1. The Causal Exclusion Argument
Jaegwon Kim (1998, 2000, 2005) has argued that mental properties are epiphenomenal if they are not, in some way, reducible to physical properties. The target of his so-called Causal Exclusion Argument is any position that simultaneously affirms the following four claims: (i) the mental is ontologically distinct from and irreducible with respect to the physical, (ii) the mental is causally efficacious, (iii) the physical world is causally closed – every physical event at time t has a physical cause at t, and (iv) no single event has more than one sufficient cause unless it is an instance of overdetermination. I will call these commitments DISTINCTNESS, EFFICACY, CLOSURE and EXCLUSION respectively. Kim believes that these commitments cannot be simultaneously affirmed and argues, in short, that all the causal work in our world is done by the physical and there is nothing left for the mental to do.
To make this a bit more concrete consider a physical event X, say the rising of my arm. I would like to think that it was caused by the instantiation of a mental property M1, say my desire to raise my arm. Since we know that X is a physical event and the physical world is causally closed it follows that there is a physical cause for X, say the instantiation of some physical property P1 in my brain. Can the instantiation of M1 and P1 both be causes of X? If we assume that mental and physical properties do not overdetermine their effects it seems, according to EXCLUSION, that one of the putative causes must be excluded. Kim goes on to argue, of course, that M1 and not P1 should be excluded. But why think that M1 and P1 do not overdetermine X? Kim argues:
Apart form the implausible consequence that it makes every case of mental causation a case of overdetermination, this approach encounters two difficulties: first, in making a physical cause available to substitute for every mental cause, it appears to make mental causes dispensable in any case; second, the approach may come into conflict with the physical causal closure. (Kim, 2000, p. 45)
Kim rejects overdetermination for at least three reasons: (i) it is implausible, (ii) it makes the mental cause dispensable, and (iii) it violates CLOSURE. Though I am unconvinced by any of these reasons I will only develop a response to (iii) in what follows.
Kim’s third reason is that the overdetermination option violates CLOSURE – or at least it violates CLOSURE in relevant nearby worlds. On the assumption that nearby worlds are ones that are as similar to the actual world as possible in terms of the events that actually obtain in them, it seems that nearby worlds where one of the overdetermining causes is missing should nevertheless be worlds that include the effect. Consider a burning warehouse. In the actual world there is an effect E, the burning warehouse, that is overdetermined by two causes, C1 and C2, a lightning bolt and a cigarette bud. How should we characterize the nearest worlds where C1 does not obtain? These worlds, it seems, are worlds where C2 and E still occur since C2 is not tied to C1 in any way and C2 remains a sufficient cause for E. That is, the nearest non-C1 worlds are worlds where the relevant cigarette bud still causes the warehouse fire.
Returning to the mental case Kim argues that we ought to characterize the nearest non-P1 worlds in similar fashion. Since we’re assuming that nearby worlds are as similar to the actual world as possible in terms of the events that actually obtain in them, it follows that the nearest non-P1 worlds would still include the instantiation of M1 and X. Given that X is overdetermined by P1 and M1 in the actual world it seems that X will also obtain in the nearest non-P1 worlds. Therefore, the nearest non-P1 worlds are worlds where M1 still exists and continues to be a sufficient cause for X. This, Kim argues, is an undesirable result for it implies that a minor counterfactual change in the world, namely the absence of P1, would entail a violation of CLOSURE since we are forced to accept the existence of a physical event, X, which has, as its only cause, something non-physical, M1. That is, the nearest non-P1 worlds will violate CLOSURE.
Here it seems that Kim makes a blunder. As some have pointed out, the nearest non-P1 worlds are nevertheless worlds where supervenience obtains. It follows that M1 will still have a subvenient base. Though P1 may be absent in the nearest non-P1 worlds it will be replaced by a similar physical property, say P*. Consequently, in nearby non-P1 worlds P* will be a sufficient cause for X. So long as the nearest non-P1 worlds are supervenience worlds there will still be some physical property with equal claim on being a sufficient cause for X. In this way CLOSURE is not compromised.
In response Kim concedes this point but argues if CLOSURE is not violated because of supervenience then the very meaningfulness of the claim that M1 and P1 overdetermine X is threatened. If all the nearby non-P1 worlds are worlds where supervenience holds then it is difficult to understand what it could mean to say that X is the result of overdetermination. That is, the attribution of overdetermination is vacuous since the putative causes, M1 and P1, cannot come apart. Consequently, we would not know what it could mean for both putative causes to be independently causally sufficient for the effect. He writes:
… if of necessity any [non-P1] world is ipso facto a [non-M1] world – what significance can we attach to the claim that P1 and M1 are each an overdetermining sufficient cause of [X]? That in addition to P1, M1 also is a sufficient cause of [X]? (Kim, 2005, p. 46)
We might capture Kim’s reasoning in terms of the following dilemma:
(D1) It is either the case that the nearest possible non-P1 worlds are (i) worlds where P1 is simply absent or (ii) worlds where P1 is replaced by a different physical property P*.
(D2) If (i) is the case then the nearest possible non-P1 worlds violate CLOSURE.
(D3) If (ii) is the case then the overdetermination option is vacuous.
Both horns are undesirable so, Kim argues, the overdetermination option must be abandoned. Since I am reasonably convinced of (D2) my goal is to undermine (D3).
To begin mounting a response it is important to get clear on Kim’s logic behind (D3). His comments seem to be grounded in a straightforward counterfactual test for identifying instances of overdetermination. Suppose there are two putative causes, C1 and C2, of a single effect, E. The truth of the following two counterfactual conditionals seems to secure C1 and C2’s being overdetermining causes of E.
(CF1) If C1 had happened without C2 then E would still have happened.
(CF2) If C2 had happened without C1 then E would still have happened.
The joint truth of (CF1) and (CF2) ensures that there is ‘significance’ to the claim that both C1 and C2 are each overdetermining sufficient causes of E for it demonstrates that C1 and C2 are causally sufficient for E independent of each other. C1 would still cause E in the absence of C2 and vice versa. There are, perhaps, general worries as to the sufficiency of this test for determining instances of overdetermination. For even if the counterfactual conditionals come out true there may be certain principled reasons for denying that C1 and C2 are causes of E. These worries depend, perhaps, on deep and controversial analyses of causation that are beyond the scope of this essay. Setting these worries aside, I think it is safe to say that the joint truth of (CF1) and (CF2) is at least a necessary condition for C1 and C2 to be overdetermining causes of E.
Technically speaking, however, there are two different ways conditionals can come out true. Generally, we take a conditional to be true if both the antecedent and the consequent are true. But a conditional is also true, regardless of the truth value of the consequent, if the antecedent is false. To distinguish these we might say that a conditional is vacuously true when it is true because its antecedent is false and we might say that a conditional is non-vacuously true when it is true because its antecedent and consequent are both true. This distinction is relevant because in the mental case we have two putative causes that are inseparable. As a result putting them through the counterfactual test will result in both (CF1) and (CF2) coming out true since their antecedents will be false. So the counterfactual test, as it stands, cannot be what Kim has in mind. To tighten the counterfactual test to match Kim’s intuitions we must say that the joint non-vacuous truth of (CF1) and (CF2) is a necessary condition for overdetermination. Since (CF1) and (CF2) cannot come out non-vacuously true in the mental case Kim claims that M1 and P1 cannot overdetermine E. He writes:
The usual notion of overdetermination involves two or more separate and independent causal chains intersecting at a common effect. Because of Supervenience, however, that is not the kind of situation we have here. In this sense, this is not a case of genuine causal overdetermination, and Exclusion applies in a straightforward way. (Kim, 2005, p. 48)
There is something intuitively satisfying about this way of construing overdetermination. After all, unless the putative causes are independent and separable there are no possible worlds to verify that the putative causes are, in and of themselves, independently sufficient.
Let’s assume that Kim is right and agree that overdetermination loses any substantive meaning in light of the fact that M1 and P1 can never come apart. The only way, it seems, to make sense of the independent causal sufficiency of M1 and P1 is to see if M1 or P1 could cause X independently of each other. That is, some form of verifiability is critical to Kim’s assessment of overdetermination. But if some form of verifiability is essential to understanding causality, doesn’t this put pressure on CLOSURE itself? How, one might ask, are we to verify that every physical event at time t has an independent, sufficient physical cause at t if the physical causes we are interested in can never occur without their non-physical counterparts? Given Kim’s commitment to supervenience we know that P1 can never occur without the instantiation of M1. But if they can never come apart how can the claim that P1 is independently, causally sufficient for X have any meaning? If the claim that mental properties overdetermine certain physical events is meaningless in light of the necessary co-extensiveness of mental and physical properties it seems the same can be said for the relevant physical properties. How are we to make sense of the causal sufficiency of P1 since it is necessarily co-extensive with M1?
Another way to look at this, perhaps, is to ask what it could mean for Kim to claim that M1 is causally epiphenomenal with respect to X? A straightforward test might be captured with the following two counterfactual conditionals:
(CF3) If C1 had happened without C2 then E would still have happened.
(CF4) If C2 had happened without C1 then E would not have happened.
If we have two putative causes, C1 and C2, of a given effect, E, and ceteris paribus the presence of C2, given the presence of C1, is irrelevant to the presence of E it seems that we have a way of deciding whether or not C2 is epiphenomenal with respect to E. At the very least the non-vacuous truth of (CF3) and (CF4) seems to be a plausible necessary condition for C2’s being epiphenomenal with respect to E. I realize, as with the counterfactual test for overdetermination, that the counterfactual test for epiphenomenalism is not without its difficulties. But if we can borrow the logic of Kim’s rejoinder it seems we have a way of deflecting epiphenomenal worries regarding M1. Unless (CF3) and (CF4) come out non-vacuously true in the mental case it seems any claim that M1 is epiphenomenal with respect to X is robbed of any substantive meaning. At best Kim’s rejoinder results in a stalemate and there is no decisive reason, following Kim’s own logic, to think that positing M1’s causal sufficiency for X is any less vacuous than positing M1’s being epiphenomenal with respect to X.
5. Additional Worries
I think, however, the situation is worse than it first appears because of Kim’s commitment to multiple realization. If you recall, Kim argued that nearby non-P1 worlds are worlds where the instantiation of M1 and X still obtain. Though he initially considered this a violation of CLOSURE he later conceded that CLOSURE would still be preserved assuming nearby non-P1 worlds were supervenience worlds. Kim’s acknowledgement of this point, it is safe to say, is grounded in his commitment to multiple realizability for he goes on to admit that non-P1 worlds still contain some physical property, P*, that M1 supervenes on. His conclusion is that although the causal chain from M1 to X coincides with the causal chain from P1 to X in the actual world the causal chain from M1 to X in the nearest possible non-P1 worlds must coincide with the causal chain from P* to X. He writes:
To be a cause of [X], [M1] must ride piggyback on the physical causal chains – distinct ones depending on which physical property subserves [M1] on a given occasion, in the same world or in other possible worlds. (Kim, 2005, p. 48)
If we carefully revisit the counterfactual test for overdetermination mentioned above it is not difficult to see, contrary to Kim’s implicit commitments, that only one of the counterfactuals comes out vacuously true. If we substitute C1 and C2 with M1 and P1 and E with X we get the following counterfactuals:
(CF5) If M1 had happened without P1 then X would still have happened.
(CF6) If P1 had happened without M1 then X would still have happened.
It is evident that (CF5) would, even according to Kim, come out true and not merely vacuously true. While it is clear that Kim requires that both counterfactuals be non-vacuously true I think Kim’s concession puts the causal relevance of M1 in a favorable light with respect to P1. At least there are possible non-P1 worlds where instantiations of M1 are followed by X. There are, however, no non-M1 worlds where instantiations of P1 are followed by X since there are no non-M1 worlds where P1 is present. Consequently we can give meaning to the claim that M1 is a cause of X independently of P1 but not vice versa.
Furthermore, a necessitarian dualist, as opposed to a non-reductive physicalist, can point out that supervenience is an asymmetric relation. Yes, the necessitarian dualist is committed to the presence of certain mental properties given the presence of the relevant physical properties. The necessitarian dualist is also committed to the presence of the relevant physical properties given the presence of certain mental properties in all nomologically possible worlds. But the necessitarian dualist is not committed to the presence of the relevant physical properties given the presence of certain mental properties tout court. In fact the necessitarian dualist is not committed to the presence of any physical properties at all given the presence of mental properties. There are, after all, possible worlds comprised purely of ectoplasmic entities that contain mentality in the absence of anything physical. Moreover there is no reason why we should deny the possibility of worlds with a mixture of physical and non-physical particulars, properties, and relations that enjoy inter-domain causal transactions. In such worlds M1 may cause X in the absence of any subvening physical properties whatsoever. A world like this would surely violate CLOSURE but it is not at all clear that it violates supervenience construed as an asymmetric relation. After all, most physicalists grant such possibilities. In fact the acknowledgment of such possibilities has played a critical role in shaping the ‘standard’ formulation of physicalism. The ‘standard’ formulation was designed to make the conservative claim that the truth of physicalism only entails that any minimal physical duplicate (and not any physical duplicate) of the actual world is a duplicate simpliciter.
Perhaps nearby non-P1 worlds are worlds where M1 supervenes on some physical property but it is clear that there are more distant non-P1 worlds where M1, unaccompanied by any physical properties whatsoever, continues to be a sufficient cause of X. If anything this raises worries, given Kim’s commitments, over the causal relevance of certain physical properties. At least there are non-P1 worlds where instantiations of M1 are still followed by instantiations of X. There are, however, no non-M1 worlds (not even extremely distant non-M1 worlds) where instantiations of P1 are followed by X. There is, then, a reasonable sense in which M1 is an independent sufficient cause for X. At least M1 is a candidate for such a designation. It is clear, however, that P1 cannot receive such a designation since P1 will, of necessity, be accompanied by M1. (CF5) can be non-vacuously true but (CF6) can only be vacuously so.
It seems to me that there is a way of rebutting Kim’s third reason against the overdetermination response to the Causal Exclusion Argument. Given my skepticism over Kim's first two reasons it is not unreasonable to think that mental causes overdetermine certain physical events.
Crisp, Thomas and Ted Warfield. (2001) Kim’s Master Argument. Nous 35: 304-316.
Jackson, Frank. (2000) From Metaphysics to Ethics: A Defense of Conceptual Analysis. Oxford: Oxford University Press.
Kim, Jaegwon. (1995). Supervenience and Mind. Cambridge: Cambridge University Press.
Kim, Jaegwon. (1998). Philosophy of Mind. Boulder, Colorado: Westview Press.
Kim, Jaegwon. (2000). Mind in a Physical World. Cambridge, Mass.: MIT Press.
Kim, Jaegwon. (2005). Physicalism or Something Near Enough. Princeton: Princeton University Press.
Stoljar, Daniel. (2010). Physicalism. New York: Routledge.