Functionalism and Psychological Reductionism: Friends, Not Foes Andrew Melnyk University of Missouri 1. Introduction Some opponents of psychological reductionism oppose it because they reject physicalism about the mind; they hold that psychological properties, and perhaps even minds themselves, are at most causally or nomologically dependent on the physical properties of brains and their physical environments.1 However, there are other opponents of psychological reductionism who accept physicalism, understanding minds and their properties to be, in some suitably broad sense, physical. It is to these physicalist opponents of psychological reductionism that I’ll speak in this paper. Here are four claims that such opponents typically make by way of complaint about psychological reductionism: (1) Psychological phenomena are often multiply realized. (2) The explanations that psychologists supply are more than merely heuristic: they identify objectively existing explanatory factors that the explanations of neurophysiologists (or molecular biologists, or whoever) fail to identify. (3) It is possible for psychologists to discover psychological entities, psychological processes, and the psychological generalizations that govern them without either investigating or already knowing about the underlying physical (e.g., neurophysiological or biochemical) phenomena. (4) It is possible for discoveries in psychology to correct claims made by those who study the underlying physical phenomena. Now such claims as these are plausible in their own right. But they have often been seen too—and sensibly—as natural consequences of a broadly functionalist picture of psychological phenomena.2 According to this picture, the physical entities of which nature is ultimately composed can come, in accordance with physical laws, to exhibit certain fully objective patterns of organization. But although these patterns are patterns in physical entities, and in nothing else, they are still independent of physical entities in the sense that the very same patterns can be exhibited in systems with greatly varying physical compositions (claim 1). Consequently, these patterns can be described and explained in their own right, and it is precisely the role of the special sciences, including psychology, to do so (claim 2). The entities studied by psychology, then, are functional (though physically realized) entities: they are entities whose existence just is the existence of some or other (physical) entity or complex of (physical) entities that exhibits such-and-such a pattern, where patterns can be specified by reference to causal or computational roles, or indeed in other ways. Thus, psychologists who investigate these entities do not necessarily require knowledge framed using the conceptual repertoires of physical sciences such as neurophysiology or biochemistry (claim 3). However, precisely because psychological phenomena are patterns in physical phenomena, the discovery of psychological phenomena that couldn’t be realized in physical phenomena as we currently conceive them would be evidence that our current conception of physical phenomena is incorrect (claim 4).3 Since these four claims issue from a broadly functionalist picture of psychological phenomena, it’s tempting to think that such a picture is inconsistent with psychological reductionism; and many philosophers have yielded, and continue to yield, to this temptation (e.g., classically, Fodor 1974 and Putnam 1975, Ch. 14; and, recently, Ross and Spurrett 2004). But, I claim, they’ve been wrong to do so, for, as I’ll argue in this paper, a broadly functionalist picture of psychological phenomena is quite consistent with at least one interesting thesis of psychological reductionism. I’ll work toward a formulation of this thesis of psychological reductionism in Section 2. Then, in Section 3, I’ll argue that psychological reductionism, thus formulated, is consistent with a broadly functionalist picture of psychological phenomena by showing how it is consistent with the four supposedly anti-reductionist claims stated above.4 2. Reducibility and Psychological Reductionism In this section, I’ll first describe the kind of reducibility presupposed by the interesting thesis of psychological reductionism mentioned above. (There may well be other kinds of reducibility, but if there are, I won’t be talking about them.) Then I’ll formulate the thesis. 2.1 Reducibility. The first thing to say about the kind of reducibility presupposed by the thesis of psychological reductionism that we’ll eventually formulate is that it’s a relation between contingent phenomena of one kind (e.g., biological phenomena) and contingent phenomena of another kind (e.g., chemical phenomena). The phenomena fall into two categories: first, particular phenomena, such as the existence of object-tokens, property-instances, and process-tokens; second, general phenomena, such as the holding of ceteris paribus regularities. (So my use of “phenomena” is very liberal.) What sort of relation is reducibility? It’s best viewed, I suggest, as a special kind of non-causal, synchronic explainability. So, for example, biological phenomena are reducible to chemical phenomena iff biological phenomena can be non-causally, synchronically explained—in a special way—by chemical phenomena, including chemical regularities. For the chemical phenomena to explain the biological phenomena in my intended sense of ‘synchronically’, some (but not necessarily all) of the chemical phenomena must be simultaneous with the biological phenomena they explain. For biological phenomena to be non-causally, synchronically explained by chemical phenomena in the special way mentioned above, two conditions must be met. The first condition is that purely chemical phenomena must be the only contingent phenomena involved in the explaining. Therefore, we wouldn’t have reducibility if the non-causal, synchronic explanation of biological phenomena by chemical phenomena required the holding of a fundamental chemical-to-biological law of emergence. Because the holding of such a law, though a contingent phenomenon, wouldn’t be a purely chemical phenomenon (since it would be partly biological), we wouldn’t have the reducibility of biological to chemical phenomena. The second condition for explanation in the special way mentioned above is that the metaphysical necessitation of biological phenomena by the chemical phenomena that explain them must be the metaphysical necessity of relevant object- or property-identities, e.g., identities between biological properties and chemical properties, or between biological properties and chemically-realized functional properties. Therefore, if biological phenomena are reducible to chemical phenomena, the metaphysical necessitation of the biological phenomena by the chemical phenomena that explain them can’t be any kind of brute (i.e., inexplicable) metaphysical necessitation between chemical phenomena and entirely distinct biological phenomena. Such brute metaphysical necessitation between distinct existences may in the end make no sense, as Hume famously thought; but if it does make sense, then it cannot be part of a genuinely reductive explanation. The reason is that if we learn that phenomena of one kind are reducible to phenomena of another kind, then we have discovered that we can shrink our overall ontological commitments; we no longer need to believe in the reducible phenomena as something over and above the reducing phenomena, and we no longer need to treat laws that link the reducing phenomena to the reducible phenomena as fundamental laws of nature. But if the allegedly reducing phenomena brute-metaphysically necessitate the allegedly reducible phenomena, then the latter phenomena are still something over and above the allegedly reducing phenomena, and so we have no genuine reducibility. 2.2 Psychological Reductionism. Any thesis of reductionism that invokes the kind of reducibility just described will obviously assert that phenomena of one kind are reducible to phenomena of another kind. But what are the two kinds of phenomena in the case of psychological reductionism? Let us ask first which phenomena psychological reductionism should assert to be reducible. The obvious answer is “Psychological phenomena—psychological objects, properties, processes, regularities, and so on”, but we need to say more precisely what these are. A natural first thought is to tie psychological phenomena to current psychology, and to take ‘psychological phenomena’ to refer to all those phenomena asserted to exist, or of the kinds asserted to exist, in the consensus theories of current psychology.5 But this proposal suffers from two problems. First, since the consensus theories of current psychology are bound to contain errors in what they assert to exist, some of the phenomena apparently picked out by the proposal will turn out not to exist. But only real phenomena are reducible in the present sense, since reducibility is a kind of explainability, and only genuine phenomena can be explained. Hence, if psychological reductionism claims the reducibility of all psychological phenomena, then on the present proposal it will be committed—absurdly—to the reducibility of certain non-entities. Second, since the consensus theories of current psychology are bound to be incomplete, in the sense of failing to mention some phenomena truly asserted to exist, or of kinds truly asserted to exist, in future psychology, a thesis of psychological reductionism that merely asserts the reducibility of all the phenomena asserted to exist, or of the kinds asserted to exist, in the consensus theories of current psychology will fail to assert the reducibility of all psychological phenomena. Let us therefore take ‘psychological phenomena’ to refer to all the phenomena truly asserted to exist, or of the kinds truly asserted to exist, in both the consensus theories of current psychology and the consensus theories of psychology at each point in its future. The requirement that psychological phenomena be those truly asserted to exist in certain theories avoids the problem resulting from the likely falsehood of some psychological theories, present and future. And the reference to future psychological theories avoids the problem of failing to include within the scope of psychological reductionism all the phenomena that it should include. It might be objected to this proposal that we have no principled way of telling whether a future science is psychology, given that we are assuming that psychology might change the content of its theories over time. But there is a way of tracking psychology over time that doesn’t rely on the content of its theories: we can count any future science as psychology if it seeks, at least in the first instance, to explain human behavior—and perhaps also to give an account of the nature of such paradigmatic mental phenomena as beliefs, desires, and sensations. Let us now ask to which phenomena psychological reductionism should claim that psychological phenomena are reducible. A tempting first response is to suggest that psychological reductionism should claim that psychological phenomena are reducible to neurophysiological phenomena that exist entirely within the skin of the thinker. But this response is inadvisable at least for the following reason.6 If psychological reductionism asserts the reducibility of psychological phenomena to neurophysiological phenomena of this sort, then it renders itself incompatible with at least two kinds of plausible psychological hypothesis. The first kind of hypothesis identifies the representational content of an organism’s mental states with those states’ standing in certain relations to phenomena that are external to the organism (see, e.g., Fodor 1990). If a hypothesis of this first (i.e., semantically externalist) kind is true, then not all psychological phenomena are reducible to neurophysiological phenomena. The second kind of plausible psychological hypothesis aims to “biologize” psychological phenomena by treating them as essentially possessing certain biological functions (see, e.g., Millikan 1984, Lycan 1987, Ch. 4, Dretske 1988). But the possession by an organism of a biological function is plausibly thought to require the organism to have had a certain evolutionary history in a certain environment. If so, then the truth of a hypothesis of this second kind would also entail that psychological phenomena are not reducible to (narrow) neurophysiological phenomena. A better response to our question, then, is to say that psychological reductionism should claim that psychological phenomena are reducible to non-psychological phenomena. Non-psychological phenomena would obviously include (narrow) neurophysiological phenomena, but would also include phenomena in the environment, or in the history, of the organism that exhibits the psychological phenomena. This suggestion thus makes room for the reducibility of psychological phenomena that don’t supervene solely on intrinsic and simultaneous features of the organism exhibiting the psychological phenomena.7 However, we must construe “non-psychological phenomena” carefully. We can’t take the term to refer to all phenomena not identical with psychological phenomena, since in that case psychological reductionism would have to be deemed false if psychological phenomena turned out to be identical with (say) neurophysiological phenomena! So we should take ‘non-psychological phenomena’ to refer to those phenomena truly asserted to exist, or of the kinds truly asserted to exist, in the consensus theories of sciences other than psychology, both current and future. And, for this purpose, we should include our common-sense view of the physical world as an honorary science, since a reductive explanation of psychological phenomena might well have to invoke everyday objects such as cars and human bodies. Let me now pull the various threads together. The interesting thesis of psychological reductionism that was promised in the introduction can be formulated like this: [Psychological Reductionism] All those phenomena truly asserted to exist, or of the kinds truly asserted to exist, in both the consensus theories of current psychology and the consensus theories of psychology at each point in its future are synchronically explainable by contingent phenomena that include only those phenomena truly asserted to exist, or of the kinds truly asserted to exist, in the consensus theories of sciences other than psychology, both current and future; but such explainability may not depend on any brute metaphysical necessitation of the psychological phenomena by entirely distinct phenomena. All uses of ‘psychological reductionism’ in the rest of this paper will refer to this thesis. 3. The Consistency of Psychological Reductionism With the Functionalist Picture Let me now argue that psychological reductionism is consistent with a broadly functionalist picture of psychological phenomena by showing that it is consistent with the four supposedly anti-reductionist claims that issue from the broadly functionalist picture. I’ll consider each claim in turn. 3.1 The first supposedly anti-reductionist claim was that psychological phenomena are often multiply realized. However, psychological reductionism is consistent with this claim because the multiple realization of psychological phenomena does not prevent their possessing an explanation, in terms of non-psychological phenomena, of the special sort required by psychological reductionism. To show that this is so, I will outline schematic representations of reductive explanations of multiply realized psychological phenomena. As already noted, psychological phenomena fall into the two categories of particular phenomena, such as psychological object-tokens and property-instances, and general phenomena, such as the holding of ceteris paribus psychological regularities. I will take each category of psychological phenomena separately, starting with particular phenomena. 3.1.1 Suppose, then, that a psychological state-type, P, is multiply realized by non-psychological phenomena. (Roughly, what this amounts to is, first, that P is a functional state-type of a particular sort, the state-type (say) that is tokened iff some or other state-type is tokened that meets a certain specific condition, C; and, second, that, on at least two occasions on which P is tokened, it is distinct non-psychological state-types that, on each occasion, meet condition C.) Suppose, further, that P is tokened on a particular occasion. Here is how, in highly idealized form, we could represent a reductive explanation of this particular psychological phenomenon, even though P is a multiply realized psychological state-type: (1) There is a tokening of a non-psychological state-type, N. (2) There hold certain non-psychological regularities and (perhaps) there are tokenings of non-psychological types other than N. So (3) there is a tokening of a state-type that meets condition C. (4) There being a tokening of some or other state-type that meets condition C just is there being a tokening of P. So (5) there is a tokening of P. Premisses (1) and (2) together describe the non-psychological phenomena to which the tokening of P is reducible.8 Premiss (1) describes the local non-psychological reduction base of the tokening of P, while premiss (2) describes the existence of whatever non-local non-psychological environmental or historical phenomena are required, and the holding of whatever non-psychological regularities are required, for the tokening of N to meet condition C and hence to realize a tokening of P. Therefore, (3) follows from (1) and (2). Premiss (4) affirms the identity of the psychological state-type P with a certain functional state-type. From (4) and (3) there follows the conclusion (5), which reports the tokening of P that is to be reductively explained. I first presented a reductive derivation of this kind in (Melnyk 1995), developing the idea in (Melnyk 2003, Ch.3), where it is defended much more fully. Jaegwon Kim has also defended a very similar idea, most recently in his (2005, Ch.4). However, there are three important differences between Kim’s position and mine. First, I insist that the identity of reduced types with functional types will almost always be an a posteriori matter, while Kim treats such identities as knowable a priori, presumably through a priori conceptual analysis. Second, Kim holds that reductive explanations somehow license the identification of instances of reduced properties with instances of reducing properties; I don’t share this view. Third, Kim doesn’t discuss how to reduce regularities, whereas I have done so in the sources cited, and will do so again below. Now, the explanation that this derivation represents is indeed of the special kind described in my earlier account of reducibility, and hence qualifies as reductive.9 First, it appeals in part to a particular non-psychological phenomenon (i.e., the tokening of N) that is simultaneous with the particular psychological phenomenon being explained. Second, the only contingent explanatory factors that the explanation appeals to are non-psychological phenomena. It is true that in the representation of the explanation premiss (4) is required to serve as a bridge law to connect the disparate vocabularies in which representations of non-psychological and psychological phenomena are couched; but because premiss (4), though not a priori, still expresses a necessary truth, it cites no explanatory factor in addition to those cited in premisses (1) and (2).10 Third, the explanation does not require the holding of a relation of brute metaphysical necessitation between distinct existences (i.e., between non-psychological and psychological phenomena). It’s true that premiss (4) expresses a necessary identity, as already noted, but identity is a relation between an entity and itself, not between an entity and some distinct entity. Furthermore, and crucially, the reductive explanation represented by the derivation above is fully consistent with the multiple realizability of psychological state-type P. For, although the reductive explanation requires the identity of P with a functional state-type, it does not require the identity of P with any specific neurophysiological or biochemical or physical state-type. So it leaves open the possibility that some non-psychological state-type N*, distinct from N, should, given the right non-psychological environment and regularities, also meet condition C and hence that a token of N* should constitute the local reduction base of a distinct tokening of P—which would mean, of course, that P was multiply realized. 3.1.2 Let me now show that the reducibility of general psychological phenomena (i.e., of psychological regularities) is likewise consistent with the multiple realizability of psychological kinds. Suppose that it is a psychological regularity (perhaps a ceteris paribus law) that P-tokens are followed by P*-tokens. Then, if P and P* are multiply realized psychological types, different instances of the regularity (i.e., different sequences each consisting of a particular P-token followed by a particular P*-token) will be differently realized by non-psychological phenomena. How, then, can the regularity have a reductive explanation? It can do so, I say, if every instance of the regularity has a reductive explanation; no more is required.11 There is no conflict with multiple realizability, because the respective reductive explanations of the various instances of the regularity need not each appeal to the same kind of non-psychological realizers. But how can an instance of a regularity have a reductive explanation? An instance of a regularity consists of a P-token and then a P*-token, and can have a reductive explanation composed of three elements. The first element is the reductive explanation, of the sort described above, of the existence of the P-token. The second element is a different reductive explanation, also of the sort described above, of the existence of the P*-token. The third element is a causal explanation of the later non-psychological phenomenon (by which the P*-token was reductively explained) by reference to the earlier non-psychological phenomenon (by which the P-token was reductively explained). The three elements taken together constitute an explanation of the psychological regularity: an earlier non-psychological phenomenon both realizes (and hence necessitates) the simultaneous P-token and causes the later non-psychological phenomenon that realizes (and hence necessitates) the P*-token. So, given the existence of the earlier non-psychological phenomenon and the holding of the non-psychological laws, there must exist the P-token followed by the P*-token. And, as reflection will soon show, this explanation of an instance of a psychological regularity clearly meets the conditions for being reductive given in Section 2.1.12 3.2 The second supposedly anti-reductionist claim was that the explanations that psychologists supply are more than merely heuristic, since they identify objectively existing explanatory factors that the explanations of neurophysiologists (or molecular biologists, or whoever) fail to identify. I’ll argue, however, that this claim is quite consistent with psychological reductionism. But I shall do so indirectly, by considering three arguments for fearing that the claim is inconsistent with psychological reductionism and showing that none of them is sound. The burden of proof will then lie on my opponent to make a positive case for inconsistency. 3.2.1 The first argument holds that if psychological reductionism is true, then psychological phenomena are not real; and that if psychological phenomena are not real, then explanations in psychological vocabulary that cite such phenomena can at best be of heuristic value, since non-existent phenomena can explain nothing. But this argument fails. Its first premise assumes that if something is reducible, then it doesn’t really exist, but the assumption is false. Suppose that you are first acquainted with dry-stone walls from too great a distance to be able to determine their composition; you would surely be wrong to conclude that they don’t really exist just because you later discover that they are reducible to the stones that compose them. Any number of other similar examples could easily be produced to show that, at least if our ordinary thought on such matters is correct, reducibility doesn’t entail non-existence. To the contrary, in fact: reducibility entails existence. Since reducibility is a kind of explainability, and since only what exists can be explained, the reducibility of psychological phenomena actually entails that they do exist. So psychological reductionism doesn’t require that psychological claims must be dismissed as non-explanatory on the grounds that the phenomena they describe are unreal. 3.2.2 According to a second argument, psychological reductionism entails that all psychological phenomena already have an explanation—a reductive explanation—and hence that there is no room for any further explanation such as psychology might supply; so any apparent explanations of psychological phenomena in psychological vocabulary must be treated as heuristic at best. Now this argument clearly rests on the assumption that a single phenomenon can have only one explanation. But why assume that? Let me consider just one possible reason: if (say) psychological phenomena have multiple explanations, then, implausibly, psychological phenomena will be causally overdetermined, since any given psychological phenomenon will have two simultaneous causes—a psychological cause and a non-psychological cause, this latter cause being the cause of the non-psychological phenomenon to which the psychological phenomenon is reducible. Causal overdetermination, however, is objectionable only if the two simultaneous causes are entirely independent of one another, as they are in the classic examples such as the case where a victim is simultaneously hit by two separate bullets each of which would have sufficed alone to cause death; for only then can one reasonably worry, for example, whether some principle of parsimony (e.g., Ockham’s Razor) has been violated. However, given psychological reductionism, the case of a psychological phenomenon that has both a psychological cause and a simultaneous non-psychological cause is importantly different from the two-bullets case precisely because the two causes, unlike the two bullets, are not entirely independent of one another. For, given psychological reductionism, the first cause of the psychological phenomenon, being itself a psychological phenomenon, must itself be reducible to some simultaneous non-psychological phenomenon; and this simultaneous non-psychological phenomenon must be (or be part of) the second cause, i.e., the non-psychological cause of the non-psychological realizer of the psychological phenomenon. Thus, since the first cause must (metaphysically, not causally) exist given that the second cause does, it can’t be treated as some additional posit in gratuitous violation of Ockham’s Razor. So, if psychological reductionism is true, then, precisely because it’s true, psychological phenomena can have psychological causes and simultaneous non-psychological causes, but without overdetermination of the arguably objectionable kind present in the two-bullets case. There is a variant of the second argument, however, that must also be considered. According to it, psychological reductionism entails that all psychological phenomena already have an explanation—a reductive explanation—and hence that there is no point in any further explanation in psychological vocabulary; further explanation isn’t impossible, just needless. In response to this argument, let me sketch two very abstract ways of thinking about the nature of explanation, both of which are consistent with the idea that multiple explanations of a single phenomenon do have a point. According to the first, to explain some phenomenon is to show that it was necessary (or probable) given some earlier phenomenon; and one can do so by citing some earlier phenomenon and subsuming both it and the explanandum under some law. If so, then a psychological phenomenon can be shown to be necessary, and hence explained, by citing an earlier psychological phenomenon and an appropriate psychological law. But the psychological phenomenon can also be shown to be necessary, and hence explained, by citing an earlier non-psychological phenomenon which, because of a non-psychological law, causes the non-psychological phenomenon to which the original psychological explanandum is reducible. Neither demonstration has any just claim to be the (i.e., the only) explanation of the explanandum. According to the second way of thinking about explanation, to explain some phenomenon is to exhibit it (e.g., by subsuming it under a law) as fitting into some kind of pattern. But if so, then there seems no reason why a single phenomenon should not fit into more than one kind of pattern, in which case we can view a psychological explanation and a reductive explanation of the same psychological phenomenon as exhibiting its fitting into two different, but both entirely real, patterns. So far in this sub-section I have argued that, even given psychological reductionism, explanations such as psychology supplies cite real phenomena, and are neither impossible nor pointless. But a further question remains, namely, whether such explanations are indispensable, not just heuristically but metaphysically (see, e.g., Schouten and Looren de Jong 1999, 246). No simple answer is possible. On the one hand, because explanations in psychological vocabulary of psychological phenomena represent those phenomena as fitting into patterns that are fully objective features of the world, whereas reductive explanations of the same phenomena, in a different vocabulary, don’t represent them as fitting into those patterns but rather into others, explanations in psychological vocabulary really do provide explanatory insight that reductive explanations miss. On the other hand, it must be possible in principle (though not in practice) somehow to use non-psychological vocabulary to describe those patterns into which psychological explanations describe psychological phenomena as fitting. For if even this in-principle possibility is denied, physicalism seems to have been abandoned; at the very least, if you endorse physicalism but still think there are contingent items in the world that can’t, even in principle, be described using non-psychological vocabulary, then you owe a plausible account of why you still qualify as a physicalist, an account that’s precise enough to show that by denying this in-principle possibility you haven’t contradicted yourself. However, if non-psychological vocabulary were used to describe the patterns into which psychological explanations describe psychological phenomena as fitting, those patterns would be the very same patterns into which psychology describes psychological phenomena as fitting; so, the possibility in principle of describing those patterns non-psychologically doesn’t detract from psychology’s claim to provide explanatory insight. 3.2.3 According to a third argument, psychological reductionism entails that the explanations that psychologists supply aren’t genuine because such explanations can’t be causal explanations. If psychological reductionism is true, then all psychological phenomena are reducible to non-psychological phenomena. But since all non-psychological phenomena have sufficient non-psychological causes, it seems to follow that among psychological phenomena no genuine causal relations hold, even though they appear to. Jaegwon Kim has long defended this kind of argument, though in his formulation a supervenience thesis plays the role played here by psychological reductionism (see, for his most recent version, Kim 2005, Ch. 2).13 But in fact the conclusion of this argument doesn’t follow from its premises; the so-called “causal exclusion principle” on which it relies seems to be just false. Consider our judgment of an analogous case. Suppose that as I walk beside a large dry-stone wall it falls, bruising me extensively; clearly my extensive bruising was caused by the falling wall. Now, because the wall is entirely made up of stones, the falling wall is reducible to many falling (and doubtless interacting) stones; similarly, the extensive bruising is reducible to many sub-areas of bruising. Now, we could in principle give a sufficient explanation of every single sub-area of my extensive bruising by reference to the career of some or other particular falling stone. But our ability to do so, and the reducibility which makes it possible, cast no doubt at all on our original judgment that my extensive bruising was caused by the falling wall. At least if our ordinary thinking about the matter can be trusted, the causal explanation of my extensive bruising that cites the falling wall can peacefully co-exist with the “reductive” causal explanation that cites the many falling stones that individually cause the many sub-areas of bruising that constitute my extensive bruising. Likewise, I assume, for the case of causal explanations framed in psychological and in non-psychological terms: causal explanations of both kinds can co-exist even if psychological reductionism is true.14 3.3 The third supposedly anti-reductionist claim was that it is possible for psychologists to discover psychological entities, psychological processes, and the psychological generalizations that govern them without either investigating or knowing about the underlying physical (e.g., neurophysiological or biochemical) phenomena. However, psychological reductionism doesn’t rule out this possibility. To see why, it will help to explore more deeply what psychological reductionism is, and what it is not. What it is is a thesis, with a truth-value, concerning the relationship between the property-instances, events, objects, and so on (correctly) described by psychology and the property-instances, events, objects, and so on (correctly) described by sciences other than psychology. If true, it contributes to addressing a larger question in which everyone curious about what the world is like—and hence, one would have thought, every scientist—should be interested: on the basis of the evidence currently available to us, what is the relationship between the various segments of reality investigated by the many sciences? Suppose someone asks us what the world—at least the contingent world—is like, and that, as good naturalists, we respond by reading out the contents of all of our science textbooks, adding that this answer, though certain to be wrong in many respects, is nonetheless the best we presently have. There is still a question as yet unanswered, concerning how the various accounts of the world given in the various textbooks fit together to form a unified account of the contingent world as a whole. It is to this question that psychological reductionism offers part of an answer. Accordingly, both the question and the answer can be said to belong to metaphysics—if metaphysics is the part of philosophy that aims to understand what the world is like, but at a higher level of abstraction than that afforded by particular branches of science—but to metaphysics of a kind that philosophers of science should welcome. What psychological reductionism is not is a thesis about the relationship between certain theories (e.g., psychological ones and neuroscientific ones), either at a time or over time, although an account of inter-theoretic reduction for true theories could be constructed from it.15 Nor is psychological reductionism a thesis about the actual relationship between psychology as an activity or practice and other sciences considered as activities or practices. The nature of this relationship is certainly a fascinating question for sociologists and even philosophers of science, but it is not the question to which psychological reductionism is addressed. Finally, psychological reductionism, being a descriptive thesis, is not a methodological directive, which would be prescriptive. Once the character of psychological reductionism has been properly understood, I think it becomes clear that it doesn’t entail anything about how discoveries in psychology can or cannot be made. A fortiori, it doesn’t entail that psychologists can’t make psychological discoveries unless they investigate or know about the non-psychological phenomena to which psychological phenomena are reducible. In further support of this contention, consider the analogous case of water, which is presumably reducible in the present sense to chemical entities. There is much that could have been discovered about the macro-features of water (e.g., its viscosity, specific heat, boiling and freezing points, necessity for life, role in chemical reactions) by someone who neither investigated nor knew about the atoms of hydrogen and oxygen to which it is reducible. The position is no different, I suggest, for reducible psychological phenomena. 3.4 The fourth supposedly anti-reductionist claim was that it is possible for discoveries in psychology to correct claims made by those who study the underlying physical phenomena (see, e.g., Schouten and Looren de Jong 1999, 256-8). This possibility, too, is quite consistent with psychological reductionism, as you would expect given that psychological reductionism doesn’t even aim to comment on the relationship between the practice of psychology and that of other sciences. The puzzling thing is why anyone would have expected psychological reductionism to rule out this possibility. The answer, I suggest, is that psychological reductionism indeed entails that certain non-psychological phenomena are metaphysically prior to psychological phenomena, and from this someone might infer that the relevant non-psychological sciences are therefore epistemologically prior to psychology, with correction flowing only upwards, from non-psychological claims to psychological ones. But such an inference would be unsound, as I’ll now explain. Imagine embarking on a particular psychological inquiry while assigning a significant probability to psychological reductionism—perhaps on the ground that other psychological phenomena, and other common-sense but non-psychological phenomena, have as a matter of fact turned out to be reducible in the pertinent sense. You are therefore confident that the psychological phenomena you are aiming to discover do in fact mesh in a certain way with (i.e., are in fact reducible to) whatever the relevant simultaneous non-psychological phenomena turn out to be. For this reason, it counts against a proposed psychological hypothesis if the phenomena it postulates can’t mesh with (what our best theories claim to be) the relevant non-psychological phenomena; it doesn’t, of course, conclusively falsify the psychological hypothesis, but it does count against it. But for the very same reason, it also counts against a relevant non-psychological hypothesis if the non-psychological phenomena that it postulates can’t mesh with (what our best psychological theories claim to be) the relevant psychological phenomena. Just as the assumption of psychological reductionism makes it reasonable for correction to flow upwards, from non-psychological to psychological claims, so also the same assumption makes it reasonable for correction to flow downwards, from psychological to non-psychological claims. In fact, it is hard to see how anything other than some sort of metaphysical reductionism could underwrite the reasonableness of the two-way flow of correction between psychology and the relevant non-psychological sciences. For suppose that psychology and, say, the neurosciences investigated domains that were entirely independent of one another except for the holding of certain fundamental cross-scientific laws; this is the radically non-reductionist picture of the relations between psychology and the neurosciences that psychophysical dualists must favor.16 Then, presumably, psychology and the neurosciences would—and should—develop quite independently of one another, except for when it was necessary to discover such fundamental laws as might hold between their respective domains. Precisely because psychological phenomena were not reducible to non-psychological phenomena, there would be no rationale for testing psychological hypotheses by examining neuroscientific phenomena, or vice versa. The methodological claim that psychology and the neurosciences should co-evolve would therefore appear to presuppose some version of metaphysical reductionism. REFERENCES Dretske, F. (1988). Explaining Behavior: Reasons in a World of Causes. Cambridge, MA: The MIT Press. Fodor, J. (1974). Special Sciences, or the Disunity of Science as a Working Hypothesis. Synthèse, 28, 97-115. Fodor, J. (1990). A Theory of Content and Other Essays. Cambridge, MA: The MIT Press. Kim, J. (2005). Physicalism, or Something Near Enough. Princeton, NJ: Princeton University Press. Looren de Jong, H. (2002). Levels of explanation in biological psychology. Philosophical Psychology, 15, 441-462. Lycan, W. (1987). Consciousness. Cambridge, MA: The MIT Press. Melnyk, A. (1995). Two Cheers for Reductionism; or, The Dim Prospects for Non-Reductive Materialism. Philosophy of Science, 62, 370-388. Melnyk, A. (2003). A Physicalist Manifesto: Thoroughly Modern Materialism. Cambridge: Cambridge University Press. Millikan, R. G. (1984). Language, Thought, and Other Biological Categories: New Foundations for Realism. Cambridge, MA: The MIT Press. Putnam, H. (1975). Mind, Language, and Reality: Philosophical Papers. Vol. 2. Cambridge: Cambridge University Press. Ross, D., & Spurrett, D. (2004). What to say to a skeptical metaphysician: A defense manual for cognitive and behavioral scientists. Behavioral and Brain Sciences, 27, 603-627. Schouten, M., & Looren de Jong, H. (1999). Reduction, elimination, and levels: the case of the LTP-learning link. Philosophical Psychology, 12, 237-262. NOTES 1 There may be a form of dualism that holds that psychological properties, and perhaps even minds themselves, are metaphysically dependent on the physical properties of brains and their physical environments (see Melnyk 2003, 58-62). 2 The picture is broadly functionalist because it isn’t committed to any particular hypotheses about the functional character of psychological phenomena. 3 Contra Huib Looren de Jong, who holds that functionalism can’t allow the possibility of biological constraints on psychological theorizing (Looren de Jong 2002, 457). 4 I shan’t here argue that this form of psychological reductionism is true, but evidence for it is, in effect, presented in (Melnyk 2003, Ch. 6). 5 Consensus theories of current psychology are those that are currently the objects of consensus among scientific psychologists. A good way of determining which theories these are is to examine the contents of widely used textbooks in psychology. 6 Another possible reason is this: in specifying which phenomena should count as the neurophysiological ones, we would need to appeal to what is asserted to exist by neurophysiological theories; but in that case we would run into familiar difficulties concerning whether the intended theories belong to current or to future (or even to ideal) neurophysiology. 7 Thus, the possession by psychological phenomena of biological functions is not in itself a reason to doubt psychological reductionism, pace (Schouten and Looren de Jong 1999, 240-46). 8 Note that premisses (1) and (2) needn’t, and typically won’t, describe the totality of non-psychological phenomena simultaneous with the tokening of P. The descriptions they give will be selective, confining themselves to relevant non-psychological phenomena, i.e., those relevant to the realization of a token of P. Therefore, the non-psychological phenomena cited in a psychological reduction (if it takes the form sketched in the text) can’t be charged with indiscriminately including irrelevant non-psychological detail. And no obstacle to reduction is created by “multiple supervenience”, i.e., “the capacity of one and the same physical substrate to support a number of different higher-level properties” (Schouten and Looren de Jong 1999, 241; see also Ross and Spurrett 2004, Sect. 3.2), since reductions of (tokens of) different higher-level properties will cite different aspects of the same physical substrate. 9 Here, I am just assuming that this derivation represents a genuine explanation. For defense, see (Melnyk 2003, 98-102). 10 I take this point from Jaegwon Kim (Kim 2005, 131-9). 11 This claim is highly contentious. For defense, see (Melnyk 2003, 104-109). 12 The extent of a psychological regularity can perfectly well be restricted spatio-temporally, as Looren de Long plausibly suggests is true of biological regularities (Looren de Jong 2002, 453). However, restricted regularities can be reductively explained along the lines I’ve sketched—by explaining each of their instances—just as well as unrestricted regularities can. 13 By the way, Kim wouldn’t regard psychological reductionism as strong enough for true reductionism, since it doesn’t entail the holding of token or type identities between psychological and non-psychological phenomena. 14 For an account of causation that tries to show how this co-existence is possible, see (Melnyk 2003, Ch. 4). 15 Perhaps this: theory T1 is inter-theoretically reducible to theory T2 iff the ontology of T1 is metaphysically reducible to the ontology of T2. 16 At least those psychophysical dualists who don’t hold that a science of the mind is impossible in principle. PAGE 25