Ckdislitx I't'tti U:379 386 BRIDGES BETWEEN EVOLUTIONARY PATTERN AND PROCESS Jonathan A. Coddington' 1 Department ?/ Entomology, \ alumni Museum of .Xalural History, Smithsonian Institution. Washington. !)(. 20M), I ..S..I. "Musi ul ihr riiMiiictivr rharai icrisiiis of higher laxa an* surHv adaptations tluii evolved l>v natural M'li'iiiiiii..." htunsm.i l!)79: +34 . All ul* liinlotrv i omprK us In mobilize th.ti tir^ailisni.i an- nut tipiimaUv designed, tlial mativ Iraturrs .nc mil adaptive, .nut ili.n species mav diller lm reasons other than natural srlii'iion." r'utuvnta I'lHli; _>.*>4 . Uc Pinna and Sallcs ask lor clarification on how cladistic analysis tests adaptational hypotheses de Pinna and Sallcs. 1990; Coddington, 1988 . They also [joint out that an alternative optimization exists lor character 3 in my 19881 fig. 2. Ynp. OK, I'll admit it. All readers of the previous paper are hereby instructed to believe that a new laxon has been discovered, identical to taxon D except for some convincing autapomorphies which need not concern us. Based on this new information, the previously reported optimization wins by one and only one step. Hey, close don't count in fladistics! Passing on to a more profound.issue. I have to confess a long-standing difficulty with the allegedly important distinction between "states of becoming" and "states of being". II something already is. I would guess it probably became in order to be. and if it is only becoming, it had to become from somewhere, and probably to somewhere as well, not to mention be while it is becoming. I am not sure whether I view adaptation as becoming rather than being or whether the distinction makes a difference. Mosi of de Pinna and Salles's remaining arguments tend to the same point: however desirable, no amount of mapping characters on cladograms has yet provided an "empirical bridge", "bridge concept", "empirical link", or even "true and empirical relations" to processes such as natural selection or process theories such as adaptation. Although I think the jargon peculiar to theories about theories often obfuscates simple issues anyone care to define "individual" in 50 pages or less.3 . I hope I understand more or less what they mean by "empirical bridge". In fact. I think the argument reduces to the question. "What is a bridge?" When bushwhacking through chasm-filled forests real or theoretical . people do differ on what ihey accept as trustworthy bridges (e.g. rock hoppers, waders. fallen tree walkers, vine swingers . 11 is a fair question. I will first summarize what I said, and then use the rhinoceratid example, or at least a fictive extrapolation of it. to demonstrate the tests. To review the backbone of the 1988 argument: 1. Adaptation is generally agrted to be the modification of existing features by natural selection so as to increase fitness of the organism. 2. In cladistic terms, the novel modification is an apomorphv. Hence adaptations ari' a special class of apomorphv. and the study or analysis of their historical dimension requires recognition as such. n7 lit- 51X17" '?MP|U)7'I * nXSiri'im " ? I'"'0 'IT*' Will! Hrnnig Sneirtv 380 J. A. CODDINGTON 3. Adaptation makes at least two special claims about apomorphies. la) Natural selection is responsible for their origin the initial role of mutation aside i. ib) The novel modification (apomorphy) is better than the pre-existing feature plesiomorphy). Better means that increased fitness is conferred by the apomorphy relative to the plesiomorphy. 4. Selection operates on the performance (broadly construed | of the apomorphy in its environment (broadly construed). 5. Taken together, points 2-4 imply a variety of ways in which predictions deduced from adaptational hypotheses can be compared to real, yes, empirical, data on performance effect on litness or its corollaries, reproductive success, efficiency of design or function, etc.j and character distributions. 6. Other things being equal, it helps if the adaptational hypothesis claims that selection is responsible lor both the origin and maintenance of the feature. Otherwise the hypothesis eschews a uniformitarian assumption about history. That makes any lest of an historical hypothesis with current data ambiguous. Requirements of the claim of origin and maintenance can be relaxed, but doing so sacrifices some measure of testability Coddington 1988: 1 7-19, fig. 8). In my view surviving such a series of tests does not "prove" adaptation, it just fails to disprove it. I, at least, am comfortable with identifying such a procedure as a test, and calling it science. These comparisons are without doubt tests, and rigorous ones at that. I demonstrated that several classical adaptationist stories fail to survive these trials. Given 1-4, what, precisely, is adaptation? It seems that the theory of adaptation contributes two unique components to the explanation of the origin and persistence of apomorphies. These are the stipulation that natural selection is involved at least in the origin of the feature, and a prediction about the performance, or function, of the apomorphy relative to the plesiomorphy. Hence the most restrictive cladistic definition of adaptation would be apomorphic function due to natural selection. It is worth reiterating here that competent adaptationist hypotheses will generally have the form: "the derived trait M , arose at time (/) in the stem lineage of taxa C, D and E via selection for the derived function, F,, with respect to the primitive trait \10 with primitive function F() in taxa A and B Coddington 4988: 4 and fig. I; see below as welli. However, adaptationist theory rarely has, and need not, couch its hypotheses in cladistic terms point 2, above), although my whole discussion tried to emphasize the scientific efficiency that would result if it did. Personally, I think adaptationist hypotheses concerning events at or above the species level that ignore cladistic context do not deserve the name; others may feel that this is too restrictive. At a minimum, however, even adaptationists wholly ignorant of cladistic methods would require a precise description of the putative adaptation, a measure of its performance, and predicted outcomes of comparisons between organisms or populations bearing the adaptation and control groups. As de Pinna and Salles point out, some of the tests I outlined address whether any- given adaptational hypothesis makes sense in a cladistic context. Whether these stipulations are regarded as tests, or merely as necessary prerequisites for admission to the company of competent scientific hypotheses, is a quibble. Because they can cause hypotheses to be rejected outright or seriously modified, they do function as tests. Also with regard to point 2, I pointed out early on that "adaptation is one possible V FORL'M: PATTERN-PROCESS BRIDGES 581 general explanation of empirically observed -A napomorphy among many*'. Thereafter I focussed on alleged adaptions! explanations ofsynapomorphies as a subset of all possible explanations of synapomorphies. To say that I operationally "equated" adaptation and synapomorphy is therefore an error. Because all hypotheses about the origin of adaptations must concern apomorphies. I discussed adaptations from that perspective. With regard to point 4 above. I am puzzled that de Pinna and Salles are puzzled that I placed ?emphasis on natural selection as a necessary component in the definition ol adaptation". Adaptation um natural selection at some level or at some point disembowels the Darwinian explanation for descent with modification. I agree that natural selection as a theory has nothing to do with cladistic methods. Their mutual independence gives strength to the test. I hope no more needs to be said on that point. In contrast to the methodological nature of cladistics, "process theories" are causal explanations of evolutionary change. They make predictions about how and why particular features change through evolutionary time. Cladistics outlines the structure of the test to assess the truth of that statement, and the process theory explains how and ?before" changed into ""after". I restricted the argument to change involving peciation. in which case the "before" and "after" versions of the feature are, in systematic terms, homologous characters or alternate "states" of a homolog. I did this mainly because explaining the existence or persistence ofsynapomorphies is a central question in macroevolution, and because cladistics presumes hierarchy which, with few- exceptions, is hard to document within demes. I very much doubt if adaptation is the sole .tnser to the big question. Significantly, cladistics has nothing to say on the subject, nor should it. given that cladistics is merely a method. That is both its strength and ultimately a source of sterility Coddington. 1985.1. In m\ view, therefore, the most testable theories about adaptations at the species level or above oiler detailed explanations for the origin and maintenance of apomorphies. It is the responsibilitv of the adaptationist hypothesis not the cladistics) to specify as exactly as possible the nature of the adaptation, an appropriate measure of function, and to predict the outcome of pair-wise comparisons between taxa bearing the derived and the primitive trait. Many extant' hypotheses of adaptation are of this form, even if the outgroup. the primitive trait, and the function of the primitive trait are implicit. (?iven the abo\!e. the bridge requested by de Pinna and Salles seems obvious to me. If natural selection,, or any process theory, predicts its own particular polarity or ordination of evolutionary change, and if cladistics provides an independent polarity, then fruitful science can be done by investigating the whys and wherefores of agreement or disagreement between predictions of process theories and observed phylogenetic pattern. The possibility of conflicting polarities is a test. If the polarities agree in all respects, the predictions of the process theory have survived the test. The theory may be elaboraied and applied to other test cases in order to assess its generality. Further, perhaps ingenious, deductions can be made about the relation between the posited cause of the diici ted change morphological, physiological, behavioral, genetic, biochemical, etc. . and these new deductions can be tested in the same taxa. If the predictions do not survive the test, the prediction as slated is falsified. In my experience, failed predictions .ire rarely accepted as inflicting mortal wounds on their parent theories. The authors regroup, studv the new evidence, and issue a modified version of the theory or modified deduction from it that accounts for the previously falsifying evidence. The modified prediction is necessarilv more detailed, but that just affords further chances for new tests. 382 J. A. C:()l)l)lN(;r()N lor de Pinna and Sallcs to condemn this son of scientific progress as "ad lux" is not very realistic and perhaps extreme. In order to use the rhinoceraticl example cited In de Pinna and Salles, one has to rework it. heeanse die hypothesis presented bv I.ewontin 1978 lails even the rudimentarv test ofhomology, as I pointed out. It seems pointless to argue about what I.euontin meant; I and presumably most other readers interpreted his example 1978: 228 and figure on p. 225 to mean dial the ancestral rhino had neither the one-horned nor the two-horned condition frequency ofalleles A and B is zero at the origin, p. 225 . and that two species "evolve" = change in gene frequency along "alternative evolutionary paths'" . I.ewontin, 1978: 225' to fixation of these alleles as mutually exclusive apomorphies of distinct species. Lew on tin said predator defense. De Pinna and Salles ask how you lest it. Although two adaptive hypotheses are involved due to the mistake in homology determination i, let us focus on the less general of the two and presume that the hypothesis ol predator defense still stands lor the origin and maintenance of one horn in the Indian rhino. The cladogram is 'Diceros. (.'eralotherium , Dieerorhinus. Punjabitherium, (iaindatherium. Rhinoceros)" (Groves, 1983). The sister taxon of Rhinoceros is either the extinct two- hoi tied Punjabitherium or (iaindatherium. I assume we have no little or no reliable evidence on their predators or defensive behavior, although paleobiologists ran deduce some amazing things. The next outgroup is the two-homed Dieerorhinus. the Asian rhino not the African rhino, as I.ewontin implied', and the next is the two-horned Diceros- Ceratolherium clade. Because frontal horns are homologous as dermal outgrowths above the frontal bone and nasal horns likewise above the nasal bone, and because Rhinoceros lacks the frontal horn present in the outgroups, the hypothesis concerns secondary loss of the frontal horn, not gain of a nasal horn, as I.ewontin implied. The complete adaptive hypothesis cf. general form above therefore is: Rhinoceros lost the frontal horn present in the outgroup taxa due to natural selection acting to reduce the incidence or severity ol predatorv attacks through increased efficiency of a single horn .is a predatory defense, as compared to the plesiomorphic two-horned condition in the outgroups. First, gather information on predators of Rhinoceros and Dieerorhinus especially, but do not ignore Diceros or Ceralolhenum. as it makes a stronger case to establish the retained primitive function and context of the two-horned condition though parsimony and the doublet on the cladogram. Hope that predators of all genera are ecological or behavioral equivalents, so that from thefhino's point of view, it does not much tnatter whether they are being attacked in Africa. India, or Asia. The tests are at least the following, not counting the tests derived directly from the implied cladistie context which, bv the way. falsified the original formulation : 1. If no rhinoeeratid uses its horn at any ontogenetic stage to deter predators, stop. The adaptional In pothesis is falsified because the statement of horn function was awry. 2. If for all predators. Rhinoceros suffers more predation implying net fitness decrease than its outgroups. stop. The adaptive hypothesis as stated is falsified. because selection cannot have acted on the two-horned condition to yield the one- horned condition via a decrease in fitness. ;5, If for some predators. Rhinoceros does s u Her less predation or less serious predation sav bv tigers, since tigers are sv mpatrie v\ ith rhinos . focus on detailed analysis of the defensive behaviors of rhinos and of offensive behaviors of tigers and their FORUM: PATTERN-PROCESS BRIDGES 383 outgroups to specify exactlv how tigers and horns interact during a predatory encounter. II it turns out that Indian tigers arc apomorphic in their predatory behavior such that Rhinoceros is at less risk. stop. The adaptive hypothesis is Falsified because Rhinoceros apparently lost its frontal horn through the reduction of stabilizing selection to maintain two horns, not directional selection for a superior adaptive state. 4. If tigers have not changed their attack behavior in any significant way, find out exacly how the absence of the frontal horn enables Rhinoceros to evade predation or reduce its severity. Perhaps the frontal horn interferes with infliction of maximum damage by the nasal horn. Perhaps the frontal horn, or its proximity to the nasal, affords the tiger a belter grip on the rhino during the attack sequence. Perhaps the frontal horn area is more frequently damaged than the nasal area, and thus two-horned rhinos stiller greater mortality as a result of secondary infection. The last conjecture is intriguing because it raises the possibility that the loss of the horn may be an adaptation to prevent secondary infection, and not solely against predation. This might be especially relevant if frontal areas are injured through other agents than tiger attacks. These questions can be addressed with data, and are the stuff of good research. 5. Make all six pair-wise comparisons among taxa neglecting the fossil taxa because ue assumed that comparable data were lacking . The adaptive hypothesis as stated predicts no significant differences among any two-horned rhinos, and always a significant difference between Rhinoceros and any two- horned rhino. Deviations from these predictions may lead merely to further research on the transformation series among horns in rhinoceratids and its ecological implications, or they could weaken the adaptive hypothesis or make it equivocal. The worst case of the latter might be that if either C.eratotherium or Diceros horn function is used as the null hypothesis against which Rhinoceros horn function is judged, the adaptational hypothesis cannot be rejected, but the opposite happens it Dicerorhinus is used. Assuming that some creditable explanation emerges from the above, the adaptive hv pothesis has survived the test. One now has a lot ol data on the outgroup condition, its plesiomorphic function in a plcsiomorphic context deterrence ol predation bv tigers . and its apomorphic function in the same context. Based on competent analysis of adequate samples of predatory interactions between all species of rhino and the tiger, as well as detailed study of the use of the horn and the attack behavior of tigers and other cats, the adaptive hypothesis cannot be rejected. This is getting increasingly unrealistic because you could not possibly get the permits to do this kind of work with endangered species, to say nothing of problems of sample size....) Note that none of the above 10 or so lests emerges solelv from the cladistic side of the "bridge", although cladistics is central to each. They all derive chiefly from the adaptational hypothesis. Relative to the general statement of a competent adaptational hypothesis above . de Pinna and Salles comment 1990: . "although he allows that M mav be an inferior solution lor the design problem posed by F0, no reason is given why one should consider this situation as less likely than its inverse". I lake this to mean that they do not understand why the theory of adaptation predicts that M? should he an inferior design solution to the design problem posed by F,. The reason is simply that selection is a e 384 ]. A. CODDINGTON directional, polarized process. Adaptive innovations arise because thev are in some sense better. Also, "lie proposes that the out groups A and B in his example: are the null hypothesis to test derived state M , as an adaptation in the ingroup. Evidently this is not I he case: the out groups are the null hypothesis when proposing simply that state M , is derived relative to state M,, '. Perhaps de Pinna and Salles missed the point that the theory of adaptation makes predictions about comparisons of performance and efficiency of plesiomorphies and apomorphies. It is the posited relationship of apomorphy to apomorphic function that makes these cladistic tests more than an excercise in out group comparison or character state optimization. Cladistics, by itself, makes no predictions as to the outcome of these comparisons; adaptation does. I think I have shown how cladistics. and, most particularly the synergy of adaptation viewed in cladistic context, can be used to test adaptation. It is (airly clear what the theory ol adaptation contributes to cladistics: predictions about comparisons on which cladistics, if by cladistics we mean merely hierarchical pattern mappings, is silent. Adaptation contributes a causal explanation of apomorphy. As I said Coddington. 1985, 1988) cladistics without process theory verges on sterility. Cladistics contributes the essential independent estimate of character change, organizes the program of research on function and performance, and, crucially, identifies which taxa provide th best controls, or null hypotheses, against which data on apomorphic function should be judged. Using present data, whether on character distributions, functions, or ecological interactions, to reach back in time to reconstruct history will never achieve the ris^or of. say, experiments in which all evidence suggests that past, present, and future apples always have, do, and will lall down. History is a series of unique events and it is harder to explain unique events than replicable ones. Given that difficulty, I do not know whether de Pinna and Salles would trust themselves to the bridge outlined above or not. especially given their inclination to view anything mapped on cladograms as only pattern, and never relevant to process. The introductory quotes from the major text on evolutionary biology Futuyma, 1979, 1986 makes it clear why we should try. Anyway, the history of life is a whole lot more interesting than the direction favored by apples. Other authors Donoghue. 1989: Greene. 1986: Maddison, 1990i, have argued that convergences, parallelisms, and other such instances of homoplasy can shed light on features as adaptations. In a particular way. I agree, and certainly would not dismiss studies of exceptional or surprising instances of convergent evolution if carried out carefully and logically. Indeed, the homology approach and the convergence approach relate as part to whole. Studies of homoplas\ can build to a generality that studies of unique events never will. There are, however, several difficulties with exclusive reliance on an approach tint not only requires homoplasy, but whose strength is actually proportional to homoplasv. The above authors did not advocate exclusive reliance on studies ol homoplasy, but it seems worthwhile to point out some of the pitfalls. One supposed strony; point of the convergence approach is that each instance of homoplasy on the cladogram is regarded as independent, and therefore statistical procedures can be applied to test lor significance I Maddison, 19901. Leaving aside the issue of whether the assumption of independence is well-conceived, the strength of the convergence approach is proportional to the number of independent trains, transformations, or losses on the cladogram. This result is at least ironic: the worse the fit of the character is to the cladogram. the better chance it has to be accepted as an adaptation. Since statistics on small sample sizes is difficult, the method may impel one to broaden the definition of the feature, a slipperv slope at best. FORUM: PATTERN-PROCESS BRIDGES 385 Second, even supposing a cogent definition and an adequate sample size, a method that ignores the homology approach i.e.. ingroup-outgroup comparisons of function, as described above can show only superficial correlation as the result of the analysis. Nothing thus far has provided evidence that adaptation explains the homoplasy better than other possible explanations ol evolutionary change. Adaptation remains untested. In order to link the theory of adaptation to apomorphy, it seems to me that one must have evidence on function. Strictly speaking, to have two or more instances of adaptation, one needs evidence on function in each. Third, even supposing a statistically significant association between characters or whatever, the homology approach can falsify any single instance counted as confirmation by the convergence approach. A conclusion based solely on the observation of convergence will always be vulnerable to test by the homology approach, whereas the opposite is less likelv. Even though nine out of 10 instances of convergence fail the homology-based tests, the tenth cannot have less status as an adaptation only because it is unique. If one insists on replication of adaptation in order to invoke it as an explanation of evolutionary history, one places the study of unique synapomorphies by far the largest class of interesting and significant evolutionary events; bevond the reach of science. Luckily the homology approach and the convergence approach are compatible. Compilations of careful research on apomorphies similar enough to justify the term "convergent" may substantiate general evolutionary trends. At the other extreme, no amount ol hastily compiled and superficially considered "convergences" can sum to a convincing demonstration of adaptive value if each instance fails the sort of tests outlined above. After a history of work on adaptation marked by sloppy thinking and armchair speculation Goud and Lewontin, 1979i, the theory of adaptation as a general explanation ofdescent with modification perhaps has more need of meticulous attention to detail than broad but shallow demonstrations of correlation. However, the approaches complement each other: both need precise and independently cor- roborated cladistic context to function. Acknowledgments I thank Charles Griswold, Michael Donoghue, and an anonymous reviewer for omments and help in improving the manuscript. REFERENCES CODDINGTON. J. A. 1985. Review of The Explanation of Organic Diversity: The Comparative Method and Adaptations for Mating. Cladisiics I: 102-107. CoDDtXC.TON. J. A. 1988. Cladistic tests of adaptational hypotheses. Cladistirs +: 1-20. OF. PINNA. M. C. C. AND SAII.ES. L. O. 1990. Cladistic tests of adaptational hypotheses: a reply to Coddington. Cladistirs 6: 373-377. DoNOGHL'E. M. J. 1989. Phytogenies and the analysis of evolutionary sequences, with examples from seed plants. Evolution 4:5: 1 137 ? 1 156. I't'TCYMA. I). J. 1979. Evolutionary Biology. Sinauer. New York. FCTL'YMA. D. J. 1986. Evolutionary Biology. 2nd cdn. Sinauer. New York. Goi'LD. S. J. AND R. C. LEWONTIN. 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proc. R. Soc. Loud. B. 205: 581-598. GREENE, H. VV. 1986. Diet and arborcality in the emerald monitor. I'aranus prasinus, with comments on the study of adaptation. Fieldiana Zool. 31: 1-12. 386 j A. CODDINGTON GROVES. C:. P. 198:5. Phytogeny <>l the living species of rhinos. Z. Zool. Svst. Evolut -lorsch 21- 293-313. I.KWOMIN. R. C. 1978. Adaptation. Sci. Amcr. 2:59: 212-230. MADDISOS. \V. I> 1990. A method lor testing the correlated evolution of two binary characters: Arc mains or losses concentrated on a certain branch of a phvlogenctic tree? Evolution 44: 539-557. f Received for publication 1 May 1990; accepted I July 1990)