Volume 77, 4 2002 The Quarterly Reviewibdev.berkeley.edu › labs › wake ›...

Volume 77, No. 4 December 2002

383

The Quarterly Review of Biology, December 2002, Vol. 77, No. 4

Copyright � 2002 by The University of Chicago. All rights reserved.

0033-5770/2002/7704-0001$15.00

The Quarterly Reviewof Biology

INTEGRATING HISTORICAL AND MECHANISTIC BIOLOGYENHANCES THE STUDY OF ADAPTATION

Kellar Autumn

Department of Biology, Lewis and Clark College

Portland, Oregon 97219–7899 USA

e-mail: [email protected]

Michael J. Ryan

Section of Integrative Biology C0930, University of Texas

Austin, Texas 78712 USA


David B. Wake

Museum of Vertebrate Zoology and Department of Integrative Biology, University of California

Berkeley, California 94720–3160 USA


keywordsintegrative biology, mechanism, evolution, selection, phylogeny

abstractAdding a causal, mechanistic dimension to the study of character evolution will increase the strength

of inferences regarding the evolutionary history of characters and their adaptive consequences. Thisapproach has the advantage of illuminating mechanism and testing evolutionary hypotheses rig-orously. We consider the advantages of combining mechanistic and historical biology in the studyof behavior, physiology, and development. We present six examples to illustrate the advantages:(1) preexisting biases in sound perception in frogs; (2) preexisting biases in visual cues in swordtail

384 Volume 77THE QUARTERLY REVIEW OF BIOLOGY

fishes; (3) exploitation of prey location behavior for attraction of mates in water mites; (4) heterospecificmating in asexual molly fishes; (5) developmental foundation of morphological diversification inamphibian digits; and (6) locomotor performance at low temperature and the evolution of nocturnalityin geckos. In each of these examples, the dominant role of history, combined with organismal integra-tion, makes ignoring history a risky proposition.

UNDERSTANDING HOW LIFE works isthe fundamental goal of biology. As

modern biologists, we strive for a mechanisticexplanation of phenomena that range frommolecules to ecosystems. Mechanistic under-standing involves distinguishing reproducibleand testable causal patterns from noncausalor nontestable associations (Brandon 1996).With the advent of objective methods for gen-erating and testing hypotheses of phylogeneticrelationships (Hennig 1966), researchersactively began to incorporate phylogeneticperspectives into many areas of biologicalresearch focused on mechanistic understand-ing, including (as general examples) behav-ior, life history, functional and developmentalmorphology, and comparative physiology(Lauder 1982, 1990, 1991; Stearns 1992;Arnold 1993, 1994b; Niewiarowski 2001).This has meant that these and other fields,long isolated from systematic biology, gainedhistorical perspective that has enriched inter-pretations, sometimes opening new pathwaysfor understanding how history and biologicalmechanisms interact (e.g., Galis 1996). Onthe one hand, there has been enthusiasticintegration of historical perspectives withmechanistic biology, as attested by a large lit-erature that includes numerous symposia vol-umes (e.g., Harvey et al. 1996), collections ofessays (e.g., Martins 1996), book chapters(e.g., Lauder 1991), and monographs (e.g.,Brooks and McLennan 1991; Harvey andPagel 1991; Garland and Carter 1994; Federet al. 2000), in addition to a large number ofresearch papers. On the other hand, authorshave argued that the new methods are fun-damentally a distraction, and that theyobscure the central issues in fields such asbehavioral evolution (Reeve and Sherman1993, 2001) and comparative physiology(Mangum and Hochachka 1998). Our objec-tive in this paper is to illuminate how intri-cately history and mechanism are integrated,and to show the value of a phylogenetic per-spective in understanding patterns of evolu-

tion in behavior, morphology, and physiology.We present case studies that show how his-torical perspectives enable better understand-ing of mechanisms. We also argue that deepunderstanding of history and mechanism isinvaluable for interpretation of results thatotherwise could be seen as the workings ofsome ongoing dynamic (Ghiselin 1999). Inparticular we wish to counter the view thatadequate understanding of complex patternsof evolution can be obtained by using anapproach that relies solely on surrogates orproxies for fitness, such as performance orappearance, instead of the underlying and webelieve more significant biological features oforganisms.

In the following sections, we begin first bydefining mechanism and history. Second, wepresent a logical treatment of the conflict thatexists between integrative (historical, mecha-nistic) and purely adaptive approaches. Test-ability is a major advantage of the integrativeapproach we advocate. Next, we outline a pro-tocol for testing integrative hypotheses andshow how purely adaptive explanations arenot scientifically valid unless the integrativeexplanation is first shown to be false. We usesix examples from behavior, development,and physiology to illustrate the power andnecessity of an integrative approach in thestudy of complex integrated systems.

Mechanism and HistorySince mechanistic and historical biology

are central concepts in our argument, wedefine what we mean by mechanism and his-tory, and show how they can be integrated inanswering evolutionary questions. Mecha-nism and history can be thought of as twodimensions or axes (Autumn et al. 1999) thateach contain important information, but bythemselves may be insufficient to address evo-lutionary questions involving complex inte-grated systems.

We use “mechanism” broadly to meanexperimentally established, testable, causal

December 2002 385HISTORICAL AND MECHANISTIC BIOLOGY

relationships among parts (Lauder 1991; Bran-don 1996). The appropriate level or levels inthe biological hierarchy for a mechanisticexplanation will depend on the question( Jacob 1977). Since causal interactions mayoccur in both upward (small to large) anddownward (large to small) directions (Bock1989; Autumn 1995; Brandon 1996), it is notalways possible to reduce the function oforganisms to simple molecular causes (Sava-geau 1991). For this reason, mechanismsdescribe interactions that range from mole-cules to ecosystems.

Figure 1A shows a simple mechanism con-sisting of a lever arm, gears, and a belt. Eachpart of the mechanism is related causally tothe production of torque, the measure of per-formance in this example. Variation in torqueis a mechanistic consequence of variation inthe moment arm and variation in the gearratio. Let us imagine that there are two ver-sions of this mechanism: A and B. A and B aresimilar in all respects except that one of thegears in B is larger than in A. Is the largergear in B an adaptation for increasing torque?If the parts of A and B were laid out randomlyon a table, there would be no way to predictwhether a larger gear would increase torquebecause the order of gears in the mechanismwill determine if increasing gear sizeincreases or decreases torque. By addinginformation on how the mechanism is puttogether and how it works, one can predictthe effects of a larger gear with a great dealof certainty. The strength of the predictionwill depend on the strength of experimentaland theoretical support for the mechanism inquestion. In this example, the causal effect ofvariation in the form of each part can be pre-dicted by the equation:

Torque Moment ArmSize of Gear

Size of Gear= ⋅

2

1(1)

or, where

Size of Gear

Size of GearGear Ratio

1

2= (2)

and

TorqueMoment Arm

Gear Ratio=

(3)

Accordingly, only by knowing how a mecha-nism is assembled, and by understanding theprinciples that underlie its operation, canone predict how changing one of the partswill change performance. If the only questionis, “how does the mechanism work,” onecould use this dimension only. Evolutionaryquestions involve time and directionalchange, however. The mechanistic axis doesnot represent how the parts change, nor whatactual patterns exist in nature. Even if we cansay with certainty that changing the size ofGear 2 increases torque, we cannot saywhether the direction of change was fromlarge to small or from small to large. Themechanistic dimension lacks history.

We represent history as a second dimen-sion that is complementary to the causaldimension. The historical axis shows how theparts have changed over evolutionary time. InFigure 1B, evolution has resulted in smallergears, a larger moment arm, and an increasein torque. Any attempt to attribute theincrease in torque to the inferred evolution-ary changes in the parts would be purely cor-relational. Such inferences would not be test-able without adding information about howthe mechanism is assembled. In fact, if thegear ratio remained constant (as shown inFigure 1B), change in gear size would beindependent from change in torque, andonly changes in moment arm would have anyeffect on torque. Thus, by integrating themechanistic and historical axes, the evolutionof the system can be fully understood, andunfounded inferences can be avoided.

Testing the Mechanistic andHistorical Explanation

The general form of most controversy thatsurrounds explanations based on history andmechanism versus those based on currentutility alone has been as follows: one study(e.g., Wake 1991) argues that characters A(e.g., rules of development in salamanders)and B (e.g., small body size) mechanisticallyconstrain character C (e.g., number of digits),


Figure 1. Mechanistic and Historical AxesA. Mechanistic axis. This represents the causal interactions among the parts. Here the lever and the gears

work together to produce a torque. Variation in torque is a mechanistic consequence of variation in the momentarm and variation in the gear ratio. This axis does not tell us how these variables change and what actualpatterns exist in nature. This dimension lacks history. B. Historical axis. This represents patterns of change inisolated parts over evolutionary time. By itself, this axis cannot provide strong inferences about the mechanisticcauses of change. In this example, ancestral state is large gears, small moment arm, and low torque performance.In the derived condition, the gears have become smaller, the moment arm has become larger, and there is anincrease in torque. Why did torque increase? One could explain this by correlated change in any of the variables.Possibly the smaller gears increase torque, or possibly the larger arm increases torque. It is not possible to tellwithout adding the mechanistic dimension. In this example the gear ratio remains the same, so change in gearsizes have no effect on torque.


and that because of the historical patterns ofchange in those characters, the ultimatecause of the current character state (e.g., dig-ital reduction) was historical and acted mech-anistically through A and B (see Arnold1994a). An alternative approach (e.g., Reeveand Sherman 1993, 2001) argues that currentfitness differences are both the proximateand ultimate cause of the current characterstates. How can the controversy be resolved?We suggest that the controversy cannot beresolved by theorizing about current fitness,proxies for fitness, or performance differ-ences in populations—not even by actuallymeasuring fitness. Fitness differences are ared herring in challenging mechanistic andhistorical hypotheses because mechanismand history determine the range of charactervariation that is possible. The challenge to themechanistic hypothesis is to show that thehypothesized mechanism is incorrect. Logi-cally, this should be done in an experimentalcontext since one must demonstrate thatcharacters A and B do not mechanisticallyconstrain character C. Similarly, the appro-priate way to challenge an historical hypoth-esis is to reject it on phylogenetic grounds.The challenge is to show that another his-torical hypothesis is more likely.

Since the goal of this paper is to highlightthe importance of integrating mechanisticand historical biology to address evolutionaryquestions, it is reasonable to ask if we wouldhave reached our conclusions without inte-grating mechanism and history. In each ofthe six case studies we present, a false conclu-sion would have been reached without anintegration of mechanism and history. In thefollowing sections, we present a summary ofthe six case studies, and identify the appro-priate tests of the conclusions they reach. Ineach case, we present an evolutionary conclu-sion based on an integration of mechanisticand historical hypotheses, and an alternatehypothesis based only on maximizing fitnessin current populations. We show that in eachcase, the alternate fitness-based hypothesiscannot be true unless the mechanistic or his-torical hypotheses are shown to be false (Fig-ure 1).

The mechanistic and historical dimensionsof the six examples show how and why mech-

anism and history must be integrated inorder to reach a robust and meaningful con-clusion. Figure 3 places each example in thecontext of the model shown in Figure 2. Acommon theme in these examples is that therange of phenotypes available for selection isconstrained by the mechanisms that relatethe characters in question, and by the phylo-genetic history of the taxa in question.

Frog BehaviorCommunication systems provide fertile

ground for integrative studies. Evolutionarybiologists want to know how mate recognitioncan reproductively isolate taxa, and Lorenz’suse of display behavior in a phylogeneticanalysis of the anatine ducks presaged thecurrent practice of mapping behavioralcharacteristics onto phylogenies (Lorenz1941). Animal communication is one of thesocial behaviors of great interest to behav-iorists for over two millennia (Aristotle,translated by D’A W Thompson 1918; Dar-win 1873). Neuroethology, the study of neuralmechanisms of behavior, often focuses oncommunication; examples include the func-tion and evolution of the song controlnucleus in songbirds (Konishi 1994), and themechanisms that underlie signal decoding ina variety of insects and frogs (Gerhardt 1994).

Communication systems are not only ame-nable to an integrative study but actuallyrequire such an approach for reliable inter-pretation of their evolution. The simplestcommunication system is a dyadic interactionof signal and receiver. Signal production,however, is not isolated from the rest of theanimals’ biology. In some insects such as Dro-sophila, song patterns produced by wing beat-ing are inextricably linked to a morphologyand physiology involved in and perhaps ini-tially derived for self-powered flight (Ewingand Miyan 1986). The vocalizations of tetra-pods are intertwined with the respiratory sys-tem (Gans 1973), and many visual stimuli,such as the long feathers of peacocks andwidowbirds, are elaborations of morphologiesinvolved in critical aspects of the animals’ sur-vival strategy, such as flight (Balmford et al.1993). The sensory modality involved in sig-nal reception is often used for other func-


Figure 2. Diagram Illustrating How Integrating History and Mechanism Can DistinguishBetween Evolutionary Causes and Evolutionary Side Effects

A. Alternate hypothesis based on maximizing fitness in current populations. Taxa 1 and 2 differ in a featureC that is associated with a difference in performance. Without mechanistic or historical knowledge, the infer-ence is that the evolution and maintenance of the features in their respective taxa are driven by selectionmaximizing fitness within populations. B. With additional information about the mechanistic basis of variationin character C, the effects caused by stasis and change in other characters can be assessed. With additionalinformation about the phylogenetic history of character C and its mechanistic basis, character polarity (ancestraland derived states) can be determined. By integrating mechanism and history, what was initially (see part A)thought to be an evolutionary cause can now be seen to be an evolutionary effect. C. The appropriate tests ofthe hypotheses outlined in part B, and the appropriate way to distinguish between the scenarios of parts A andB. The alternate (fitness based) hypotheses suggested by part A can only be valid if the mechanism or historyoutlined in part B is incorrect. Arguments based on current fitness in populations do not have any bearing onthe hypotheses outlined in part B.


tions as well. Some moths, for example, havethe ability to detect ultrasonic sounds, whichcould function in conspecific communicationas well as detection of predatory bats, thefunction for which it might have been initiallyderived (Conner 1987). Thus one cannotunderstand the evolution of the signals usedto communicate and the mechanisms used todecode the signals solely in the context of thecommunication system.

The two central questions in sexual selec-tion are: (1) how does signal variation inmales influence their attractiveness tofemales; and (2) why do females find somesignals more attractive than others. In thetungara frog, Physalaemus pustulosus, malescan add a component, a chuck, to the basicwhine component of the advertisement call.Females prefer whines with chucks to whineswithout chucks. Bigger males produce lowerfrequency chucks, and females prefer lowerfrequency chucks. This results in larger maleshaving greater mating success. But femalesbenefit as well; by mating with larger males,females have more of their eggs fertilized(Ryan 1985). An interpretation based solelyon the current function of this system is thatmale tungara frogs evolved chucks to signallarge size as females coevolved a preferencefor such calls. A closer examination of thesensory mechanisms involved in signal pro-cessing, together with knowledge of the phy-logenetic relationships and signal variation ofclose relatives, yields a different interpreta-tion, however.

Frogs have two peripheral auditory organswhose frequency tuning contributes impor-tantly to decoding acoustic signals. Theamphibian papilla (AP) is more sensitive tolower frequencies, usually below 1500 Hz,while the basilar papilla (BP) is more sensitiveto higher frequencies. In the tungara frog,the AP’s most sensitive frequency is ca. 700 Hzwhile that of the BP is 2100 Hz (Ryan et al.1990). The whine has most of its energy below1000 Hz, while the chuck has most of itsenergy above 1500 Hz. Phonotaxis experi-ments combined with the neurophysiologicaldata show that the AP is critical in the initialdecoding of the whine, while the BP is criticalin the initial decoding of the chuck (Wilczyn-ski et al. 1995).

Although the tuning of the AP and BPtends to match the dominant frequencies ofthe whine and chuck, respectively, the BP istuned on average slightly below the dominantfrequency of the average chuck in the popu-lation: 2100 Hz for the BP and 2500 Hz forthe call. This suggests that the female’s behav-ioral preferences for lower frequency chucksmight result from the better match of thecalls to the sensitivity of the BP. Computermodels integrating an average tuning curveand digitized calls drawn at random from apopulations of tungara frogs support this con-tention (Ryan et al. 1990), as do a number ofphonotaxis experiments that use syntheticstimuli (Wilczynski et al. 1995). Thus thecombination of studies of function and mech-anism shows that females gain a reproductiveadvantage by preferring the lower frequencychucks of bigger males, and this preferenceis mediated by the tuning of the BP.

Chucks are derived within one of the twoclades (the P. pustulosus-P. petersi clade) of thePhysalaemus pustulosus species group (Ryanand Drewes 1990; Cannatella et al. 1998). Ofthe ca. 40 species in the genus, only somepopulations of the tungara frog’s sister spe-cies, P. petersi (which might include more thanone species; cf. Cannatella et al. 1998), areknown to add a call suffix. Results of behav-ioral experiments were combined with phy-logenetic information to determine if femalepreference for chucks evolved in concert withchucks (the coevolution hypothesis) or iffemales had a preexisting preference forchucks that was exploited by males (sensoryexploitation hypothesis; both hypothesesreviewed in Kirkpatrick and Ryan 1991; Shaw1995; Ryan 1998). These hypotheses can bediscriminated by reconstructing the phylo-genetic history of the species group and infer-ring the historical sequence by which chucksand preferences for chucks evolved.

P. coloradorum is a member of the Physalae-mus pustulosus species group. It is a memberof a clade of species west of the Andes thatdiverged from the P. pustulosus-P. petersi cladeafter the evolution of the chuck (Ryan andDrewes 1990; Cannatella et al. 1998). Phon-otaxis experiments show that P. coloradorumfemales prefer their own calls with P. pustulo-sus chucks added over the normal chuckless


calls of their conspecific males (Ryan andRand 1993). Both P. pustulosus and P. colora-dorum prefer calls with chucks. Parsimonysuggests that this preference was inheritedfrom a common ancestor, although it is pos-sible that females of P. coloradorum coinciden-tally evolved the same preference for traitsnot existing in their own males.

Studies of the population biology of tun-gara frogs show that females gain a reproduc-tive advantage by preferring whines withlower frequency chucks. The phylogeneticanalysis, however, rejects the hypothesis thatfemale tungara frogs evolved this preference.Instead, both the preference for chucks andthe neural tuning that biases females towardlower frequency chucks are present in a closerelative, P. coloradorum, and appear to havearisen prior to the divergence of these spe-cies. This preexisting bias was responsible forthe selection that favored chucks when theyarose. At this point, it is not clear why thereis a preexisting preference for chucks. Theanswer might be adaptive; chucks mightincrease the ability of females to detect thecall in background noise or locate it in space.It might not be obviously adaptive; the BP isa frequency channel that is not used in com-munication by most species in the Physalae-mus pustulosus species group. Once speciesrecognition is released by reception of thewhine, further acoustic stimulation maymerely enhance the physiological andbehavioral response of the receiver.

Knowledge of the mechanisms involved incall preference can prevent misguidedspeculation about other adaptive scenarios.Reeve and Sherman (1993:25) suggestedthat P. coloradorum as well as P. pustulosusshould be selected to prefer low frequencycalls since these calls should be produced bylarger males. They suggest that this sharedpreference could explain the presence of sim-ilarly tuned auditory responses. As notedabove, however, frogs have two auditory papil-lae sensitive to airborne sound. The amphib-ian papilla is tuned to lower sounds, and inboth P. pustulosus and P. coloradorum it is max-imally sensitive to the species’ whinelikeadvertisement call. The basilar papilla istuned to high frequency sounds, and is criti-cal in sensing the P. pustulosus chuck. The P.

coloradorum call does not contain frequenciesin the maximum sensitivity range of its ownBP. Thus a shared preference for low fre-quency whines in both species, which involvesthe AP, could not explain why P. coloradorumand P. pustulosus have BP organs with similartuning. Nevertheless, Reeve and Sherman’scomment, however misguided, is in the spiritof the approach we recommend—knowledgeof the animal’s entire biology might be nec-essary to understand specific functions; it isimportant, however, to get the biology right.

Why do female tungara frogs prefer calls withchucks to calls without chucks and preferlower frequency chucks to higher frequencychucks?

Mechanism: The tuning of the basilar papilla(BP) is sensitive to the dominant frequencyof the chuck, and is critical in its initial neuralprocessing. The BP of the female is tuned torespond to the frequency range of the chuck,and is more responsive to slightly lower thanaverage frequency chucks in the population.Since large males make lower frequencychucks, females prefer larger males. As a con-sequence, females gain a reproductive advan-tage.

History: The tuning of the BP is similaramong species in the Physalaemus speciesgroup, and thus represents an ancestral char-acter state. The tungara frog is well nestedwithin Physalaemus, and the chuck representsa derived character state. The tuning of theBP and therefore the preference for chucksexisted prior to the evolution of the chucksthemselves.

Integration of mechanism and history:Females gain a reproductive advantage bypreferring whines with lower frequencychucks, but selection for this female prefer-ence is not the evolutionary cause. Instead,the reproductive advantage gained by pref-erence for whines with lower frequencychucks is an evolutionary effect due to theway frog brains work and how they haveevolved.

Testing the conclusion: Since females do havegreater reproductive success when they matewith larger males, one could have concluded


that this was a selective advantage over thosethat did not prefer larger males, that this wasthe cause of preference for chucks, and thatthe evolution of AP and BP tuning was drivenby this advantage. This alternate hypothesisbased on current fitness maximization is con-tradicted by the mechanistic and phyloge-netic hypotheses. Choice of larger males andreproductive advantage are the evolutionaryside effects, not the evolutionary cause (Fig-ure 3). In fact there is no advantage becausethere is no alternative to preference for largermales. Selection for choosing larger males isnot possible unless the mechanistic or the his-torical hypotheses can be rejected. Chal-lenges to this conclusion are likely to comefrom advances in the neurobiology of frogs,or from a systematic revision of Physalaemus.

Swordtail BehaviorBehaviors associated with the “sword” of

fishes of genus Xiphophorus (which alsoincludes swordless platyfish) are important inmate recognition. Basolo (1990a) showedthat females of X. helleri preferred males withlonger tails. Rosenthal and Evans (1998) usedvideo playbacks to show that this female pref-erence was based on the sword itself and notother correlated characters. Basolo (1990b,1995) also appended swords to two species ofplatyfish, X. maculatus and X. variatus. Inboth of the species, females preferred thesworded conspecifics to normal swordlessones. A phylogenetic analysis of swordtails byRosen (1979) and Rauchenberger et al.(1990) argued for monophyly of swordtails.Since swords are present in this group andabsent in platyfish and other fishes in the fam-ily Poeciliidae, Basolo argued that there wasa preexisting bias for swords, and male sword-tails exploited this female bias when evolvingthe caudal extension. Meyer et al. (1994) pre-sented an alternative interpretation of thephylogenetic relationships within Xiphophorusthat questioned the preexisting bias interpre-tation of Basolo, although recent analyses(Borowsky et al. 1995; Marcus and McCune1999) tend to be more supportive of the pre-vious phylogenetic hypotheses (Rosen 1979;Rauchenberger et al. 1990). Regardless of therelationships within the genus, Basolo’s inter-

pretation was supported by experiments thatshow female preference of males with artifi-cially added swords in the closely relatedgenus Priapella.

Females of many species prefer largermales (Ryan and Keddy-Hector 1992). Thesword is only one of several phenotypes thatcould enhance male attractiveness; perhapslarge body size is what many females appearto find so attractive. In a study using videoplaybacks, Rosenthal and Evans (1998)manipulated the male phenotype in other-wise identical sequences of a courting male.In addition to the obvious control experi-ments, they removed the sword from themale but increased the male’s size (maintain-ing natural proportions) such that it was thesame length as the male with the sword.Females showed no difference in preferencebetween the sworded and swordless malethat were the same body length. This sup-ports the contention that evolution of thesword might exploit a preexisting bias forlarge size that might be widespread throughthe poeciliids.

Why do female swordtails prefer males withlonger swords?

Mechanism: Female platyfish prefer largemales. Females measure male size by estimat-ing total body length, and the presence of asword increases total body length. Thus,female swordtails prefer sworded males oversworded or swordless males of shorter bodylength.

History: Preference for large males predatesthe evolution of swordtails. Swordtails aremonophyletic and share a common swordedancestor. Thus, preference for sword predatesevolution of the sword.

Integration of mechanism and history:Females prefer males with longer swordsbecause the sword exploits a preexisting biastoward preference for large males.

Testing the conclusion: An alternate hypoth-esis based on current fitness maximization isthat females are being selected to choose“superior” males, which possess large and/orlarger swords. The mechanistic and phyloge-netic analyses contradict the alternative


hypothesis that female swordtails evolved thispreference as a result of selection for malequality. Selection for choosing “superior”males identified by larger swords is not pos-sible unless the mechanism or the history isfalse. Challenges to this conclusion are likelyto come from advances in the study of platy-fish behavior, or from a systematic revision ofXiphophorus and related taxa.

Water Mite BehaviorPreexisting bias for male courtship signals

in water mites may result from selection in a

context unrelated to mate choice (Proctor1991, 1993). Male water mites (Neumania pap-illator) produce water surface vibrations thatattract females. Females approach thesevibrations, and upon reaching their sourcethey are courted and mated by the male.These mites hunt at the water surface forcopepods swimming below. Male and femalemites find their prey by localizing the sourceof the water vibrations generated by the loco-motion of the copepods. The water surfacevibrations produced by the courting malesand the swimming prey are similar in form.

Figure 3. Key to Figure 2 for the Six Examples Presented in the TextIn each example, the mechanistic and historical hypotheses contradict the alternate adaptive hypothesis

(Figure 2A) based on current selective processes. Ancestral characters (Aa) and derived characters (Bd) deter-mine mechanistically the state of character C, which in turn determines performance (P). Thus, derived per-formance (Pd) is the evolutionary effect, and the derived characters (Bd) are the evolutionary cause. Therefore,it is not possible for selection to directly affect the state of character C (denoted by Ø), and the adaptivehypothesis based only on current fitness maximization must be false unless the historical or mechanistic hypoth-eses can be rejected.


Foraging and sex are linked: hungry femalesare more likely to mate since they are morelikely to be attracted to the water surfacevibrations (Proctor 1991, 1993). Responses tosensory input have multiple functions, butknowing this does not establish whether aparticular response arose simultaneously inboth functions (foraging and sex) or evolvedinitially in one. A phylogenetic analysis showedthat the use of water surface vibrations in for-aging preceded the use of similar vibrations inmale courtship. Thus, the most parsimonioushypothesis is that response to water surfacevibrations evolved initially in foraging and wasthen exploited for male courtship.

Why are female water mites attracted to watersurface vibrations created by males?

Mechanism: Water mites locate copepod preyby the water surface vibrations they create inthe water when they move. Male water mitescreate water surface vibrations that match thefrequency of those created by the copepodprey. Males obtain matings when female mitesare attracted to the water surface vibrationscaused by males.

History: The use of water surface vibrationsin foraging preceded the use of similar vibra-tions in the attraction of females by males.

Integration of mechanism and history: Malesgain a fitness advantage by exploiting a pre-existing sensory bias in females.

Testing the conclusion: The mechanistic andphylogenetic analyses contradict the alterna-tive adaptive hypothesis that female watermites evolved their preference as a result ofselection for locating males. Any reproductiveadvantage for the female gained by male pro-duction of water surface vibrations is an evo-lutionary side effect, not the evolutionarycause. Challenges to this conclusion are likelyto come from advances in the study of behav-ior, and/or a systematic revision, of watermites.

Molly BehaviorHistorical analysis can reveal the existence

of currently functionless traits. It often isassumed that behavioral traits must be adap-tive. An alternative is that the presence of

these traits might be due to evolutionary per-sistence rather than current maintenance byselection. The Amazon molly (Poecilia for-mosa), a small poeciliid fish, is an all-femalespecies that reproduces by gynogenesis. It isthought to have evolved from hybridizationbetween a male P. latipinna and a femaleP. mexicana (Turner 1982). Female Amazonsmate with males of either one of these speciesto obtain the sperm necessary for embryo-genesis, but this sperm is not incorporatedinto the genome of the offspring. Female Ama-zon mollies, like other poeciliid fishes andmany other organisms, show a preference forlarge males that is statistically indistinguishablefrom that of its two parental species, P. lati-pinna and P. mexicana. The evolutionary main-tenance of these preferences might beexplained by direct selection if mating withlarger males has an immediate influence onfemale reproductive success. Thus femalesshould mate with larger males if these malesprovide more or better resources such as foodor nesting sites, are better fathers because ofincreased protection of the young or mate, orif larger males fertilize more eggs. Alterna-tively, indirect selection can result in largemale mating preferences if there is geneticvariation for the preference for large size, andif there is linkage disequilibrium betweengenes associated with these preferences andthose associated with large male size. Underthis scenario of Fisher’s runaway sexual selec-tion theory, the preference increases in fre-quency as a correlated response to direct evo-lution of the trait (Andersson 1994).

The peculiar mating system of Amazonmollies provides an opportunity to test theindirect and direct selection hypotheses forthe maintenance of this female preference.The indirect selection hypothesis is immedi-ately rejected. Since there are no males in thespecies there can be no correlated evolutionof trait and preference. Some of the directselection hypotheses can also be rejectedsince the males mated by Amazon molliesoffer no parental care or resources to eventheir conspecific mates, let alone heterospe-cifics. Large male size does not influencefemale reproductive success (Marler andRyan 1997). There is no relationship betweenbody size of male P. latipinna and the fecund-


ity of females they mated. This eliminates theobvious explanations for the evolutionarymaintenance of mating preference for largemales, and supports an alternative hypothesisthat this preference was inherited from theparental species and has persisted in Amazonmollies without being maintained by selec-tion.

Why do asexual all-female mollies preferlarger males?

Mechanism: Females of both sexual and asex-ual species of molly prefer to mate with largemales. Mating with heterospecific males isrequired for embryogenesis in asexual Ama-zon mollies, but sperm is not incorporatedinto the genome of the offspring of Amazonmollies.

History: Preference for large males is anancestral character state of the species thatgave rise to the Amazon molly. Asexual repro-duction evolved in the Amazon molly.

Integration of mechanism and history:Female Amazon mollies prefer large malesbecause preference for large males was inher-ited from their sexual ancestors, and notbecause selection currently maintains thispreference.

Testing the conclusion: The mechanistic andphylogenetic analyses contradict the alterna-tive adaptive hypothesis that Amazon molliesevolved a preference for large males as aresult of selection for male quality. The con-clusion based on integrating mechanism andhistory is strong because there is no fitnessadvantage of choosing larger males. Selectionfor choosing large males is not possible unlessthe mechanism and the history are false.Challenges to this conclusion are not likely,unless it can be shown that sperm is incor-porated into the genome of the Amazonmolly, and that the Amazon molly did notevolve from species that already exhibit apreference for large males.

Amphibian Limb DevelopmentThe pentadactyl limb is highly conserved

in the evolution of terrestrial vertebrates. Theearliest amphibians had more than five digits(Coates 1994), but very early in the history of

tetrapods, evolution reduced the numbers ofdigits down to four or five in the forelimb andfive in the hind limb. Those numbers char-acterize most generalized tetrapods that havelived, with exceptions involving furtherreductions. Both genetic and developmentalmechanical constraints have been proposedas reasons for this conservation, and as expla-nations as to why polydactyly is so uncommonand unstable (Shubin et al. 1997). Reductionsin digital number frequently are adaptive innature (e.g., the mechanically specializedskeleton of the bird wing is reduced to threehighly modified digits; Burke and Feduccia1997). There are also alternative explana-tions for digital reduction, however.

Digits are numbered from I to V, countingfrom preaxial to postaxial. When digitalreductions occur they typically involve lossesof the outer digits, I or V, or occasionally Iand II. Only one instance of the loss of bothdigits IV and V is well documented, in thehand of theropod dinosaurs (Sereno andNovas 1992). Developmental constraints maycause the outermost digits to be lost first.Proximal portions of the limb are formed firstduring morphogenesis, then more distal por-tions condense and finally digits appear. Adistinct developmental axis is apparent infrogs and amniotes that extends through thepostaxial limb, specifically through digit IV. Adigital arch then extends preaxially, anddigits III, II, and I are formed in temporalsequence, with digit V usually being addedrelatively early and in an opposite direction(Shubin and Alberch 1986). Amphibians dif-fer from amniotes in that no living species hasmore than four digits on the forelimb. Sala-manders differ further both from frogs andall amniotes that possess digits in that the axisof the developing limb extends through digitsI and II rather than digit IV. The digital archof salamanders starts at the primordium of anelement unique to amphibians, the basalecommune, and grows postaxially so that thesequence of digit formation is I-II (nearsimultaneous)-III-IV-V (Shubin and Alberch1986).

Digital reduction has proceeded beyondthe four digits of the forelimb in both frogsand salamanders. In frogs, miniaturized spe-cies (and only such species) have only four


toes, and in salamanders reduction to four orfewer digits has occurred repeatedly in phy-logenetically distinct lineages (Duellman andTrueb 1986). Digital reduction is always pre-axial in frogs, whereas in salamanders it isinvariably postaxial (Shubin and Wake 1996).

Digital reduction has been studied exten-sively, and explanations fall into two generaland somewhat competing categories: adap-tive (e.g., Lande 1978) and incidental toother phenomena (e.g., Alberch and Gale1985; Wake 1991). We propose a resolutionto this controversy that illustrates the syner-gism between history, mechanism, and adap-tation that we advocate.

Alberch and Gale (1985) argued that digitalreduction in both frogs and salamanders is aresult of developmental truncation, in whichthe digit formed last is that which is lost as asimple consequence of the failure of the digitto undergo morphogenesis. Limb develop-ment has been portrayed as a series of seg-mentation and bifurcation events duringwhich cell masses must reach a certain size (interms of numbers of cells) before proceedingthrough either event (Oster et al. 1988).Experimental reductions in the rate of celldivision resulted in the absence of the mostpreaxial digit in the frog studied, and of themost postaxial digit in the salamander studied,in accordance with expectations from theShubin-Alberch model of limb morphogene-sis. A cause of digital loss is thus reasonablywell known. Reduction in the numbers of cellsin a primordium can be accomplished as anincidental outcome of reduction in overallorganismal size, which results in fewer cells incell aggregations and organ primordia, provid-ing cell size remains constant. If cell sizeincreases and body size remains constant,there also will be fewer cells in primordia. Sal-amanders have enormous genomes. Thesmallest salamander genome is larger thanthat of all other tetrapods (except a fewAustralian myobatrachid frogs), and there isa well-documented positive relationshipbetween cell size and genome size (Roth etal. 1994). A small reduction in the size of anorganism that has a large genome, or a smallincrease in the size of the genome in anorganism of constant body size, would havethe side effect of reducing the number of cellsin a primordium; this effect would accen-

tuate through development, having its great-est impact near the end of limb morphogen-esis when the last digits are forming. Aminiaturized species with a large genome size(e.g., the four-toed plethodontid genus Batra-choseps; Sessions and Larson 1987) would, ineffect, have a double dose and be a likely can-didate for digital reduction if this formula-tion is correct. These observations show thatthere is a well-understood mechanism inamphibians for explaining digital reduction.

History provides a means of testing whetherthis proximal understanding is a sufficientexplanation for the ultimate causes of digitalreduction in amphibians. For example, onecan predict that miniaturized frogs belongingto different phylogenetic lineages might losethe first toe but never the fifth. This is, in fact,the case, because miniaturized terminal taxain phylogenetically distant lineages (such asspecies of Psyllophryne of the family Brachy-cephalidae, and Mertensophryne of the familyBufonidae; Alberch and Gale 1985) haveundergone digital loss. In frog genera thatbelong to different families, phalangealreduction in the first digit occurs. Phyloge-netic analysis shows that miniaturization hasoccurred independently, and that digitalreduction has evolved independently.Whereas only miniaturized species of frogshave lost a digit, there are many miniatur-ized frogs that retain the ancestral numberof digits. Thus, while miniaturizationincreases the likelihood of digital reduction,it does not dictate it.

Using similar logic, one can predict thatminiaturized salamanders, especially thosewith large genomes, belonging to differentphylogenetic lineages might lose the fifth toebut never the first. Such is the case, as dem-onstrated by Wake (1991), who showed thatminiaturized terminal taxa in three distinctlineages of the family Plethodontidae had lostthe fifth toe. Furthermore Wake showed thatrare variant animals in other miniaturizedlineages are sometimes found that have onlyfour toes, and occasionally they are asymmet-rical in a single organism. Miniaturized spe-cies in other families (e.g., Salamandrina ter-digitata of the family Salamandridae) alsohave only four toes, and invariably it is thefifth that is absent. But not all miniaturized


plethodontids have four toes. And a furthercomplication is that there are some relativelylarge species that have lost digits (membersof the families Amphiumidae and Proteidae)and even limbs (members of the family Sir-enidae). These species are not only large,however, they are also paedomorphic—allretain either larval morphology as adults, orsome larval traits such as open gill slits asadults—and all have very large genomes (Nec-turus of the Proteidae has the largest tetrapodgenome). Thus, using the arguments of Han-ken and Wake (1993), they can be said to bebiologically miniaturized, if not physicallyminiaturized. Necturus has only four toes (V isabsent), and Proteus (also Proteidae) has onlytwo, with digits III, IV, and V being absent, aspredicted by the Shubin-Alberch model. Dip-noans are basal outgroups of tetrapods thathave fins with a central axis, but their fins areremarkably reduced in relation to ancestralfishes that had fins that resembled limbs farmore than do those of modern dipnoans(Shubin 1995). Again, there is an importantcorrelation. Living dipnoans have by far thelargest genomes of any vertebrates, and theyare paedomorphic relative to more basalextinct taxa (Bemis 1984).

Miniaturization must often be adaptive inamphibians (Wake 1991), and we postulatethat the advantages of miniaturization withrespect to habitat use, predation avoidance,early age at first reproduction, or other fac-tors are more than sufficient to offset any dis-advantage that might arise from digital loss.Reeve and Sherman (1993) urge that oneadopt a strict adaptationist perspective inwhich one seeks an immediate adaptiveadvantage to digital reduction, but evolution-ary reduction has always posed difficulties forstrictly adaptationist explanations. They spec-ulate that alternative phenotypes might havereduced fitness because of possible disrup-tions of the original developmental program.In essence, they want to change the questionfrom constraint on the production of form toconstraint on adaptation (Amundson 2001).Such a perspective might be appropriate forlineages such as lizards, where stages in limband digital reduction can be found in livingtaxa that appear to be elongating and adapt-ing slowly to snakelike locomotor behavior.

Lande (1978) argued that it might take mil-lions of years for distal to proximal limbreduction, but in cases where such reductionwas geologically rapid, only weak selectionpressures were necessary. The situation infrogs and salamanders appears to be quite dif-ferent, however. In salamanders, digits comeand go as complete organs, as witness thecases of asymmetry found in single individ-uals (Wake 1991) and extreme variationfound within and among populations of a sin-gle species (e.g., the hynobiid salamanderHynobius lichenatus; Hasumi and Iwasawa1993). From our perspective it is far more sat-isfying to seek an explanation that takes fullaccount of mechanism, history, and adapta-tion, rather than automatically give prece-dence to direct adaptation alone, which risksthe loss of useful information that other per-spectives can provide.

Why do some small salamanders have onlyfour toes?

Mechanism: Development of the limbs of ver-tebrates entails several cell-level morphoge-netic processes. Cells destined to give rise tomesopodial skeletal elements cluster togetherto form condensations. As the condensationsgrow in size by accretion (but most impor-tantly by cell division), they either segment orbifurcate; they segment as they elongate, andas they round out they bifurcate. There is abias in the direction of cell proliferation thatis determined by positional information inthe limb itself. The combined effects of seg-mentation and bifurcation are the proximalmechanisms that underlie sequential forma-tion of phalanges and digits as well as themesopodial elements. Salamanders havemuch larger genomes than other tetrapods,and within the Plethodontidae, members ofthe tribe Bolitoglossini—a deeply nestedclade to which Batrachoseps belongs—have thelargest genomes of terrestrial salamanders.Furthermore, there is an empirical relation-ship between genome size and cell volume,with volume increasing in direct proportionto haploid genome size. Small adult size com-bined with large cell size means that conden-sations will have few cells, and in some partsof the developing limb perhaps an insuffi-


cient number to undergo segmentation orbifurcation.

History: A synapomorphy for the Order Cau-data (salamanders) is a switch from postaxialto preaxial dominance in limb development(Shubin 1995; Shubin and Wake 1996). Infrogs and amniotes the first digit to appear isnumber four. A digital arch forms and growsin a preaxial direction, bifurcating and seg-menting to give rise to digits three, two, andone, in that order. Before this sequence iscompleted, digit five forms in a postaxial direc-tion. In salamanders there is precocial devel-opment of digits one and two, and from themthe digital arch grows posteriorly, bifurcatingand segmenting to give rise sequentially todigits three, four, and five. Thus digit one isthe last formed in frogs and amniotes, but digitfive is the last formed in salamanders.

Salamanders also have experienced a dra-matic increase in genome size relative to allother tetrapods, and this has major implica-tions for limb development (Sessions andLarson 1987). Because salamanders are rela-tively small as juvenile (and adult) organisms,the number of cells is dramatically reducedrelative to numbers in other tetrapods ofcomparable size (Hanken and Wake 1993;Roth et al. 1994). Condensations of cells donot form unless a certain minimal numberare present, and when condensations doform they do not bifurcate or segment untila certain indeterminate number of cells arepresent.

Integration of mechanism and history: Thecombination of an historical shift in posi-tional dominance in limb development andan historical increase in genome and cell size,in conjunction with particular morphoge-netic mechanisms and small overall size, leadsto the repeated evolution of four-toed sala-manders in which the missing toe is the lastformed, or number five. Digital reductionalso occurs in very small frogs, even in absenceof large genomes and cells, and it is invariablythe first digit that is lost. These patterns arepredictable from a combined understandingof history and mechanism. Small frogs andsmall salamanders do not invariably lose adigit; doubtless there is a role for stabilizingselection in maintaining the general design

principles of the limbs. But in the absence ofsustained selection digits are lost as a result ofa default process related to the combinationof history and mechanism.

Testing the Conclusion: Without the integra-tion of mechanism and history, one couldconclude that digitally reduced salamandershave a fitness advantage over individuals with5 toes, and that this fitness advantage under-lies the pattern of digital reduction. Instead,an understanding of the rules of develop-ment and the phylogenetic history of sala-manders and frogs strongly suggests that anydifferences in performance due to digitalreduction are evolutionary side effects. Selec-tion cannot act directly on toe numberbecause variation in toe number is limited bythe number of cells in the developing limbbud, which is in turn a mechanistic conse-quence of cell number and cell size. Chal-lenges to this conclusion are likely to comefrom advances in developmental biology, andfrom systematic revisions of the Amphibia.

Gecko Locomotionthe nocturnality paradox

In ectotherms, the rate of physiologicalprocesses decreases exponentially as bodytemperature drops below the thermal opti-mum. Diurnal lizards use behavioral ther-moregulation to maintain body temperaturesnear their thermal optima, typically 35–45�C(Cowles and Bogert 1944; Avery 1982; Huey1982). As one would predict from an adaptiveperspective, the thermal optima for a varietyof physiological and performance variablesare typically close to the temperatures diurnallizards experience in nature (Huey 1982).The evolution of nocturnality is interestingbecause nocturnal lizards actively forage attemperatures 10–35�C below the body tem-peratures (and thermal optima) of typicalheliothermic, diurnal lizards. Nocturnal activ-ity thus represents a significant challenge tothe evolutionist. For example, given typicalrate-temperature effects (Bennett 1982), arelatively moderate nocturnal body tempera-ture of 25�C would result in a 50–75% decreasein performance in a diurnal lizard. From anadaptive perspective, one would expect lizardsto evolve thermal optima that are similar to


body temperatures during activity (Huey et al.1989). Therefore nocturnal lizards shouldhave thermal optima that are similar to thebody temperatures they experience at night.In other words, both nocturnal and diurnallizards should function best at their respectiveactivity temperatures. If the thermal optima ofnocturnal lizards have decreased, it followsthat nocturnal lizards should have evolvedgreater performance than diurnal lizards atlow temperature, and to have done this at thecost of reducing performance at high tem-perature.

Although this model makes sense as athought experiment, the data strongly contra-dict it. Nocturnal lizards are capable ofgreater locomotor performance at low tem-peratures than are comparable diurnal liz-ards (Autumn et al. 1994; Autumn et al. 1997;Autumn et al. 1999), but low temperaturesremain suboptimal for many physiologicalfunctions, including locomotion (Huey et al.1989; Autumn and Full 1994; Autumn et al.1994; Autumn and Denardo 1995; Autumn etal. 1997; Autumn et al. 1999). The evolutionof nocturnality seems to represent a paradox:geckos have evolved greater performance atlow temperatures yet their performance isgreater at high temperatures they never expe-rience during activity. How can this paradoxbe resolved? Is more information neededabout the behavioral ecology of geckos tounderstand the selective forces involved? Inthis case, behavioral ecology cannot answerthe question.

A traditional phylogenetic analysis cannotresolve the paradox either. As in the gearratio example (Figure 1), simply knowingthe order of events on a cladogram is notsufficient information to infer the causalbasis of an evolutionary change (Lauder1991). Instead, the resolution of the noctur-nality paradox requires the integration ofthe physiological mechanisms that underliesustained locomotion with the evolutionaryhistory of the gecko clade.

mechanistic parameters underlyingvariation in sustained locomotor

performanceThe capacity for an animal to sustain loco-

motion is known as endurance capacity. The

physiology underlying variation in endurancecapacity is similar in humans (Brooks andFahey 1985), lizards (Bennett 1982), amphib-ians (Gatten et al. 1992), and invertebrates(Full 1997). Locomotion for periods longerthan approximately 10 minutes is sustained byaerobic metabolism. The maximum rate ofaerobic metabolism, or VO2max, plays a largepart in determining the endurance capacity ofan animal. The speed of locomotion at whichVO2max is reached is termed the maximum aer-obic speed (MAS), which sets the upper limitfor sustainable locomotion and is largelyresponsible for determining endurance capac-ity. VO2max is only one of two variables that setthe MAS, however. The minimum cost of loco-motion (Cmin) is equivalent to the inverse offuel economy. Animals with greater fuel econ-omy (lower Cmin) have greater MAS becausethey cover more ground per unit of fuel. Therelationship between these variables can berepresented by the equation,

MASy

C=

−�VO2 0max

min

, (4)

where the idling cost (y0) is approximately 1to 1.5� the resting metabolic rate.

There is a cascade of effects (Figure 4)from the large decrease in body temperatureassociated with the evolution of nocturnality.In order to quantify what effect this had onperformance, it is necessary to account forthe other related variables in the system.Since aerobic capacity is strongly temperaturedependent, colder lizards will have lowerendurance, all else being equal. Both fueleconomy and aerobic capacity are stronglydependent on body mass, so one must factorout body mass in comparisons of MAS amonglizards that differ in size. This is not as simpleas it might seem. In order to explain a changein performance (MAS), it is necessary toaccount for the separate effects of change inthe three physiological variables that deter-mine MAS: y0, VO2max, and Cmin.

Estimates of Mechanistic Parameters areSensitive to Phylogenetic Bias

To make matters more complex, the rela-tionships between mass, aerobic capacity, and


Figure 4. Outline of the Integration of the Environmental and Physiological VariablesUnderlying Endurance Capacity in Nocturnal Lizards

For a detailed explanation, see Autumn et al. (1999).

Cmin are allometric, and are based on mea-surements of a phylogenetically biased sam-ple of species. Measurement of MAS, y0,VO2max, and Cmin in lizards is extremely timeconsuming, and members of some lizard taxaare not sufficiently cooperative to run steadilyon a treadmill. Not surprisingly, the sampleof species for which all of the variables havebeen measured remains relatively small andphylogenetically biased. In particular, vara-noid species thought to have unusually highVO2max dominate the sample of diurnal lizarddata. As Felsenstein (1985) and many others(e.g., Harvey and Pagel 1991; Garland 1992;)have emphasized, species cannot be treatedas independent data points. In order to com-pare species that differ in body mass, it is nec-essary to filter out the phylogenetic biascaused by sampling species that are differen-tially related by common ancestry. The esti-mates of the allometric coefficients may beconfounded with phylogenetic effects (Gar-land and Ives 2000), yet precise values of theallometric coefficients are necessary if ani-mals that differ in body mass are to be com-pared. Accordingly, an integration of mech-anism (allometric effects) and history

(phylogenetic effects) is necessary to resolvethe paradox of how nocturnal lizards couldhave evolved increased endurance at low tem-perature and yet remain suboptimal.

In order to use a phylogenetically biasedsample of lizard taxa to tease apart the causaleffects of variation in temperature, bodymass, y0, VO2max, Cmin, and MAS, an explicitintegration of mechanism and history is criti-cal (Autumn et al. 1999). This involves a phy-logenetic analysis of mechanism, in contrastto a traditional phylogenetic analysis of iso-lated characters.

resolving the nocturnality paradoxwith an integration of mechanism

and historyLizards are ancestrally diurnal, are physio-

logically optimized for diurnal temperatures,and are therefore capable of maximal perfor-mance in the environment they experience innature. Consider the effect of an ecologicalshift to a nocturnal environment on theancestral lizard physiology: low temperaturesreduce y0 and VO2max, but Cmin is thermallyinsensitive ( John-Alder and Bennett 1981;John-Alder et al. 1983; Bennett and John-


Alder 1984; Lighton and Feener 1989; Fulland Tullis 1990; Autumn et al. 1994; but seeWeinstein and Full 1994 for crabs). The effectof a reduced y0 and VO2max is a reduced MAS.This explains why diurnal lizards have poorendurance at low temperature. Two questionsremain: (1) Why do geckos have relativelypoor endurance at low temperature? In otherwords, why are geckos suboptimal at the tem-peratures they experience during activity?(2) Why do geckos have increased enduranceand MAS at low temperature relative to com-parable diurnal lizards? What is the physio-logical basis for the increased performancecapacity and can it explain why geckos remainsuboptimal at low temperatures? Answeringthese questions requires a comparison of y0,VO2max, Cmin, and MAS in lizards that differgreatly in body mass and phylogenetic his-tory.

Evolution of the Maximal Rateof Oxygen Consumption

Evolution of an increased VO2max at low tem-perature would increase MAS and endur-ance, and would explain question (2) above.Optimality theory predicts that the thermaloptimum for VO2max should coincide withactivity temperatures. VO2max is strongly tem-perature sensitive in diurnal lizards; a 10�Cdecrease in body temperatures causes a 50–75% decrease in VO2max (Bennett 1982). VO2max

is also strongly mass dependent. Allometricanalysis has played a central role in the com-parative physiology of metabolism (e.g., Klei-ber 1961), and would seem to be the propertool for the job. A comparison accounting forbody mass and temperature should reveal ifgeckos evolved an increased VO2max (Figure5A). Recent studies, however, that focuseddeeply on particular lizard clades (varanids:Thompson and Withers 1997; geckos:Autumn et al. 1999) have inadvertently cre-ated a phylogenetic bias in the VO2max data set.Varanids are thought to have unusually highVO2max, and the inclusion of six relativelyclosely related varanid species (Thompsonand Withers 1997) may amount to “phyloge-netic pseudoreplication” of a single evolu-tionary event rather than six independent val-ues (Garland and Adolph 1994). Fortunately,

advances in phylogenetic comparative meth-ods (Felsenstein 1985; Garland et al. 1992;Garland and Ives 2000) provide a simple solu-tion. Instead of treating species as indepen-dent samples, the method of phylogeneticallyindependent contrasts compares pairs of sis-ter taxa (Garland et al. 1993; Figure 5B). Themethod reduces the influence of sister taxathat have similar values by common ancestry.For example, the contrasts between varanidsister taxa (each with high VO2max) might benumerically small, while the single contrastbetween the ancestor of varanids and its sistertaxon might be numerically large. A phylo-genetically correct analysis reveals, surpris-ingly, that when geckos evolved nocturnality,they did not evolve an increased VO2max at lowtemperature. In fact, there is no evidence thatthe thermal optimum for VO2max decreased atall (Autumn and Full 1994). VO2max at 25�C isnearly identical in both nocturnal and diur-nal lizards, once the effects of body mass andphylogeny are accounted for. This answersquestion (1) of how geckos are suboptimalfor endurance at low temperature: low tem-peratures cause a decrease in VO2max in lizards,and geckos are no exception. The answer toquestion (2) of how geckos have increasedperformance at low temperature requires ameasurement of the other variables thataffect MAS.

Evolution of the Minimum Costof Locomotion

One of the most compelling questions inthe fields of physiology and biomechanics is:why does fuel economy differ among ani-mals? This question can be divided into twoclasses of mechanism: mass dependent andmass independent. This is an important dis-tinction because variation that is mass inde-pendent is likely to be the result of a differentmechanism than variation that is mass depen-dent. Figure 6A shows the energetic cost oflocomotion—or the inverse of fuel econ-omy—versus body mass in mammals and liz-ards. Note that the metabolic cost scalesstrongly with body mass. Lower values on they-axis mean better fuel economy. Larger ani-mals have better fuel economy. The lizardshave a slope that is 33% lower than the mam-


Figure 5. Scaling of Maximum Rate of Aerobic MetabolismA. Allometric analysis of body mass and VO2max (Q10 scaled to 35�C) in diurnal and nocturnal lizards. Phylo-

genetic pseudoreplication of varanid species (numeral 1) makes interpretation of differences between geckos(diamonds) and diurnal lizards (circles) difficult. B. Phylogenetically independent contrasts of body mass andVO2max (at activity temperatures). Solid line represents contrasts between diurnal lizard sister taxa (squares).Contrasts between gecko sister taxa fall within the 95% confidence limits of the diurnal lizard regression,indicating that body mass has a similar effect on VO2max in all lizards. Once the confounding effect of a phylo-genetically biased sample (see A) is removed, it is evident that VO2max (at activity temperatures) decreased sub-stantially in the gecko clade (triangle), and that this decrease is fully acounted for by the decrease in bodytemperature associated with the evolution of nocturnality. (Modified from Autumn et al. 1999.)


nation of suboptimality and increased perfor-mance at low temperature in geckos seemedparadoxical. An integration of mechanismand history reveals that there is no paradoxafter all. Low temperatures are suboptimalfor geckos because VO2max is thermally sensi-tive, and the thermal optimum for VO2max issimilar in nocturnal and diurnal lizards.Geckos have greater MAS at low temperaturebecause they evolved a low Cmin. The increasein performance is substantial: nocturnalgeckos at low temperature are operating atabout 2–3 times the MAS of diurnal lizards atlow temperature (Autumn et al. 1999). MASin a gecko at low temperature is only about50% of the MAS of a diurnal lizard at hightemperature, however.

Why are geckos suboptimal at the tempera-tures they experience during activity, and yethave increased endurance and MAS at lowtemperature relative to comparable diurnallizards?

Mechanism: The maximal rate of aerobicmetabolism (VO2max) is strongly thermally sen-sitive in ectotherms. The thermal optimumfor VO2max is high (30–45�C) in lizards, andgeckos are no exception. The minimum costof locomotion (Cmin) is thermally insensitivein all terrestrial vertebrates. Endurancecapacity is determined largely by the maxi-mum aerobic speed (MAS), which in turn isdependent on Cmin and VO2max by the equa-

tion, MASy

C=

−�VO2 0max

min. The Cmin and VO2max, and

thus MAS, scale allometrically with body mass.

History: The thermal sensitivity and the ther-mal optimum for VO2max remained largely evo-lutionarily static in lizards in general, andgeckos in particular. Accounting for a phylo-genetically biased sample and for differencesin body mass, geckos and other lizards havesimilar VO2max at low nocturnal temperatures.Accounting for a phylogenetically biased sam-ple and for differences in body mass, geckosevolved Cmin 1/2 to 1/3 that of other lizards.Geckos evolved a reduced MAS when theyevolved to be active at low nocturnal tem-peratures, but not as low a value of MAS aspredicted for diurnal lizards at the same lowtemperatures.

mals. This means that fuel economy in lizardsmay have a fundamentally different relation-ship to body mass than mammals. This wouldstrongly challenge Kram and Taylor’s (1989)mass-dependent hypothesis that ground con-tact time determines energetic cost of loco-motion. The difference in slope implies thatthe mass-dependent mechanism that relatesbody mass to Cmin is different in mammals andlizards. The next step would be to study ener-getics, muscle function, and biomechanics tofind out what is the cause of difference inmass dependence in lizards. This could takea lifetime and a substantial amount of grantsupport—especially since each data point onFigure 6A can take 3 months to collect. Whenwe separate the data on nocturnal geckos(Figure 6B) from the data on diurnal lizards,however, it becomes clear that the differencein slope between lizards and mammals (Fig-ure 6A) was a phylogenetic artifact. This isdue to the fact that geckos and diurnal lizardsdiffer greatly in fuel economy. Body mass hassimilar effects on fuel economy in bothgeckos and diurnal lizards, however. A phy-logenetically correct independent contrastsanalysis (Figure 6C) adds statistical support tothis conclusion. Only by considering historycan we reveal that the mass dependence is thesame within the geckos and diurnal lizards,and that mass-independent mechanisms maybe important in explaining the difference inCmin between geckos and diurnal lizards.Without considering phylogeny one couldhave invested a huge amount of time andresources looking for a mass-dependent mech-anism that does not exist. In this example,ignoring history leads one down an unproduc-tive research direction while embracing his-tory elucidates mechanism.

Thus, Cmin in lizards does not represent anunusual mass-dependent scaling relationship;rather, geckos have a low Cmin. A low Cmin is ashared, derived character of the gecko clade(Autumn et al. 1999). This answers question(2) of how geckos have increased perfor-mance at low temperature: since Cmin is tem-perature independent, it has the effect ofincreasing MAS (and endurance capacity) atall temperatures (Figure 7).

Without an understanding of the mecha-nisms that underlie endurance, the combi-

Figure 6. Scaling of Minimum Cost of LocomotionA. Allometric relationship between body mass and minimum cost of locomotion (Cmin). Circles represent data

for lizards (Autumn et al. 1999). The allometric slope in lizards (-0.21; solid line) is 33% lower than the slopemeasured in mammals (-0.31; Taylor et al. 1970). B. Allometric relationship between body mass and Cmin, withinthe gecko clade (squares), and within nongeckos (circles). The allometric slope in geckos and nongeckos issimilar, but geckos have a lower Cmin than nongeckos of a given body mass. C. Phylogenetically independentcontrasts in log Cmin and log body mass. The solid line represents a regression through contrasts among non-gecko sister taxa (circles). Contrasts among gecko sister taxa fall within the 95% confidence limits of theregression, indicating that evolution in body mass is associated with a similar change in Cmin in geckos andnongeckos. The contrast between geckos and nongeckos fell outside the confidence limits of the regression,indicating that a low Cmin is a shared derived character of geckos. (Modified from Autumn 1999 and Autumnet al. 1999.)



Figure 7. Theoretical Illustration of theResolution of the NocturnalityParadox

The solid curve represents the diurnal ancestor ofgeckos, and the dotted curve represents nocturnalgeckos. The dashed arrows show how performancedecreased because of the thermal sensitivity of VO2max,and also increased due to the adaptive effect ofdecreasing Cmin. A decreased Cmin increases perfor-mance at all temperatures, but since the geckos areactive at low temperature, performance is still sub-maximal.

an evolutionary side effect, not the evolution-ary cause, because the major changes tookplace 160 million years ago (Autumn et al.1999), and the most thermally sensitive vari-able (VO2max) did not evolve. Challenges to thisconclusion are likely to come from advancesin physiology and biomechanics of terrestriallocomotion, and from systematic revisions ofthe Squamata.

ConclusionThe historical and mechanistic approaches

we present enhance understanding of pat-terns and processes of organismal evolution.Understanding the historical and mechanis-tic foundations of phenotypes has the poten-tial to increase the accuracy and efficiency ofresearch. Perhaps most importantly, we pres-ent a protocol for testing historical andmechanistic explanations. In order for ourfields to progress, we must move beyondrhetoric toward rigorous tests of our hypoth-eses.

Objections to the use of historical approacheshave taken two fundamentally different direc-tions. First, a great strength of decades ofresearch in behavioral biology, functional mor-phology, and organismal physiology has beenthe strong focus on proximate causal mecha-nisms. Mapping of traits on a phylogenetichypothesis has been viewed as a regressiveencouragement of correlational approachesthat threaten the genuine progress that hasbeen associated with the focus on causal mech-anisms (Mangum and Hochachka 1998).

Second, the incorporation of evolutionaryperspectives in fields such as social behaviorhas led to a sharp focus on phenotypes andtheir contribution to fitness. This trend hascontributed to ongoing debates over themeaning and definition of adaptation. We didnot directly address the issue of whetheradaptation should be defined historically orby current utility. In the context of the casestudies we presented here, the definition ofadaptation is less important than a robustunderstanding of why organismal traits takethe form they do. We are especially con-cerned that exclusion of historical andmechanistic biology from adaptive explana-tions leads to the substitution of a superficial

Integration of mechanism and history:Geckos gained a performance advantage atlow temperature when they evolved a lowCmin, but remain suboptimal for sustainedlocomotion at the temperatures they experi-ence during activity. Suboptimality is a his-torical legacy of high thermal sensitivity andthermal optimum of VO2max inherited by thediurnal ancestor of geckos that lived approx-imately 160 million years ago.

Testing the conclusion: An alternate hypoth-esis based on current fitness maximization isthat current selection is responsible for sub-maximal performance at low temperature ingeckos. Without the integration of mecha-nism and history, one could have concludedthat balancing selection or phylogenetic iner-tia (Huey et al. 1989) was responsible for theparadoxically increased but submaximal per-formance at low temperature in geckos.Instead, the mechanistic and historical evi-dence strongly suggests that suboptimality is


are generally the ones that involve complexsystems. Such systems, however, do not lendthemselves to easy answers based on thoughtexperiments. Conclusions made withoutknowledge of the causal linkage among theparts of a complex integrated system areuntestable and are likely to be false. Even ifresearchers are mechanistically oriented sothat evolutionary questions do not interestthem, phylogenetic methods may be neces-sary to make valid comparisons among spe-cies. Even if researchers are interested inadaptation and not in mechanism, under-standing of mechanism may be necessary toreach a robust and rigorous answer. Let usnot slow the advancement of evolutionarybiology by assuming that what is outside ourfield is either simple or irrelevant. Instead, letus progress more rapidly by integratingmechanistic and historical biology in thestudy of evolution.

acknowledgmentsWe thank the following for their comments and sug-gestions: Sanford Autumn, Ham Ferris, Wendy Han-sen, Michael Kearney, Anne Peattie, Bill Rottschaefer,members of the Wakelab discussion group, DouglasFutuyma, an anonymous editor, and two anonymousreferees. Paul Sherman kindly provided a preview ofhis paper.

lineage effects. Pages 124–168 in Phylogenetics andEcology, edited by P Eggleton and R I Vane Wright.London: Academic Press.

Autumn K. 1995. Performance at low temperatureand the evolution of nocturnality in lizards [PhDthesis]. Berkeley (CA): University of California atBerkeley.

Autumn K. 1999. Secondarily diurnal geckos return tocost of locomotion typical of diurnal lizards. Phys-iological and Biochemical Zoology 72:339–351.

Autumn K, Denardo D F. 1995. Behavioral thermoreg-ulation increases growth-rate in a nocturnal lizard.Journal of Herpetology 29:157–162.

Autumn K, Farley C T, Emshwiller M, Full R J. 1997.Low cost of locomotion in the banded gecko: a testof the nocturnality hypothesis. Physiological Zoology70:660–669.

Autumn K, Full R J. 1994. Phylogenetic patterns ofnocturnality and physiological capacity in geckos.Physiologist 37:A-61.

Autumn K, Jindrich J, DeNardo D, Mueller R. 1999.Locomotor performance at low temperature and

and largely untestable narrative (see O’Hara1988) for deep understanding of the traititself by assigning the significance of variantsin the trait to fitness a priori. Every aspect ofthe phenotype potentially contributes to fit-ness, so some workers (e.g., Reeve and Sher-man 1993) interpret an overly historical per-spective as fostering the idea that features oforganisms persist that are nonadaptive. Whilediscovery of nonadaptive traits is a possibility,it is by no means a foregone conclusion of anintegrative approach.

In the case studies we presented, there wasa synergy between mechanistic and historicalbiology that led to discoveries that would havebeen impossible without this approach. Webelieve that this integrative approach willadvance the field of evolutionary biologymore rapidly than an approach targetedsolely at fitness and local adaptation becausemore information is used to reach conclu-sions, and because conclusions will be easilytestable. Biologists should not be threatenedby the opportunity to use new tools (e.g., newphylogenetic comparative methods; Schluteret al. 1997; Garland and Ives 2000) to answermechanistic questions, while at the same timeincreasing the strength of their evolutionaryconclusions. The most interesting questions

REFERENCES

Alberch P, Gale E A. 1985. A developmental analysisof an evolutionary trend: digital reduction inamphibians. Evolution 39:8–23.

Amundson R. 2001. Adaptation and development: onthe lack of common ground. Pages 303–334 inAdaptationism and Optimality, edited by S H Orzackand E Sober. Cambridge: Cambridge UniversityPress.

Andersson M. 1994. Sexual Selection. Princeton (NJ):Princeton University Press.

Aristotle. Historia animalium. Translated by D’A WThompson. Oxford: Clarendon Press. 1918.

Arnold E N. 1993. Historical changes in the ecologyand behavior of semaphore geckos (Pristurus, Gek-konidae) and their relatives. Journal of Zoology229:353–384.

Arnold E N. 1994a. Investigating the evolutionaryeffects of one feature on another: does musclespread suppress caudal autotomy in lizards? Journalof Zoology 232:505–523.

Arnold E N. 1994b. Investigating the origins of per-formance advantage: adaptation, exaptation and


ship of the arctiid moth, Cycnia tenera. Experientia43:1029–1031.

Cowles R B, Bogert C M. 1944. A preliminary study ofthe thermal requirements of desert reptiles. Bulle-tin of the American Museum of Natural History 83:265–296.

Darwin C. 1873. Expression of the Emotions in Man andAnimals. London: Murray.

Duellman W E, Trueb L. 1986. Biology of Amphibians.New York: McGraw-Hill.

Ewing A W, Miyan J A. 1986. Sexual selection, sexualisolation and the evolution of song in the Drosoph-ila repleta group of species. Animal Behaviour 34:421–429.

Feder M E, Bennett A F, Huey R B. 2000. Evolutionaryphysiology. Annual Review of Ecology and Systematics31:315–341.

Felsenstein J. 1985. Phylogenies and the comparativemethod. American Naturalist 125:1–15.

Full R J. 1997. Invertebrate locomotor systems. Pages853–930 in Handbook of Physiology, Section 13: Com-parative Physiology, Volume 2, edited by W H Dan-tzler. Oxford: Oxford University Press.

Full R J, Tullis A. 1990. Capacity for sustained terres-trial locomotion in an insect: energetics, thermaldependence and kinematics. Journal of ComparativePhysiology B 160:573–581.

Galis F. 1996. The application of functional morphol-ogy to evolutionary studies. TREE 11:124–129.

Gans C. 1973. Sound production in the Salientia:mechanism and evolution of the emitter. AmericanZoologist 13:1179–1194.

Garland T. 1992. Rate tests for phenotypic evolutionusing phylogenetically independent contrasts.American Naturalist 140:509–519.

Garland T, Adolph S C. 1994. Why not to do two-speciescomparative studies: limitations on inferring adap-tation. Physiological Zoology 67:797–828.

Garland T, Carter P A. 1994. Evolutionary physiology.Annual Review of Physiology 56:579–621.

Garland T, Dickerman A W, Janis C M, Jones J A. 1993.Phylogenetic analysis of covariance by computersimulation. Systematic Biology 42:265–292.

Garland T, Harvey P H, Ives A R. 1992. Proceduresfor the analysis of comparative data using phylo-genetically independent contrasts. Systematic Biol-ogy 41:8–32.

Garland T, Ives A R. 2000. Using the past to predictthe present: confidence intervals for regressionequations in phylogenetic comparative methods.American Naturalist 155:346–364.

Gatten R E, Miller K, Full R J. 1992. Energetics at restand during locomotion. Pages 314–377 in Environ-mental Physiology of the Amphibians, edited by M EFeder and W W Burggren. Chicago (IL): Univer-sity of Chicago Press.

Gerhardt H C. 1994. The evolution of vocalization in

the evolution of nocturnality in geckos. Evolution53:580–599.

Autumn K, Weinstein R B, Full R J. 1994. Low cost oflocomotion increases performance at low tem-perature in a nocturnal lizard. Physiological Zoology67:238–262.

Avery R A. 1982. Field studies of body temperaturesand thermoregulation. Pages 93–166 in Biology ofthe Reptilia, Volume 12, edited by C Gans and F HPough. London: Academic Press.

Balmford A, Thomas A L R, Jones I L. 1993. Aerody-namics and the evolution of long tails in birds.Nature 361:628–631.

Basolo A L. 1990a. Female preference for male swordlength in the green swordtail, Xiphophorus helleri(Pisces: Poeciliidae). Animal Behaviour 40:332–338.

Basolo A L. 1990b. Female preference predates theevolution of the sword in swordtail fish. Science250:808–810.

Basolo A L. 1995. A further examination of a pre-exist-ing bias favouring a sword in the genus Xiphopho-rus. Animal Behaviour 50:365–375.

Bemis W E. 1984. Paedomorphosis and the evolutionof the Dipnoi. Paleobiology 10:293–307.

Bennett A F. 1982. The energetics of reptilian activity.Pages 155–199 in Biology of Reptilia, Volume 13,edited by C Gans and F H Pough. New York: Aca-demic Press.

Bennett A F, John-Alder H B. 1984. The effect of bodytemperature on the locomotory energetics of liz-ards. Journal of Comparative Physiology B 155:21–27.

Bock W J. 1989. Organisms as functional machines:a connectivity explanation. American Zoologist 29:1119–1132.

Borowsky R L, McClelland M, Welsh J. 1995. Arbitrarilyprimed DNA fingerprinting for phylogenetic recon-struction in vertebrates: the Xiphophorus model.Molecular Biology and Evolution 12:1022–1032.

Brandon R N. 1996. Concepts and Methods in EvolutionaryBiology. Cambridge: Cambridge University Press.

Brooks D R, McLennan D A. 1991. Phylogeny, Ecology,and Behavior: A Research Program in Comparative Biol-ogy. Chicago (IL): University of Chicago Press.

Brooks G A, Fahey T D. 1985. Exercise Physiology: HumanBiogenergetics and Its Applications. New York: Mac-millan.

Burke A C, Feduccia A. 1997. Developmental patternsand the identification of homologies in the avianhand. Science 278:666–668.

Cannatella D C, Hillis D M, Chippendale P, Weigt L,Rand A S, Ryan M J. 1998. Phylogeny of frogs ofthe Physalaemus pustulosus species group, with anexamination of data incongruence. Systematic Biol-ogy 47:311–335.

Coates M I. 1994. The origin of vertebrate limbs. Devel-opment (Supplement):169–180.

Conner W E. 1987. Ultrasound: its role in the court-


context. Annual Review of Ecology and. Systematics21:317–340.

Lauder G V. 1991. Biomechanics and evolution: inte-grating physical and historical biology in the studyof complex systems. Pages 1–19 in Biomechanics inEvolution, edited by J M V Rayner and R J Wooten.Cambridge: Cambridge University Press.

Lighton J R B, Feener D H. 1989. A comparison ofenergetics and ventilation of desert ants duringvoluntary and forced locomotion. Nature 342:174–175.

Lorenz K. 1941. Comparative studies of the motor pat-terns of Anatinae. Studies in Animal and HumanBehavior 2:106–114.

Mangum C P, Hochachka P W. 1998. New directionsin comparative physiology and biochemistry:mechanisms, adaptations, and evolution. Physiolog-ical Zoology 71:471–484.

Marcus J M, McCune A R. 1999. Ontogeny and phy-logeny in the northern swordtail clade of Xipho-phorus. Systematic Biology 48:491–522.

Marler C A, Ryan M J. 1997. Origin and maintenanceof a female mating preference. Evolution 51:1244–1248.

Martins E P, editor. 1996. Phylogenies and the Compara-tive Method in Animal Behavior. Oxford: Oxford Uni-versity Press.

Meyer A, Morrissey J M, Schartl M. 1994. Recurrentorigin of a sexually selected trait in Xiphophorusfishes inferred from a molecular phylogeny. Nature368:539–542.

Niewiarowski P H. 2001. Energy budgets, growth rates,and thermal constraints: toward an integrativeapproach to the study of life history variation.American Naturalist 157:421–433.

O’Hara R. 1988. Homage to Clio, or toward an his-torical philosophy for evolutionary biology. System-atic Zoology 37:142–155.

Oster G, Shubin N, Murray J, Alberch P. 1988. Evolu-tion and morphogenetic rules: the shape of thevertebrate limb in ontogeny and phylogeny. Evo-lution 42:862–884.

Proctor H C. 1991. Courtship in the water mite Neu-mania papillator: males capitalize on female adap-tations for predation. Animal Behaviour 42:589–598.

Proctor H C. 1993. Sensory exploitation and the evo-lution of male mating behaviour: a cladistic testusing water mites (Acari: Parasitengona). AnimalBehaviour 44:745–752.

Rauchenberger M, Kallman K D, Morizot D C. 1990.Monophyly and geography of the Panuco Basinswordtails (genus Xiphophorus) with descriptionsof four new species. American Museum Novitates2974:1–41.

Reeve H K, Sherman P W. 1993. Adaptation and the

frogs and toads. Annual Review of Ecology and System-atics 25:293–324.

Ghiselin M T. 1999. Progress and the economy ofnature. Journal of Bioeconomics 1:35–45.

Hanken J, Wake D B. 1993. Miniaturization of bodysize: organismal consequences and evolutionarysignificance. Annual Review of Ecology and Systematics24:501–519.

Harvey P H, Leigh-Brown A J, Maynard Smith J, NeeS, editors. 1996. New Uses for New Phylogenies.Oxford: Oxford University Press.

Harvey P H, Pagel M D. 1991. The Comparative Methodin Evolutionary Biology. Oxford: Oxford UniversityPress.

Hasumi M, Iwasawa H. 1993. Geographic variation inthe pes of the salamander Hynobius lichenatus: acomparison with tetradactyl Hynobius hidamontanusand pentadactyl Hynobius nigrescens. Zoological Sci-ence 10:1017–1027.

Hennig W. 1966. Phylogenetic Systematics. Campaign(IL): University of Illinois Press.

Huey R B. 1982. Temperature, physiology, and theecology of reptiles. Pages 25–91 in Biology of Repti-lia, Volume 12, edited by C Gans and F H Pough.New York: Academic Press.

Huey R B, Niewiarowski P H, Kaufmann J, Herron J C.1989. Thermal biology of nocturnal ectotherms: issprint performance of geckos maximal at low bodytemperatures? Physiological Zoology 62:488–504.

Jacob F. 1977. Evolution and tinkering. Science 196:1161–1166.

John-Alder H B, Bennett A F. 1981. Thermal depen-dence of endurance and locomotory energetics ina lizard. American Journal of Physiology 241:R342-R349.

John-Alder H B, Lowe C H, Bennett A F. 1983. Ther-mal dependence of locomotory energetics andaerobic capacity of the gila monster (Heloderma sus-pectum). Journal of Comparative Physiology B 151:119–126.

Kirkpatrick M, Ryan M J. 1991. The paradox of the lekand the evolution of mating preferences. Nature350:33–38.

Kleiber M. 1961. The Fire of Life. New York: Wiley.Konishi M. 1994. An outline of recent advances in

birdsong neurobiology. Brain, Behavior and Evolu-tion 44:279–285.

Kram R, Taylor C R. 1989. Why does metabolic rateincrease with running speed? Proc. Int. Union phys-iol. Sci. 17:430–431.

Lande R. 1978. Evolutionary mechanisms of limb lossin tetrapods. Evolution 33:73–92.

Lauder G V. 1982. Historical biology and the problemof design. Journal of Theoretical Biology 97:57–67.

Lauder G V. 1990. Functional morphology and system-atics: studying functional patterns in an historical

Sereno P C, Novas F E. 1992. The complete skull andskeleton of an early dinosaur. Science 258:1137–1140.

Sessions S K, Larson A. 1987. Developmental corre-lates of genome size in plethodontid salamandersand their implications for genome evolution. Evo-lution 41:1239–1251.

Shaw K. 1995. Phylogenetic tests of the sensory exploi-tation model of sexual selection. Trends in Ecology& Evolution 10:117–120.

Shubin N. 1995. The evolution of paired fins and theorigin of tetrapod limbs. Evolutionary Biology28:39–95.

Shubin, N, Alberch P. 1986. A morphogenetic approachto the origin and basic organization of the tetrapodlimb. Evolutionary Biology 20:319–387.

Shubin N, Tabin C, Carroll S. 1997. Fossils, genes andthe evolution of animal limbs. Nature 388:639–648.

Shubin N, Wake D B. 1996. Phylogeny, variation, andmorphological integration. American Zoology 36:51–60.

Stearns S C. 1992. The Evolution of Life Histories. Oxford:Oxford University Press.

Taylor C R, Schmidt-Nielsen K, Raab J L. 1970. Scalingof energetic cost of running to body size in mam-mals. American Journal of Physiology 219(9):1104–1107.

Thompson G G, Withers P C. 1997. Standard and max-imal metabolic rates of goannas (Squamata: Var-anidae). Physiological Zoology 70:307–323.

Turner B J. 1982. The evolutionary genetics of a uni-sexual fish, Poecilia formosa. Pages 265–305 in Mech-anisms of Speciation, edited by C Barigozzi. NewYork: Alan R. Liss.

Wake D B. 1991. Homoplasy: the result of naturalselection or evidence of design limitations? Ameri-can Naturalist 138:543–567.

Weinstein R B, Full R J. 1994. Thermal dependenceof locomotor energetics and endurance capacity inthe ghost crab, Ocypode quadrata. Physiological Zool-ogy 67:855–872.

Wilczynski W, Rand A S, Ryan M J. 1995. The process-ing of spectral cues by the call analysis system ofthe tungara frog, Physalaemus pustulosus. AnimalBehaviour 49:911–929.

goals of evolutionary research. Quarterly Review ofBiology 68:1–31.

Reeve H K, Sherman P W. 2001. Optimality and phy-logeny: a critique of current thought. Pages 64–113 in Adaptationism and Optimality, edited by S HOrzack and E Sober. Cambridge: Cambridge Uni-versity Press.

Rosen D E. 1979. Fishes from the uplands and inter-montane basins of Guatemala: revisionary studiesand comparative geography. Bulletin of the AmericanMuseum of Natural History 162:263–375.

Rosenthal G G, Evans C S. 1998. Female preferencefor swords in Xiphophorus helleri reflects a bias forlarge apparent size. Proceedings of the National Acad-emy of Sciences of the United States of America 85:4431–4436.

Roth G, Blanke J, Wake D B. 1994. Cell size predictsmorphological complexity in the brains of frogs andsalamanders. Proceedings of the National Academy of Sci-ences of the United States of America 91:4796–4800.

Ryan M J. 1985. The Tungara Frog, A Study in SexualSelection and Communication. Chicago (IL): Univer-sity of Chicago Press.

Ryan M J. 1998. Receiver biases, sexual selection andthe evolution of sex differences. Science 281:1999–2003.

Ryan M J, Drewes R C. 1990. Vocal morphology of thePhysalaemus pustulosus species group (family Lep-todactylidae): morphological response to sexualselection for complex calls. Biological Journal of theLinnean Society 40:37–52.

Ryan M J, Fox J H, Wilczynski W, Rand A S. 1990.Sexual selection for sensory exploitation in thefrog Physalaemus pustulosus. Nature 343:66–67.

Ryan M J, Keddy-Hector A. 1992. Directional patternsof female mate choice and the role of sensorybiases. American Naturalist 139:S4-S35.

Ryan M J, Rand A S. 1993. Sexual selection and signalevolution: the ghost of biases past. PhilosophicalTransactions of the Royal Society, Series B 340:187–195.

Savageau M A. 1991. Reconstructionist molecular biol-ogy. New Biologist 3:190–198.

Schluter D, Price T, Mooers A O, Ludwig D. 1997.Likelihood of ancestor states in adaptive radiation.Evolution 51:1699–1711.


Volume 77, 4 2002 The Quarterly Reviewibdev.berkeley.edu › labs › wake ›...

Documents

Transcript of Volume 77, 4 2002 The Quarterly Reviewibdev.berkeley.edu › labs › wake ›...