How can we fit human beings into a scientific or naturalistic metaphysics? We cannot, many philosophers contend, and intentionality is the most basic reason why. A scientific or naturalistic account must be a causal account, they believe, but no causal account can possibly do justice to the intentionality of our thoughts. The intentionality of our thoughts, as of anything else, is a matter of their being directed upon or about something, and this directedness or aboutness cannot be any merely causal affair. For one thing, our thoughts are often about things that do not exist (Santa Claus, unicorns, the ether), and there can be no causal relation to a nonexistent. But even when we are thinking about something that does exist, there may be no causal relation between us and it (as lately seen of Ammon). In either case, the argument continues, that the thought is about something, and what it is about, cannot be determined by our behavior or brain states or anything of the sort. For these will often be causally unaffected by what the thought is about, so that they can be causally the same yet the thought can be about something else or nothing at all. Physical duplicates can be about or represent quite different things.

Furthermore, in order to understand our fellow human beings and to explain why they do certain things, we must often attribute various thoughts to them. Why did Jones go into the garden? Because he thought there was a unicorn there. Why did Columbus sail west? Because he believed he could reach the East that way. The only remotely adequate explanatory theory of human beings we have is one in which such intentional properties play an essential and irreducible role. Call the theory "folk psychology," if you like, but what could conceivably rival it?

Rivals there are, indeed superiors, or so it is claimed. According to behaviorists, insofar as it makes sense to ascribe a belief or other intentional state to Jones, Jones' having the belief is simply a matter of how Jones responds, or at least is disposed to respond, to various stimuli or input. The relevant kinds of response can include verbal responses. Suppose Jones is disposed to respond with "Yes" when asked whether there is a unicorn in the garden. This disposition is not only evidence that Jones believes there is a unicorn in the garden. In addition, the disposition to answer "Yes" and to behave accordingly is just what it is to have the belief. So too for the rest of Jones' intentional states. Those who insist that the intentional involves something beyond behavioral dispositions are being unscientific and obscurantist.

Identity theorists, who are not at all unscientific, do hold that the intentional involves something beyond the behavioral. Intentional and other states of the organism, insofar as it makes sense to talk about them, are not to be identified with dispositions to respond in certain ways to sensory input. Rather, they are identical with physical states of the organism, including the inner workings, structure and physical composition of the central nervous system.

Functionalists, who likewise are not unscientific, reject both behaviorism and physical identity theories. Intentional and other states of the organism are to be identified neither with the organism's dispositions to respond in certain ways to sensory input nor with its physical states. Instead, they are to be identified with its functional states. A functional state of an organism or machine is first of all an abstract property that can be realized in many ways, much as addition can be performed in many ways by different calculators with their different circuitry, materials, and programs. The same functional state could be realized in many different kinds of "stuff," including nonphysical stuff. Thus functionalists are not necessarily physicalists, though many of them are.

Beyond its abstractness, a functional state is defined by its causal relations not only to sensory input and motor output, as with the behaviorists, but also to other internal states. Thus the causal role of a functional state can be extremely complex. But it need not be very complex to resist definition in purely behaviorist terms. There is a mathematical result to the effect that even a state so simple as to involve only primitive feedback cannot be defined behavioristically. Since most if not all animal nervous systems involve feedback, pure behaviorism would seem not even to be in the running as an account of animal behavior, let alone of something so complex as the human animal.

How do physical identity theories and functionalism fare with beliefs? Let us concentrate on functionalism. For if functionalism fails to account for beliefs, so do identity theories. The reason is that identity theories are narrower than functionalism, since they identify the belief state with a physical state, rather than regard the physical state as just one way the belief state might be realized or embodied. According to functionalists, to have a particular belief is just to be in a certain functional state. To believe that there is a unicorn in the garden is for your central nervous system, primarily your brain, to be in or realize a particular state defined by certain causal relations to sensory input, other brain states, and motor output. To insist that the belief involves more than this is to be, if not unscientific or obscurantist, at least soft-headed. How could the existence or the individuation of a psychological state involve something that is not a matter of causal relations to sensory input, other brain states, and motor output?

One reply is that even if having a particular belief requires your brain to be in a certain functional state, it also requires more. The belief, after all, is directed upon or about something. Functionalism gives no account of this directedness or aboutness. Consider two beliefs, or the two patterns of neuronal activity they are realized in. Suppose they are the same as regards their causal relations to sensory input, other inner states, and motor output. Two such beliefs are thus exactly the same as regards their causal powers. What prevents them from being about quite different things?

Functionalists will tend to regard this sort of question as unintelligible. Two such beliefs cannot be about different things. What a belief is about, if the notion is to make any sense, is just a matter of its causal relations to input, other states, and output -- nothing more, nothing less. This includes causal relations to distant objects, via their causal effects on the organism and its input. Therefore, what a belief is about, what it is "directed upon," is determined without remainder by its causal powers, meaning its causal role. It makes no sense, or at any rate no empirical sense, to imagine two beliefs that are the same as regards their causal powers and yet are about different things.

This is a move in the direction of eliminitivism. Whatever cannot be fitted into the framework, whatever cannot be appropriately reduced, is to be eliminated as the sad remnant of primitive folk psychology. But even though the move is eliminativist, it is not necessarily wrong. Scientific advances often force us to give up cherished heirlooms. Such was the fate of the earth-centered universe, of the brain as an organ for cooling the blood, of the eyes' seeing by sending out beams, of phlogiston, of an earth and its species lately created much as we see them now, and so on. Why not also this mysterious directedness or aboutness that allegedly goes beyond the causal powers of a thing?

It will not do simply to reply by appealing to our intuitions. True, many people have strong intuitions to the effect that some thoughts and acts necessarily involve a directedness or aboutness that transcends any causal account. But so too did people have strong intuitions about an earth-centered universe and the origin of species. What we call "intuitions" are nothing more than entrenched ways of conceptualizing things, ways rooted in theories that share the fallibility of all theories.

This applies even to the Ammon example in Chapter 5. Such intuitive obviousness as the example may have could turn out to derive from a misplaced confidence in folk psychology. We may well need a better theory than folk psychology, a better theory of our psychological attributes, including the attribute of having a particular true belief. Such a theory might entail that our psychological attributes must after all be determined solely by matters to which we can be causally related, on pain of the attributes' being judged unreal if they are not so determined. Functionalism claims to be just such a theory. Champions of the reality of irreducible intentionality therefore need a less question-begging reply than appeals to intuition.

One such reply challenges not only functionalism and related theories, but physical identity theories, and challenges them at a very elementary level. Minimally, functionalism ought to provide an account of routine functional properties, whether of mechanisms or organisms. But consider again a heart. To be a heart is to have the proper function of pumping blood. The biological property of being a heart is evidently a functional property. According to functionalism, this functional property must be a state defined by certain causal relations to input, output and other states of the thing that is a heart. That is, what makes a thing a heart are its actual causal powers.

According to Millikan, the trouble here is double. First, there are things that have these causal powers but are not hearts. Water pumps are capable of pumping blood, but they are not hearts. Likewise, "Devices have now been designed that in fact pump blood in people, but these are only artificial hearts, not members of the biological category `heart.' " Even an instant duplicate of your heart, with therefore precisely the same causal powers, would not be a heart, as we saw in §5.3.3. Second, there are things that are hearts but do not have the powers. There are malformed hearts, diseased hearts, injured hearts. Biological functional properties are a matter not of actual causal powers but of a history. No connective generalization holds between the actual causal powers of a thing and its proper function.

Functionalism looks only to the causal powers of things, not to their histories. It must therefore fail as an account of the biological functions not only of hearts but of lungs, eyes, brains, if Millikan is right. Now suppose it turns out that the intentionality of a thing -- of a certain distribution of neuronal activity, say -- is a particular kind of biological function of the thing (where biological function, we recall from §3.3, need not be innate, evolved or wired-in, but can be learned, indeed learned very quickly). Then we should expect functionalism and related theories to have a tough time accounting for intentionality without eliminating it. For "intentionality does not reside in a mechanism but in its history." Indeed, from Millikan's point of view, functionalism and its kin are on the wrong track entirely.

But is intentionality a species of biological function? According to Millikan's theory it is. Anything that has a proper function is supposed to do something. Hearts are supposed to pump blood, hoverfly males to catch conspecific females, bee dances to map onto a direction, sentences to map onto configurations of things in the world. Each of these organisms, behaviors or devices has a purpose, something it is meant to do, "which something can be described, yet which something may or may not be." But this is the traditional earmark of the intentional. What is intentional is supposed to stand in relation to something else -- that which it intends or means or is meant to do -- even if that something does not exist or never comes about. Thus "in the broadest possible sense of `intentionality,' any device with a proper function might be said to display `intentionality.' "

But in a narrower sense of `intentionality', only things like beliefs or sentences have intentionality. Why suppose that their intentionality is a matter of their being members of certain biological or proper-function categories? One reason is that we acquired the ability to have beliefs and use sentences by way of a long evolutionary process. The ability to produce and understand sentences with "aboutness" has survival value, which is one of the marks of biological proper function. Another and more important reason is that beliefs and sentences can be defective, which is a normative matter. They can be false, they can fail to be about what they are supposed to be about, and they can fail in other ways. Anything that can be defective displays the characteristic mark that all things defined by proper-function categories display.... What is defective is, just, that which is not what it should be or cannot do what it should do, hence is something defined by its "shoulds" rather than by its "coulds" and "woulds." A third reason, according to Millikan, is that the resulting theory of intentionality has greater explanatory power than its rivals, greater coherence, and a greater capacity to "unravel paradox and produce understanding."

How does the theory work? A review of §3.3, on biosemantics, helps us see how a belief or sentence can be a device that exhibits intentionality derived from the proper function of speech-producing and speech-interpreting devices that are a reproduction of an ancestor in a family of devices, which family historically proliferated because a critical mass of the ancestors performed a certain function or functions. In particular, in the case of sentences the derived functions are performed by way of the sentences' mapping onto conditions or world-affairs in conformity to certain rules, and thereby having a certain content. The sentences, together with the beliefs they express, thereby exhibit the sort of aboutness or ofness or directedness associated with intentionality. What the sentence is about or represents is the configuration or world-affair onto which it is supposed to map in conformity to a certain rule. Biosemantics therefore includes a theory of representation.

Another consequence of biosemantics is that sentences and other devices can exhibit intentionality even if they are neither produced nor interpreted by a conscious being. Indeed, contrary to Brentano and Husserl, "the problem of understanding intentionality can and should be divorced from the problem of understanding consciousness." This means that a naturalist account of intentionality need not also provide an account of consciousness -- of pains, of other sensations, of what it is like to fall in love, and so on -- though of course eventually some account must be given of them as well.

It also means that even though we are often conscious or aware of the intentionality of our thoughts, the awareness affords no guarantee that the thoughts are about something, no guarantee that they are not empty or senseless. For their being about something is in part an external matter, which depends on how other thoughts and the devices that produce them performed in a history over which our present awareness has no say. Indeed, for essentially the same reason our very awareness of something is itself partly an external relation, "the inside of the awareness -- the feeling part -- giving no absolute guarantee that it is the inside of a genuine awareness of relation." meaning-rationalism, as we learned to call it in §3.3, is therefore false. "Consciousness ... does not contain within it or directly before it an object of consciousness.... There is nothing diaphanous about consciousness."

Intentionality, then, being a biological proper-function category, is a real phenomenon, every bit as real as other biological categories. The aboutness of a belief is as real as the proper function of the heart, as real as the proper function of a bee-dance to map onto a certain direction to nectar. And, like the proper function of the heart or of the dance, neither aboutness in particular nor intentionality in general can be reduced to or identified with the causal powers or physical structure of the devices that exhibit it. One must look to their histories. Moreover, biosemantics licenses the Ammon example of the previous section. For biosemantics explains how the truth-conditions of a sentence or belief can include matters not connected to it by any causal chain. According to biosemantics, functionalism and related theories have some explaining to do.

What about philosophies that deny the possibility of any naturalistic account of intentionality? Their argument was chiefly that any such account must be a causal account. But Millikan's is a naturalistic account that not only is not a causal account, in the relevant sense, but explains why no such account will do. On her view, the mistake of those who deny the possibility of a naturalistic account is to suppose that any such account must be causal. True, they have been abetted in this supposition by those who insist that intentional concepts have no place in science because they have no place in causal explanation. But from Millikan's point of view, this is just the flip-side of the same mistake.

What about nonreductive physicalism? What account can it give of intentionality? How are the intentional properties of an act or thought supposed to emerge from or be determined by the way things are at the level of the underlying physical entities? These of course are the pointed questions raised at the end of the previous chapter.

Millikan's theory, though naturalistic, is neutral between physicalist and nonphysicalist metaphysics. Nevertheless, nonreductive physicalists, armed with her theory, might reply to the pointed questions as follows. The proper function of a device, as seen, is determined by whether the device is a reproduction of an ancestor in a family of devices, which family historically proliferated because a critical mass of ancestors performed that function (or were correlated with it). But whether one thing is in the intended technical sense a reproduction of another is determined ultimately by whether certain physical causal relations hold between them; so too for whether one thing is an ancestor of another via reproduction. Likewise, whether a critical mass of devices in the reproductive family performed the function is determined ultimately by the physical conditions both of the devices and of their surroundings. Unfavorable physical surroundings "select out" devices that fail to perform. Ill-suited devices fail to be reproduced as often as other devices, which tend to crowd them out.

It follows that proper function is determined ultimately by the physical properties both of the devices, including their ancestor devices, and of their surroundings in a history, even though no connective generalization holds between a device's own physical properties or relations and its proper function. So too for the special case of proper function we call intentionality. That an item is about or means or "intends" something, and what it is about or means or "intends," are determined by certain physical properties and surroundings. It is determined not just by the physical properties and surroundings of both the device that produced them and the device's ancestors, but by those of their physical properties and surroundings in virtue of which a critical mass of the ancestors performed a certain function by way of producing items that mapped onto certain configurations. The specific physical phenomena that determine what a belief is about -- those that form the least set that suffices to determine this -- occur over time, in various places, certainly not all in the head. The nonreductive physicalist is free to deny that beliefs that are physical duplicates must have the same content. In this way physicalists could argue that they can "make room for the requisite concept of content." So too for nonphysicalists: if the things and properties they take as the unifiers determine these matters of proper function, they also thereby determine intentionality.

This account relies on the fact that the determination relation is transitive. That is, if A determines B, and B determines C, then A determines C. If a metaphysician's unifiers determine matters of proper function, and matters of proper function determine intentionality, then the unifiers determine intentionality. Because determination entails no connective or bridge generalizations or laws, there need be no such generalizations or laws from properties at a lower level to those at a higher (though of course other kinds of bridge principles connect them, as in Millikan's theory). There is no need to say --and strong reason to deny -- that whatever has certain unifying properties, relations or causal powers also has a certain proper function.

In particular, there need be no physical explanation of proper function or of intentionality that relies on such generalizations. Nonetheless, the nonreductive physicalist claims that an important kind of explanation or understanding is achieved when we see that and how the physical properties of devices and their surroundings in a history determine matters of proper function, and thereby determine intentionality. In this way and in this sense, there is an explanation of how and why the intentional is determined by the physical, even though there is no explanation based on connective generalizations from physical properties to intentional properties. In this sense, it is not "a brute and unexplainable ... primitive fact about the world" that the one determines the other. Or so the nonreductive physicalist will try to argue, as will nonphysicalist metaphysicians who wish to integrate biosemantics with what they take to be the unifying entities and properties.