Appeared in: Studies in the History and Philosophy of Biological and Biomedical Sciences; 41 (2010):

Transcription

1 Appeared in: Studies in the History and Philosophy of Biological and Biomedical Sciences; 41 (2010): PENULTIMATE DRAFT What Can Natural Selection Explain? Ulrich E. Stegmann Department of Philosophy, University of Aberdeen, Old Brewery, High Street, Aberdeen, AB24 3UB Abstract One approach to assess the explanatory power of natural selection is to ask what type of facts it can explain. The standard list of explananda includes facts like trait frequencies or the survival of particular organisms. Here I argue that this list is incomplete: natural selection can also explain a specific kind of individual-level fact that involves traits. The ability of selection to explain this sort of fact ( trait facts ) vindicates the explanatory commitments of empirical studies on microevolution. Trait facts must be distinguished from a closely related kind of fact, i.e. the fact that a particular individual x has one trait rather than another. Whether or not selection can explain the latter type of fact is highly controversial. According to the so-called Negative View it cannot be explained by selection. I defend the Negative View against Nanay s (2005) objection. Key Words Explananda of natural selection, explanation of traits, adaptation, Sober-Neander controversy, the Negative View. 1. Introduction What kinds of facts can natural selection explain? Here is a standard list of things that at least some authors believe selection can explain: 1. The dynamics of trait frequencies in populations across time, i.e. their change or stagnation (Sober, 1984); 1

2 2. The composition of a population at a particular point in time (Sober, 1995, p. 384), e.g. the fact that 90% of the population are G-individuals and 10% are B-individuals (individuals with trait G or B, respectively); 3. The origin of traits in a population, in addition to their spread and maintenance as acknowledge in (1) (Forber, 2005). 4. An individual s survival, its reproductive success (Sober, 1984, p. 152) and its existence (Sober, 1995, p. 388); 5. The fact that a particular individual has trait G rather than trait B. It is uncontroversial that selection can explain facts (1) and (2). It is also generally accepted that selection can explain (4), at least as long as selection is regarded as a causal process 1. However, a long-standing controversy surrounds (5). Many authors deny that selection explains the traits of individuals (Sober, 1984; 1995; Walsh, 1998; Lewens, 2001; Pust, 2001; 2004), a position known as the Negative View 2. Others insist that selection can explain this fact in special circumstances (e.g., Neander, 1995; Matthen, 1999; Forber, 2005; Nanay, 2005). More recently, Forber (2005) has added (3) to the list, arguing convincingly that under some circumstances selection explains the origin of traits. It is easy to loose sight of the significance of this debate, so much so that some may wonder whether it has any. So let me say why I think it is important, but also where its relevance may have been overstated. First, the list of explananda is not a haphazard list of empirical facts for which we have good evidence that selection explains them. Debating such a list might well be tedious and pointless. Rather, the list, and the debate surrounding it, concerns the kinds of facts selection can explain. Identifying the kinds of facts selection can explain is imperative if the goal is to understand the explanatory power of natural selection. And surely, understanding the explanatory power of such a fundamental notion is a worthy goal for both philosophers of science and biologists. The debate is essential for this project. Second, scrutinizing the Negative View in particular is justified because it is deeply counter-intuitive. It suggests that strong selection for, say, long necks in giraffes over many generations does nothing whatsoever to explain why a particular giraffe has a long neck, despite its manifest causal impact in the past. That the Negative View is so counter-intuitive reveals that there is more to natural selection than meets the eye, which makes the debate all the more pressing. In the end I believe the Negative View is correct. But it would be a mistake to take this as vindicating the view that evolutionary biology is somehow not interested in particular individuals. Some branches of this discipline are not, but others are. Behaviour ecology, in 2

3 particular, takes profound interest in the properties of individual organisms, as is vividly documented in recent work on, for instance, personality traits (Réale et al., 2007) and the social significance of long-term memory of elephant matriarchs (McComb et al., 2001; Foley et al., 2008). Indeed, while the Negative View is correct, evolutionary biology is nonetheless committed to explaining a certain kind of fact concerning traits at the individual level, or so I will argue. The significance of the debate has been overstated in other respects. It has been argued that the Negative View implies two far-reaching consequences, i.e. that selection cannot explain adaptations (e.g., Neander, 1995; Nanay, 2005) and that, on a grander philosophical scale, it dissolves the basis of teleosemantic theories of mental content (Walsh, 1998; Nanay, 2005). I believe it implies neither. Admittedly, there is something right in claiming that selection cannot explain adaptations on the Negative View. According to the Negative View, selection cannot explain why a particular individual has trait G rather than B; so, when G is an adaptation, it cannot explain why the individual has adaptation G, rather than B. However, the claim that selection cannot explain adaptations tout court follows only on the further assumption that explaining adaptations consists in explaining why a particular individual has adaptation G rather than B. That assumption is questionable. Explaining adaptations may reasonably be interpreted as involving the explanation of population-level facts regarding adaptations, especially the spread and eventual prevalence of an adaptation (Sober, 1984) or the composition of populations at a given point in time (Sober, 1995). 3 The view to the contrary may spring from a familiar way of construing adaptations. Adaptations are often taken to be traits whose possession by particular individuals is due to selection (cf. this giraffe has a long neck because of past selection for long necks ). 4 Consequently, if selection cannot explain why individuals have a trait, then that trait is not an adaptation. From this vantage point the implication of the Negative View appears even worse than the critics allege. It is not so much that selection cannot explain adaptations, but rather that there are no adaptations in the first place, because selection lacks that crucial bit of explanatory power. But, of course, we may and should revise our concept of adaptation to fit the explanatory status selection actually has. 5 With such a refinement in place, there is no more reason to demand that the only genuine explanations of adaptations be explanations of why an individual has adaptation G. Does the Negative View undermine the very project of accounting for mental content in terms of evolutionary function? Consider what the various versions of teleosemantics require. 3

4 What they require, among other things, is that token mental states, or the token mechanisms producing them, have evolutionary functions. But as just argued, token states or mechanisms can be adaptations without selection being able to explain why particular individuals have them. Hence, since teleosemantics is not tied to the truth or falsity of the Negative View, the debate does not have quite the ramifications some think it has. Finally, a word about an issue that has complicated the debate about the explanatory status of natural selection: how the identity of individual organisms is determined. One prominent way of delineating individuals is origin essentialism, which is the view that an individual s identity is determined by parentage or, in other words, that a given individual could not have been generated by different parents. Since the Negative View relies on origin essentialism (Matthen, 1999; Pust, 2001), the debate has partly turned on whether in evolutionary biology this view is plausible (Lewens, 2001; Pust, 2001; 2004) or not (Matthen, 1999; 2003; Forber, 2005). But the debate about the explananda of selection is broader than this. As will be shown here, important issues remain to be settled even on the assumption of origin essentialism, not least because a more recent attack on the Negative View does not target this assumption (Nanay, 2005). I therefore adopt origin essentialism for the purposes of this paper. 6 Here is the plan. In the first section I argue that selection can explain an individual-level fact that has been overlooked thus far. In other words, I believe the present list of explananda is incomplete. One consequence, explored in the second section, is that the Negative View can be shown to account for certain fine-grained explanatory commitments entailed by some studies of microevolution. Another consequence is to appreciate that the new fact can itself assume a unique explanatory role under some conditions. The last section defends the Negative View against Nanay s (2005) objection. 2. Selection explains trait facts Consider a hypothetical uniparental organism (fig. 1) 7. Individuals are born consecutively into lineages of this organism, so that there will be one second, one third, one twenty-fourth member, and so on. Definite descriptions like the lineage s second member, abbreviated in figure 1 as numerals (e.g. 2), are used to denote whichever particular individual happens to be the relevant member of a given lineage. The numerals are therefore non-rigid designators. Finally, lineages of this organism are individuated by their founding individuals, e.g. 1 in figure 1. 1 denotes the same individual in figures 1A and 1B. The two figures depict alternative variants of the same lineage, which differ in their trajectories after the second generation. Figure 4

5 1A depicts the actual lineage, in which only 3 reproduces (in generation II). Figure 1B depicts the lineage s trajectory if, contrary to fact, 2 had reproduced instead of 3. [Space for figure 1] Quite obviously, distinct individuals can become any given member depending on the circumstances (with the exception of 1). For example, in the actual circumstances only 3 reproduces and the lineage continues with an offspring of 3 (fig. 1A). The lineage s next (fourth) member is therefore an offspring of 3. By contrast, if only 2 had reproduced, then the fourth member would have been an offspring of 2 and, hence, a different individual (fig. 1B). Which organism reproduces has consequences for how the traits G and B are distributed (the parents of our hypothetical organisms simply pass on to their offspring whichever trait they happen to have; 3 is the exception insofar as a mutation caused it to develop G). Since in the actual lineage (fig. 1A) one of 3 s offspring becomes the fourth member, and since all of 3 s offspring have G, 4 will be a G-individual. If, however, 2 had reproduced instead of 3, the fourth member would have been a B-individual, because 2 s offspring would have had B (fig. 1B). In other words, the fourth member may be a G- or a B-individual depending on the circumstances. [Space for figure 2] In the actual circumstances, the property of being the fourth member is exemplified by a G-individual (rigidly denoted as a 1 in fig. 2B). But if circumstances had been different, then the fourth member would have been a B-individual (rigidly denoted as a 2 in fig. 2B). Facts like the fourth member s being a G-individual, rather than a B-individual, are called trait facts in what follows. For comparison, figure 2A illustrates what might be termed a Sober-fact: an individual has trait G rather than B. We can now ask: why is the fourth member in figure 1A a G- rather than a B-individual? The answer is that all of 3 s offspring have G and 4 happens to be one of 3 s offspring. Hence 4 is a G-individual. Now, one of the explanatory factors may easily depend on natural selection. More specifically, the fact that the fourth member is one of 3 s offspring may depend on selection. We only need to assume that generation II experiences strong selection for G, so strong that only G-individuals can reproduce. Since in that generation only 3 has G, only 3 will reproduce. The next (fourth) member of the lineage will therefore be an offspring of 3. So, in 5

6 virtue of explaining why the fourth member is an offspring of 3, natural selection contributes to the explanation of why the fourth member is a G-individual. What if selection were less than absolute, allowing 2 to generate a few offspring alongside 3? If the fourth member would still be one of 3 s offspring, rather than one of 2 s, then this would simply be an effect of 3 giving birth a bit earlier than 2. Since selection has no influence on the order of births, selection would not determine that the fourth member stems from 3 (rather than 2). So for selection to explain why the fourth member is an offspring of 3, it may seem as if we must endorse the unrealistically strong assumption that selection is an all-ornothing affair. 8 However, this worry is overstated. First, all-or-nothing selection may occur, even if rarely. So there is a set of circumstances in which selection explains trait facts. Saying that selection can explain trait facts is not to say it always does. Second, even when selection is not all-ornothing it can explain the probability of a certain member stemming from a particular parent. For example, if due to selection 3 contributed 90 % of generation III, then selection would account for the probability being 0.9 that the next (fourth) member stems from 3 (rather than 2). And given that all of 3 s offspring have G, selection would also make it likely that 4 would be a G-individual, rather than a B-individual. Further, selection is responsible for the probability being 1.0 that 4 is a G-individual in the deterministic case initially presented. A different worry concerns the relation between the referent of the fourth member and what Matthen (1999; 2003) calls receptacles. At least on one interpretation, Matthen s receptacles are best construed as abstract entities that can be filled by different genomes, which in turn may be identified with distinct individuals (Lewens, 2001, p. 594). It may therefore seem as if there is not much of a difference between receptacles and genealogical positions, since both are like slots that can be filled with different individuals. But the similarity is only superficial. Matthen (1999; 2003) rejects origin essentialism, instead embracing the view that different parents may generate the very same individual. Such individuals are receptacles. That is, receptacles are individuals whose identity does not depend on parentage (Matthen, 2003, p. 306), and who instead have a very thin essence (p. 307). By contrast, the fourth member is neither a name for an individual with wide identity conditions, nor for a particular slot or genealogical position that exists independently of the individual filling it. It rather denotes an ordinary individual, i.e. as individuated by its parentage, in virtue of its having the property of being the fourth member. Consequently, whereas for Matthen selection explains Sober-facts involving thinly specified individuals, I suggest it explains trait facts involving ordinary individuals. 6

7 3. Implications Once it is accepted that selection explains trait facts it can be seen that the Negative View accounts for the fine-grained explanatory commitments associated with some empirical studies in microevolution. This is the first implication I will explore in this section. Studies of microevolution in the wild have documented cases in which the crossgeneration change in the mean value of a phenotypic trait is due to directional selection for the trait in question. For example, selection for larger weight in a species of Darwin finches resulted in birds of the post-selection generation being heavier, on average, than birds of the preselection cohort (Grant and Grant, 1995, and references therein). Similarly, members of a guppy species matured at a later age after 18 generations of selection for increased age at maturity (Reznick et al., 1997). In some studies, notably Peter and Rosemary Grant s long-term project on Darwin finches, organisms were marked individually. This enabled tracing parent-offspring relations and documenting high heritabilities (e.g., larger-billed individuals produced larger-billed offspring). It was also shown that individuals with, say, larger bills survived and reproduced because of selection (they were better at opening hard seeds, which became the only food available during a severe draught). Taken together, these data imply that selection can in principle explain why a specific member of a lineage is a heavier or large-billed individual, rather than a lighter or small-billed individual. In other words, the level of detailed information available in some microevolutionary studies implies the possibility that selection can explain trait facts. This degree of explanatory resolution is not readily accounted for by the explanatory powers of selection as presently conceived by the Negative View. On the one hand, proponents of the Negative View accept that selection may explain facts concerning individuals. But these facts have nothing to do with traits: they concern an individual s survival, its reproductive success (Sober, 1984, p. 152) and its existence (Sober, 1995, p. 388). On the other hand, it is acknowledged that selection can explain facts regarding traits. But these facts do not pertain to individuals: what selection explains about traits are population-level facts, e.g. that a population is entirely composed of X-individuals rather than of Y-individuals (Sober, 1995, p. 384). Thus, the Negative View has not yet been explicitly construed as allowing selection to explain individual-level facts concerning traits. In conclusion, although selection cannot explain why an individual has A instead of B on the Negative View, it nevertheless allows selection to explain a different individual-level fact 7

8 concerning traits. This restores some degree of explanatory power at the level of individuals, which selection seemed to have lost. I have made liberal use of the distinction between individual-level and population-level facts, but more needs to be said about it. Intuitively, the distinction is clear enough: trait facts are individual-level facts 9 because a proposition like the fourth member is a G-individual asserts something about the individual that satisfies the definite description. It expresses a fact concerning that individual, just as the current president of France has dark hair asserts something about the man Nikolas Sarkozy. However, when attempting to specify what underpins the distinction, complications arise. The distinction does not appear to rely on a difference in the relevant truthmakers. 10 So I suggest appealing to the subjects of predication instead: if the subjects are individuals, the fact is individual-level; but if they are populations, then the fact is population-level. Thus, Lonesome George s surviving to reproductive age is an individual-level fact because has survived to reproductive age is predicated of an individual. Similarly, being taller than is predicated of individuals, as is is a G-individual. Both rigid and non-rigid designators may be used in subject position to denote individuals. By contrast, contains 70% females is predicated of a population, not of individuals. Of course, the population is composed of individuals and they contribute to making the proposition true, but they are not the subject of predication; contains 70% females is not predicated of any individual. 11 So far I argued that acknowledging trait facts helps restoring some of the fine-grained explanatory force of selection. The second implication to emphasize is that under some conditions trait facts themselves can have unique causal and explanatory relevance. Such conditions could be met, for example, in birds with obligate siblicide. Some eagles, pelicans, penguins, vultures, cranes, and owls lay two eggs per clutch, although only the first-hatched nestling survives. The older nestling, being larger and heavier, outcompetes the younger over access to food, eventually killing it (e.g., Edwards and Collopy, 1983; Anderson, 1990). Now, suppose the older nestling, member 50 of the lineage, expresses the advantageous novel trait G due to a rare mutation and kills its younger sibling, 51. G spreads in subsequent generations because it is advantageous and its first bearer, 50, survives to reproduce. 12 In this scenario, the evolutionary consequences uniquely depend on the trait fact that 50 is a G-individual. If 51 (rather than 50) had been the G-individual, the new trait would not have spread because 51 was destined to die and no other individual in the population had G. Furthermore, even though it may now be Polly who has G, G s spread does not depend on Polly s having G. For Polly might well have been 51, in which case her having G would have been inconsequential (because she 8

9 would have been killed), and any G-individual other than Polly would have ensured G s spread provided it had been 50 instead of her. Trait facts can therefore make causal and explanatory contributions that neither population-level nor Sober-facts can make Defending the Negative View In the previous section I argued that the standard list of explananda of natural selection is incomplete. We should add trait facts. In this section I defend the view that Sober-facts are not proper explananda for selection. As mentioned in the introduction, whether selection can explain Sober-facts has been discussed for many years. Some authors insist that selection can explain this sort of fact at least in special circumstances, i.e. when selection is cumulative (e.g., Neander, 1995), when reproduction is sexual (Matthen, 1999; 2003), or when resources are limited (Nanay, 2005). Neander s and Matthen s arguments have already been discussed extensively elsewhere and found wanting 14. I will therefore be concerned with Nanay s argument. Nanay (2005) asks us to consider a case in which an individual x has trait A, there is selection for A in x s population, and the population also contains B-individuals. His argument then aims to show that x s having A ultimately depends on selection. Here is the argument (paraphrased from Nanay, 2005, p ) 15 : 1) X s having A depends counterfactually on x s mother having A and surviving to reproductive age. 2) The mother s survival depends counterfactually on the death of B-individuals. 3) The death of B-individuals depends counterfactually on selection for A. Hence, x s having A depends counterfactually on selection for A. Selection therefore partly explains x s having A. Setting aside other issues 16, my worry here is with the first premise. I believe that the first premise is ambiguous because it does not specify the relevant contrast class, and that proper disambiguation breaks the chain of counterfactual dependence on selection. The phrase x s having A counterfactually depends on from the first premise asserts a counterfactual dependence. But it does not specify whether what depends is (1) x s, rather than y s, having A, or (2) x s having, rather than lacking, A, or (3) x s having A, rather than some other trait B. The debate about whether selection can explain Sober-facts only concerns contrast (3) (Pust, 2001; the significance of contrast classes to explanatory projects in general has been emphasized for instance by Van Fraassen, 1980) 17. So, in order for Nanay s argument to work against Sober, the first premise must be disambiguated accordingly: 9

10 1*. X s having A, rather than B, depends counterfactually on x s mother having A and surviving to reproductive age 18. The first thing to note is that premise 1* makes x s having A rather than B (or x s having A, for short) counterfactually dependent on a conjunction of two conditions. Both must obtain for x to have A instead of B. The truth of 1* therefore implies that if one of the conditions did not obtain, then it would not be the case that x had A, rather than B. We can assess 1* on the basis of this implication. Unfortunately, it is not entirely clear what the truth conditions are for counterfactuals whose antecedent is a proposition of the form A rather than B. This in itself weakens Nanay s argument somewhat. It seems the best we can do, short of fully working out the semantics of these counterfactuals, is to impose the following constraint: the negation of A rather than B is equivalent to B rather than A. This constraint holds quite generally for such counterfactuals. Consider the following example: my remaining dry (rather than getting wet) during rain counterfactually depends on my using an umbrella (rather than not using it). I would get wet if I did not use it. So, the counterfactual asserts that if the relevant condition was not satisfied (if I did not use my umbrella), then the contrasting case of the antecedent would be realized (I would get wet). It seems this is the whole point of asserting that my remaining dry counterfactually depends on my using the umbrella. With this constraint in place, we can now assess 1* by asking whether x would have B instead of A if either one of the two conditions did not hold. One of the conditions is that x s mother survives to reproductive age. So is it true that if the mother had died before reaching reproductive age, then x would have B instead of A? Clearly not. If the mother had died, the very same x would not have been born (assuming origin essentialism) and x would not have B or any other trait; a fortiori, x would not have B instead of A. 19 Premise 1* therefore falsely asserts a counterfactual dependence both on x s mother having A and on her survival. Now, the condition that allegedly links x s having A to selection (via the chain of counterfactuals expressed in premises 2 and 3) is the mother s survival. But since x s having A does not counterfactually depend on the mother s survival, it does not depend on selection either. These considerations therefore sever the purported chain of counterfactual dependencies. Further, since the mother s survival is the only condition linked to selection, it does not help to point out that x s having A does depend on the other condition, i.e. on x s mother having A. 20 One worry still needs to be addressed. X needs to have trait A in order to have A rather than B. If x did not have A in the first place, there is no sense in which it could have A instead of another trait. Furthermore, for x to have A, x s mother must survive to reproductive age; x would 10

11 not be born otherwise. Apparently then, if the mother had died, x would not have A. Does this not show that x s having A counterfactually depends on the mother s survival, after all? All this shows is that the mother s survival and x s existence are preconditions for x s having A, albeit preconditions not unique to x s having A. They are also preconditions for the opposite fact, x s having B rather than A: if x s mother had died, x would not have existed and it would not have any traits (e.g., x would not have B as opposed to A). The mother s survival and x s existence are therefore preconditions common to x s having either A or B. They do not decide between the two alternatives. 21 Acknowledgments Bence Nanay, Phyllis McKay, Mark Textor, Eliott Sober, Kim Sterelny, and Dennis Walsh commented very helpfully on earlier versions of this paper. Audiences at the Stirling Workshop in Philosophy of Biology and the First Graduate Conference on the Philosophy of the Natural Sciences (both in May 2005) provided valuable feedback. Financial support from the British Society for the Philosophy of Science and the British Academy is gratefully acknowledged. 11

12 Figures and legends 4 (G) 5 (G) 4 (B) 5 (B) III 2 (B) 3 (G) Selection 2 (B) 3 (G) II 1 (B) 1 (B) I A B Figure 1. Different trajectories of a lineage of a uniparental organism. Numbers denote (non-rigidly) individual organisms; B and G denote distinct types of traits. (A) Selection favours G in generation II such that only 3 reproduces. (B) Trajectory of the same lineage if selection had favoured B instead of G, such that only 2 would have reproduced. 12

13 G B A a 1 G 4 B a 1 a 2 B Figure 2. Comparing Sober-facts with trait facts. Squares represent particular individuals, circles represent properties; solid lines represent actually obtaining features, whereas dotted lines represent non-actual features. (A) A Sober-fact: individual a 1 has trait G rather than trait B. (B) A trait fact: the lineage s fourth member is a G-individual (here: a 1 ), rather than B- individual (a 2 ). 13

14 References Anderson, D. J. (1990). Evolution of Obligate Siblicide in Boobies. I. A Test of the Insurance- Egg Hypothesis. The American Naturalist, 135(3), Edwards, T. C. & Collopy, M. W. (1983). Obligate and Facultative Brood Reduction in Eagles: An Examination of Factors that Influence Fratricide. The Auk, 100, Foley, C., Pettorelli, N. & Foley, L. (2008). Severe drought and calf survival in elephants. Biology Letters, 4, Forber, P. (2005). On the explanatory roles of natural selection. Biology and Philosophy, 20, Grant, P. R. & Grant, B. R. (1995). Predicting microevolutionary responses to directional selection on heritable variation. Evolution, 49(2), Lewens, T. (2001). Sex and Selection: A Reply to Matthen. British Journal for the Philosophy of Science, 52, Matthen, M. (1999). Evolution, Wisconsin Style: Selection and the Explanation of Individual Traits. British Journal for the Philosophy of Science, 50, Matthen, M. (2002). Origins Are Not Essences in Evolutionary Systematics. Canadian Journal of Philosophy, 32, Matthen, M. (2003). Is Sex Really Necessary? And other Questions for Lewens. British Journal for the Philosophy of Science, 54, Matthen, M. & Ariew, A. (2002). Two Ways of Thinking About Fitness and Natural Selection. Journal of Philosophy, 49(2), McComb, K., Moss, C., Durant, S. M., Baker, L. & Sayialel, S. (2001). Matriarchs as repositories of social knowledge in African Elephants. Science, 292, Nanay, B. (2005). Can cumulative selection explain adaptation? Philosophy of Science, 72, Neander, K. (1995). Pruning the Tree of Life. British Journal for the Philosophy of Science, 46, Pust, J. (2001). Natural Selection Explanation and Origin Essentialism. Canadian Journal of Philosophy, 31, Pust, J. (2004). Natural selection and the traits of individual organisms. Biology and Philosophy, 19,

15 Réale, D., Reader, S. M., Sol, D., McDougall, P. T. & Dingemanse, N. J. (2007). Integrating animal temperament with ecology and evolution. Biological Reviews of the Cambridge Philosophical Society, 82, Reznick, D. N., Shaw, F. H., Rodd, F. H. & Shaw, R. G. (1997). Evaluation of the Rate of Evolution in Natural Populations of Guppies (Poecilia reticulata). Science, 275, Sober, E. (1984). The Nature of Selection: Evolutionary Theory in Philosophical Focus. Cambridge, MA: MIT Press. Sober, E. (1995). Natural Selection and Distributive Explanation. British Journal for the Philosophy of Science, 46, Sober, E. (2000). Philosophy of Biology. Boulder etc.: Westview Press. Van Fraassen, B. (1980). The Scientific Image. Oxford: Clarendon Press. Walsh, D. (1998). The Scope of Selection: Sober and Neander on What Natural Selection Explains. Australasian Journal of Philosophy, 76, Walsh, D. (2007). The Pomp of Superfluous Causes: The Interpretation of Evolutionary Theory. Philosophy of Science, 74, Footnotes 1 The causal understanding of selection has recently come under attack (Matthen and Ariew, 2002; Walsh, 2007). 2 This is in line with Pust s (2001; 2004) use of the Negative View, but not with Neander s (1995), who uses this expression for the view that selection cannot explain how genetic plans for traits originated. 3 Selection explains, for example, why beaver populations are composed of flat-tailed individuals (rather than of rod-shaped ones): flat tails are an adaptation for, presumably, effective swimming. Selection might even be said to explain why species or populations have certain adaptations, if having adaptation X is understood in the derived sense of being composed of X-individuals (the European beaver has a flat tail in the sense that all current Castor fiber populations are composed of flat-tailed individuals). 4 Here is a definition of adaptation easily interpretable along these lines: Characteristic c is an adaptation for doing task t in a population if and only if members of the population now have c because, ancestrally, there was selection for having c and c conferred a fitness advantage because it performed task t. (Sober, 2000), p. 85. On a natural (though uncharitable) reading 15

16 of this definition, it is one of the conditions for trait c s being an adaptation that selection explains of individual population members why they have c. The reading is uncharitable because the phrase members of the population now have c because, ancestrally, there was selection for having c can, and should, be construed as stating that selection is responsible for the current population s being composed entirely of c-individuals (Sober, pers. comm.). 5 Consider Sober s (1984), p. 208, earlier definition: 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. According to this definition, selection causes traits to become prevalent in populations; it does not cause members of the population to have them. 6 Furthermore, origin essentialism has underpinned much of the discussion to date, and it is still an intuitive and plausible view despites its difficulties. In particular, it can withstand Matthen s (1999; 2002) objections (Pust, 2001; 2004). Moreover, it is far from obvious whether surrendering origin essentialism offers a less demanding or even metaphysically neutral (Matthen, 2003), p. 303, account of natural selection, especially if this involves commitment to entities such as receptacles. 7 This is the example Sober (1995) used to illustrate the Negative View. Figure 1a contains the first three generations from Sober s original figure, which comprises five generations and a total of seventeen individuals (his figure 1 on p. 386, Sober, 1995). Note that Sober uses the numerals as rigid designators. 8 An objection of this sort was raised by Phyllis McKay. 9 Although trait facts are individual-level facts, they may entail population-level facts. Suppose in a population of 100 individuals, members 56, 57, and 93 (and only they) had G, rather than B. It then follows that the population would have the property of consisting of 3 G-individuals and 97 B-individuals. Note that the reverse does not hold. It does not follow from population-level facts like population size and trait frequencies which members of the population have which traits (unless trait frequency is 100 %). 10 Traditionally, facts are the state of affairs that make propositions true. Accordingly, one might suggest that when a proposition is made true by just one individual it expresses an individual-level fact, whereas if its truthmaker involves several individuals, then it expresses a population-level fact. This works well with some propositions. For example, Lonesome George has survived to reproductive age is made true only by Lonesome George, whereas This population contains 70% females is made true by a group of individuals. However, the appeal 16

17 to truthmakers fails when relational properties are involved. For all we know, Jane s being taller than Jill is a fact concerning, at least, Jane. It is thus an individual-level fact. But the proposition is made true by both Jane and Jill, not by Jane alone. 11 This also helps to make sense of propositions like Boris belongs to population P, where belongs to is predicated of both an individual and a population. We have no qualms regarding this proposition as expressing a fact regarding Boris the individual (that he belongs to population P) and, at the same time, a fact concerning P the population (that P contains Boris). 12 This is a case of cumulative selection for G. Note that not even cumulative selection explains why a distant descendant of 50 has G. As Sober (1995), p , argued against Neander (1995), in each generation selection does not explain why the offspring has the trait is does; repeated rounds of selection do not change this. 13 Here I assume that the explanatory interest is in giving a fine-grained explanation of why a descendant population is composed of G-individuals. If the explanatory interest were different, it may well be explanatorily irrelevant that the G-ancestor was the 50 th member. It may suffice to establish that at least some ancestor or other had G. 14 Neander s arguments have been discussed by Sober (1995), Walsh (1998), Pust (2001), and Forber (2005). For replies to Matthen s work see Pust (2001; 2004) and Lewens (2001), as well as Matthen s (2003) response. 15 For reasons of simplicity I paraphrase Nanay s probabilistic counterfactuals in deterministic terms. 16 Apart from the contentious issue of the transitivity of counterfactuals, to which Nanay has a response, one may have doubts about the second premise. Nanay suggests that the scarcity of resources implies that individuals can become more successful in exploiting a resource only at the expense of other individuals. Hence the idea that x s mother could survive only if the death of a B-individual renders a resource accessible for x s mother (Nanay puts this in probabilistic terms). But not all competition over limited resources takes this form (Sober, 1984), p. 17. So the argument s second premise may not hold even if it is rendered probabilistically. 17 Walsh (1998) used a scope ambiguity in propositions like selection can explain why giraffes have long necks to make a similar point. 18 Note that the restated first premise is consistent with origin essentialism. According to origin essentialism, one and the same individual may have traits other than it actually has (Pust, 2004). 19 Perhaps it is meaningless to say that x lacks B when x not even exists. After all, for x to lack B seems to require that x instantiates the second-order property of lacking B, and x must exist in 17

18 order to instantiate that property. But the relevant counterfactual can easily be reformulated: if x s mother had died, x would exist such that x would have B instead of A. Again, this counterfactual is false. 20 Suppose that x s mother had B instead of A and that she survived (only one of the conditions is changed in order to assess counterfactual dependence). Since she survived, she would generate x and pass on trait B to her offspring. So under these circumstances, x would have B instead of A. 21 The mother s having A is very different in this respect. It is a factor on which x s having A depends, but x s having B does not. X s having B counterfactually depends on the mother s having B (not on her having A). Note that modifying the first premise to x s existence depends counterfactually on x s mother having A and (thus) being capable of reproduction does not change the result for the reason outlined in the main text. The conclusion would be that x s existence, but not x s having A, depends on selection for A. 18