Fitness as a Second-Order Functional Property

organisms. It is not a physical property but is a supervenient property (Rosenberg 1982;

Sober 1984; Bouchard and Rosenberg 2004). It is a posit that grounds the population-genetic notion of fitness, which Matthen and Ariew aptly point out is simply the (actual or expected) growth rate of a (within-population group of organisms of the same) genotype or trait-type. And it is supposed by some philosophers including Bouchard and Rosenberg to be causally efficacious with respect to reproductive success. These are the common grounds for an account of the nature of fitness. Actually, they already strongly suggest that this supposed property is a second-order functional property, although no philosophers have ever recognised that. To expound and consolidate the thesis, we first consider what fitness is not.

First, fitness is not what it grounds. The population-genetic notion of fitness is either a descriptor of change (actual growth rate) or an estimator of change (expected growth rate). Either it picks out the whole change/process or it is a mere concept. In either case, it does not designate a property. So the fitness we’re talking about cannot be identified with it. Note that the reason lies in that it cannot be said to designate a property, rather than that it applies to organism-groups as opposed to individual organisms. A group version of the fitness we’re talking about is readily conceivable, which would be a property of organism-groups and would directly ground the notion of growth rate, actual or expected. Of course, there are additional metaphysical or methodological difficulties for accepting such a property. The point here is just that the sort of thing that can be possibly identified with fitness has to be a property, not a change (however it is described) or a mere concept devised to estimate a change.

Second, fitness is not actual or expected reproductive success of an organism. It cannot be identified with expected reproductive success because the latter is also an

estimator and hence a mere concept.¹ And it cannot be identified with reproductive success for a very obvious reason: The latter is its supposed effect. Furthermore, even if we don’t suppose that fitness is causally efficacious with respect to reproductive success, identifying fitness with reproductive success amounts to saying that fitter organisms are those having greater reproductive success, which simply renders the (modified) PNS tautological (the probability part aside, which will be addressed in the next section).

Thus, if we want the PNS to be a law, and/or if we want fitness to be a causal property, then fitness definitely has to be kept distinct from reproductive success.

Third, fitness is not the trait-complex that is causally responsible for an organism’s reproductive success. The reason is simple: Organisms of different types have different trait-complexes that are causally responsible for their respective reproductive success.

This holds not only for organisms of different species but also for organisms of different trait-types (or genotypes) of the same species or within the same population. And this is precisely why fitness has to be regarded as a supervenient property that supervenes upon a set of complexes. Identifying fitness with each of the different trait-complexes in the subvenient set either violates the principle of transitivity of identity (so long as fitness is still recognised as a property) or simply eliminates fitness altogether.

This is a key reason why it is generally acknowledged that supervenient properties are ontologically irreducible to their subvenient properties. Besides, for those who insist that selection is essentially characterised by the PNS, elimination of fitness is out of the question, or otherwise their account of selection cannot even get off the ground.

1 Beatty (1984) claims that population-genetic fitness, which he equates with expected reproductive success, is a property (specifically a propensity). This claim does not take population-genetic fitness literally; that is, it is already a theory of fitness. The literal interpretation of population-genetic fitness, underscored by Matthen and Ariew and also accepted by Bouchard and Rosenberg, is that it is the estimator of (actual) growth rate (or reproductive success). An estimator is not itself a property. So Beatty’s claim should be understood as a posit of a property named “fitness” that proprietarily grounds fitness-the-estimator-of-growth-rate. A better semantics of the fitness-talk in population genetics, which is compatible with the plain idea of fitness as expected reproductive success yet does not simply take it

Despite these, however, it is quite surprising that some philosophers who assert that fitness is a supervenient property should say something that strongly suggests that they identify fitness with trait-complexes. For example, fitness has been said to be the property of “overall adaptedness” of an organism in its environment (Brandon 1978), to have “components” which are an organism’s traits (Sober 1984), and to be the “overall solution to the design-problems set by the environment” (Bouchard and Rosenberg 2004). The major problem with all these arises from the fact that fitness is taken to be some sort of sum total. A property certainly cannot be a mereological sum. So the only way to make sense of the talk of fitness as an overall something or a composite is to understand it to be saying that fitness is a conjunctive property the conjuncts of which are also properties. For Sober such a conjunct-property is straightforwardly a trait. For Bouchard and Rosenberg it is, on the face of it, a solution to the design-problem set by certain aspect of an environment. I take it that such a solution is nothing but one or more aspects, i.e. traits, of an organism. For Brandon, it is also a trait since, according to Byerly and Michod (1991a), Brandon shifts the sense of “adaptedness” between expected reproductive success and “adaptive capacities”, which are simply traits that causally contribute to the reproductive success of an organism of a certain type. The trouble here is not that conjunctive properties are not acceptable. They are, and we’d better have them even in our case. The real problem is that the conjunction of traits (or of partial adaptedness or of partial/particular design-problem solution should these are not just traits but some exotic properties) that is identified with fitness is type-specific.

Thinking fitness in terms of a conjunctive trait is exactly identifying fitness with each of its subvenient properties. The result is that fitness cannot be a property, and thus there is no fitness in the sense we’re talking about. The talk of fitness as a total/overall something is contradictory to the idea that fitness is a supervenient property. If we want

to keep fitness alive, the former has to be abandoned.

Now, fitness is not only a supervenient property but also a functional property. The functional nature of fitness has never been laid bare. It comes to the fore when we ask:

What exactly is the subvenient property-set for fitness? That set cannot be determined by enumerating its elements because the subvenient property-set (or realiser-set) for a supervenient property (or functional property) is characteristically an open set (Fodor 1997). It has to be determined by the similarity among the trait-complexes, or conjunctive traits, that are members of it. What the similarity consists in is quite obvious: The trait-complexes that are subvenient properties for fitness are all causally responsible for reproductive success. This similarity effectively works as a specification of a causal/functional role, and any trait-complex that satisfies this role-specification is a subvenient property or, better, a realiser of that causal/functional role. And, above all, fitness itself is essentially that causal/functional role. That is to say, fitness is a functional property. Also, since it has multiple realisers, it is a multiply realised functional property.²

I’ll stick to the talk of realisation instead of supervenience. And I’ll adopt the order view about realised functional properties and understand fitness to be the second-order functional property of having some trait-complex or other that is causally responsible for reproductive success. This is consistent with the idea that fitness

2 Standard functionalism, i.e. causal-role functionalism (which is also called analytical functionalism), is adopted here. The functionalism about fitness is intended to encompass a wide variety of proposed understandings of fitness, including the propensity/disposition interpretation, the (categorical) property view (to which Bouchard and Rosenberg’s proposal belongs) and the ability/capacity talk (e.g. Haug 2007). For in all these understandings, fitness is defined (or should be taken as defined) essentially by reference to its supposed effect, i.e. reproductive success, which is uncontroversially also an effect of indefinitely many different trait-complexes. This is where supervenience and functionalism enter the scene. Also, it follows from functionalism about fitness that we take the functionalist view about disposition/propensity if fitness can be appropriately considered a disposition/propensity (which will not be questioned in the present thesis, but see Byerly and Michod (1991a, 1991b) for an objection).

Functionalism about disposition/propensity has prima facie plausibility in our case since the propsensity interpretation of fitness, too, accepts the supservenience of fitness upon traits, which are themselves also properties and the instantiation of which is simlultaneous with the instantiation of fitness. These invite

supervenes upon trait-complexes. Moreover, according to some philosophers, the notion of a realised or second-order property helps understand why supervenience, which is at bottom a pattern of concurrent property-instantiations, ever holds at all. For the supervenient property, here fitness, is instantiated in virtue of the instantiation of one of its subvenient properties, in our case a trait-complex that is causally responsible for reproductive success. This is the main reason why the talk of realisation is more preferable. Plus, an idea very akin to that of property-ordering will prove to be a key to a satisfactory characterisation of selection in the next chapter. Nonetheless, the following arguments against the causal efficacy of fitness depend only upon its being a supervenient functional property. So we can take advantage of the notion of a second-order functional property without loss of generality of our arguments.

A note on property-ordering in our case before we proceed. Orders of properties are always assigned relatively. We take fitness to be a second-order property and all traits and trait-complexes first-order ones. This only signifies that fitness is one order above the latter. It surely doesn’t imply that there are no properties one or more orders below traits. Nor does it imply that there is no ordering among traits. I believe that there is a plurality of intersecting ordering hierarchies below fitness. Yet none of these is our concern. The only ordering contrast relevant to the current debate is between fitness and traits. Also, first-order properties in our case are not all physical properties. Some traits, like shape and colour, are physical properties or readily reducible to physical properties.

Other traits, like visual acuity and cold tolerance, are capacities or dispositions and are themselves functional properties and thus irreducible to other properties. Given the causal exclusion argument presented later, these latter traits cannot be causal properties either. But their physical bases (one or more orders below), which are still traits, are nonetheless causal properties. Thus, when saying that realisers of fitness are causally

responsible for reproductive success, it is to be understood that the traits constitutive of a realiser are either physical traits in the beginning or physical traits that are physical bases of some functionally identified traits. So, although for functionally identified traits we’ll directly mention them rather than their physical bases, there is no problem asserting that the first-order, realiser properties in our discussion are causally efficacious with respect to reproductive success in the final analysis. For this reason, we’ll continue to take all traits as belonging to one and the same order for convenience.