Journal of 4(4):604-608, December 1984

PHYLOGENETIC METHODS AND THE EARLY HISTORY OF

In a recent review of early , Carroll groups must have arisen within Carroll's "Protoro- (1982) criticized the methods of phylogenetic (cladis- thyridae" (= Protorothyrididae). Carroll, however, is tic, Hennigian) . Specifically, he concluded concerned with entirely different characters, specifi- that several of the most important assumptions of this cally, those "in which the most primitive known mem- methodology were contradicted by the evidence from bers of each of the derived groups are more advanced early amniotes and that these contradictions resulted than the protowthyrids" (Carroll, 1982:100, italics from the evolutionary process of in added). But if all the other groups of early amniotes this group. Our opinions concerning the utility of the are derived independently from "protorothyrids," then phylogenetic method contrast strongly with those of any such derived characters shared by two or more of Dr. Carroll. We feel that the problems he encountered these groups must by definition be convergent. There- result from his misunderstanding of the phylogenetic fore, we do not find it suprising that Carroll (1982:102) method rather than from inadequacies of this method finds no obvious congruence among these characters. itself Dr. Carroll is one of the most knowledgable stu- Critics of the phylogenetic method (Mayr, 1981; dents of early amniote evolution; here we present nei- Carroll, 1982) argue that evolutionary parallelism and ther a detailed reanalysis of his characters nor a precise convergence pose particular problems for this method. alternative to his phylogeny of this group. Instead, we Platnick (1977) demonstrates clearly that the opposite argue that Carroll's own procedure was formulated such is true, namely that analysis of shared apomorphies that his negative conclusions about the assumptions provides an unambiguous method of identifying cases of the phylogenetic method were inevitable. We con- involving convergence. Such a determination is made tend that many of Carroll's objections to the phylo- by considering the distribution of the feature in ques- genetic method rest on faulty reasoning. tion in relation to those of other derived features among Carroll's most severe criticism of the phylogenetic the taxa concerned. For a given character distribution, method seems to be its ". . . assumption that the ma- parallelism/convergence can be shown to be either nec- jority of derived features possessed in common by re- essary or unnecessary to account for the distribution lated groups are the result of inheritance from a com- of particular derived characters among the terminal mon ancestor, rather than from convergence" (Carroll, taxa (Platnick, 1977). 1982:104; but see Farris, 1983). For the early amniotes, Convergence and parallelism refer to similarities de- Carroll (1982:102) finds that the distribution of de- rived independently rather than those inherited from rived features forms "an almost random mosaic" a common ancestor. By definition then, an hypothesis among his basic taxa. This leads him to conclude (Car- of relationship must be established in order to support roll, 1982:104) that "most of the derived character a claim of convergence. As Gaffney (1979) aptly states states recognized in two or more of the derived groups the case, "An argument for parallelism or convergence must be attributed to convergence." Thus, an impor- is simply an argument (often unstated) in favor of one tant assumption of the phylogenetic method seems to phylogeny over an alternative one." Carroll (1982) pro- be violated. vides no support for his alternative hypothesis of re- We certainly agree that rampant convergence can lationships. He presents no characters indicating that lead the systematist to incorrect phylogenetic conclu- any particular "protorothyrid" shared a more recent sions. However, if Carroll's scenario about early am- common ancestor with one of his "derived" groups niote evolution is correct and his prior assumptions than with other "protorothyrids." Such evidence would about basic taxa are accepted, then he has chosen to be necessary to support his claim that 1) these "de- discuss only those characters that must be convergent rived" groups were actually "derived from protoro- (see Fig. 1). In Carroll's scenario (p. 106), the time span thyrids," and 2) that any particular derived characters from the Early Pennsylvanian to the Early shared by two or more of these "derived" groups (but saw the appearance of "a succession of derived groups not by "protorothyrids") are therefore convergent. that may all be traced to the protorothyrid stock." Given that Carroll's hypothesis about early amniote Therefore, any true synapomorphies evidencing a close phylogeny is true, then all of the derived characters relationship among two or more of these "derived" shared by two or more of his derivative (non-proto-

604 1984 by The Society of rothyrid) taxa are necessarily convergent. It is also pos- sible, however, that Carroll's phylogeny is incorrect and that at least some of the derived characters shared by two or more of the other amniote taxa are true synapomorphies, characters shared because they were inherited from a common ancestor with a similar con- dition. Carroll dismisses this possibility, presumably because he doubts the significance of similar derived features shared by groups with different adaptive pat- terns in demonstrating a close common ancestry (see below). Such a superficial dismissal of characters seems to us to be ill-founded. Possible relationships suggested by shared, derived features are often inappropriately discounted when the derived features in question are thought to be adaptive for a particular lifestyle. It is argued that such features are not phylogenetically meaningful because they are rotorothyridae" predisposed to convergence. We do not deny the pos- sibility of convergence; however, it is not logical to conclude summarily that such shared, adaptive fea- tures are convergent and, therefore, that the suggested FIGURE 1. Phylogenetic diagram depicting derivation of relationships do not exist. Because a phylogeny is re- various "derived" amniote groups (shaded bubbles A-E) from quired to demonstrate convergence in a particular an "ancestral Protorothyridae" (unshaded bubble) as hy- character, the synapomorphic of such a shared, pothesized by Carroll (1982). Only derived characters that derived feature and the relationships suggested by it arose within "Protorothyridae" (bars with asterisks) can pro- cannot be dismissed until other derived features are vide evidence for a close relationship between two or more found to evidence different relationships, that is until of the "derived" lineages. Nevertheless, Carroll (1982) dis- an alternative phylogeny is proposed and supported. cusses only derived characters not present in any protoro- Many of Carroll's arguments are strongly tied to the thyrid (bars without asterisks). Given these relationships, any concept of , particularly to the concepts of such characters shared by two or more of the groups A-E the adaptive zone and of adaptive radiation (see also are necessarily convergent. Van Valen, 1971, 1978). For example, Carroll (1982: 104-105, italics in original) states: ory of evolution predicts that living things form a hi- "A distinction should be made in the relative sig- erarchy of groups within groups. Because evolutionary nificance of similar derived character states within descent is accompanied by the divergence of lineages, groups that share a common adaptive and structural the more inclusive the group, the less likely that its pattern, and between groups that have adapted to members will be very similar in adaptation. This does different ways of . not, however, preclude the possibility of recognizing The presence of shared derived characters can be more inclusive groups. Carroll certainly recognizes the useful in recognizing members oí particular groups. group Amniota, despite the very different adaptive . . . On the other hand, similar derived characters zones occupied by its component subgroups. We pre- present in groups with fundamentally different adap- dict that groups less inclusive than Amniota, but con- tive patterns may be of less significance in demon- taining two or more of Carroll's basic taxa, will even- strating a close common ancestry." tually be recognized. Some progress is already being made along these lines (GafFney and McKenna, 1979; We find this approach undesirable because it is so Reisz, 1981). dependent on subjective initial decisions of what con- Another perplexing aspect of the passage quoted stitutes an adaptively unified group. For example, the above is Carroll's seeming implication that one can "" are said to have contained both large discover little more about relationships than the mem- and small carnivores and two major groups of herbi- bership of groups already assumed to exist. Carroll's vores (Carroll, 1982:93)•groups seemingly very dif- statement that shared, derived characters are useful for ferent in adaptation. Had Carroll followed his own recognizing the members of particular groups is logi- reasoning and recognized these adaptive types as his cal•provided that the groups are already character- basic taxa, he would never have been able to conclude ized. Carroll's examples show that his "groups" are his that together they formed a larger group, his "pely- basic taxa (pelycosaurs, captorhinids, etc.), whose in- cosaurs." Carroll (1982:93) again contradicts himself terrelationships are the object of investigation. But by accepting an exclusive common ancestry among Carroll doubts the significance of derived characters groups even more divergent in their : shared by members of these different "groups" in dem- "pelycosaurs," "," and . The the- onstrating a close common ancestry between them.

JVP 4(4), December 1984 605 Thus, he faces the dilemma of being unable to dem- "In any case, I cannot accept Patterson's contention onstrate anything more than what he has initially as- that no useful information is conveyed by statements sumed to be true, namely, the existence of his basic referring to paraphyletic groups. Presumably he taxa as separate entities. would argue that the statement "Amniota originated Carroll never explains why derived characters shared from the Amphibia" is a meaningless statement, by groups that differ fundamentally in adaptive pattern simply because the Amphibia are a paraphyletic tax- might be insignificant in demonstrating a close com- on; nevertheless, he would agree that the Tetrápoda mon ancestry. Groups differing fundamentally in ad- are a natural monophyletic group, a , within aptation, such as the early amniote taxa discussed here, which are nested the Amniota as a subordinate clade. would be subject to dissimilar selective forces. There- This means•if an evolutionary interpretation be fore, it seems unlikely that the derived characters shared permitted•that the Amniota originated within the by such groups were shaped convergently by adapta- Tetrápoda. But they could not have originated from tion to similar selective forces. A more reasonable ex- themselves; they must have orginated from other planation for two or more of the groups sharing these . And the only other tetrapods, by defini- derived characters is inheritance from a common tion, are the Amphibia!" ancestor, i.e., a close relationship among the groups. Charig misses the point. Given that 1) Tetrápoda is One major difference between the traditional ap- monophyletic, 2) Amniota is monophyletic, and 3) proach advocated by Carroll (1982) and Hennig's (1966) Amniota is a subset of (i.e., originated within) Tetrápo- phylogenetic method is recognition of paraphyletic da, no additional phylogenetic information is con- groups in the traditional approach. Although justifi- veyed by 4) naming a group ("Amphibia") for non- cations for the formal recognition of such taxa are often amniote tetrapods and declaring that amniotes couched in biological terms, that is, in terms of theories originated from them. This would simply be saying about the evolutionary process (e.g., adaptation), para- that amniotes originated from tetrapods that were not phyletic groups only disguise a lack of knowledge. Per- amniotes. The statement "Amniota originated from haps there really are no characters that will serve to the Amphibia" is not meaningless, but given that am- link the major groups of amniotes, and we cannot say niotes are a subset of tetrapods it is certainly redun- how they are related to one another. This is a precise dant. statement about our lack of knowledge. Some of these Carroll offers an argument purporting to demon- groups, however, must share a more recent common strate that "Protorothyridae" (formerly the Romeri- ancestor with one another than they do with any of idae, now excluding Romerid) is a natural group. This the other groups (even if that common ancestor did argument, however, suffers from his failure to distin- not look very different from the most recent common guish between the meanings of two different pairs of ancestor of all of them). Therefore, when Carroll (p. terms, primitive and derived versus generalized and 107) hypothesizes that each of these lineages "may specialized. The former pair refers to the relative time have evolved separately from the conservative pro- of origination of characters in a transformation series, torothyrid stock," he has not solved the problem. In- the latter is an ecological concept referring to the breadth stead, a lack of knowledge is obscured by the formal of resources used. Carroll seems to equate specialized term Protorothyridae. For now, ironically, we "know" with derived, as can be seen in the following passage the evolutionary history of the amniotes (i.e., adaptive (Carroll, 1982:91, italics in original): radiation) even though we have no idea how its subgroups are interrelated. "Although protorothyrids exhibit the primitive Paraphyletic higher taxa, such as "Protorothyridae," character state for nearly all the skeletal features rec- are often called on to serve as "ancestors" when po- ognized among early , this should not be con- tential ancestors are unknown. But if species are the sidered as defining the group on the basis of prim- most inclusive actively evolving entities (Wiley, 1979), itive characters. then they are also the most inclusive entities that can Judged on the basis of Paleozoic , es- be ancestors. Supraspecific taxa, which do not evolve pecially the advanced anthracosaurs, the protoro- except as a result of the evolution of their constituent thyrids are highly specialized in their small body species (Hull, 1980), cannot give rise to anything, i.e., size, relatively small mobile skull, and light limbs. they are not biologically meaningful ancestors (Wiley, Of equal importance are their probable habitat and 1979, 1981). Often no known species meets the spec- dietary specializations, resembling those of primi- ifications of an ancestor in all of its characters. In such tive modem . This adaptive pattern charac- cases paraphyletic higher taxa are very accommodat- terizes the protorothyrids after a period of some 35 ing, for they can always be defined in such a way as to million years. This is as specialized a way of life, fit the character specifications of an ancestor (although relatively [sic] to that of other early tetrapods, as the they will not meet the requirement that an ancestor is carnivorous way of life of early pelycosaurs, or the a species). Unfortunately, such paraphyletic "ances- durophageous specializations of captorhinids." tral" higher taxa convey a false impression of knowl- Carroll's argument fails when one realizes that it is edge, as is revealed in the following passage from Char- possible for a character to be both primitive and spe- ig (1982:431, italics in original): cialized. It does not matter that "protorothyrids" are

606 JVP 4(4), December 1984 specialized as small insectivores. If the other amniote groups were derived from "protorothyrids" as Carroll claims (but does not support), then the characters of "protorothyrids" as a whole can only be primitive rel- ative to those of other amniotes. Another problem with the passage quoted above is Carroll's apparent failure to accept the consequences of the relative nature of the terms primitive and de- rived. "Protorothyrids" may be derived in certain fea- tures relative to Paleozoic "amphibians," but so are all amniotes. Carroll (1982:91) admits that "proto- rothyrids" "... share few if any advanced character states, relative to other early amniotes." Therefore, given that "protorothyrids" are amniotes, they can only be distinguished from other amniotes by their primi- tive (unmodified) characters. We do not understand how Carroll can claim otherwise. Perhaps what Carroll really considers to be natural about paraphyletic groups is that the members of such groups supposedly share common adaptive morpho- FIGURE 2. Time, , and . See text for types (see also Van Valen, 1971, 1978). But the mor- explanation. phologies seen in the living world do not fall on a single line of increasing adaptation. Part of modem evolu- tionary theory is the concept of (Mayr, The failure of traditional evolutionary systematists 1969), in which different characters undergo change in to appreciate the difference between paraphyletic and different groups and the same characters change at dif- monophyletic groups allows them to offer a curious ferent rates in different groups. The practice of rec- argument in favor of the natural status of paraphyletic ognizing paraphyletic grades as formal taxa is com- taxa. Carroll (1982:87) writes: "Early amniotes may patible with notions of evolutionary progress from be viewed today as a phylogenetically heterogeneous "lower" to "higher" forms or with the pre-evolution- assemblage (or paraphyletic group), for they include ary scala naturae, it is incompatible with modem evo- the ultimate ancestors of mammals and , as well lutionary theory. as of the modem squamates, and . A curious feature of traditional evolutionary system- Yet, within the context of the Paleozoic, they were a atics is its failure to discriminate consistently between monophyletic group, with a close common ancestry." two very different kinds of groups, paraphyletic groups The argument usually continues: if these organisms and strictly monophyletic groups. The traditional evo- formed a natural, monophyletic group in the Paleozoic, lutionary systematist considers both of these to be then they can still be viewed as a natural group today. monophyletic in a broader sense (Ashlock, 1979; Mayr, Surely the mere passage of time cannot change the 1974), and, therefore, to be both natural and equally natural status of a group. valid as formal taxa in phylogenetic classifications. In order to evaluate this argument, let us first assume There are, however, important differences between that the phylogeny of the organisms involved is known paraphyletic and strictly monophyletic groups, differ- (Fig. 2). At time t,, A, perhaps the "Protoro- ences that call into question the status of paraphyletic thyridae," is indeed a natural, monophyletic group. At groups as anything but artificial constructs. ij, taxon A is paraphyletic and, in the opinion of many The ambiguity of traditional evolutionary classifi- systematists, no longer natural. How has time changed cations is seen upon examination of a few of the taxa the status of the group? It hasn't. At t,, "Protoro- that this system has produced. Here we find in seem- thyridae" is synonymous with Amniota, and both are ingly equivalent (monophyletic, in the broad sense) natural (monophyletic) groups. At ij, Amniota (taxa taxa an unexpected difference in violability. Given suf- in bubble X on the right side of the figure) is still a ficient evolutionary change, it is permissible to remove natural (monophyletic) group. "Protorothyridae" and lineages from certain (paraphyletic) taxa. For example, Amniota, however, are no longer synonymous. In- Aves and Mammalia are said to have evolved from stead, the direct descendants of the former have been the paraphyletic "Reptilia" as has Amniota from the placed in different groups. Even though some "pro- paraphyletic "Amphibia." But lineages never evolve torothyrids" (e.g., II) are more closely related to some out of other (monophyletic) taxa. Even though caecil- nonprotorothyrids (Ila) than they are to other "pro- ians and most lack legs, they are and will always torothyrids" (I or III), these nonprotorothyrids are be tetrapods. Since traditional evolutionary system- placed in different groups (B, C, D). It is not the mere atists grant both paraphyletic and strictly monophy- passage of time that has changed the status of the group, letic taxa equivalent status, they seem not to appreciate but the systematist's (not nature's) decision to remove this fundamental difference. certain direct descendants from the group.

JVP 4(4), December 1984 607 There is another problem with paraphyletic grade REFERENCES taxa, namely that by creating such taxa the systematist invites misuse of classifications by biologists untrained Ashlock, P. D. 1979. An evolutionary systematist's view in systematics (although even trained systematists are of classification. Systematic Zoology 28:441-450. Carroll, R. L. 1982. Early evolution of reptiles. Annual guilty of these errors). These biologists can now talk Review of Ecology and Systematics 13:87-109. about such things as variation, geographic and tem- Chang, A. J. 1982. Systematics in biology: a fundamental poral distribution, and of taxa whose fates comparison of some major schools of thought; pp. 363- have been determined as much by human predilections 440 in Joysey, K. A. and Friday, A. E. (eds.). Problems as by biological processes. Can we honestly say that of Phylogenetic Reconstruction. Academic Press, Lon- after 35 million years the "Protorothyridae" became don. extinct even though some of them may have been the Farris, J. S. 1983. The logical basis of phylogenetic analysis; direct ancestors of the vast array of living amniotes? pp. 7-36 in Platnick, N. I. and Funk, V. A. (eds.), Ad- What does it mean to label this group "conservative" vances in . Volume 2. Proceedings of the 2nd (Carroll, 1982:103) if some of its members eventually Meeting of the Willi Hennig Society. Columbia Uni- versity Press, New York. gave rise to organisms as different as bats, whales, tur- Gaffney, E. S. 1979. monophyly: a phylogenetic tles, snakes, and birds? analysis. Bulletin of Carnegie Museum of Natural His- Throughout this paper we have referred to Hennig's tory 13:92-105. (1966) method as the phylogenetic method, as he orig- and McKenna, M. C. 1979. A late Permian cap- inally called it. Although this method is currently more torhinid from Rhodesia. American Museum Novitates widely known as cladism or cladistics, we have delib- no. 2688:1-15. erately avoided these terms. Too often cladism is con- Hennig, W. 1966. Phylogenetic Systematics. University of sidered to represent an alternative to traditional phy- Illinois Press, Urbana, 263 pp. logenetic methods (e.g., Carroll, 1982). Although this Hull, D. L. 1980. Individuality and selection. Annual Re- view of Ecology and Systematics 11:311-332. may be true of some of cladism's more recent subdis- Mayr, E. 1969. Principles of Systematic Zoology. Mc- ciplines (Nelson and Platnick, 1981; Patterson, 1982), Graw-Hill, New York, 428 pp. the phylogenetic method that Hennig formulated is 1974. Cladistic analysis or cladistic classification. not. Instead, Hennig's method is an elaboration of the Zeitschrift für zoologische Systematik und - kinds of evidence than can logically be used to make forschung 12:94-128. phylogenetic statements. As such Hennig's precepts are 1981. Biological classification: toward a synthesis not an alternative to traditional phylogenetic methods, of opposing methodologies. Science 214:510-516. but a refinement of them. Although many of his critics Nelson, G. and Platnick, N. 1981. Systematics and Bio- do not appreciate this fact, Hennig's method remains geography. Columbia University Press, New York, 567 a necessary part of any phylogenetic systematic study. pp. Patterson, C. 1982. Morphological characters and homol- Acknowledgements•We wish to thank M. J. Dono- ogy; pp. 21-74 in Joysey, K. A. and Friday, A. E. (eds.). ghue, L. J. Flynn, E. S. Gaffney, M. J. Novacek, and Problems of Phylogenetic Reconstruction. Academic R. R. Reisz for their careful reading and helpful crit- Press, London. icism of earlier drafts. Platnick, N. I. 1977. Parallehsm in phylogeny reconstruc- tion. Systemadc Zoology 26:93-96. ANDRÉ R. WYSS' Reisz, R. R. 1981. A from the Pennsylvanian Department of Geology, Columbia University, and of Kansas. University of Kansas Museum of Natural Department of Vertebrate Paleontology, History Special Publication no. 7, 74 pp. American Museum of Natural History Van Valen, L. 1971. Adaptive zones and the orders of Central Park West at 79th St. mammals. Evolution 25:420-428. New York, New York 10024 1978. Why not to be a cladist. Evolutionary Theory 3:285-299. KEVIN DE QUEIROZ' Wiley, E. O. 1979. Ancestors, species, and cladograms• Department of Zoology and remarks on the symposium; pp. 211-225 in Cracraft, J. Museum of Vertebrate Zoology and Eldredge, N. (eds.), Phylogenetic Analysis and Pa- University of California leontology. Columbia University Press, New York. Berkeley, California 94720 1981. : The Theory and Practice of Phylogenetic Systematics. John Wiley and Sons, New 'Both writers share equal authorship. York, 439 pp.

608 JVP 4(4), December 1984