3. The Account of Selection as a Type of Causal Process Characterised by
3.4. The Causal Exclusion Problem
Being a second-order or a supervenient property, fitness is also a target of the standard version of the causal exclusion argument. By “standard version” I mean the one that is systematised by Kim (1997, 2003) in showing that non-reductivist physicalist mental properties, and more generally all second-order or supervenient properties, are causally inefficacious. It has four general premises. As will be shown immediately, any of them either holds true for our case or is a widely agreed metaphysical principle.
Hence, the consequence that fitness is causally inefficacious with respect to reproductive success is inevitable.
The first premise is the irreducibility of the properties the causal efficacy of which is in question to any of the properties the causal efficacy of which is accepted and is to exclude that of the former properties. Fitness is irreducible to any trait-complex. And it is beyond doubt that trait-complexes are causally efficacious with respect to reproductive success. So, apparently, the relation between fitness and trait-complexes satisfies the irreducibility thesis.
The second is the supervenience or realisation thesis: The causal-efficacy-to-be-excluded properties are supervenient upon or realised by the properties the causal
efficacy of which is to exclude the former’s. Fitness supervenes upon trait-complexes and is multiply realised by the latter. Thus this thesis also holds true in our case.
The third premise is what Kim (2003) refers to as “weak physical (causal) closure”.
In the discussion of mental causation, it is formulated in this way: If a physical event has a cause at an earlier time, then it has a physical cause at that time. Note that it alone doesn’t exclude the possibility that the physical event has a mental cause at that time, because it doesn’t say that the physical cause is the only cause then (and so it is intended to be distinguished from “strong closure”, by which a physical event can only have physical causes). Its parallel in our case is: If there are some properties possessed by an organism that are causally responsible for its reproductive success (given all relevant environmental and background conditions), then some of its traits are causally responsible for that. This is very much a truism. And it is also a case of physical closure because traits, environmental and background conditions and reproductive success can all be regarded as belonging to the physical. As mentioned in the end of §3.2, when talking about causal efficacy, traits are either physical traits in the beginning or physical traits that are physical bases of some functionally identified traits. Environmental and background conditions here are all physical conditions. Reproductive success is basically a matter of total amount of offspring reproduced. It is usually thought of as a property of an organism but it is better considered an aspect of an organism’s reproductive results. Either way, since it has to do with amount of objects, it may be regarded as physical or reducible to the physical. For the sake of argument, we just treat it to be physical. The only important thing is that fitness is neither a trait nor a physical property; it is distinctively functional. Nonetheless, as in the case of mental causation, the truth of weak physical closure doesn’t by itself preclude the possibility that the same organism’s fitness is causally responsible for its reproductive success.
The last premise is the causal exclusion principle: Barring genuine cases of causal overdetermination, an event or property-instance (or an aspect of a result) cannot have more than one sufficient cause at any earlier time. This is a metaphysical principle. It is intuitively acceptable and is widely acknowledged. Clearly, in conjunction with the closure thesis, it is what forces us to rule out the causal efficacy of all second-order or supervenient properties.
A step-by-step derivation of the exclusion argument against the causal efficacy of fitness analogous to the one given by Kim in the case of mental causation is due here.
Suppose, given all relevant environmental and background conditions, an organism’s fitness is causally responsible for its reproductive success. By the supervenience or realisation thesis, that organism also has some trait-complex, say Tc, upon which its fitness depends. By the irreducibility thesis fitness is distinct from Tc. Suppose its Tc is also causally responsible for its reproductive success. By the exclusion principle, since this is not a genuine case of causal overdetermination, either its fitness or its Tc, but not both, is causally responsible for its reproductive success. But which? Suppose its fitness is. Then, by the closure thesis, it must have some traits, either conjointly Tc or otherwise, that are causally responsible for its reproductive success. Then the exclusion principle applies again. The closure thesis and the exclusion principle would alternately apply unless and until some trait-complex is taken to be causally responsible and fitness not.
And we should stop the regress as early as possible, by taking to be causally responsible the trait-complex that is first available as an option in the argument, i.e. Tc. Therefore, it is the organism’s Tc, but not its fitness, that is causally responsible for its reproductive success. Thus completes the exclusion argument against the causal efficacy of fitness.
In so far as fitness is a second-order or a supervenient property, there is nothing specifically about it which one can resort to in order to avoid the causal exclusion
problem. And none of the four premises seems abandonable. It’s not an option for those who recognise fitness as a property/disposition/propensity to give up the irreducibility and/or the supervenience/realisation theses, for otherwise fitness would not be the sort of thing they advocate and for them that would amount to the elimination of fitness. The closure thesis is self-evident. The exclusion principle cannot be discarded either since its purpose is to prohibit positing superfluous causes. So, if one still insists upon the causal efficacy of fitness, one has to demonstrate that the exclusion argument, in its general from, is somehow mistaken. To my knowledge, general objections that have been raised against the exclusion argument fall into three categories. Although this is not the place to give a full examination of them, before closing this chapter I want to briefly discuss them in order to point out that they are not enough to refute the exclusion argument.
First, many authors require that overdetermining causes should occur independently (Schlosser 2006; Shapiro and Sober 2007; Raatikainen 2010; Shapiro 2010; Carey 2011). The reason is that in genuine cases of physical causal overdetermination, the overdetermining causes are independent causes. This idea of independent occurrence, or something near, is regarded by those authors as constitutive of the very notion of causal overdetermination. Since the instantiation of a supervenient property depends upon the instantiation of one of its subvenient properties, an instance of a subvenient property of a supervenient one and the concurrent, dependent instance of that supervenient property are thus not deemed overdetermining causes. There is, then, no question of causal overdetermination in such occasions, and the causal efficacy of supervenient properties become compatible with that of subvenient properties. The exclusion principle is thus rendered irrelevant/inapplicable to second-order or supervenient properties, or it is viewed as problematic as it stands.
This line of objection is obviously ad hoc. It doesn’t give any reason as to why dependent but distinct causes cannot be overdetermining causes. Besides, it intimates that the notion of causal overdetermination can always be tailored to one’s purpose and thus can hardly be the real issue. The rationale underlying the exclusion principle is essentially a version of Occam’s razor. When it is applied to properties, its content has been perspicuously articulated by Rives: “[W]e [should] not posit any more causally efficacious properties than we need in order to account for the causal powers of particulars” (Rives 2005:25-26). It doesn’t hinge upon the notion of causal overdetermination, and it is no less intuitively acceptable and no less legitimate an ontological principle than the exclusion principle. We can directly appeal to it instead of the exclusion principle in the exclusion argument without being bothered by the notion of causal overdetermination, and the conclusion is the same. Unless one explicitly shows what is wrong with it, there is no reason not to accept it and its implications.
Second, some philosophers find the standard talk of realised properties as second-order properties unsatisfactory and put forward alternative conceptions of realisation or other notions to understand the relation between realised and realiser properties, for the purpose of solving the exclusion problem. Chief among them is Shoemaker (2001, 2007), whose subset theory of realisation says, roughly, that a property’s being realised by a physical property is such that its causal profile is a proper subset of the causal profile of its physical realiser. Thus realised properties are first-order properties, and the causal powers which a realised property bestows upon an object are the same as some of the causal powers which its realiser property bestows upon the same object. Since realised properties confer causal powers upon objects, they are ipso facto causally efficacious. Similar strategies include taking the relation between mental and physical properties as that between determinables and determinates (Yablo 1992), along with its
various descendents which propose such metaphysical notions as “inclusion” (Schlosser 2006) and “co-instance” (MacDonald 2007) to explicate the relation between instances of realised properties and instances of realiser properties. In all of these, instances of realised properties and the concurrent instances of the corresponding realiser properties are tightly related in such a way that the causal efficacy of the former just coincides with (part of) the causal efficacy of the latter. Because of this, realiser properties are reckoned as causally efficacious properties.
Without going into detail about each proposal and its specific problems, I agree with Kim (2010) that this line of strategy assumes the conclusion from the very start: The theories or notions that are supposed to be able to explicate the relation between realised and realiser properties or between their instances have presupposed that realised properties do confer causal powers. If the exclusion problem is real, it won’t help to redefine “realisation” or to devise new concepts so that the exclusion argument as it stands doesn’t seem to go under alternative accounts of the relation between realised and realiser properties or between their tokens. According to Kim, the original problem of whether realised properties have causal efficacy is just verbally transformed, in the subset model of realisation, to the problem of whether they are “realised” at all, that is, whether they do have causal profiles or confer causal powers at all. In addition, theories modelled upon the determinable-determinate relation are poor ones because determinables are subject to ontological elimination (Heil 1999; Gillett and Rives 2005).
And, if any alternative to the second-order view about realised properties is aimed at motivating the idea that, since the instantiation of a realised property depends upon the instantiation of any of its realisers, the causal efficacy of an instance of the realised property just depends upon that of the depended instance of the corresponding realiser but instances of realised properties nonetheless have causal efficacy, then that is plainly
in vain. That is something like the notion of “supervenient causation” or the principle of
“causal inheritance” which Kim himself advocated in the past but has long renounced.
As he observes, supervenient causation and inherited causal efficacy are still subject to the exclusion problem, unless one simply stipulates that they are not so (Kim 2003).
The third objection is the “generalisation problem” raised by Block (2003), which concerns the overwhelming consequence of the exclusion argument. It initially arises against the backdrop of the idea of a layered ontology, in which the psychological constitutes an autonomous level and mental properties are regarded as higher-level properties. If mental properties’ causal efficacy is excluded by physical properties’
causal efficacy as dictated by the exclusion argument (which is originally directed at mental properties), then, by the same token, all other higher-level properties’ causal efficacy should also be excluded by lower-level properties’ causal efficacy. The result is that only the properties of fundamental particles can be causally efficacious, given that the level of fundamental particles is the bottom level for the conventional layered view of reality. This is intended as a reductio: Since the exclusion argument leads to such an implausible consequence, it should be abandoned and one need not worry about the causal efficacy of mental and all other higher-level properties.
It is always possible to rebut such an objection by insisting that we cannot but bite the bullet in so far as the argument is sound. But we have to ask the more substantial question: Does the exclusion argument generalise to higher-level properties? It is crucial to make the distinction between second-order properties and higher-level properties.
There are properties that are essentially functionally identified and those that are not.
When the causal-role that is identified with a functional property is satisfied by each and every property from a set of properties, the first property is said to be a property over, or an order above, the latter set of properties. Thus the order-distinction is purposively a
distinction between functional properties and their realisers. On the other hand, the level-distinction is used much less systematically. In the usual level-talk, including the talk of a layered ontology, the level-distinctions are really a mixture of order-distinctions and macro-micro order-distinctions (Kim 2002). The latter sort of distinction is based upon the mereological relation: Macro-objects are mereologically composed of (not always homogeneous) micro-objects, and macro- and micro-properties are respectively properties of macro- and micro-objects. Such a distinction is evidently very different from the order-distinction, and it is useful to distinguish them apart by reserving the notion of level for it.
Now, as observed by Kim (1997), the order-distinction does not track the level-distinction. It is part of such notions as supervenience, realisation and hence property-ordering that the two sets of properties bearing these relations are properties of the same objects, i.e. properties of the same level. And properties of different levels can not be possessed by the same objects at a fixed level. The implication is obvious: Properties of a given level cannot be supervenient upon or realised by properties of a different level.
As a result, the exclusion argument does not apply to higher-level properties, because the supervenience/realisation thesis does not hold for properties that are of different levels. This should come as no surprise: An object’s causal powers, which are endowed by its own properties, are distinct from any of its micro-constituents’ or its proper part’s causal powers, which are endowed by the latter’s own properties. Hence the causal efficacy of higher-level properties is not excluded by the causal efficacy of lower-level properties. So, none of these three lines of objections is well founded.
3.5. Summary and Prospect
chapter. If there is an organismal property called “fitness” that grounds the talk of expected reproductive success or expected growth rate, then it is a second-order (or supervenient) functional property. Because it is a functional property, it, as well as the dependency between it and reproductive success, suffers from the problem of metaphysically necessary dependency. The same goes for its determinates, difference/variation in its determinates and all associated dependencies. Therefore, fitness-levels and fitness-difference/variation are not causally efficacious with respect to degrees of reproductive success and difference/variation in reproductive success, respectively. Nor is the PNS a causal law. All these remain true even when the connections between fitness-levels and degrees in reproductive success are taken as probabilistic or even indeterministic. Moreover, because fitness is a second-order property, it cannot escape the causal exclusion problem. Thus, again, fitness, etc., are not causally efficacious properties and the PNS is not a causal law.
If one thinks that selection is characterised by the PNS or any version of the “fitness law”, then selection cannot be a causal process-type, even though each instance of selection is no doubt a causal process. Conversely, since selection is considered a causal process-type, it cannot be characterised by a type of connection that has fitness or its relative as its causal term. Fortunately, a fitness-free characterisation of selection not only is possible but also has been actually proposed. That proposal is not problem-free as it stands but it is good start. At least it is not troubled by the problems raised in this chapter. Lastly, I believe that the property of fitness has neither theoretical function nor explanatory utility and hence is redundant. If a fitness-free characterisation of selection is plausible, that would itself be a cogent reason to dispense with fitness. As we shall see in the next chapter, this is indeed the case.