1. The phrase is due to Bickle (1998). However, we will not here be engaging with Bickle's interesting thesis, which has enough direct empirical content to be a piece of cognitive science in its own right. The philosopher who has done most to inspire the backwash is Jaegwon Kim, and his most influential argument, as given in Kim (1998), will be the target of our discussion.

2. Or, at least, so philosophers often say. As a claim about actual behaviorist psychologists, this claim is largely nonsense, flatly untrue of, for example, E. C. Tolman or Karl Lashley. But the importance given to knocking over this straw man in the history of the rise of functionalism is indisputable, and is what is of relevance to us here. We would encourage more footnotes like this one in the philosophical literature, however.

3. Functionalism, thus understood, can be a kind of behaviorism – just one allowing for some intermediate behaviors between stimulus and response. Our own favored variety of functionalism is in fact of this behaviorist sort; but this will not play any direct role in our argument in this paper.

4. We use this example because it is standard in the literature we are describing. We should point out, though, that the philosophers who introduced it knew from the outset that it was at best neurologically implausible, and were using it as a place-holder for some imagined future reduction of a psychological to a neurological state.

5. The same argument structure has also been used (e.g., Horgan 1997) to argue for functionalism within a species (on the basis of significant neural and other differences between conspecifics), or, over time, within a single individual.

6. There are various particular ways of being a realizer functionalist in the broad sense indicated here. One particularly strong way is via the “functional analysis” strategy associated with Armstrong and Lewis, as discussed in section 2.2. Another way, canonically defended in Pylyshyn (1984), requires that types be individuated either by reference to intrinsic properties of members of the type, or by reference to intrinsic properties of independently specifiable sets of tokens of the type.

7. On our broad conception of realizer functionalism; see note (6) above.

8. Kim (1998) is not the first expression of the problem, merely an elegant, sophisticated and up to date version of it. Yablo (1992) is a clear and widely cited statement of the issues as of the early 1990s, and the papers in Kim (1993) show many of the lines of argument and thinking that lead up to Kim (1998).

9. We assume throughout that we are talking amongst people who would regard the admission of supernatural causes into science as the end of the world.

10. Talk of “enhancing” is somewhat sloppy, as Marras (2002) points out, but the details need not detain us here.

11. The philosophical literature on explanation is enormous, and so some philosophers might object to our announcing that we can boil it down to consideration of just two approaches. A few meta-comments on that literature are therefore in order. It divides naturally into two piles. The first pile, concerned directly with the way in which the search for explanation descriptively and normatively guides scientific activity, really does mainly revolve around the dialectic established by Kitcher's and Salmon's long argument with each other (see Kitcher & Salmon 1989). The second pile, highlights of which include van Fraassen (1980), Garfinkel (1981), and Achinstein (1983), concerns the logic of explanatory statements. Both piles descend from the classic work on explanation in philosophy of science by Hempel (1965), that, in the way of positivism, saw these two concerns as indistinguishable. To a post-positivist of whatever stripe, however, they are distinct, and to a considerable extent orthogonal. That is, just about any combination of views from the first and second debates can be made compatible (see, e.g., Kincaid 1997, whose work depends on subtle recombinations of them). For our purposes in this paper, only the first set of issues about explanation are directly relevant. Non-philosophers are cautioned, however, against taking our summary as a mini-survey of the whole literature on the subject.

12. As Batterman (2000, p. 118) notes, “Kim's argument won't go through unless the causal properties of the macroproperties just are the resultant (or ‘sum’) of the microstructural properties.”

13. For a sample of the literature urging this perspective, see McClamrock (1995); Wilson (1995); Clark (1997); and, especially influential with respect to what we say here, Dennett (1991a). Pet-tit (1993) provides the most systematic, though very cautious, investigation of these ideas.

14. According to Menzies (1988) this line of argument was suggested by Lewis.

15. That is, a property possessed by an object (such as dormativity) in virtue of its having some more basic (e.g., chemical) properties.

16. We are especially indebted to Ponce's treatment here.

17. Clapp (2001) successfully argues that some leading defenses of the autonomy of special sciences, such as Fodor's (1974), are guilty of this lapse of metaphysical seriousness. We should therefore note explicitly that none of our arguments in this paper depend on the idea that the kinds of any special science must be preserved as kinds just because people find it useful to think with the concepts they represent. Indeed, on one interpretation this is what “taking metaphysics seriously” in our sense here means.

18. Ross finds it necessary to make Dennett's idea more systematic because Dennett's own account, as Ross explains, leaves too many doors open to emergentist and, in other places, instrumentalistic, readings. On the other hand, Dennett's paper surpasses Ross’ in anticipation of a theme to which we will shortly be devoting much attention: the relationship between reductionism and a scientifically unsophisticated understanding of causation. See, especially, Dennett's footnote 11, and compare this with sections 4 and 5 of the present paper.

19. Stich (1983) devoted a book to arguing that this sort of picture, intended as a way of reconciling a plausible cognitive scientific typology of states with folk psychology, could not work. Kim must believe that Stich is wrong about this.

20. Cartwright (1983; 1999) has famously argued that the world is not a single, working machine, but is instead “dappled,” by which she means ontologically disunified. Dupré (1993) has urged a similar thesis. For reasons given in Spurrett (1999; 2001a) we reject this conclusion. The fact that science is never finished, and therefore never completely unified, may mean that its current description of the world at any given time will always be of a world that is “dappled”; but to derive as a metaphysical conclusion the claim that the world is dappled is to simply abandon the regulative ideal that informs Salmon's project, and, for that matter, Kim’s. Answering Kim this way would simply amount to shrugging off the significance of realist metaphysics, another way of trying to have lunch for free.

21. The “something,” we would say, is indeed fundamental structure; Ross (2000) takes it to be the network of Schrödingerstyle negentropic relations. That network is our favorite candidate for universal glue.

22. Kitcher develops, at length, additional criticisms based on counterexamples to Salmon's technical criteria for distinguishing genuine causal processes from pseudo-processes. We will not incorporate these into our summary here, because they contribute little to the issues relevant to our discussion, and because even if Salmon's apparatus is repaired so as to block the counterexamples, Kitcher's main critique is unaffected.

23. It is common outside philosophy for Dennett to be called a “reductionist” because he analyzes intentionality and consciousness without recourse to any entities or processes incompatible with the causal closure of physics. However, Dennett in fact denies, like us, that there is any general relation between physics and special sciences stronger than global supervenience; and in the context of most debates in philosophy of science, this makes him as anti-reductionist as the recent tradition allows. Thus, for example, when Kincaid (1997) defends anti-reductionism, he feels he needs to spend a few pages showing that he need not go as far in that direction as the “radical” Dennett. Ross (2000) explains in detail the sense in which Dennett's anti-reductionism is radical.

24. Philosophers typically grant that our current physical theories are open to revision, so the point here is slightly more complicated. Still, for philosophers of mind an ideal physicist is generally assumed to be making unproblematically causal claims, whereas an ideal economist, say, would need to do additional philosophical work over and above her economics to justify thinking of her claims as causal.

25. We owe this insight to Andrei Rodin.

26. The logic of this, and comparison of the causation concept's role in different branches of science, is made formally explicit in a recent paper by Thalos (2002).

27. We allude to the earlier references to the work of Nottale (1993; 2000).

28. This point has also been vividly argued by Dennett (1991a).