Fitness and the Principle of Natural Selection (PNS)

Bouchard and Rosenberg’s view of selection is neatly summarised by a variant of the so-called “principle of natural selection” which they endorse. It reads: “(x)(y)(E) [[If] x and y are competing organisms in generation n, and x is fitter than y in [environment] E, then probably (there is some generation n’, at which x has more descendents than y)]”

(Bouchard and Rosenberg 2004: 695). It is intended as a law, and the only law in

biology for that matter (Rosenberg and McShea 2007). Obviously it is considered a probabilistic law. And, above all, the fitness that figures in it is said to be a causal property and the law itself is deemed a causal law. For the sake of argument, in this chapter we’ll assume that the fitness-talk in the domain of selection and evolution does pick out a property called “fitness” and that (this variant of) the PNS is indeed a law.

My goal then is to show that fitness cannot be a causal property and, by extension, the PNS cannot be a causal law. If this is the case, and if we think that selection can and should be characterised in causal terms, then it cannot be characterised by fitness and any associated laws or regularities.

Some modifications and clarifications have to be made about the PNS in the first place. The indefinitely long time-frame in which the effect is specified is pointless. For the PNS to be at least a prima facie causal law, the effect has to be the fitter organism’s reproducing more offspring than the less fit one. That is, the time-frame has to be limited to a single generation. A property-instance cannot be causally responsible for whatever happened long after it had ceased to exist. If an organism’s fitness can be causally responsible for some reproductive results at all, it can be so only for that organism’s own reproductive success, i.e. the amount of that organism’s (immediate) offspring, and similarly for an organism’s being fitter than another. It may be said to be linked to the amount of that organism’s distant descendants at a remote generation via a very long causal chain. But that doesn’t imply that it is itself causally linked to the latter, for causal relation is not generally transitive. In our case, the environment and other causal conditions may and often vary from generation to generation, and partly because of this different generations of descendants of the same organisms may not and often do not have the same fitnesses or fitness-variations. For these reasons, fitness-instances cannot be causally responsible for any aspect of non-immediate reproductive results and,

therefore, an organism’s being fitter than another cannot have as its causal consequence the former’s having more descendents than the latter if those descendants are not reproduced by them.

Besides, if one merely wants to accommodate the idea that fitter organisms do not always reproduce more offspring, because some background causal conditions do not always obtain and/or because fitness-instances are genuinely indeterministic in relation to reproductive successes, the probability operator contained in the consequent-clause of the PNS has already done the job. If, however, one just intends to talk about series of differences between the amounts of two organisms’ respective descendants over generations, then the PNS is actually not characterising selection but is describing trends of selection, i.e. series of selection-instances. Since descendants of the fitter are not always fitter than descendants of the less fit by more or less the same degree, especially over generations, there is little reason to concatenate what are ordinarily thought of as successive yet distinct instances of selection into a single process. And, even in cases where (most) descendants of the fitter ancestor are actually fitter than (most of the) descendants of the less fit ancestor (by more or less the same degree) over several generations and, moreover, in some of those generations the descendants of that fitter ancestor do outnumber those of the less fit one, what can be causally responsible for such outnumbering in each of those generations is the larger groups’ parents’ being (generally) fitter than the smaller groups’ parents in the previous generation, rather than the former’s distant ancestor’s being fitter than the latter’s in the starting generation, or the former’s being descendants of that fitter ancestor while the latter are descendants of the less fit one. Different generations of organisms are simply involved in different

“rounds” of selection. The indefinitely long time-frame in the consequent-clause makes no sense and has to be replaced by the single-generation time-frame, or otherwise the

PNS cannot be said to characterise selection and cannot be a (prima facie) causal law.

The case for the antecedent-clause is much trickier. It picks out pairs of (life-histories of) competing organisms living in a common environment and having the relation of fitter-than. But Bouchard and Rosenberg didn’t clarify under what circumstances are two organisms competing ones or how environments are supposed to be individuated in their talk of selection. Depending upon whether the two organisms are allowed to be members of different populations, these may pose special difficulties for their account of selection, since the ordinary notion of selection neither counts upon the idea of competition nor demands a special principle of individuating environments. The ordinary notion applies to the whole of the organisms of a population: For all practical purposes, a biological population is nothing but a group of conspecific organisms that is relatively isolated from other groups of organisms in respect of gene flow and/or environment. While this does not in itself implies that organisms of the same population typically interact to a considerable degree or interact more frequently/closely with one another than with non-members, it does however suggest so, especially for animals (e.g.

mating behaviours). But it by no means suggests that they compete in any straightforward sense, locally or globally, between individuals or between groups within the population, unless one resorts to a notion of competition in which organisms sharing the same environment compete for the (limited) resources in it, and hence for their respective survival and reproductive success. Such a notion is obviously redundant in the common talk of selection since members of the same population must live in the same environment and so must compete in that trivial sense. Meanwhile, the environment for a population is understandably coarse-grained, so that organisms of the same population can be said to share the same environment yet there can also be inhomogeneities within the environment where they live. Such a coarse-grained way of

individuating environments is inherent in the practical notion of a biological population.

The ordinary talk of selection simply builds upon the latter along with the associated way of individuating environments.

Now, if the antecedent of the PNS applies to all and only organism-pairs within the same population that enter into the relation of fitter-than, saying that such organism-pairs are competing organisms is as redundant as in the case of the ordinary talk, and the environment for each of those organism-pairs is exactly the same coarse-grained environment for the population to which it belongs. In that case, as we shall see immediately, Bouchard and Rosenberg’s account isn’t fundamentally different from the ordinary population-based talk. If, however, two organisms belonging to the same population can be non-competing organisms or be said to live in different environments so that the antecedent of the PNS does not apply to them, or two organisms belonging to different populations that live in partially overlapping (coarse-grained) environments (hence members of the two populations are of different species) can be competing organisms and one of them can be said to be fitter than the other in a certain finer-grained environment shared by them so that the antecedent of the PNS does apply to them, then Bouchard and Rosenberg’s account is distinct from the population-based notion of selection. We certainly don’t want to say that two organisms of very different species are competing organisms even if they share a very fine-grained environment or there are frequent/close interactions between them. Also, it probably makes little sense to say that one of them is fitter than the other in such a scenario or that they (and they alone) are involved in a selection-token. Without a general criterion for discriminating competing organism-pairs from non-competing ones and another for sorting out when two organisms share the same environment and when not, the scope of the PNS is simply left open. Then, unfavourable cases like those mentioned above cannot be ruled

out and thus are potential threats for Bouchard and Rosenberg’s account of selection if they intend it to be different from the population-based notion (which I believe is the case). Note that I didn’t mean that there cannot be such criteria. But responsibility lies with the advocates of the PNS to spell them out in order for us to further evaluate it.

For the sake of argument, however, we’ll assume that the antecedent of the PNS applies to all and only organism-pairs within the same population that enter into the relation of fitter-than. For the thesis to be argued later, that fitness is causally inefficacious with respect to reproductive success, is independent of the precise scope of the PNS because, for any proposed scope of it, the thesis holds true for each and every instance of it. So we’ll drop the notion of competition and take environments to be coarse-grained so that all members of the same population (at the same generation) are considered to live in the same environment. With some verbal rearrangements, the revised PNS now reads: For any two organisms of the same population (at the same generation), if one is fitter than the other in the environment they live, then the former is likely to reproduce more offspring (i.e. to have greater reproductive success) than the latter. Quantification over environments is abandoned because it is problematic to quantify over causal conditions and because it is readily understandable from the antecedent-clause that the PNS is not restricted to any specific set of environments. And, since the PNS is advertised as a causal law, the revised PNS can be thus abbreviated:

An organism’s being fitter than another within the same population is usually responsible for its having greater reproductive success than the latter. Or, equivalently:

Difference in fitness between two organisms of the same population is usually causally responsible for difference in reproductive success between them.

While further modifications will be made in later discussions, for now we can be content with the current re-formulation. It is formally very similar to the regularity

characterising the ordinary, population-based and fitness-dependent notion of selection, which, if also regarded as a causal regularity, can be expressed as: Variation in fitness amongst organisms of the same population is causally responsible for the variation in reproductive success among them. This gives us an opportunity to see that Bouchard and Rosenberg’s view about selection is not really distinct from the ordinary, population-based talk. In the ordinary notion, a selection-instance is a population-wide (as well as generation-long) process, a macro-process that involves all and only organisms of a population at a generation. For Bouchard and Rosenberg, however, a selection-instance is a pair-wise process, involving only a pair of organisms of a population at a generation. The population-wide process is then regarded as an aggregation of such pair-wise processes. But, while they claim that the pair-wise processes are causal processes, they are reluctant to accept that the aggregated process is a causal process. They take the population-wide process to be characterised by variations/differences in population-genetic fitnesses, which they think are mathematical/statistical constructs each derived from the reproductive results of one or more groups of organisms of the same genotype or trait-type, are attributed to organism-groups and are not causal properties. Only the “ecological-comparative” fitness they advocate, i.e. the fitter-than relation, is explicitly said to be a causal property.

Presumably, the talk of the fitter-than relation presupposes and is based upon the existence of a “non-comparative” fitness that is attributed to individual organisms. If fitter-than is a binary causal property, then there is no reason why the fitness that is attributed to individual organisms, an ontologically more fundamental unary property, cannot be a causal property. Suppose the population-genetic notion of fitness can be interpreted as the group version of such a notion of organismal fitness. Then Bouchard and Rosenberg owe us an explanation as to why the former cannot pick out a causal

property or even a property at all. If group fitness is itself a causal property, then, given that fitter-than is a causal property, there is a high chance that variation/difference in group fitness also represents a property, say, of the whole population. If so, the regularity between variation in group fitness and variation in group reproductive success may be deemed a causal regularity and the sort of population-wide process characterised by it a type of causal process. Thus, Bouchard and Rosenberg have actually opened this possibility, if their claim that fitter-than is a causal property is true.

But the ordinary, population-based talk of selection we’re talking about does not rest upon the notion of population-genetic fitness or that of group fitness. The notion of fitness it builds upon can only be applied to individual organisms and, naturally, can only be the same as the one which Bouchard and Rosenberg advocate or, more precisely, presuppose. So we don’t have to worry about whether group fitness itself is a causal property or whether it is a property at all. And Bouchard and Rosenberg are completely silent about whether variation in fitness amongst all organisms of a population is a causal property or not. Again, if variation/difference in fitness between two organisms is a causal property, I don’t see why variation in fitness among many organisms cannot also be a causal property. Such a property of course cannot be a relation between organisms, because populations are far from being constituted by the same number of organisms and it is a metaphysical requirement that all instances of a relation should have the same number of terms. As a workaround, we follow Millstein (2006) in saying that variation in a property amongst all organisms of a population is a unary property of the population (which she calls a population-level property).

Here, one may resort to something like a micro-macro causal exclusion argument and conclude that variation in fitness amongst many organisms is not a causal property given that fitter-than is. The trouble is that if that argument goes, then fitter-than is not a

causal property either on condition that fitness is. In the same vein, one may say that it is unnecessary to posit the property of variation-in-fitness-among since, if posited, it is clearly ontologically reducible to fitter-than and nothing seems to be explainable by citing it without also being explainable by citing the more basic fitter-than relation. But, likewise, if that argument is correct, then we don’t have to posit fitter-than either, since it is evidently ontologically reducible to fitness and there is nothing such that it is explainable by citing fitter-than but not by citing fitness. This is the line of thought that we’ll pursue in the next chapter in rejecting that property-variations themselves are causal properties or that they are properties at all. But the point here is that Bouchard and Rosenberg’s account of selection and the ordinary, population-based talk are in the same boat regarding the causal efficacy and the redundancy of the supposed causal property that characterises selection in respective views. If in our case there is a dividing line between causal properties, laws/regularities and process-types, on the one hand, and non-causal ones on the other hand, it should lie between the organismal ones and the pair-wise and population-wide ones. Hence, for our purposes, Bouchard and Rosenberg’s account and the ordinary talk do not represent two distinct views.

So, whether fitter-than is a causal property, whether the PNS is a causal law, and whether the sort of pair-wise process characterised by the PNS is a type of causal process, all depend upon whether fitness is causally efficacious with respect reproductive success, assuming that there is indeed such an organismal property. But what sort of property is fitness, after all? Obviously an adequate account of the nature of fitness is in demand. And the next section is devoted to it.