How to Study Adaptation (and Why To Do It That Way)€¦ · HOW TO STUDY ADAPTATION (AND WHY TO DO...

HOW TO STUDY ADAPTATION (AND WHY TO DO IT THAT WAY)

Mark E. OlsonInstituto de Biologıa, Universidad Nacional Autonoma de Mexico, Tercer Circuito

s/n de Ciudad UniversitariaMexico DF 04510 Mexico

e-mail: [email protected]

Alfonso Arroyo-SantosFacultad de Filosofıa y Letras, Universidad Nacional Autonoma de Mexico, Circuito Interior

s/n de Ciudad UniversitariaMexico DF 04510 Mexico

Centro de Informacion Geoprospectiva, Berlın 29B, Del Carmen CoyoacanMexico DF 04100 Mexico

e-mail: [email protected]

keywordsabduction, Bayesianism, comparative methods, deduction, optimality models,

population biology

abstractSome adaptationist explanations are regarded as maximally solid and others fanciful just-so

stories. Just-so stories are explanations based on very little evidence. Lack of evidence leads tocircular-sounding reasoning: “this trait was shaped by selection in unseen ancestral populations andthis selection must have occurred because the trait is present.” Well-supported adaptationist explana-tions include evidence that is not only abundant but selected from comparative, populational, andoptimality perspectives, the three adaptationist subdisciplines. Each subdiscipline obtains its broadrelevance in evolutionary biology via assumptions that can only be tested with the methods of the othersubdisciplines. However, even in the best-supported explanations, assumptions regarding variation,heritability, and fitness in unseen ancestral populations are always present. These assumptions areaccepted given how well they would explain the data if they were true. This means that some degreeof “circularity” is present in all evolutionary explanations. Evolutionary explanation corresponds notto a deductive structure, as biologists usually assert, but instead to ones such as abduction orBayesianism. With these structures in mind, we show the way to a healthier view of “circularity” inevolutionary biology and why integration across the comparative, populational, and optimalityapproaches is necessary.

The Quarterly Review of Biology, June 2015, Vol. 90, No. 2

Copyright © 2015 by The University of Chicago Press. All rights reserved.

0033-5770/2015/9002-0003$15.00

Volume 90, No. 2 June 2015THE QUARTERLY REVIEW OF BIOLOGY

167

This content downloaded from 132.248.9.8 on Wed, 13 May 2015 21:29:32 PMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

Introduction

THE study of adaptation aims to under-stand the fit between organismal form

and function across the living world. Someinferences of adaptation are accepted assolid, almost without question. Many aredismissed as just-so stories, while others areaccused of being circular arguments. Herewe outline the difference between a just-sostory and a solid, widely accepted adaptation-ist explanation. We examine the “circularity”that is so often criticized in evolutionary bi-ology, why it is there, and the important partit plays in all evolutionary explanations. Weshow that the difference between a just-sostory and a well-accepted adaptationist expla-nation is the amount of direct evidence avail-able. Like all attempts to infer things aboutthe deep past, adaptationist explanations aremost trusted when they have a lot of evi-dence from a diversity of sources. That moreevidence is better is obvious. Not so obviousis that some strategies for getting this infor-mation are better than others. Even less ob-vious is that some degree of “circularity” isnecessarily present in all adaptationist expla-nations, no matter how well supported. Thiscircularity can be thought of in terms of in-ference types such as induction, abduction,or Bayesianism. Whatever the name of infer-ence type applied, the need for a diversity ofevidence leads us to conclude by calling forintegration across the adaptationist subdisci-plines. We start by exploring “just-so stories”and the three main adaptationist subdis-ciplines before seeing how to structuremaximally supported explanations of thefit between form and function across liv-ing things.

just-so storiesAdaptationist scenarios are often criticized

as “just-so stories.” The term comes from thetitle of Rudyard Kipling’s (1902) children’sbook of origin stories. In the context of ad-aptation, it is a derogatory term, implyingthat a given adaptationist explanation is un-falsifiable, fanciful, and is accepted not be-cause of evidence, but based on plausibilityalone (Lennox 1991; Durrant and Haig2001). Reference to Kipling and just-so stories

is often attributed to Gould and Lewontin’s1979 spandrels paper (Alcock 1998; Hull2001; Hall 2002; Travis 2003; Sosis 2009;Frost-Arnold 2010), even by Gould himself(1997, 2002). However, the spandrels papermakes no mention of Kipling and does notuse the “just-so story” term, although otheressays by Gould (1977, 1978, 1997, 2002) do.Just-so stories are of interest here becausethey reveal the structure of little-trusted ad-aptationist explanations and thus the way toones regarded as solid.

Adaptationist just-so stories are criticizedfor two reasons, one being “circularity” andthe other their freedom to proliferate. Just-sostories are criticized for “circularity” becausethe presence of a given trait in current or-ganisms is used as the sole evidence to inferheritable variation in the trait in an ancestralpopulation and a selective regime thatfavored some variants over others. This un-observed selective scenario explains the pres-ence of the observed trait, and the onlyevidence for the selective scenario is traitpresence (e.g., Griffiths 1996; Frost-Arnold2010). Gould (1996) called the giraffe’s neckthe “canonical just-so story” because the storyis so often repeated. The story consists of thenotion that selection favored variants withrelatively long necks in short-necked ances-tral giraffe populations as a result of theirgreater ability to obtain food. Giraffe necksare long as a result of this unobservable se-lection on heritable ancestral variation inneck length, and this selection must haveoccurred because giraffe necks are long andbecause giraffes today can eat leaves from talltrees (Figure 1). The relative lack of infor-mation that leads to this “circular” structurealso allows different potential explanationsto proliferate.

Multiple accounts of selection in thedistant past can be devised to explain the pres-ence of any trait. In the absence of informa-tion beyond simple trait presence, it is hardto choose the best from among these alter-native accounts. For example, the long neckof the giraffe might have been favored inreaching high leaves. Alternatively, perhapsmales with longer necks prevailed in battlesfor females; perhaps long necks in males de-velopmentally imply long ones in females,

168 Volume 90THE QUARTERLY REVIEW OF BIOLOGY



and so all giraffes have long necks. Basedonly on the trait “long neck” and that thenecks are at least occasionally involved inreaching high leaves as well as in male-malebattles, it is hard to pick one or the other ofthese explanations as the best one (Gould1996; Simmons and Scheepers 1996). Whenthe available data are unable to distinguishconvincingly between two or more hypotheses,the hypotheses are said to be underdeter-mined by the data (Ladyman 2002; Dietrichand Skipper 2007). The literature on adapta-tion is full of examples of underdetermination(Forber 2009). For example, carrion flowersmight have been favored in drylands with adearth of bees and an abundance of flies. Butherbivores are apparently driven away by thescent of rotting flesh, so maybe stinky flowers

are instead an herbivore deterrent in plant-scarce drylands (Lev-Yadun et al. 2009). Thecolorful peeling bark of tropical trees in manyplant families has been seen as an adaptationpermitting photosynthesis of the living bark(Franco-Vizcaıno et al. 1990), a mechanism toshed epiphytes and thereby reduce mechani-cal loads (Stevens 1987), or even as a attractantof fruit dispersers (Rzedowski and Kruse 1979).MacColl (2011) details no less than six under-determined adaptive explanations for the ar-mor plates of sticklebacks. Evidence beyondsimple trait presence is needed to choosebetween hypotheses and to minimize the “cir-cularity” of just-so explanations. But some strat-egies for gathering evidence are better thanothers. To see why, it is necessary first to exam-ine the three main adaptationist approaches.

“Circularity” and the ThreeAdaptationist Subdisciplines

Approaches for studying adaptationfall into three subdisciplinary categories:comparative, populational, and optimality.Just-so stories, understood as “circular” argu-ments with little direct evidence, can befound in all of these approaches. The intentof this section is to describe the generalitiesof each of the three approaches briefly, to beable to examine how each is associated with“circularity” when little direct evidence isavailable. After, we will show that some “cir-cularity” is in fact natural and necessary in allexplanations involving adaptation, henceour scare quotes. Importantly, we show thatthe assumptions being accepted using “circu-lar” reasoning are those that give each ap-proach its broad relevance in evolutionarybiology. We then argue that recognition ofthese assumptions points the way to morerobust adaptationist explanations by simulta-neous use of the three approaches.

the comparative methodThe comparative method detects adapta-

tion through convergence (Losos 2011). Abasic version of comparative studies, per-haps the one underpinning most state-ments about adaptation, is the qualitativeobservation of similar organismal features insimilar selective contexts. An example is the

Figure 1. Circularity and the Giraffe NeckJust-So Story

Giraffes in present-day populations use their longnecks to reach leaves from tall trees. The presence oflong necks is explained as the result of unobservedand unobservable selection in ancestral populationsin the distant past. It is assumed that there was onceheritable variation in short-necked ancestral giraffepopulations, and that this variation had fitness conse-quences. Specifically, longer-necked individuals werefavored because of greater access to food. This entireselective scenario, variation, heritability, fitness, andall is accepted as true because giraffes today at leastsometimes use their necks to reach food from talltrees. The selective scenario in turn explains whygirafffes have long necks. An adaptationist scenariowith little direct evidence beyond the pattern to beexplained is known as a just-so story.

June 2015 169HOW TO STUDY ADAPTATION



observation that aquatic animals often havestreamlined shapes and fins, regardless ofwhether they are fish, whales, ichthyosaurs,eurypterids, or squids. Convergence thinkingfinds a quantitative expression in methodsthat seek statistical associations in cross-speciesdata (Bell 1989; Martins 2000; Blomberg etal. 2003). These include both studies ofhow organismal attributes change predict-ably across environmental gradients, suchas the global negative relationship betweenthe toothiness of plant leaves and temper-ature (Peppe et al. 2011), as well as be-tween organismal attributes, such as bonelength-diameter scaling (Christiansen 1999;Swartz and Middleton 2008; Kilbourne andMackovicky 2012). Other examples of com-parative approaches include those that aimto detect deviations from neutral substitutionpatterns in multiple molecular alignments(Nielsen 2009). Across this methodologicaldiversity, the use of cross-species variationunites all comparative methods.

It is easy to construct “circular” just-so sto-ries based on comparative data. These “cir-cular” stories anchor comparative methodsas a fundamental source of information forconstructing evolutionary explanations. Ifthe pattern in Figure 2 is regarded as reflect-ing adaptation, then it is implied that theoccupied part of the plot corresponds tocombinations of X and Y that are of higherfitness than the surrounding space (Arnold2005). This view effectively asserts that “thisspace is filled in nature; because selectionfavors variants with high fitness, this spacemust be of high fitness. I know that this is thespace corresponding to high fitness, becauseit is filled” (Figure 2).

Without assumptions regarding evolution-ary process, comparative data would be nomore than blank descriptions of how traitvalues are distributed. Assumptions aboutpopulation-level phenomena such as devel-opmentally possible variation, heritability,and fitness are the vital glue that connectcomparative patterns to notions of evolution-ary process and thereby give comparative pat-terns relevance beyond simple description(Table 1). However, based on the pattern inFigure 2 alone, any adaptive explanation is

rightly considered a just-so story in that it ishard to choose the adaptationist scenarioover other potential explanations. For exam-ple, the pattern might be observed becausethe empty spaces are developmental impos-sibilities, even though they would be of muchhigher fitness than the observed morpholo-gies (Olson 2012). The pattern might bedue to drift or other chance alignments(Brandon and Fleming 2014). Based on thepattern in Figure 2 alone, all of these expla-nations will have the “circular” structureshown there. Similar things happen with theother adaptationist approaches.

Figure 2. Adaptationist Explanations andCircularity: General Case

Points on the graph refer to mean species traitvalues, and the line an allometric regression fit. The“adaptation” view sees the entire space defined by themean values as potentially accessible in ontogeny, butthat the configurations corresponding to the emptyspaces are eliminated by selection. The “limited de-velopmental potential” view sees allometry as themanifestation of a lack of developmental alternatives.Both perspectives make untested assumptions: adap-tationist reasoning regarding empty spaces is shownabove the scaling line, and thinking in terms of de-velopmental potential below. Both loops can be readstarting at a. or b., i.e., “a. This space is empty. There-fore, b. These morphologies must be developmentallypossible but of low fitness,” or “b. These morpholo-gies must be developmentally inaccessible, thereforea. This space is empty.” In both cases a. is used to inferb., which in turn is inferred based on a. The means tostrengthen these “circular” inferences is via additionallayers of evidence (see Figure 3).




populational approachesAnother approach for studying adaptation

focuses on the population level. Popula-tional approaches include a wide range oftools for testing hypotheses of adaptation,from detailed studies of reproductive biologyor intrapopulational variation to quantitativegenetics (e.g., Lande and Arnold 1983; Bell2008; Olson 2012). These studies reason thatbecause natural selection acts on interindi-vidual variation, the population level is thedomain appropriate for studying adaptation.These studies focus directly on variation, her-itability, and fitness between potentially com-peting individuals. In a particularly completestudy, Travers et al. (2003) documented vari-ation in the curvature of floral nectar spurs

in populations of the jewelweed Impatienscapensis. They found that the spurs, tubularprojections from the backs of the flowersthat attract pollinators with sugary nectar,varied in projecting almost straight back toalmost completely recurved, with the tip fac-ing the front of the flower. For a trait to besubject to selection, variation must be herita-ble, and many techniques are available forestimating the degree to which offspringtend to resemble their parents in a giventrait. Travers et al. (2003) estimated herita-bility using a selfing protocol followed by aregression of progeny spur curvature onparental curvature. They found a markedtendency for parental curvature to predictprogeny curvature. In addition to being her-

TABLE 1The three principal approaches for studying adaptation, some typically cited advantages and disadvantages,

and the key assumptions that give each method its relevance to evolutionary biology in general

Definition Advantages Disadvantages Key assumption

Comparative/convergence

The convergence onsimilar morphologiesin similar selectivecontexts fromancestors with differentstates suggestsadaptation

Studies species in naturethat are thedescendants ofnatural evolutionaryprocesses; examinespatterns applicableacross evolutionarilyrelevant timespansand many species

Does not examinefitness or heritabilitydirectly; often relieson ancestralcharacter statereconstructions orassumptions of tempoand mode that areimpossible to test

Comparative patternsare produced bypopulation-levelprocesses,involvingdevelopmentalvariation,heritability, anddifferential fitness

Populational Studies the raw materialof selection directly,i.e., fitness/peformancedifferences associatedwith heritable within-species trait variation

The focal approach fordirectly examiningintraspecific variation,heritability, and thefitness impact of thisvariation

Examines relativelyminor characters thathave not gone tofixation;extrapolation ofresults to multiplespecies and largetimescales debated

The population-levelprocesses beingstudied are thoseshaping the entirediversity of life

Optimality Predicts theconfiguration(s)maximizing aperformance/fitnesscriterion given generalbiophysical principlesand a set of competingconsiderations;concordance betweenmodel and naturesuggests adaptation

Based on models thatexplicitly incorporatecompeting demandson an organism; evena lack of model-naturecorrespondence isuseful because ithighlights elementsthat need to beconsidered; explicitlyincludesfitness/performanceindices

The process of selectionof variables is oftencriticized; inaddition, there is nodirective emergingfrom nature toindicate where thecutoff in fit betweennature and themodel should betaken as congruentwith modelpredictions or not

Adaptation is theonly plausibleexplanation fortrait optimality




itable, a trait subject to selection must beassociated with differences in fitness. Traverset al. (2003) studied the way that differentspur curvatures resulted in different hum-mingbird visitation times. They found thatflowers with more recurved spurs were asso-ciated with longer visitation times and moregrains of pollen carried away. That spur cur-vature is heritable and associated with differ-ential reproductive success is compatiblewith the hypothesis that curvature can besubject to selection and that some predict-able patterns of variation in curvature couldbe adaptive.

As compared to comparative approaches,populational methods invoke a different butequally important set of “circular” assump-tions. Like comparative methods, theseassumptions also have to do with the funda-mental justification that gives populationalmethods their general relevance in evolu-tionary biology. Populational methods in-volve detailed studies of very geographicallyand restricted sets of organisms under oftenunusual situations, e.g., purebred lines, overshort times. This approach is of direct impor-tance mostly to applied activities such asplant or animal breeding, in which humanswish to produce a given selective responsein a given time (Pigliucci and Schlichting1997). The relevance of populational studiesto evolutionary biology at large is only via theassumption that population-level processesidentical to those being studied in fact playimportant roles in generating the patterns oftrait distribution observed over geologicaltime and across clades (Table 1). This as-sumption is exactly the one invoked in forg-ing the link between the fossil record andpopulation genetics of the Modern Synthesis(e.g., Simpson 1953). In this way, the greatjust-so story of population biology is that verylocal population-level phenomena are insome way isomorphic with the factors shap-ing life on Earth at large. In populationalstudies, “circularity” takes the form that her-itable variation with fitness consequencesshaping local situations is taken as an expla-nation of the organismal form-function fitglobally, and the form-function fit is taken asconfirming population-level selection as theshaping factor. Based only on population-

level data, this assumption is as much ajust-so story as the unobserved variation andfitness in the comparative example above(Figure 2). Missing from both the compar-ative and populational approaches are ex-plicit notions of the biophysical reasons be-hind a given variant being favored,information provided by the optimalityperspective.

optimality modelingOptimality methods examine the ways that

performance or fitness differences emergeas the result of predictable biophysicalprinciples. Given a series of competing con-siderations, optimality models predict thecombination or combinations that maximizefitness or some other performance criterion(Parker and Maynard-Smith 1990; Vincentand Brown 2005; Potochnik 2009). For ex-ample, the most influential optimality modelof the past 20 years is that of West, Brown,and Enquist (West et al. 1997; WBE). WBEasks what organismal geometry simultane-ously maximizes metabolite exchange sur-face area and minimizes transport distancesand therefore transport energy investment.For a given amount of tissue, metabolite ex-change area is maximal in a plane and trans-port distances are minimized in a sphere, sothese functions cannot be globally maximalsimultaneously. The intermediate “least bad”configuration is a fractal branching one, asfound in trees, lungs, kidneys, and mostother transport systems that innervate livingthings. Optimality models make no neces-sary reference to a given level of biologicalorganization. That any given model does sois a contingent fact of that model, not a prop-erty of optimality models in general. For ex-ample, the exact same interpretation of WBEas applied to plant vasculature (West et al.1999; Petit and Anfodillo 2009) has beentested with reference to individual plants(Bettiati et al. 2012), within populations ofthe same species (Petit et al. 2010), andacross the flowering plants (Olson et al.2014; see also West et al. 2002). In all of thesecases, the empirical results, down to the allo-metric scaling exponents, are identical withthose predicted by the model. Coincidencebetween optimality predictions and nature




by chance and not via the action of selectionseems very unlikely.

However unlikely chance model-naturecoincidence might seem, explanations builtonly on optimality models also involve “cir-cular” acceptance of assumptions (Griffiths1996). Coincidence between optimality pre-dictions and nature that are due to chanceor any other nonadaptive process is regardedas so unlikely that the trait must be due toselection (Orzack and Sober 1994). An opti-mality just-so explanation takes the form thatcoincidence between model and nature im-plies selection; this unobserved selection inturn explains why there is coincidence be-tween the model and nature. As we will show,this sort of “circularity” can be minimizedbut never eliminated.

Adaptationism: A VirtuousCircularity

Whether comparative, populational, oroptimality, all of the approaches for studyingadaptation have built-in “circular” assump-tions, and these assumptions are the onesthat justify each method as being of evolu-tionary relevance (Table 1). Without theseassumptions, each of these methods pro-duces only descriptive accounts of verylimited local interest. When combined, anadaptationist explanation that includescomparative, populational, and optimalitydata is always considered well supported andmuch more than a just-so story (cf. Forber2010). As we will show, however, even thebest-supported explanation still involves “cir-cularity.”

Although “loops” of reasoning are easy todetect in the just-so examples in Figures 2and 3A, they are still present even in thebest-supported studies of adaptation (Figure3B). If a given pattern has an adaptive cause,then by definition at some time in the past,not just the observable present or momentscaptured in fossil traces, heritable variationwith fitness consequences was present (Leroiet al. 1994; Forber and Griffith 2011; includ-ing assimilable plastic variation, see West-Eberhard 2003). Selection on this variation isassumed to have led to the pattern observedtoday (Figure 3B). All adaptationist explana-tions at some point invoke these assumptions

about unseen and unexaminable sets oforganisms. That almost all swimming or-ganisms have streamlined bodies and finscertainly suggests that these features areadaptive. Their being adaptive means that insome ancestral populations there was varia-tion leading to differential survivorshipand reproduction (Scriven 1959). Thesepopulations will never be seen, but thatthey must have existed is accepted becausetheir having existed would explain the dataso well if it were true (Figure 3). As moredirect evidence is gathered (from Figures3A to 3B), the relative importance of “cir-cular” loops diminishes. But no matter howmuch direct evidence is accumulated orwhat method is used, the existence of un-seen populations is assumed. That traitstoday are distributed the way that they aresuggests that these populations must haveexisted, and the assumed existence ofthese populations helps explain why traitsare distributed the way that they are. Thisapparent “circularity” is what we mean by“loopy” (cf. Rieppel 2003). “Loopiness”does not undermine the solidity of rea-soning regarding adaptation. Because somuch information is available from so manydifferent sources, the notion that the pres-ence of fins in aquatic animals involves ad-aptation seems as solid an assertion as can behoped for in science (Figure 4).

In fact, scientific explanations in gen-eral, not just evolutionary ones, have aloopy structure. This statement from as-tronomy is an excellent example of loopyreasoning: “[T]he transmission spectrum of[the super-Earth exoplanet] GJ 1214b [is ob-served to be featureless] at near-infraredwavelengths . . . [and its] atmosphere mustcontain clouds to be consistent with thedata” (Kreidberg et al. 2014:69). The as-sumption that the planet has clouds is ac-cepted because it would explain the data bestif it were true. The featureless near-infraredtransmission spectrum is observed becausethere are clouds; there must be clouds be-cause of the featureless spectrum. The au-thors marshal other layers of direct evidencein favor of their interpretation of a cloudyplanet, building an explanatory structure ex-




actly analogous to that in Figure 3B, loopsand all.

Loopiness is well documented by philoso-phers of biology. Griffiths’ (1996) “adapta-tionist abduction” (see also Ruse 1975;Sterelny and Griffiths 1999; Durrant and

Haig 2001) is an account of “loopy” reason-ing in terms of an inferential strategy knownas abduction or inference to the best expla-nation (Lipton 2008). Griffiths’s accountmaps the “loops” of reasoning that optimal-ity studies use to construct adaptationist ex-

Figure 3.




planations. The notion of abduction was in-troduced by Charles Sanders Peirce in 1903as a type of inference of the form

The fact C is observed;If A were true, C would be a matter ofcourse,Hence, there is reason to suspect that A istrue.

The assumption A is accepted because itwould explain the data so well if it were true,exactly the way assumptions are accepted inFigure 3. Different authors have proposedepistemic and quantitative criteria for evalu-ating abductive statements (for a review seeDouven 2011), but whatever way that abduc-tive statements are judged, they involve“loopy” interplay between the phenomenonto be explained and the explanation itself.This interplay, in the form of assump-tions that are accepted as a function of howwell they would explain the data if they weretrue, is precisely how explanations in evolu-tionary biology are constructed (Haig andDurrant 2002; Ladyman 2002 offers afriendly introduction to inference types).Abduction is a form of reasoning that corre-sponds well to the way that studies of adap-tation are genuinely carried out, but it is byno means the only one.

Loopiness can also be found in familiarstatistical procedures. For example, the es-sence of Bayesian statistical methods is thatconfidence in a given hypothesis is strength-ened in the light of new evidence. In Bayes-ianism, the probabilities involved are read asa measure of belief in a given hypothesis.

Evidence allows scientists to confirm or dis-confirm the belief they have in the hypothe-sis (Table 2; Fisher 1985; Okasha 2000). Withits back-and-forth relation between theposterior probability, the priors, and thehypotheses under consideration, Bayesianreasoning is an excellent example of loopyreasoning.

Our proposal to recognize “loopiness” ismotivated not so much by the need to tagadaptationist reasoning as “abductive,”“Bayesian,” or any other term, but becauserecognition of loopiness shows how to makeevolutionary explanations stronger. We havetried to show that the construction of scien-tific explanations is “loopy” by arguing firstand foremost from common sense biologicalpractice. The aim in the present section wasto show briefly that this take is not just ourpersonal view. Instead, loopiness is old newsto philosophers who study the way that sci-entific explanations are constructed. It is,however, news to most biologists, and there-fore has important consequences for every-day biological practice. At this point, basedon the reasoning above and their own expe-rience, many biologists will be convincedthat evolutionary explanations are built witha loopy structure and will want to know whatthis means for the study of adaptation. Theycan safely skip the next section. Others willbe left wondering why evolutionary biologistsspend so much time insisting that their sci-ence is deductive, nonloopy, and even Pop-perian. The next section gives a little moredetail for these readers.

Figure 3. The “Loopy” Structure of Adaptationist Explanations: General CaseA. It is easy to see that, when very little evidence is available, adaptive explanations of a given pattern have

an element of “circularity” or loopy nature. This structure involves loops of reasoning in which, of the possibleexplanations (e.g., adaptation, limited developmental potential, drift/chance), adaptation is chosen as seem-ing the most likely. Declaring a given pattern as the result of adaptation immediately implies assuming manythings about variation, heritability, performance, and fitness. These assumptions are accepted given how wellthey would explain the data if they were true. Biologists often call these adaptive explanations “just-so stories”and demand additional evidence. B. The “loopy” structure of adaptationist explanations persists even whenabundant direct evidence is available. Explanations with diverse sources of direct evidence seem as solid as anyin any branch of science. For example, the presence of fins in aquatic animals seems certain to involveadaptation (Figure 4). However, these explanations still require acceptance of assumptions based on how wellthey would explain the data if they were true. Adding more layers of direct evidence diminishes the relativeimportance of “loops” of reasoning, but they never disappear entirely.




Five Myths of Hypothetico-Deductive Evolutionary BiologyThe arguments above run counter to

more than 50 years of biological tradition.Evolutionary biology has a long-standing cus-tom of regarding all forms of “circularity”with suspicion (Table 3; Hull 1967), reject-ing “induction,” and affirming biology’sdeductive, Popperian, and falsificationistnature. In fact, like science in general,evolutionary biology is neither deductive norPopperian. Induction and “circularity” in theloopy sense we use here are the not the badwords biologists traditionally make them outto be. Terms like “hypothetico-deductive”and “falsificationism” imply very differentthings in biology than they do in their orig-

inal contexts in the philosophy of sciencefrom which biologists co-opted them. Be-cause traditional biological positions mightmake many biologists resist accepting “loopi-ness,” in this section, we give some additionaldetail. Because these notions are pervasive inbiology and confusing, we give some orderby treating them as five “myths” of hypothetico-deductive evolutionary biology.

Myth 1: Deduction is the standard of goodscience. The importance of deduction versusloopy thinking has had two different trajec-tories in the philosophy of science and inbiology. The deductive vision of science hasbeen controversial in the philosophy of sci-ence since the birth of the field in the early20th century with the Vienna Circle. The

Figure 4. The “Loopy” Structure of Adaptationist Explanations: The Fins and Fusiform Bodies ofAquatic Animals

That the possession of fins and a streamlined shape represent the effects of selection in an aquaticenvironment seems certain. This explanation is so solid because there is an abundance of evidence from acrossthe three main adaptationist subdisciplinary approaches. There is the comparative observation that unrelatedaquatic animals, such as squid, whales, fish, and eurypterids, have or had streamlined bodies and fins. From apopulational point of view, it is clear that there is heritable variation in many body and fin traits, and that thisvariation is associated with performance differences, as in domestic goldfish breeds. That selection on thesetraits can be operative now strongly suggests that it also did in unobserved ancestral populations. Moreover,optimality models based on fluid mechanics illuminate the biomechanical basis for performance differencesbetween variants. But no matter how much direct evidence accumulates, some reasoning “loops” remain. Atsome point in the distant past, there were presumably populations without these traits, and in which they arose,varied, and were favored. These ancestral populations are impossible to observe. The assumptions regardingtheir characteristics are accepted because they would explain the data so well if they were true.




Vienna Circle was formed by a group of phi-losophers who are regarded largely as de-fenders of a deductive vision of science(Uebel 2007). This vision had its crowningmoment with the deductive-nomological(DN) model developed by Hempel and Op-penheim in the late 1940s. The “deductive”part of the DN model specified that scientificexplanations are deductive arguments inwhich the phenomenon to be explained, theexplanandum, is the logical consequence of aset of premises, the explanans. The “nomo-logical” aspect of the DN model said that theexplanans must contain at least one law ofnature, nomos being Greek for “law.” Ac-cordingly, for any valid deduction, acceptingthe premises of an argument implies accept-ing the conclusion. The DN model was crit-icized from the outset. Not all philosophers

believed that all of science could really be fitinto a deductive mold (Uebel 2007). In ad-dition, it is not entirely clear what a “law ofnature” actually is, or how to tell one from ageneralization or a model (Salmon 1989).Another criticism was that perfectly valid de-ductions can result in perfectly invalid expla-nations, e.g., the conclusion that the heightof a flagpole is caused by its shadow (seeBromberger 1966). As a result, philosophersquickly got over the notion that science hasto be entirely deductive (see, for example,Scriven 1959; Salmon 1989; Ladyman 2002).In the ensuing decades, the DN model hasbeen replaced by a consensus viewing scien-tific explanations as “loopy” (Sober 2008),not deductive.

To show that studying adaptation cannotbe a deductive enterprise, we can compare

TABLE 2Major types of inference

Type ofinference

A conclusion is correctbecause. . .

Formalrepresentation* Example

Kind ofexplanation

Deduction If the premises are true, theconclusion is true

If A then B.A.--------------------------------Then B.

If it rains the floor is wet.It rains.----------------------------The floor is wet.

Pattern

Induction Numerous observations fromnumerous sources such asexperimental results,statistical analyses, andprevious informationstrengthen the cogency ofa conclusion

If A then probably B.A.|||||||Then probably B.

80% of the time, when it rains,the floor gets wet.

It rains.|||||||There is an 80% chance that

the floor is wet.

Pattern/Loopy

Abduction Theory and availableevidence make it likelythat a conclusion iscorrect

A.If B obtained then A

would be a matterof course.

B.|||||||Then A.

The floor is wet.If it rained, then the floor

being wet would be a matterof course.

It rained.|||||||Rain is likely the reason why

the floor is wet.

Loopy

Bayesianism Belief in a conclusionincreases as more andmore relevant evidence isgathered

P(A�B)�P(B�A)*P(A)P(B)

I believe that it rained (A).The floor is wet (B).Then my belief that it rained

has been increased giventhe available evidence

P(A�B)P(A).

Loopy

*Following convention, deductive inferences are written with the premises and the conclusion being separated by a single lineto indicate that they are “truth preserving,” i.e., that given the truth of the premises the conclusion will be true as well. Byconvention two lines indicate nontruth preserving arguments, such as those that are upheld by probability and loopy reasoning.Inductive arguments are often though not necessarily probabilistic, given that there are many different ways to compute thestrength of a given conclusion. Bayesianism is included to exemplify a popular form of reasoning using probabilities, but thesame example can be generalized to other statistical procedures, for example, to Neyman-Pearson hypothesis testing. InBayesianism, beliefs are quantified in probabilistic terms. P(A�B) is read as “the probability of A given B.”




“loopy” explanations with what we will call“pattern” explanations. Pattern explanationsare considered correct if they conform to aparticular structure or pattern (Nagel 1961;Schaffner 1993; Strevens 2008). This is thecase of deductive explanations, which must

conform to a pattern given by the rules offormal logic. If an argument conforms tosuch a pattern, then it is certain that its con-clusion is correct (see, for example, Hempel1965). In contrast to loopy ones, explana-tions built on deduction do not shore up

TABLE 3Examples of the ways that biologists discuss circularity in evolutionary biology and especially in the study of

adaptation

Authors Year Quotes

Waterman 1962 “[C]ircularity is inherent in the methodology of science since one must proceed from datato construct or model and thence back to new data or from model to data and back tomodel again. . . . In a well-developed science a multiplicity of such intersecting circularpathways form a coherent system of consistent relations” (p. 549).

van der Steen andBoontje

1973 A critique of the view that definitions of “homology” in terms of common ancestry representcircularity (homology is manifest as similarity due to common ancestry; common ancestryis inferred due to similarity).

Peters 1976 Because stressful habitats are identified by low species diversity, “‘the stability-timehypothesis’ [which specifies that nonstressful habitats give rise to higher species diversity]cannot be accepted as a scientific theory as it now stands” (p. 10).

Raven 1976 It is circular to infer homology between chromosomes from pairing experiments, andexplain pairing because of homology.

Tattersall andEldredge

1977 “[M]uch of the reasoning that goes into . . . [phylogeny] construction is circular: the manyelements involved feed back upon each other in an extremely intricate way” (p. 205).

Stevens 1980 It is circular to use distributions to inform the reconstruction of the phylogeneticrelationships between species and then make inferences regarding the evolution ofdistributions on the basis of the resulting phylogeny (see also Schaefer and Lauder 1986).

Tyler 1986 “[A]ccording to Popper, the difficulty the historical sciences face, whether the biologicalsciences or the social sciences, is that the systems they study can only be identifiedthrough change. And yet it is the changes themselves, rather than the systems, which arethe main object of interest. Hence there is an unavoidable circularity in the historicalsciences” (p. 727).

Landres et al. 1988 “[C]ircularity arises when using indicators to predict habitat conditions, because the initialchoice of the indicator depended on those habitat conditions” (p. 320).

Sage et al. 1993 “Another approach [to test the accuracy of methods of reconstructing the evolutionaryrelationships between species] has been to use computer simulations to generateevolutionary divergence in sets of genes. These simulated data can be used to evaluate theefficacy of various computer algorithms to reproduce the simulated genetic history.Unfortunately, the assumptions used to simulate the data can often be matched almostexactly by the assumptions of the algorithm used” (pp. 545–546).

Blackstone 1995 “Further, constructing a hypothetical ancestral form by assembling suites of shared primitivecharacters introduces an element of circularity and can have unintended results such aserecting paraphyletic taxa (e.g., see discussion of the ‘hypothetical ancestral mollusk’ inBrusca and Brusca 1990)” (p. 786).

Neal et al. 1998 “More sophisticated bees are said to be found on more complex flowers. The argumentsometimes becomes circular because the bees are often classified by the flowers they visit,rather than by experimental tests of learning ability” (p. 362).

Pennington et al. 2006 “[S]tudies [of the ages of clades] that rely too heavily on single geological calibrations . . .have been criticized . . . for their circularity” (p. 607).

Waters and Craw 2006 New Zealand inherited its flora and fauna when the great southern landmass Gondwanabroke up; this is inferred from New Zealand sharing lineages with other southernlandmasses. The similarity is explained by the breakup, the breakup inferred from thesharing of lineages.




faith in the underlying assumptions. For ex-ample, in the addition of 1�1, the conclu-sion of “2” does not increase confidence indefinitions of what “1” and “�” mean. In realscience, conclusions do increase confidencein assumptions (see Lipton 2008), as in thecloudy exoplanet example above, and in Fig-ures 3 and 4. In adaptationist studies, faith inassumptions is clearly a function of how wellthey explain the data. This quote from Dar-win could be a textbook illustration of loopyacceptance of assumptions given their abilityto explain the data: “It can hardly be sup-posed that a false theory would explain, in sosatisfactory a manner as does the theory ofnatural selection, the several large classes of

facts above specified” (Darwin [1876] 2009:471). Moreover, in real-life studies of adap-tation, instead of deductive certainties andineluctable laws there are at best statisticalprobabilities and likelihoods (e.g., Scriven1959; Ruse 1975; Rieppel 2003; Sober 2008).Explanations involving adaptation are thusloopy, not deductive (Figures 3 and 4).

Even though biology cannot be deductive,in the late 1960s biologists consolidated afirm tradition of describing their proceduresas “deductive,” ascribing this position tocertain philosophers, and have stuck withthe practice to the present day (Table 4).Some prominent examples include Ghis-elin (1966) and Medawar (1967), both of

TABLE 4How biologists classify evolutionary inference

Author YearEvolutionary biology (mostly the study

of adaptation) is. . . Cites

Ghiselin 1966 “obviously hypothetico-deductive” (p. 210) PopperMedawar 1967 (revision of

a 1963 original)hypothetico-deductive

Ghiselin 1969 “Biology . . . is a hypothetico-deductive,predictive, deterministic, and nomotheticscience” (p. xiii, preface to the 1984edition)

Popper, Mill, Whewell

Williams 1970 deductivePlatnick and

Gaffney1978 hypothetico-deductive Popper

Gould 1980 abductive, defined as “the creative grabbingand amalgamation of disparate conceptsinto bold ideas that could be formulatedfor testing” (p. 102), but see definition inTable 2

mentions Peirce

Jaksic 1981 hypothetico-deductive Popper (1959)Mayr 1982 hypothetico-deductive Ghiselin (1969) among

others; mentionsHempel and Popper

Fisher 1985 hypothetico-deductive, also “strong inference”(p. 131)

Platt (1964)

Calow 1987 hypothetico-deductiveBock 1988, 1994 deductive-nomological and “historical-

narrative” (p. 205)Murray 1992 hypothetico-deductive Newton, PopperThornhill 1996 hypothetico-deductive HempelAnelli 2006 hypothetico-deductiveEldredge 2006 hypothetico-deductiveAyala 2009 hypothetico-deductive Popper (1959);

Hempel (1965)McKnight 2009 “hypothesis-driven” with “inductive inquiry”

(p. 818)Ayala (2009)

Sulloway 2009 hypothetico-deductive Ghiselin (1969)




whom stated again and again that biologyis built on deduction. Their approach wasmirrored exactly by Mayr in his influentialwork The Growth of Biological Thought(1982). Gould (1980) even entitled a paperstating the aspiration of paleobiology to be-come a “nomothetic” discipine. Ghiselinsaid, echoing Hempel and Oppenheim, that“Biology . . . is a hypothetico-deductive, predic-tive, deterministic, and nomothetic science”(Ghiselin 1984:xiii). These exact argumentsare repeated by biologists to the present, per-haps most prominently by Ayala (2009). Thus,in exactly the period in which philosopherswere acutely seeing the limitations of deductiveaccounts of science and looking beyond themto “loopy” structures, the tradition of citing de-duction was hardening in biology and to alarge extent remains to this day. In this way, theposition in biology regarding deduction hasfollowed a very different trajectory from that inthe philosophy of science. However, it is clearthat explanations in evolutionary biology, likescience in general, are loopy and cannot bebuilt entirely on deduction. What, then, of the“hypothetico-deductive method”?

Myth 2: The hypothetico-deductivemethod in biology is deductive. Even thoughbiologists very often use the term “hypothetico-deductive method” (Table 4), their use doesnot really denote a deductive, nonloopymethod. As used by biologists, this methodconsists of three steps. The first is the gener-ation of a hypothesis. In the second, whichpotentially involves a deductive operation,predictions are generated from the hypoth-esis. Most often, these are of a nondeductive,probabilistic nature (e.g., “if X is true, then Yshould be common”). The third step in-volves empirically testing the predictions.This step is also not deductive (Table 2).Deciding whether observations conform ornot to predictions is always a probabilisticeffort. Generating and testing predictions inthe context of a hypothesis is a firm part ofeveryday science, but it results in loopy ex-planations, never deductively structuredones. In addition to being not deductive,science is also not Popperian.

Myth 3: Science is Popperian. Following asimilarly divergent trajectory as statementsabout deduction, there is a long tradition of

biologists saying that what they do is Poppe-rian (Table 4; Panchen 1992; Holcomb1996; Sterelny and Griffiths 1999; Haig andDurrant 2002; Ladyman 2002; Rieppel 2003;Morange 2009; Lancaster 2011). Biologistsgo to great lengths to show that Popper sup-ports one or another position (e.g., Wiley1975; Platnick and Gaffney 1978; Jaksic 1981;de Queiroz and Poe 2003; Ayala 2009). It isnot clear why Popper has been made a “pa-tron saint of science” (Ruse 1979:287), butRuse (1979, 2005) suggests that it is the sim-plicity of Popper’s scheme, that the schememakes scientists look daring and clever, andthat it fulfills desires to see progress in sci-ence. These considerations notwithstanding,the essence of Popper’s vision was a deduc-tive one, meaning that there is no way thatscience in general and evolutionary biologyparticular can be Popperian.

Instead of trying to justify their practicesby appealing to the supposed authority of agiven philosopher, biologists need to selectadaptationist explanations that they considersatisfactory and then dissect the structure ofthese explanations. If a given structure pro-duces an explanation that is considered sat-isfactory in the field, then this is the properstarting point for building stronger studies ofadaptation regardless of whether they con-form to a given philosopher’s stipulation ofwhat science should be (cf. Rieppel 2003).These explanatory structures turn out to beloopy, not Popperian. Biologists also do notuse falsification in the sense that Popper in-tended.

Myth 4: Biologists are Popperian falsifica-tionists. Popper’s falsificationist scheme wasa deductive one. Confirming or falsifying hy-potheses correspond to two different rules ofdeductive inference known as modus ponens(latin for “the way that affirms by affirming”)and modus tollens (“the way that denies bydenying”; see Sober 2008). Modus ponens isthe idea that if the condition “if P then Q” istrue, and P is the case, then Q must be thecase (see Table 2 for examples). Contrary tomodus ponens, modus tollens is the idea that ifthe condition “if P then Q” is true, and Q isnot the case, then P cannot be the case.Popper famously said that science proceedsthrough instances of modus tollens, a process




he called falsificationism. His idea was that itmakes more sense to proceed via modus tol-lens given that positive predictions or confir-mations of particular instances may be hardto come by but disconfirming hypothesesshould be more achievable. Either way, Pop-per envisioned the use of deduction, neverstatistics.

But most biologists thinking in falsifica-tionist terms use statistics. For example,biologists often identify a set of possible expla-nations and then use statistical tests to excludethem one by one to see which seems to bethe best one (e.g., Templeton 2009). Statis-tical operations such as hypothesis testingare probabilistic and by definition nonde-ductive. For Popper, falsification meant con-structing a deductive argument using modustollens. Because statistical null hypothesis test-ing is not deductive, there is a margin forerror. This margin for error is quantified viathe statistical significance of the test in reject-ing a given hypothesis. Statistical significancethresholds are arbitrary, with the choice ofP�0.05 as a threshold for significance beingone of convention but not salient from na-ture. The arbitrariness of these thresholds isone reason why philosopher Elliot Sobernotes that there is “no such thing as proba-bilistic modus tollens” (Sober 2008:192). Thismeans that the procedures that biologists useand call falsification (see, e.g., Forber 2011),although an important and accepted part ofscientific practice, are definitely not deduc-tive and definitely not Popperian. There isno reason to expect them to be, given theloopiness of real scientific explanation.

Myth 5: Modern evolutionary biologytakes a stand against “induction.” In the writ-ings of authors such as Ghiselin (1966,1969), Medawar (1967), Gould (1980), Mayr(1982), Ayala (2009), and the many who fol-low them, “induction” is presented as therandom collection of facts in the hope that auniversal generality will spring unaided fromthe data. That this is a caricature is revealedby the fact that even these authors admit thatno one really proceeds in this way (e.g., Ghis-elin 1969:4; Gould 1980:97). In reality, aninduction is simply an inference whose con-clusions are associated with some level oferror (Table 2; in deduction, if the premises

are true, the conclusion is true; Hull 1973;Ladyman 2002). Because all of their conclu-sions are associated with some degree ofuncertainty, every adaptationist study, likescience generally, involves some flavor ofinduction. Statements such as “the confi-dence interval of the metabolic rate-bodymass scaling slope includes 3/4,” “feathersoriginally evolved in the context of thermo-regulation,” or “the broad sense heritabilityof this trait is significantly different fromzero,” no matter how well supported clearlyall involve some degree of uncertainty. Allare expressions of induction. Aware thatmost of science proceeds through induction,philosophers and mathematicians have longbeen interested in coming up with degreesof certainty or likelihood functions to quan-tify the relation between evidence and hy-potheses (see Popper 1959; Hacking 1976;Sober 1988; de Queiroz and Poe 2003). Be-ing associated with uncertainty, loopy abduc-tive reasoning is usually regarded as a type ofinduction. “Induction” is not the ingenuous“idyll” (Gould 1980:97) that so many authorshave made it out to be. Instead, nondeduc-tive, loopy procedures are the heart and soulof science.

To summarize, these “myths” illustrate fiveimportant points that might make biologistsresist the idea that loopy explanations arethe true structure of evolutionary reasoning.1. The idea that science is deductive hasbeen refuted in philosophy of science sinceearly on but continues, incorrectly, to receivelip service in biology to this day. 2. Much ofthis lip service is in the form of referencesto the “hypothetico-deductive method.” Inbiology this simply means generating hy-potheses, deriving predictions from these hy-potheses, and testing them. It does notcorrespond to a purely deductive approach.3. Popper’s scheme was a deductive one; sci-ence in general and evolutionary biology inparticular are loopy and so cannot be Pop-perian. 4. Falsification in biology is not thePopperian version because what biologistscall “falsification” is statistical and not thedeductive procedure that Popper stipulated.5. Induction is not random data collection,but instead a type of inference in which theconclusion is associated with some uncer-




tainty, like practically all operations in realscience.

These misconceptions have had majorconsequences for evolutionary biology. Biol-ogists can often be found accusing eachother of “circularity” and disqualifying legit-imate “loopy” explanations (Table 3; seeHull 1967). A lack of appreciation of loopi-ness is the only condition that could permittaking the “tautology” of natural selectionseriously for so long in evolutionary biology(Peters 1976; Campbell and Robert 2005).The debate regarding adaptationism (e.g.,Gould and Lewontin 1979; Orzack and So-ber 1994) could only have continued so longbecause practitioners of comparative, popu-lational, and optimality methods work inisolation. This isolation, in turn, can onlyexist if biologists regard their individualsubdisciplinary approaches as autonomouslysufficient for studying adaptation. The per-ception of autonomous sufficiency is encour-aged by the misapplied “deduction” label: ifa given method is regarded as being “deduc-tive,” and deduction implies certainty of con-clusions, then there is no reason to combineresults across adaptationist subdisciplines. Itis time to for biologists to recognize andvalue the true “loopy” structure of scientificexplanations. It is time to recognize the com-plementary nature of the data the three ad-aptationist approaches generate, and forgetrue cooperation across comparative, popu-lational, and optimality perspectives.

How to Study Adaptation: The VitalComplementarity of Comparative,

Populational, and OptimalityApproaches

Although there is no conceptual reason todo so, populational and comparative biolo-gists tend to work in isolation and even crit-icize each other (Oakley 2009). Quantitativegeneticists can be heard accusing theircomparative biologist colleagues of study-ing fitness by “intuition and clairvoyance.”Comparative biologists concede that quan-titative genetic studies may be interestingfor their detail but note that they can onlyfocus on traits “so trivial” that they have notgone to fixation—surely the aim of evolu-tionary biology is to account for the great

patterns of trait variation across all of life,and not just inconsequential local varia-tion. As for studies of adaptation from theoptimality modeling perspective, they arecaricatured as the naive view of inexorableprogress to the best of all possible solu-tions. The result of this mutual aversion isthat, for the most part, proponents of com-parative, populational, and optimality ap-proaches work separately (Oakley 2009;Hadfield and Nakagawa 2010; cf. Parkerand Maynard-Smith 1990; Harvey and Pa-gel 1991; Falconer and Mackay 1996).

Amid this general separation, adaptation-ists do occasionally call for integration (e.g.,Fisher 1985; Endler 1986; Wake and Larson1987; Brandon 1990; Leroi et al. 1994; Sin-ervo and Basolo 1996; Baum and Donoghue2001; Durrant and Haig 2001; Matos et al.2004). For example, Larson and Losos(1996) proposed a procedure for testingadaptive hypotheses integrating various lay-ers of direct evidence. Their methodologyinvolves a series of steps sequentially exam-ining trait heritability, trait individuality/quasi-independence, restricted versus ampledevelopmental potential, and comparisonsof ancestral versus contemporary selective re-gimes, as well as documenting the relativeperformance of variants. Recognizing the“loopy” nature of adaptive explanationsmakes clear why schemes that integratemultiple lines of evidence generate satisfy-ing results (Figure 3).

Integration provides satisfying results be-cause, rather than one method being supe-rior to another, they in fact provide equallyimportant pieces of the adaptation puzzle,pieces that are moreover complementary(Table 5). They are complementary becauseimportant evidence not provided by onemethod is provided by the others (Table 1,Figure 3). Comparative methods are the onlyones that address the true products of realevolutionary diversification in the wild, onevolutionarily relevant time scales beyondthe ecological moment. However, the com-parative method leaves unexamined the de-tails of heritability, variation, and fitness,details that only populational methods canaddress. In addition, which variant hashigher fitness should be predictable given




TABLE 5Comparative, populational, and optimality approaches provide complementary sources of direct evidence

regarding hypotheses of adaptation, with none having a more privileged perspective than any other

Comparative studies Populational studies Optimality studies

Variantspresent/producible inancestral populations

no direct evidence no direct evidence no direct evidence

Variants were heritable/assimilable in ancestralpopulations

no direct evidence (butnote the loopyassumptionsregarding heritabilityin the case ofsynapomorphy)

no direct evidence no direct evidence

Variants differed in fitnessin ancestral populations


Variants differed inperformance inancestral populations


Intrapopulational variantscurrently produced/producible

no direct evidence study variation acrosspopulations within aspecies, additive geneticvariance

no direct evidence

Intrapopulational variantsare currently heritable

no direct evidence quantitative geneticmeasurements ofheritability

no direct evidence

Intrapopulational variantsvary in fitness

no direct evidence studies of survivorship, matingsuccess, fecundity, responseto selection

no direct evidence

Population-level processesplausibly produceinterspecific patterns

cross species organism-environment or trait-trait (allometry)relationships


Difference in performanceunderstandable basedon functionalgeneralizations,engineering principles

no direct evidence no direct evidence the optimality approach broadlyconstrued is required forthinking in terms of functionalgeneralizations

Variants fill morphospaceevenly or there areconstraints that may,even in the absence ofselection, lead topatterns of traitassociation

studies of how speciesfill morphospace,includingcomparativeembryology

studies of how variants naturaland induced, includingteratologies, fillmorphospace; artificialselection

predict the range of morphologiesthat should be observed; maypredict “holes” in morphospace

Trait (quasi-)independence (i.e., thetrait is a “part” that canbe subject to selection)

study how traits varyindependently acrossspecies

study how traits varyindependently in ontogeny,G matrix

explicit focus on functionallycoupled and competing traits

Current utility/function compare performanceof differentcharacter states

compare fitness of variants ina population

generate explicit expectationsregarding performancedifferences between variants

Arose for its currentfunction in its currentselective context

compare performanceof apomorphic statein current selectivecontext withplesiomorphic state





considerations of biomechanical perfor-mance or energetic efficiency. The onlymethod that examines this aspect directly isthe optimality approach. All of these aspects,including the patterns of trait distributionacross clades and continents, population-levelprocesses, and optimality considerations, areessential for a maximally supported adapta-tionist explanation (summarized in Table 5).Intensive studies from just one of thesepoints of view cannot provide all of the layersof direct evidence needed to diminish therelative importance of loops of “circular” rea-soning caused by acceptance of assumptionsabout the unobservable past (Figures 3 and4; Griffiths 1996; see also Figure 3 of Ruse1975). The shortcomings of each methodare almost perfectly filled by the strengths ofthe others (cf. Ghiselin 1969:21; Forber andGriffith 2011). As a result, the best-supportedadaptationist explanations have not just anabundance of information, but informationcarefully drawn from across the three adap-tationist approaches.

Maximally supported adaptationist expla-nations require evidence from comparative,populational, and optimality approaches.This requirement highlights from the outsetwhich adaptationist studies are likely to havefewer layers of direct evidence available.Studies of single species or unique structuresare important examples. Such traits cannotbe studied using comparative approaches,because the putatively adaptive states areunique (cf. Maddison and FitzJohn 2015).When the traits are fixed within populations,the typical tools of populational studies areunavailable. In humans, experimental meth-ods such as surgical intervention or selectivebreeding are unethical (Ruse 1979). As aresult, many aspects of humans continue tobe debated, such as the female orgasm, hu-man language, or rape (Travis 2003; Lloyd2005; Nielsen 2009; MacColl 2011). To theextent that less information is available, inmany cases it will continue to be hard todistinguish between different alternative ex-planations to decide which is the likeliest(Forber 2009). By asking what information isideally needed to generate a given explana-tion, a maximally robust explanation can beconstructed. Because its history is so vexed

and it is of such broad interest, we have fo-cused on adaptation here. However, a simi-lar search for the optimal combination oflayers of direct evidence can be used to guidethe effort to turn any evolutionary just-sostory into a well-supported explanation.

externalist versus internalist just-so stories

The traditional perspective of the ModernSynthesis is that variation is ample and ob-served morphologies represent the winnow-ing effects of selection (Amundson 1994;Jablonka and Lamb 2005). Because environ-mental factors “external” to the organism areviewed as determining which variants pros-per, this adaptation-driven view is oftenreferred to as externalism. In contrast, inter-nalism is the notion that interactions be-tween parts of developmental systems aresuch that developmental possibilities are se-verely limited and therefore the domain ofaction possible for natural selection is quiterestricted (Alberch 1989). Up to now, wehave focused on externalist just-so stories.

Internalist just-so stories are, however, justas easy to tell (Figure 2). Likewise, internalistexplanations can be constructed just as ro-bustly as externalist ones (Sober 1996). Infact, because examining developmental po-tential is essential for testing adaptationisthypotheses (Table 5), and because rulingout an externalist explanation is essential forshoring up an internalist one, the externalistand internalist approaches are both neces-sarily built in to a maximally robust loopyexplanation and really not separate perspec-tives at all (cf. West-Eberhard 2003; Schwenkand Wagner 2004).

Whether starting from an internalist oran externalist perspective, biologists test thedevelopmental accessibility of apparentlyempty patches of morphospace (e.g., aboveand below the line in Figure 2) via a numberof approaches (Olson 2012). These includedetailed studies of embryology and artificialselection or other types of manipulation(Sinervo and Basolo 1996; Bell 2008;Frankino et al. 2009; Vedel et al. 2010). Com-parative studies can pit the performance orfitness of species with different characterstates against each other in different selective




contexts (Larson and Losos 1996; Losos2011). Finding that developmental possibili-ties are sufficiently wide as to permit manyother possible trait combinations, and thatthe “empty space” morphologies are inferiorin performance, are findings that help shoreup the idea that a pattern is an adaptive one.Developmental potential is thus a centralpart of any adaptation based explanation (cf.Forber 2010).

These detailed depictions of possible de-velopmental variation, and study of the rela-tive performance of variation natural andconstructed, helps overcome the constraint-adaptation dichotomy. A move away from asimple constraint-adaptation dichotomy isone of the most promising aspects of loopyexplanations built on a correct selection ofsources of evidence. Moving away from thedichotomy is in part desirable because thevagueness of the term “constraint” makes itof little use in evolutionary thinking (see thecatalog of meanings in Olson 2012). Moreimportantly, the dichotomy is unacceptableas an explanatory formula because both se-lection and constraints, whatever the defini-tion used, are involved in the generation ofany given pattern in nature (Fusco 2001;West-Eberhard 2003; Schwenk and Wagner2004; Minelli 2009; Badyaev 2011). For ex-ample, although selection might cull fromthe possible, resulting in a narrow range ofcommonly observed morphologies, as in Fig-ure 2, factors such as minimum developmen-tally possible cell dimensions can limit thedomain of the possible (see, for example,Amundson 1994). As a result, it is meaning-less to ask whether the pattern in Figure 2 is“caused by adaptation or constraint” becauseevery pattern in the living world is the resultof both, however “constraint” is defined. Byshowing the way explicitly away from thisunsatisfactory dichotomy with clear ques-tions and a battery of empirical tools, studiesthat draw on multiple layers of direct evi-dence provide ever more satisfactory expla-nations of organismal form. One reason thatthese explanations are so satisfactory is that,when detailed information regarding thecauses of a given pattern are available, dich-tomous classing of “adaptation versus con-straint” is unnecessary and uninformative

(e.g., Malagon et al. 2014). Correctly repre-senting the way that robust explanations areconstructed also should improve public un-derstanding of science.

embracing loopiness: improvingscientific practice and the

communication of evolutionThat adaptationist explanations are neces-

sarily loopy helps reorient discussions of “cir-cularity” in evolutionary biology. It shedslight on the long tradition of accusing theentire study of adaptation as resting on tau-tology, i.e., circular reasoning. A popular ver-sion of this criticism goes that natural selectionis the survival of the fittest, and the fittest arethose that survive. The phenomenon to beexplained is part of the premises, therebyrendering the formulation circular (Peters1976; Bowler 1984). When the loopy struc-ture of evolutionary explanation is recog-nized, the debate over adaptationism as atautology appears to be predicated on theincorrect interpretation of “loops” as fatalflaws rather than natural and necessary. Anexplanation involving adaptation includesmany more layers than just “fit” and “sur-vival” (Hull 1969). As in any evolutionaryexplanation, loops of reasoning are present,e.g., the form-function fit is explained byselection and selection is identified as animportant process because of the globalform-function fit. The situation in Figure 3B,even though it includes “loops,” is hardly atautology. In fact, across evolutionary biologyat large, most of the accusations of “circular-ity” that biologists sling at each other almostalways simply refer to loops of reasoning inabductive/Bayesian reasoning (Table 3).

Rather than accusing one another ofcircularity whenever “loops” are detected, bi-ologists can more profitably discuss how rick-ety a reasoning loop is versus how well sup-ported it might be, and what additional datawould be desirable. In Table 3, the state-ments of Waterman (1962) and Tattersalland Eldredge (1977) are very close to theaccount we offer here. Most of the accusa-tions of circularity in Table 3 bear reevaluat-ing, to ask whether they might be reasonable“loopy” explanations that await testing viathe accumulation of more layers of direct




evidence, moving them from the situation inFigure 3A to that in Figure 3B. For example,Neal et al. (1998; Table 3) discussed waysthat floral morphology, scent, and rewardsfit with pollinators. They noted that “sophis-ticated” bees, understood as those withgreater manipulation skills, learning abil-ity, or sensory perception, are often said topollinate more complex and difficult-to-negotiate flowers. The problem is that somebees are identified as “sophisticated” pre-cisely because they pollinate complex flow-ers. This reasoning certainly sounds circular.However, additional layers of direct evidencecan be generated to see how well the expla-nations, assumptions and all, fit the observedpatterns (see also Nielsen 2009; O’Malleyand Koonin 2011). Neal et al. (1998) suggestthat independent tests of learning ability ofbees should be compared with floral com-plexity to see if there is a correspondence.Such studies could be conducted across spe-cies, within populations with floral variation,or even with artificially manipulated flowers.Any of these could potentially strengthen thesophisticated bee-complex flower hypothe-sis. This means that there is nothing flawedabout the structure of the hypothesis, justthat it is in a preliminary stage. Some exam-ples might be more perniciously circular.Using assumptions to simulate data for vali-dating a method built with those same as-sumptions might be such a case (Sage et al.1993 in Table 3). By recognizing the loopystructure of evolutionary explanation, biolo-gists can more effectively guide efforts to dis-tinguish flawed reasoning from legitimatelyloopy explanations.

The lack of clarity regarding the loopystructure of adaptationist explanations notonly affects science but also the way biolo-gists present evolutionary biology to the pub-lic at large. That scientists lack clarity regard-ing the structure of the explanations thatthey themselves strive to construct has exac-erbated public misunderstanding of how sci-ence works. Prominent authors have beenimplying for decades that evolutionary biol-ogy proceeds via deduction (e.g., Ghiselin1966; Ayala 2009), and therefore producesdeductive certainties. This entrenched tradi-tion has in many ways played into the hands

of critics, such as advocates of intelligent de-sign (Oakley 2009; Lancaster 2011). Publicdemands for certain “proof” and criticism of“circularity” in debates over evolution (andothers such as global climate change) arefueled by the notion that science producesdeductive certainty, almost always failing totake into account the loopy nature of thelegitimate explanations in these fields.

ConclusionStudies of adaptation necessarily require

the sort of loopy reasoning depicted in Fig-ure 3B (Holcomb 1996). Recognizing howadaptationist explanations are structured inactual practice helps give clarity to problemsthat have plagued biology, such as debatesover tautology/circularity, and resolve falseconflicts, such as the mutual scorn that oftencharacterizes the adherents of the compara-tive, population/quantitative genetics, andoptimality approaches (e.g., Calow 1987; Le-roi et al. 1994; cf. Zimmermann 1983:2 withHaberlandt 1914:12). Instead, as providersof complementary sources of direct evi-dence, no single approach has a monopolyon tests of adaptation (Table 5). An under-standing of the real, loopy structure of evo-lutionary explanation encourages biologiststo discuss truly substantial issues awaiting at-tention, such as how to identify the popula-tion of hypotheses from which to select the“best” explanation (Forber 2010), how scien-tists know the best explanation when they seeit, what observations most need explaining,or how best to weave disparate sources ofevidence into a single explanation. By ac-cepting that studies of adaptation requiremultiple types of direct evidence, evolution-ary biologists can improve current researchpractice by implementing a long-overdue in-tegration of comparative, populational, andoptimality approaches.

acknowledgments

Funded by Consejo Nacional de Ciencia y Tecnologıaproject 132404; support from the Sydney Centre for theFoundations of Science is gratefully acknowledged.Comments and assistance of Paul Griffiths, Karola Stotz,Maureen O’Malley, Francisco Vergara-Silva, DavidBraddon-Mitchell, Kristie Miller, Samuel Baron, Emanu-ele Serrelli, Steven Orzack, Arnon Levy, Pierrick Bour-




rat, Rod Taveira, Susanne Renner, Karl Rollings, WayneMaddison, Cesar Domınguez, Juan Fornori, Karina

Boege, Gerardo Arevalo, two anonymous reviewers, andthe editor are gratefully acknowledged.

REFERENCES

Alberch P. 1989. The logic of monsters: evidence forinternal constraint in development and evolution.Geobios 22 (Supplement 2):21–57.

Alcock J. 1998. Unpunctuated equilibrium in the Nat-ural History essays of Stephen Jay Gould. Evolutionand Human Behavior 19:321–336.

Amundson R. 1994. Two concepts of constraint: ad-aptationism and the challenge from developmen-tal biology. Philosophy of Science 61:556–578.

Anelli C. M. 2006. Darwininan theory in historicalcontext and its defense by B. D. Walsh: what is pastis prologue. American Entomologist 52:11–19.

Arnold S. J. 2005. The ultimate causes of phenotypicintegration: lost in translation. Evolution 59:2059–2061.

Ayala F. J. 2009. Darwin and the scientific method.Proceedings of the National Academy of Sciences of theUnited States of America 106:10033–10039.

Badyaev A. V. 2011. Origin of the fittest: link betweenemergent variation and evolutionary change as acritical question in evolutionary biology. Proceed-ings of the Royal Society B: Biological Sciences 278:1921–1929.

Baum D. A., Donoghue M. J. 2001. A likelihoodframework for the phylogenetic analysis of adap-tation. Pages 24–44 in Adaptationism and Optimal-ity, edited by S. H. Orzack and E. Sober. Cam-bridge (United Kingdom): Cambridge UniversityPress.

Bell G. 1989. A comparative method. American Natu-ralist 133:553–571.

Bell G. 2008. Selection: The Mechanism of Evolution.Second edition. Oxford (United Kingdom): Ox-ford University Press.

Bettiati D., Petit G., Anfodillo T. 2012. Testing theequi-resistance principle of the xylem transportsystem in a small ash tree: empirical support fromanatomical analyses. Tree Physiology 32:171–177.

Blackstone N. W. 1995. Perspective: a units-of-evolution perspective on the endosymbiont theoryof the origin of the mitochondrion. Evolution 49:785–796.

Blomberg S. P., Garland T. Jr., Ives A. R. 2003. Testingfor phylogenetic signal in comparative data: be-havioral traits are more labile. Evolution 57:717–745.

Bock W. J. 1988. The nature of explanations in mor-phology. American Zoologist 28:205–215.

Bock W. J. 1994. Concepts and methods in ecomor-phology. Journal of Biosciences 19:403–413.

Bowler P. J. 1984. Evolution: The History of an Idea.

Berkeley (California): University of CaliforniaPress.

Brandon R., Fleming L. 2014. Drift sometimes domi-nates selection, and vice versa: a reply to Clatter-buck, Sober and Lewontin. Biology and Philosophy29:577–585.

Brandon R. N. 1990. Adaptation and Environment.Princeton (New Jersey): Princeton UniversityPress.

Bromberger S. 1966. Why questions. Pages 86–111 inMind and Cosmos: Essays in Contemporary Science andPhilosophy, edited by R. G. Colodny. Pittsburgh(Pennsylvania): University of Pittsburgh Press.

Calow P. 1987. Towards a definition of functionalecology. Functional Ecology 1:57–61.

Campbell R., Robert J. S. 2005. The structure of evo-lution by natural selection. Biology and Philosophy20:673–696.

Christiansen P. 1999. Scaling of the limb long bonesto body mass in terrestrial mammals. Journal ofMorphology 239:167–190.

Darwin C. 1876. On the Origin of Species, sixth edition,edited by J. Endersby. [2009 reprint.] Cambridge(United Kingdom): Cambridge University Press.

de Querioz K., Poe S. 2003. Failed refutations: furthercomments on parsimony and likelihood methodsand their relationship to Popper’s degree of cor-roboration. Systematic Biology 52:352–367.

Dietrich M. R., Skipper R. A. Jr. 2007. Manipulatingunderdetermination in scientific controversy: thecase of the molecular clock. Perspectives on Science15:295–326.

Douven I. 2011. Abduction. The Stanford Encyclopedia ofPhilosophy, spring 2011 edition, edited by E. N. Zalta.http://plato.stanford.edu/archives/spr2011/entries/abduction/. (Accessed 15 August 2011.)

Durrant R., Haig B. D. 2001. How to pursue theadaptationist program in psychology. PhilosophicalPsychology 14:357–380.

Eldredge N. 2006. Confessions of a Darwinist. VirginiaQuarterly Review 82:32–53.

Endler J. A. 1986. Natural Selection in the Wild. Prince-ton (New Jersey): Princeton University Press.

Falconer D. S., Mackay T. F. C. 1996. Introduction toQuantitative Genetics. Fourth edition. Essex (UnitedKingdom): Pearson.

Fisher D. C. 1985. Evolutionary morphology: beyondthe analogous, the anecdotal, and the ad hoc.Paleobiology 11:120–138.

Forber P. 2009. Spandrels and a pervasive problem ofevidence. Biology and Philosophy 24:247–266.

Forber P. 2010. Confirmation and explaining how



http://plato.stanford.edu/archives/spr2011/entries/abduction/

http://plato.stanford.edu/archives/spr2011/entries/abduction/


possible. Studies in History and Philosophy of SciencePart C: Studies in History and Philosophy of Biologicaland Biomedical Sciences 41:32–40.

Forber P. 2011. Reconceiving eliminative inference.Philosophy of Science 78:185–208.

Forber P., Griffith E. 2011. Historical reconstruction:gaining epistemic access to the deep past. Philoso-phy and Theory in Biology 3:e203.

Franco-Vizcaıno E., Goldstein G., Ting I. P. 1990.Comparative gas exchange of leaves and bark inthree stem succulents of Baja California. AmericanJournal of Botany 77:1272–1278.

Frankino W. A., Emlen D. J., Shingleton A. W. 2009.Experimental approaches to studying the evolu-tion of animal form: the shape of things to come.Pages 419–478 in Experimental Evolution: Concepts,Methods, and Applications of Selection Experiments, ed-ited by T. R. Garland Jr. and M. R. Rose. Berkeley(California): University of California Press.

Frost-Arnold G. 2010. The no-miracles argument forrealism: inference to an unacceptable explana-tion. Philosophy of Science 77:35–58.

Fusco G. 2001. How many processes are responsiblefor phenotypic evolution? Evolution and Develop-ment 3:279–286.

Ghiselin M. T. 1966. On psychologism in the logic oftaxonomic controversies. Systematic Zoology 15:207–215.

Ghiselin M. T. 1969. The Triumph of the DarwinianMethod. Berkeley (California): University of Cali-fornia Press.

Ghiselin M. T. 1984. The Triumph of the DarwinianMethod. Chicago (Illinois): University of ChicagoPress.

Gould S. J. 1977. The return of hopeful monsters.Natural History 86:22–30.

Gould S. J. 1978. Sociobiology: the art of storytelling.New Scientist 16 November:530–533.

Gould S. J. 1980. The promise of paleobiology as anomothetic, evolutionary discipline. Paleobiology6:96–118.

Gould S. J. 1996. The tallest tale. Natural History 105:18–23 and 54–57.

Gould S. J. 1997. The exaptive excellence of spandrelsas a term and prototype. Proceedings of the NationalAcademy of Sciences of the United States of America94:10750–10755.

Gould S. J. 2002. The Structure of Evolutionary Theory.Cambridge (Massachusetts): Belknap Press ofHarvard University Press.

Gould S. J., Lewontin R. 1979. The spandrels of SanMarco and the Panglossian paradigm: a critique ofthe adaptionist programme. Proceedings of the RoyalSociety of London B: Biological Sciences 205:581–598.

Griffiths P. E. 1996. The historical turn in the study ofadaptation. British Journal for the Philosophy of Science47:511–532.

Haberlandt G. 1914. Physiological Plant Anatomy. Lon-don (United Kingom): Macmillan.

Hacking I. 1976. Logic of Statistical Inference. Cam-bridge (United Kingdom): Cambridge UniversityPress.

Hadfield J. D., Nakagawa S. 2010. General quantita-tive genetic methods for comparative biology: phy-logenies, taxonomies and multi-trait models forcontinuous and categorical characters. Journal ofEvolutionary Biology 23:494–508.

Haig B. D., Durrant R. 2002. Adaptationism and in-ference to the best explanation. Behavioral andBrain Sciences 25:520–521.

Hall W. 2002. Taking Darwin seriously: more thantelling just so stories. Addiction 97:472–473.

Harvey P. H., Pagel M. D. 1991. The ComparativeMethod in Evolutionary Biology. Oxford (UnitedKingdom): Oxford University Press.

Hempel C. G. 1965. Aspects of Scientific Explanation, andOther Essays in the Philosophy of Science. New York:Free Press.

Holcomb H. R. III. 1996. Just so stories and inferenceto the best explanation in evolutionary psychol-ogy. Minds and Machines 6:525–540.

Hull D. L. 1967. Certainty and circularity in evolution-ary taxonomy. Evolution 21:174–189.

Hull D. L. 1969. What philosophy of biology is not.Journal of the History of Biology 2:241–268.

Hull D. L. 1973. Darwin and His Critics: The Reception ofDarwin’s Theory of Evolution by the Scientific Commu-nity. Cambridge (Massachusetts): Harvard Univer-sity Press.

Hull D. L. 2001. Science and Selection: Essays on Biolog-ical Evolution and the Philosophy of Science. Cam-bridge (United Kingdom): Cambridge UniversityPress.

Jablonka E., Lamb M. J. 2005. Evolution in Four Dimen-sions: Genetic, Epigenetic, Behavioral, and SymbolicVariation in the History of Life. Cambridge (Massa-chusetts): MIT Press.

Jaksic F. M. 1981. Recognition of morphological ad-aptations in animals: the hypothetico-deductivemethod. BioScience 31:667–670.

Kilbourne B. M., Makovicky P. J. 2012. Postnatal longbone growth in terrestrial placental mammals: al-lometry, life history, and organismal traits. Journalof Morphology 273:1111–1126.

Kipling R. J. 1902. Just So Stories. Garden City (NewYork): Doubleday Page and Company.

Kreidberg L., Bean J. L., Desert J.-M., Benneke B.,Deming D., Stevenson K. B., Seager S., Berta-Thompson Z., Seifahrt A., Homeier D. 2014.Clouds in the atmosphere of the super-Earth exo-planet GJ 1214b. Nature 505:69–72.

Ladyman J. 2002. Understanding the Philosophy of Sci-ence. London (United Kingdom): Routledge.

Lancaster J. A. T. 2011. The semantic structure of




evolutionary biology as an argument against intel-ligent design. Zygon 46:26–46.

Lande R., Arnold S. J. 1983. The measurement ofselection on correlated characters. Evolution 37:1210–1226.

Landres P. B., Verner J., Thomas J. W. 1988. Ecolog-ical uses of vertebrate indicator species: a critique.Conservation Biology 2:316–328.

Larson A., Losos J. B. 1996. Phylogenetic systematicsof adaptation. Pages 187–219 in Adaptation, editedby M. R. Rose and G. V. Lauder. San Diego (Cal-ifornia): Academic Press.

Lennox J. G. 1991. Darwinian thought experiments: afunction for just-so stories. Pages 223–245 inThought Experiments in Science and Philosophy, editedby T. Horowitz and G. J. Massey. Savage (Mary-land): Rowman and Littlefield.

Leroi A. M., Rose M. R., Lauder G. V. 1994. What doesthe comparative method reveal about adaptation?American Naturalist 143:381–402.

Lev-Yadun S., Ne’eman G., Shanas U. 2009. A sheepin wolf’s clothing: do carrion and dung odours offlowers not only attract pollinators but also deterherbivores? BioEssays 31:84–88.

Lipton P. 2008. Inference to the best explanation.Pages 193–202 in The Routledge Companion to thePhilosophy of Science, edited by S. Psillos and M.Curd. London (United Kingdom): Routledge.

Lloyd E. A. 2005. The Case of the Female Orgasm: Bias inthe Science of Evolution. Cambridge (Massachu-setts): Harvard University Press.

Losos J. B. 2011. Convergence, adaptation, and con-straint. Evolution 65:1827–1840.

MacColl A. D. C. 2011. The ecological causes of evo-lution. Trends in Ecology and Evolution 26:514–522.

Maddison W. P., FitzJohn R. G. 2015. The unsolvedchallenge to phylogenetic correlation tests for cat-egorical characters. Systematic Biology 64:127–136.

Malagon J. N., Ahuja A., Sivapatham G., Hung J., LeeJ. Munoz-Gomez S. A., Atallah J., Singh R. S.,Larsen E. 2014. Evolution of Drosophila sex comblength illustrates the inextricable interplay be-tween selection and variation. Proceedings of theNational Academy of Sciences of the United States ofAmerica 111:E4103–E4109.

Martins E. P. 2000. Adaptation and the comparativemethod. Trends in Ecology and Evolution 15:296–299.

Matos M., Simoes P., Duarte A., Rego C., Avelar T.,Rose M. R. 2004. Convergence to a novel environ-ment: comparative method versus experimentalevolution. Evolution 58:1503–1510.

Mayr E. 1982. The Growth of Biological Thought: Diver-sity, Evolution, and Inheritance. Cambridge (Massa-chusetts): Belknap Press of Harvard UniversityPress.

McKnight S. L. 2009. Unconventional wisdom. Cell138:817–819.

Medawar P. B. 1967. The Art of the Soluble. London(United Kingdom): Methuen and Company.

Minelli A. 2009. Forms of Becoming: The EvolutionaryBiology of Development. Princeton (New Jersey):Princeton University Press.

Morange M. 2009. Articulating different modes ofexplanation: the present boundary in biologicalresearch. Pages 15–26 in Mapping the Future of Bi-ology: Evolving Concepts and Theories, edited by A.Barberousse, M. Morange, and T. Pradeu. Berlin(Germany): Springer Science.

Murray B. G. Jr. 1992. The evolution of clutch size: areply to Wootton, Young, and Winkler. Evolution46:1581–1584.

Nagel E. 1961. The Structure of Science: Problems in theLogic of Scientific Explanation. New York: Harcourt,Brace and World.

Neal P. R., Dafni A., Giurfa M. 1998. Floral symmetryand its role in plant-pollinator systems: terminol-ogy, distribution, and hypotheses. Annual Review ofEcology and Systematics 29:345–373.

Nielsen R. 2009. Adaptionism—30 years after Gouldand Lewontin. Evolution 63:2487–2490.

Oakley T. H. 2009. A critique of experimental phylo-genetics. Pages 659–669 in Experimental Evolution:Concepts, Methods, and Applications of Selection Exper-iments, edited by T. Garland Jr. and M. R. Rose.Berkeley (California): University of CaliforniaPress.

Okasha S. 2000. Van Fraassen’s critique of inferenceto the best explanation. Studies in History and Phi-losophy of Science 31:691–710.

Olson M. E. 2012. The developmental renaissance inadaptationism. Trends in Ecology and Evolution 27:278–287.

Olson M. E., Anfodillo T., Rosell J. A., Petit G., CrivellaroA., Isnard S., Leon-Gomez C., Alvarado-CardenasL. O., Castorena M. 2014. Universal hydraulics of theflowering plants: vessel diameter scales with stemlength across angiosperm lineages, habits and cli-mates. Ecology Letters 17:988–997.

O’Malley M. A., Koonin E. V. 2011. How stands theTree of Life a century and a half after The Origin?Biology Direct 6:32.

Orzack S. H., Sober E. 1994. Optimality models andthe test of adaptationism. American Naturalist 143:361–380.

Panchen A. L. 1992. Classification, Evolution, and theNature of Biology. Cambridge (United Kingdom):Cambridge University Press.

Parker G. A., Maynard Smith J. 1990. Optimality the-ory in evolutionary biology. Nature 348:27–33.

Peirce C. S. 1903. Harvard Lectures on Pragmatism. CP5.189 in Collected Papers of Charles Sanders Peirce,edited by C. Hartshorne, P. Weiss, and A. Burks.




Cambridge (Massachusetts): Harvard UniversityPress.

Pennington R. T., Richardson J. E., Lavin M. 2006.Insights into the historical construction of species-rich biomes from dated plant phylogenies, neutralecological theory and phylogenetic communitystructure. New Phytologist 172:605–616.

Peppe D. J., Royer D. L., Cariglino B., Oliver S. Y.,Newman S., Leight E., Enikolopov G., Fernandez-Burgos M., Herrera F., Adams J. M., Correa E.,Currano E. D., Erickson J. M., Hinojosa L. F.,Hoganson J. W., Iglesias A., Jaramillo C. A., John-son K. R., Jordan G. J., Kraft N. J. B., LovelockE. C., Lusk C. H., Niinemets U., Penuelas J., Rap-son G., Wing S. L., Wright I. J. 2011. Sensitivity ofleaf size and shape to climate: global patterns andpaleoclimatic applications. New Phytologist 190:724–739.

Peters R. H. 1976. Tautology in evolution and ecol-ogy. American Naturalist 110:1–12.

Petit G., Anfodillo T. 2009. Plant physiology in theoryand practice: an analysis of the WBE model forvascular plants. Journal of Theoretical Biology 259:1–4.

Petit G., Pfautsch S., Anfodillo T., Adams M. A. 2010.The challenge of tree height in Eucalyptus regnans:when xylem tapering overcomes hydraulic resis-tance. New Phytologist 187:1146–1153.

Pigliucci M., Schlichting C. D. 1997. On the limits ofquantitative genetics for the study of phenotypicevolution. Acta Biotheoretica 45:143–160.

Platnick N. I., Gaffney E. S. 1978. Evolutionary biol-ogy: a Popperian perspective. Systematic Zoology 27:138–141.

Platt J. R. 1964. Strong inference. Science 146:347–353.Popper K. R. 1959. The Logic of Scientific Discovery.

London (United Kingdom): Hutchinson.Potochnik A. 2009. Optimality modeling in a subop-

timal world. Biology and Philosophy 24:183–197.Raven P. H. 1976. Systematics and plant population

biology. Systematic Botany 1:284–316.Rieppel O. 2003. Popper and systematics. Systematic

Biology 52:259–271.Ruse M. 1975. Charles Darwin’s theory of evolution:

an analysis. Journal of the History of Biology 8:219–241.

Ruse M. 1979. Sociobiology: Sense or Nonsense? Secondedition. Dordrecht (The Netherlands): D. ReidelPublishing Company.

Ruse M. 2005. The Evolution-Creation Struggle. Cam-bridge (Massachusetts): Harvard University Press.

Rzedowski J., Kruse H. 1979. Algunas tendencias evo-lutiv as en Busera (Burseraceae). Taxon 28:103–116.

Sage R. D., Atchley W. R., Capanna E. 1993. Housemice as models in systematic biology. SystematicBiology 42:523–561.

Salmon W. C. 1989. Four decades of scientific expla-nation. Pages 3–219 in Scientific Explanation, editedby P. Kitcher and W. C. Salmon. Minneapolis(Minnesota): University of Minnesota Press.

Schaefer S. A., Lauder G. V. 1986. Historical transfor-mation of functional design: evolutionary mor-phology of feeding mechanisms in loricarioid cat-fishes. Systematic Zoology 35:489–508.

Schaffner K. 1993. Discovery and Explanation in Biologyand Medicine. Chicago (Illinois): University of Chi-cago Press.

Schwenk K., Wagner G. P. 2004. The relativism ofconstraints on phenotypic evolution. Pages 390–408 in Phenotypic Integration: Studying the Ecology andEvolution of Complex Phenotypes, edited by M. Pigli-ucci and K. Preston. Oxford (United Kingdom):Oxford University Press.

Scriven M. 1959. Explanation and prediction in evo-lutionary theory. Science 130:477–482.

Simmons R. E., Scheepers L. 1996. Winning by aneck: sexual selection in the evolution of giraffe.American Naturalist 148:771–786.

Simpson G. G. 1953. The Major Features of Evolution.New York: Columbia University Press.

Sinervo B., Basolo A. L. 1996. Testing adaptation us-ing phenotypic manipulations. Pages 149–185 inAdaptation, edited by M. R. Rose and G. V. Lauder.San Diego (California): Academic Press.

Sober E. 1988. Reconstructing the Past: Parsimony, Evo-lution, and Inference. Cambridge (Masachusetts):MIT Press.

Sober E. 1996. Six sayings about adaptationism. Phil-osophical Exchange 26:41–57.

Sober E. 2008. Evidence and Evolution: The Logic Behindthe Science. Cambridge (United Kingdom): Cam-bridge University Press.

Sosis R. 2009. The adaptationist-byproduct debate onthe evolution of religion: five misunderstandingsof the adaptationist program. Journal of Cognitionand Culture 9:315–332.

Sterenly K., Griffiths P. E. 1999. Sex and Death: AnIntroduction to Philosophy of Biology. Chicago (Illi-nois): University of Chicago Press.

Stevens G. C. 1987. Lianas as structural parasites: theBursera simaruba example. Ecology 68:77–81.

Stevens P. F. 1980. Evolutionary polarity of characterstates. Annual Review of Ecology and Systematics 11:333–358.

Strevens M. 2008. Depth: An Account of Scientific Expla-nation. Cambridge (Massachusetts): Harvard Uni-versity Press.

Sulloway F. J. 2009. Why Darwin rejected intelligentdesign. Journal of Biosciences 34:173–183.

Swartz S. M., Middleton K. M. 2008. Biomechanics ofthe bat limb skeleton: scaling, material propertiesand mechanics. Cells Tissues Organs 187:59–84.

Tattersall I., Eldredge N. 1977. Fact, theory, and fan-




tasy in human paleontology: controversy in thestudy of human evolution reflects inadequacies inthe formulation of hypotheses more than it doesthe supposed inadequacies of the fossil record.American Scientist 65:204–211.

Templeton A. R. 2009. Statistical hypothesis testing inintraspecific phylogeography: nested clade phylo-geographical analysis vs. approximate Bayesiancomputation. Molecular Ecology 18:319–331.

Thornhill R. 1996. The study of adpatation. Pages107–127 in Readings in Animal Cognition, edited byM. Bekoff and D. Jamieson. Cambridge (Massa-chusetts): MIT Press.

Travers S. E., Temeles E. J., Pan I. 2003. The relation-ship between nectar spur curvature in jewelweed(Impatiens capensis) and pollen removal by hum-mingbird pollinators. Canadian Journal of Botany81:164–170.

Travis C. B. 2003. Talking evolution and selling dif-ference. Pages 3–27 in Evolution, Gender, and Rape,edited by C. B. Travis. Cambridge (Massachu-setts): MIT Press.

Tyler G. W. 1986. Operational research and evolu-tion. Journal of the Operational Research Society 37:725–733.

Uebel T. 2007. Vienna Circle. The Stanford Encyclopediaof Philosophy, edited by E. N. Zalta. http://plato.stanford.edu/entries/vienna-circle/. (Ac-cessed 4 February 2009.)

van der Steen W. J., Boontje W. 1973. Phylogeneticversus phenetic taxonomy: a reappraisal. SystematicZoology 22:55–63.

Vedel V., Apostolou Z., Arthur W., Akam M., Brena C.2010. An early temperature-sensitive period for theplasticity of segment number in the centipede Strigamiamaritima. Evolution and Development 12:347–352.

Vincent T. L., Brown J. S. 2005. Evolutionary GameTheory, Natural Selection, and Darwinian Dynamics.Cambridge (United Kingdom): Cambridge Uni-versity Press.

Wake D. B., Larson A. 1987. Multidimensional analy-sis of an evolving lineage. Science 238:42–48.

Waterman T. H. 1962. Revolution for biology. Ameri-can Scientist 50:548–569.

Waters J. M., Craw D. 2006. Goodbye Gondwana? NewZealand biogeography, geology, and the problemof circularity. Systematic Biology 55:351–356.

West G. B., Brown J. H., Enquist B. J. 1997. A generalmodel for the origin of allometric scaling laws inbiology. Science 276:122–126.

West G. B., Brown J. H., Enquist B. J. 1999. A generalmodel for the structure and allometry of plantvascular systems. Nature 400:664–667.

West G. B., Woodruff W. H., Brown J. H. 2002. Allo-metric scaling of metabolic rate from moleculesand mitochondria to cells and mammals. Proceed-ings of the National Academy of Sciences of the UnitedStates of America 99:2473–2478.

West-Eberhard M. J. 2003. Developmental Plasticity andEvolution. Oxford (United Kingdom): Oxford Uni-versity Press.

Wiley E. O. 1975. Karl R. Popper, systematics, and classifi-cation: a reply to Walter Bock and other evolutionarytaxonomists. Systematic Zoology 24:233–243.

Williams M. B. 1970. Deducing the consequences ofevolution: a mathematical model. Journal of Theo-retical Biology 29:343–385.

Zimmermann M. H. 1983. Xylem Structure and the As-cent of Sap. Berlin (Germany): Springer-Verlag.

Handling Editor: Daniel E. Dykhuizen



http://plato.stanford.edu/entries/vienna-circle/

http://plato.stanford.edu/entries/vienna-circle/


How to Study Adaptation (and Why To Do It That Way)€¦ · HOW TO STUDY ADAPTATION (AND WHY TO DO...

Documents

Transcript of How to Study Adaptation (and Why To Do It That Way)€¦ · HOW TO STUDY ADAPTATION (AND WHY TO DO...