Should Methodological Adaptationism be Replaced?

Abstract: Elizabeth Lloyd has objected to Methodological Adaptationism (MA) on the grounds that it misleads Science, so she is trying to motivate the idea that it can be replaced with an Evolutionary Factors (EF) framework. Her main argument is that the latter framework is more fruitful for scientific discoveries than MA. However, fruitfulness is a vague term, and she uses it as such—therefore, using a recent explication of fruitfulness, I shall assess her claims. Highlighting junk DNA and evolutionary medicine, I will show how MA is more fruitful than the EF approach because we have epistemic and non-epistemic reasons to privilege adaptive answers. In essence, privileging these adaptive answers circumvents most of Lloyd’s criticism, so I conclude that MA should not be replaced.

Introduction

In the 40 years since Gould and Lewontin’s landmark paper “The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme”, the biological community is thought to have, by and large, adapted to their criticism. Gould and Lewontin had served several charges against the adaptationist camp, namely ad hoc rationalizations, questionable standards of evidence, and paying lip service to alternatives. Some biologists downplayed the issues raised, while others were more conciliatory and tried to make adaptationist reasoning more robust. Following their paper, there has been a great deal of

development in the field—that is, from defining different types of adaptationism to more scrutiny of just so stories.

While the debate has centred around evidentiary standards, Elizabeth Lloyd has recently raised investigative issues around MA. Essentially, she argues that lessons from Gould and Lewontin are harder to implement because adaptive thinking begins with initial research questions that lead to “closed-mindedness, and the inability to see alternatives, or evaluate evidence”1 and, therefore, to scientific failure. Lloyd has expanded upon the objections of the original Spandrels paper, and she has also raised some significant novel challenges to MA. Her argument is that there are reasons to replace MA with an alternative: the EF approach. I shall argue against Lloyd’s claim that MA should be replaced by the EF approach.

Lloyd’s Argument Against Methodological Adaptationism

After Gould and Lewinton’s paper, three distinct varieties of adaptationism have been put forward. MA is the most popular one. Defining MA, Peter Godfrey-Smith says, “The best way for scientists to approach biological systems is to look for features of adaptation and good design. Adaptation is a good ‘organizing concept’ for evolutionary research.” 2 Note that this claim is not the same as the idea that the biological world is saturated with adaptations, or that selection is a prevalent and powerful force with little constraints (this is known as empirical adaptationism). Under MA, it may even be assumed that very few adaptations actually exist. In essence, there are two types of MA: strong and weak. The latter sees adaptationism as one strategy to understand biological traits. The Strong form is the view that assumes adaptationism as both a starting point and the only route to discover the actual status

1 Lloyd 2015, p343

2 Godfrey-Smith 2001, p337

of traits, regardless of whether or not they turn out to be adaptive. Although Lloyd does not make use of the distinction between strong and weak forms of adaptationism, it is clear that her claims are against the stronger form. Lloyd argues that asking different questions can expand or constrain classes of answers, which she refers to as the “logic of research questions.” Her depiction of MA inspired the following research question: What is the function of this trait?

The answers allotted by the aforesaid question are all in the form of the following:

The function of the trait is X. The function of the trait is Y. The function of the trait is Z. Etc.

Lloyd acknowledges that proponents of MA accept that if a functional answer fails then a non-adaptive explanation can be explored. However in practice she thinks non-adaptive answers never seen the light of day. She contrasts MA with the EF approach. Although EF is used by some biologists Llyod believes MA is still subscribed to by the majority. Llyod explains EF research question as:

“What Evolutionary Factors account for the form and distribution of this trait?” Or rather, “Does this trait have a function?”

This question has a series of possible distinct answers (that can be considered in any order, but, usually, in practice, the adaptive answers go first):

Possible Answers:

A: This trait occurs in the population because it has the function F (i.e., the trait is an adaptation).

A: This trait occurs in the population because it has the function G (i.e., the trait is an adaptation).

A: This trait occurs widely in this population because it is genetically linked to a trait that is highly adaptive in this species (genetic linkage or correlation).

A: This trait has its current form largely because of an ancestral pattern (phyletic inertia). A: This trait has its current form and distribution because of pleiotropy with a trait that was under natural selection (pleiotropy or byproduct).

A: This trait has its current form and distribution because it is a byproduct (or bonus) of a trait that is strongly selected in the opposite sex of this species (byproduct or

bonus of an adaptation).

A: This trait has its current form and distribution because of some combination of the above factors.

Etc.3

It is clear that the EF approach permits more types of answers. Thus, Lloyd believes this approach should replace MA as the default method because it is more fruitful. She sees EF as the best of both worlds. Functional answers are available as a first go-to algorithm and, afterward, if these do not work, non-adaptive answers are tested. Lloyd makes her case against MA by pointing out problems that she thinks mislead science and, therefore, make it less fruitful as a research programme. To that belief, Lloyd raises five specific problems with the MA programme; (1) The disappearance of the “onerous burden of proof”; (2) mistaking

3 Lloyd 2015, p346

non-adaptive alternatives as mutually exclusive instead of as complementary or cooperative accounts; (3) the lack of a “stopping rule” for functional answers; (4) loss of ability to evaluate and weigh evidence for alternative causal hypotheses; and (5) treating non-adaptive answers as a “Null” Hypotheses.4 These problems, she contends, make MA less fruitful as a research programme.

Disappearance of the Onerous Burden of Proof

Lloyd raises the issue pertaining to the disappearance of the “Onerous Burden of Proof”. She defines evolutionary adaptations as current traits that developed in the past history of the organism, which is due to selective pressures being applied to an assortment of phenotypes. Lloyd claims adaptationists have set a high standard for themselves in order to detect traits that are adaptations. That is, they have never had a problem with embracing the burden of proof and, clearly, they have specified what it entails. Lloyd cites Paul Andrews and George Williams as acknowledging that the demonstration of adaptation “carries an onerous burden of proof.” Williams suggested some qualities of a trait in order to detect design—such as precision, efficiency, and economy5—however, what he actually desired was the development of “sets of objective criteria [of special design].”6 Lloyd points out that Williams, in practice, accepted something less than the rigorous objective criteria he proposed to detect adaptations. However, Williams is not alone in this because, in general, adaptationists only pay lip service to the burden of proof with which they are supposed to work. Nevertheless, for Lloyd, this is not even the worse part because their research question—“What is the function of this trait?”—has hidden the burden of proof. That is, they have replaced the assumption of

4 Lloyd 2015, p347–356

5 Williams 1966, p503

6 Williams 1966, p9

adaptation—which was only supposed to be a research heuristic—with a claim about its actual existence. Another issue she raises concerns the definition of adaptation. Skipping Elliot Sober’s widespread definition for detecting adaptations, the adaptationists are using a watered down definition. Sober’s definition links the function as well as the selective process that lead to it in a clear way, explaining, “A is an adaptation for task T in population P if and only if A became prevalent in P because there was selection for A, where the selective advantage of A was due to the fact that A helped perform task T.” 7 Knowing the difficulties this would cause them, adaptationists are motivated to decoupling this history-laden definition. Behavioural ecologists Reeve and Sherman argue in favor of such a revision. They think the problem inherent within Sober’s definition is devastating for methodological adaptationists, such as themselves. Reeve and Sherman believe Sober’s definition “may be sufficient to implicate a trait as an adaptation,” but “such criteria are not necessary to recognize adaptations.”8 If Sober’s definition is used, the majority of traits that are currently considered to be adaptations will be downgraded to non-adaptations. Stripping traits of their adaptive status would be a result of Sober’s definition requiring us to know that the trait spread through the population via natural selection working on varying phenotypes. Thus, Reeve and Sherman argue for a separation of the current function and the historical process that lead to said function. For the methodological adaptationist camp, work would be much easier if Reeve and Sherman’s definition were to be adopted. Their definition is that, “An adaptation is a phenotypic variant that results in the highest fitness among a specified set of variants in a given environment.”9 Of course, one should note that this definition is elastic enough to count the majority of traits as adaptations. That is, inference from a trait’s current function to its past selection is sufficient for a trait to be classified as an adaptation. Under

7 Reeve & Sherman, 1993, p7

8 Ibid

9 Reeve & Sherman 1993, p1

Sober’s definition, such an inference would not be allowed. Even before Reeve and Sherman’s definition was penned, methodological were applying it in practice. Llyod’s problem is, more or less, that adaptationists are guilty of shrinking their own “onerous burden of proof.” All told, this perk cannot be enjoyed because she desires Sober’s definition to be employed.

Mistaking Complementary Accounts as Mutually Exclusive

Methodological adaptationists see non-adaptive answers as mutually exclusive to adaptative ones. According to Lloyd, this approach puts adaptationists at a disadvantage. Within the EF framework, hypotheses of spandrels, genetic and developmental constraints, drift, and exaptations can be welcomed as complementary accounts. These limit the selective options and, also, they are needed in order to explain how an optimum adaptation is achieved.

Conversely, adaptationists have set up a false dichotomy—that is, either a trait is an adaptation or it is not. Adaptive hypothesis for a trait has to be considered first, and then, if repeated attempts fail, a trait may be considered a non-adaptation. For a non-adaptive hypothesis to be considered, it is necessary for an adaptive hypothesis to be generated and exhaustively invalidated. Adaptationists have set up adaptive and non-adaptive answers as exhaustive disjuncts. Essentially, increasing confidence in adaptive answers decreases confidence in non-adaptive ones. According to Lloyd, such a belief regarding confidence is a logical claim, not an empirical one. As such, it evades the empirical considerations. However, confidence should rely upon empirical data, so this will help us to detect the true status of traits. Therefore, Lloyd thinks adaptationists mislead science by using this staunch dualist approach.

The Lack of a Stopping Rule

Regarding the stopping rule problem, Lloyd reignites the complaints of Gould and Lewontin. They raised two interconnected issues pertaining to the continuous generation of adaptive hypotheses and the infamous just so storytelling. While adaptationists have claimed to have adapted their approach to these problems, Lloyd believes lessons have not been learned because the blocking of non-adaptive answers as well as inadequate evidentiary standards are still in practice. Adaptationists have proposed that, by falsifying adaptive hypotheses, they are increasing confidence in non-adaptive answers whilst, eventually, switching over. Lloyd argues that, in practice, this does not work because adaptative accounts are endlessly generated. The reasoning for this is that adaptationists still fail to distinguish between speculative hypotheticals—that are consistent with natural selection—and evidence-based hypotheses. Regardless, Lloyd does not think much progress has actually been achieved on this front—that is, except for granting lip service to the existence of these problems.

To further develop the stopping rule problem, Lloyd provides the adaption of the glass tree frog as an example. Because the environment within which this species resides has green leaves, it was thought that glass tree frogs—akin to other tree frogs—evolved to be green. However, an interesting difference is the fact that, while regular tree frogs acquired their colour through the absorption of sections of the visible light spectrum, glass tree frogs acquired their green by refracting light. That said, it was not clear if this was a relevant difference that had implications for their respective adaptations. Nevertheless, both types of frog were investigated under infrared light. It was later discovered that the leaves the regular tree frogs rest upon reflect infrared light, but the frogs themselves absorb it. Conversely the glass tree frogs like the leafs that they sit on reflect infrared and thereby become completely

disguised. Additionally, it was found that pit vipers were sharing the environment with glass tree frogs. This is significant because those particular snakes hunt using infrared sensitivity, so a refurbished adaptive account was formed. Clearly, this is an additional problem for the methodological adaptationist—that is, because one good adaptive account is not enough to warrant pausing the search for newer adaptive accounts. The stoppage problem has been reinforced by Lloyd in a novel way.

Loss of Ability to Evaluate and Weigh Evidence

In the rare cases when a methodological adaptationist considers non-adaptive answers, they still face an uphill struggle. The ability to weigh evidence from non-adaptive and adaptive hypotheses is sometimes lost. According to Lloyd, MA uses a lens that distorts what is actually going on. To explain this, she highlights her flagship case: the female organism.10 Over the years, Lloyd has scrupulously eliminated every adaptive account for this trait. She argues it is a byproduct (her preferred term is “fantastic bonus”) that is due to the stabilizing selection of the male organism. In 1979, Donald Symons was the first to present evidence of the female organism being a byproduct. Symons view was not singularly based upon the rejection of adaptive accounts, for it also had independent evidence. To Lloyd’s disappointment, Symons’s view has either been dismissed or outright ignored as a baseless a priori assertion. Reasoning for this is that the methodological adaptationist research question blocks any non-functional answer from appearing. Lloyd cities Andrews’s reaction as a standard misplaced methodological adaptationist attitude. Andrews contended that the argument for a byproduct account was invalid because it lacked a strategy to reject the

10 Lloyd 2006

adaptive hypothesis.11 Regardless, the positive evidence for the byproduct view—which included developmental symmetries—were invisible to adaptationists. This is because their radars do not detect drift, persistent lack of genetic variation, and rigid genetic correlations as alternative explanations. If anything, these viable evolutionary causes are substituted as tools that can aid in understanding the optimality of adaptations. In cases where the adaptationist does acknowledge the presence of these factors, they make unwarranted assumptions regarding them. Lloyd highlights how adaptationists deny the possibility that genetic correlations cannot be broken, and, also, that such correlations are assumed too insufficient to be a complete explanation for the distribution of traits. Instead, they attempt to explore the prospect that the genetically correlated traits are selectively maintained over potential alternatives. Thus, because of their initial research question, adaptationist eyes are colour blind to causally efficacious factors that are non-adaptive, which is why Lloyd thinks they can neither properly nor fairly weigh evidence.

Treating Alternatives as Nulls

Questions in the form of “What is the function of a trait x?” only limits answers to functional ones. In effect, this causes non-adaptive answers to be classified as nulls. Because nulls are negative results contrasting a positive alternative hypothesis, independent evidence cannot be accumulated for them. Lloyd stresses this would be a mistake because drift, phyletic inertia, and genetic constraints are alternative causal accounts for the form and distribution of traits. Therefore, they have their own independent evidence. While some adaptive researchers explicitly name non-adaptive hypotheses as nulls, other practitioners simply treat them as such by adhering to the function style questioning. In statistical analysis, while nulls are used

11 Andrews 2002a, p499

in a formal way, within adaptive camps, it is intended to be in an informal way. Relegating non-adaptive answers to nulls means only evidence for correlation with fitness counts as actual evidence for a positive hypothesis. While the null is simply a non-correlation with fitness, it is akin to a barren result as well as a scientific surrender.12 Thus, independent evidence for non-adaptive answers is concealed and only becomes apparent within the EF approach. This framework initiates its enquiry with a particular question (“Does this trait have a function?”), but it can descend into another question (“What is the function of this trait?”), which, at that point, is the same function question that Lloyd claims is misleading science. Realizing this problem, she presents a simple solution. Both questions may look syntactically identical, but this does not necessarily mean that they are, logically, asking identical questions. Consider how the same question can be formed by adopting syntactically distinct sentences. Structural wording is a trifle bit logically arbitrary. In order to know what the research question logically entails, we must see what answers it permits. Although both competing approaches may be asking questions about function, the actual essence is unveiled when we catalog the live options that are actually allowed under each framework. EF allows non-adaptive answers while methodological adaptationists only pay lip service to such answers. But it is not enough to examine the verbal structure of questions, which means we need to analyze the logic of the research question itself. The EF framework does not lead to non-adaptive answers being labelled as null hypotheses—rather, it permits them to garner evidence that is independent and visible. However, the null hypothesis status consigned to non-adaptive answers is the result of the functional questioning of the methodological adaptationist. Contrary to the claims of adaptationists, their question is not a harmless preference to first explore functional questions. Therefore, researchers are soberly mislead when they are exploring the actual status of traits.

12Alcock 1987

There is much debate regarding how to appraise competing scientific theories. However, Lloyd is clear that an EF approach is more fruitful and, hence, should be our default paradigm in evolutionary biology. To date, Lloyd’s claims have gone unchallenged, so, below, what I intend to do is give a clear route that can undermine her central argument.

Lloyd’s argument can be summarized as:

1. MA in practice does not allow non-adaptive answers.
2. EF allows both adaptationist answers and non-adaptive answers.
3. Between competing scientific frameworks, we should use fruitfulness as a criteria when deciding between them.
4. The EF approach is a more fruitful approach than MA because it allows more answers.

Conclusion: The EF framework should replace MA.

Defending Methodological Adaptationism

Lloyd has pointed out some problems with MA and, as a result, she wants it to be rejected as the default framework. Aside from her stance on using Sober’s onerous definition of adaptations, I largely agree with Lloyd in regard to the problems she raises. Indeed, under MA, adaptive answers are given an immense privilege over non-adaptive ones. However, I do not agree that this gives us justification to replace MA with EF. The reason for this is that we

have epistemic and non-epistemic reasons to privilege MA. Lloyd wants us to start research on biological features by assuming that the traits we see are selectively neutral—that is, we do not know their actual status. Conversely, adaptationists ask us to reject all adaptationist explanations before we can even consider a non-selective account. Lloyd assumes the explicit privileging of MA is unwarranted, but I will argue against this. Here is my argument:

1. It is incorrect to assume the EF framework covers all the benefits of MA.
2. We have epistemic reasons for knowing adaptive answers are more fruitful than non-adaptive ones.
3. We have non-epistemic reasons to choose adaptive answers over non-adaptive ones.

Therefore, MA should not be replaced

What counts as an Adaptation?

Before I lay out the case for MA being more epistemically fruitful than EF, I am going to explain why I disagree with Lloyd with her regarding her use of Sober’s definition. It is important to first clarify this point because, if we accept the Sober definition, much of what we consider adaptations would be relegated of their status, hence the fruitfulness of MA would hit a stumbling block. Lloyd points out that the “onerous burden of proof” that was put forth by prominent adaptationists is not put into practice. She is correct to note that Sober’s conservative definition to detect adaptationists is ignored while Reeve and Sherman’s more liberal one is employed. However, I do not see this as a problem. It could be argued that the adaptationists who proposed the onerous burden of proof were simply wrong to propose it. It is not abnormal for scientists to sometimes put forward criteria in order to confirm their

conclusions, which are later revised according to practical considerations. For example, the imperfection of the fossil record would change the way we set the criteria to confirm gradualism. A hypothetically perfect and preserved paleontological record would make the criteria stricter. In fact, this is one of the early problems that Darwin noted in On The Origin of Species. The so-called Cambrian explosion—the appearance of fossils in the Cambrian period that seemed to suddenly appear (without intermediate transitional fossils)—challenged what one would expect if evolution gradually took place. Of course, due to the poor paleontological record, the phenomenon was explained by Darwin himself (as well as others after him) in such a fashion: “Why then is not every geological formation and every stratum full of such intermediate links? Geology assuredly does not reveal any such finely-graduated organic chain; and this, perhaps, is the most obvious and serious objection which can be urged against the theory. The explanation lies, as I believe, in the extreme imperfection of the geological record.”13 It would be unfair to criticise gradualism by the fact that most species in that period never fossilized. Moreover, it is not reasonable to expect a linear fossil record to show each gradual step.

In our case, there is no reason to suppose that the “onerous burden of proof” is necessary at all. Lloyd did not make an effort to explain why Sober’s definition is the correct one. Perhaps she assumed that, because Sober’s definition is widespread, it should be taken for granted.

Maybe this seemed to be a reasonable approach—from her perspective, at least—but it is still questionable. Lloyd may argue that Reeve and Sherman need to provide sound reasoning in order to accept their definition—that is, apart from speaking about the difficulties of using Sober’s definition and, also, the ease of using their own. Reeve and Sherman did not actually justify why their definition is correct, either. However, I think a case can be made for their

13 Darwin 1860, p280

definition, which would be, more or less, a case against Sober’s definition. Starting with an example that Lloyd herself gives:

In the ancestral population of anteaters, for instance, which resembled armadillos, tongue length was likely highly variable, with high fitness values accruing to those anteater-ancestors that might be able to reach into ant nests with their long tongues and eat the most ants, and were thus most able to pursue their food resources. These longer-tongued anteater-ancestors—eventually with their 25-inch-long tongues—would represent the best fit—or closest-to-best fit—to their environment.

The anteater example thus presents a good instance of a natural selection explanation that reinforces or produces an adaptation.14

Applying Sober’s definition, we would require more information before we could class the anteater tongue as an adaptation. Essentially, his definition requires that we know that a trait T was selected in population P. Although Lloyd writes that the tongue length was “likely highly variable,” that is not quite enough. How does she know that was the case? It simply cannot be taken as self-evident. A historical record of anteaters with various tongue lengths—with only the longer ones surviving—would be necessary in order to, according to Sober’s definition, detect an adaptation. Surveying the paleontological record of the suborder Vermilingua (which is formed of the four extant mammals commonly referred to as anteaters) is a problematic task because anteaters diverged from sloths in the Early Eocene. Fossils of anteaters in this period are sparse.15 The earliest fossils of the extinct and extant genera of anteaters is very difficult to reconstruct. Some fossils have been discovered to have long muzzles (and therefore long tongues), but we do not have a clear range of sizes. Therefore,

14 Lloyd 2015, p347–356

15 McDonald 2008, p64–72

under the very definition to which Lloyd is subscribing, a trait that she accepts is an adaptation cannot be classified as one. Conversely, Reeve and Sherman’s definition would allow us to count the anteater tongue as an adaptation. To do so, all that would be needed would be a simple inference from current function to a selective causation. The incompleteness and fragmentary nature of the anteater fossil record is no different from that of other animals, so detecting adaptations using Sober’s definition would be a general problem. Thus, we have good reason to trust the Reeve and Sherman definition. Obviously, Lloyd may object that Reeve and Sherman’s definition casts the net too far. For instance, she could point out that a fitness enhancing trait does not necessarily have to be a product of selective pressures. Such traits have been labelled as exaptations.16 Exaptations have been defined in two ways: (1) “[A] feature, now useful to an organism, that did not arise as an adaptation for its present role, but was subsequently co-opted for its current function”17 and

(2) “features that now enhance fitness, but were not built by natural selection for their current role.”18 Her point would be valid as there is a clear risk of non-adaptive traits being labelled as adaptive ones. While it is true that the definition used by methodological adaptationists involves such a risk, there is, likewise, a similar risk involved with Sober’s definition. In order to make it easier to decide between the two definitions, we can introduce Lloyd’s own belief about adaptation in nature. In relation to the adaptation of the anteater muzzle, she notes, “I take it as given that our living world is filled with examples of such adaptations.”19 If she maintains this assumption about the biosphere, one would expect her to advocate for Reeve and Sherman’s definition. Indeed, it is somewhat of an endeavour to comprehend why Lloyd would adopt Sober’s definition when doing so would cause the en masse disappearance of traits that are currently labeled as adaptations, which, of course, would not

16 Gould 1982

17 Gould 1991, p43

18 Ibid

19 Lloyd, 2015, p344

fit her assumption of abundance of adaptations. That said, I would sympathise if we had a much clearer fossil record maybe Sober’s definition would not be too demanding. However, due to the poor paleontological record, it would be impractical to deploy such a definition—in fact, it would seem uncontroversial to bypass Sober’s definition.

Although Reeve and Sherman penned their definition two decades ago, its actual use has been widespread, going all the way back to the mid 19th century. Essentially, Darwin indulged in the type of adaptive thinking that Lloyd is complaining about. Indeed, he was using the loose definition that Reeve and Sherman propose for making predictions. In 1862, Darwin came across a rare orchid flower from Madagascar called Angraecum sesquipedale. It had a bizarrely deep nectar reservoir. Darwin candidly made two bold predictions. First, he claimed that the unusual plant had developed an adaptation and, second, that there was a

yet-to-be-discovered long-tongued moth with which the flower had a co-evolutionary relationship.20 This supposed moth had never been seen, so Darwin’s account was criticized by entomologists. However, he was still adamant that the moth existed and had a tongue of 10–11 inches. After Darwin’s death in 1903, 20 years later, the illusive moth was finally found. It is important to note how the structure of the flower was enough for him to label both the orchid and the moth as having adapted to each other, for there was no history of the moth or the orchid flower that he had at hand in order to make this claim. Under Sober’s definition, Darwin would not have been able to call this an adaptation. Biologists since Darwin have used a loose definition of adaptation (such as that of Reeve and Sherman, which is based on current utility without a history of the trait within the population) to make novel discoveries, so to give that up would not make sense—that is, unless Lloyd happened to have a good argument, which, so far, she has not made. Even if we assume that Lloyd can successfully

20 Darwin 1862, p197–203

argue against Reeve and Sherman’s definition—whilst projecting Sober’s definition as the correct one—there is still a way to undermine her. One thing that is clear is how Reeve and Sherman’s definition carries the risk of labelling non-adaptations as adaptations while, in the same context, Sober’s definition has the opposite risk of labeling adaptations as

non-adaptations. As we shall see later, the epistemic and non-epistemic benefits of detecting adaptations is worth the false positives that Reeve and Sherman’s definition may cause.

What is Fruitfulness?

Debates considering the evaluation of competing scientific theories and approaches are an unsettled matter. However, there is a general agreement upon which list of values should compose scientific appraisal, and these include consistency, accuracy, simplicity, scope, and fruitfulness.21 Debates among philosophers regarding the performance and definitions of these values persists. In essence, fruitfulness is a vague term and Lloyd uses it as such.

Despite the fact that fruitfulness is included in most lists of scientific virtues, it has been deprived of philosophical attention that has been, instead, granted to other virtues. One can easily find papers on simplicity, but very little work has been done to elucidate fruitfulness as well as its role in the assessment of scientific methodologies and theories. This is not inconsequential, as Silvia Ivani points out, “the problem with fruitfulness is that it can be easily ascribed to many programs because its definition is loose and no clear strategy for assessing it is provided.”22 Recently, Ivani (2019) has worked to develop a more precise definition of fruitfulness. Her definition is the “ability of research programs to

21 Khun, 1977, p320–339

22 Ivani 2019, p3

develop”23—that is, where development is when we advance our understanding of the world. Specifically, Ivani has epistemic goods—such as a novel hypothesis in mind—because a fruitful programme develops by gifting us with quality innovative hypotheses to test. Having methods to assess the fruitfulness of programmes is important, too. Here, Ivani suggests that, in our evaluation of research programs, we should turn to research questions and discovery heuristics. Both these tools are used by scientists in order to arrange and outline research as well as generate novel hypotheses. Research questions and discovery heuristics broaden, direct, and set the parameters of the content of research programmes.

Wording of research questions is crucial. Science is, essentially, a question answering process, so the questions we ask play a vital role in expanding our understanding of the world. Questions set the direction for what is and is not of interest, and where to begin our quest. As Sven Lundstedt puts it, “initial scientific questions, like first impressions, carry a great deal of weight in shaping the direction of a system of thought.”24 For example, consider the research questions below:

1. What is the evolutionary function of human guilt?
2. Is human guilt an evolutionary byproduct?
3. What is the link between smoking and stress?

While each of these questions allows a class of answers, they still forbid others. Hence, questions are vital to measure fruitfulness because they constrain or broaden the development of research programmes. Moreover, research questions guide research methodology—sample

23 Ivani 2019, p6

24 Lundstedt 1968, p229

size, what data is important, what data is not important—and they fundamentally limit answers that count as legitimate. The formulation of research questions depends on the background assumptions of scientists. As Ivani explains, these “assumptions can be of different sorts, e.g. moral, political, and methodological, and their importance can vary across contexts.”25

Research questions are composed of three dimensions, which are focus, aim and orientation. In the case of (b), the focus would be on psychological traits. The aim is to see if there is a function to the particular trait, orientation is the path that the study shall follow in order to develop. Questions (a) and (b) are based upon differing assumptions pertaining to the role of selection in the formulation of human emotions, which the former may assume to be empirical or explanatory adaptationism while the latter will not. All told, question (a) cannot be answered with “Human guilt is a spandrel” because the aim of the question only allows functional answers. However, question (b) would allow such an answer because the question’s orientation allows for both functional and non-adaptive answers. A question such as (c) might be based on a moral assumption that human well-being is our primary concern, and that we should better our understanding of those things that harm our health. Of course, only certain types of answers would be legitimate for this question—for example, “There is no correlation between smoking and stress” would not count because it is against the aim of the research question. A different research question allows a unique set of answers that are not available under other research questions. Hence, some research questions would be more fruitful than others because they permit relevant answers that other questions do not. The second tool for assessing fruitfulness is discovery heuristics. A research programme is said to be fruitful if it employs reliable heuristics. Being reliable means generating well-designed

25 Ivani 2019, p7

and rigorous hypotheses that are free of faulty reasoning. Moreover, even if a formulated hypothesis turns out to be false, this does not mean the research heuristic is unreliable. As long as the hypotheses generated are testable and we have reason to assume they may be true, that is enough to determine a reliable heuristic is being utilized. Research heuristics that produce numerous testable and novel hypotheses are more reliable than those that produce less—but quality also matters. If a research heuristic is producing a large number of sloppy yet creative hypotheses, it would not be considered reliable. It is important to carefully think about the fruitfulness of discovery heuristics because sometimes it takes the generation of a large number of hypotheses (the majority of which may turn out to be false) to come up with the right one.

Ivani’s original contribution to clarifying fruitfulness shall be beneficial whilst evaluating the competing frameworks under critique. I will argue Ivani’s explication of fruitfulness as well as the tools she provides for assessing programme works in favour of MA. This is because it helps us to develop programmes better than the EF framework. Lloyd’s reasoning for believing that EF is more fruitful than MA is twofold. Firstly, she thinks EF covers the functional answers MA provides and, secondly, that EF gives access to answers that are not allowed under MA. I am going to argue that the former is untrue and that the latter is insignificant and inconsequential when one looks at the overall benefits of MA. All said, these benefits are epistemic and non-epistemic. Below, I will cover the fruitfulness of MA and EF, which I shall execute via Ivani’s definition.

Fruitfulness of MA and EF

In relation to research, MA asks “What is the function of this trait?” Due to the wording of the question, functional answers—that is, the trait has a function of x, y, z—are favoured. The

focus of this question is biological traits, so its aim is to discover for what function natural selection has shaped the trait. The orientation of the question makes it clear that the trait, from the starting point of enquiry, is an adaptation. As such, from the onset, MA practitioners discriminate against non-adaptive answers, and they see functional answers as mutually exclusive to non-adaptive ones. The rejection of functional answers is considered to be necessary before non-adaptive answers are allowed in—therefore, in doing so, adaptationists are treating non-functional explanations as statistical nulls. This is not to say that the MA framework never allows non-functional answers, it does, but only temporarily. For instance, Ersnt Mayr writes that the adaptationist “must first attempt to explain biological phenomena and processes as the product of natural selection. Only after all attempts to do so have failed, is he justified in designating the unexplained residue tentatively as a product of chance.”26 There is an asymmetry here that is worth highlighting because, when a trait is labelled an adaptation, it is done so categorically, however, when a trait is labelled as a non-adaptation, it is only temporary. The discovery heuristics of MA—the strategies for developing novel hypothesis—are of two kinds: adaptive thinking and reverse engineering. Adaptive thinking is when scientists posit a specific adaptive problem in our evolutionary past and, subsequently, generate a hypothesis about the existence of a particular trait in order to solve that problem. After that, they then search for the said trait. Reverse engineering involves biologists studying a trait and working backwards to determine the particular types of functions that the trait would have potentially served for our ancestors. Both of these research heuristics produce adaptive hypotheses—that is, explanations of a trait in terms of selection pressures. However, Lloyd would claim these impoverish the quality of a well-designed hypothesis because there is a strong assumption of selective optimality. MA heuristics ignore or just give lip service to environmental constraints, developmental restrictions, and

26 Mayr 1983, p326

byproducts. As such, some have argued that, by monopolizing the explanations, they are using unreliable heuristics.27 However, others have claimed these heuristics are useful when studying some cases, but they perform poorly when studying others.28 Still, others have argued that adaptationist heuristics may produce some sloppy hypotheses, but there is no reason to think they systematically do so.29 Despite criticism, adaptationists continue to persevere. They highlight that MA is fruitful because it has accounted for previously unexplained phenomena and, also, it has solved novel problems. They would further add that it is not the case that they totally ignore constraints, for these are important in order to explain how new optimum peaks are reached30—what is more, the idea of selection without constraints is meaningless.31 All said, the fruitfulness of MA is, in part, due to its understanding of how non-adaptive factors work within adaptive modelling.

Within evolutionary psychology, adaptationist thinking is championed as a productive and fruitful programme.32 MA has also been used in evolutionary medicine to help identify and cure diseases.33 Indeed, MA is a very fruitful programme that generates novel hypotheses whilst stoking the creative imagination of scientists. Despite some criticism, the fruitfulness of MA is accepted by all—including Lloyd—so the question arises, why would she want to replace something that is working? While Lloyd agrees that MA is fruitful, she does not see EF as any less fruitful because it includes the benefits of adaptive answers. To that effect, she says, “But Methodological Adaptationism is so useful! Surely you are not advocating sacrificing our most fruitful research tool?! And no, I am not doing so, since the EF

27 Griffiths 2001 p309–325

28 Green 2014

29 Machery 2011, p232–246

30 Wade 2016

31 Rosenberg 2008, p75

32 Lewis 2017, Cosmides and Tooby 1994, Pinker 2002, Buss and Reeve 2003

33 Nesse and Stearns, 2008

framework includes the use of adaptation, and the search for connections to fitness or function, as a first ‘go to’ algorithm.”34 Lloyd finds MA unsatisfactory because it focuses on function, so non-adaptive answers get left by the wayside. This is problematic because traits that are non-adaptive are either mislabelled or only tentatively accepted as non-adaptive.

Lloyd complains that this can make us miss the actual status of traits, which she thinks leads to scientific failure. Conversely, the EF question—“What Evolutionary Factors account for the form and distribution of this trait?”—would give functional answers a limited time before allowing non-adaptive answers to float in. For Lloyd, EF is the best of both worlds because it includes adaptive answers (although she puts them first, she says the order does not matter) and non-adaptive answers. She does not consider adaptive ones to be mutually exclusive to non-adaptive ones, but, rather, that many factors can be simultaneously invoked to explain a trait. The focus of the EF question is also biological traits, but its aim and orientation differ from the MA question. Its aim is to find out what factors account for a trait, and its orientation is pluralistic. As such, EF allows genetic linkage, phyletic inertia, pleiotropy, and other such non-adaptive answers. In terms of prima facie, having more answers on the list allows EF to seem more fruitful than MA as a research programme. Therefore, if Lloyd is correct about EF including adaptationist answers, it seems indisputable that EF has a significant advantage. While MA may desirably develop the content of programmes, EF also does so. Thus, EF would be better at developing, improving, amending, and extending novel hypotheses because there are more explanations for the creative imagination of scientists to explore. If this is the case, why do MA practitioners have so much confidence in it as opposed to EF? The reason seems to be in the background assumption of the MA research question. That is, scientists make the methodological assumption that focusing on functional answers has been the only trustworthy generator of novel hypotheses, so moving away from

34 Lloyd 2015, p356

such as practice would be detrimental. Mayr and others have vocally opposed EF on these very grounds. However, if Lloyd is correct, it means EF includes the package of functional answers that are associated with MA and, hence, this methodological assumption would turn out to be false. In short, EF would be more fruitful than MA and there would be no disputing it.

Logically, the next question one might ask is “What about the research heuristics that EF practitioners would employ?” This is a difficult question to answer because Lloyd and others have not provided precise heuristics of any sort. It may be a type of Bauplan thinking and reverse engineering, biologists could start by examining a trait and determining what developmental or environmental restrictions would have been in place to keep the trait from evolving in a different way. Alternatively, they could also imagine what type of constraints would have been prevalent in the evolutionary past, and then, from there, venture out to find them. Perhaps this is not what Lloyd has in mind, it may simply be the case that Lloyd allows adaptive thinking and reverse engineering, yet, has a stricter criteria regarding the detection of functional capacities and selective explanations. Either way, research heuristics under EF prima facie would be more fruitful than MA because it allows a vaster pool of explanations (adaptive and non-adaptive). To this, MA practitioners would reply that, they do allow

non-adaptive answers, MA is not committed to the idea that the biological world is saturated with adaptations. Instead, they merely begin with an assumption that the trait they are studying is an adaptation and that this starting heuristic is the best way to find the actual status of the trait. They start with the assumption that a trait has been optimally shaped by selection and deviation from this optimal can help us get an idea of where to look for other factors. As mentioned earlier, Lloyd would say this is merely hypothetical in practice because adaptive answers do not budge to allow non-adaptive explanations. In relation to that, I would

agree with Lloyd that adaptationists cannot adequately make the case that they, in practice, allow non-adaptive answers as alternative explanations. I think she is right about MA practitioners only seeing non-adaptive factors as fodder that merely fits within an adaptive explanation.

Thus far, it seems as if Lloyd’s critique of MA is successful. That is, MA practitioners have a lack of a stopping rule for functional answers, they miss out when it comes to invoking

non-adaptive answers, they treat adaptive and non-adaptive as mutually exclusive, and they sometimes lose the ability to weigh evidence properly. Clearly, if we go by what we know so far, EF is more fruitful than MA. However, there is one crucial point that Lloyd is wrong about, and it is this that will give us the opening to directly challenge her framework. Lloyd claims that EF covers adaptationist answers and, also, even treats them as a priority. Although she is partially correct, EF does not cover all potential functional answers. Lloyd knows this is true and, what is more, she would not disagree because she not only speaks about a stopping rule for functional answers, but also writes against staunch persistent adaptive reasoning. Lloyd thinks a stoppage rule is harmless, but that is not true because it causes us to miss out on some adaptive answers, which is a significant problem. I believe the reason Lloyd thinks a stoppage rule is harmless is because she possesses a false methodological assumption that functions are easy to detect.

Lloyd challenges the persistent functional questions that MA practitioners revel in. However, it is precisely this type of tenacious adaptationist thinking that, despite a large number of initial failures, has led to functional discoveries. As Tim Lewens explains, “It is true, of course, that many traits turn out to have rather complex functions that a less persistent

biologist would never have noticed.”35 The very reason Lloyd desires to replace MA is its endless stubborn adaptive reasoning—yet, in some cases, this does lead to the discovery of elusive adaptations. In fact, Lloyd explicitly challenges Mayr because he thinks that, even if we constantly fail to find functional answers, we “tentatively” label those traits as

non-functional. Lloyd dismisses this asymmetry because she does not want tentative labels for non-adaptive answers—instead, she desires them to be categorically accepted in a manner akin to adaptive traits. Because EF has the potential to miss out some functional answers, it is not true that it includes the benefits of MA. This is significant because it means MA can produce epistemic goods that are unavailable under the EF framework, so the idea that the latter is more fruitful because it includes both adaptive and non-adaptive answers is incorrect.

Although the EF approach denies the opportunity to bank hard-earned adaptationist discoveries, Lloyd could respond that MA risks denying some non-adaptive traits their actual status. Her prize horse, the female organism, would be a case in point. Now, it seems we are at a deadlock, because both camps claim there is a risk in each approach. However, to properly appraise the frameworks under critique, we also need to clarify the impact of each risk. Lloyd and Naomi Oreskes’s recent work (2018) on anthropogenic climate change (ACC) would be a helpful place to start. They compared the two rival approaches (Risk-based and Storyline) to the detection and attribution (D&A) of the effect of ACC. Each approach had a different research question that Lloyd and Oreskes argued would lead to different risks. As the authors explain:

35 Lewens 2009, p170

Much of the discussion of risk-based approaches to D&A makes the claim, either explicitly or implicitly, that it would be wrong to overstate the contribution of ACC. Advocates for this approach are thus in effect arguing that a type I error (claiming something that is not the case or overstating an effect, also known as a false positive) is more serious than a type II error (missing something that is the case or understating an effect, also known as a false negative).36

Similarly, in our case, the MA and EF framework leads to different risks. Determining which risk is more important would help us ascertain the merits of each approach. And so, the question arises: “More important in what way?” In terms of the detection and attribution of ACC, it is quite clear that the two approaches have a direct impact on society’s concern with ACC. Therefore, in deciding what risk type is more damaging, we would be required to view it from a sociological perspective, which as the authors admit tells us that “there is no ‘right’ or ‘wrong’ approach to D&A in any absolute sense, but rather that in different contexts, society may have a greater or lesser concern with errors of a particular type. How we view the relative risk of overestimation versus underestimation of harm is context-dependent.”37 In our case, it is not that simple. Lloyd would posit that our goal is to get a true picture of the biosphere as well as a clear understanding of the processes that generates traits. She could argue that MA overestimates the effect of natural selection and, therefore, risks misleading us about the true nature of traits. Although EF misses out on some functional answers—because of the fact that it as a framework includes functional and non-functional answers—it is less risky when it comes to providing us with a false status of traits. The risk of accepting MA may lead to non-adaptive traits being mislabelled as functional or provincially functional, which would be a type 1 error. Conversely, we have reason to assume that the EF approach

36 Lloyd & Oreskes 2018, p316

37 Lloyd Oreskes 2018, p321

may lead to cases of type 2 errors—that is, where selective effects are downplayed and traits are labelled as non-functional, which would be in spite of the fact that they are shaped by selection. Without delving any further into the risks of each framework, we now have one small victory against Lloyd’s argument for replacing MA with EF. At the very least, we can say that Lloyd’s claim to categorically replace MA has been blunted. Similar to her own stance on the rival approaches to D&A, there is no outright wrong or right framework because it is all context-dependent. Society’s concerns largely dictate what risks are important and, therefore, what framework may be favoured. Lloyd must concede this point because it is clear—from her paper on ACC—that rival programmes are neither right nor wrong in an absolute sense, which means she needs to admit the same is true about EF and MA. If she does not, she would be inconsistent and would, thus, need to provide reasoning for why, in the case at hand, we have right and wrong approaches while, in ACC, we do not. Currently, it seems as if we have arrived at an impasse. Based upon what we know so far, both MA and EF have risks, so there is no clear winner. However, I think there is a clear way of showing how MA is superior to EF. In the next section, we shall examine cases of junk DNA and evolutionary medicine. I will show that the mislabelling of junk DNA was due to EF, and that only MA practitioners brought about its revision from “junk” to “functional”.

Also, I will show that the success of evolutionary medicine is due to adaptationist thinking. Both these cases have significant epistemic and non-epistemic consequences, so they will help us comprehend why MA is a more fruitful framework than EF.

Epistemic and Non-Epistemic Benefits of MA

Let us now consider the case of junk DNA. Biologists studying the genome found that large sections were doing nothing, and repeated attempts to find function came back empty. Only

1% of our DNA was found to code for proteins, so the remaining 99% was assumed to be meaningless garbage leftover after natural selection. Explaining this, biologist Wojciech Makalowski remarked, “For decades, scientists were puzzled by this phenomenon. With no obvious function, the noncoding portion of the genome was declared useless or sometimes called ‘selfish DNA,’ existing only for itself without contributing to an organism’s fitness.”38 In 1972, geneticist Susumu Ohno named these non-functional DNA as junk. The label stuck and was eventually accepted by the majority of biologists. In essence, junk DNA was defined as having two unique properties: “(1) It arises when a DNA sequence spreads by forming additional copies of itself within the genome. (2) It makes no specific contribution to the phenotype.”39 In the case of junk DNA, not only was there a failure to find function, but these noncoding regions were deemed detrimental to hold. In 1980, biologists Ford Doolittle and Carmen Sapienza wrote, “. . . there are two reasons to doubt that they arose or are maintained by selection pressures for such evolutionary functions. First, DNAs without immediate phenotypic benefit are of no immediate selective advantage to their possessor. Excess DNA should represent an energetic burden, and some of the activities of transposable elements are frankly destructive.”40 Significantly, biologists were discouraged from questioning the label of junk, and even eminent scientists—such as Francis Crick—openly called out the folly of obsessively hunting for functions. However, a small group of biologists committed to staunch adaptationism persisted, and, ultimately, their struggle was not in vain. As Makalowski points out:

Although very catchy, the term “junk DNA” repelled mainstream researchers from studying noncoding genetic material for many years. After all, who would like

38 Makalowski 2019

39 Orgel Crick 1980, p1

40 Doolittle & Sapienza, 1980, p602

to dig through genomic garbage? Thankfully, though, there are some clochards who, at the risk of being ridiculed, explore unpopular territories. And it is because of them that in the early 1990s, the view of junk DNA, especially repetitive elements, began to change.41

Junk DNA was still accepted by the majority of biologists until 2012. In that year, a landmark study by an international panel of scientists—published by the public research project ENCODE (Encyclopedia of DNA Elements)—had shown that the majority of the noncoding DNA was functional. Junk DNA was not actually junk—that is, it was found to be playing a crucial role in regulating the coding parts of the genome. What was previously thought to be meaningless rubbish had vital information for the understanding of many common diseases. Clarifying the implications of these findings, science writer Katherine Harmon says, “An international consortium of hundreds of scientists has now deciphered a large portion of the strange language of this junk DNA and found it to be not junk at all. Rather it contains important signals for regulating our genes, determining disease risk, height and many of the other complex aspects of human biology that make each one of us different.”42 It is not an exaggeration to say the revision of junk DNA opened up an epistemic treasure trove for biologists to explore. Indeed, there has been a paradigm shift in the way biologists now view noncoding DNA, which is far from meaningless rubbish because it is now considered valuable information that can radically improve our understanding. Since 2012, there has been a flurry of activities enhancing our understanding of genetics, inheritance, and diseases. Most recently (2019), a study found that noncoding elements play a critical role in biological

41Makalowski 2019

42 Harmon 2012

processes, and, therefore, knowing this “is necessary for understanding the underlying mechanisms of the diseases and to design effective treatments.”43

Considering the abundance of the noncoding regions as well as the fact that non-functional answers are only temporarily accepted, MA reasoning would never categorically label these noncoding DNA elements as junk. Under a strict MA regime, the absence of evidence for function is never evidence for absence. As mentioned earlier, EF and MA are both live approaches that are used by biologists, and it is Lloyd’s aim to have biologists decisively drop the latter for the former. Acceptance of non-functional junk DNA is exactly what one would expect from an EF approach because, essentially, this framework gives functional answers a limited space for exploration before deciding that a trait has a non-adaptive status categorically instead of tentatively. Proponents of EF did not encourage the lone clochards to continue the struggle against the odds in order to revise the status of junk DNA. Rather, it was a stubborn and persistent faith in MA that lead to this discovery. Everything that Lloyd suggests is wrong with MA—its lack of a stopping rule for functional answers, its privileging of adaptationism, its treatment of non-adaptive answers as nulls, and so on—turned out to be instrumental in the success of causing this revolution in genomics. According to the EF research question and discovery heuristics, Noncoding DNA would have been labeled as “junk” a long time ago (which it was), and there would be no room to redeem it of that status. EF practitioners have a serious problem to consider—that is, their reason for the rejection of MA was mislabelling of traits—yet that is exactly what they have done in this case, and it was far from inconsequential. At this point, in this discussion, we now have a clear epistemic reason to favour MA over EF. In short, MA has led to a goldmine of pursuits for scientists to explore while, conversely, EF’s labelling of junk was both a science stopper and a blockage

43 Noorul, Annette, & Chen 2019, p1

of progress in genomics. Llyod may respond that the door swings both ways, EF can lead to mislabelling a functional trait as non-adaptive and this had epistemic consequences, but the vice versa is also true. MA may lead to non-functional traits being labelled as functional and this has epistemic consequences too. I disagree with the symmetry that Llyod may try and draw. What epistemic benefits have we had from labelling a trait as non-functional? When a trait is found to be an adaptation there is a clear opening for further research as in the case above. When function is mentioned, ‘design’ has been found, this means design specifications are invoked, which means types of design problems need to inferred, which leads to selective pressures to be invoked and so on. Conversely when a trait is labelled

non-functional it is a science stopper. What further research can be done if a trait is labelled as junk, a spandrel, a product of drift, phyletic inertia, developmental or environmental constraints? Nothing or not much. The buck stops right there, there is no further major leads to pursue, it’s practically a dead end epistemically.

Lloyd has empirical and methodological assumptions behind her choice of EF. All told, she believes adaptations are ubiquitous and should be easy to spot, which is why she wants to give them limited time. As we have seen with the case of junk DNA, it is not true that functions can always be easily found. It took a long time—decades, in fact—to find functions for the DNA that was supposed to be junk. Lloyd’s methodological assumption that adaptations are easy to spot is wrong. Also, because non-functional DNA was labelled as junk in the 1970s, no one bothered to go further and do a detailed evaluation for the cause of such junk—that is, determine which specific non-adaptive factors lead to this, what proportion of the junk is due to drift, what amount is due to duplication, and what is constrained due to the Bauplan of the genome. No one cared because there was no benefit to such superfluous history. However, if one considers the flurry of work that has been done

since we discovered the junk was not junk, there are further studies, more research funding, an increased incentive to expand our understanding, and an epistemic return on investment with functional traits that do not exist with non-functional ones.

Clearly, these new findings pertaining to DNA also have enormous implications for our

well-being. Non-epistemic consequences call for the introduction of non-epistemic values in science, and these should affect the choice of methodology. I will argue that the cost of having too many false negatives in relation to functional traits is a greater risk than having too many false positives. This is because adaptive traits can help us find novel and previously hidden information that is important for well-being. Conversely, I believe that labelling a trait non-functional does not have this level of instrumental worth. If this is correct, it means it is in our interest to privilege MA over the EF framework—that is, for both epistemic and

non-epistemic reasons. Privileging MA shall circumvent most of Lloyd’s criticism of the programme.

There is an inductive risk involved in the process of labelling traits, and there are important non-epistemic consequences to these frameworks. Therefore, MA should not be replaced as the default framework. As mentioned earlier, research questions have background assumptions, and, of course, Lloyd would say that MA is based upon a false methodological assumption—namely, that starting with adaptationism as an organising principle would reveal the true status of traits. If this is the assumption, I agree that it is false—but it is also a moot point. A better methodological assumption would be that, because adaptive traits have epistemic and non-epistemic benefits in a way that non-adaptive traits don’t—we need to focus on the former and give little or no attention to the latter. What is more, we can invoke a moral background assumption because there are medical benefits to discovering functional

traits, and functional thinking (like in the case above) helps us to develop medicines, understand diseases, and comprehend inheritance. By contrast, discovering non-functional traits means nothing, medically, except a null result with no tangible benefits. Therefore, we have a moral duty to privilege the system that gives us medical benefits because the very reason we do science is for human well-being—that is, regardless of the risk of mislabeling of traits (false positives). This moral assumption is rather sensible because scientists are personally driven by a desire for human well-being, and, also, science research is granted based on multiple criteria, one of which is the practical benefits of projects. ENCODE is a case in point because although as a project it cost $400 million, the revision of junk DNA has given us promising potential avenues to radically improve human well-being. If this was not the case, the money would be unjustified. In fact consider if rather than revising the status of junk DNA, ENCODE instead reconfirmed it and we were given a detailed history of each non-adaptive factor as well as what role it played in leading to the garbage island that this noncoding region was supposed to be, there would be no public or scientific interest in knowing such trivial information. It would have been seen as an utter waste of time, money, and expertise. Indeed, scientists would not have been so keen to publish or even do research if ENCODE was a project to uncover the genomic mysteries of, say, anteaters! The very reason it is of importance and the public is willing to fund it—and scientists willing to labour—is the fact that this is about humanity and it’s well-being.

Another concrete example of the benefits of MA is the field of evolutionary medicine (also known as Darwinian medicine). Evolutionary medicine is the application of evolutionary biology to solve issues in medicine. Adaptationists wanted medicine to be integrated with evolutionary biology, so they came together to develop this field. It is important to highlight that it has only been since the 1990s that this field officially came into being. Before then,

medicine was largely independent of evolutionary thinking. At the forefront of this movement was Mayr, for he desired adaptationism to be centre stage in this new field. It is clear that evolutionary medicine—as it is practised today—is still based upon staunch MA. In a recent paper, Randolph Nesse and Stephen Stearns write, “At the core of evolutionary medicine is recognition that diseases need both proximate explanations of bodily mechanisms and evolutionary explanations of why natural selection has left the body vulnerable to disease.”44 Recently, there was a meticulous study that sought to “systematically elicit core principles from a diverse panel of experts in evolutionary medicine,”45 which found that, among practitioners of evolutionary medicine, 86.5% identified natural selection as a core principle that has “shaped all aspects of our biology, and results in adaptations.”46 And so, there is no doubt that evolutionary medicine—as it is currently practised—is based upon an MA framework instead of an EF one. Therefore, the successes of evolutionary medicine is due to MA. Indeed, by giving us new perspectives on diseases, cures and prevention, MA has directly benefited us. What is more, recent publications by proponents of evolutionary medicine highlight how adaptive accounts have been instrumental in understanding vulnerability to diseases, nutrition and development, miscarriage, cancer, auto-immune problems, mental disorders, and so on. Its practitioners argue that its success is due to its capacity to posit new research questions and to give us an integrated framework to synthesise medical knowledge with evolutionary history. In relation to this, Nesse points out five benefits of applying evolutionary thinking to medicine:

1. Expanding evolution’s contribution to existing research enterprises that rely on it (e.g., genetics, infectious disease, and research on aging).
   44 Nesse & Stearns 2007, p3045 Grunspan 2018, p146 Ibid
2. Providing a theoretical foundation for epidemiology and public health.
3. Heuristic value: formulating new questions about disease that motivate new studies.
4. Unifying research from different disciples.
5. Providing a framework for understanding disease from the perspective of evolutionary as well as proximate biology.47

Regarding each of these areas, Nesse goes into quite a bit of detail, with examples of successes and, also, potential avenues for future findings that would have medical benefits. As an example, here is a research question that evolutionary medicine could pose: “Why has natural selection left this species vulnerable to diseases?”48 According to Nesse, the potential answers can be put into six broad categories:

Natural selection is slow

1. Mismatch: Our bodies were shaped for environments far different from those we live in, and the mismatch gives rise to much disease.
2. Co-evolution with fast evolving pathogens: Because their generation times are so much shorter, pathogens evolve much faster than we can, so evolution cannot provide perfect protection against infection.There are limits to what selection can shape
3. Constraints: There is much that selection simply cannot do, such as starting a design from scratch to fix a design defect such as the vessels in the eyeball running between the light and the retina, or creating a gene replication method that never makes mistakes.
4. Trade-offs: Inevitable trade-offs make every trait suboptimal; for instance, if our vision was as acute as that of the eagle, our color vision and field of vision would be worse.47 Nesse 2007, p42448 Nesse 2007, p422Common difficulties in understanding what selection shapes
5. Reproduction at the expense of health: Natural selection increases the frequency of genes that yield a net increase in reproduction even if they compromise health and longevity.
6. Defences: Defences such as pain, fever, nausea, vomiting, and fatigue are not problems, but useful responses shaped by natural selection.49

Certainly, there is a great epistemic leap with the new research questions that evolutionary medicine allows, and this, obviously, has tangible medical benefits. We have almost 30 years of evidence to show the instrumental value of adaptationism in medicine. Moreover, the success of evolutionary medicine has been acknowledged and well-documented by experts from various medical fields.50 And yet, how has the EF framework lead to medical benefits? It has not, and there is no literature to show that it has. Acknowledging non-adaptive factors in evolution (except within an MA framework) may have added to our understanding of the biosphere, but it has failed to directly benefit us in a practical sense. A pluralistic account such as EF would have hidden some useful relationships and avenues for medical research. This is because EF does not grant the adaptive answers MA does, and it is these very elusive adaptive explanations that can have significant consequences for our well-being. If evolutionary medicine was based on an EF framework, would it have the success that we have today? If one wants to argue that it can be just as successful, they would, at best, give us hypothetical scenarios of how acknowledging non-adaptive factors would aid the development of medicines. Of course, it is possible to run an evolutionary medicine programme under an EF framework, and only then will it be clear whether or not it is better than the way it is practised today. Currently, there is no empirical evidence for that claim.

Evolutionary medicine does acknowledge a range of non-adaptive factors, but it does so,

49 Nesse 2007, p423

50 Stearns 2015; Gluckman 2016

strictly, within an MA framework—that is, in the exact way Mayr recommended. Adaptations are assumed to be optimal and, when it is found that a trait is suboptimal, non-adaptive factors are invoked to explain this divergence.

Because MA has already provided us with enormous medical benefits, and because it has the potential to do a lot more, it would be a moral duty to accept, endorse, and practice it.

Therefore, the moral assumption is well-founded. We should favor MA for the non-epistemic benefits it grants us. In addition to the moral, the methodological background assumption makes us choose the MA framework to be more fruitful. Having said that, the moral assumption has more weight than the methodological one, the reason why is that the latter is about the epistemic goods that can be gifted via a trait being labelled functional—while the moral assumption carries with it life changing implications. Imagine if, 40 years ago, we discovered that junk wasn’t actually junk. On the surface, it seems unlikely that we would have been able to make this discovery, however, if we imagine that the entire biological community was committed to hard-nosed adaptationism, it does not seem so far-fetched.

Indeed, it was not technological limitations that stopped biologists from looking for functions, it was the belief that noncoding DNA is non-functional. If we had, decades earlier, discovered what ENCODE did in 2012, we would, by now, have advanced significantly in our understanding of diseases, which, of course, would have directly impacted our

well-being. Missing out on traits that are non-adaptive is not as important as mislabeling adaptive traits as non-adaptive. That is, the risk of underestimating natural selection is greater than overestimating it. And so, the moral assumption gives us decisive reason to privilege MA over EF. What is more, the benefits of MA allow us to circumvent Lloyd’s criticism. Her problems with MA—its lack of stopping rule, mutually exclusive treatment of adaptive and non-adaptive answers, and treating non-functional alternatives as nulls—turn out to be a

virtue rather than a vice because there is epistemic and non-epistemic reasons to privilege MA, which is why it is a more fruitful as a system and should not be replaced.

Conclusion

Lloyd has strongly critiqued MA and desires to replace it with the EF framework. She believes it is more fruitful than MA because it has more answers available, and, also, that EF includes adaptationist explanations as a first-go to algorithm. I have argued against this because EF misses out on some adaptationist answers. Likewise, under MA, non-adaptive answers are missed out. Although there is a risk of overestimating functional answers in MA and underestimation in EF, I have argued that this does not lead to a deadlock because we have reasons to assume adaptive answers are more important than pluralistic answers. The reasoning behind this is the fact that adaptive answers have epistemic and non-epistemic benefits that non-adaptive ones don’t. To that regard, I used the example of junk DNA and evolutionary medicine to illustrate this asymmetry. In knowing this, we must privilege adaptive answers because they are more fruitful. Indeed, doing so would dampen Lloyd’s criticism of MA, for what she thinks is a disadvantage works out, instead, to be an advantage. Therefore, Lloyd’s argument to replace MA with EF does not work.

Bibliography

Alcock, J. (1987) Ardent adaptationism. Nat Hist 96:4.

Alvergne, A. Jenkinson, C., & Faurie, C. (2016). Evolutionary Thinking in Medicine: From Research to Policy and Practice (Advances in the Evolutionary Analysis of Human Behaviour), Springer. 1st ed.

Andrews, P.W., Gangestad, S.W., & Matthews, D. (2002a) Adaptationism—how to carry out an exaptationist program. Behav Brain Sci 25(4):489–504.

Bausman, W., & Halina, M. (2014). Not null enough: pseudo-null hypotheses in community ecology and comparative psychology. Biology and Philosophy, 33:30.

Boland, R. (2017). Non-coding RNA: It’s Not Junk, Dig Dis Sci 62: 1107.

Buss, D.M., & Kern Reeve, H. (2003). Evolutionary psychology and developmental dynamics: Comment on Lickliter and Honeycutt (2003). Psychological Bulletin, 129(6), 848–853.

Carey, N. (2015), junk DNA: A Journey Through the Dark Matter of the Genome. London: Icon Books Ltd

Cosmides, L., & Tooby, J. (1994). Beyond intuition and instinct blindness: Toward an evolutionarily rigorous cognitive science. Cognition, 50, 41–77.

Darwin, C., (1860). Origin of Species, second British edition.

Darwin, C., (1860), On the various contrivances by which British and foreign orchids are fertilised by insects, and on the good effects of intercrossing. London: John Murray.

Doolittle, F., & Sapienza C., (1980). Selfish genes, the phenotype paradigm and genome evolution, Nature 284, 601–603

Douglas, H., (2000). Inductive Risk and Values in Science, Philosophy of Science, Vol. 67, No. 4, pp. 559–579

Elliott, K.C., McKaughan, D.J. (2014). Nonepistemic Values and the Multiple Goals of Science. Philosophy of Science. Vol. 81, No. 1. pp. 1–21.

Germain, P.L., Ratti, E., & Boem. (2014). Junk or functional DNA? ENCODE and the function controversy. Biology & Philosophy, Vol 29, Issue 6, pp 807–831.

Gluckman, P., Beedle, A., & Buklijas, T. (2016). Principles of Evolutionary Medicine Paperback. Oxford: Oxford University Press.

Godfrey-Smith, P. (2001). Three kinds of adaptationism. In: Orzack SH, Sober E (eds) Adaptationism and optimality. Cambridge University Press, Cambridge, pp 344–362.

Gould, S., & Lewontin, R. (1979). The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings of the Royal Society of London, Series B. Vol. 205, No. 1161: 581–598.

Gould & Vrba (1982) Exaptation: a missing term in the science of form. Paleobiology 8:4–15.

Green, S. (2014). A philosophical evaluation of adaptationism as a heuristic strategy. Acta Biotheoretica. 62, 479–498.

Griffiths, P. (2001). From adaptive heuristic to phylogenetic perspective: Some lessons from the evolutionary psychology of emotion. In H. Holcomb (Ed.) Conceptual challenges in evolutionary psychology Dordrecht, Springer.

Grunspan, D.Z., Nesse, R.M., Barnes, M.E., & Brownell, S.E. (2018). Core principles of evolutionary medicine: A Delphi study. Evolution, Medicine, and Public Health, Volume 2018, Issue 1, Pages 13–23.

Grunspan, D., Nesse R., Barnes E., & Brownell, S. (2018). Core principles of evolutionary medicine, Evol Med Public Health. 2018; 2018(1): 13–23.

Ivani, S. (2019). What we (should) talk about when we talk about fruitfulness. European Journal for Philosophy of Science. 9:4.

Kuhn, T. (1977). Objectivity, value judgement, and theory choice. In The essential tension. Selected studies in scientific tradition and change, Chicago, University of Chicago Press.

Lewens, T. (2009). Seven types of adaptationism. Biology and Philosophy. Biology and Philosophy 24 (2):161–182

Lewis, D.M., Al-Shawaf, L., Conroy-Beam, D., Asao, K., & Buss, D.M. (2017). Evolutionary psychology: A how-to guide. American Psychologist, 72(4), 353–373.

Lloyd, E.A. (2006). The Case of the Female Orgasm: Bias in the Science of Evolution. Massachusetts: Harvard University Press.

Lloyd, E.A. (2015). Adaptationism and the Logic of Research Questions: How to Think Clearly About Evolutionary Causes. Biological Theory. Volume 10, Issue 4, pp 343–362.

Lloyd, E.A., & Oreskes N. (2018). Climate Change Attribution: When Is It Appropriate to Accept New Methods?. Earth’s Future. Vol 6. Issue 3. Pages 311–325.

Lundstedt, S. (1968). A note on asking questions. The Journal of General Psychology, 79(2), 229–239.

Machery, E. (2011). Discovery and confirmation in evolutionary psychology. In Jesse J. Prinz (Ed.), The Oxford handbook of philosophy of psychology. Oxford: Oxford: Oxford University press.

Makalowski, W. (2019). What is junk DNA, and what is it worth? Retrieved from URL: https://www.scientificamerican.com/article/what-is-junk-dna-and-what/?redirect=1.

Accessed: [7/8/2019]

Mayr, E (1983). How to carry out the adaptationist program? Am Nat 121:324–334.

McDonald, H.G., Vizcaíno, S.F., & Bargo, M.S. (2008). “Skeletal anatomy and fossil history of the Vermilingua”. In Vizcaíno, S.F.; Loughry, W.J. The Biology of the Xenarthra.

McGrath, A., Chen, Y.P., & Amin, N. (2019). Evaluation of deep learning in non-coding RNA classification. Noorul Amin, Nature Machine Intelligence, 246–256.

Nesse, R.M., & Stearns, S.C. (2007). The Great Opportunity: Evolutionary Applications to Medicine and Public Health. Evolutionary Applications.

Nesse, R.M. (2007). The Importance of Evolution for Medicine, Edited by: Trevathan Wenda R., Smith E. O., J. McKenna James, Evolutionary Medicine, Second Edition, New York, Oxford University Press.

Orgel, L.E, & Crick, F.H.C. (1980). Selfish DNA: the ultimate parasite. Nature 284:604–607.

Pigliucci, M., & Kaplan, J. (2000). The Fall and Rise of Dr. Pangloss: adaptationism and the Spandrels paper 20 years later, Trends in Ecology and Evolution 15 (2):66–77.

Pinker, S. (2002). The Blank slate: The modern denial of human nature. New York: Viking.

Reeve, H.K., & Sherman, P.W. (1993). Adaptation and the goals of evolutionary research. Q Rev Biol 68(1):1–32.

Rosenberg, A. (2008). Philosophy of biology: A Contemporary Introduction (Routledge Contemporary Introductions to Philosophy). New York: Routledge.

Sober, E. (2000). Philosophy of Biology, 2nd Edition (Dimensions of Philosophy). New York: Westview Press.

Stearns, S. (2015). A Primer of Evolutionary Medicine Paperback. Oxford: Oxford University Press.

Valles, S.A. (2012). 27: 241. Evolutionary medicine at twenty: rethinking adaptationism and disease. Biology & Philosophy, March 2012, Volume 27, Issue 2, pp 241–261.

Wade, M.J. (2016). Adaptation in metapopulations: how interactions change evolution. University of Chicago Press. Chicago.

Williams, G.C.1., & Nesse, R.M. (1991). Q Rev Biol. Mar;66(1):1–22.The Dawn of Darwinian Medicine.

Williams, G.C. (1966). Adaptation and selection. Princeton University Press, Princeton.