Chapter Eleven—
Intentionality Without Vindication, Psychology Without Naturalization

The argument thus far in the book may be summarized briefly as follows: (1) Advocates of CTM have claimed that viewing the mind as a computer allows us not only to make advances in empirical psychology, but also to satisfy the more specifically philosophical desiderata of supplying a (naturalistic) account of intentionality and vindicating intentional psychology by demonstrating its compatibility with causal-nomological psychological explanation, materialism, and the generality of physics. (2) CTM fails to make good on its claims to produce these philosophical results. But (3) a "bowdlerized" version of CTM can nonetheless provide a framework for an interesting empirical research programme in computational psychology, because what is needed for the "good-making" qualities internal to a science is much weaker than what is needed for strong naturalization or vindication. Psychology could attain a significant degree of internal mathematical maturity without any demonstrable connections between psychological categories and the categories of the physical sciences. And it could attain a great deal of connective maturity through localizations that were empirically adequate yet metaphysically contingent and epistemically opaque. Empirical science is largely blind to metaphysical modalities stronger than empirical adequacy, while questions about the metaphysical nature of mind-body relations are precisely the sorts of things that are of importance for strong naturalization and vindication. In short, science and metaphysics enjoy a substantial degree of mutual autonomy.

But if such issues are not the practicing scientist's concern, they cer-

tainly are the concern of philosophers. Or, more precisely, while there is an important kind of philosophy of science that examines precisely what particular sciences are about, there are also broader philosophical questions of a metaphysical bent as well. And neither the legitimacy nor the importance of the metaphysical questions is undercut by the fact that they are not questions for the empirical scientist. The reason that CTM was initially of interest was that it seemed to offer solutions to some problems that many philosophers in this century have regarded as difficult and important ones. Even if we can do good science without answers to the questions, it nonetheless behooves us as philosophers to see where we are left if our bowdlerized interpretation of computational psychology leaves them unresolved.

The reader will recall that there were two main issues that CTM sought to address. The first was to provide an account of the intentionality of mental states—and in particular, to do so in naturalistic, or nonintentional, terms. The second was to "vindicate" intentional psychology. And here there were three principal concerns: (1) that intentional explanation be (or at least point to) causal-nomological explanation, (2) that intentional realism be compatible with materialism, and (3) that intentional realism be compatible with the generality of physics. For a compatibility proof , however, you need demonstrable identities between mental and naturalistic states, and I have argued (a ) that BCTM does not provide this and (b ) that it is not to be expected. Luckily, empirical correlations of a weak metaphysical nature are good enough for science. But those who are concerned about the original motivating problems on philosophical grounds may still have reason to be worried about the resulting picture on those same grounds. It is thus my intention to address such concerns in this chapter.

11.1—
The Central Problem of Modern Philosophy

It is not merely a matter of convenience that we divide the history of philosophy into "ancient," "medieval," and "modern" epochs. In each of these transitions, there was a distinctive new synthesis that arose to accommodate new foundational assumptions. Medieval philosophers needed to find a synthesis between the philosophy and science of the ancient world and the precepts of revealed religion. Modern philosophers needed to accommodate the changes in world view that accompanied the emergence of what has come to be called modern science in the sixteenth

through eighteenth centuries. The medieval synthesis embraced a largely Aristotelian science whose teleological nature was applied indifferently to physics, biology, and psychology. There was no mind-body problem for the medieval mind, because the medieval notions of body and soul, while distinct, were not in conflict. Modern science, however, employed very different modes of description and explanation. On the Galilean view of science, embraced in different ways by philosophers as different as Hobbes and Descartes, the physical world is basically just a large collection of bodies in motion. These motions can be exactly described by mathematics, and the behavior of complex bodies is a consequence of (and can be derived from) the motions of their simple parts. Indeed, the guiding metaphor for many writers of this period was that of geometric proof: to understand a natural phenomenon, one first "resolves" it into its simple parts (corresponding to the definitions and axioms) and then "composes" the complex phenomenon from the simple (corresponding to geometric constructions and the proof of a theorem). The resolutive step requires hypothesis testing of some sort or other, but the compositive step is regarded as being necessary and epistemically transparent.

The problem comes in how to integrate this view of physical nature with a larger philosophical picture that includes mental, moral, and social phenomena as well. Hobbes sketched a strongly naturalistic programme in which politics is derived from psychology, complex psychological phenomena are derived from simpler ones, and simple ones are identified with motions of the body. But he never explained how this crucial identification is supposed to proceed, much less showed us how to derive phantasms from bodily motions in quasi-geometric fashion. And Descartes (who was, incidentally, the most important promoter of mechanistic physics, and even mechanistic psychology with respect to things like reflex action and perception [see Treatise on Man, AT XI.202]) points out at least three significant differences between the mental and the physical: the unmediated first-person access to mental states, the metaphysical "real distinction" between body and mind due to incommensurable essential properties, and the inability to give mechanistic explanations of the faculties of reason, language, and the will.

Both of these writers—Descartes by argument and Hobbes by exampie—provide early lessons in the fundamental problem facing modern metaphysics: that of how to find a single overarching philosophical framework that accommodates both our best way of talking about the physical world and the most natural ways of talking about the mind (i.e., in mentalistic terms). Now just how you describe this problem depends

very heavily on your metaphysical assumptions. For example, if you buy into the kind of substantival metaphysics assumed by Descartes, you may be inclined to see the issue as one of how many kinds of substances there are, while this way of formulating the question may seem unintelligible if you reject the substantival metaphysics on which it depends. Perhaps the most neutral way of describing the problem is as follows: we have natural and successful ways of talking about the mind and about physical nature—different kinds of discourse, if you will—and they seem to be incommensurable. Neither seems to be reducible to the other, and there does not seem to be any common denominator that unites them. The problem is not that they are inconsistent . It is merely that they are not unified. Let us call this apparent chasm between mentalistic and physical discourse the "Cartesian Gap."

Now there are at least two kinds of problems that have traditionally led philosophers to feel uneasy about the Cartesian Gap. One is that the whole philosophical instinct is directed towards finding a world view that unifies all of our discourses. The philosophical impulse is an impulse towards unification. And hence there is something ugly and unsatisfying about the Cartesian Gap. The second problem is that we have clear intuitions to the effect that there are relationships between the physical and the mental—and important ones at that!—that we would like to be able to describe and explain. On the one hand, there is voluntary causation: volitions would seem to be causes of actions. On the other hand, there is perception, in which events in sensory nerves would seem to be causes of perceptual states. In addition, as was already clear to Descartes and his contemporaries, thoughts seem to bear some kind of special and intimate relationship to events in the brain. Action, perception, and the localization of mental functions are all things that we would like to be able to talk about. Incommensurably separate discourses about mind and matter leave out some of the phenomena that we would most like to explain. To remain content with the Cartesian Gap is to remain content with some amount of mystery, and indeed with a substantially greater helping of it than most philosophers are inclined to be content with.

11.2—
The "Received View"

There are, of course, many kinds of philosophical theories directed at solving or dis solving this problem: various kinds of dualism, materialism, and idealism, as well as linguistic and social theories that rework the apparently fundamental metaphysical issues into epiphenomena of

language, cognition, or sociality. CTM and many of its favorite adversaries (behaviorism, reductionism, eliminativism) share a grounding in a particular materialist approach to the Cartesian Gap: namely, a normative claim to the effect that particular kinds of connections to physicalistic discourse are necessary conditions for the legitimation of mentalistic discourse . One might call this view normative naturalism . CTM's advocates differ with behaviorists and reductionists on the nature of the needed connections. They differ with eliminativists on the issue of whether the grounding of the mental in the world of nature can in fact be accomplished. However, they generally share the assumption that if push comes to shove between naturalism and intentional realism, it is intentional realism that should be abandoned.^[1]

Now one might well think that, since CTM's original appeal drew in large measure from its ability to address concerns that arose directly out of positivist, behaviorist, and reductionist projects, the perceived need to address those concerns would not long survive the demise of the projects that spawned them. But almost no one believes the verification theory of meaning anymore or embraces a reductionism of the form popular in the 1950s, and even methodological behaviorists are increasingly difficult to find. But concerns about vindicating intentional psychology live on. Indeed, the current orthodoxy in philosophy of mind—the "Received View," if you will—seems to treat as axiomatic the claims (1) that materialism is true, (2) that there are no mentalistic properties that are fundamental (as opposed to derived from more primitive physical properties), (3) that sciences must deal in causal-nomological explanations, (4) that the only legitimate entities are those that appear in the explanatory inventory of some natural science, and hence (5) that, however useful or well-confirmed mentalistic ascriptions and explanations may be, they are nonetheless in the position of needing to be justified on metaphysical grounds. According to the Received View, it is our discourses about the mind (whether scientific or commonsensical) that must answer to a materialist and naturalistic metaphysics and a causal-nomological view of science, and not the other way around. Fodor, for example, writes,

The deepest motivation for intentional irrealism derives . . . from a certain ontological intuition: that there is no place for intentional categories in a physicalistic view of the world; that the intentional can't be naturalized . . . .

. . . It's hard to see . . . how one can be a Realist about intentionality without also being, to some extent or other, a Reductionist. If the semantic and the intentional are real properties of things, it must be in virtue of their iden-

tity with (or maybe their supervenience on?) properties that are themselves neither intentional nor semantic. If aboutness is real, it must be really something else. (Fodor 1987: 97, 98)

Thus it becomes an intelligible move in the game to claim such otherwise outrageous things as that mental states are "explanatory posits" of "folk psychology" and might be "eliminated" if this "folk psychology" cannot be made into a rigorous causal-nomological science.

11.3—
Dialectical Possibilities

How does one arrive at the Received View? One story about the underlying train of thought goes as follows: You start by looking at a few physical sciences (particularly mechanics, thermodynamics, and chemistry) and a few notable episodes of scientific accomplishment involving intertheoretic connections (the atomic theory, the derivation of thermodynamic equations from statistical mechanics, etc.). On the basis of observations of these actual domains of scientific research and actual interdomain connections, you form second -order theories about the proper form of all scientific discourse (or even all metaphysically respectable discourse), and about the proper relations between discourses. The resulting view is one in which the objects of the special sciences are differentiated by the kinds of physical processes they study, and more particularly by the levels of complexity of the processes they study. The more basic sciences study simpler objects that form the proper parts of objects studied by the higher-level sciences, and ultimately you should be able to explain the higher-level properties as derivative from the lower-level properties. Let us call this picture the "Hierarchical Picture." The Hierarchical Picture is a second-order theory about the canonical form of discourses in the special sciences (and indeed for all discourses speaking about real objects) and about the connections between them. Because its paradigm examples all involve straightforwardly physical objects, a materialistic inventory seems implicit in the model. The Received View then applies the Hierarchical Picture as a norm for looking at actual discourse about the mind and actual attempts to form connections between mentalistic discourse and other kinds of discourse, such as that of neuroscience.

This is arguably a very charitable way of interpreting the emergence of the Received View. One might well point out, for example, that the essentials of the Hierarchical Picture were already present in Hobbes,

[Full Size]

Figure 14

who tried to make the Galilean method of resolution and composition into the basis for a metaphysics long before most of the scientific accomplishments that might plausibly be thought to support the Hierarchical Picture. One might well take the view that it is the Hierarchical Picture that is driving the interpretation of science, and not vice versa. Or, one might point out the many advances in special sciences that would not have taken place had the scientists of the day placed a higher priority upon conformity with the Hierarchical Picture than with discoveries in their respective local discourses. But it is, I think, the kind of story that adherents of the Received View like to tell, and we can grant it for present purposes.

What I wish to point out about this story is the complex dialectic between four separate kinds of concerns (see fig. 14):

(1) the state internal to each of the separate discourses (e.g., the actual state of commonsense discourses about action, neuroscience, psychophysics, physics),

(2) one's theories about the forms individual discourse do or ought to take (e.g., the philosophy of psychology, the philosophy of physics),

(3) the state of connections between discourses (actual intertheoretic reductions, localizations, correlations), and

(4) one's theories about the forms connections between discourses do or ought to take (e.g., reductionism, supervenience, the Neutral Project).

Behind this prevailing naturalistic current in contemporary philosophy of mind is an assumption that the status of our discourse about the mind is to be held up to the litmus of conformity with metatheories about the nature of particular discourses and about the connections one ought to be able to find between discourses. In particular, it is assumed that real intentional states would play a role in a causal-nomological science of the mind, and that they would ultimately be derivative from nonintentional phenomena in much the way that, say, thermodynamics is derivative from statistical mechanics. Thus this Received View canonizes certain metatheoretical assumptions about the nature of sciences and the connections between science and ontology, and then applies these assumptions as norms for judging the form of actual work in psychology and connections forged between psychology and other domains.

While the dialectic underlying the Received View may seem very compelling, it is by no means the only serious approach that has been taken. Recent philosophy of science, for example, has increasingly moved towards an approach that looks first and foremost at the actual dynamics of discourse within the special sciences—with the result, for instance, that the teleological categories of evolutionary biology have won increasing acceptance against older mechanistic objections. On another front, writers like Ryle and Wittgenstein have insisted that intentional explanation is not scientific discourse, or even protoscientific discourse, at all, and should not be forced into that model. And continental philosophy has decided that the Geisteswissenschaften or "sciences of mind and culture" are fundamentally different in their form than the natural sciences, and have adjusted their metatheoretical views in light of this observation.

This brings up two questions. First, does the Hierarchical Picture really capture the form of current work even in the natural sciences? And, second, is the healthiest dialectic between (a ) actual practice in the local discourses of the special sciences (and, for that matter, nonscientific discourses) and (b ) our metatheories of the same, one in which the metatheories are applied as a test for legitimacy of the assumptions of special discourses—or should things, perhaps, be the other way around? And to these one might well add a third question: namely, do all of the considerations relevant to the assessment of specifically scientific discourses apply more generally as well? In other words, must discourse be scientific to be respectable, and must object kinds appear in scientific explanations in order to be real?

In the remainder of this chapter, I shall deal principally with the sec-

ond question regarding the healthiest form of the dialectic between the successes of local discourses and our metatheoretical views about the forms of those discourses and the relations between them. I shall argue that the Received View has the proper dialectical relationship backwards, even by standards expressed in the pieties of many of its proponents. To wit, it is principally our metatheories that stand in the dock against successes of special discourses, and not vice versa. And, as a consequence, intentional categories stand in no need of vindication. Along the way, I shall point to a very few of the possible examples of important work in the physical sciences that have strayed from the course the Received View might have urged. I shall also make a case for separating science and ontology in such a fashion that the metaphysical legitimacy of mental categories in no way depends upon the possibility of a science that deals with them.

My plan, then, is to look at the assumptions underlying the perceived need for a vindication of intentional psychology, and to show how the dialectic that drives the perception of this need for vindication is wrongheaded. I shall first consider concerns about extending the causal-nomological model of explanation to psychology, and then treat of the desires for materialism and the generality of physics together.

11.4—
Psychology, the Mental, and Causal-Nomological Explanation

First, consider the issue of whether intentional psychology can be made into a causal-nomological science. This issue concerns the dialectic between (a ) the actual form of our understanding (both commonsensical and scientific) of mental states and processes and (b ) the prevailing view that scientific discourse is causal and nomological in character. There are really two separate issues here, of course, though they tend to get conflated: (1) the issue of whether there can be a psychological science that is causal and nomological in character, and (2) the issue of whether entry into nomic causal relations is relevant to ontology . The relationship between science and ontology will be discussed later. Here I shall concentrate on the dialectic between the causal-nomological picture of science and the actual state of our discourse about the mental.

There are two basic ways of coming to the conclusion that intentional explanation cannot be causal explanation based on causal-nomological regularities. The first way is to think that intentional explanation is part

and parcel of an enterprise called "folk psychology" which is an attempt at protoscientific explanation of behavior, and then conclude that it fails in its attempt to capture causal regularities in the intentional vocabulary. This view that commonsense intentional psychology is an attempt at causal-nomological explanation, sometimes called the "theory-theory," is fairly prevalent in cognitive science circles. It will be discussed in the following section. Before turning to it, it is important to see that there is also another way to come to the conclusion that intentional explanations cannot be causal-nomological explanations: namely, to believe that they were never intended as such in the first place .

11.4.1—
The Verstehen Argument

The claim that mentalistic discourse does not even attempt to provide causal explanations has had important advocates in both the analytic and the continental traditions. It includes the Verstehen tradition in the Geisteswissenschaften on the continent, Ryle and Wittgenstein, and lately a related sort of view has attracted interest from analytic philosophers like Alvin Goldman (1992, 1993a, 1993b,). On this view, intentional explanation involves giving reasons, and giving reasons is different from giving causes. The kind of "explanation" we are involved in when we allude to people's beliefs and desires is not causal explanation, but something like interpretation . The goal of investigation framed in the intentional idiom is not knowledge of general laws of thought, but in Verstehen, or interpretive understanding, of human beings.

In a certain way, I am very sympathetic to this view, especially in the ways in which it stands in contrast to the theory-theory. It is no doubt true that "commonsense psychology" involves a "theory" in the weak sense that we make generalizations about how people are likely to think and act, and that these expectations, were we to express them, would be expressed in the intentional vocabulary. But it seems a bit dubious to equate this with an attempt to formulate causal laws, and very highly questionable to refer to intentional states as "theoretical entities" as that expression is sometimes used in the philosophy of science (i.e., to signify entities posited through a process of retroduction). Particularly when we refer to dispositional states such as beliefs, it seems quite reasonable to say that what we are doing is trying to paint a picture of the person's thoughts that makes sense of their words and deeds, and it would be somewhat strained to say that we are trying to provide the causes of their

actions, much less that we are tacitly assuming universal laws formulable in the intentional idiom.

Nevertheless, it seems important here to separate two very different sorts of issues. We might, on the one hand, concentrate on issues relating to the nature of commonsense intentional psychology. On the other hand, we might undertake quite a different sort of inquiry into the prospects for a causal-nomological intentional psychology. It is useful here to distinguish four very different questions:

(1) Is the theory-theory an accurate representation of the nature of commonsense psychology? (That is, is commonsense psychology really an attempt at a scientific theory based on causal laws?)

(2) Is there a viable enterprise of interpretive psychology whose goal is Verstehen?

(3) If so, should we call it "science"?

(4) Is there a viable causal-nomological psychology that makes use of generalizations at the level of cognitive states and processes?

The reason that it is useful to separate these questions is that the answer to the fourth question—our main focus here—is really quite independent of the answers one gives to the other three. In point of fact, I think that one might have to give a somewhat mixed answer to the first question, since the characterization of commonsense views is likely to be a somewhat complicated undertaking; but I am inclined to side more with Ryle and Wittgenstein here than with the theory-theory. I take it that the second question can best be answered by looking at what is accomplished by attempts at interpretive psychology (e.g., psychoanalysis as interpreted by Ricoeur [1970]) in the long run. The third question is in large measure a matter of wrangling over words. In some ways, I think the German approach has some real advantages: if you have two categories of Wissenschaften you can then ask serious questions about how they differ; whereas if your distinction is between "science" and "nonscience," there is an unfortunate tendency to assume that everything that is nonscience is also nonserious and nonrigorous, and that all such enterprises are nonserious and nonrigorous in the same ways. All in all, though, I find this fight less interesting than I used to, and do not intend to belabor it here.

What seems crucially important here is that the viability of a causal-nomological intentional psychology is in no way threatened either by the

possibility of interpretive psychology or by the possibility that commonsense psychology is not accurately conceived of as protoscience. One might demonstrate this point by way of a thought experiment. Suppose, for example, that someone were convinced by writers like Wittgenstein and Ryle that the "because" of intentional explanation (in ordinary usage) is not the "because" of causation. Suppose, furthermore, that she believed that ordinary intentional explanation was part of a project of interpretive understanding (Verstehen ), and that there were disciplines—the Geisteswissenschaften —that attempted to approach such interpretative understanding in a systematic way. She might think, for example, that Ricoeur's Freud was an active and even an occasionally successful practitioner of such a discipline. Suppose now that one day she comes across something like Colby's (1975) attempts to implement models of Freudian theories in the form of computer programs. Upon examining these she might well feel that she has discovered a new possibility for explanation in the intentional idiom: in addition to the possibility of an interpretive psychology, it might be possible to systematically explain causal relationships among intentional states as well. It might even turn out that there are important connections between them—for example, that rules for deriving new cognitive state tokens are so formed as to preserve truth and maximize coherence and relevance. Our fictitious person could thus discover the idea of a nomological intentional psychology as a project quite orthogonal to commonsense psychology and interpretive psychology, but in no way in competition with them.

What I think this shows is that neither the Ryle-Wittgenstein view of intentional explanation nor the Geisteswissenschaft approach to psychology should be viewed as counting against the viability of a nomological intentional psychology. These views would, of course, undercut the claim that the success of commonsense psychology in allowing us to predict one another's behavior in day-to-day affairs provides direct evidence for the possibility of a nomological intentional psychology. But it need not rule out providing such evidence more indirectly: if we are successful at interpreting others, it might be in part because we have a kind of model in our heads of how people are likely to think and act given certain beliefs and desires, and this at least suggests the possibility of a nomological account of how different beliefs and desires in combination will produce other cognitive states and, ultimately, behaviors. In short, even if "folk psychology" is a dubious philosophical reconstruction, this does no fundamental damage to the project of nomological intentional psychology.

11.4.2—
Intentional Categories and Causal Relevance

Establishing compatibility between causal explanation and Verstehen (as two separate projects), however, is the easy part of the job. The hard part lies in making a case that intentional psychology is susceptible to being made into a "science" in the prevalent usage of the word in America, where it suggests a causal and nomological character. One important problem lies in the fact that intentional psychology individuates its objects semantically. But if it is to be a science, one would wish for the individuation of terms to take place in a fashion that captures causal regularities: if a belief that a cat is in the yard is used to explain different behaviors than a belief that a unicorn is in the yard, and we are talking about a kind of causal explanation, we should expect the difference in the causal powers of the two beliefs to be intimately connected to their difference in content. But it is hard to see how this might be the case, and it has generally seemed that there are only three basic possibilities: (1) there is some interpreter that is sensitive to semantic properties and is also the locus of the causal powers, (2) semantic properties are themselves causally efficacious, or (3) semantic properties are linked to some other kinds of properties that are causally efficacious. The problem with (1) is that it leads to a homuncular regress. The problem with (2) is that semantic properties just do not seem like things that have causal powers. The computer paradigm takes option (3) with respect to semiotic-semantic properties, which are unproblematically linked to nonsemantic properties through interpretive conventions, but it is not clear how mental-semantic or MR-semantic properties might similarly be coordinated with some other properties that are causally efficacious, or what those properties might be.

The issue here is one of trying to see how something might be the case —that is, how it might be that causal regularities of cognition could run parallel to semantic properties of cognitive states. And the issue of "seeing how X might be the case" shapes up very differently depending on what kind of evidence we have that X actually is the case. For example, many people in Newton's day (and perhaps many of us today) found the picture he presents of gravitation to be problematic because it involves action at a distance. The familiar paradigm of causal interaction had long been one of contact interaction, and it seemed—indeed, it still seems—hard to see how bodies could exert influence somewhere they are not. This difficulty in seeing how it could be so might have been seen as a compelling argument against causation that does not involve contact

interactions, except for the fact that there was overwhelming evidence that such causal influence did take place. (Indeed, of the fundamental "forces" posited by contemporary physics, none of them aside from mechanical force involves contact interaction.)

I think that the same kind of benefit could, in principle, accrue to psychology: if there were to emerge an intentional psychology framed as a set of laws governing reasoning, and it had a sufficient degree of predictive accuracy, this would provide strong evidence that intentional states do have the kind of causal role assigned to them by such a theory. And it would provide such evidence regardless of whether we can see a mechanism that could account for such causation . The particular form of intentional causation could be fundamental, after all, like the particulars of gravitation or electromagnetism or of why particles behave in the precise way they behave when they collide. Or intentional causation could be opaque to us without being fundamental—it could be that we can have evidence that there are causal regularities at the level of intentional explanation, and it might also be the case that these are emergent out of some more basic kinds of regularities without our being able to know just what the relationship between levels is. (Even if computational formalisms are appropriate for psychology, it seems all too likely that we shall never know the details of interlevel relationships in detail.) In such an eventuality, its status would be not unlike that of Newtonian description of gravitational attraction, which supplied a nomological description (and hence conferred mathematical maturity) without explaining this behavior by positing an underlying mechanism (and hence did not supply additional connective maturity).

The first part of the answer to the second objection to intentional psychology, then, is this: computational psychology provides one of the first chances we have really had for making any realistic attempt at building nomological theories of cognition that treat intentional states as causally relevant in reasoning and behavior.^[2] The programme is relatively young, and any possible model of the mind rich enough to test for predictive accuracy with respect to cognition would necessarily be orders of magnitude more complex than most of the fundamental laws in other sciences.^[3] (Though many researchers seem to think that some of the initial results seem promising.) The obvious moral to draw is that we ought to let cognitive psychology mature as best it may, and see whether it does provide compelling evidence that semantic properties of intentional states seem to be at least correlated with causal regularities.

There is also a second response to this objection—one that draws more

directly upon what is made explicit in BCTM. BCTM makes a fairly bald-faced assertion to the effect that intentional states are realized through naturalistic properties in a systematic way, such that (a ) intentional modality is realized through functional relations and (b ) mental-semantic properties are realized through some naturalistic properties (labeled "MR-semantic properties" until their true identity is discovered) possessed by cognitive counters. And it is important not to be fooled by the "semantic" in 'MR-semantic'. Remember that all 'MR-semantic properties' means is "those properties of cognitive counters, whatever they turn out to be, through which content is realized." In particular, the "semantic" in 'MR-semantic' is not so robust that our intuitions that "semantic properties" cannot be causally efficacious should transfer to MR-semantic properties. MR-semantic properties, if they are anything at all, are just naturalistic properties whose real identity we have not discovered, and so they seem perfectly respectable as possible explainers of causal regularities (though whether they capture the right regularities to make our intentional explanations into causal laws remains to be seen). If the research programme associated with BCTM can be carried out, then there will be a causal, nomological science to be carried out, at least at the level of the realizing system . That is, the system of cognitive counters will have lawlike causal regularities, and the MR-semantic properties through which mental-semantic properties are realized will be at least correlated with the causal properties of the cognitive counters of which they are the properties.

The difficulty here is this: the level of description for the realizing system is not a level of psychological description per se. What can the causal, nomological character of the realizing system show us about the realized system, the system of intentional states and processes? Does the causal, nomological character of the realizing system automatically accrue to the system it realizes? Or does it, perhaps, suggest that the realized system also has such a character? There are, I think, two parts to the answer. First, even if we could not construe the system of intentional states and processes as a causal and nomological system, it would matter a great deal if we could show that this system is realized through a system that did have these virtues. For example, if there were overwhelming Rylean objections to interpreting intentional states as being even the sorts of things that can enter causal relations, the psychologist can at least take heart at the news that there is some other system of states and properties—those through which intentional states and processes are realized—that can enter into such relations, and moreover, that whenever one picks

out intentional states and processes, one picks out the states and processes through which they are realized as well. (If you go to the racetrack, you can bet on the horse or bet on the jockey on the horse; they win or lose together!)

But I think one can make a case that at least some of the properties of the realizing system can also accrue to the realized system as well. Take our example of the Victoria Crown. The property of being the Victoria Crown is realized through a particular bit of matter. Let us say it weighs fifty pounds. Now there is nothing about the property of being the Victoria Crown that entails weighing fifty pounds. But the object that is the Victoria Crown does weigh fifty pounds, and does so because the matter of which it is composed has that mass. This property of the realizing matter accrues to the realized object as well. Similarly, suppose Jones's generosity is realized through his giving $100 to the Presiding Bishop's Fund for World Relief on Pentecost. The act takes place at a particular time and has a particular beneficiary. Now we should not say that Jones's generosity is an event taking place at a particular time, but we might say that it was exercised at a particular time. And we should surely say that the Presiding Bishop's Fund was the "beneficiary of Jones's generosity" and not just the recipient of $100. So it seems that, at least in some limited ways, the realized system can take on some of the properties of the realizing system. (It does not follow, of course, that the properties of the realizing system accrue to the realized property —the property of generosity does not take on new implications because of how it is realized.)

A detailed examination of the various possible relations between realizing and realized systems would probably require some very careful metaphysical investigation. It seems quite reasonable, however, to suppose the following: if (1) a mental state X of type M is realized through a natural state of type N , and (2) it is a law that N 's cause O 's under condition C , and (3) C obtains, then (4) an O will come about, and (5) X may be said to be a cause of O . (And this holds even if we do not say "X is an N " but only "X is realized through an N -token.") Here the causal powers of the instantiating type N accrue to the realized individual X , but not to the realized property M .

Now none of this precludes the possibility of saying that intentional states have causal powers in their own right, and not just by virtue of how they are realized. The point is merely that, if computational psychology as described by BCTM can be carried out at all, there will be some naturalistic system through which intentional states and processes are realized, and this system can be causal and nomological in ways that

are not at all problematic. If there are problems about intentional and semantic properties being of the right sort to ground causal regularities, perhaps this gap can be filled by way of appeal to the realizing system.

To summarize, there really are some prima facie difficulties with the attempt to construe intentional explanation as lawlike causal explanation, and hence to make intentional psychology into a "science" in the sense of the word that implies such causal lawlike explanation. The issue of the nomological character of intentional explanation, however, is best settled by letting the project of intentional psychology, supplemented by resources of the computer paradigm, flourish as best it may and seeing whether the project will pan out in the end. And if it does produce what look like nomological regularities, this in itself provides substantial warrant for suspecting that these regularities are causal in nature as well, whether or not we can find or even imagine an underlying mechanism that could account for the causality . In any case, if BCTM can be carried out, there will also be a system describable in wholly naturalistic terms through which the system of intentional states and processes is realized. And this system (a ) can unproblematically be viewed as causal in nature, (b ) is in one-to-one (or even many-to-one) correspondence with the system of intentional states and processes, and (c ) may even confer its causal properties upon the things it realizes.

11.4.3—
Need Mentalistic Discourse Be Scientific to Be Legitimate?

To tell the awful truth, though, I have my doubts about whether there can be a causal-nomological science of the intentional. Apart from all the problems in formulating actual theories with so many hard-to-isolate, mutually dependent variables, I share the Wittgensteinian suspicion that ordinary-language belief ascriptions are not causal explanations; and hence, whatever a computational psychology might do, it would not render whatever ordinary language is doing in such cases scientific, but add a new kind of discourse, perhaps only loosely inspired by the original. I also believe in free will, which seems in tension with a thoroughgoing nomological psychology. There are also other reasons that I find harder to articulate. But none of this makes me doubt the existence or legitimacy of mental states . And this is because I do not think that the considerations that exclude an object from a specifically scientific ontology (i.e., the domain of a science) exclude it from ontology generally. I will discuss some parts of this issue here and others later in the chapter.

First, different kinds of discourse have different purposes, and the conditions for legitimation for a given kind of discourse vary with its purpose. The natural sciences aim at describing and explaining the regularities of nature. The good-making qualities of the natural sciences are thus conditioned by the practical constraints governing what counts as a good enough explanation, the need for truth-conditional evaluation, and the availability of real regularities to be found in nature. Most of our human discourses, however, have different sets of constraints. Many speech acts do not have truth conditions at all, but other kinds of felicity conditions (see Austin 1962; Searle 1969). Indeed some entire language games may lack truth conditions, though of course they have felicity conditions of some kind. For example, if two people play mental chess, saying "King to Queen one" constitutes a move, not an assertion. Now some people would go so far as to think that mental state ascriptions are similarly nonassertoric. And perhaps some mental state ascriptions are not assertions and do not have truth conditions. But I think many of them are assertions and are subject to truth-conditional evaluation.

What I think they might lack is a nomic character. Categories of nonscientific discourses can be legitimate and can pick out real objects even if those categories do not pick out things that are the subjects of natural laws. Dollars can be bills or coins, and there are no physical laws regarding dollars; yet I do not conclude that I am broke.^[4] I suspect that there are no laws regarding Toyotas, but I do not feel trepidation when I look out in the driveway every morning in fear that my car was an illusion. I suspect that there are no natural laws applying only to jigs, but I am not dissuaded from buying recordings of Irish music as a result. There are no physical laws regarding numbers, but that does not excuse me from balancing my check book properly or tallying my taxes. Nor do my doubts about natural laws governing dollars (or Toyotas or jigs) give me reason to accuse my employer (or my car dealer or the Bothy Band) of fraud. In general, the failure of a category to appear in a science does not cause me to doubt the reality of things to which that category is supposed to apply, or to doubt the legitimacy of the category. Indeed, the only cases in which I am inclined to make inferences in anything like this way are those in which an object-kind is introduced by hypothesis within the context of a scientific theory (e.g., phlogiston). I do, of course, doubt the existence of trolls and unicorns. But the reason I doubt them is that there seems to be no reliable evidence of their existence. It is not because there are no theories in which the kind "troll" or

"unicorn" plays a nomic role. If I were inclined to doubt trolls and unicorns on that basis, I would have to doubt dogs and Toyotas as well. In short, what you need to qualify a category for legitimacy in a scientific theory is something far stronger than what you need to qualify it for mere ontological legitimacy. Scientific kinds are reasonably orderly. Ontology is a motley.

11.5—
Intentionality, Materialism, and the Generality of Physics

A separate set of issues lies upon a second axis: the relationship between the current state of discourse about the mental and metatheories about the relationships one ought to find between separate discourses. CTM was supposed to "vindicate" intentional psychology by demonstrating the compatibility of intentional realism with materialism and the generality of physics. Such a vindication is only necessary, however, on the assumption that compatibility with materialism and the generality of physics is a condition for the legitimacy of psychology as a science or of mental states as real entities. This position takes a particular picture of how various discourses do or ought to fit together—a picture in which all special discourses are in some sense a special case of physics—and applies it as a norm for evaluating the current state of discourse in psychology and commonsense discourse about the mental. (Indeed, if you reverse the dialectical relation, so that the commitment to mental states is held constant and the claims of materialism are weighed in the balance, computers might be thought to provide a "vindication of materialism" instead of a vindication of intentional psychology.)

In order to better assess the strength of arguments against intentional realism based upon concerns about materialism and the generality of physics, it may be helpful to make explicit some of the prevalent views about ontology and its relation to science in the context in which the apparent need for vindication arises. In particular, I shall discuss three views whose popularity is to no small extent traceable to Quine's influence.^[5] First, Quine is, of course, well known for the "desert landscape" plea for the impoverishment of the ontological inventory and the association of this view with a presumption in favor of some form of materialistic monism. (I personally have always found it curious that people found this view appealing, and even more curious that they regarded an expression of taste as an argument. I like lush landscapes and seashores; and, more to the point, would be disinclined to try to describe a rain-

forest as a desert just because I liked deserts.) Second, Quine was influential in the dissemination of the now widespread view that all ontological questions are essentially questions to be answered by science. And, finally, Quine seems to bear much responsibility for the currency of the view that the business—indeed, the only business—of ontology is to provide an inventory of the basic kinds of things . I think that most philosophers who do not have one foot (or at least a few toes) outside of late-twentieth-century analytic philosophy simply take this last view for granted. For an eye-opening discussion of how ontology has been viewed historically, I would point them to the entry on ontology in the Encyclopedia of Philosophy .^[6] In brief, ontology has traditionally involved not only the question "What (kinds of) things are there?" but a number of questions about the nature of being —for example, "What is it to be a thing?" "What is it for X and Y to be the same thing?" "Is 'being' said in the same sense about different kinds of things?"

For the record, I think that all of these Quinean views are very deeply wrong. However, I shall not undertake a full-scale assault upon Quinean ontology here. Instead, I shall confine myself to making two much more modest points. First, there are certain ways that ontology can, in principle, extend beyond what science can talk about, with the consequence that intentional realism can be ontologically respectable even if intentional psychology cannot become a mature science. Second, if one takes seriously the notion that science ought to guide ontology, one ought to take the attitude that a successful psychology committed to intentional realism would give us warrant for believing in intentional states and processes, whether they be compatible with materialism and the generality of physics or not—and indeed that if push comes to shove between intentional realism and materialism and the generality of physics, it is the latter that stand in the dock. Hence intentional psychology does not stand in need of vindication on this score.

11.5.1—
Why Ontology Extends Beyond Science

There are several reasons that ontology extends beyond science. First, ontology has traditionally been concerned with questions above and beyond those concerned purely with inventory. It has been concerned with questions about the nature of being and unity, and with such issues as whether there are different kinds of principles for "being" and for "being one thing" for different kinds of "objects." Granted that there is a kind of "being" that applies to material simples, the question remains of whether

there are other kinds of "being" that apply to other things. Peter van Inwagen (1990, 1993), for instance, has recently defended the position that there is also a principle of individuation to be found in something like the "transcendental unity of apperception"—namely, that subjects capable of a certain kind of thought are thereby things in a sense denied to mere physical aggregates. Aristotle (Metaphysics, IV-VIII) thought that "being" was said in a primary sense only of members of what we today might call "natural kinds," that it was said in derivative senses of properties and relations, and that it was somewhat of a stretch to say it of mere matter at all. Charles Sanders Peirce (and perhaps the late Plato) saw three realms of being: matter, abstract objects such as mathematical objects, and minds.

It is important to see that such thinkers were not simply engaged in projects of empirical science that somehow went awry. They are engaged, in part, in linguistic and conceptual analysis; but insofar as our ideas of being and unity reflect how things really are, they are engaged in attempts to clarify the nature of being and unity as well. Such a project may have results that stand outside of empirical science: for example, if there is a sense of "being" that applies to abstract objects, those objects stand outside of the realm of empirical science. If van Inwagen is right that there is a principle of being and unity that attaches to a thing by virtue of being a thinker, then there is a reason to include thinkers in your basic ontology even if wherever there is a thinker there is also a body of a specific kind (physical or functional) as well. That is, if there is a principled reason for allowing things other than physical simples to count as basic kinds, then one cannot simply assume in advance that the simplest inventory needed for science (by that meaning the inventory of simplest parts employed in our scientific theories) is all we need for ontology.

A second reason for seeing ontology as extending beyond science is that some things that might be thought of as objects and unities seem to stand outside of the domain of empirical investigation—most notably, abstract objects such as numbers and sets. I take it that there are serious questions about the ontological status of abstract entities. It may or may not be possible to settle such questions at all; but it seems clear that empirical science cannot settle them, nor can such answers as might emerge from the analysis of empirical science. Perhaps there are also other sorts of objects (or possible objects) that are similarly unsuitable for empirical investigation of the sort conducted in the sciences—for example, God, angels, objective values—in which case there are other ontological issues that fall outside of the domain of science.

A third reason (already touched upon) that ontology can extend beyond

science attaches to the systematic character of science. Scientific description applies to the universe only insofar as it is describable in ways involving systematic features of structure and causation. Many of our object categories, however, are not implicated in such systematic relations: the category "lamb chop" has little to do with how to carve nature at the joints and much to do with how we carve meat at the joints. Yet there are surely sensible questions to be asked about the senses in which it is right to say a lamb chop "exists" and is "one thing," how these differ from the senses in which "unity" and "being" may be predicated of living animals, simple particles, or numbers, and why a lamb chop is not just a "heap" of atoms. Such questions fall within the scope of what has historically been called "ontology," they seem like sensible questions, and they may shed some light on the questions that do have something to with science. Moreover, the inventory view of ontology accepts as a premise one possible answer to questions of the sort raised here—that is, it is one contender in a field of possible answers about how we ought best to talk about various sorts of things that common sense treats as "objects." Hence there are ontological questions (i.e., questions about what ontological approach to adopt) that cannot be addressed within the inventory approach.

This last consideration brings up the possibility that there are ontological questions to be asked about intentional states that are not questions about the analysis of intentional psychology as a science. If one takes it that one has reason to believe there are either dispositional beliefs or inner episodes such as occurrent judgments, one might sensibly ask ontological questions about them: for example, is "being" applied to such states as it is applied to objects such as living beings, or as it is applied to properties or states, or in some other fashion? And these questions are no less proper if there is no science of intentional states. In short, even if intentional psychology fails as science, this does not have the implication that intentional states are any less ontologically respectable than any number of other things that do not fall into the categories used for explanation in an ideal science—that is, they are no worse off than dogs or lamb chops or numbers. (Likewise, the consciously accessible occurrent states are no more "theoretical" in nature than are dogs, lamb chops, or numbers.)

11.5.2—
Intentional Psychology, Ontology, Generality

If this last sort of consideration is helpful with respect to intentional realism generally, it does nothing for intentional psychology as an attempt

at a rigorous science. Here we need to explore the relationship between the project of intentional psychology and commitments many have to materialism and to the generality of physics. Now while Quineans officially state that what we should allow into our ontological inventory is the simplest inventory needed for science, in practice their emphasis tends to be on the simplicity rather than upon what is needed for science. That is, there are really three claims to be distinguished here:

(1) Ontology ought to include all entities that are required for scientific explanation.

(2) Ontology ought to include only entities that are required for scientific explanation.

(3) Given a choice between two different scientific pictures, one should opt for the one which posits the fewest basic entities.

I wish to argue, first, that anyone who accepts (1) ought to take the success of a special science as evidence that the entities it describes and posits actually exist and, second, that if push comes to shove, getting good explanations at the right level of description is a more important value than is having a very simple inventory.

Different sciences are distinguished from one another in large measure by the proprietary vocabularies they use and the descriptive and explanatory categories they employ. Categories such as "fault line," "high pressure system," "predator," and "desire" are employed by geology, meteorology, biology, and psychology, respectively. For these special sciences to give the kinds of explanations they need to give, they in some sense need such categories, and in many cases need to posit "entities" corresponding to them. But what ontological conclusions ought we to draw from the success of a science? Perhaps we can draw only prima facie conclusions, but it seems that we ought to count the success of a theory as evidence for the existence of the objects to which the theory is committed. Such commitments can, of course, be undermined by competitor theories; and theories can come into conflict with one another. But if we are to take the enterprise of the special sciences seriously at all, we have to be willing to entertain a prima facie commitment to both the things they explain (earthquakes, storms, wolves eating sheep, decision making) and the things invoked to explain them (fault lines, high pressure fronts, predation, desires, etc.).

Now to all of this the Quinean has a perfectly straightforward response: namely, that he is quite willing to admit that fault lines and preda-

tors "exist" in some sense, but that ontology, in his sense, is not interested in the rich abundance of being that admittedly is to be found in the universe, but in the basic entities, properties, and relations out of which that abundance is generated—the things out of which everything else is made and which are themselves reducible no further. And thus it is important to separate two different issues about ontological status: (a) the distinction between legitimate entities and pseudo -entities, and (b) the distinction between basic entities and compound entities. For it is clear that there are all sorts of entities—rabbits, rabbit forelegs, buttonholes, Corinthian columns, Wagner operas—that are unproblematically real (as opposed to fictitious or unreal), yet are neither included in nor reducible to the explanatory vocabulary of the sciences. The "desert landscape" approach does not work like the replacement of phlogiston with oxygen, but like the explanation that water is H₂ O. The point is not that, in some intuitive sense, there "aren't any" rabbits or buttonholes—or, for that matter, mental states and processes. The point, rather, is that simplicity of basic ontological inventory is to be viewed as a virtue for a theory, and that rabbits and buttonholes (and perhaps even mental states) are complex phenomena whose ultimate parts are all of a very few kinds—namely, the kinds of basic particles recognized by an ideally completed physics.

There are, of course, several ways one could interpret this kind of principle of simplicity. The extreme position is that of thinking there is an a priori case for monism. However, a more sensible way of looking at the principle of simplicity is to see it as a kind of maxim or guiding principle for doing science. Ontological parsimony might be seen as one of the "good-making" qualities of science, a part of the "elegance" that has apparently proven a good guide to finding viable theories in physics in this century. Such a principle must, however, be played off against other principles: a theory with a larger basic inventory might well be preferable to a more frugal theory if it also has greater explanatory power or more elegant laws. There is a point at which a "taste for desert landscapes" would cease to be a reasonable inclination towards elegance and begin to degenerate into a mania for monism. In particular, if one truly believes that entities that are needed for science are thereby ontologically warranted as well, it is important not to dictate to science in advance what entities it is allowed to need.

Now if one takes this point seriously, the whole rationale behind criticizing intentional psychology on the basis of a possible incompatibility with materialism seems wrongheaded. If science were really to dictate

ontology, the proper strategy would seem to be to wait and see if we could get a good explanatory psychology that could capture the relevant generalities about thought and behavior, then see what entities it was ultimately committed to. Writers like Fodor (1975, 1987) and Pylyshyn (1984) have argued (quite persuasively to my mind) that there is a broad range of psychological phenomena for which the only kinds of explanations we have that seem to capture the right generalizations are cast in the intentional idiom. The explanatory success of such theories, and the lack of competitor theories, they rightly argue, provides significant warrant for assuming, provisionally but with confidence, the existence of the entities posited in the explanations.

This, however, makes a case only that intentional states are legitimate entities rather than pseudo-entities, not that they are ontologically basic. The point to be made there, however, seems perfectly straightforward: if you are methodologically committed to letting what is needed by science into your ontological inventory, and you have a psychology that provides warrant for the existence of intentional states, then you have at least a prima facie commitment to whatever kinds of things intentional states turn out to be. If they can be accommodated within a materialist inventory, hooray for simplicity. But if they cannot be so accommodated, so much the worse for materialism . It is one thing to be committed to letting science determine what falls within the basic inventory; it is quite another to let science do so, but only so long as the results are consistent with materialism . On the one approach, the "vindication" of intentional psychology will stand or fall with its explanatory success, and the question of whether intentional states are basic in the inventory will be answered by analysis of the relationship between the resulting psychology and other sciences. On the other approach, the "vindication" of intentional psychology would consist in its being held to a standard of ontological orthodoxy. Unless some compelling a priori argument for materialism can be marshaled, it is hard to see why either science or ontology ought to be held to such a standard.

I confess that I have never found anything attractive about materialism in any case, but it seems to me that even those who do find it attractive ought to consider the following scenario very carefully: suppose that computational psychology (or some other research programme) were to bring intentional explanation to a stage of considerable mathematical and connective maturity and to supply general explanations that displayed a good measure of predictive accuracy. This is, I think, the scenario that most advocates of CTM think is suggested by current research.

Suppose, further, that an analysis of the resulting psychology revealed a commitment to something over and above what we were committed to by physics. Do we then (a ) throw out our psychology even though it is respectable in relation to the values internal to science, or do we (b ) decide that we now have good scientific grounds for rejecting materialism? I believe that (b ) would be the more reasonable course to take in such an eventuality. But perhaps more to the point, if (b ) is the more reasonable option to choose if push comes to shove between intentional psychology and materialism and the generality of physics, then it is likewise wrong to hold intentional psychology to proving its compatibility with materialism in advance. If a successful intentional psychology could call materialism into question, it is quite wrongheaded to expect intentional psychology to justify itself in advance by demonstrating compatibility with materialism.

11.6—
The Commitments of the Special Sciences

There is also another and more fundamental way of opposing the Quinean view of the relationship of ontology and the sciences. The Quinean view seems to equate "the entities to which science is committed" with the basic components out of which the objects described by the sciences are composed. One can, of course, equivocate on words such as 'object' and 'being' so that this view is necessarily true. But there is also a sense in which the "objects" to which the special sciences are "committed" are not the basic particles of the physicist, but things like high pressure fronts, fault lines, and paranoid delusions . For the generalizations of the special sciences are cast in vocabularies that are distinctive of those sciences, and unless generalizations can be made at those particular levels, one ceases to have explanations that are specifically meteorological, geological, or psychological. Now if one thinks that one is ontologically committed to the things one quantifies over in the statement of the laws of ultimately completed sciences, and the special sciences need to be formulated in ways that quantify over things other than the simple objects of physics, it would seem that the level of objecthood to which science commits us is not restricted to that of fundamental particles, but to objects of any of the types that are required for scientific explanation.

It is tempting to think that the distinction between type and token physicalism will take the bite out of this problem posed by the special sciences, but I think this is not fully true. We have seen in recent years

that some kinds of explanation that seem useful in such fields as computer science, biology, and psychology—notably functional explanation and explanation involving such categories as "selection" and "adaptation"—cannot be reduced to statements in physics. And thus any laws formulable in terms of such explanations cannot be translated into laws cast wholly in the vocabulary of physics, with the result that the kinds of explanation that are now flourishing in several of the special sciences are incompatible with type physicalism. But token physicalism is seen as avoiding this problem, since what it claims is not that all object and event types correspond to physical types, but merely that every individual object has a physical description and every individual event is predictable under a physical description by way of laws cast in the vocabulary of physics.

It does seem correct to say that the kinds of explanation we see in the special sciences do not have the prima facie incompatibility with token physicalism that they have with type physicalism. And token physicalism may or may not be true—it is not really my concern to argue that question here. What I do wish to argue is two things: First, the special sciences need to quantify over things that fall into the kinds picked out by their proprietary vocabularies, and when their categories have a one-to-indefinitely-many realization relation (as in the case of functional kinds), there is no way of specifying the same classes of objects in the vocabulary of physics. The second thing I wish to argue is that the success of a special science should urge upon us a commitment to its theoretical posits that is at least as strong as is our commitment to the propositions (1) that all of the things it names have descriptions in the vocabulary of physics or (2) that the events it describes could be predicted by laws cast in the vocabulary of physics. Hence the success of intentional psychology would commit us to the existence of intentional states regardless of whether this result was inconsistent with materialism. That is, it is token physicalism and the generality of physics that need to be tested against the special sciences, and not vice versa .

To repeat an earlier point, claims made for materialism and the generality of physics may be taken in at least two ways: they can be taken as metaphysical theses or as maxims guiding science. If they are taken as theses, they can be supported either on the basis of a priori claims or on the basis of considerations involving empirical theories. Now it is true that someone who was convinced she had an a priori argument for materialism and the generality of physics might have strong reasons to hold the special sciences to the task of proving themselves compatible with

these positions. (Likewise, it seems true that someone convinced of an a priori argument against materialism might not take any scientific evidence as warrant for materialist conclusions.) But it is not my impression that most modern materialists embrace these theses on the basis of a priori considerations. (Which is just as well, as the denial of materialism does not appear to involve one in contradiction.) Instead, there are two basic lines of argumentation that one tends to hear in support of such views: (1) the argument from simplicity (the "desert landscape" approach), and (2) the argument from the "collective evidence of modern science." In order to argue for materialism and the generality of physics on grounds such as these, it is necessary to adopt the premise that the results of the sciences can authoritatively determine the answers to ontological questions. But if the results of the sciences can be used to argue for materialism, they can, in principle, be used to argue against materialism as well, and likewise for the generality of physics. Considerations of simplicity might favor a materialist theory, but only if none of the explanatory force of the special sciences is bought at the price of incompatibility with materialism. Then one has to choose between a simpler theory that does not explain well and a more profligate theory that has greater explanatory scope. The criterion of simplicity does not tell us how to choose between theories that differ with respect to explanatory power. Likewise, if a proof that the entities required for a special science are just physical composites counts as evidence for materialism, a proof that such entities are not physical simples or composites ought to count against materialism, at least if one really thinks that ontology ought to be accommodating to successful science. And this would seem to commit us to pursuing promising candidates in the special sciences first and drawing our ontological conclusions afterwards, rather than the other way around.

Nonetheless, one might have some reason to view materialism and the generality of physics as having more of a normative status than some other kinds of claims. More specifically, one might wish to regard them not so much as claims at all, but as something on the order of maxims of scientific theory construction . That is, we have a kind of picture of what we think the overall story about the world ought to look like, and we quite reasonably try very hard to make the scientific and metaphysical stories we tell about the world exhibit the virtues of this picture. For example, we think the universe is orderly and rational (i.e., its order is of a sort comprehensible to our rational faculties), and hence we tend to look for ways to replace lots of piecemeal generalizations with a single

overarching principle. We try to explain new phenomena in terms of laws with which we are already familiar. And, more to the point, we think it would be very elegant if we had a model of the world in which there are a few basic simple entities that interact in a few well-defined ways, such that we can describe all other objects as combinations of these simples and predict all events in terms of the laws governing interactions of the simple units. We thus approach the actual scientific theories we have with an eye towards goals such as subsumption of laws under more general laws, microexplanation of phenomena at one level in terms of the interactions of their components, simplification of the ontological inventory by analyzing objects into their constituents, and so on. The limiting case of this sort of procedure is a science in which basic entities are all of one basic sort (e.g., material bodies) or a few basic sorts (leptons, mesons—substitute this year's list of basic particles) and all events that take place can be given an explanation at the level of interactions between the simple units—that is, the case in which an atomistic materialism and the generality of physics hold true.

But it is important to see that what we have here is a set of maxims for scientific theorizing that is guided by a particular view of what a picture of the world should look like. It would be a grave error to pass subtly from the view that we ought to try very hard to see whether the entities posited by psychology are physical entities, to the claims (a ) that we have shown that they are physical entities, or (b ) that they must be physical entities. It is simply false to say that any such thing has been shown. There is a serious and long-standing discussion about the question of the unity of the sciences, and it is unresolved. Deeply held pictures of what explanations ought ultimately to look like are easily mistaken for necessary truths, or truths that have been demonstrated satisfactorily. But these assumptions about what an ultimate theory would look like—a picture that looks a lot like Tractarian metaphysics minus the Tractarian account of language—might well prove incorrect. Our assumptions about what scientific explanation ought to look like have very often been wrong in the past, and one ought not bank on them too heavily.

In particular, one ought to find a way of applying methodological maxims in a way that does not prevent seeing what is really out there. For example, the maxims should be directed towards taking the special sciences that do in fact develop and attempting to unify and simplify them to such extent as proves possible. They should not be oriented towards

assuming in advance what degree of simplicity and unification one wants and then discounting theories inconsistent with these somewhat arbitrary desiderata. But this requires that we allow the special sciences to flourish and that we take descriptive and explanatory success seriously, even when it conflicts with our ontological views. In other words, it means that mature sciences should shape our views about materialism and the generality of physics rather than the other way around.^[7]

If this is the case, however, there is no need to vindicate intentional psychology against charges of incompatibility with materialism and the generality of physics. It may or may not prove to be the case that intentional psychology is thus incompatible. But it is not in need of vindication for two reasons. First, compatibility could only definitively be established for a psychology in a form far more mature than its present form. Second, commitments to materialism and the generality of physics are not things that have already been established as true against which scientific theories need to be tested; rather, it is the results of successful science that will determine whether materialism and the generality of physics are in fact correct. Of course, psychology and other special sciences may never reach a stage of maturity at which such claims can properly be assessed; but let us judge the success or failure of this maturation by standards internal to science, and not by tests of metaphysical orthodoxy.

11.7—
Final Words

While the final picture I have presented does not do all of the things that CTM was touted as doing, I hope that I have made a case (1) that computational psychology could, in principle, do some very important things for empirical science and (2) that the specifically philosophical desiderata of strongly naturalizing intentionality and vindicating intentional psychology should never have been viewed as imperatives in the first place.

On the one hand, computation may provide the mathematical resources for a successful psychology of cognition. On purely scientific grounds, this is a good thing even if it does not bring a solution to the mind-body problem in its wake. And on metaphysical grounds, those inclined to strong naturalism on the basis of what they have seen happening in the sciences can only be consistent by holding their strong naturalism accountable to the ultimate state of psychology, and not the other way around. As for the vindication of the mental, I am inclined to view the situation in the following way: If Smith accuses Jones of trespassing

on Smith's property, Jones may very well need to vindicate himself by showing that he did not break the law. But if Smith accuses Jones of trespassing on Jones's own property, there is no need for vindication, because walking on your own property is not a crime. Similarly with intentional states: there is no need to vindicate them, because failure to conform with materialism and the generality of physics are not philosophical crimes. Strong naturalizers may not like unreduced mental kinds, much as Smith may not like Jones's domestic perambulations. But you need more than a violation of taste or ideology to call for a vindication. (Of course, one might well have concerns in the opposite direction. That is, one might view successes in intentional psychology, and the inability of nearly four centuries of modern philosophers to reduce the mental to the physical, as casting substantial doubt upon materialism and strong naturalism. If that is the real concern of naturalizers, it seems to me to be a concern that is well founded.)

Where does one go from here? I think that the results of this inquiry point in two directions. First, there are questions that are purely about the philosophy of psychology: For example, how do rival research programmes in psychology today confer good-making qualities upon the psychological enterprise? To what are they committed? What are their underlying methodological assumptions? In short, it is important to do careful case studies in contemporary psychology just as it is to do other case studies in the history and philosophy of science. Second, the present discussion of the relationship between psychology and the metaphysics of the mind has only scratched the surface. If "naturalism" has become a kind of shibboleth in recent times, this disguises the fact that there is enormous variety in the kinds of projects that are called "naturalistic." I have given some reasons here for skepticism about naturalistic platitudes, and suggested that the Neutral Project is all that one really needs for science and all one is likely to get in metaphysics. But really this is not the end of that topic but only the beginning. We need a much more serious examination of the roots of contemporary naturalism and the assumptions it encodes. We need a more thorough examination of how well its normative models of explanation and of intertheoretic relations are supported by actual examples of scientific theories even in the natural sciences. We need a more systematic categorization of kinds of "explanation," and an application of these to actual work done, both in psychology and the other sciences.

My hope is that this book will have separated what is truly useful about the computer paradigm from false hopes based upon incautious

uses of language. If I am right, both science and metaphysics will be better off if they see that beyond a certain point they must follow separate paths. I do not claim to know whether the computational path in psychology will end up leading to Oz or leading nowhere. But I think we will be better off if we turn around the map.