Very few opponents of realizer-state functionalism deny this claim. Very few opponents try to argue that the semantics of realizer-state functionalism doesn’t work. What they argue instead is that realizer-state functionalism cannot account for multiple realizability construed in a metaphysically robust way – that is, construed as involving a realization relation that obtains between distinct natural properties.211 These opponents are right, I have argued: realizer-state functionalism does deny the existence of natural and multiply realizable mental properties. Demonstrating that there is some distinct semantic conception of multiple realizability with which realizer-state functionalism is compatible does nothing to address this point. Realizer-state functionalists would be far better served by simply granting their opponents’ claim that their view is incompatible with (the metaphysical conception of) multiple realizability and then trying to demonstrate that this doesn’t add up to a decisive objection against their view.
Maybe realizer-state functionalists can even take the present point and work it up into an argument in their favor. Presently, it’s not entirely clear what the metaphysical relation of realization amounts to exactly. As I mentioned above, a number of competing accounts of the relation have been offered by philosophers, and none have yet gained widespread acceptance. Perhaps it’s a virtue of realizer-state functionalism that it doesn’t require a metaphysically robust realization relation. Realizer-state functionalist don’t need such a relation since on their view all natural mental properties are (contingently) identical to natural physical properties, not realized by them. Unlike realization, we presently have a pretty good understanding of the identity relation. Whether or not an argument along these lines can be made very compelling, this response would at least genuinely grapple with the issue that the opponents of realizer-state functionalism raise.
I have not yet tried to show what the problem is with denying the existence of natural and multiply realizable mental properties. I have tried to show only that realizer-state functionalists are in fact committed to such a denial. We will use this as a launching point for my master objection to realizer-state functionalism in Chapter 7. First, though, I want to consider a few alternative forms of reductive physicalism.
6.3 Kim’s Disjunction Identity Theory
Kim is both a reductionist of sorts and a causal functionalist of sorts. Thus, his position inevitably bears certain important similarities to realizer-state functionalism, as several commentators have noted.212 However, there are also important differences between Kim and realizer-state functionalists. One difference centers around the distinction we have drawn between first order physical properties and second order causal-functional properties. Realizer-state functionalists accept this distinction, exploiting it, even, when they introduce causal-functional properties in their attempt to respond to the Common Feature Objection. Kim, by contrast, denies the distinction’s legitimacy. Kim writes that when we set aside multiple realizability concerns, if
F is the property of having some property that meets specification H [e.g., a specification of some causal role], and P is the property that meets H . . . [then] F is the property of having P. But in general, the property of having Q = property Q. It follows then that F is P.213 That this conclusion is highly contentious may not be immediately obvious. Even without relying on multiple realizability, there are different ways philosophers have tried to resist Kim’s proposed property identification. Ned Block argues that while P is an intrinsic first order property, F is an extrinsic second order property (as all causal-functional properties are). If this is right, then by the non-identity of discernibles it would seem to follow that P and F are non-identical.214 A second way is by arguing that contrary to what Kim asserts at the end of the above passage, it is not generally the case that the property of having Q = the property Q. According to Jackson, it is not the case, for instance, that the property of having the sky’s color = the sky’s color.215
For my purposes here, I’m not especially interested in whether such responses to Kim are ultimately right or wrong. I mention them only to help us get a handle on what Kim is claiming in the above passage, which requires that we see that his claim is not at all trivial. I am willing to provisionally grant Kim that when multiple realizability considerations are left aside, F would be identical to P. Less important than whether this identity really holds is Kim’s underlying thought here, which is that existentially quantifying over a domain of properties cannot literally produce a new property. To think otherwise would be to believe in “sheer magic,” according to Kim.216
If P is the only property meeting specification H, then the expression ‘having some property which meets H’ shouldn’t be regarded as a device for picking out some property which is distinct from P, according to Kim. It should be regarded instead as a device for picking out P in a new way, a “second order way” we might say. Thus, Kim: “it is less misleading to speak of second order descriptions or designators of properties, or second order concepts, than second order properties.”217 By extension, it is less misleading to speak of causal-functional descriptions or designators of physical properties, or causal-functional concepts, than causal-functional properties. But if there aren’t really distinct causal-functional properties, then there is no distinction to be drawn in the first place between causal-functional properties and physical properties.218
If F is identical to P (as I’m willing to grant) when we set multiple realizability considerations aside, what happens when we don’t set such considerations aside? What happens to F when H is a specification met by both P and P*, where these two properties are non-identical? Kim’s view on how to answer this question has evolved in recent years. To help us work out his present view, I want to begin by considering one potential answer to this question which Kim has rejected in earlier works, including even his recent (1998) Mind and World, but which he seriously entertains in his more recent (2002) précis to Mind in a Physical World: that of identifying F with the disjunction of those properties meeting H. Call this view the disjunction identity theory.
6.3.1 INTRODUCING THE DISJUNCTION IDENTITY THEORY219
The disjunction identity theory is a reductionistic view that merits attention for its own sake. However, we should bear in mind throughout our discussion of it that at no point does Kim fully commit himself to it. Some of my reasons for discussing the view, then, are instrumental: I hope to use the disjunction identity theory as a tool to get us clear on Kim’s core views – those to which he is fully committed.
Here is what a disjunction identity theorist would say about the pain example we have been considering. Pain is identical to the property of having some property that occupies pain’s characteristic causal role. This property is identical to the property which occupies that role if there is just one such property, or to disjunction of those properties which occupy the role if there is more than one. In our example, both PH and PM occupy pain’s causal role. If we now assume that no other property does so as well, then pain is identical to the disjunctive property (PH or PM).
Let’s now try to locate the disjunction identity theory with respect to the larger issues we have been considering throughout the chapter. Is the disjunction identity theory reductionistic? Well, if the set of physical properties is closed under the operation of property disjunction, then the disjunction identity theory entails that mental properties are identical (and so reducible) to certain physical properties – namely, to disjunctive ones. There may be ways to resist the claim that the set of physical properties is closed under disjunction, but I don’t want to challenge the point here. I want to grant that the claim is correct and thus that the disjunction identity theory is reductionistic.
6.3.2 A MODIFIED CAUSAL EXCLUSION ARGUMENT
Notice, though, that the sort of reducibility just conceded does not obviously ensure the causal efficacy of mental properties. Pain, for instance, will be causally efficacious only if (PH or PM) is. Many philosophers deny that disjunctive properties are ever causally efficacious however. This should make us sensitive to a possibility we had not previously considered: the identification of mental properties with physical properties does by itself guarantee the causal efficacy of those mental properties. What’s needed is that mental properties be identified with the right physical properties – that is, with causally efficacious ones. It’s not at all clear that the disjunction identity theory succeeds in doing this. Putting the point in terms of the six theses set out last chapter: the negation of (Irreducibility) does not entail the truth of (Mental Causes), even assuming the truth of the other four theses.
In connection, there is reason to worry that the disjunction identity theory is susceptible to a modified version of the causal exclusion argument. Consider the following two principles about disjunctive properties. First, disjunctive properties can never be instantiated in the absence of an instantiation of at least one of their property disjuncts. So for instance, (PH or PM) can never be instantiated by an entity that instantiates neither PH nor PM. This first principle, I take it, is absolutely uncontroversial. Second, there are no effects caused by the instantiation of a disjunctive property but not caused by the instantiation of one of that disjunctive property’s disjuncts. So, for instance, suppose that Marvin the Martian instantiates PM and thereby instantiates (PH or PM). According to this second principle, there won’t be any effects that are caused by Marvin’s instantiation of (PH or PM) but not by his instantiation of PM. This second principle is more controversial than the first, but still strikes me as acceptable.220
Consider then an instantiation of (PH or PM) by some entity x which causes an effect e. Given our first principle regarding disjunctive properties, it follows that x must also instantiate either PH or PM. Suppose x instantiates PM. Given our second principle about disjunctive properties, if e is caused by x’s instantiation of (PH or PM) then e must also be caused by x’s instantiation of PM. Now (PH or PM) and PH are of course non-identical properties. Does it make any sense to say that disjunctive properties are reducible to their property disjuncts? Does it make any sense to say, for instance, that (PH or PM) is reducible to both PH and PM, where PH and PM are non-identical? Let’s first suppose that this does not make sense. Then (PH or PM) is irreducible to PM. But then by (Competition) it follows that since e is caused by both an instantiation of (PH or PM) and an instantiation of PM, e is causally overdetermined. This conclusion will generalize to all effects of mental causes, which is inconsistent with the truth of (No Overdetermination). In this case the disjunction identity theory will have just as much of a problem with causal exclusion as do nonreductive forms of physicalism.221
Let’s now suppose that it does make sense to speak of disjunctive properties being reducible to their property disjuncts, and that in fact (PH or PM) is reducible to PM. Then (Competition) will be inapplicable here. We can’t use it to show that e is overdetermined if it is caused by both an instantiation of (PH or PM) and an instantiation of PM. The disjunction identity theory is not out of the woods just yet though. (Competition) provides only a sufficient condition for causal overdetermination, not a necessary one. It is fair for us to ask why we don’t have overdetermination in the present case, given the non-identity of (PH or PM) and PM.
A disjunction identity theorist does not adequately answer this question merely by pointing out that (PH or PM) is reducible to PM. If pain were identical to PM, it would be clear enough why there is no causal overdetermination. In the present case, though, there is reduction without identity. What is the nature of this non-identity reduction relation such that it is not causal overdetermination engendering? This is something that needs to be explained. Property reduction without identification is not well enough understood in advance that one can simply take for granted that it solves all causal exclusion difficulties.
6.3.3 INSTANTIATION IDENTITIES WITHOUT PROPERTY IDENTITIES
We can read Kim as in effect responding to our question in the following passage.
What I argue in Mind in a Physical World, perhaps not entirely explicitly, is that for the physicalist what happens with M, what we say about the status of M, doesn’t really matter . . . What’s important is the fact that every M [instantiation] is a Pi [instantiation], for some i.222 In saying that it doesn’t matter what happens to M, Kim partly means to be contrasting his view with those of realizer-state functionalists and other type-type identity theorists. According to Kim it doesn’t matter with respect to causal exclusion whether mental properties are identified with first order physical properties like PM, or whether they are identified with disjunctions of first order physical properties like (PH or PM), or even whether they are ultimately eliminated (more on this below). What matters is just that we are able to identify mental property instantiations with first order physical property instantiations whose causal efficacy is not in doubt.
There is a question, though, of whether anything short of identifying mental properties with first order physical properties will license the sort of property instantiation identifications that Kim says are needed. To see how the disjunction identity theory at least potentially could yield such property instantiation identifications, consider the following principle.
The Disjunctive Property Instantiation Principle (DPIP): The instantiations of disjunctive properties are numerically identical to the instantiations of their property disjuncts. If D is the disjunctive property (P1 or P2 or . . . or Pn), then every instantiation of P1, and every instantiation of P2, and . . . and every instantiation of Pn, will be numerically identical to an instantiation of D, and every instantiation of D will be numerically identical to either an instantiation of P1, or an instantiation of P2, or . . . or an instantiation of Pn.
Unless the (DPIP) is true, I don’t see how both the disjunction identity theory could be true and yet every M instantiation could be identical to a Pi instantiation for some i. Closely related to this point, once we assume that the (DPIP) is true we can see why there is no threat of causal exclusion in our case in which the effect e is caused by “both” a PM instantiation and a (PH or PM) instantiation: the given PM instantiation is numerically identical to the given (PH or PM) instantiation. Since there is just one property instantiation here, there is no threat of e being causally overdetermined. The upshot of this all is that if the sense in which (PH or PM) is reducible both to PH and to PM without being identical to either is just that every instantiation of (PH or PM) is numerically identical to an instantiation of PH or PM, as follows from the (DPIP), then we can see why this non-identity reduction relation isn’t causal overdetermination engendering. We can see how the disjunction identity theory manages to avoid causal exclusion troubles.223
Now at least at first glance, the (DPIP) is not completely implausible. If Marvin the Martian instantiates PM, it seems reasonable to say that there aren’t really two things going on: Marvin’s instantiation of PMand, in addition, his instantiation of (PH or PM). However, the (DPIP) is incompatible with the identity condition for events originally proposed by Kim. According to the original identity condition, events e and e* are identical just in case the individual, time, and property that are constitutive of e are identical to the individual, time, and property (respectively) that are constitutive of e*.224 So either the (DPIP) must be rejected or else Kim’s original identity condition for events must be amended. Consider then the following proposed amendment: events e and e* are identical just in case the individual, time, and natural property that are constitutive of e are identical to the individual, time, and natural property that are constitutive of e*. In an earlier writing, Kim has in fact entertained something like this amendment.225 If disjunctive properties are ipso facto unnatural – as is widely held, and as Kim himself holds even as he entertains the disjunction identity theory – then the (DPIP) is consistent with this amended version of Kim’s identity condition for events.
6.3.4 A NEW CAUSAL PROBLEM
The (DPIP) was introduced partly to respond to the causal exclusion worries we were raising for the disjunction identity theory. However, the (DPIP) gives rise to a new sort of causal problem. Kim holds, as many do, that the causal status of properties is derivative from the causal status of their instantiations: a property is causally efficacious only in the sense that its instantiations are causes. Assuming that this is correct, consider an event which is my alarm clock’s beeping at 7:00 a.m., and suppose that this event causes me to wake up. By the (DPIP), this event is numerically identical to the instantiations of the following properties by my alarm clock at 7:00 a.m.: that of (beeping or failing to beep), that of (beeping or failing to beep or being epiphenomenal), that of (beeping or failing to beep or being epiphenomenal or being a banana), and so on, for each of the infinitely many disjunctive properties that can be constructed having the property of beeping as one of their property disjuncts. If, as we are supposing, the causal status of properties is derivative from the causal status of their instantiations, then there seems to be nothing to distinguish the causal status of the beeping in this case from that of these various disjunctive properties. This is counterintuitive though: my alarm clock’s beeping seems to directly causally matter to my waking up in a way that its (beeping or failing to beep or being epiphenomenal or being a banana) does not.
Here we are encountering a special case of a general problem faced by proposed solutions to the causal exclusion problem. Pre-theoretically, an ideal account of mental causation would establish the causal efficacy of mental properties without generalizing to establish the causal efficacy of a great many other properties which aren’t obviously causally efficacious. In other words, the trick is to devise an account that bestows the honor of causal efficacy on mental properties without cheapening the honor by bestowing in on a motley crew of other properties as well. The disjunction identity theory seems to be unable to do this. If the property pain has no better causal status vis-à-vis my wincing than the property (beeping or failing to beep or being epiphenomenal or being a banana) has vis-à-vis my waking up, this by itself seems to be a blow to mental causation. If the principle that the causal status of properties is derivative from the causal status of their instantiations together with the (DPIP) seem to suggest otherwise, then perhaps this gives us a good reason to reject one of these two things – most likely, the (DPIP).
I regard this as a serious problem for the disjunction identity theory, though perhaps not a decisive one. For the sake of the master argument against reductionism that I will be presenting Chapter 7, I can afford to provisionally grant that perhaps this problem can be solved. My argument will focus on our grounds for thinking that natural and multiply realizable properties exist, so let’s now conclude this section by seeing how the disjunction identity theory fares on this front.
6.3.5 THE DISJUNCTION IDENTITY THEORY AND MULTIPLE REALIZABILITY
Assuming the (DPIP), the disjunction identity theory is compatible with the existence of causally efficacious and multiply realizable mental properties. According to the disjunction identity theory, there is a causally efficacious property that is shared by pained humans and pained Martians, the property (PH or PM).226 At least as it is developed by Kim, however, the disjunction identity theory denies the existence of natural and multiply realizable mental properties. Again, even as he entertains the theory, Kim explicitly denies that disjunctive properties are natural, contending that they are neither projectible nor able to figure in natural laws. This is the most important conclusion of the present section: like realizer-state functionalism, Kim’s disjunction identity theory denies the existence of natural and multiply realizable mental properties.227