4. Selection with Respect to a Trait as a Type of Process and Its
4.2. The Causal Inefficacy of Population-Level Properties
In justifying the ordinary talk of selection in §2.3, I argued that every case of selection is a macro-causal process and affirmed that selection is a type of process. Now, in light of the above arguments, the latter affirmation has to be corrected: Selection with respect to a particular trait, as opposed to selection tout court, is a type of process, and selections with respect to different traits are different types of process. But, selection with respect to T1, for example, is nonetheless a process-type, and any token of it (and/or of selection with respect to any other trait as well) is still a macro-causal process.
So far, however, I have asserted the latter only in the minimal sense of a population-wide causal process as a collection of all and only those organismal causal processes each of which involves some organisms of a given population within a given generation.
Also, asserting this alone is far from specifying selection with respect to T1 as a (unique) type of macro-process, let alone a type of macro-causal process. By contrast, Millstein’s account of selection, that selection with respect to T1 is such process that variation-in-T1
is causally responsible for variation in reproductive success with respect to T1, is obviously over and above that. It is essentially about the type. It effectively specifies selection with respect to T1 as a type of macro-causal process. It asserts that a token of selection with respect to T1 is a macro-process in the sense of involving a population of organisms that instantiates the population-level property variation-in-T1 and having the collective reproductive results that exhibit the collective feature variation in reproductive success with respect to T1. And it is tantamount to claiming that a token of selection with respect to T1 is a causal process in virtue of its constitutive property-token variation-in-T1’s being causally responsible for the tokening of variation in reproductive success with respect to T1 in its collective reproductive results.
Now, if variation in T1 is not properly considered a property, then selection with
respect to T1 cannot be a type of macro-process in Millstein’s sense of a population-level process. If it is indeed a property, but its instantiation is not causally efficacious with respect to the occurrence/tokening of variation in reproductive success with respect to T1, then selection with respect to T1 cannot be the type of causal process intended by her. In this section, we assume that variation in T1 is a (population-level) property and proceed to ask whether it is a causally efficacious property. We’ll see that such a supposed property has no causal efficacy at all, in consideration of the exclusion-style argument again, or something very akin to it. So, whilst it may be fine to characterise selection with respect to T1 in terms of the “variation in T1—variation in reproductive success with respect to T1” generalisation/law, it is incorrect to regard the latter as a causal generalisation/law or, which is the same, to say that tokens of selection with respect to T1 are (population-level) causal processes in virtue of the respective instances of variation-in-T1 involved being causally responsible for the respective tokening of variation in reproductive success with respect to T1.
We begin by first considering Haug’s (2007) defence of Millstein’s view against Matthen and Ariew’s formal-pattern account. Continuing their talk of selection as being
“realised”, Haug proposes a notion of “causal realisation” of process and claims that under that notion, tokens of selection (with respect to a particular trait, say T1) are causal processes in virtue of their being tokens of selection in Millstein’s sense of a type of population-level causal process. The notion of causal realisation is based upon Gillett’s (2003) “dimensioned” view of realisation, by which a higher-level property is realised by a set or combination of lower-level properties and relation(s) just in case all objects instantiating the former property instantiate it in virtue of their (proper) parts or (micro-)constituents instantiating the latter properties and having the latter relation(s).
This is expanded by Haug into his notion of causal realisation: A (macro-)causal
process that is a (macro-)causal connection between a higher-level property-instance and another, is said to be causally realised by a plurality of (micro-)causal processes that are each a (micro-)causal connection between a lower-level property-instance and another, just in case the two higher-level property-instances that are the cause-term and the effect-term, respectively, of the first (macro-)causal connection are realised, respectively, by the two combinations of lower-level property-instances that are the cause-terms and the effect-terms, respectively, of the latter set of (micro-)causal connections. He explicitly asserts that a population’s variation-in-T1 is realised by the combination of its members’ T1-determinates, and presumably holds that a population’s variation-in-reproductive-success(-with-respect-to-T1) is realised by the combination of its members’ respective degrees of reproductive success. These, together with his idea of causal realisation of process, allow him to say that a token of selection with respect to T1, a population-level causal process that is a causal connection between an instance of variation-in-T1 and an instance of variation-in-reproductive-success-with-respect-to-T1, is causally realised by a plurality of organismal-level causal processes, each of which is presumably a causal connection between the T1-determinate of an organism (that is a member of the given population) and its degree of reproductive success. This, then, is taken by him as having established that, in spite of the fact that tokens of selection with respect to T1 are realised by sets of organismal-level causal processes, they are causally realised and therefore are themselves population-level causal processes.
Against Haug’s argumentation I raise three criticisms. First, he totally misses Matthen and Ariew’s idea of selection, yet takes for granted their misguided talk of realisation. He apparently thinks that Matthen and Ariew maintain that selection as normally understood is not a causal type because it is realised. This is certainly wrong.
As we see in chapter two, when Matthen and Ariew say that selection is realised in
physically different substrates, what they mean by “selection” is a formal pattern, not a (supposedly) causal type. They never claim that selection as ordinarily understood is multiply realised. I have shown that there is nothing functional in their idea of selection, and thus their talk of realisation is actually an abuse of that notion. And the reason that selection as understood by them is not a causal type lies in the fact that the content of their notion of selection is an arithmetic theorem, a reason that they recognise and has nothing whatsoever to do with realisation. As regards the conventional talk of selection as a causal process, however, it is their fault to think that it is inconsistent with the constitution thesis, that a case of selection is wholly constituted by (certain) organismal-level occurrences. But they nonetheless don’t affirm that a case of selection is not a causal process because it is realised, as opposed to being constituted, by a collection of organismal-level occurrences, or causal processes. So, their (mistaken) reason against the causal-process view of selection as normally understood does not hinge upon their (misguided) talk of realisation, either. Taken as it is, then, Haug’s argument is not quite a defence against Matthen and Ariew, contrary to what he himself conceives.
Second, Haug’s own talk of realisation, while different from Matthen and Ariew’s, is misguided too. In the standard talk, whatever is realisable is multiply realisable and is a functional property, whatever realises a functional property is itself a single (physical) property and different realisers of the same functional property are mutually distinct properties. It is fine to speak of “process realisation” in Haug’s sense since it is nothing but a definition that builds upon the common talk of property realisation. We may also grant the dimensioned view of realisation upon which his whole talk of selection as realised process depends (since, as Kim himself says, one can freely define “realisation”
so long as the resulting notion is useful in elucidating philosophical problems). Yet, even if trait-variations and variation in reproductive success are deemed properties, they
are by no means functional properties. Then in exactly what sense can a property like variation-in-T1 be said to be multiply realisable? That is, what distinguishes between two supposedly different realisers of it?
Haug just leaves this fundamental question unanswered because, when talking about variation-in-T1’s realiser, he is talking about token-combination of property-tokens instead of type of combination of property-tokens. Any realisation relation is essentially a type-type relation. It’s alright to take realisation as a token-token relation at bottom, but then what is important is realisation type, and to identify a realisation type is to identify the pair of types each pair of which is a realised-realiser pair. So, token-realisers of (tokens of) the same supposedly multiply realisable type must themselves be subsumed under different types in order for the latter to be multiply realisable. Given that every token-realiser of variation-in-T1 is a combination of tokens of two or more distinct T1-determinates (that are possessed by organisms of the population within the same generation), the question translates into: Under what different types are subsumed such different token combinations of property-tokens? Do two such combinations that do not comprise tokens of exactly the same T1-determinates belong to different types?
Do two such combinations that comprise tokens of exactly the same T1-determinates but not exactly the same numbers of tokens for all those T1-determinates belong to different types? There just doesn’t seem to be any set of different laws/regularities or different types of explanation that requires us to recognise the allegedly different types so differentiated. On top of that, there is an obvious sense in which all such combinations of property-tokens belong to the same type: They are all combinations of tokens of two or more distinct T1-determinates. Therefore, unless there is any cogent reason to the contrary, variational properties like variation-in-T1 are not multiply realisable and hence are not realised properties at all. Then, by Haug’s definition, variational process-types
like selection with respect to T1 are not realised process-types, either. And soon we’ll see that the relation between variation-in-T1 and any definite combination of tokens of multiple T1-determinates is actually something other than realisation.
Third, whether or not population-level properties like variation-in-T1 are realised properties, Haug really fails to establish that they are causally efficacious properties in the face of a version of the exclusion problem which he himself recognises. He admits that organismal-level occurrences are causally sufficient for the tokening of variation in reproductive success with respect to T1, and hence that some organismal property-tokens, among them the most important ones being organisms’ different T1-determinates (tokens), together with some environmental and background conditions, are causally sufficient for the tokening of the latter. Also, he affirms that variation-in-T1 is not identical to any specific combination of particular numbers of tokens of particular T1 -determinates, and that a token of variation-in-T1 is distinct from the co-present token-combination of tokens of plural T1-determinates that he says realises the former. And, whilst it is wrong to apply the notion of (dimensioned) realisation to variation-in-T1 and a combination of two or more different T1-determinates, no doubt the instantiation of variation-in-T1 by a population of organisms depends upon the instantiations of two or more, but no matter which, of the different T1-determinates by its members (regardless of the number of instances of each T1-determinate instantiated). Haug recognises that the three assertions above, plus the exclusion principle or some equivalent, make up a version of the exclusion argument against the causal efficacy of variation-in-T1. He quickly dismisses it, however. He believes that properties like variation-in-T1 “do not compete with their realizers for causal sufficiency because the properties they involve are abstract ‘logical parts’ of the properties that realize them (in the same way that being red is part of being scarlet […])” (Haug 2007: 439). How is this to be taken and is it
plausible at all?
In so far as he applies “being a logical part of” both to variation-in-T1 and a definite combination of tokens of multiple T1-determinates and to being-red and being-scarlet, his idea of a property being a “logical part” of another or a combination of some others is essentially Yablo’s (1992) and Schlosser’s (2006) idea that a mental property or a determinable is “metaphysically included” in, or is a “metaphysical part” of, a certain subvenient/realiser physical property of that mental property or of a certain determinate of that determinable. The latter idea is originally intended to apply to any pair of distinct properties the instantiation of one of which depends upon the concurrent instantiation of the other. It is held that a property that is a “metaphysical part” of another cannot compete against the latter for causal efficacy, just as a determinable cannot compete against any of its determinates for causal efficacy. Grant that it is legitimate to broaden the whole talk of metaphysical inclusion in a way parallel to the dimensioned view of realisation, so that it is also applicable to a higher-level property and a distinct but depended combination of lower-level properties/relation instantiated by the constituents of the possessors of that higher-level property. This way, Haug can have recourse to the talk of metaphysical inclusion, maintaining that variation-in-T1 is a metaphysical/logical part of any definite combination of tokens of multiple T1-determinates and so cannot compete against the latter for causal efficacy.
But the truth is that the talk of metaphysical inclusion cannot dissolve any exclusion-style argument, for it offers virtually no reason as to why a metaphysically included property and any including counterpart property or property/relation-combination cannot compete for causal efficacy. In fact, it just stipulates that the latter is the case. It is also begging the question since metaphysically included properties are nothing other than the targets of the exclusion-style argument. And it is ad hoc because it amounts to
saying that the exclusion principle is not to cover any property that is distinct from but dependent upon another property or property/relation-combination. Further, so long as determinables are reckoned as prima facie causal properties, it is thoroughly wrong to presume that a determinable and any of its determinates do not compete for causal efficacy. Consequently, the analogies or comparisons with such a “model case”, be they appropriate or not, cannot serve the purpose of sustaining or motivating the idea that physicalist mental properties, realised functional properties or properties like variation-in-T1 do not suffer from the exclusion problem. So, contrary to what Haug believes, invoking the talk of metaphysical inclusion and comparing variation-in-T1 to being-red are far from being able to secure the causal efficacy of population-level properties from the threat of the exclusion problem.
Now, granted that variation-in-T1 is a property of a population, it is undoubtedly a higher-level property relative to the T1-determinates of organisms. Since it is not a functional property and, as I have specifically argued, not a realised property, it should be a higher-level physical property. In §3.4, we observed that higher-level physical properties are not affected by the exclusion-style argument. Yet, on the other hand, we’ve also seen that variation-in-T1 is not causally efficacious with respect to variation in reproductive success with respect to T1, since definite combinations of tokens of multiple T1-determinates are and they are each distinct from but depended upon by variation-in-T1. This is certainly an exclusion problem, which Haug attempts but fails to dismiss. But isn’t variation-in-T1 a higher-level physical property and, consequently, immune to the exclusion problem?
No, it indeed has the exclusion problem. However, that isn’t because it is a realised property (which it is not) or a higher-level property (which is irrelevant). The true reason is that it is treated as essentially a determinable. What Millstein calls
population-level properties, viz. variations in particular traits and variations in reproductive success with respect to particular traits if the latter variations are also deemed properties, are explicitly taken by Haug as determinables. This seems a very natural interpretation. For the moment suppose it is the only plausible one. Then, each such property is mere variation in a certain organismal determinable amongst organisms of a population, as opposed to a determinate variation that is essentially an exact number distribution of organisms of a population by the different determinates of that organismal determinable.
On the assumption that determinables are properties, no determinables are causally efficacious properties. This is the case because determinates alone are causally sufficient for any effects, or accepting them alone as causally efficacious properties are sufficient for accounting for all causal powers, and any determinable is distinct from any of its determinates yet its instantiation is nonetheless dependent upon the instantiation of any of its determinates. To avoid overdetermination of effects or of causal powers, then, we should not grant causal efficacy to determinables. This, clearly, is an exclusion-style argument against the causal efficacy of determinables. It affects any determinable, whether it is a functional or physical one and no matter of which level it is, for being a determinable is a matter over and above these. The so-called population-level properties such as variation-in-T1 are determinables apart from being higher-level physical properties, and are causally inefficacious because, and only because, they are determinables.
With the understanding that variation-in-T1 is a determinable, we can now see clearly the correct relation between it and a definite combination of tokens of multiple T1 -determinates: They are related as a higher-level physical determinable and what may be called a lower-level determinate of it or, better, a combination of lower-level physical properties that is identical to a determinate of it that is of the same level as it is. Haug
apparently mistakes this determinable-determinate plus level relation for the trans-level realisation relation, thus misapplying the dimensioned view of realisation to the likes of variation-in-T1. There are indeed several similarities between a higher-level determinable and a higher-level functional property. Like a higher-level determinable, a functional property is also distinct from yet dependent upon any of the properties, or combinations of properties/relations, the instances of which are causally sufficient for all the effects its instances are presumed to cause, or are sufficient for accounting for all the causal powers its instances are supposed to account for. Accordingly, both of them are targets of the exclusion-style argument and are hence causally inefficacious. And they are both so independent of their being higher-level properties and regardless of whether their causal rivals (realisers and determinates, respectively) are of the same level as they are or of a lower level. We can always identify the realisers of a functional property or the determinates of a determinable at the same level as the functional
apparently mistakes this determinable-determinate plus level relation for the trans-level realisation relation, thus misapplying the dimensioned view of realisation to the likes of variation-in-T1. There are indeed several similarities between a higher-level determinable and a higher-level functional property. Like a higher-level determinable, a functional property is also distinct from yet dependent upon any of the properties, or combinations of properties/relations, the instances of which are causally sufficient for all the effects its instances are presumed to cause, or are sufficient for accounting for all the causal powers its instances are supposed to account for. Accordingly, both of them are targets of the exclusion-style argument and are hence causally inefficacious. And they are both so independent of their being higher-level properties and regardless of whether their causal rivals (realisers and determinates, respectively) are of the same level as they are or of a lower level. We can always identify the realisers of a functional property or the determinates of a determinable at the same level as the functional