1. Introduction
Evolution by natural selection (ENS) is often conceived as requiring three essential elements: variation, heredity, and fitness differences. Lewontin (Reference Lewontin1970) provides perhaps the most well-known formulation of this ‘classical account’. For Lewontin, the three components are as follows:
1. Different individuals in a population have different morphologies, physiologies, and behaviors (phenotypic variation).
2. Different phenotypes have different rates of survival and reproduction in different environments (differential fitness).
3. There is a correlation between parents and offspring in the contribution of each to future generations (fitness is heritable).
While subsequent authors have attempted to refine and improve this account, its essentials are widely accepted (Hull Reference Hull1980; Sober Reference Sober1984; Endler Reference Endler1986; Ridley Reference Ridley1996; Stephens Reference Stephens2004; Okasha Reference Okasha2006). On this view, these three elements define what ENS is and give the conditions under which it operates (Godfrey-Smith Reference Godfrey-Smith2007). This is generally thought to be the case, regardless of what sort of biological system is involved—heritable variation in fitness is the recipe for ENS whether we are considering organisms, species, or prebiotic molecules (Lewontin Reference Lewontin1970). And insofar as one entertains the idea of evolution by natural selection in nonbiological contexts, the key consideration is whether some appropriate analogue of heritable variation in fitness can be found (Sober Reference Sober1991/2006; Hull, Langman, and Glenn Reference Hull, Langman and Glenn2001; Hodgson and Knudson Reference Hodgson and Knudsen2008).
I will argue herein that this classical account of ENS is flawed and should be abandoned. The relationship between selection, heredity, and variation must be dramatically reconceived. My analysis will focus on some problem cases, inspired by and based on similar cases considered in Godfrey-Smith's Darwinian Populations (Reference Godfrey-Smith2009). The conclusions I draw from the analysis, however, differ substantially from those of Godfrey-Smith. Godfrey-Smith suggests that the problem cases show that it is difficult for accounts of ENS to say both what will produce ENS and what it means to be a case of evolution by natural selection (20). Adequate descriptions of what it means for something to count as ENS do not do a good job of giving conditions when ENS will actually occur. By contrast, I suggest that the problem cases show that the classical account does not adequately describe ENS at all. It is not a case of trading off descriptive versus predictive adequacy: it is a case where the proposed definition is subtly but irreparably flawed. In particular, I use problem cases to show two difficulties for the classical account that are not considered by Godfrey-Smith: the evolutionary effect of variation in heritability and the possibility of ENS when heritability is 0.
I first consider the role of heredity in ENS. I argue that ENS can occur even when there is no tendency for offspring to resemble their parents more than a randomly chosen member of the population. This substantially undermines the view that heredity is a real requirement for ENS. Next I show that variation in heritability can drive evolution in just the same way that variation in fecundity or survivability can. This suggests that the classical account of ENS is too narrow, providing an artificially circumscribed role for heredity in some respects, even as it overstates the indispensability of heredity.
The reconception of heredity demanded by this analysis raises further difficulties with the classical conception of fitness. This is traditionally viewed as the propensity of an individual to survive and reproduce or as the average such propensity of all individuals that share a trait. However, since predictable evolutionary change can arise from a variety of sources other than differences in the propensity of individuals to survive and reproduce, the tendency of a trait to spread or decline cannot be identified with average individual fitness. This is problematic for reasons that I discuss in detail.
Instead, I suggest that we conceive evolution as arising out of the operation of distinct causal mechanisms, none of which can be identified with ENS. ENS can be understood in two related senses: as an all-things-considered measure of the tendency of traits to become more or less prevalent vis-à-vis their alternatives (broad trait fitness) or as the predictable tendency of a specific causal mechanism to advantage a trait (competitive advantage). In this sense, ENS can be understood as a description of the statistical tendencies of the evolving system, similarly to the statistical conception advocated by Walsh, Lewens, and Ariew (Reference Walsh, Lewens and Ariew2002) and Matthen and Ariew (Reference Matthen and Ariew2002, Reference Matthen and Ariew2005, Reference Matthen and Ariew2009).
2. The Role of Heredity
On the ‘classical’ account of ENS, heredity is one of three essential components, along with variation and fitness differences. Each element can be seen as playing a specific role: variety acts as the raw material, which then is shaped by fitness differences (‘selection’, or the strength thereof), and heredity preserves the effect. Without heredity, the differences in fitness can be as large as one likes, and hence the strength of ‘selection’ can be as powerful as you like, but no lasting change will result. If there is weak heredity, then the lasting impact of selection will also be correspondingly weak. In this conception, heredity gives ‘traction’ for selection to operate, as without heredity, even a powerful engine of selection will just spin its wheels without going anywhere.
I will illustrate the major failings of the classical account by considering three problem cases. The first case shows that, contrary to the classical picture, ENS can proceed even when there is no heredity. The second and third cases show how variation in heritability can drive evolutionary change, giving heredity a completely different role than merely giving traction to selection. Taken together, the cases suggest a fundamental reconception of the relation between selection, heredity, and evolution.
Case 1: Selection without Heredity
The concept of heredity is strongly connected with the idea of correlation between parent and offspring. For example, Godfrey-Smith (Reference Godfrey-Smith2009) claims that “a Darwinian process requires that parents produce offspring who are similar to them” (172), where similarity is ultimately understood via some sort of statistical measure of the correlation between parent and offspring. ‘Heritability’ is, for Godfrey-Smith, heredity made numerically precise in this way; any particular measure of heredity is problematic, but there is a family of them we can employ as the situation warrants (172). He rejects the idea that heredity must involve actual replication, or even that parents have to consistently produce offspring of the same type, on the grounds that it can be sufficient for ENS that there is a small correlation between parent and offspring. For example, if tall animals are fitter than short ones, as long as there is some tendency for the offspring of tall animals to be taller than average, ENS may result. By comparison with such authors as Hull (Reference Hull1980) or Sober (Reference Sober1984), Godfrey-Smith's conception of heredity is relatively weak and undemanding.
If heredity is an essential element of ENS, then presumably a system without heredity cannot evolve by natural selection. However, on analysis, it becomes clear that we can have ENS, even when there is no tendency for offspring to resemble their parents in the weak sense demanded by Godfrey-Smith.
Consider a population for whom offspring character is independent of parent character. Knowing the character of the parent gives no predictive value with regard to the offspring character; the distribution of offspring characters is the same, regardless of the character of the parent. This means that the covariation between parent character and offspring character is 0. The linear regression of parent character on offspring character has slope 0. Heritability is 0. However, it is still possible that the population evolves by natural selection. Contrary to the classical view, ENS does not require heredity, if heredity involves any correlation between parent and offspring character.
Suppose that our population has just two characters, A and B, and reproduction occurs in discrete generations. Both variants have fecundity 1; that is, they have 1 offspring if they survive to reproduce. All members of the population are subject to a chance of mortality in between reproductive episodes: character A has a 30% mortality chance; character B has a 70% mortality chance. But individuals that survive to reproduce are included in the population at the next time step, which is again subject to mortality risk before the next round of reproduction.
The fitness of A is higher because of character A's superior survivability. This is because 30% of type As and 70% of type Bs die before they reproduce (giving them individual fitnesses of 0), whereas 70% of As and 30% of Bs have one offspring but also themselves continue to exist in the population potentially to reproduce again, giving them fitness of 2. The formula for calculating fitness here is s(1+f), where s is the chance of surviving to reproduce and f is the fecundity, giving type A a fitness of 1.4 versus 0.6 for type B.
ENS, however, occurs in this case, despite the fact that heritability is 0. The population moves to an equilibrium frequency of around 66% A and 34% B.Footnote 1 There is no tendency for offspring to be similar to their parents, but the population still evolves due to type A's superior fitness.
So what is going on here? Basically, what is giving ENS traction (i.e., what permits variation in survivability to have a persistent effect across generations) is not heredity but the constant tendency for organisms to stay the same type over time. Alice the A organism's offspring may have no greater tendency to be Alice-like than if they had been offspring of Brenda, but Alice herself has a tendency to remain Alice-ish over time. Alice-type organisms therefore accumulate in the population because of the superior ability of type As to survive when compared with type Bs.
Fitness, as generally understood, has two components: survivability and fecundity. These components are generally treated as interchangeable, but this example shows that they are not—at least, not always. Fecundity differences need to be conjoined with heritability, in order to lead to evolutionary change. Survivability differences do not. All that survivability requires is a tendency for the individual to retain its type over time. And survivability can suffice to produce evolutionary change not just in systems with no reproduction (a case considered by Godfrey-Smith Reference Godfrey-Smith2009) but also in systems that include reproduction with low, or 0, heritability. It might be doubted (as does Godfrey-Smith) whether systems that do not involve reproduction evolve in a reasonable sense of the word; however, this concern does not apply to the case we are considering. The population in our case will evolve to a stable equilibrium with a preponderance of type As over type Bs, and this preponderance is due to the superior fitness of As over Bs, despite the fact that there is no heritability.
A very significant consequence of this is that in any case where individuals potentially survive for longer than the average generation time, treating fitness as a single measure fails to adequately predict evolutionary change, unless heredity is perfect (i.e., offspring are guaranteed to be identical to parents, as in idealized asexual orthogenesis or cell mitosis). This can be demonstrated fairly easily in the case of discrete generations with asexual reproduction.
When individuals die after reproducing, average individual fitness 
, where 
is the average probability of survival to reproduction, and f is the expected number of offspring. If the proportion of the offspring that possess the same trait as their parents is h, then 
gives the proportional change in the absolute numbers of individuals with the trait from generation to generation.Footnote 2 That is to say, however many individuals there are of the type, we multiply that number by 
to get the expected number of individual offspring of that type in the next generation. In this case, it makes no difference to what extent survival and fecundity contribute to fitness. If individuals survive reproduction, however, the resultant proportional change in the quantity of the type is 
. This is because the subsequent generation is made up of both individual offspring, a fraction 
of which share their parent's type, and surviving individual parents, who will in this simple model retain their type by default. This change cannot be expressed in terms of 
; 
must be decomposed into survival and fecundity to specify the change. If heredity is less than 1, the absolute change in the quantity of the type simply cannot be determined from fitness, without knowing the individual contributions of survival and fecundity, and neither can the proportional change with regard to other such types as may exist in the population. Hence fitness, construed as a causal propensity of individuals, becomes a useless quantity in determining the evolution of the population.
Cases in which individuals survive reproduction and heritability is less than perfect are, to put it mildly, common in biology. The fact that dealing with overlapping generations is more complicated should not obscure the fact that overlapping generations is a core case that our theory of ENS must adequately handle. It does not take a very sophisticated or complex analysis of overlapping generations to show that it does not. On the conventional understanding of fitness, heritable variation in fitness cannot predict evolutionary change given overlapping generations, despite the fact that the only confounding factor is imperfect heredity, and its effect is incorporated into the model. This raises a very serious problem for the classical account. Perhaps ‘fitness’—the two-headed amalgam of survival and reproduction—should not be given the central role that has traditionally been allocated to it.
Case 2: Systematic Hereditary Bias
While we have just seen that heredity is not required for ENS, heredity can itself be a source of evolutionary change just as fitness differences can, a source commonly disregarded. Consider a case in which a continuously varying trait such as height is positively correlated with fitness. Ordinarily we would expect the population to gradually get taller and taller. However, suppose that there is a systematic hereditary bias in the population such that children are shorter than their parents. If this bias is equal to the selective advantage, then no evolutionary change will result, and we see a ‘problem case’ for Godfrey-Smith, where one has heritable variation in fitness but no evolutionary change. This particular problem case can be solved (for some purposes, at least) by observing that “the inheritance system would produce change alone if there were no fitness differences” (Reference Godfrey-Smith2009, 170). Hence we can say that heritable variation in fitness would have produced ENS; it is just that the bias in the inheritance system is exactly counteracting the evolution due to natural selection, so we do not observe any actual change.
However, this also illustrates that an inheritance system can produce evolutionary change quite independently from the change due to ‘selection’ as conventionally understood. Rather than simply acting as a handmaiden to selection giving it assistance to work, inheritance here is acting on its own account to change the population.Footnote 3 We could easily set the model so that the population changes toward shorter and shorter individuals, despite the selective advantage of height. In the terminology of Sober (Reference Sober1984), heredity seems to be acting as a separate force from selection, which is a quite different role than that which is conventionally envisaged for it.
Case 3: Heritability as Fitness
Perhaps even more interesting than the evolutionary effect of hereditary bias is the evolutionary effect of variation in heritability. A more heritable trait spreads at the expense of a less-heritable trait, without requiring any variation in fitness as conventionally understood. Consider a simple case in which a population of bacteria is evenly divided among individual phenotypes A and B, who reproduce by mitosis in discrete generations. Phenotype A individuals are fitter, surviving until division up to 75% of the time (giving them fitness 1.5), whereas only half of B individuals survive until mitosis, giving them fitness 1.Footnote 4 The B phenotype is 100% heritable: B individuals always have B offspring. The A phenotype is less heritable, however; 5/6 of the offspring of A bacteria are A, but 1/6 are B.Footnote 5 In this case, overall population size grows, but the proportion of A bacteria and B bacteria remains exactly the same. If A is any less heritable, it will actually decline in the population, despite its higher fitness.
Furthermore, we can consider a case in which the fitnesses of all the variants are identical, but they differ with regard to their heritability. So every individual has an identical expected number of offspring, but trait A is 90% heritable, and B is 70% heritable—10% of the offspring of As are B, and 30% of the offspring of Bs are A. This population will move toward an equilibrium state where 75% of the population is A and 25% is B. We cannot explain this in terms of fitness, at least not in the standard sense of fitness as propensity to survive and reproduce (Mills and Beatty Reference Mills and Beatty1979). But it is clearly not drift, insofar as this is perfectly predictable. Phenotype A has a systematic advantage over B that is not constituted by any difference in fecundity or survivability.
At this point, we run into a terminological difficulty. It is clear from the example that a trait with high heritability has a competitive advantage vis-à-vis a trait with lower heritability.Footnote 6 This effect can counteract or outweigh the effect of superior survivability or fecundity. In the traditional vocabulary, however, superior survivability and fecundity together constitute an organism's fitness, and heritability serves only to give traction for fitness to bring about evolutionary change. But this is inadequate. It should be clear by this point that this traditional picture generally assumes that the heritability of each trait is equal.Footnote 7 But this assumption is an idealization that does not even adequately describe biological reality, let alone ENS in the more general case.Footnote 8 Heritability has been seen as merely giving traction to selection because it has been implicitly assumed in accounts of ENS that there is no variation with regard to heritability. Once one considers the evolutionary effect of variation in heritability, it becomes clear that it provides a competitive advantage every bit as predictable and potentially significant as variation in fitness.
It is somewhat difficult to make room for the competitive advantage of high heritability in the standard terminology, however, since it presupposes that ENS results only from variation in fitness. I wish to emphasize: variation in heritability is a fundamental source of evolutionary change, and it is only where there is no variation in heritability that there is no tendency for the more heritable variants to become more prevalent than the less heritable ones. But it would not be standard to consider a high-heritability trait to be fitter than a low-heritability trait if there were otherwise no differences in fecundity or survivability. This is despite the fact that being more heritable represents a real competitive advantage that can explain the success of a trait against its rivals. Further, it seems likely that this ‘heritability superiority’ has historically played an important role in the evolution of life.
That the classical definition of ENS seems to ignore the possibility of variation in heritability suggests that the account of natural selection that focuses on fitness in terms of survival and reproduction may be biased toward an overly simplified and biologically specific scenario. Darwin conceived of the driving force of natural selection in terms of the struggle of individual organisms to survive and reproduce, with favorable variants more likely to be preserved and then passed on through the hereditary mechanism (whatever it was). But in the more general case, we can explain the origin and distribution of traits through any mechanism that confers a competitive advantage on one trait over its rival. This may be survivability or fecundity or heredity or other possibilities such as resistance to mutation or likelihood of migration.Footnote 9 The question becomes whether it is important to bracket off survivability and fecundity as the components of selection and treat other factors that may advantage the spread of one variant as something other than selection. I will argue the opposite: that there is nothing special about survivability and fecundity and that there are good reasons to treat all cases of competitive advantage as cases of ENS.
3. Rethinking the Classical Account
Our discussion has shown two ways in which the classical approach to defining ENS in terms of heritable variation in fitness fails. It accords both too much and too little importance to heredity. On the one hand, we have seen that heredity plays no essential role in ENS; heredity is not required for fitness differences to produce evolution. When the selective advantage is due to variation in survivability, heredity plays no necessary role. Full Darwinian explanations can occur with perfect heritability and with no heritability at all.
On the other hand, the classical account fails to recognize the extent to which heritability can figure in evolution in ways that precisely parallel more familiar appeals to individual fitness, thereby giving heredity too little credit as a source of evolutionary change. It is advantageous for traits to have higher heritability, in precisely the same way that it is advantageous for a trait to promote high fecundity. The question therefore arises how to parse the evolutionary contribution of variation in heritability.
We are used to thinking of natural selection as one source of evolutionary change among many—the component of change that is due to heritable fitness differences. On the traditional account, ENS is just such an evolutionary change as is due to variation in the survivability and fecundity of individuals. Since variation in heritability is not variation in survivability or fecundity, it would appear that it is not a case of ENS. However, I will suggest that this simple answer is inadequate. I advocate a substantially revised account of ENS, which I will try to show better accommodates the relationship between survivability, fecundity, and heredity.
Let me begin by highlighting an issue that arises with the standard account of ENS. If we believe, as initially seems reasonable, that the fitnesses of traits should allow us to predict the change in frequency of a trait (Mills and Beatty Reference Mills and Beatty1979), we have a problem when there is variation in heritability. Recall from section 2 that superior heritability is a competitive advantage in and of itself. The average fitness of individuals of type A might be higher than that of individuals of type B, but type B might nevertheless increase in the population if B is more heritable than A. Fitnesses in the classical sense are therefore inadequate for us to predict whether the trait will increase or decline in any case in which there is variation in the heritability of the trait. If we conceive the fitness of a trait as just the average of the individual fitnesses, the fittest trait may well decline in the population, and predictably so. This is despite the fact that there are no confounding factors aside from the variation in heritability.
This may not immediately seem to be a problem. Sober (Reference Sober1984) states that there are many sources of evolution, and a trait may have a tendency to increase in the population, due to forces other than ENS. If these forces outweigh ENS, then the fittest trait will decline. We restrict application of the term ENS to such evolutionary changes as result from fitness differences (in the sense of differences in propensities to survive and reproduce), bracketing off other sources of evolutionary change such as variation in heritability or propensity to migrate, or what have you, as other evolutionary forces. ENS, then, explains population change only when no “conditions obtain which would interfere with the actualization of the descendant contribution propensities” (Mills and Beatty Reference Mills and Beatty1979, 275). The variation in heritability represents just such an interfering condition—a distinct force that acts independently of ENS, thereby explaining the failure of ENS to by itself suffice to predict evolutionary change. This, it might be urged, solves the problem and accords with biological usage and the historical sense of the term tracing back to Darwin. However, I think the suggestion has such serious problems that it should be rejected, and we should understand ENS quite differently. I will discuss three major difficulties with this conception.
First, the effect of variation in the heritability of a trait cannot plausibly be bracketed from the effect of selection operating on the trait. We already recognize that a trait that is much fitter but only weakly heritable (for whatever reason) will spread more slowly, or not at all. This is why the classical account sees ENS as heritable variation in fitness. The strength of ENS is not simply equal to the variance in fitness; it is conditioned by the degree to which the trait is heritable. This is part of why heredity is included in the classical recipe for ENS—heredity provides the ‘traction’ for fitness differences to result in ENS. So even in the classical sense of ENS, we cannot separate the degree of heritability of a trait from the effect of ENS, since the heritability of the trait is included in the effect of ENS.Footnote 10
Furthermore, insofar as there is variation in the heritability of a trait, this variation will influence the response to selection of the different traits. Since we must include the degree of heritability in determining the overall ENS effect, any variation in heritability will influence this value, to the advantage of the more heritable variants. Hence, the effect of any variation in heritability will be surreptitiously included in our determination of the effect of selection in cases where we are not entirely idealizing away imperfect heredity. So since we must include the effect of heritability variation in the effect of ENS, we cannot separate heritability variation from the force of ENS.
The second difficulty is that fitness is itself a confused amalgamation of distinct evolutionary processes: survival and reproduction. It makes more sense to amalgamate heritability with reproduction and treat survival separately, but this is incompatible with the classical account. Modeling ‘fitness’ as a single measure is an idealization that under most circumstances leads to inaccurate predictions. Recall that earlier we saw that with overlapping generations, knowing the fitnesses of the types does not suffice to predict evolution even when there are no confounding factors aside from imperfect heredity. This is, more or less, because the portion of fitness that is due to fecundity is affected by the degree to which heredity is imperfect, whereas the survival portion is not: disentangling this requires treating them separately.
Predicting evolutionary change therefore requires that we decompose fitness into its components in all but the simplest and most idealized scenarios. Insofar as survival and reproduction have to be treated separately in even moderately realistic scenarios, it suggests that fitness (which agglomerates them indiscriminately) lacks explanatory relevance. If fitness is ultimately an unwieldy and artificial kludge of two really distinct processes, how can it play such a central role in ENS, itself the central concept of evolution?
The more natural division is to group heredity and fecundity together as aspects of the process of reproduction and treat survival separately. The effect of reproduction on the spread of a trait is, after all, a combination of the fecundity of the reproduction and the mechanism of heredity. Variation with regard to either will advantage the more fecund or more heritable trait. Privileging fecundity differences as selective and ignoring the potential role of heritability differences is understandable from a historical perspective, but it seems to have little conceptual justification. Both fecundity and heritability are required to quantify the evolutionary impact of reproduction, and variation with respect to either constitutes a competitive advantage that can predictably change the distribution of a population over time. A more accurate account of ENS must therefore distinguish between survival and reproduction, rather than lumping them together as ‘fitness’, and must treat the role of heredity as specifically related to reproduction, as both a potential source of change in its own right, and as the enabler for variation in fecundity to have an evolutionary effect.
The third, broadest concern is that treating ENS as a force distinct from variation in heritability and all other forces conceptually separates the fitness of the trait from the expected success of the trait. However, it is our expectations about the success of the trait that are most central—not just to evolution but to the concept of evolution by natural selection—whereas the fitness of the trait, as we have seen, has dubious utility. Out-reproducing and out-surviving rivals is one way that a trait can become more prevalent over time, but it is the concept of predictable increase in prevalence that actually does the work in evolutionary theory. ENS explains the diversity of organic form, the transformation of species over time, and the development of complex adaptations. Each of these explanations proceeds by showing how a consistent tendency of a trait-type to grow in the population leads eventually to the remarkable facts of diversity and adaptation—what Godfrey-Smith calls ‘origin explanations’ (Reference Godfrey-Smith2009, 42). This predictable tendency to grow may, but need not, be due to competitive advantage with regard to survival and reproduction.Footnote 11
If a trait spreads in a population via its superior heritability, its predictable success can then serve as a basis for a new slight mutation, which further improves heritability, and so on. Perhaps this is how the robustness of the developmental system to mutational insults evolved, or how the various mechanisms that safeguard the fidelity of DNA duplication in mitosis spread. Since superior heritability is in itself a competitive advantage for a trait, no further explanation is necessarily needed in terms of the spread of alleles that promote the faithful transcription of an organism's genes. We do not need to suppose that lineages with that allele became more widespread (although that also seems likely due to the assumed negative fitness associated with transcription errors); the allele will itself become more prevalent in the population because a higher fraction of lineal descendents will possess it than the rival alleles in alternative lineages.
It seems to me that this type of evolutionary spread shares the relevant explanatory structure of ENS. The problem is that conceiving natural selection purely in terms of reproduction and survival rules out considering this as a case of ENS, which means that the relevant explanatory fact—namely, the competitive advantage of a trait in terms of its tendency to spread with respect to rivals—becomes conceptually independent from ENS. So separating ENS from evolutionary success loses the connection between ENS and ‘origin explanations’ of diversity and adaptation. To me, this is a serious problem: my view is that the concept of competitive advantage is more central to ENS than survival and reproduction, precisely because it is competitive advantage that does the explanatory work in Darwinism. If the DNA code or aspects thereof became prevalent in the early history of life because of superior heritability compared to alternatives, then I think we should say that selection favored it. But this idea requires that we substantially revise our definition of ENS.
These considerations strongly suggest that we cannot simply bracket off heredity as another force alongside natural selection. At the least, the effect of ENS will include variation in heritability alongside variation in survivability and fecundity. Except in very idealized models, we have to distinguish between the two components of fitness, treating fecundity and heredity together and distinguishing them from survivability. However, this still divorces ENS from the key explanatory role of competitive advantage. Therefore, I suggest we instead adopt a new account of fitness, heredity, and selection that avoids the manifold problems of the classical account and elucidates the relation between selection and evolution in a clearer, more elegant way.
4. Interacting Mechanisms
Ultimately there are only a few ways that populations can change. Individuals can be subtracted from the population—ability to avoid this is ‘survival’ in the conventional understanding, but emigration also counts. Individuals can be added to the population. Reproduction and immigration correspond to this mode of population change. Individuals in the population can change the traits they have, which corresponds more or less to mutation in the standard view. Any of these processes can cause evolutionary change and might over time predictably cause some type to become more prevalent in the population. Any of these processes might therefore give one type a competitive advantage over an alternative.
ENS on the classical account limits itself to changes in the population due to survival and reproduction. If there is heritable variation in the propensity of various types to survive and reproduce, there will be a tendency for the population to evolve such that these ‘fitter’ types become more prevalent—although this tendency might be counteracted or overwhelmed by other processes. So even though one trait might be ‘fitter’ in the sense that individuals with the trait tend to survive or reproduce better than the alternative, the overall expectation as to which trait will increase in the population must take into account all biasing effects from all processes causing evolutionary change. This trait is ‘fitter’ in the sense that we expect it to increase in the population compared to its alternatives as the result of its all-things-considered competitive advantage vis-à-vis said alternatives. Let us call the first sense of fitness ‘narrow’ trait fitness and the second sense ‘broad’ trait fitness.
Broad trait fitnesses are what are most relevant to predicting evolution. Variation in broad trait fitness, together with an appropriate source of novel variation, is sufficient to explain the adaptation of organisms to environment, the complexity of the eye, and the diversity of organic life. It is for this reason that I prefer to identify ENS as variation in broad trait fitness. Narrow trait fitness is interesting because it is an extremely prevalent contributor to broad trait fitnesses. Because of the central role of survival and reproduction in biological evolution, not to mention the history of the term going back to Darwin, the focus on narrow trait fitness is perfectly understandable. However, understanding evolution abstractly—especially if we are interested in extending evolutionary ideas to other domains than biology—requires us to identify broad trait fitness as the central explanatory concept.
Evolution, on this account, is best understood in terms of the interaction of distinct mechanisms of change. In any given population, we can conceive of change as resulting from the operation of one or more causal mechanisms that act to add or subtract members from the population or change the traits of already existing members. Such mechanisms may be treated in a fine-grained way—we may decompose mortality into predation, resource competition, and so forth—or at the coarse-grained level of individual fitness. Any such mechanisms that produce the systematic increase of one variant over another constitute a competitive advantage for that trait. A trait may be advantaged by one evolutionary mechanism yet disadvantaged by another. The aggregate of these effects will be broad trait fitness, our all-things-considered measure of expected evolutionary change.
To some extent, what we choose to call ENS after all this discussion is a matter of preference. One can continue to identify ENS with something like heritable variation in narrow trait fitness. As we have seen, however, heredity is not necessary for ENS, and the evolutionary effects of variation in heritability will be included in the effects of ENS whether they are explicitly recognized or not. That is because modeling reproduction requires that we model both fecundity and heredity, since in order to model the associated evolutionary change, we must determine not only how many new individuals enter the population but also what traits these individuals possess. This suggests that the concept of narrow trait fitness would have to be amended to include heritability, since the evolutionary effect of variation in heritability will be included in any rate of change that involves the fecundity of individuals. Even then, this amended account of ENS would apply only to the most idealized circumstances.Footnote 12 And it has the further disadvantage that it gives ENS a derivative explanatory role, since complex adaptations, organic diversity, and fit of organism to environment are explained ultimately by variation in broad trait fitness (plus drift and novelty).
I prefer to identify evolution by natural selection with variation in broad trait fitness because of its central explanatory role. On my account, evolution by natural selection occurs when there are variants in the population such that there is an overall causal propensity for some of the variants to increase relative to others. This obviates the difficulties with the classical account emphasized in Godfrey-Smith (Reference Godfrey-Smith2007, Reference Godfrey-Smith2009), in which selection seems to be subject to ‘problem cases’ in which the requisites are met but selection mysteriously fails to occur, as well as the various problems that have been discussed herein.
More specifically, any distinct evolutionary mechanism that tends to bias the distribution of traits in the population is a selective mechanism and will be contributing to the overall selection for some traits and against others. An evolutionary mechanism is understood as a distinct real-world phenomenon that causes individuals in the population to enter it, leave it, or change their traits. This usage is similar to that in Beatty (Reference Beatty1984) and Millstein (Reference Millstein2002), in identifying a mechanism that discriminates among the types in the population as constitutive of selection. It diverges from them, however, in not identifying drift as a process of indiscriminate selection or indeed as any process at all.
Evolution by natural selection, on this account, then, is not contrasted with mutation or migration, and it does not necessarily involve heredity or narrow fitness differences (although it does require variation in broad trait fitnesses). If it is contrasted with anything, it is contrasted with drift, for ENS gives us an overall expectation of how the population will evolve, and drift measures the extent to which the evolution of the population diverges from what would be expected. Insofar as any of the individual mechanisms of change operate probabilistically, there will be scope for change to differ from our expectations and therefore for drift to occur.
5. Conclusion
Heritable variation in fitness is an entrenched, venerable conception of the conditions for evolution by natural selection. It captures some of the most significant and distinctive aspects of Darwin's insight into the nature of biological change. It conceives variation as serving as the raw material for selection to operate, with heredity preserving the resultant ‘winnowing’ or ‘sorting’, thus gradually building upon itself for generation after generation. It is an elegant picture in many respects, but as we have seen, it cannot account for the full complexities of the relationship between fitness and heredity.
The classical account of evolution by natural selection should therefore be abandoned. Heritable variation in narrow fitness is not necessary for evolution by natural selection, nor is it sufficient. Our ultimate picture should be quite different from the three-element classical view. Heredity itself figures in evolution by natural selection insofar as reproduction figures. The classical account mistakes a specific biological process with a general evolutionary requirement. Heredity is one way that traits can spread, but there are other ways that do not involve any heredity at all. This means both that heredity is not required for evolution by natural selection to occur and that variation in heredity can lead to evolution by natural selection even if there are no differences in survivability or fecundity.
Any mechanism of change that affects a population can potentially advantage one trait over a rival, causing a difference in broad trait fitness, and hence bring about evolution by natural selection, in the sense I recommend. ENS is the tendency of the population to change in response to these biasing mechanisms. This picture solves the difficulties of the classical view and offers a general account of evolution that can be applied to any sort of system whatsoever, not just those with reproduction and inheritance. It therefore offers a superior alternative to the classical account both for biology and for evolutionary theory more generally.
