Supervenience Argument

The further problem is that the Identity Condition does not appear to be consistent with Exclusion—at least not if we understand “overdetermination” in a particular way, which seems to us natural. Exclusion implies that if an event has two distinct synchronous causes, then it is overdetermined by them. Brutus’s killing of Caesar (call this event “BKC”) and Brutus’s murdering Caesar (call this “BMC”) are synchronous events that have many common effects. But clearly BKC and BMC cannot overdetermine their common effects if we assume, as we have done, that overdeterminers must be independent at least in that each could have occurred without the other (the murder could not have occurred without the killing). Now, because murdering and killing are two distinct properties (relations), the Identity Condition implies that BKC ≠ BMC. But because BKC and BMC are synchronous events with at least one common effect, and BKC and BMC do not overdetermine any of their effects, Exclusion implies that BKC = BMC.²⁴ This is not an outright logical inconsistency: no contradiction follows from the set {Exclusion, Identity Condition} alone. To avoid having to choose between Exclusion and the Identity Condition, and thus giving up his supervenience argument, Kim can argue that: (a) BKC and BMC are not synchronous, (b) BKC and BMC have no common effects, (c) BKC and BMC do, after all, overdetermine all of their common effects, or (d) the property of killing is the same as the property of murdering. None of these seems very promising to us.

There is, however, a fifth alternative: (e) Kim could reply that the event description notation “[x, P, t]” used in the formulation of the Identity Condition is not to be understood in terms of the commonsense idea of an object having a property at a time, but in terms of a technical notion of a “constitutive property” of an event. The idea would be that the “P” in “[x, P, t]” always denotes a special, unique property “constitutive” of the event denoted by “[x, P, t]”, and that not every property x has at t can be a constitutive property of an event. It should be clear how this distinction would enable Kim to dodge the objection about BKC and BMC: the application of the Identity Condition in our argument for the distinctness of BKC and BMC would simply be invalid, because the argument does not have premises stating that killing and murdering are constitutive properties and hence we are not entitled to conclude that there exist any such events as [Brutus, Caesar, kills, t] and [Brutus, Caesar, murders, t] for us to apply the Identity Condition to. (Of course, if the Identity Condition and Exclusion are both to be maintained, it had also better be the case that at least one of killing and murdering is not a constitutive property, or else the argument could be supplemented with additional true premises to yield a refutation of the Exclusion/Identity Condition combination. We do not, however, have any arguments to show that murdering and killing are constitutive properties.)

We believe that Kim would choose alternative (e) for coping with the present problem—the distinction between constitutive and other properties of events is, in fact, drawn by Kim in his 1976. Although in that paper Kim resists the identification of such events as BKC and BMC,²⁵ he is open to the identification of other events involving the instantiation of distinct properties by an object.²⁶ But choosing alternative (e) comes at a price: To begin, Kim would have to reject the principle that [x, P, t] exists if and only if x has P at t (an unrestricted form of what Kim 1976, 35, calls the “existence condition” for events).²⁷ But if this principle is rejected, the meaning of Kim’s event description notation is no longer clear: to understand it we would need an account of just what it is that makes some properties “constitutive” and others not, and none has been provided by Kim. We do not claim that such an account cannot be provided—the salient point is, rather, that for anyone who is worried about the coherence of the Exclusion/Identity Condition combination, as we are, the acceptability of Kim’s supervenience arguments will depend on his ability to produce such an account.

What’s more, if Kim chooses (e), a new logical problem is introduced into the argument. The problem is that option (e) renders not only our argument against the coherence of the Exclusion/Identity condition combination but also Kim’s own supervenience arguments invalid, for exactly the same reason: if the Identity Condition is to be applied to m, M, p, and P in the supervenience arguments, the arguments will require two additional premises to remain valid: 1) that M is a constitutive property of m, and 2) that P is a constitutive property of p. Perhaps it will turn out that both claims fall out of the correct account of constitutive properties—even though since writing his 1976, Kim has indicated that, on pain of having to revise his property-exemplification account of events, he may have to deny that mental properties can be constitutive properties of events!²⁸ However that may be, it should be clear that if option (e) is chosen, some further work on Kim’s part will be required to make the supervenience arguments convincing.

8. A problem about mental quausation?

Kim is well aware, of course, that there are token physicalists who are not type physicalists. Does he think his supervenience argument has anything to say to them? Surprisingly, he does. In footnote 9 in ch. 2 of PSNE and elsewhere (e.g. Kim 1998, ch. 4) Kim hints that his argument could be reconstructed so as to have bite against nonreductive token physicalist positions. Unfortunately, however, Kim never gives us an explicit reconstruction. Though it would be obviously unfair to criticize an argument never explicitly formulated, we would still like to register the reason for our scepticism that a plausible argument along the lines Kim hints at can be made. Kim says (PSNE, 42, n. 9) that in the reconstructed argument,