Download 1 Review of Godfrey-‐Smith`s, Darwinian Populations and Natural

Review of Godfrey-­‐Smith’s, Darwinian Populations and Natural Selection Hardcover, $33.82, 224 pages Anya Plutynski, Associate Professor Department of Philosophy, University of Utah Correspondence: [email protected] Peter Godfrey-­‐Smith’s Darwinian Populations and Natural Selection is a detailed examination of the meaning of natural selection, and the scope of Darwinian explanations. The book explores the criteria for what may count as a “Darwinian population,” or, which collections of entities have the capacity to undergo evolution via natural selection. Godfrey-­‐Smith deploys these criteria in addressing: (a) the twin problems of reproduction and individuation of biological entities, (b) the persistent “gene’s eye view,” (c) the levels and units of selection problem, and (d) the evolution of cultural artifacts and behaviors. This review provides an overview of Godfrey-­‐Smith's framework for representing populations, a discussion of his critique of the "agential" view of evolution, and a critical assessment of his analysis of cultural evolution. Introduction Natural selection is an extremely powerful process – so powerful, in fact, that it is often tempting to deploy it in explaining phenomena as wide-­‐ranging as the persistence of blue eyes, the origins or persistence of religious belief, or, the history of science. One long-­‐standing debate among both critics and advocates of Darwin’s theory concerns the scope of Darwinian explanations, and how we are to draw the line. Peter Godfrey-­‐Smith’s Darwinian Populations and Natural Selection is a detailed examination of this question. The book explores the criteria for what may count as a “Darwinian population,” or, which collections of entities have the capacity to undergo evolution via natural selection (p. 6). Drawing upon his answer to this question, Godfrey-­‐Smith examines and provides his own view on the following long standing issues in philosophy of biology: (a) the twin problems of defining reproduction and individuation of biological entities, (b) the persistent “gene’s eye 1 view,” (c) the levels and units of selection problem, and (d) the possibility of the evolution of cultural artifacts and behaviors. Ever since Darwin, biologists have offered summaries of the “ingredients” necessary for natural selection: heritable variation, which leads to differences in reproductive output (or, fitness). Godfrey-­‐Smith starts with these “minimal” criteria, but argues that “classical” or “standard summaries do not cover all cases, and do not suffice to predict change” (p. 20). A careful discussion of exceptions to the rule in Chapter 2 sets the stage for the rest of the book. Godfrey-­‐Smith uses these three criteria as a first pass at a framework that may help us discriminate marginal from “paradigmatic” cases of Darwinian populations. There are, Godfrey-­‐Smith argues, “minimal” criteria that make a population of entities capable of evolving via natural selection; “A Darwinian population in the minimal sense is a collection of causally connected individual things in which there is variation in character, which leads to differences in reproductive output.”(p. 39) This summary sounds rather close to the formulation given above; but the devil is in the details. What makes a population of individuals causally connected? When and how must the entities in question vary and reproduce? What, for that matter, is “reproduction” and how do we discriminate cases of genuine reproduction from mere “growth”? We need to resolve these problems if we wish to demarcate “Darwinian populations”. Unpacking these details makes up the first five chapters of the book. Rather than review all of the discussion of these chapters, it may be helpful to first draw out some central themes. First, in his pivotal chapter 3, Godfrey-­‐Smith provides a framework for representing populations in the world that are “closer” to a “paradigmatic” Darwinian population, and those which are “further” away. He uses a multi-­‐dimensional spatial diagram with axes that represent qualitative measures of “closeness” to “paradigm” cases. Among these features are: “fidelity of heredity,” (H) “abundance of variation,” (V) “continuity, or smoothness of the fitness landscape,” (C) and “dependence of reproductive differences on intrinsic character” (S) (p. 63). The diagram may be used as a “heuristic device” for exploring both the relationship between these features in particular populations, and comparing different populations, over evolutionary time. Different populations will be located at different positions in this multidimensional space. The heuristic also frames Godfrey-­‐Smith’s solutions to questions that have plagued philosophers of biology for decades, such as the levels and units of selection problem. More details on this point will be explored in Section 1, below. Second, Godfrey-­‐Smith argues that various psychological habits of thought can, and in fact, have led us astray in characterizing Darwinian populations. The view, for instance, that selection requires “replicators,” as well as what Godfrey-­‐Smith calls 2 the “agential” view of evolution, according to which one should think of evolution “in terms of a contest between entities that have purposes, strategies, and agendas,”(p. 10) are central examples of what he takes to be psychologically powerful, but powerfully misleading characterizations of natural selection. More detail on this point will be provided in section 2, below. Third, Godfrey-­‐Smith uses vivid examples to emphasize how the biological world is, packed with “marginal cases, precursors, and not-­‐quites”(p. 108). To the great frustration of philosophers (and formal modelers), the world is not made up of spherical cows. While this will not be altogether surprising to most readers, one of Godfrey-­‐Smith’s central aims seems to be to resist formal solutions to many of the long-­‐standing problems in philosophy of biology. While he acknowledges the power that simple models of evolutionary patterns and process can offer, he resists deploying formal models in solutions to philosophical problems – problems to do with clarifying concepts, making distinctions, or identifying fault lines in debates. Thus, formal models are relegated in this book to the appendix, which happens to be a very useful compendium of the relevant theoretical work at stake in much of the debates about selection. Godfrey-­‐Smith does a wonderful job of showing why his view of Darwinian populations as ‘more’ and ‘less’ paradigmatic is to be expected, at least in part, because of the process of evolution itself. As he remarks in the opening chapter of the book, “When a population evolves… it also changes, as a consequence, how it evolves, the manner in which further change can come about.”(p. 8). That is, evolution itself can change the process of evolutionary change: what is changing, and how. “De-­‐Darwinization” is a useful way of characterizing how a population may, via the process of evolution itself, move further away from paradigmatic cases of evolutionary change. The “de-­‐Darwinizing” of lower level populations of entities can be affected by generational bottlenecks or segregation of the germ line from somatic cells. Such processes reduce competition at lower levels, such as between cells in multicellular organisms. De-­‐Darwinization explains, at least in part, why we don’t all have cancer (at least much more frequently than we do)! Godfrey-­‐Smith is a pragmatist, but he’s not a conventionalist; the truth is out there. There are things that natural selection can explain, but we need to be watchful about when and how. Godfrey-­‐Smith’s diagram gives us a way of “seeing” how natural selection covers phenomena much more wide-­‐ranging than the paradigmatic cases of evolution in multicellular eukaryotes. Slime molds, Volvox carteri, and champagne grapes make more frequent appearances here than Darwin’s finches. The picture that emerges is one of a dynamic, diverse, biological world, with no uniform mode of variation, heredity, or reproduction, and thus, no uniform measure of “reproductive success.” This complicates the problem of demarcating Darwinian populations, exactly as it should. However, Godfrey-­‐Smith, despite the blooming buzzing confusion, can and does rule some cases out. 3 In the following sections, I will first discuss Godfrey-­‐Smith’s spatial representation of more and less “paradigmatic” Darwinian populations, and some useful applications of this spatial representation, as well as some questions it raises. Second, I will discuss his criticisms of the “Gene’s eye view,” and briefly discuss his definition of “reproduction”, a central component of his argument. And, finally, I will discuss how his model helps us better understand cultural evolution. Part I: Spatial Representation To grasp what Godfrey-­‐Smith is up to in this book, one needs to have a very solid understanding of his spatial diagram. His spatial diagram is the way in which he provides a structural representation of the relationships between paradigmatic and marginal cases of Darwinian populations. His diagram will set the terms of debate and, in fact, help frame his solutions to disputes about levels and units of selection, as well as his take on cultural evolution. What’s important for our purposes is to grasp three key points: -­‐
First, as mentioned above, one’s “position” in the field of “more” and “less” paradigmatic cases of Darwinian populations. All populations in his field must satisfy “minimal” conditions, but these conditions are very “broad and permissive.”(p. 40) -­‐
Second, the axes of the diagram represent features of more and less paradigmatic Darwinian populations – but only “qualitatively”, and, while Godfrey-­‐Smith only pictures a three-­‐dimensional space, the space is intended to include multiple dimensions. -­‐
Third, and finally, the position of a population in the field, or, relative value of different features of a population (“low S” or “high C”) will allow us to predict both how frequently such a population is likely to result in novelty (or, Darwinian processes that figure in “origins” as opposed to mere “distribution” explanations), the relative roles and significance of drift, and even developmental features of the entities in the population, such as extent of modularity. So, what are the axes of this diagram that will have such magical powers? First, fidelity of heredity is one factor that determines how “paradigmatic” Darwinian a population is. In order for change to be cumulative, and thus, for it to be possible for complex adaptive structures to arise, one needs fairly reliable processes of inheritance. However, what makes for a “reliable process of inheritance” can be fairly open-­‐ended. Godfrey-­‐Smith remarks that (a) heredity need not require genes; (b) replication is not necessary for heredity. In other words, as long as there is “heritability” and this has some degree of “fidelity,” Godfrey-­‐Smith argues, selection can go forward, whatever the mechanistic basis, and whether or not “replicators” are the engine of heredity. 4 This point is worth emphasizing, especially as it pushes against a strong tradition in biology of emphasizing the gene as the exclusive means of inheritance. That an entity is "replicated" and "persists" over many generations is neither necessary nor sufficient for something to be of evolutionary interest, according to Godfrey-­‐Smith. We are familiar with the paradigm case of genes, but that some thing is “replicated” is, of course, not necessary for resemblance between parents and offspring. There are a variety of mechanisms of inheritance apart from genes – various forms of epigenetic inheritance jump to mind, but social learning may be a highly reliable means of transmitting information from one generation to the next. All that matters, really, is heritability: the resemblance of parents and offspring, through whatever means you like. Godfrey-­‐Smith grants that genes are important, but his point, recall, is to both find a way of demarcating “paradigmatic” from “marginal” cases, and how to rule some cases out as not meeting even “minimal” conditions. That “heritability” has a much broader meaning than is often supposed by fans of the “gene’s eye view” will turn out to have important implications for whether and under what circumstances “culture” evolves. The second axis is variation; evolution by natural selection, as is well known, requires that there be sufficient variation. However, according to Godfrey-­‐Smith, mere abundance of variation is not sufficient for significant adaptive change. The variation should be slight in extent, (not involving “huge jumps in the space of phenotypic variation,”) not too “biased” in any direction, and ideally, traits should be “quasi-­‐independent.” In a famous paper on adapation in 1985, Richard Lewontin argued that one feature of an organism that leads to greater capacity to evolve complex adaptations is that traits be not too tightly correlated or coupled to one another. “Quasi-­‐independence” is supposed to make evolution of complex adaptations more likely, because it is easier to “tinker” with and “fine-­‐tune” a trait if change in this trait can occur independently of change in others. Godfrey-­‐Smith uses a single qualitative measure, “V” to describe “abundance of variation” (p. 47). One wonders, however, whether “abundance” tracks one thing or several – additive genetic variance, variance slight in extent, “unbiased” variation, or, “quasi-­‐independent” variation? Moreover, might there be trade-­‐offs between these different features of variation? Are abundance, lack of bias, and quasi-­‐
independence correlated? If not, should these be separate “axes” on the spatial representation of Darwinian populations? Perhaps this is an empirical matter and should be left to the biologists; but it would have been interesting to explore in somewhat more detail the mechanistic bases of different aspects of variation, or its “abundance”, and how they are related. A third axis along which populations might tend toward more “paradigmatic” and less “paradigmatic” is extent of causal interaction in a population. Godfrey-­‐Smith borrows the concept of α from ecology (roughly, a measure of the extent to which two populations grow in a density-­‐dependent way), to represent the “overall 5 measure of reproductive competition…” or, more precisely, “the extent to which adding reproductive success to one individual reduces another’s.”(p. 52) This fairly broad sense of competition captures at least one important way in which individuals in a putative Darwinian population may be causally connected, such that both sexual and asexual populations, single or even multiple species could count as a ‘population’ using this measure. Godfrey-­‐Smith further argues that competition can yield novelty, when the “winning” variant’s absolute numbers increases backgrounds against which new variation may arise. However, competition could have the very opposite effect – a “fight to the death” that reduces absolute numbers of both variants, thus reducing the variety of backgrounds against which a new variant might arise (p. 51). Thus, it seems that degree to which a population is glued by competition can be both a driver of evolutionary novelty, and a detriment. Here is one place where the axis does not seem to measure the direction of “Darwinization” in a straightforward manner. A further axis is “S”, “the extent to which differences in reproductive output in a population depend on intrinsic features of the members of the population.” (p. 53) This appeal to “intrinsic” features is supposed to allow us to distinguish change in reproductive output due to features of the organism or entity in question – its fitness – and changes due to extrinsic features – which may include anything from location, to being someone’s cousin. In other words, “extrinsic” causes are somehow located “outside” the organism. A population whose trajectory of change is driven by “S” is more “paradigmatically” Darwinian. Godfrey-­‐Smith explains, “S is the extent to which “realized” fitness differences in a population are tied to differences in intrinsic character”(p. 54). One odd result of this is that populations more subject to extrinsic, chancy, and “drifty” phenomena such as lightning strikes turn out to be “less paradigmatically Darwinian.” On it’s face, it seems odd to say that simply being subject to drift makes a population less “Darwinian.” However, recall that Godfrey-­‐Smith’s aim is to identify populations as more or less likely to evolve via natural selection, not take a stance on the relative significance of drift v. selection in the evolution of adaptive novelty. Or, is it? On p. 40, Godfrey-­‐Smith more or less equates “being paradigmatically Darwinian” to being more “evolvable”; where, evolvability is supposed to represent how likely a population is to give rise to “significant novelty,” or “complex and adapted structures.” Usually, those who speak of “evolvability” have in mind a complex of traits (modularity, robustness) that are supposed to yield greater “capacity to generate adaptive novelty” or capacity to generate “complex adaptation.” Godfrey-­‐Smith seems to suggest that being more and less “Darwinian” in the narrow sense of “subject to natural selection” and being more or less “evolvable” correspond. 6 However, it seems that drift could play a very significant role in the evolution of adaptive novelty. Some have argued that population bottlenecks have been important in the generating adaptive novelty, in generating, say, gene duplication events which put selection pressure on “redundancy” at the genomic (or molecular developmental) level. Indeed, drift (arguably) leads to many interesting long-­‐term evolutionary consequences. This leads to the more general question: what does Godfrey-­‐Smith mean, exactly, by “paradigmatic”? Does he have in mind “of historical importance to the history of life,” or, to the history of Darwinism, or both? Why is “Darwinian” evolution taken to be more or less equivalent to “evolution by natural selection”? One is reminded here of debates among biologists about who the “real” Darwinian is, but might have been worth addressing, briefly, why he chose this particular term, if only as a matter of history or sociology of biology. However, to return to the argument about “S”, Godfrey-­‐Smith makes an interesting case for his view of S as describing “intrinsic” features by discussing the case of cells in a multi-­‐cellular organism. The cells in our bodies, while making up a population (in the “minimal” sense), are less “Darwinian” than whole organisms competing for survival. This is so because whether or not a cell might have long-­‐term downstream evolutionary consequences has to do more with accidents of where they end up (in the germ line or soma), than with their “intrinsic” character. Even when a cell acquires a mutation that permits it to divide rapidly, and take over its neighboring cells, (as in cancer), such a cell is, strictly speaking, an evolutionary dead-­‐end. Genetically identical cells in the human body may have very different evolutionary consequences due to accidental features such as location (cells in some tissues are more likely to proliferate as cancer than others), so, this is a case of a “low-­‐S” or “marginal” Darwinian population. Though one could ask if, at least in some cases, location is a reflection of “intrinsic” properties – as when organisms are vying for position in a space of limited resources. While cancer is a clear-­‐cut case of “marginal” Darwinian evolution, it does seem odd to suggest that populations of entities whose fate (in general) more subject to chance extrinsic factors are less “Darwinian”. Surely drift may factor into the evolution of novelty, on a case-­‐by-­‐case basis, and over shorter versus longer time-­‐
spans? And, even if individual cells are less subject to fate, it seems that if a whole population is subject to chance events, this can, over the long term, have interesting evolutionary consequences (to do with adaptive novelties), due to extrinsic factors, like whether an adaptive niche opens up. This seems to be a case where being “Darwinian” in the narrow sense of subject to selection pulls apart from the broader sense of being more “evolvable.” Lynch has argued (2007), for instance, that drift has played an important role in the evolution of adaptive novelties in multicellular eukaryotes. His passive emergence thesis is that drift in multicellular eukaryotes (due to frequent population bottlenecks) resulted in “intrinsically deleterious genome-­‐level changes…” (duplications, etc.), that “impose[d] selection pressure for cellular defense 7 mechanisms” such as various forms of developmental buffering – robustness or plasticity. In other words, drift yields a “problem” at the genomic level that creates selection pressure at the level of development. Whatever the merits of this hypothesis, it may have been interesting for Godfrey-­‐Smith to examine in some more detail when and if being subject to chance “extrinsic” events is always, or only sometimes, a way of compromising a population’s evolutionary potential. The final feature he considers is “C” – this is a measure of “continuity” or “the degree to which small changes in an organism’s phenotype lead to small changes in its fitness.”(p. 57) We ran across the compatriot of this concept in the discussion of variation, above. Lewontin argued that complex adaptation is more likely to evolve in a population where traits are ‘quasi-­‐independent’, but he also offered a second condition: change in phenotype should affect fitness proportionally. Godfrey-­‐Smith deploys this idea here, and appeals to Wright’s concept of a “fitness landscape” to picture the relationship between variation and fitness. A “smooth” landscape is one where similar phenotypes are associated with similar fitnesses; a “rugged” landscape is one where similar phenotypes are associated with very different fitness values. Godfrey-­‐Smith argues that “smoother” landscapes are more “Darwinian” in character, presumably because it is easier to traverse them via selection. One temptation, on first reading this section, would be to assimilate smoothness of the landscape with extent of epistasis for fitness. Greater epistasis for fitness (less additive variance in fitness) can compromise the extent to which selection may act on a population. However, according to Godfrey-­‐Smith (pers. comm.) this temptation should be resisted. What makes a landscape more or less smooth could have to do with genetics, but it could also have to do with development, or, with the environment in which a population finds itself. Indeed, Godfrey-­‐Smith suggests that C describes the relation “between (realized) fitness and everything about an organism – internal structure, place of birth, history of movement from one location to another, who it interacts with, and so on.”(p. 60). This particular suggestion may induce puzzlement at first. On the one hand, “C” is a measure of the relationship between change in phenotype and change in fitness. Godfrey-­‐Smith’s argument is that a phenotype is a composite of all the factors that “feed into” an entity’s fitness – and this might include place of birth, or history of movement, if we wish to be inclusive. Godfrey-­‐Smith has hit upon an important point – one that has been more or less passed over in much of the literature on evolvability. Many of the advocates of evolvability speak of it as if it is some intrinsic (and particularly, developmental or molecular) property of the organisms or class of organisms in question. Multicellular eukaryotes are supposed to be more “evolvable” because they are more modular, or have features that are more “quasi-­‐independent”. Population geneticists counter that the exclusive focus on development fails to recognize how features of the genetics of populations are relevant to the effective response to selection. With this concept of continuity, Godfrey-­‐Smith puts evolvability squarely back into its larger context; whether and 8 how quickly an organism evolves complex adaptations is determined not merely by population genetics (additive variance in fitness), or even development, but also ecology. The above are only four of many potentially relevant features of a population. Where it all comes together is when Godfrey-­‐Smith deploys his axes to mark off more and less paradigmatic Darwinian populations. It’s worth exploring two consequences of his spatial diagram where one might expect some dispute among various philosophers; first, concerning drift, and second, concerning levels of selection. First, Godfrey-­‐Smith argues that the relative significance of drift in a population is something that emerges out of picturing populations along axes of “fitness” and “continuity” (S and C), or, it is, he says, a “bonus” of his representation of the space of Darwinian populations. In other words, where there is BOTH low C and low S, this “looks most like drift,” but he is careful to say that drift cannot “be explained in terms of S and C.”(p. 61) This is a bit puzzling, and deserves some unpacking. Godfrey-­‐Smith resists both the claim that drift is “force”, and the claim that it is a measure of our lack of knowledge of the causal details leading to the apparently “chance” deaths: “drift is not a “force” in the biological world, but it is not a mere reflection of our ignorance either.”(p. 62) Drift, in Godfrey-­‐Smith’s view, is not a “distinct factor” from selection, where change is due to these two “forces” acting independently, and, (as is often said) in opposing directions. Of course, he grants that in smaller populations, chance may be expected to play more of a role in change in distribution gene frequencies over time than, say, in larger populations. This simple function of population size as a clear predictor is one argument for drift as a cause, or force, acting in “opposition” to selection. However, while population size is a good operational measure or means of predicting “drift”, it is not all that we need to discriminate a genuine case of drift from selection. One way of putting this point is to say that while population size is a great “lever” for drift, it’s not an explanation of drift. Godfrey-­‐Smith argues that what makes a purported case “look” more or less “like” drift is a product of at least two factors: whether and to what extent change in a population is due to some “intrinsic” fitness difference, as well as the extent to which small changes in phenotype make a difference to fitness (“ruggedness” of landscape). In other words, his way of addressing drift is, like much else in the book, a matter of discriminating among more and less “paradigmatic” cases. This particular assessment has much to recommend it, but will be puzzling for many biological readers. It is not so puzzling that S and drift will be inversely related; when differences in reproductive success are not due to intrinsic factors, we have a case where whatever is causing change in a population is not selection, but something else (usually assumed to be drift!). It’s less immediately clear how and why low “C” and drift are related. Godfrey-­‐Smith offers the following rationale: “Organisms can have their reproductive output affected by internal accident, as well 9 as through external, lightning-­‐like events. So, let us introduce a role for C as well.”(p. 60-­‐1) But, are internal accidents contributors to “drift”? If an “internal” accident does not yield long-­‐term evolutionary consequences, how is it a contributor to drift? Godfrey-­‐Smith defines C as “when small changes in an organism’s phenotype lead to small changes in fitness.”(p. 56). But, it seems that small changes in phenotype might have both small and large effects on fitness in a population subject to a great deal of drift, or none at all. Why is a ‘low C’ population more “drifty”? This definition of C, “when small changes in an organism’s phenotype lead to small changes in its fitness” (on p. 57) seems, on its face, rather different from the definition Godfrey-­‐Smith offers a on p. 61, “the relation between realized fitness and everything about an organism – internal structure, place of birth, history of movement from one location to another, who it interacts with.” However, Godfrey-­‐
Smith’s aim seems to be to point out that continuity of a landscape is a product not only of an organisms’ internal properties, but also to behavioral and historical factors. An organism’s “phenotype” in this sense, is more broad and inclusive -­‐ “everything” about an organism – so that when changes happen in a population that cannot be predicted on the basis of any and all such features, we have a phenomena that “looks like drift.” That is, only on this much more inclusive sense of “phenotype” does it seem that C is importantly relevant to drift. A further application or consequence of Godfrey-­‐Smith’s spatial diagram comes much later, in Chapter 6. There is much to recommend this chapter (and particularly the illuminating discussion of transitions in evolution). But, one conclusion that deserves particular attention is the matter of when we ought to take something to count as a case of group selection. Godfrey-­‐Smith begins by reviewing some of the debate about group selection, and offers an exceptionally clear analysis of much of the literature on the evolution of altruism. He points out that what matters to the evolution of altruism is “correlated interaction” – interaction between individuals sharing some behavior. Moreover, this correlation need not be due to shared genetics. Surely, preferential interaction among kin sharing altruistic behaviors is one way for altruism to evolve. However, it is not the only way: “a behavior flourishes in an evolutionary context if somehow it differentially generates fitness benefits for those with a tendency to transmit the behavior to their offspring.”(p. 121) In other words, that those benefits go to the actor, his or her relatives, is only two of three possible ways for behaviors to evolve. Actors only need to display this behavior preferentially towards those who have a tendency to pass on the behavior when they reproduce. This point nicely illustrates how the “long-­‐term beneficiary” model of the evolution of altruism misses the point. It’s not necessary that there is a “gene” that “benefits” from this process; what’s necessary is only that (whatever the hereditary basis) there are degrees of altruism in a population, and the more altruistic tend to display their behavior toward types that will pass on similar behaviors. 10 Suppose we have a collection of groups of individuals that are themselves reproducing and competing; or, a case of potentially “multi-­‐level” selection. When shall we say that the ‘higher level’ group constitutes a Darwinian population? Godfrey-­‐Smith falls on the conservative side here. That is, he takes Wilson’s “trait-­‐
group” models case, or, “MLS1” cases, where groups are ephemeral, and do not themselves reproduce, to be at best “marginal” cases of group selection. Since the fitness of the group is not measured in terms of its production of “offspring groups” but only as a sum of the fitnesses of individual members, he argues that this is not a “paradigmatic” case of group selection. In other words, for Godfrey-­‐Smith, group selection requires that groups not only vary, but also reproduce as groups, and inherit features from other groups. Thus, he parts ways with Sober and Wilson, among others. Some may object that Godfrey-­‐Smith’s strategy here seems to be to use his diagram to dodge the problem. Why can’t he just take a side? Why is it a matter of “marginal” versus “paradigmatic”? Others may object that his spatial model lacks precision; there is no universal “scale” of comparison along which we can meaningfully represent different instances of “S” or “C.” Moreover, S and C seem to not have a clean, orthogonal relationship – it’s hard to picture how we would position various actual populations relative to one another in such a space. It would have been nice if Godfrey-­‐Smith had discussed some more examples in this context, but of course, Godfrey-­‐Smith can simply point out that these are qualitative, not quantitative measures. Godfrey-­‐Smith’s aim is to articulate the minimum conditions on what may evolve, and then describe (loosely) the variety of continua along which populations may reside. The point is that there is no one feature of populations that is “essential” or that we should “foreground” in considering the problem of evolution by natural selection. This leads us naturally to his discussion of the gene’s eye view. Part II: Gene’s Eye View Chapter 7 is Godfrey-­‐Smith's critique of the "gene's eye view". He begins by carefully distinguishing three different versions of the 'gene's eye view': (a) that in special cases, genes are the units of selection, (b) that all cases of biological evolution can be represented from the gene's eye perspective, and (c) that selection acts all and only on replicators (genes), and it is a mistake to take anything else to be the unit or target of selection. He grants a variety of special cases where genes are indeed the target or unit of selection -­‐ that is, where gene-­‐level description of selection processes is appropriate -­‐ since what is being selected for are genes that are active in their own reproduction, and where their fitness is indeed a property of the genes themselves. These include "selfish" genetic elements, such as transposons, or cases of "meiotic drive" -­‐ these are genes that manage to spread to different places in the genome, using various elements of the chromosome's own replication machinery. Here are clear cases 11 where "genetic success" is not tracking reproductive success of actual organisms, but refers to the gene's own success at reproduction. So, while in a few, select cases, it does make sense to treat genes as Darwinian populations, Godfrey-­‐Smith is less persuaded by the second two versions of a "gene's eye view" -­‐ for at least three reasons. First, representing populations of organisms' evolutionary dynamics from the gene's eye perspective is simply a choice about what features of the process to 'foreground' and which to 'background'. There is no principled reason to prefer the representation that foregrounds genes. Of course, gene action is often causally responsible for patterns of inheritance in organismic populations. However, it is the organisms that are reproducing. That is, it is the relative success of whole "collections" of actual genes that happen to be packaged in discrete organisms that we are "tracking" with a "gene's eye view." In other words, there is a bit of sleight of hand at play in the claim that all evolutionary processes can be "represented" from a gene's eye view; since what we care about is often the success of whole organisms, we "count" copies of genes in terms of organismic success, as opposed to, say, the physical copies of the genes themselves. Our "book-­‐keeping" is always through the lens of whole organisms' success, even when ostensibly counting "genes." Godfrey-­‐Smith's second reason for resisting (b) and (c) is that genes are not such discrete, independent, persistent objects as the idealized picture offered up by the "gene's eye view" would have us believe. Genes are, after all, rather fuzzy natural kinds; chromosomes and nucleotides are "naturally" bounded, but genes are not. Estimates of the number of genes in the chromosome of various organisms are plagued by problems of how we determine which of potential elements to count. Should we only count cistrons -­‐ elements responsible for specific proteins -­‐ or, include regulatory elements as well? Do we take into account that crossing over in meiosis does not respect cistron boundaries? Eukaryotic chromosomes are largely composed of non-­‐coding regions -­‐ if we've not (yet) recognized a function of such regions, should we simply ignore them? Advocates of the gene's eye view will say that the size of a gene in the evolutionary sense depends upon the particular rate of crossing-­‐over; if there are more crossing-­‐
over events, there are more genes. That is, the smallest unit that is passed on without being broken up without crossing over would seem to be the likely candidate, since what we are supposed to care about is what ‘persists’. Godfrey-­‐
Smith thinks persistence is irrelevant; what matters to whether a population counts as Darwinian is change over time, due to heritable variation in fitness; natural selection can act over the short or the long term. There’s no rule that says selection over the short term is uninteresting or unimportant. I suspect that many evolutionary biologists will be dubious about this – surely evolutionary biologists do care about the long term, which is why genes have been such important and interesting objects. In response to this, I suppose Godfrey-­‐Smith would again raise the objection that genes don’t come in discrete packages with labels – what counts 12 as a “gene” is a difficult matter to determine, absent context. Do we care about gene products – and which gene products? Why these gene products and not others? Williams suggested, as a sort of solution to this dilemma, that the strength of selection will determine which units count as genes; "selection bias equal or several times the rate of endogenous change" in an allele will determine the unit in question. But, Godfrey-­‐Smith points out that this means of distinguishing evolutionary genes causes all that is solid to melt into air; since selection differentials change over time, "genes will fade in and out of existence." In sum, "The question, "how many evolutionary gene tokens in humans?" is turning out to be unanswerable -­‐ not because we do not know the facts well enough, but because there is no definite number to learn."(p. 138). While Godfrey-­‐Smith grants that while using genes for "book-­‐keeping" is a pragmatic tool -­‐ one that may be deployed for specific purposes in specific contexts -­‐ if we are concerned to characterize the "real entities that undergo the kind of change Darwin described", then genetic book-­‐keeping will simply not do the job. Third and finally, Godfrey-­‐Smith points out that the very feature of chromosomes that permits a gene's eye view -­‐ crossing-­‐over -­‐ is an evolved feature of organisms. So, treating this as "the" measure of evolutionary units seems a particularly presentist bias. If most of the history of life consisted of entities whose form of reproduction was rather unlike our own, particular, sexual form, then taking meiosis and crossing over to measure the true "units" of evolution seems odd. We would, in effect, have to rule out all sorts of other forms of shuffling and exchange of hereditary material -­‐ from bacterial conjugation to RNA world scenarios. So, reproduction is not merely a matter of passing on “genes.” In chapter 5, Godfrey-­‐
Smith distinguishes different modes of reproduction. He first distinguishes three broad families of entities that may reproduce – collectives, simple reproducers, and “scaffolded” reproducers. “Collectives” are simply reproducing entities that have parts, which may themselves reproduce; this includes herds of buffalo, and all multicellular organisms. “Simple” reproducers are “lowest levels entities that reproduce largely under their own steam – using their own machinery, in conjunction with external sources of energy and raw material.” This category includes bacterial cells. Finally, “scaffolded” reproducers “get reproduced as part of the reproduction of some larger unit.” Thus, viruses and chromosomes are “scaffolded” reproducers. One of the issues that may confuse readers is what we may and may not count as “scaffolded” reproducers. Godfrey-­‐Smith makes clear that many cases in the biological world fall between these categories; the biological world is, as we know well by this point in the book, recalcitrant to precise distinctions. Nonetheless, this set of categories of reproduction permits us to rule out some cases as reproduction, and rule in others as genuine cases, according to Godfrey-­‐Smith. 13 For something to count as reproduction, he argues, there needs to be a causal relationship of some sort between “parent” and “offspring”, where this causal relation can be a matter of either material contribution or determination of structure or form, or both. Thus, on this account, retroviruses, prions, and LINE transposons may reproduce, but only in the “formal” sense that they contribute to the form of the “offspring”; there is a “chain of material influence” but no material “overlap” in these marginal cases. Godfrey-­‐Smith uses his three “families” of reproducers to usefully distinguish reproducers from the idea of “replicators” familiar from Maynard Smith and Szathmary (1995). He argues that their definition of “replicator” is too broad, permitting enzymes to count, insofar as they “can arise only if there is a preexisting structure of the same kind in the vicinity.” Since enzymes do not have “parent” enzymes, in Godfrey-­‐Smith’s view, they cannot count as reproducers. On its face, this makes perfect sense – for there to be heritability, there needs to be entities (parents-­‐offspring) that resemble one another, to a greater or lesser extent. But, in the case of some “scaffolded” reproducers, it may be a challenge to identify “parents” in a non-­‐question begging way. Godfrey-­‐Smith argues that prions and retroviruses are scaffolded reproducers, but enzymes are not. Yet, both kinds of entities reproduce in ways that require “pre-­‐existing structures of some kind in the vicinity,” to “scaffold” their reproduction. The second part of the chapter is a particularly illuminating discussion of the features of collectives that make them more and less likely to evolve as a Darwinian population. Bottlenecks, germ-­‐soma distinctions, and relative “integration” are, in Godfrey-­‐Smith’s view, features that make a collective more or less likely to possess relevant features for evolution. The idea of “de-­‐Darwinization” serves as a helpful way to picture, evolutionarily, the relative significance of bottlenecks and germ-­‐
soma distinctions to the evolution of a population via natural selection. Part III: Cultural Evolution Does culture evolve? This is the subject of the final chapter of Godfrey-­‐Smith's book, and is, it seems, one of the main motivators for the expansive view of Darwinian populations he is defending. The idea of cultural "evolution" is, of course, historically quite an old one -­‐ Spencer and other 19th Century thinkers predating and post-­‐dating Darwin imagined that there was some sense in which cultures (or, societies, forms of government, or even religious, moral or cultural norms) might be in competition, leading to the eventual winnowing of less successful forms. Two long-­‐standing problems in assessing such claims concern what the "unit" is that evolves, and how its "fitness" is measured. Godfrey-­‐Smith takes a characteristically middle-­‐ground view on the cultural evolution debate: "Darwinism is not likely to unify and transform the social sciences, in the way enthusiasts have claimed. But culture has Darwinian roots, and generates additional Darwinian phenomena when 14 circumstances cooperate." (p. 148) So, when do circumstances cooperate? Godfrey-­‐Smith begins by briefly distinguishing two senses of cultural evolution, what I will call, for the sake of clarity, CE1 and CE2. One form he characterizes as biological evolution, with a "special set of inheritance mechanisms." In CE1, the entities that make up a Darwinian population are biological individuals, where the phenotypes in question are cultural traits characteristic of individuals in the population -­‐ i.e., individual phenotypes (skills, habits, vocabularies). In CE2, cultural variants themselves -­‐ behaviors, ideas, artifacts, or words, might engage in “reproduction” in some sense. Godfrey-­‐Smith explains how both senses of 'cultural evolution' so described might evolve at different levels. In the first sense, either individuals may be more or less fit, in virtue of their possession of particular cultural phenotypes, or groups may be more or less fit, in virtue of the greater proportion of them possessing such a phenotype. In the second sense, the cultural variants themselves could count as "individual level" habits, or group habits. It seems clear that both senses of evolution depart from "paradigmatic" Darwinian populations in several senses. First, it's clear that the inheritance of cultural traits is, at best, a messy business. Second, when we begin to speak of groups of individuals sharing a cultural trait, it's less clear when this is a case of ordinary biological evolution with odd inheritance mechanisms, and when the first category of cultural evolution begins to blend into the second. In other words, it becomes rather difficult to discern where the "individual" is that is "reproducing." Needless to say, it also becomes rather difficult to determine relative "fitness." Godfrey-­‐Smith, to his credit, is rather upfront about this problem. He does a wonderful job of explaining how even some biological phenomena (a hemoglobin molecule, for example), are too tightly integrated to be seen as a population of entities -­‐ the atoms in the molecule. Thus, "integration" is key to demarcating individuals -­‐ where integration is measured (loosely) by considering the extent to which parts are structured, heterogeneous, and more or less capable of autonomous activity. We standardly treat organisms as individuals (humans are clearly our commonplace example), even though individual organisms are themselves often difficult to individuate and "isolate" from their environment; slime molds, lichens, and strongly symbiotic species, such as some jellyfish and their luminescent bacteria, to name a few, are hard to classify as singletons. Likewise, Godfrey-­‐Smith remarks on the many ways that cultural phenomena can be more or less "populational in character" -­‐ that is, some cultural traits are direct products of individual human activities, and some are population level properties or products. Moreover, the same practice or artifact may change its status; some products of community-­‐level activity may subsequently become ways of dissolving community into hierarchies with asymmetric roles. 15 Godfrey-­‐Smith is aware of the resistance to the idea that "cultural change is a Darwinian process," for the reasons one might expect. Historicist conceptions of social or cultural change -­‐ where some endpoint is viewed as inevitable or ideal -­‐ have led to some of the most horrifying totalitarian states in history. Needless to say, Godfrey-­‐Smith is not in sympathy with this family of views. In keeping with the pragmatic flavor of the book, Godfrey-­‐Smith is using the idea of treating cultural traits as capable of evolving in some sense as an interesting heuristic -­‐ one way of viewing the problem of culture, certainly not the only way, and certainly departing in interesting and informative ways from evolution in "paradigmatic" Darwinian populations. So, one of the most helpful things that Godfrey-­‐Smith does in the book is classify forms and varieties of cultural evolution, based on the variety of entities that may evolve (cultural phenotype, v. the actual cultural variant itself) and forms of “inheritance” (“vertical,” “horizontal,” “oblique”). There are some clear challenges with spelling out exactly how and whether a cultural trait is “heritable” – surely, social learning or imitation is one way in which such traits are “passed on.” When do we have “group level reproduction” under CE1? Here, Godfrey-­‐Smith deploys the same argument he gave in the chapter on levels of selection. Groups that grow and persist (do not go extinct) in virtue of their possession of some group level trait (such as division of labor or other forms of social organization) are not genuinely evolving via selection. Rather, genuine “group level” selection requires differential reproduction – the production, presumably, of more groups of the same kind, not simply persistence. How do we demarcate cases of reproduction under CE2 – how do cultural traits themselves reproduce? Here, Godfrey-­‐Smith draws a useful analogy with what he earlier termed “formal reproduction” – or, reproduction by prions, retroviruses, or LINE transposons, which use “scaffolding” and involve no or very little “material overlap.” In just the way that retroviruses may not reproduce without invading their host and deploying the tools there made available, so too, cultural traits may not “reproduce themselves” without the assistance of persons who can use turntables, type on computers, or sport just the right hat at church. Reproduction here depends importantly on adopters picking up habits, but the “causal responsibility” for the unit’s replication is “mixed”, or, if you like, distributed over both the church ladies and the colorful Easter hats themselves. Easter hats need church ladies with a special flair; in Godfrey-­‐Smith’s language, “adopting the trait” is a matter of the disposition of the adoptee as well as the cultural variant itself. Here, though, the issue of what may count as “parents” and what “offspring” seems again a deeply difficult question to answer. Fitness seems an especially troubling problem under CE2. This is where the examples seem, to say the least, to pull away from the paradigm. In what sense is a “cultural trait” more fit? How shall we measure its fitness (or the fitness of its bearers?)? Is fitness simply a matter of more instances of the variant “spreading”? 16 What shall we say about the (not infrequent) modification of both ideas and artifacts? Godfrey-­‐Smith has several helpful suggestions. Drawing upon Grisemer’s discussion of biological reproduction, he explains how, for “scaffolded” reproducers, (those which depend on outside mechanisms for their reproduction), reproductive success depends importantly on the surrounding conditions. In other words, “fitness” in this case is strongly “extrinsic” – what Godfrey-­‐Smith treats as a marginal case. For artifacts, he argues, “recurrence and success-­‐guided proliferation of some tool is not enough for a Darwinian process. What is required is that each instance of the artifact have a small number of “parent” instances, so that asexual lineages or family trees are formed, with a pattern of heredity across links.”(p. 155) This insistence on Godfrey-­‐Smith’s part seemed uncharacteristic, but he attempts to clarify with an example: the building of Polynesian canoes. Godfrey-­‐Smith explains that what is required for a population of such an artifact to evolve is not simply the proliferation of boats, but the “relevant parent-­‐offspring relation” – i.e., some causal relationship between “parent” boats (including multiple parents) and “offspring” boats. He even grants that this causal relationship can involve appropriation from multiple sources (“this mast is copied from this boat; this rudder-­‐blade from that one.”) Yet, he concludes, “once general intelligence intervenes in such a way that a vague and disparate set of models all make blended and customized contributions to the new boat, net, or hut, the Darwinian pattern is lost.”(p. 155) This insistence seems a bit odd, coming directly after his very permissive account of “scaffolded” “formal” reproduction; it seems, to say the least, a very difficult demarcation to make. When has “general intelligence” intervened too much? What counts as mere appropriation from multiple sources, versus true “inheritance,” and how shall we draw the line between this and “blended and customized contributions”? Here the reader runs ashore on a sandy bank, a rather shifting shoreline between the firm ground of Darwinian cultural evolution and the sea of novel intellect-­‐directed change. There is, according to Godfrey-­‐Smith, a line to be drawn between genuine cases of artifact or behavioral evolution and cases where “people combine too many sources of information and manipulate that information too intelligently,” (p. 160) but this line is, needless to say, difficult to draw. The last half of Godfrey-­‐Smith’s chapter on cultural evolution offers a helpful (rather psychoanalytic) discussion and critique of two opposing schools of thought on the matter of how behaviors evolve. On the one hand, Skyrms’s model of behavioral updating is, clearly, an idealized description of moral learning – we do not simply imitate our successful peers in real cultural interactions, or use algorithmic rules of updating in decision, but live in complicated social hierarchies with asymmetrical social roles. Nonetheless, Skyrms’s model is, Godfrey-­‐Smith thinks, a useful way for picturing behavioral evolution, and, it turns out, cultural evolution in both the first (CE1), and the second of two senses, (CE2), described above. 17 Under Skyrms’s game-­‐theoretical model, different payoffs are assigned to different behaviors in an idealized population – in this case, a social network, where one has associations with only near neighbors, faced with the same set of limited behavioral choices. Skyrms was able to demonstrate that under different versions of this model, cooperative behavior was a relatively easy outcome to generate, a result which has been used to argue that “the moral sense” is “evolvable.” Godfrey-­‐Smith is less interested in the interpretation of these models than in the structure of the models themselves; that is, he is interested in whether and in what sense Skryms’s model counts as a case of cultural evolution, and how close, or far, this case is from the paradigmatic Darwinian ones. Clearly, in this case, behavior is heritable in the “horizontal” rather than “vertical” sense; that is, behaviors are “inherited” via copying in a social network. (Though, of course it’s possible that parents may pass behavioral dispositions to their biological offspring either genetically or otherwise (vertical inheritance)). Further, such behaviors have fitness consequences, and more successful rules of thumb might form “lineages” and thus family trees, with one or another rule of behavior proving to be more or less successful. Alternatively, one could “combine” rules – via a “smart” imitation rule, such as acting on various rules that aggregate local responses of an individual’s neighbors in the previous time step. This is a “sexual” rather than “asexual” form of “reproduction”; insofar as the individual is combining information from past interactions. Godfrey-­‐Smith nicely sums up all these various scenarios, taking adjustable parameters to measure the extent to which a process such as this is “more or less Darwinian.” (p. 158) What’s intriguing about Godfrey-­‐Smith’s discussion is that he remarks how such as CE2 process might occur “on top of” a CE1 process; an individual might be born with one or another behavioral disposition, making him or her likely to update in one or another direction along the spectrum of conformity, cooperation, or what have you, and then such dispositional forms might themselves evolve, via a more “paradigmatic” (i.e., biological) form of evolution. “In this scenario,” Godfrey-­‐Smith explains, “Darwinian processes give rise to a rule for social learning, and that rule may or may not have a Darwinian character itself.”(p. 159) In other words, here seems to be a case where CE1 feeds into or supports a CE2; evolution is acting here at a variety of levels, certainly sequentially, and possibly simultaneously. Darwinian evolution of the more “paradigmatic” sort (individual reproducers’ particular behavioral phenotype evolving), might yield evolution of the less paradigmatic sort, (i.e., actual cultural variants themselves evolving). This scenario naturally leads to the question of whether and how certain cultural variants might be somehow more or less ‘successful’ in some absolute sense? Could there be “selfish” memes – or, cultural variants that are particularly good at getting a hold of our consciousness, whether or not these serve our evolutionary interests? While Godfrey-­‐Smith grants “particular cultural circumstances produce phenomena of this kind” such as “situations of overload” where people are subject to such a “soup of cultural variants,” “discrete cultural fragments – brands, icons, compactly 18 named political entities” might gain ascendance. Godfrey-­‐Smith is skeptical, however, of attempts to generally characterize “memes” as more or less “fit” – or “selfish”. Why? In short, because Godfrey-­‐Smith believes that this “agential” view of memes “induce[s]” us “to see a complex phenomena” through an agential, and thus, inaccurately simple lens. Though he grants that taking a “gene’s eye view” on “memes” is tempting, the assignment of “hidden agendas” simplifies a complex picture, which requires, he thinks, a much more deft hand. While he grants Dennett’s critique of Skyrm’s rational choice model, insofar as we are not “Cartesian loc[i] of well-­‐being,” he resists the move of attributing the evolution of cultural norms to more and less “selfish memes.” Why he resists this has a rather Ockham’s razor feel; just as he resisted the agential view of evolution, insofar as positing “hidden agents” does not clearly do any new explanatory work, he thinks that giving memes “selfish agendas” does no more than redescribe a problem. We have not explained, but described, evolutionary success of ideas. We may not be perfectly rational, but neither are we at the whim of insidiously “selfish” replicators – whether genes or ideas. Conclusions: In Chapter 1, Godfrey-­‐Smith introduces an interesting distinction between science, philosophy of science, and “philosophy of nature.” The last is “taking science as developed by scientists, and working out what its real message is, especially for larger questions about our place in nature” (p. 3). This work, he explains, is a combination of the last two; he is interested not only in the world per se, or how scientists come to understand that world, but in how scientific understanding may illuminate our place in the word. Philosophy of nature “refines, clarifies and makes explicit the picture that science is giving us of the natural world and our place in it”(p. 3). Does the book does achieve this end? I would say so. Godfrey-­‐Smith provides a useful unpacking and critique of the “classical” approach to natural selection – a very frequently invoked set of summaries of the central conditions necessary for a population of entities to evolve. Along the way, he tackles the replicator/vehicle/interactor distinction, the levels of selection debate, population thinking v. essentialism, the gene’s eye view, reproduction, and much else besides. His approach is to start with the science, and explore what it means, where concepts require refinement, or clarification. I would recommend that any aspiring philosopher of biology to read this book. It’s a great introduction to the field, and would serve as a useful central text in a graduate seminar. In addition, of course, it will be a great source of discussion for professional philosophers of biology, as well to general philosophers of science, philosophers of mind and philosophers with an interest in naturalized approaches to knowledge and behavior. 19 The book also proves an interesting contrast to Okasha’s Evolution and Levels of Selection. While their solutions to some of the same problems, when all is said and done, are very similar, the form of argument, and style of exposition is quite different. It would be useful to teach the two books alongside one another in a graduate seminar. Indeed, comparing the two books requires us to think carefully about where philosophy of biology needs to go next, and how. Is reliance on formal models limiting, or enlightening? What should be the relationship between philosophers of biology and practicing scientists – both formal modelers and others? What is or should be the role of biological examples in exposition of philosophical problems? What motivates philosophy of biology? What are the central problems that the field should be tackling? These two books (set alongside one another) require one to think carefully about method, aim, and scope of the philosophy of biology, today, and in the future. Some might critique this book as being too “narrow” – or, focusing too exclusively on evolution, to the exclusion of development or molecular biology. This objection has been raised frequently in recent philosophy of biology. While I’m sympathetic with this critique, Godfrey-­‐Smith’s account is so careful, detailed, and systematic, that one comes away having a fresh sense of how all these long-­‐standing debates hang-­‐together. Godfrey-­‐Smith is interested in characterizing evolutionary theory in a broad enough fashion so as to permit inclusion of the kinds of entities that are often left out – e.g., the “marginal” individuals we find both prior to and during major transitions – or, indeed, any transitional stages where standard or “classical” “agential” views of evolution run aground of the messy biological world. His representation of the space of Darwinian populations and accompanying tool of “de-­‐Darwinization” are useful heuristics. Bibliography: Lewontin, Richard. 1985. "Adaptation" in Levins, R. and Lewontin, R. eds. The Dialectical Biologist. Cambridge, MA. Harvard University Press. 65-­‐84. Lynch, Michael. 2007. Origins of Genome Architecture. Sinauer Associates. Godfrey-­‐Smith, Peter. 2009. Darwinian Populations and Natural Selection. Oxford: Oxford University Press. Maynard-­‐Smith, John and Szathmary, E. 1995. The Major Transitions in Evolution. Oxford: W.H. Freeman. Okasha, Samir. 2006. Evolution and the Levels of Selection. Oxford: Oxford University 20 Press. 21