until thework of Gauthier et al. (1988) that the firsttruly global consideration of early amniotephylogeny utilized that methodology inconjunction with commercially-availablemicrocomputer software (PAUP 2.4) for

587EARLY AMNIOTE LOCOMOTOR STRATEGIES

analysis of Paleozoic tetrapod interrelation-ships. As such, this study became a bench-mark for subsequent investigators. Gauthieret al. (1988) corroborated earlier phyloge-netic work that had focused on more re-stricted branches of the amniote tree (e.g.,Heaton, 1980; Reisz, 1981; Heaton andReisz, 1986), and they provided the stron-gest evidence to date for their identificationof the closest fossil relatives of amniotes(diadectomorphs and Solendonsaurus).More importantly, Gauthier et al. (1988)were able to draw attention to several for-merly problematic amniote groups by iden-tifying an unsuspected evolutionary radia-tion of amniotes that had no modern rep-resentatives. Interestingly, Gauthier et al.(1988) had little faith that their enigmaticassemblage would sustain careful scrutinyby later workers, and christened the groupwith the informal name ‘‘parareptiles.’’Reisz and Laurin (1991) later proposed thatthe closest relatives of turtles were to befound among these ‘‘parareptiles’’ (specifi-cally: procolophonoids), thereby bringingtaxonomic legitimacy to the group. How-ever, a detailed analysis did not appear untilseveral years later (Laurin and Reisz, 1995).In the meantime, Lee (1993) also an-nounced that turtles had affinities with‘‘parareptiles,’’ but regarded pareiasaursrather than procolophonoids as the closestrelatives of testudines, a hypothesis he ex-amined in greater detail in a subsequentpublication (Lee, 1995).

The study of Laurin and Reisz (1995) isthe second global treatment of early amni-ote phylogeny. Those authors corroboratedthe validity of Gauthier et al.’s (1988) cladeof ‘‘parareptiles,’’ which they recognizedofficially by giving it Olson’s (1947) nomenParareptilia, and they provided strong evi-dence that procolophonid parareptiles werethe closest fossil relatives of turtles. In rec-ognizing turtles as parareptiles, Laurin andReisz (1995) greatly augmented the taxo-nomic composition of Reptilia. They alsoidentified mesosaurs as the only non-syn-apsid amniotes to fall outside of this en-larged concept of Reptilia, but the additionof new data and a slight modification totheir data matrix (Modesto, 1999) results inmesosaurs having the same relative position

among other reptiles as reported by Gau-thier et al. (1988).

The position of turtles remains hotly con-tested. Lee (1997) advocated an originamong pareiasaurs, whereas Rieppel anddeBraga (1996; see also deBraga and Riep-pel, 1997; Rieppel and Reisz, 1999) pro-posed the more unorthodox hypothesis thatturtles are aberrant lepidosauromorph diap-sids. The likely phylogenetic position ofturtles is not critical to our review, primar-ily because they make their first appearanceonly in the Lower Mesozoic. However, theview that turtles are parareptiles is followedhere because it appears to be the generalconsensus, and it allows us to retain thephylogenetic taxonomy that concerns earlyamniotes and has become established in therecent literature (e.g., deBraga and Reisz,1996; Modesto, 1999). It should be notedthat the larger analysis of deBraga andRieppel (1997) did not alter the relation-ships among Paleozoic taxa outlined byLaurin and Reisz (1995).

Several recent studies have examined theinterrelationships of tetrapods in globalanalyses and are crucial to considerations ofthe origins of locomotory features in am-niotes. Parenthetically, it is important topoint out that although the above survey ofstudies of reptilian origins appears criticalof Carroll’s (1969a, b, 1970) work, he hasmore recently adopted a phylogenetic sys-tematic approach wherein his proposed hy-potheses of relationship amongst basal am-niotes differ significantly from his earlierwork. In one such study (Carroll, 1995), hesuggested lepospondyls as closer relativesof amniotes than such reptiliomorph an-amniotes as seymouriamorphs and anthra-cosaurs, a view supported by Laurin andReisz (1997) and Laurin (1998).

A cladogram summarizing the currentconsensus of the interrelationships of Paleo-zoic amniotes and their nearest relatives isshown in Figure 1. By mapping character-state distributions onto this phylogeneticframework, functional hypotheses and sce-narios for taxa spanning the amphibian toamniote transition may be tested. In this re-view we examine the major locomotorystrategies exhibited by basal amniotes and

588 S. S. SUMIDA AND S. MODESTO

FIG. 1. Consensus of cotylosaurian phylogeny drawnfrom Gauthier et al. (1988), Laurin and Reisz (1995),deBraga and Reisz (1996), and Modesto (2000). Pri-mary vertebral forms (s, swollen and expanded neuralarches; n, narrow neural arches) are placed on branch-es they first appear on this cladogram; alternative sce-narios within Eureptilia result from accelerated (A) ordelayed (D) transformation options and are indicatedby reverse image capital letters. See Laurin and Reisz(1995) and Modesto (2000) for stratigraphic distribu-tions for most of these taxa. See Figure 6 for justifi-cation of plesiomorphic state for Amniota.

their anamniote sister taxon Diadectomor-pha.

PHYLOGENETIC CONTEXT: AMNIOTE

OUTGROUPS AND BASAL AMNIOTES

Schultze (1987) and Hopson (1991) haveargued convincingly that defensible phylo-genetic analyses must include, wheneverpossible, all members of a group being con-sidered. If this is impractical, it is impera-tive that basal members of any clade be em-ployed in an analysis in order to avoid bi-asing hypotheses of transformation betweenclades on highly derived, terminal taxa. Bylogical extension, a consideration of func-tional transformations must also includesuch basal taxa. Thus, we begin our func-tional survey by examining locomotory fea-tures in basal amniotes, as well as basal rep-resentatives of the tetrapod clades closest toamniotes as a necessary prerequisite for es-tablishing ancestral locomotory abilities inAmniota. Although the interrelationships ofmore distant ‘‘reptiliomorph’’ taxa contin-ues to be debated, the Diadectomorpha havebeen confirmed as the sister group of am-

niotes in a variety of recent studies and re-views (Gauthier et al., 1988; Lombard andSumida, 1992; Laurin and Reisz, 1997).Berman et al. (1992) even went so far as tosuggest that diadectomorphs share a morerecent common ancestor with synapsidsthan with any other tetrapod group, makingthem de facto amniotes. Diadectomorphsare known from the Permo-Carboniferousof both Europe and North America (Ber-man et al., 1997).

In general morphology the diadecto-morph postcranial skeleton suggests a highdegree of terrestriality (Sumida, 1997). Thebest preserved forms possess highly ex-panded neural arches that range about threeto four times the diameter of the centrumin transverse width (Sumida, 1990). Thearches can be described as ‘‘swollen’’ inthat their dorsal surfaces are conspicuouslyconvex when viewed from anterior or pos-terior aspects. The great breadth of the arch-es, together with the approximately hori-zontal orientation of the zygapophyseal fac-ets, suggest strongly that the diadectomorphvertebral column primarily facilitated lat-eral bending during locomotion. Diadecto-morphs feature accessory intervertebral ar-ticulations that would have further restrict-ed movements of the vertebral column tothe horizontal plane. Apart from some le-pospondyls (microsaurs), diadectomorphsare the only anamniotes known to exhibitalternation of neural spine height and struc-ture (Sumida, 1990). (Seymouriamorphsexhibit such alternation, but it is extremelyinconsistent in pattern.) Although the incli-nation of the zygapophyseal surfaces is ata slight angle to the horizontal plane in re-gions of alternation, it is not so extremelyexpressed as to suggest that diadectomorphsengaged in other than predominantly lateralbending during locomotion.

The appendicular elements are robustlyconstructed. Propodials (humeri and femo-ra) lack shafts and have massive proximaland distal heads. Epipodial segments mea-sure 70% or greater that of humeri and fem-ora. The diadectomorph manus and pes areshort and broad with stout digits. Together,the axial and appendicular skeletons sug-gest that although appendicular elementsare relatively longer and better developed


than in anamniotes (Sumida, 1997), the lo-comotory strategy of these animals contin-ued to accentuate the plesiomorphic, trans-verse undulatory action of the vertebral col-umn.

Synapsids

Paleozoic synapsids have been the focusof many large-scale treatments over the pastcentury and their anatomy and interrelation-ships are well known (Romer and Price,1940; Reisz, 1986). We concentrate here onthe basal, non-therapsid synapsids, knowncolloquially as ‘‘pelycosaurs.’’ These earlysynapsids span a temporal range from theLate Carboniferous through to the LatePermian. The overwhelming majority of‘‘pelycosaurs’’ are from North Americanand European deposits; a single species isknown from southern Africa. Although anamphibious habitus is thought to have aris-en in one group, Ophiacodontidae, mostearly synapsids are regarded generally tohave been highly terrestrial animals. Nar-row neural arches are found in the over-whelming majority of taxa. Although theterm ‘‘narrow’’ is never defined in phylo-genetic studies that use neural arch widthas a phylogenetic character, the term here isapplied to arches whose zygapophyses donot extend laterally beyond the vertebralcentrum. The sole synapsid in which nar-row arches are absent is the small ophia-codontid Varanosaurus, where they are ex-panded transversely and feature horizontalzygapophyseal articulating facets. The de-gree of dorsolateral expansion of the neuralarches in Varanosaurus is not as great asthat in diadectomorphs (Sumida, 1990);nonetheless, these two features of the dorsalneural arches suggest that undulatorymovements of the vertebral column wereemphasized during locomotion in this form(Sumida, 1989). Considering that Varano-saurus is nested deeply within Synapsida(Fig. 1), its locomotory strategy is best re-garded as a derived for synapsids.

In all narrow-arched synapsids the artic-ulating facets are tilted throughout the col-umn. In caseids and ophiacodontids theyare tilted an average of 30 degrees to thehorizontal, whereas in edaphosaurids andsphenacodonts the facets are tilted even fur-

ther, averaging 40 degrees (Reisz, 1986).The great angulation of the zygapophysealfacets in these taxa suggests that, in addi-tion to the plesiomorphic side-to-sidemovement, rotation of the vertebral columnwas a general feature of locomotory actionin synapsids. Tall neural spines are also acharacteristic of early synapsids. In basaltaxa (caseosaurs, varanopseids) the dorso-ventral measures of the spines are only alittle greater than their anteroposteriorlength. In the more crownward forms, theheight of the spines is at least twice theirlength; in edaphosaurids and some sphen-acodontids the neural spines are greatlyelongated and form ‘‘sails’’ on the back ofthe animal. The function of such ‘‘sails’’has been the subject of much speculationand modeling with thermodynamic controlthe favored hypothesized function (Haack,1986). The possible role of the relativelyshorter spines of closely related forms hasnot been explored. The tall neural spines ofearly synapsids, which in life would havebeen connected to neighboring spines byequally tall interspinous ligaments, mayhave served to prevent hyper-rotation of thevertebral column during locomotion.

The appendicular skeleton of basal syn-apsids is markedly less robust in construc-tion than that of diadectomorphs. The rel-ative sizes of the propodial and epipodialelements are essentially the same as inthose anamniotes. The manus and pes areshort, broad structures in large-bodied taxa,particularly caseids, ophiacodontids andlarge sphenacodontids, but they appear lon-ger and more gracile in taxa at the smallerend of the size spectrum (e.g., Varanops,Haptodus). The variance in the generalbuild of the manus and pes is presumablydue to scaling with respect to the widerange in body sizes seen among synapsids.Data from the axial skeleton suggest thatsynapsids adopted a locomotory strategythat incorporated a substantial degree of ro-tary action into vertebral column. Why syn-apsids acquired this modification of generalvertebral movement is not entirely clear, butit could be related to the slightly more elon-gated nature of the vertebral column ascompared to diadectomorphs and such rep-tiles as captorhinids (Reisz, 1986). Vara-


nosaurus represents an exception to thisstrategy in its independent adoption of thepattern described for diadectomorphs. Thelatter have relatively shorter bodies, yet thevertebral column in Varanosaurus does notappear to be relatively shorter than that ofother synapsids (Sumida, 1989). Accord-ingly, the role of vertebral rotation in syn-apsid locomotion might be better investi-gated in terms of its effect on limb exten-sion and recovery. This is a postulate thatis beyond the scope of this review; how-ever, it is interesting to note that Ritter(1996) has pointed out that lateral rotatorymovement plays an important role in limbrecovery mechanics in a variety of extantlepidosaurs.

Eureptilia

Whereas the Protorothyrididae was onceconsidered the basal stock for all amniotes(e.g., Carroll, 1969a, b, 1970) a series ofstudies (Heaton and Reisz, 1986; Lombardand Sumida, 1992; Laurin and Reisz, 1995;Sumida, 1997) now indicate that it is nestedwell within the Eureptilia (sensu Gauthieret al., 1988). Eureptilia includes the Cap-torhinidae, Protorothyrididae, Araeosceli-dia, and Neodiapsida, with the latter twotaxa forming the clade Diapsida.

The postcranial skeletons of captorhinidreptiles demonstrate clearly terrestrial fea-tures. In general, their axial and appendic-ular skeletons are robustly constructed. Ver-tebral neural arches are ‘‘swollen’’ and ex-panded laterally beyond the lateral marginsof the centrum, though not to the degreeseen in diadectomorphs. Whereas the neuralarches may measure as much as four timesthe width of the centrum in diadectomophs,those of captorhinids usually measure ap-proximately 1.2 to 2 times the width of thecentrum. Greater relative measures of theneural arch are associated with a more cau-dad position in the presacral column. Vir-tually all genera of captorhinids for whichadequate materials are known demonstratealternation of neural spine height and struc-ture. Sumida (1990) pointed out that in thisgroup zygapophyseal angle is greater in re-gions of alternation; however, the angula-tion is still significantly less than in narrow-arched synapsids, protorothyridids, and ar-

aeoscelidians (including basal diapsids). Inregions of the captorhinid column where al-ternation is either absent or muted, zyga-pophyseal surfaces are nearly parallel to thehorizontal plane. Despite the minor regionaloccurrence of zygapophyseal tilting in re-gions of extreme neural spine alternation,captorhinids still appear to have retained aplesiomorphic pattern dominated by lateralundulatory movement of the vertebral col-umn.

Although not as large as those of pely-cosaurian-grade synapsids, the appendicularelements of captorhinids demonstrate com-parable development of markers of muscu-lar development in propodial and epipodialsegments. Particularly well-developedstructures common to captorhinids, diadec-tomorphs, and synapsids include the inser-tion for the latissimus dorsi, the humeralentepicondyle for antebrachial flexors, andan extremely well developed olecranon pro-cess for attachment of triceps musculature(Sumida, 1997). Although not as large asmost diadectomorphs, it is noteworthy thatcertain larger captorhinids such as Labido-saurus, Labidosaurikos, and other moradi-saurine captorhinids do demonstrate diadec-tomorph-like proportions.

Of araeoscelidians and protorothyridids,only certain members of the former sharevertebral features and proportions in com-mon with captorhinids. Araeoscelidians arecharacterized by elongate ‘‘cervical’’ ver-tebrae, but those of the mid-dorsal regionare extremely similar to the laterally ex-panded dorsal vertebrae of captorhinids.The araeoscelids Araeoscelis and Zarcasau-rus and at least one specimen of the prim-itive diapsid Petrolacosaurus (University ofKansas Vertebrate Paleontology Collectionsspecimen 33605) possess expanded neuralarches and alternation of neural spine struc-ture. As in captorhinids, there is minimaltilt to the zygapophyseal surfaces, and whenpresent is usually less than 15 degrees inAraeoscelis and Zarcasaurus (Sumida,1990), and less than 10 degrees in Petro-lacosaurus (Reisz, 1981). Whereas somespecimens of Petrolacosaurus retain mod-erately swollen neural arches, most exam-ples depart from the pattern of conical neu-ral spines alternating with tall narrow ones


to being consistently elongate and nearlyquadrangular in section. This condition re-calls the approximate shape of the neuralspines of basal synapsids, but they are rel-atively smaller when compared that group.

All araeoscelidians for which data areavailable possess appendicular skeletonsthat are significantly more lightly built thanthose of any of the taxa described herein.Epipodials and propodials are elongate withwell-defined shafts and, when present,proximal and distal processes are not as ro-bustly developed. The femur is conspicu-ously sigmoidal. Manual and pedal ele-ments are similarly gracile and elongate. Asin certain extant lepidosaurians, the fourthdigit of the manus and pes is significantlymore elongate than the others, a conditionsuggestive of rotation of podal elements toallow enhanced lateral pushoff (Sumida,1997). Basal araeoscelidians appear to havebeen an early experiment as a small, grac-ile, lizard-like organism. However, theirvertebral structure appears to have retainedthe primitive undulatory movement of theaxial column.

Other than basal synapsids, protorothyr-idids are the only basal group of amniotesto possess conspicuously narrow neuralarches. In concert with this, their zygapo-physeal articulations are inclined between30 and 40 degrees (Carroll, 1969a, c), yettheir neural spines are relatively less tallthan in pelycosaurs. Appendicular elementsof protorothyridids are also more gracile inconstruction than those of diadectomorphs,captorhinids, or parareptiles. Only meso-saurs and araeoscelidians demonstrate morelightly built limbs than protorothyridids. Asin araeoscelidians, manual and pedal con-struction of protorothyridids is gracile andthe fourth digit of both is conspicuouslyelongate. Protorothyridids appear to havedeveloped a gracile, lizard-like habitus thatincluded significant axial rotation.

Anapsid reptiles

In concert with the ongoing debate onturtle origins, the clade of basal reptilescomprising parareptiles and mesosaurs, andnow known as Anapsida (sensu Gauthier etal., 1988; see Modesto, 1999), has receivedmuch attention in recent systematic studies

and there is considerable disagreement con-cerning the interrelationships of the constit-uent taxa. If we disregard the phylogeneticposition of turtles, the remaining disagree-ment does not bear on this review, inas-much the postcrania of the group of con-tention, Lanthanosuchidae, are not knownin adequate detail (Ivakhnenko, 1980). Thevast majority of anapsids are restricted tothe Permian, with only the procolophonidssurviving into the succeeding Mesozoic.The most basal anapsids were limited tosouthern distributions in the Pangean su-percontinent, but some advanced lineageswere restricted to northern portions of Pan-gea, whereas the more crownward taxa,specifically procolophonoids and pareia-saurs, separately attained cosmopolitan dis-tributions (Modesto, 2000).

Among the three major Paleozoic amni-ote groups, Anapsida is remarkable for abroad scale of locomotory adaptations. Themost basal anapsids, the mesosaurs, are old-est known fully aquatic reptiles. The enig-matic Eunotosaurus africanus is character-ized by a semi-rigid, turtle-like rib cage(Gow and de Klerk, 1997), one which pre-sumably necessitated a tortoise-like fashionof walking. The large, ponderous pareia-saurs also possessed a relatively short trunk,but featured the most upright limb posturesof any Paleozoic reptile group. However,most anapsids evinced a sprawling, lizard-like stance, and their skeletons are highlyindicative of terrestriality. As in diadecto-morphs, the neural arches of the vast ma-jority of anapsids are expanded transverselyand are conspicuously ‘‘swollen.’’ Howev-er, the neural arches are never wider thanthey are long, except in pareiasaurs (Figs.2, 3), where they become as greatly ex-panded as in diadectomorphs, a conver-gence suggestive of the scaling effect oflarge body size. Zygopophyseal facets areoriented roughly horizontally, which to-gether with the great breadth of the archessuggest strongly that lateral bending wasthe primary movement of the vertebral col-umn during locomotion, an interpretationthat applies to the terrestrial forms and theaquatic mesosaurs equally. In the latter,however, accessory intervertebral articula-tions are also present, which would have


FIG. 2. Single dorsal vertebra of the Late Paleozoictetrapod Diadectes demonstrating ‘‘swollen’’ construc-tion of the neural arch. (A) cranial, (B) left lateral, and(C) caudal views of a single dorsal vertebra.

FIG. 4. Dorsal vertebral structure in the ophiacodon-tid eupelycosaur Varanosaurus, the only synapsid todemonstrate expanded neural arches and alternation ofneural spine height and structure. (A) left lateral viewof two articulate dorsal vertebrae. (B) anterior view oftypical dorsal vertebra with large (tall) neural spinemorph. (C) anterior view of typical dorsal vertebrawith small (low) neural spine morph. (Modified fromSumida, 1990.)

FIG. 3. Reconstructions in left lateral view of singleor paired dorsal vertebrae of representative Late Pa-leozoic tetrapod taxa. All vertebral bodies are drawnto approximately the same size to facilitate proportion-al comparison of structures. Accompanying bar scalesall equal 1 cm except where noted. In each case, ge-neric name is accompanied by a higher order taxonom-ic description to facilitate systematic consideration ofthe element. (Partially after Sumida, 1997.)

FIG. 5. Reconstructions in dorsal view of the left fe-mur of representative Late Paleozoic tetrapod taxa. Allfemoral bodies are drawn to approximately the samelength to facilitate proportional comparison of struc-tures. In each case, generic name is accompanied by ahigher order taxonomic description to facilitate sys-tematic consideration of the element. Accompanyingbar scales all equal 1 cm. (Partially after Sumida,1997.)

further restricted the mesosaurid vertebralcolumn to movements in the horizontalplane (Modesto, 1999). The single excep-tion to the presence of expanded, swollen neural arches, and thus the inferred exag-

gerated lateral flexure of the vertebral col-umn during locomotion, is Eunotosaurus,which features a reduced number of antero-posteriorly elongated trunk vertebrae (thatarticulate with the anteroposteriorly ex-panded trunk ribs, which in turn form thecarapace-like rib cage: Gow and de Klerk,1997). The serial anatomy of the dorsalspines is not well established for most an-apsid groups, but subtle alternation of neu-ral spine height is displayed by posterior-most dorsal vertebrae in tall-spined speci-mens of Mesosaurus (Modesto, 1996).


FIG. 6. Justification for interpreting the presence ofexpanded, swollen neural arches and the locomotorystrategy they imply as the ancestral condition for am-niotes. Assuming Amniota is polymorphic with re-gards to the presence of swollen neural arches (s) ornarrow arches (n), optimization using five successiveoutgroups suggests that it is more parsimonious to re-gard swollen neural arches as the primitive conditionfor amniotes. Arrow indicates the morphology that isdecisive for Amniota in accordance with the distribu-tion of the two states on this tree. Tree topology fol-lowing Laurin and Reisz (1999).

The appendicular skeleton in general ismore gracile that that seen in synapsids anddiadectomorphs. Pareiasaurs are the soleexception. Their appendicular elementsevince the robustness seen in diadecto-morphs. The relative lengths of the propo-dials and epipodials in anapsids are approx-imately the same as in other cotylosaurs(sensu Gauthier et al., 1988), except in themesosaurian pectoral limb, where the latterelements are reduced to half the length ofthe former, a reduction common to aquaticreptiles that employ the tail as a scullingorgan. The manus and pes are generallyshort and broad but are comprised of ele-ments more gracile than the diadecto-morphs. The locomotory apparatus as awhole suggests that, as with diadecto-morphs, anapsids perpetuated the plesio-morphic undulatory action of the vertebralcolumn. The two derived locomotory strat-egies that evolved within Anapsida are thepresumably turtle-like locomotion of Eu-notosaurus africanus, where sprawlinglimbs propelled a semi-rigid body, and thesemi-erect stance and gait of pareiasaurs.

DISCUSSION

Basal amniotes demonstrate two mark-edly different patterns of vertebral mechan-

ics. Optimization of the two locomotorystrategies onto the amniote consensus tree(Fig. 1) suggests that the one accentuatingsinusoidal movement of the vertebral col-umn is plesiomorphic for amniotes. Ac-cordingly, the narrow-arched condition ofsynapsids, Eunotosaurus, Paleothyris andneodiapsids was acquired independently inthese amniote taxa. Our preferred hypoth-esis is highly sensitive to the position of thesuccessive outgroups to Amniota. We haveused the phylogenetic conclusions of Laurinand Reisz (1999), the most recent workersto examine the broader scope of tetrapodinterrelationships, to identify outgroup taxa.The relative positions of Amphibia, Wes-tlothiana, and Seymouriamorpha as recog-nized by Laurin and Reisz (1999) are crit-ical to our identification of expanded neuralarches as the ancestral condition for amni-otes (Figure 6), as character optimizationsuggests that the narrow arched conditionascribed to Solenodonsaurus is an autapo-morphy of that taxon (a basal ‘‘anthraco-saur,’’ using the prioritized phylogeneticterminology of Laurin and Reisz, 1999).

The alternative to our preferred hypoth-esis, in which narrow neural arches and thelocomotory strategy that they imply is an-cestral for amniotes, is supported in a sce-nario that is marginally less parsimonious(at 6 steps) than the hypothesis advocatedhere (5 steps). In this alternative hypothesisthe evolution of expanded presacral verte-brae in reptiles becomes complicated byambiguous optimization, leading to the pos-sibility that expanded, swollen neural arch-es arose independently in anapsids, captor-hinids, and araeosceloids. That possibilitycompetes with two equally parsimoniousinterpretations: (1) that laterally-expanded,swollen neural arches arose in the ancestralreptile and reversed separately in Paleo-thyris and in neodiapsids; or (2) they arosein the ancestral reptile and reversed in theancestral romeriid (sensu Gauthier et al.,1988), only to reappear in the ancestral ar-aeosceloid.

In the more parsimonious interpretation,in which expanded neural arches is the an-cestral condition for amniotes, only the lasttwo scenarios must be considered, but onlyin part, as the only uncertainty lies in


whether narrow arches arose in the ances-tral romeriid and reversed in araeosceloids(the accelerated transformation scenario), orwhether narrow arches arose separately inPaleothyris and neodiapsids (the delayedtransformation scenario). It is currently im-possible to choose between these latter twoevolutionary histories with regards to theevolution of locomotory strategies in rom-eriid reptiles. However, a rigorous interpre-tation of the vertebral anatomy of the EarlyPermian reptile Thuringothyris mahlendorf-fae could shed some light on this problem.A fortuitous cross-section through one pre-sacral vertebra suggests that this taxon wascharacterized by expanded neural arches(Boy and Martens, 1991). Accordingly, ifThuringothyris is indeed a protorothyrididas Boy and Martens (1991) have concluded(there is some question as to the rigor oftheir phylogenetic analysis), and if it can bedemonstrated that the taxon is a basal mem-ber of Protorothyrididae, Thuringothyriscould be decisive in terms of the optimi-zation of the two locomotory strategiesonto this part of the amniote tree, for itwould support the delayed optimizationscenario which suggests that narrow archesarose independently in (some) protorothyr-idids and in neodiapsids.

Our conclusions contrast with those ofearlier workers, primarily Carroll and hisco-workers (Carroll, 1969a, b; Carroll andBaird, 1972; Clark and Carroll, 1973), whoregarded the narrow neural arches of pro-torothyridids as representing the primitivecondition for amniotes. That hypothesis,however, was the logical conclusion of theevolutionary-systematic methodology thatwas popular with vertebrate paleontologistsprior to the advent of phylogenetic system-atics. Most taxa that were assigned to Pro-torothyrididae are Late Carboniferous inage, and this includes the oldest known am-niote, Hylonomus lyelli. The early appear-ance of this taxon and other protorothrididssuggested to evolutionary systematists thatnarrow arches were ancestral for amniotes,and that it followed logically that the ex-panded, swollen vertebrae of anapsid, cap-torhinid, and araeosceloid reptiles musthave been derived from the narrow-archedcondition. Our phylogenetic perspective

suggests that this scenario is less likely thanthe alternative, in which the condition seenin anapsids, captorhinids, and araeosceloidswas inherited from a remote, anamniote an-cestor.

Our results have interesting paleobiolog-ical implications with regards to the loco-motor strategies that were inherited oradopted by various early amniotes. It ap-pears that those amniotes that evolved nar-row neural arches and adopted a locomo-tory strategy that placed less emphasis onlateral undulation of the backbone werehighly precocious as far as the fossil recordis concerned. During the Carboniferous,which records the first 15 million years ofearly amniote history, there is only a singleamniote taxon (the diapsid genus Petrola-cosaurus of the Late Stephanian) with ex-panded neural arches; all other Carbonifer-ous amniotes are assigned either to Proto-rothyrididae or to Synapsida. Other broad-arched amniotes do not appear for manymillions of years. Captorhinids appear dur-ing the late Asselian (equivalent to the earlyor middle Wolfcampian), anapsid reptilesduring the Sakmarian (late Wolfcampian),and the synapsid Varanosaurus in the Ar-tinskian (Leonardian). In the case of the an-apsid reptiles, their late appearance in thefossil record (drawn from the observationthat members of their sister group Eurepti-lia are preserved in Westphalian strata)could be explained by an early history spentin Permo-Carboniferous Gondwana, whichrecords no continental sedimentation untilthe late Kazanian. That hypothesis is sup-ported by a biogeographic analysis of an-apsids using phylogenetic frameworks (Mo-desto, 2000). The relatively late entry intothe amniote record by captorhinids, whoseprotorothyridid relatives appear in West-phalian sediments, is harder to explain, andthe same applies to the synapsid Varano-saurus, whose narrow-arched sister taxonOphiacodon appears at the close of the Ste-phanian.

Based on the construction of the neuralarches and alternation of neural spines inmost of the taxa surveyed above, Sumida(1990, 1997) suggested that the hypotheti-cal ancestral amniotes did indeed engage inundulatory locomotion. It was further spec-


ulated that the pattern of vertebral neuralspine construction and associated transver-sospinalis musculature of basal amniotesand their immediate anamniote sister taxamust have been active in other functions inaddition to undulatory movement. It wassuggested that the transversospinalis groupalso provided postural support ipsilateral tothe contact limb during undulatory loco-motion. This hypothesis of muscular andlocomotor function was, however, based onfossil taxa and was thus was not easily test-able. Furthermore, it lacked a rigorous phy-logenetic context. Given the constraints onthe survey, it remained an untested hypoth-esis. More recently, using electromyograph-ic, kinematic, and denervation studies, Rit-ter (1995) demonstrated that although ver-tebral elements may constrain the excursionarcs of the axial column, epaxial muscula-ture did not contribute to actually generat-ing undulatory movement and was insteadrestricted to the function of postural sup-port. Additional study (Ritter, 1996) furtherdemonstrated that lateral axial bending is afunction of the hypaxial musculature, par-ticularly the external oblique and rectus ab-dominus muscles.

The distribution of this pattern amongstlepidosaurian reptiles and the stabilizingfunction of the transversospinalis muscula-ture in mammals lead Ritter (1995, 1996)to suggest that this division of labor be-tween epaxial and hypaxial musculaturewas most parsimoniously interpreted as abasal amniote characteristic. If Ritter is cor-rect, it may be suggested that the preferredhypothesis of vertebral movement patternssummarized above may now be linked to ademonstrated pattern of axial muscular ac-tivity. Given the congruence between themost parsimonious pattern of observed andinferred muscular pattern, this suggests thatthe pattern of undulatory locomotion pow-ered by hypaxial musculature with epaxialmusculature acting in postural stabilityprobably extended to Diadectomorpha aswell. Such a prediction could not have beenmade with any confidence without the phy-logenetic context outlined above.

This survey of postcranial structure inLate Paleozoic terrestrial tetrapods demon-strates the utility of phylogenetic trees in

generating simple, yet powerful interpreta-tions for patterns of locomotor evolution. Aphylogeny generated independent of func-tional scenarios allows actual testing of al-ternative hypotheses of the development oflocomotor complexes without the introduc-tion of confounding circular arguments.The suggestion of using phylogenies to mapor predict the evolution of functional com-plexes appears relatively straightforward.However, despite the general acceptance ofthe importance of phylogeny as a contextfor virtually any evolutionary endeavor orinterpretation of character change overtime, the search for ancestors based on apriori assumptions of locomotor constraintstill persists in certain cases (e.g., Feduccia,1999; Ruben and Jones, 2000; Geist andFeduccia, 2000). Fortunately, the predictiveand explanatory ability of cladograms togenerate functional predictions independentof a priori functional assumptions has ren-dered such approaches as a distinct minor-ity.

ACKNOWLEDGMENTS

We thank Donald Swiderski for invitingus to contribute to the symposium proceed-ings. Although this study views certain ofRobert Carroll’s work in a critical light, weacknowledge and appreciate the fact thathis earlier studies were done in a differentmethodological context, and we gratefullyacknowledge his significant and ongoingcontributions to our understanding of theorigin of amniotes. This study was sup-ported in part by the National ResearchFoundation of South Africa and the Cali-fornia State University Tyrannosaurus rexinitiative.

REFERENCES

Berman, D. S, S. S. Sumida, and R. E. Lombard. 1992.Reinterpretation of the temporal and occipital re-gions in Diadectes and the relationships of dia-dectomorphs. J. Paleont. 66:481–499.

Berman, D. S, S. S. Sumida, and R. E. Lombard. 1997.Biogeography of primitive amniotes. In S. S.Sumida and K. L. M. Martin (eds.), Amniote ori-gins, completing the transition to land, pp. 85–140. Academic Press, San Diego.

Boy, J. A. and T. Martens. 1991. Ein neues captorhi-nomorphes Reptile aus dem thuringischen Rotlie-gend (Unter-Perm; Ost Deutschland). Palaontogis-che Zeitschrift 65:363–389.


Carroll, R. L. 1969a. Problems of the origin of reptiles.Biol. Rev. 44:393–432.

Carroll, R. L. 1969b. Origin of reptiles. In C. Gans, A.d’A. Bellairs, and T. S. Parsons (eds.), Biology ofthe Reptilia, pp. 1–44. Academic Press, NewYork.

Carroll, R. L. 1969c. A middle Pennsylvanian captor-hinomorph, and the interrelationships of primitivereptiles. J. Paleont. 43:151–170.

Carroll, R. L. 1970. The ancestry of reptiles. Phil.Trans. Roy. Soc. London Ser. B 257:267–308.

Carroll, R. L. 1982. Early evolution of reptiles. Ann.Rev. Ecol. Syst. 13:87–109.

Carroll, R. L. 1995. Problems of the phylogenetic anal-ysis of Paleozoic choanates. Bull. Mus. Natl. Hist.Paris. 17:389–445.

Carroll, R. L. and D. Baird. 1972. Carboniferous stem-reptiles of the family Romeriidae. Bull. Mus.Comp. Zool. 143:321–364.

Case, E. C. 1911. A revision of the Cotylosauria ofNorth America. Publ. Carnegie Inst. Washington,D.C. No. 415, 122 pp.

deBraga, M. and R. R. Reisz. 1996. The Early Permianreptile Acleistorhinus pteroticus and its phyloge-netic position. J. Vert. Paleont. 16:384–395.

deBraga, M. and O. Rieppel. 1997. Reptile phylogenyand the interrelationships of turtles. Zool. J. Linn.Soc. 120:281–354.

Feduccia, A. 1999. The origin and evolution of birds.2nd ed. Yale University Press, New Haven.

Gauthier, J. A., A. Kluge, and T. Rowe. 1988. Theearly evolution of the Amniota. In M. J. Benton(ed.), The phylogeny and classification of the tet-rapods, Vol. 1, Amphibians, reptiles, birds, Sys-tematics Association Special Volume No. 35A,pp. 103–155. Clarendon Press, Oxford.

Geist, N. R. and A. Feduccia. 2000. Gravity-defyingbehaviors: Identifying models for protoaves.Amer. Zool. 40:664–675.

Gow, C. E. and W. de Klerk. 1997. First record ofEunotosaurus (Amniota: Parareptilia) from theEastern Cape. Palaeont. Afr. 34:27–31.

Haack, S. C. 1986. A thermal model of the sailbackpelycosaur. Paleobiol. 12:450–458.

Heaton, M. J. 1980. The Cotylosauria: A reconsider-ation of a group of archaic tetrapods. In A. L.Panchen (ed.), The terrestrial environment and theorigin of land vertebrates, Systematics Associa-tion Special Volume No. 15, pp. 497–551. Aca-demic Press, New York.

Heaton, M. J. and R. R. Reisz. 1986. Phylogeneticrelationships of captorhinomorph reptiles. Can. J.Earth Sci. 23:402–418.

Hopson, J. A. 1991. Systematics of the non-mamma-lian Synapsida and implications of evolution insynapsids. In H.-P. Schltze and L. Trueb (eds.),The origin of higher groups of tetrapods: Contro-versy and consensus, pp. 635–693. Cornell Uni-versity Press, Ithaca.

Ivakhnenko, M. F. 1980. Lanthanosuchids from thePermian of the East European platform. Paleont.J. 14:80–90.

Laurin, M. 1998. The importance of global parsimonyand historical bias in understanding tetrapod evo-

lution. Part I. Systematics, middle ear evolution,and jaw suspension. Ann. Sci. Nat. 1:1–42.

Laurin, M. and R. R. Reisz. 1995. A reevaluation ofearly amniote phylogeny. Zool. J. Linn. Soc. 113:165–223.

Laurin, M. and R. R. Reisz. 1997. A new perspectiveon tetrapod phylogeny. In S. S. Sumida and K. L.M. Martin (eds.), Amniote origins, completing thetransition to land, pp. 9–59. Academic Press, SanDiego.

Laurin, M. and R. R. Reisz. 1999. A new study ofSolenodonsaurus janenschi, and a reconsiderationof amniote origins and stegocephalian evolution.Can. J. Earth Sci. 36:1239–1255.

Lee, M. S. Y. 1993. The origin of the turtle body plan:Bridging a famous morphological gap. Science261:1616–1620.

Lee, M. S. Y. 1995. Historical burden in systematicsand the interrelationships of ‘‘parareptiles.’’ Biol.Rev. 70:459–547.

Lee, M. S. Y. 1997. Pareiasaur phylogeny and the or-igin of turtles. Zool. J. Linn. Soc. 120:197–280.

Lombard, R. E. and S. S. Sumida. 1992. Recent pro-gress in understanding early tetrapods. Amer.Zool. 32:609–622.

Modesto, S. P. 1996. The anatomy, relationships, andpalaeoecology of Mesosaurus tenuidens and Ster-eosternum tumidum (Amniota: Mesosauridae)from the Lower Permian of Gondwana. Unpub-lished Ph.D. Thesis, University of Toronto.

Modesto, S. P. 1999. Observations on the structure ofthe Early Permian reptile Stereosternum tumidumCope. Palaeont. Afr. 35:7–19.

Modesto, S. P. 2000. Eunotosaurus africanus and theGondwanan ancestry of anapsid reptiles. Palaeont.Afr. 36. (In press)

Olson, E. C. 1947. The family Diadectidae and itsbearing on the classification of reptiles. FieldianaGeol. 11:1–53.

Parrington, F. R. 1958. The problem of classificationof reptiles. J. Linn. Soc. (Zool.) 56:99–115.

Reisz, R. R. 1980. The Pelycosauria: A review of thephylogenetic relationships. In A. L. Panchen (ed.),The terrestrial environment and the origin of landvertebrates, Systematics Association Special Vol-ume No. 15, pp. 553–592. Academic Press, NewYork.

Reisz, R. R. 1981. A diapsid reptile from the Penn-sylvanian of Kansas. Sp. Pub. Mus. Nat. Hist.Univ. Kansas 7:1–74.

Reisz, R. R. 1986. Pelycosauria. Handbuch der Pa-laeoherpetologie 17:1–102. Gustav Fischer Ver-lag, Stuttgart.

Reisz, R. R. 1997. The origin and early evolutionaryhistory of amniotes. Trends Ecol. Evol. 12:218–222.

Reisz, R. R. and M. Laurin. 1991. Owenetta and theorigin of turtles. Nature 349:324–326.

Rieppel, O. and M. deBraga. 1996. Turtles as diapsidreptiles. Nature 384:453–455.

Rieppel, O. and R. R. Reisz. 1999. The origin andearly evolution of turtles. Annu. Rev. Ecol. Syst.30:1–22.

Ritter, D. 1995. Epaxial muscle function during loco-


motion in a lizard (Varanus salvator) and the pro-posal of a key innovation in the vertebrate mus-culoskeletal system. J. Exp. Biol. 198:2477–2490.

Ritter, D. 1996. Axial muscle function during lizardlocomotion. J. Exp. Biol. 199:2499–2510.

Romer, A. S. and L. I. Price. 1940. Review of thePelycosauria. Geol. Soc. Amer. Spec. Papers 28,538 pp.

Ruben, J. and T. D. Jones. 2000. Selective factors inthe origin of fur and feathers. Amer. Zool. 40:585–598.

Schultze, H.-P. 1987. Dipnoans as sarcopterygians. InW. E. Bemis, W. W. Burggren, and N. E. Kemp(eds.), The biology and evolution of lungfishes, pp.39–74. Alan R. Liss, New York.

Sumida, S. S. 1989. Reinterpretation of vertebral struc-

ture in the Early Permian pelycosaur Varanosau-rus acutirostris (Amniota, Synapsida). J. Vert. Pa-leont. 9:419–426.

Sumida, S. S. 1991. Vertebral morphology, alternationof neural spine height, and structure in Permo-Carboniferous tetrapods, and a reappraisal ofprimitive modes of terrestrial locomotion. Univ.Calif. Publ. Zool. 122:1–135.

Sumida, S. S. 1997. Locomotor features of taxa span-ning the origin of amniotes. In S. S. Sumida andK. L. M. Martin (eds.), Amniote origins, complet-ing the transition to land, pp. 353–398. AcademicPress, San Diego.

Vaughn, P. P. 1962. The Paleozoic microsaurs as closerelatives of reptiles, again. Am. Midl. Nat. 67:79–84.

Documents

Sumida & Modesto, 2001