ST?R The Ontogenetic Method for Determining Character Polarity and its Relevance to Phylogenetic Systematics Kevin De Queiroz Systematic Zoology, Volume 34, Issue 3 (Sep., 1985), 280-299. Stable URL: http://links.jstor.org/sici?sici=0039-7989%28198509%2934%3A3%3C280%3ATOMFDC%3E2.0.CO%3B2-6 Your use of the JSTOR archive indicates your acceptance of JSTOR' s Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of ajournai or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. Systematic Zoology is published by Society of Systematic Biologists. Please contact the publisher for further permissions regarding the use of this work. Publisher contact information may be obtained at http://www.jstor.org/joumals/ssbiol.html. Systematic Zoology ?1985 Society of Systematic Biologists JSTOR and the JSTOR logo are trademarks of JSTOR, and are Registered in the U.S. Patent and Trademark Office. For more information on JSTOR contactjstor-info@umich.edu. ?2002 JSTOR http://www.j stor.org/ Mon Mar 11 14:06:55 2002 Sysf. Zool, 34(3):280-299, 1985 THE ONTOGENETIC METHOD FOR DETERMINING CHARACTER POLARITY AND ITS RELEVANCE TO PHYLOGENETIC SYSTEMATICS KEVIN DE QUEIROZ Department of Zoology and Museum of Vertebrate Zoology, University of California, Berkeley, California 94720 Abstract.?In an attempt to clarify the relevance of ontogenetic transformations for system- atics, the ontogenetic method for determining character polarity (the biogenetic law of Nelson, 1978) is analyzed from the perspective of phylogenetic systematics. In phylogenetic systematics, as defined here, the relationships sought are those of common ancestry and, thus, the concept of phylogeny is taken as an axiom from which systematic methods are deduced. This perspective has a number of consequences concerning the role of ontogenetic transformations in system- atics, among which are the following: (1) Von Baer's second law, which states that less general characters are developed from the most general, is not universally true. (2) The validity of Nelson's biogenetic law (not to be confused with other concepts of similar name) does not depend on the validity of von Baer's law. (3) As a theory about the relationship between on- togeny and phylogeny. Nelson's biogenetic law can only be tested by known character phylog- enies. However, outgroup, paleontological, and ontogenetic methods of polarity determination need not be interpreted as scientific theories; instead, they can be interpreted as theorems deduced from the axiom of phylogeny and certain auxiliary assumptions. (4) The usefulness of the ontogenetic method rests on an assumption of ancestral character retention. If ancestral characters are retained in descendant ontogenies, then ancestral characters will be more general than their phylogenetic derivatives. (5) The sequence of ontogenetic transformation is irrelevant to the usefulness of the ontogenetic method; generality is the critical factor. (6) An "ontoge- netic" method based on generality may be useful for determining evolutionary polarity when characters are instantaneous morphologies, but ontogenetic transformations rather than instan- taneous morphologies are more appropriately considered characters when attempting to deter- mine phylogenetic relationships among organisms. When ontogenetic transformations are viewed as characters, there can be no ontogenetic method for determining evolutionary character po- larity; however, the comparative phylogenetic method properly involves a comparison of on- togenetic transformations. (7) Ontogenetic polarities are different than phylogenetic polarities; the two have the relationship of part to whole, respectively. (8) For characters that exhibit ontogenetic transformation, homology is distinct from synapomorphy. (9) Finally, there is no threefold parallelism in phylogenetic systematics. Comparative anatomy, paleontology, and em- bryology are not three separate disciplines within systematics; rather, the three form a single comparative method unified in the organism by the concept of evolution. [Biogenetic law; von Baer's law; cladistics; character; evolution; generality; hom.ology; ontogenetic method; ontogeny; outgroup method; paedomorphosis; parsimony; paleontological method; phylogenetic system- atics; phylogeny; polarity; semaphoront; synapomorphy; threefold parallelism.] I consider neoteny an apparent falsifier of a more ?, . . ohvloffpnptir svstpmatirs general principle [than the biogenetic law], that ^^^^ "^^ ?\ pnyiogenetic systematics of character phylogeny .... But I consider neo- (Hennig, 1966) has been accompanied by teny an apparent falsifier in a narrow sense. At an increasing concern about one of its es- times I have suspected that it is not a falsifier at sential components?methods for deter- all, but a reflection of lack of information. One joining the evolutionary polarity of char- may doubt, for example, that any characters are ?. , , , ,.?>,- . i ? i r truly lost, rather than transformed. Apparent loss acters. Although many different kinds of may be an indication that the characters and trans- evidence have been advocated towards this formations are merely poorly understood and, end (reviewed by de Jong, 1980; Stevens, consequently, wrongly defined. The problem may jggQ)^ ^^e tWO most popular form the ba- t"uf rcllrtefind SSm?alSn- ?es of the outgroup method (Watrous and tasks which I do not undertake here [Nelson, 1978: Wheeler, 1981; Farris, 1982; MaddlSOn et 344]. al., 1984) and the ontogenetic method 280 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 281 (Nelson, 1978; Nelson and Platnick, 1981; Patterson, 1982, 1983). The initial evolutionary basis of cladis- tics sparked the current interest in meth- ods of polarity assessment. Curiously, some advocates of the ontogenetic method, called pattern cladists by Beatty (1982), suggest that a belief in, or knowledge of, evolution is superfluous to cladistics (Plat- nick, 1979; Nelson and Platnick, 1981; Pat- terson, 1982, 1983). I agree with Patterson (1982) that the use of cladistic methods does not require any assumption about evolution; however, if systematics is an at- tempt to reconstruct evolutionary history, then this assumption will dictate which cladistic methods are useful. Such a per- spective reveals that, although the onto- genetic method may be satisfactory for pattern cladism, it is not satisfactory for phylogenetic systematics. The reason for this concerns one of the most fundamental concepts of systematics, namely, the na- ture of characters. The goals of systematics are important considerations in evaluating the appropri- ateness of systematic methods and con- cepts. Consequently, I must first consider the role of the theory of evolution in sys- tematics, for this is central to the differ- ence between the goals of pattern cladism and those of phylogenetic systematics. In order to expose the inadequacies of the ontogenetic method in phylogenetic sys- tematics and to highlight the need for a change in character concepts, I next ex- plore some consequences of polarity de- termination under traditional character concepts. These traditional concepts are subsequently rejected in favor of a char- acter concept chosen to reveal phyloge- netic relationships among organisms. THE ONTOGENETIC METHOD The ontogenetic method of polarity as- sessment has been articulated by Nelson (1978:327) as a restatement of the bioge- netic law: "given an ontogenetic character transformation, from a character observed to be more general to a character observed to be less general, the more general char- acter is primitive and the less general ad- vanced." This method is said to depend not on Haeckelian recapitulation but only on the validity of von Baer's second law (Patterson, 1982, 1983), which states that "less general characters are developed from the most general, and so forth, until finally the most specialized appear" (Gould, 1977:56). Nelson (1978; repeated in Nelson and Platnick, 1981) knew of no evidence against his version of the biogenetic law and concluded that it may be generally valid. He also argued that among the on- togenetic, outgroup, and paleontological methods of polarity determination, ontog- eny was the decisive criterion because it was least easily protected from falsifica- tion by ad hoc means. I do not dispute these claims within the context of pattern cladism; however, I argue that their force is lost from a phylogenetic perspective. I also argue that, although it may be possi- ble to divorce evolution and systematics, according to the view of science adopted by many systematists, a phylogenetic per- spective is preferred if systematics is to bear a scientific relationship to the theory of evolution. THE ROLE OF THE THEORY OF EVOLUTION IN SYSTEMATICS A central difference between pattern cladism and phylogenetic systematics is the role that the theory of evolution plays in systematics. According to Nelson (1978: 336, 337), "systematics and comparative anatomy (applied to fossils, too) are pos- sible only to the extent that ontogeny is orderly," and "the theory of evolution it- self is an extrapolation of ontogeny." In other words, the hierarchical relationships among characters in ontogeny form the basis of an internested hierarchical system of taxa, from which one can generalize the theory that evolution (used throughout this paper in the sense of descent with modification without regard to its mecha- nism) explains this hierarchy of groups within groups. Under this view of the relationship be- tween evolution and systematics, cladism bears no scientific relationship to the the- 282 SYSTEMATIC ZOOLOGY VOL. 34 ory of evolution, at least under the Pop- perian view of science advocated by many systematists (Bock, 1974; Wiley, 1975; Plat- nick and Gaffney, 1977; Nelson, 1978; Gaffney, 1979; Nelson and Platnick, 1981). Although my view of science differs, I will adopt the Popperian view of science for the purpose of evaluating the relationship between evolution and cladistic system- atics. According to Popper (1968), falsifi- ability is the demarcation criterion separat- ing scientific and nonscientific theories. Under this view of science and the pattern cladists' view of evolution as a generali- zation from systematics, cladistic system- atics cannot serve as a falsifier of evolu- tion. The general pattern of a systematic hierarchy of internested sets cannot be used to corroborate the theory that evo- lution explains this pattern unless alter- native patterns (e.g., intersecting sets, nonhierarchical) are tested for a better fit to the data. Furthermore, if evolution is generalized from systematics, then sys- tematics can hardly be inconsistent with evolution. Another consequence of adopting a Popperian view of science is that there is no logical basis for generalizing evolution from systematics. According to Popper (1968), there is no inductive logic; that is, there is no justification for inferring gen- eral statements from specific observations. Therefore, in addition to the lack of po- tential falsification of evolution by pattern cladistics, there is no justification for gen- eralizing evolution from pattern cladis- tics. Although some may view this as a problem with the theory of evolution, the potential to falsify this theory by other means (see below) suggests otherwise. Ac- cording to Popper (1968), the manner in which a theory is conceived is unimpor- tant as long as the theory is falsifiable. Alternatively, one can view evolution as the basis of systematics, as did Hennig (1966). Under this view, systematics might be used in conjunction with other data, such as stratigraphy or biogeography, to test the theory of evolution. For example, a deduction from the theory of evolution is the prediction that more inclusive clades will precede less inclusive ones in the stratigraphie sequence. This prediction is upheld; all recognized clades do not arise simultaneously in the stratigraphie se- quence. Failure to pass this test would surely be strong falsification, for if all clades appeared simultaneously in the stratigraphie sequence, I doubt that any- one would seriously entertain the theory that evolution accounted for the pattern of organic diversity. Although pattern cladists remove the evolutionary basis from systematics, they replace the theory of evolution with another, the theory that "there is order in nature" (Nelson and Platnick, 1981:9). As the basis for systematics, I question whether this theory is preferable to the theory of evolution. If the theory of nat- ural order claimed only that nature is or- derly, then it might be preferable because of its greater universality. Falsification of such a theory (if possible) would rule out any kind of natural orderliness, not only orderliness resulting from evolution. However, the theory of natural order as conceived by pattern cladists is not so uni- versal as this, for they entertain only one kind of order in systematics, specifically, a hierarchy of groups nested within groups. This is the kind of order expressed by cladograms. It is also the kind of order that evolution is supposed to produce. But the theory that evolution has produced a hierarchy of groups nested within groups makes more claims about the world than does the theory that such a hierarchy merely exists (for example, evolution claims that the hierarchy has a temporal component), and theories that make more claims have more potential falsifiers (Pop- per, 1968). Finally, an evolutionary basis justifies the use of cladistic methods in biological systematics. If systematics is not based on evolution, then one must ask: "Are cladis- tic methods preferable to (for example) phenetic ones and, if so, why?" Accepting the theory of evolution as the basis for sys- tematics provides a reason for preferring cladistic methods. Unlike phenetic meth- ods, cladistic methods were formulated by Hennig (1966) with an explicit evolution- ary basis. I conclude that if a falsification- 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 283 ist perspective is adopted, and if system- atics is to bear a scientific relationship to evolution, then systematics should be based on evolution. Of course, this consideration of falsifi- ability is superfluous if one considers evo- lution to be a "fact." However, acceptance of such a proposition combined with an interest in testing theories about particu- lar evolutionary processes provides an even better reason for basing systematics on evolution. This is because phylogenetic relationships have direct bearing on many theories about evolutionary processes (El- dredge and Cracraft, 1980; Wiley, 1981). Therefore, if systematics is to hold a cen- tral role in modern comparative biology, which is said to be unified by the concept of evolution, then systematics should be based on this unifying concept but mini- mize assumptions about its underlying processes. No one has specified what it means to say that systematics is "based on" the con- cept of evolution. I will do this by defin- ing phylogenetic systematics as that sys- tematics in which the relationships sought are phylogenetic relationships and, thus, the concept of evolution, or phylogeny (and by this I mean only that organisms are related through common ancestry), serves as an axiom from which systematic methods are deduced. Attempting to refor- mulate systematic methods as deductions from the axiom of evolution should help to clarify whether a given method is use- ful in phylogenetic systematics. The pres- ent paper is an evaluation of the ontoge- netic method from this perspective. My analysis of the ontogenetic method is strictly from the perspective of phylo- genetic systematics, as I define it, with no claim that this perspective is the only val- id one. However, I have just presented reasons for thinking that it is the appro- priate one for modern comparative biolo- gy- GENERALITY AND THE BIOGENETIC LAW I wish to consider the meaning of the term "general" as it is used in von Baer's second law and Nelson's biogenetic law (i.e., the ontogenetic method). This neces- sitates a brief consideration of the nature of scientific laws. There- are at least two possible interpretations of scientific laws: (1) they may be theorems, that is, neces- sary truths deduced from axioms; or (2) they may be theories, that is, falsifiable hypotheses. As necessary truths, laws can serve to define certain of their component terms, such as "general" in the case of the two laws under consideration here. Be- cause Nelson (1978) stressed the falsifi- ability of his biogenetic law, he clearly viewed it as a scientific theory rather than a necessary truth. If von Baer's law is viewed similarly, then the meaning of "general" is not given by these laws and must be established. Von Baer's second law states that the less general characters are developed from the most general. As a scientific theory about development this seems to mean that the characters of less inclusive groups arise in development from the characters of more inclusive groups. However, if the hier- archy of groups is determined by the se- quence of character transformations in on- togeny, then this law is an unfalsifiable tautology. Von Baer's law is testable only if the hierarchy of groups and, thus, the generality of characters is determined ac- cording to some other criterion. Nelson (1978) did this by defining generality as follows: a character is more general if it occurs in both members of a pair of species; it is less general if it occurs in only one of them. I will generalize this definition ac- cording to the suggestion of W. P. Mad- dison (pers. comm.): character x is more general than character y if and only if all organisms possessing y (at some stage in ontogeny) also possess x and in addition some organisms possessing x do not pos- sess y. According to this definition of general, von Baer's law is not a law; that is, it is not universally true (Fink, 1982). De Beer (1940) cited various cases ("embryonic and larval adaptations") in which less general characters develop into more general ones. For example, the dorsal nerve cord of te- leosts arises as a solid rod and later hol- lows out; however, many other chordates develop a hollow dorsal nerve cord by 284 SYSTEMATIC ZOOLOGY VOL. 34 (A) a a or a a^^c (B) CO b a^ or b c^^b FIG. 1. Possible relationships between a character and its ontogenetic precursor: (A) precursor is more general; (B) precursor is equally general; (C) precur- sor is less general. An asterisk marks the character in question. See text for relevance to von Baer's law. folding a plate of ectodermal tissue. Hence, less general characters (solid rod and un- folded plate) develop into a more general one (hollow cord). Patterson (1983:25) claimed that "the embryonic membranes of amniotes, as de- velopments (outgrowths) or [of?] more widely distributed structures, are consis- tent with this law [von Baer's second], as is, so far as I know, every other observa- tion in vertebrate morphology." Perhaps the reason that Patterson and I disagree about the validity of von Baer's law is that we interpret it differently. One might view von Baer's law as a statement about the ontogenetic precursors of less general characters that says nothing about those of more general characters. Under this inter- pretation, von Baer's law is not falsifiable. There are only three possible situations concerning the generality of the ontoge- netic precursor of a given character (Fig. 1): the precursor can be more general, it can be equally general, or it can be less general. If the precursor is more general. von Baer's law is confirmed. However, if the precursor is of equal or lesser gener- ality, there is no reason to reject von Baer's law. In the former case, the transforma- tions either involve the same characters or wholly different ones, and there is little reason to compare them. In the case where the precursor is less general than the char- acter under consideration, one has simply asked the question "What is the generality of the precursors of less general charac- ters?" about the wrong character. Von Baer's law is falsifiable only if it is inter- preted as stating that ontogenetic trans- formations always go from more general to less general. If this interpretation is ac- cepted, von Baer's law is falsified. Overemphasizing the importance of on- togeny with respect to the way in which characters are conceived can bias our views about evolution and the validity of von Baer's law. For example, some consider similar development to be the decisive cri- terion of homology. However, this should not allow us to conclude that otherwise similar characters that differ in their mode of development are not homologous. By doing this one rules out the possibility of detecting cases in which the mode of de- velopment of a structure is modified dur- ing the course of evolution without mod- ification of the structure's final form (e.g., the development of the hollow nerve cord described above). Perhaps such occur- rences are rare because a form is often in- fluenced by the forms of its ontogenetic precursors. However, if we define away the possibility for a character to remain unmodified in phylogeny when its onto- genetic precursors are modified (e.g., the hollow nerve cords of teleosts are not ho- mologous with those of other chordates because they develop differently), then von Baer's law is not only true, it is a tau- tology and cannot be otherwise. Since von Baer's law is not universally true, it is fortunate, contrary to Patterson's (1982, 1983) view, that the ontogenetic method of Nelson (1978) does not depend on this law. As stated by Nelson (1978), the biogenetic law concerns only those cases in which the ontogenetic transfor- 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 285 mation is from more general to less gen- eral. Since the exceptions to von Baer's law (transformations from less general to more general) are not considered, whether or not these exceptions exist is irrelevant to the validity of Nelson's (1978) biogenetic law. Unlike von Baer's law. Nelson's (1978) biogenetic law is a statement about phy- logeny rather than ontogeny (which is given). Specifically, it concerns the evo- lutionary polarity of characters. Neverthe- less, use of the ontogenetic method in (nonphylogenetic) systematics does not require assuming evolution. One might use the polarity (not necessarily evolu- tionary) inherent in ontogeny to construct an internested hierarchical system of taxa. Alternatively, one might use the polarity inherent in differences in the relative gen- erality of characters to do likewise. Ac- cording to von Baer's law, ontogenetic po- larities are identical with those based on generality (and, thus, so are the resulting hierarchical systems). This is false. Ac- cording to Nelson's (1978) biogenetic law, when exceptions to von Baer's law are re- moved from consideration, ontogenetic polarities and hence those based on gen- erality are identical with evolutionary po- larities (character phylogenies). As a sci- entific theory, this law can only be tested against known character phylogenies. Nelson's (1978:327) first class of potential falsifiers. However, Nelson (1978:327) as- serted "that paleontologists are the only folk who claim to know the truth of any evolution relevant in the present context." He also claimed that the paleontological argument is fallacious. Thus, Nelson (1978) seemed to reject the only evidence that could falsify his biogenetic law as a the- ory. Although Nelson (1978) claimed that his biogenetic law is falsifiable, he did not contradict himself by dismissing phylog- eny as a potential falsifier. This is because he accepted a second class of potential fal- sifiers, conflicting character transforma- tions. For example, in one pair of taxa character x is more general than character y and x transforms into y, while in another pair of taxa y is more general than x and y transforms into x. Contrary to Nelson's (1978) claim, such conflicts do not falsify the hypothesized relationship between ontogeny and phylogeny. Instead, they falsify either hypotheses about particular character phylogenies (e.g., that x is an- cestral to y based on the first pair of taxa, but see below) or the hypothesis that char- acters are not shuffled around haphazard- ly within ontogenies during phylogeny (i.e., that ontogenies evolve in an orderly fashion). Nelson (1978) knew of no in- stances of such conflicting transforma- tions, suggesting that ontogenies do not evolve through haphazard shuffling of characters. COMPARISON OF OUTGROUP, PALEONTOLOGICAL, AND ONTOGENETIC METHODS OF POLARITY DETERMINATION I have argued that Nelson's (1978) bio- genetic law is unfalsifiable if one rejects the possibility of testing it against phylog- eny. What then is the biogenetic law? Within the domain of phylogenetic sys- tematics (as defined in this paper) and un- der traditional character concepts. Nel- son's biogenetic law is interpretable as a law in a different sense. In phylogenetic systematics, the biogenetic law is not a theory about the relationship between on- togeny and phylogeny, but a theorem about the evolutionary polarity of char- acters (which I will call the ontogenetic method) that can be deduced from the ax- iom of phylogeny. I will illustrate this by comparing the ontogenetic method with two other methods for determining evo- lutionary character polarity?the out- group and paleontological methods. When referring to "alternative" characters in the following discussion I mean either that one is a modification of the other or that both are modifications of a third character. The three methods of polarity determination are designed to reveal the ancestral and derived conditions of two (for simplicity's sake) homologous characters and can be stated as follows. The outgroup method.?Given the exis- tence of a monophyletic group within 286 SYSTEMATIC ZOOLOGY VOL. 34 which occur alternative characters, a phy- logenetic character transformation must have occurred within the group. (This log- ic applies only to cases in which the char- acters vary among the taxa whose inter- relationships are being investigated, not within them.) Thus, the character found both inside and outside of the group is ancestral; the character found only inside of the group is derived. The paleontological method.?Given alter- native homologous characters whose ex- istence is documented in fossils, the char- acter found in the oldest fossil is ancestral, the alternative derived. By definition, the ancestral character must precede the de- rived character in time. Therefore, the old- est fossil will exhibit the ancestral condi- tion if it represents either (1) an ancestor that existed before the phylogenetic char- acter transformation occurred, or (2) a lin- eage that diverged before the phylogenet- ic character transformation occurred. The ontogenetic method.?"Given an on- togenetic character transformation, from a character observed to be more general to a character observed to be less general, the more general character is primitive [an- cestral], the less general advanced [de- rived]" (Nelson, 1978:327). If the defini- tion of generality proposed in this paper is accepted, then the justification for the ontogenetic method depends on the per- sistence of ancestral characters in organ- isms with modified (derived) ontogenies. As long as this condition is met, ancestral characters will always be more general, derived ones less general. Additionally, as stated by Nelson (1978), use of the onto- genetic method is restricted to cases in which the temporal sequence of ontoge- netic transformation is from more general to less general. Nelson's ontogenetic method might also be deduced from an assumption of recapitulation, but since Nelson (1973a) considered the recapitula- tionist argument to be fallacious, I assume that his method does not have a recapit- ulationist basis. Given that systematics attempts to re- construct evolutionary history, all three methods gain their usefulness from as- sumptions additional to the axiom of phy- logeny. Denying the validity of these ad- ditional assumptions permits one either to reject the method or, given conflicting re- suits, to evade "falsification" of the meth- od (i.e., to account for the conflict). The second form of denial is clearly ad hoc, since the validity of the assumptions must have been accepted in order to use the method in the first place. Nelson (1978) viewed the three meth- ods of polarity determination as theories and evaluated them in terms of their fal- sifiability. He concluded that the ontoge- netic method was decisive, because it was less easily protected from falsification by ad hoc means. From the perspective of phylogenetic systematics, such an empha- sis on falsifiability would be misplaced. Although hypotheses about the polarities of particular characters are falsifiable the- ories, the various procedures for deter- mining character polarities are more ap- propriately viewed not as theories but as methods (theorems) deduced from the ax- iom of evolution. The value of the various methods results not from their falsifiabil- ity but from the correct results that they must produce given the validity of certain other assumptions. Their "falsifiability" can be viewed alternatively as our will- ingness to accept the validity of these additional assumptions. The less willing we are to do so, the less adequate the meth- od. Explaining away inconsistencies with ad hoc propositions that contradict the assumptions necessary for the use of a particular method amounts to an unwill- ingness to accept these assumptions. Therefore, the extent to which ad hoc propositions are used to explain away con- flicts can still be viewed as a criterion for judging the adequacy of various methods. As noted by Nelson (1978), conflicting evidence in all three methods can be side- stepped by calling seemingly equivalent characters nonequivalent (nonhomolo- gous, homoplasious). The outgroup meth- od allows for a second option in question- ing the assumption of monophyly; perhaps certain outgroups are really ingroups. The paleontological method allows for a sec- 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 287 ond option in questioning the complete- ness of the fossil record. The incomplete- ness of the fossil record raises doubts about the validity of the assumption that the oldest fossil represents either an ancestor that existed or a lineage that diverged be- fore the phylogenetic character transfor- mation in question occurred. Although this assumption could potentially be support- ed by other characters, such a practice re- duces the paleontological method to a form of the outgroup method. Interpretation of the ontogenetic meth- od as a deduction from the axiom of evo- lution has consequences that bear on the use of this method under traditional char- acter concepts. First, and contrary to Nel- son (1978), the ontogenetic method allows for a second means (in addition to the ad hoc hypothesis of homoplasy) of explain- ing away conflicting data and is similar to the outgroup and paleontological meth- ods in this respect. The example of con- flicting ontogenetic data used by Nelson (1978) relies on an arbitrary separation of data, otherwise no conflict occurs. Nelson considered two pairs of species. In one pair, character Xp^ does not transform in species A, but x^ transforms into I/B in species B. In the second pair, y^ trans- forms into XQIXV species C, but y^ does not transform in species D. If the two pairs are examined independently of one another, the data from one pair suggest that x is ancestral while those from the other sug- gest that y is ancestral; however, when all four species are examined together no po- larity decision can be made since both characters have equal generality (Voor- zanger and Van Der Steen, 1982). A simi- lar inability to use the ontogenetic method results if one of the two transforming species (B or C) does not exist. A more in- teresting case would be the existence of both transforming species but only one of the nontransforming species. In this case, the characters differ in generality so that polarity can be established. Suppose that species A, B, and C exist, but species D does not. As stated. Nelson's (1978) formulation of the ontogenetic method applies only to situations in which the ontogenetic transformation is from more to less general (in.this case from x to y). Thus, transformations such as that occurring in species C (y to x) would be ignored. This is an unnecessary restriction (see below). If we remove this restriction there is a conflict. The relative positions of the characters in different transforming ontogenies suggest one or more of three possibilities: (1) seemingly equivalent characters in different ontogenies are not equivalent; or (2) ancestral characters have not been retained in descendant ontoge- nies (Fig. 2); or (3) ontogeny does not evolve in a orderly manner (see section on generality and the biogenetic law). The ad hoc hypothesis of homoplasy is available to all three methods of polarity determi- nation discussed herein. However, rejec- tion of the assumption that ancestral char- acters have persisted in organisms with modified character ontogenies provides another ad hoc means of explaining con- flicting ontogenetic data (Fig. 2). The on- togenetic method has no fewer options for ad hoc protection than do the other meth- ods of polarity determination. MISUNDERSTANDINGS ABOUT THE ONTOGENETIC METHOD Commonality.?Adopting a phylogenetic perspective clarifies several other misunderstandings about the ontogenetic method. The first concerns the difference between ontogenetic and commonality methods. Kluge (1985) claimed that Nel- son's law is a special form of ingroup anal- ysis that uses commonness as the esti- mator of polarity. The definition of generality necessitates that more general characters are also more common than less general ones (Fig. 3A). Nevertheless, the ontogenetic method is not the same as the commonality method, for more common characters are not necessarily more gen- eral (Fig. 3B). Since commonality bears no necessary relationship to relative time of phylogenetic appearance, there is no rea- son to infer that common characters are ancestral to less common ones (compare Van Valen, 1978; de Jong, 1980; Stevens, 1980; Arnold, 1981; Watrous and Wheeler, 288 A SYSTEMATIC ZOOLOGY B y. y. VOL. 34 -^ X, xf ? y. \ ,1 ^ ' ^ A -' c w2 X^ lost \ yc It / xi lost X ?*? V z J'z ! xl--*y. FIG. 2. Ad hoc protection for the ontogenetic method. (A) Character ontogenies of three organisms which it is inferred (by generality) that x is ancestral and y is derived. (B) Provided that ancestral characters have persisted in organisms with modified ontogenies, homoplasy must be invoked to explain the known patterns of ontogenetic character transformation, that is, either x^ * x^ or i/, ?= y^. (C) If loss of ancestral characters is admitted as possible, then no homoplasy is required; y-^ and y c may stem from the same character in the common ancestor. Thus, homoplasy is not the only ad hoc means of accounting for conflicting onto- genetic data. Arrows on solid lines represent ontogenetic transformations; those on dashed lines represent phylogenetic transformations. Certain nontransforming ontogenies have been expanded redundantly in or- der to facilitate comparison. 1981; Wheeler, 1981; Wiley, 1981). In con- trast, given that ancestral characters are retained in modified ontogenies, ancestral characters are necessarily more general than their derivatives. Sequence versus generality.?A more in- teresting result of the present analysis is that the sequence of ontogenetic transfor- mation is irrelevant to the ontogenetic method. The critical factor is generality. As long as ancestral characters are re- tained in descendant ontogenies, ancestral characters will always be more general than their derivatives. Even when the on- togenetic transformation is from less gen- eral to more general one can infer that the more general character is ancestral, the less general derived. In fact, the logic that characters arising earlier in phylogeny will be more general than those arising later in phylogeny applies even to characters that are parts of different ontogenetic or phylogenetic transformations. For exam- ple, one can infer that backbones arose earlier in phylogeny than did feathers from the observation that every organism that has feathers has a backbone but not every organism that has a backbone has 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 289 feathers. A corollary of this proposition is that the ontogenetic method relies neither on Haeckelian nor on "von Baerian" re- capitulation. The notion that the sequence of character transformation in ontogeny mirrors that in phylogeny is central to both of these concepts (Lovtrup, 1978), but the sequence of character transformation in ontogeny is irrelevant to the ontogenetic method. Given that the sequence of ontogenetic transformation is irrelevant to the useful- ness of the ontogenetic method, it is cu- rious that so much emphasis is placed on this sequence. Nelson's (1978) restriction of the ontogenetic method to cases in which the ontogenetic transformation is from more to less general places unnec- essary limitations on the applicability of this method, for generality alone is suffi- cient to establish polarity (see Nelson, 1978:339). Furthermore, if one is unwill- ing to dismiss the possibility that modifi- cations of nonterminal ontogenetic stages can occur without resulting in the modi- fication of terminal stages, then knowl- edge of the ontogenetic sequence without knowledge about generality does not pro- vide evidence about evolutionary charac- ter polarity. Thus, it is useful to distin- guish between different versions of the ontogenetic method (see Nelson, 1973a). The first of these is the traditional version, which is of dubious value since it is based on the universal occurrence of phyloge- netic change through the modification of terminal ontogenetic stages (including Haeckelian and "von Baerian" recapitula- tion). I largely ignore this traditional ver- sion in the present paper. A second for- mulation of the ontogenetic method is Nelson's (1978), which is based on gener- ality but is unnecessarily restricted to cases in which the ontogenetic transformation is from a more general character to a less general one. A third version of the onto- genetic method is presented in this paper. It too is based on generality but considers the sequence of ontogenetic transforma- tion to be irrelevant. This third version of the ontogenetic method can be stated as follows: given that ancestral characters are FIG. 3. The relationship between generality and commonality. (A) Since a more general character oc- curs in all those organisms (or taxa) that possess a relatively less general character as well as in some organisms that do not possess the less general char- acter, a more general character will always be more common than a less general one. (B) In cases where organisms (or taxa) bearing the characters in ques- tion either form intersecting sets (above) or do not overlap (below), more common characters are nei- ther more general nor less general than less common ones. The size of the ellipses is proportional to the number of organisms (or taxa) included within them. retained in descendant ontogenies, ancestral characters are more general than derived char- acters. Because these last two versions of the ontogenetic method are based on gener- ality, the validity of each depends on the retention of ancestral characters in de- scendant ontogenies. Unfortunately, the universal occurrence of ancestral character retention is a dubious proposition, and thus the usefulness of these methods is questionable. By itself, the failure of an- cestral characters to persist in descendant ontogenies will result in characters of equal generality (Figs. 4 and 5) and, thus, will only render the methods inapplica- ble; it will not lead to incorrect results. However, when coupled with incomplete sampling, the loss of ancestral characters can result in derived characters having greater apparent generality than ancestral ones (Fig. 5), thus leading to incorrect in- ferences about evolutionary character po- larity. Furthermore, although ancestral characters are more general than derived characters given ancestral character reten- tion, it does not follow that more general characters are necessarily ancestral, even with complete sampling. If two characters 290 SYSTEMATIC ZOOLOGY VOL. 34 A Origin of derived character ind ioss of ancestrai character x?*y y?*x Organisms with y Organisms with z FIG. 4. Effect of the loss of ancestral characters on generality and the determination of evolutionary character polarity. In cases where the origin of the derived character (2) coincides with the loss of the ancestral character {y) in phylogeny, ancestral and derived characters occur in nonoverlapping sets of organisms and thus have equal generality. Therefore, character polarity cannot be determined on the basis of generality. Horizontal arrows represent ontoge- netic transformations; vertical arrows represent phy- logenetic transformations. arise simultaneously in phylogeny (i.e., within the ontogeny of a single organ- ism), then the phylogenetic loss of one of these characters results in the alternative character having greater generality even though it is not ancestral. Paedomorphosis.?Another result of the present analysis is that, contrary to con- cerns stated by some authors (Rieppel, 1979; Stevens, 1980; Arnold, 1981), pae- domorphosis is not particularly problem- atic for either Nelson's or my version of the ontogenetic method (Fig. 6). Use of this method depends on the retention of an- cestral characters in descendant ontoge- nies, while paedomorphosis involves the elimination of terminal characters, which may or may not be ancestral. If terminal addition or the modification of terminal ontogenetic stages are the common pat- terns in phylogeny, as is suggested by the "laws" of Haeckel and von Baer, then pae- domorphosis will rarely cause problems in application of the ontogenetic method. This is because derived rather than ances- tral characters will be eliminated. Never- Loss of ancestrai character x?*y x?*x Origin of derived character COMPLETE SAMPLING y?-*y y?*x x?-^x Organisms with nisms with y iNCOMPLETE SAMPLiNG Organisms with Organisms with y FIG. 5. Effect of the loss of ancestral characters on generality and the determination of evolutionary character polarity. In cases where the origin of the derived character (y) precedes the loss of the ances- tral character (x) in phylogeny, ancestral and derived characters occur in partially overlapping sets of or- ganisms and thus have equal generality. Therefore, if organisms having each of the three classes of char- acter ontogenies are sampled, evolutionary character polarity cannot be determined. However, if organ- isms having character ontogeny x -^ x are not sam- pled, the derived character (y) will appear to be more general than the ancestral character (x), and an in- correct polarity inference will result. theless, if a terminal character lost through paedomorphosis also happens to be ances- tral, then paedomorphosis coupled with incomplete sampling can lead to erro- neous inferences about evolutionary char- acter polarity (Fig. 7). The confusion surrounding the sup- posed problem that paedomorphosis poses for the ontogenetic method results from an inappropriate extrapolation of the on- togenetic method to cases that it does not claim to cover. As stated. Nelson's bioge- 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 291 A Paedomorphosis (loss of terminal character) A Origin of derived characte x?-^a?-^a x**b A Paedomorphosis (loss of terminal character) A Origin of derived characte x?*a?*a COMPLETE SAMPLING Organisms with a Organisms with b FIG. 6. Paedomorphosis involving loss of a de- rived character and the ontogenetic method. When phylogeny involves terminal addition or the modi- fication of terminal ontogenetic stages, subsequent paedomorphosis causes loss of a derived character (&). As long as the ancestral character (a) is retained, it will never be less general than the derived char- acter, even with incomplete sampling. Therefore, in- correct polarity inferences will not result. netic law describes the relationships be- tween instantaneous morphologies. (In- stantaneous here means brief enough so that only a single morphological character is recognized.) Given an ontogenetic char- acter transformation from morphology x to morphology y in some organisms and the presence of only morphology x in oth- ers, it is concluded that morphology x is ancestral and morphology y is derived. That paedomorphosis is responsible for some of the nontransforming ontogenies does not invalidate this conclusion in any way, for the ontogenic method makes a statement about the relationship among instantaneous forms, not among ontoge- nies. Implicit or explicit in the writings of various authors who consider paedomor- phosis to be a source of problems for the ontogenetic method is the notion that comparisons are made among adults (Gould, 1977; Van Valen, 1978; Rieppel, 1979; de Jong, 1980; Stevens, 1980; Arnold, 1981). Comparison of adult stages is an at- Organisms with Organisms with b INCOMPLETE SAMPLING Organisms with a Organisms with b FIG. 7. Paedomorphosis involving loss of an an- cestral character and the ontogenetic method. When phylogeny involves the modification of nonterminal ontogenetic stages without modification of terminal stages, subsequent paedomorphosis can cause the loss of an ancestral character (a). If organisms with all three classes of character ontogenies are sampled, the characters have equal generality and no polarity in- ference can be made. However, if certain organisms (i.e., those having A: ^ a ^ a) are not sampled, then the ancestral character (a) will appear to be less gen- eral than the derived character (b), and evolutionary polarity will be incorrectly inferred. tempt to compare organisms (ontogenies) at a standard stage, and paedomorphosis will cause problems when instantaneous morphologies whose polarities have been inferred using the ontogenetic method are used to determine phylogenetic relation- ships among organisms (Stevens, 1980; Kluge, 1985). Thus, there appear to have been at least two misunderstandings in- volved when Stevens (1980) and Arnold (1981) criticized Nelson (1978) on the grounds of paedomorphosis. First, since Stevens (1980) considered Nelson (1978) 292 SYSTEMATIC ZOOLOGY VOL. 34 to have included some form of outgroup analysis in his reformulation of the bio- genetic law, Stevens misconstrued Nel- son's meaning of generality. Second, both Stevens (1980) and Arnold (1981) criti- cized Nelson (1978) from the perspective of organism phylogenies, while the onto- genetic method, as stated, elucidates not the phylogeny of organisms but that of instantaneous morphologies. Of course, most systematists (but not necessarily most comparative anatomists) study organism phylogenies, and they commonly base these organism phylogenies on inferences about the phylogeny of instantaneous morphologies (character polarities). Nel- son (1978) did not discuss the relationship between such character phylogenies and organism phylogenies, and he also mixed instantaneous and transformational char- acter concepts in his discussion of paedo- morphosis (e.g., "If neoteny [paedomor- phosis] is assumed, the transformation from X to y is primitive, not character x alone . .." [Nelson, 1978:340]). Part of the confusion about the ontogenetic method apparently stems from a failure to clearly distinguish between organism phyloge- nies and the phylogenies of instantaneous morphologies. AN ALTERNATIVE CHARACTER CONCEPT Instantaneous morphologies and ontogenetic transformations as characters.?The source of the confusion about paedomorphosis and the ontogenetic method resides in the ba- sic assumption of what constitutes a char- acter. Under the ontogenetic method, as stated by Nelson (1978) and adopted by Nelson and Platnick (1981) and Patterson (1982, 1983), characters are instantaneous morphologies that make up ontogenetic transformations. This character concept is used widely among systematists and is what I have referred to above as the tra- ditional character concept. However, bio- logical systematics has generally been concerned with the relationships among organisms rather than those among in- stantaneous morphologies, and organis- mal morphology is ontogenetically dy- namic. Consequently, as Patterson (1983: 27) and others have indicated, "phylogeny differs from ontogeny in that it is a se- quence of ontogenies, or life cycles." Therefore, if we attempt to determine the phylogenetic relationships among organ- isms as life cycles (Danser, 1950), and if we do not wish to bias ourselves against the possibility of paedomorphosis, then instantaneous morphologies whose polar- ities are determined by the ontogenetic method should not be the characters of the phylogenetic systematist; ontogenetic transformations, or the lack thereof, should be the characters instead (see Nelson and Platnick, 1981:353). This does not mean that all characters are properly ontogenet- ic transformations, for many attributes of organisms do not transform during ontog- eny. Acceptance of the view that ontogenetic transformations are characters clearly re- veals the inappropriateness of the onto- genetic method in phylogenetic system- atics. Under the (dubious) assumptions of ancestral character retention and the non- simultaneous phylogenetic appearance of the characters in question, the ontogenetic method might be used to argue that char- acter (y) in the previous example is de- rived from character {x). Nevertheless, it cannot justify the conclusion that an or- ganism bearing character (x -> y) was de- rived from one bearing character {x -^ x), for both of these characters have equal generality. In fact, if one accepts ontoge- netic transformations as characters then the phrase "ontogenetic character trans- formation" is redundant. Instantaneous morphologies, often treated as characters, are parts of ontogenetic transformations and are thus only parts of characters. Un- der this view of characters, there can be no ontogenetic method for polarity deter- mination; the ontogenetic transformations within a character tell nothing about the evolutionary polarity of that character rel- ative to others. At this point I want to reemphasize the difference between two different kinds of polarity. Characters (as ontogenetic trans- formations) have both ontogenetic and evolutionary polarities (Nelson, 1985). 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 293 Ontogenetic polarities exist within char- acters and, in doing so, they exist between instantaneous morphologies, which are parts of characters. In contrast, evolution- ary polarities exist between characters. When speaking of methods of polarity de- termination, I refer only to evolutionary polarities. I suspect that much confusion has resulted from a failure to distinguish between different kinds of polarity, a fail- ure that is not surprising under traditional character concepts. When instantaneous morphologies are treated as characters, both ontogenetic and evolutionary polar- ities exist at the same hierarchical level (i.e., both exist between characters), and it is easy to confuse them. When ontogenetic transformations are treated as characters, ontogenetic and evolutionary polarities are clearly distinguishable since they exist at different hierarchical levels: ontogenetic polarities exist within characters, evolu- tionary polarities between them. Given the validity of the arguments pre- sented above, one gains some insight into possible reasons that proponents of the ontogenetic method adopt certain other positions. Some proponents of the onto- genetic method suggest that an evolution- ary basis for systematics is superfluous (Platnick, 1979; Nelson and Platnick, 1981; Patterson, 1982, 1983). This view stems from a confusion of ontogeny and phylog- eny, a confusion that can only result when instantaneous morphologies rather than ontogenetic transformations are used as characters. When one bears in mind that ontogenetic polarities differ from phylo- genetic polarities, it is evident that, al- though ontogenetic transformations pro- vide direct evidence about ontogenetic polarities within-characters, they are un- informative about the phylogenetic polar- ities between characters. The reliance of at least some systematists on the ontogenetic method in the first place seems related to other aspects of their character concepts. Although they accept instantaneous mor- phologies as characters, they see a trans- formational relationship among these characters (Platnick, 1978; Nelson and Platnick, 1981; Nelson, 1985). But if one removes the evolutionary basis from sys- tematics, then ontogeny is the only con- text in which a transformational relation- ship among characters is meaningful. The development of pattern cladistics was ob- viously complex, and the end result may be internally consistent. However, as a system for illuminating only the ontoge- netic relationships among instantaneous morphologies, there is little reason to in- terpret its results in the context of organ- ism phylogenies. Viewing ontogenetic transformations rather than instantaneous morphologies as characters also has consequences for phy- logenetic systematics. Under this charac- ter concept, the outgroup method is still applicable. The paleontological method is most profitably viewed as a special case of the outgroup method, and one in which the characters (transformations) are gen- erally poorly known. Although there is no longer an ontogenetic method for polarity determination, ontogenetic transforma- tions lose none of their importance since they now form the basis of characters. I do not attempt to work out all the impli- cations that viewing ontogenetic transfor- mations as characters has for the practice of phylogenetic systematics. Nevertheless, I want to discuss some points in anticipa- tion of criticism. For organisms with on- togenies, every morphological character can potentially be traced back to the zy- gote. This does not mean that the entire organism must be treated as a single char- acter, for branching in developmental pathways allows one to recognize separate characters just as one is able to recognize separate lineages of organisms even though they too stem from a single com- mon ancestor. The question then is "How large a segment of ontogeny constitutes a systematic character?" If systematic char- acters are defined as features of organisms that are used to determine the relation- ships among these organisms, then the an- swer to the question is "large enough to encompass variation that is potentially in- formative about the relationships among the organisms being studied." Therefore, there is no need for the character to in- 294 SYSTEMATIC ZOOLOGY VOL. 34 elude all the parts common to the ontog- enies of all the organisms under study. The criticism might be raised that now we are back where we started?comparing parts of ontogenies. But my point is not that we must know the ontogenies fully before we can do systematics. Instead, I want to emphasize that acceptance of the proposition that phylogeny is a sequence of ontogenies should affect our views about what kinds of attributes of organisms pro- vide information about phylogenetic re- lationships among these organisms. This perspective reveals that although a knowl- edge of ontogeny is of the utmost impor- tance in phylogenetic systematics (since ontogeny is what is modified during the course of phylogeny), the sequence of on- togenetic transformation is uninformative about phylogenetic character polarities. Using the previous example (some organ- isms exhibiting an ontogenetic transfor- mation from X to y, others exhibiting x but no further transformation), it is unclear whether organisms lacking y exhibit an ancestral or a derived condition (see Lundberg, 1973). Given the possibility of paedomorphosis, the absence of a feature can be a derived character (contrast with Nelson, 1978:340). If the absence of a fea- ture can be a character, then it is valid to use instantaneous morphologies as char- acters in phylogenetic systematics. The reason I have chosen to use ontogenetic transformations as characters instead is that this position embraces the distinction be- tween ontogeny and phylogeny. I next ex- plore some further consequences of the view that ontogenetic transformations are characters. Homology and synapomorphy.?Patterson (1982), developing an idea implicit or ex- plicit in the writings of many previous au- thors (references cited in Eldredge and Cracraft, 1980:36; Patterson, 1982), equat- ed homology and synapomorphy. In a phylogenetic context, the logic behind this equivalency is as follows: Homology, as similarity inherited from a common ancestor, includes symplesiomorphy and synapomorphy. But plesiomorphy and apomorphy are relative concepts; a sym- plesiomorphic homology is a synapomor- phy when viewed at a more inclusive hi- erarchical level. Therefore, the concept of synapomorphy subsumes the concept of symplesiomorphy, and homology equals synapomorphy. Although equating homology and syn- apomorphy may at first seem ? logical con- sequence of adopting a phylogenetic def- inition of homology, I feel that a strict equation of these two concepts (in the case of characters that have ontogenies) is based on a confusion of ontogeny with phylog- eny, a confusion rooted in traditional character concepts. Homology, as used by most authors, describes a relationship among instantaneous morphologies. Such instantaneous forms are parts of ontoge- netic transformations, which are in turn parts of organism phylogenies. For this reason, ontogenetic transformation of one instantaneous form into another establish- es their homology. Synapomorphies, on the other hand, are the characters of monophyletic groups; that is, synapomor- phies are characters placed on cladograms or phylogenetic trees where they exist as evolutionary novelties. But if my previous argument is accepted, then (regardless of what has been done in practice) ontoge- netic transformations, rather than instan- taneous morphologies, are the characters of monophyletic groups. Therefore, in contrast with homology, which describes a relationship between instantaneous forms, synapomorphy describes a relation- ship between ontogenetic transforma- tions. Homology does not equal synapo- morphy. This perspective should clarify what in- ferences can and cannot be made about or- ganisms whose ontogenies are poorly known. For example, suppose that a fossil possesses instantaneous morphology y. The ontogenetic transformation x -> y es- tablishes the homology of x with y. If the transformation x -> y is determined to be derived relative to the ontogeny in which X does not transform (on the basis of out- group comparison), then it is reasonable to consider the fossil to be the remains of an organism that possessed the transform- ing ontogeny or a modification of it. How- ever, if the fossil possesses instantaneous 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 295 morphology x, it cannot be determined whether this fossil represents an organism that belongs to the clade diagnosed by the transforming ontogeny, for both ontoge- nies contain this instantaneous morphol- ogy- Ontogeny and parsimony.?Nelson and Platnick (1981; see also Nelson, 1973a, b; Lundberg, 1973; Maddison et al., 1984) jus- tified the ontogenetic method on the grounds of parsimony: Suppose . . . that we study the ontogeny of species A and B, and discover that early in development both species have pharyngeal gill slits, which sub- sequently either remain as slits (species A) or close (species B). The young stages of these species ob- viously share a . .. general character (gill slits); in a phyletic context, the presence of gill slits in young stages can be considered primitive for the two species. But what about the adults? There are two possibilities: (1) the presence of gill slits in adults, as in young stages, is primitive, or (2) the presence of gill slits in adults is derived. With ref- erence to species A and B, possibility (1) requires that species B has gained an attribute (closed gill slits in adults) that was lacking in the common ancestor, whereas possibility (2) requires that species A has lost an attribute (closed gill slits in adults) that was present in the common ancestor. But of course, for species A to have lost the attri- bute, it must first have been acquired by the com- mon ancestor, and that required prior gain of the attribute is equivalent to the entire change im- plied by possibility (1). Thus these possibilities can be diagrammed: [diagram in text redrawn here as Fig. 8] and it can be seen that possibility (2) in- volves both character transformations required by possibility (1): the acquisition of gill slits, and the acquisition of closed gill slits in adults, plus a third character transformation, the acquisition of open gill slits in adults. Possibility (1) is more parsi- monious, and can be preferred on that basis [Nel- son and Platnick, 1981:37-38]. This justification rests on a confusion between ontogeny and phylogeny, which in turn rests on a concept of instantaneous morphologies as characters. The validity of Nelson and Platnick's argument rests on the assumption that the occurrence of gill slits must precede the closure of these slits. In ontogeny, this is obviously true. In phylogeny, however, it is true only in the most trivial sense: the first organism that closed its gill slits must have had them open earlier in its own ontogeny. When the organism, or life cycle, is considered as a whole, there is no reason to assume that a given structure must be preceded in A B slits ?? stits slits ??'- ?? closed slits closed in adults A B slits ??? slits slits ?? closed stits open in adults O slits closed in adults stits FIG. 8. Dendrograms used to illustrate justifica- tion for the ontogenetic method on the grounds of parsimony (redrawn from Nelson and Platnick, 1981: 38). phylogeny by its ontogenetic precursor. In other words, there is no reason to assume that the first organism to bear gill slits could not have closed them later in its on- togeny. Thus both hypotheses require an equal number of changes (Fig. 9) Although the preceding argument un- dermines Nelson and Platnick's (1981) parsimony justification for the ontogenet- ic method in the context of organism phy- logenies, parsimony can still be used to justify this method in the context of the phylogenies of instantaneous forms. Sup- pose that organisms in taxon A have a nontransforming ontogeny for character (instantaneous form) x and those in taxon B exhibit an ontogenetic transformation from X to y. (I avoid the example of gill slits and closed gill slits since the very de- scription of the character "closed gill slits" prohibits exceptions to von Baer's second law.) We wish to know which character, x or y, came first in phylogeny. The two possibilities are diagrammed in Figure 10 (A, B). If X is ancestral to y, a single phy- logenetic transformation is required to ac- count for the observed ontogenies (Fig. lOA). If y is ancestral to x, two changes are required (Fig. lOB). Thus, it is more par- simonious to consider x ancestral to y in that this hypothesis requires fewer phy- logenetic transformations (or in that it does 296 SYSTEMATIC ZOOLOGY VOL. 34 slits closed in adults stits that remain open slits open in adults slits that close in adults FIG. 9. Parsimony and the ontogenetic method. Since there is no reason to assume that the first or- ganism that had gill slits could not have closed them later in its ontogeny, possibilities (1) and (2) require an equal number of phylogenetic transformations. not require the postulation of hypotheti- cal organisms). However, as noted earlier, this conclusion results not from the se- quence of ontogenetic transformation but from the generality of characters. It is more parsimonious to consider x ancestral to y even if the ontogenetic transformation in taxon B is from y to x (Fig. IOC, D). Such a parsimony justification for the ontogenetic method applies only to deter- mining the polarities of instantaneous morphologies and not to the polarities of ontogenetic transformations, the appro- priate characters for examining phyloge- netic relationships among organisms. To attempt such an extrapolation is to confuse ontogeny with phylogeny. Since the ori- gin of X and y in phylogeny is potentially simultaneous (i.e., within the life cycle of a single organism), there is no need to postulate the first step (y -> y) in Figure lOB and D. Simply by asking whether x or y came first in phylogeny one has ruled out the possibility that they arise simulta- neously, and of course in ontogeny they do not. Under the ontogenetic character con- cept advocated in this paper, such parsi- mony justifications involving only two taxa are simply not applicable. Using the above example, taxon A and taxon B have differ- ent ontogenetic transformations and thus different characters. Given-this informa- tion alone, parsimony provides no reason for inferring that either character (trans- formation) is ancestral to the other. Of course, additional information may be available. This is where outgroups and parsimony are relevant. Semaphoronts.?Hennig (1966:6) defined the semaphoront as "the organism .. . dur- ing a certain, theoretically infinitely small, period of its life." He considered the se- maphoront, rather than the organism, to be the basic element of systematics. Ac- cording to Wiley (1981:119), systematics should involve the comparison of "com- parable semaphoronts," that is, specimens at similar stages in their life history. This is similar to Gould's (1977:212) suggestion that phylogeny be depicted as a sequence of organisms at comparable stages of de- velopment, traditionally adults. What are "comparable" stages? To a sys- tematist, the attainment of adulthood (maturation of the gonads) is itself a char- acter, and one whose timing can presum- ably change during phylogeny (e.g., pro- genesis of Gould [1977] and Alberch et al. [1979]) much like others. Furthermore, many developmental changes occur after the maturation of the gonads, so that even adults may not be comparable. If one must choose a single transformation as the stan- dard for comparison, attainment of adult- hood is a convenient reference point, especially for those studying paedomor- phosis. This is the one character that would seemingly never be paedomorphic itself. Paedomorphosis in the character "attain- ment of sexual maturity" would be the ul- timate evolutionary mistake. Neverthe- less, choosing any one transformation as the standard for comparison is artificial, and given that all transformations have the potential to change their timing relative to others, this practice gives no guarantee of designating comparable semaphoronts. The very practice of using semapho- ronts (whether comparable or not) as the basic units of systematics necessitates us- ing instantaneous morphologies as char- acters?a practice that I argue against. But even under the view of ontogenetic trans- 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 297 formations as characters, the semaphoront is a valuable concept. It serves to remind us that although the phylogenetic system- atist attempts to determine relationships among organisms as ontogenies, or life cycles, in practice only semaphoronts are available for study (compare Danser, 1950). I disagree with Patterson's (1983) claim that ontogenetic transformations are empiri- cisms while phylogenetic transformations are generalizations. Although some onto- genetic transformations are potentially amenable to direct observation, in practice nearly all "observations" of ontogenetic transformations are generalizations from semaphoronts. HISTORY AND THE RELEVANCE OF ONTOGENETIC TRANSFORMATIONS After submission of this paper, Gareth Nelson brought to my attention three manuscripts, which have now been pub- lished, dealing with ontogeny and char- acter polarity. Brooks and Wiley (1985) and Kluge (1985) criticized the ontogenetic method, and Nelson (1985) responded to the criticisms. Except for the addition of scattered references, I have not been able to address these papers without drastically modifying my own. Nevertheless, many of the disagreements raised in the papers by Brooks and Wiley, Kluge, and Nelson embody problems that I have attempted to clarify. I hope that my paper will be seen as a means of understanding the bases of the disagreements and perhaps reconcil- ing divergent opinions. On the surface, my position may appear more critical of the position taken by Nel- son (1985) than of those taken by Brooks and Wiley (1985) and Kluge (1985). How- ever, I feel that most of the differences be- tween Nelson's position and mine stem from a single basic philosophical differ- ence, for if I accept his view of the rela- tionship between evolution and system- atics I find little to disagree with. This view that evolution is a generalization from systematics (Nelson, 1978, 1985) coupled with a view of science that rejects induc- tive methods (Popper, 1968) may explain why Nelson, Platnick and others attempt to dissociate systematics from evolution. x-*y [B] x-*x (A) A B X (B) x^-x (A) x-^x (A) x?y (B) yi^y x-^x (A) X (B) D FIG. 10. Parsimony and the phylogeny of instan- taneous morphologies. If it is assumed that x and y do not originate simultaneously in phylogeny, then it requires fewer phylogenetic transformations to suppose that the less general morphology is de- rived?regardless of the sequence of ontogenetic transformation. Parenthetical taxa next to ontogenet- ic transformations mean only that such taxa possess the specified ontogeny, not that they are actual ancestors. Horizontal arrows represent ontogenetic transformations; vertical arrows represent phyloge- netic transformations. My own philosophy of science is some- what different, finding a role for induc- tion (in generating theories) as well as de- duction (in testing them). In this light, I would like to address two issues raised by Nelson (1985) that were apparently not treated to his satisfaction by Brooks and Wiley (1985) or by Kluge (1985): history; and the relevance of ontogenetic transfor- mations. I agree with Nelson (1973a; see also Gould, 1977; Patterson, 1983) that knowl- edge of the sequence of ontogenetic trans- formations played an important role in the history of systematics. I also agree that the theory of evolution was developed partly as a generalization from systematics (com- pare Nelson, 1978, 1985; Patterson, 1983). 298 SYSTEMATIC ZOOLOGY VOL. 34 Nevertheless, some theories resist falsifi- cation so persistently or their explanatory power is so great that their roles change during the course of history. For the case of evolution and its relationship to sys- tematics, Ghiselin (1969:83) described this change as follows: Darwin solved the problem [inherent in defining the term "natural" solely on the basis of meta- physical posits] by redefining "natural" as deriv- ative of the mechanism which underlies what was previously a mere empirical generalization about observed properties of organisms. The change he made exemplifies a basic shift in attitude. Instead of finding patterns in nature and deciding that because of their conspicuousness they seem im- portant, we discover the underlying mechanisms that impose order on natural phenomena, wheth- er we see that order or not, and then derive the structure of our classification systems from this understanding. The difference, then, lies with the decision as to what is important. It reflects a basic gulf in attitude separating idealists given to the older forms of induction, on the one hand, and empiricists who employ the hypothetico-deduc- tive method, on the other. Classification ceased to be merely descriptive and became explanatory. Ghiselin's optimism notwithstanding, the transition of evolution from theory to axiom in the context of systematics is far from complete. Two notable exceptions come to mind: first, the attempt during the evolutionary synthesis to redefine species as evolutionary units (e.g., Simpson, 1961; Mayr, 1969); and, second, Hennig's (e.g., 1966) attempt to redefine various system- atic terms and formulate systematic meth- ods as deductions from the concept of evo- lution. My paper is an attempt to make this deductive approach explicit. I am now prepared to answer Nelson's (1985) question: "What is the relevance for systematics of ontogenetic character trans- formations?" Once the axiomatization of evolution in systematics is accepted, the answer is simple: characters do not trans- form in ontogeny; ontogenetic transfor- mations are themselves the characters. CONCLUSION I have argued that certain problems in polarity determination are related to cur- rent character concepts. Perhaps these character concepts are tied to an emphasis on the explanation of adult form, an em- phasis that predates acceptance of evolu- tion as the explanation for organic diver- sity. Griffiths (1974:99) traced it back to Aristotle. An emphasis on adult form tends to focus attention on terminal ontogenetic stages and thereby encourages the use of instantaneous morphologies as characters. Another part of the problem may result from the long common history shared by systematics and comparative anatomy. These two disciplines overlap one another extensively, but perhaps not enough at- tention has been paid to their differences. An example of this neglect is the failure of comparative biologists to distinguish between two different goals?reconstruct- ing the relationships among organisms, and reconstructing the relationships among certain instantaneous morpholo- gies, usually those of particular organs or organ systems. Both kinds of studies are commonly carried out in three different ways: (1) through the comparison of the adult stages of organisms belonging to dif- ferent taxa, (2) through the study of fos- sils, and (3) through the study of ontog- eny. This tripartite approach is embodied in Agassiz's threefold parallelism among comparative anatomy, paleontology, and embryology. Unfortunately, the traditional emphasis on adult form is at odds with the goals of phylogenetic systematics. An axiom of this methodology is phylogeny itself, and phylogeny is a sequence of ontogenies, of which the adult is but a segment. If the current emphasis on adult form can be re- placed with a more complete acceptance of the ontogenetically dynamic nature of organismal morphology, our character concept will also change. Then it will be seen that the threefold parallelism is based on an artificial separation of disciplines, for comparative anatomy, paleontology, and embryology along with systematics make up a single comparative method uni- fied in the organism by the concept of evolution. ACKNOWLEDGMENTS I thank G. Nelson and C. Patterson for writing the stimulating papers that challenged me to think about the ontogenetic method. A. Bauer, M. Donoghue, J. 1985 ONTOGENETIC METHODS AND PHYLOGENETIC SYSTEMATICS 299 Gauthier, A. Larson, L. Parenti, K. Thomas, and D. Wake provided helpful comments on earlier drafts of this paper. Reviews by W. Fink and G. Nelson were both detailed and constructive. I am especially grateful to my friend J. Gauthier for countless dis- cussions that influenced many of the ideas embodied in this paper and to W. Maddison for his extremely thorough and insightful reviews. None of these peo- ple are responsible for the flaws that undoubtedly remain. REFERENCES ALBERCH, P., S. J. GOULD, G. F. OSTER, AND D. B. WAKE. 1979. Size and shape in ontogeny and phylogeny. Paleobiology, 5:296-317. ARNOLD, E. N. 1981. Estimating phylogenies at low taxonomic levels. Z. Zool. Syst. Evolutionsforsch., 19:1-35. BEATTY, J. 1982. Classes and cladists. Syst. Zool., 31: 25-34. DE BEER, G. R. 1940. Embryos and ancestors. Oxford Univ. Press, London. BOCK, W. J. 1974. Philosophical foundations of clas- sical evolutionary taxonomy. Syst. Zool., 22:375- 392. BROOKS, D. R., AND E. O. WILEY. 1985. Theories and methods in different approaches to phylogenetic systematics. Cladistics, 1:1-11. DANSER, B. H. 1950. A theory of systematics. Bibl. Biotheor., 4:117-180. ELDREDGE, N., AND J. CRACRAFT. 1980. Phylogenetic patterns and the evolutionary process. Columbia Univ. Press, New York. FARRIS, J. S. 1982. Outgroups and parsimony. Syst. Zool., 31:328-334. FINK, W. L. 1982. The conceptual relationship be- tween ontogeny and phylogeny. Paleobiology, 8: 254-264. GAFFNEY, E. S. 1979. An introduction to the logic of phylogeny reconstruction. Pages 79-111 in Phy- logenetic analysis and paleontology (J. Cracraft and N. Eldredge, eds.). Columbia Univ. Press, New York. GHISELIN, M. T. 1969. The triumph of the Darwinian method. Univ. California Press, Berkeley. GOULD, S. J. 1977. Ontogeny and phylogeny. Har- vard Univ. Press, Cambridge, Massachusetts. GRIFFITHS, G. C. D. 1974. On the foundations of bio- logical systematics. Acta Biotheor., 23:85-131. HENNIG, W. 1966. Phylogenetic systematics. Univ. Illinois Press, Urbana. DE JONG, R. 1980. Some tools for evolutionary and phylogenetic studies. Z. Zool. Syst. Evolutions- forsch., 18:1-23. KLUGE, A. G. 1985. Ontogeny and phylogenetic sys- tematics. Cladistics, 1:13-27. L0VTRUP, S. 1978. On von Baerian and Haeckelian recapitulation. Syst. Zool., 27:348-352. LuNDBERG, J. G. 1973. More on primitiveness, higher level phylogenies and ontogenetic transforma- tions. Syst. Zool., 22:327-329. MADDISON, W. P., M. J. DONOGHUE, AND D. R. MAD- DISON. 1984. Outgroup analysis and parsimony. Syst. Zool., 33:83-103. MAYR, E. 1969. The biological meaning of species. Biol. J. Linn. Soc, 1:311-320. NELSON, G. 1973a. The higher level phylogeny of vertebrates. Syst. Zool., 22:87-91. NELSON, G. 1973b. Negative gains and positive loss- es: A reply to J. G. Lundberg. Syst. Zool., 22:330. NELSON, G. 1978. Ontogeny, phylogeny, paleontol- ogy, and the biogenetic law. Syst. Zool., 27:324- 345. NELSON, G. 1985. Outgroups and ontogeny. Cladis- tics, 1:29-45. NELSON, G., AND N. PLATNICK. 1981. Systematics and biogeography. Columbia Univ. Press, New York. PATTERSON, C. 1982. Morphological characters and homology. Pages 21-74 in Problems in phyloge- netic reconstruction (K. A. Joysey and A. E. Friday, eds.). Systematics Association Special Volume No. 21. Academic Press, London. PATTERSON, C. 1983. How does phylogeny differ from ontogeny? Pages 1-31 in Development and evolution (B. C. Goodwin, N. Holder, and C. C. Wylie, eds.). British Society for Developmental Bi- ology Symposium 6. Cambridge Univ. Press, Cam- bridge. PLATNICK, N. I. 1978. Gaps and prediction in clas- sification. Syst. Zool., 27:472-474. PLATNICK, N. I. 1979. Philosophy and the transfor- mation of cladistics. Syst. Zool., 28:537-546. PLATNICK, N. I., AND E. S. GAFFNEY. 1977. System- atics: A Popperian perspective. Syst. Zool., 26:360- 365. POPPER, K. R. 1968. The logic of scientific discovery. Harper and Rowe, New York. RiEPPEL, O. 1979. Ontogeny and the recognition of primitive character states. Z. Zool. Syst. Evolu- tionsforsch., 17:57-61. SIMPSON, G. G. 1961. Principles of animal taxono- my. Columbia Univ. Press, New York. STEVENS, P. F. 1980. Evolutionary polarity of char- acter states. Annu. Rev. Ecol. Syst., 11:333-358. VAN VALEN, L. 1978. Why not to be a cladist. Evol. Theory, 3:285-299. VOORZANGER, B., AND W. J. VAN DER STEEN. 1982. New perspectives on the biogenetic law. Syst. Zool., 31:202-205. WATROUS, L. E., AND Q. D. WHEELER. 1981. The out- group comparison method of character analysis. Syst. Zool., 30:1-11. WHEELER, Q. D. 1981. The ins and outs of character analysis: A response to Crisci and Stuessy. Syst. Bot., 6:297-306. WILEY, E. O. 1975. Karl R. Popper, systematics, and classification?A reply to Walter Bock and other evolutionary taxonomists. Syst. Zool., 24:233-243. WILEY, E. O. 1981. Phylogenetics. John Wiley and Sons, New York. Received 4 January 1985; accepted 5 April 1985.