We can ask the same kinds of ‘constitutive questions’ about the natures of mental states as we have been asking about the nature of intentional action. Attempting to answer such questions is one of the traditional aims of the philosophy of mind.
Three traditional kinds of constitutive theories of mental states: (i) identity theories, (ii) behaviorist theories, (iii) functionalist theories.
Intuitive arguments against theories of types (i) and (ii).
The broadest characterization of functionalist theories of a mental state m is: the class of theories that says that what it is for an agent to be in m is for that agent to be in some internal state x with a property F correlated by the theory with m.
The sense in which this kind of theory differs from identity theories; its compatibility with intuitions about multiple realizability. The sense in which this kind of theory differs from behaviorism; the pull behind the idea that, e.g., beliefs and desires are inner states.
Let psychologism about action be the view that the story about what it is to perform an intentional action will involve appeal to some of the agent’s psychological states (most likely, some combination of her beliefs, desires, and intentions). On this broad characterization, almost every theorist we have read (save Anscombe) holds a psychologistic view of action.
One good question for proponents of psychologism is: if you take intentional action to be partly explained by, e.g., facts about belief, then what is it for an agent to have a belief? Two options: (a) take belief as primitive, (b) give a substantive theory of belief.
If the proponent of psychologism takes option (b), then the natural way to go will be via a functionalist theory of belief. Then our question is: is functionalism about belief compatible with psychologism about action?
On the surface, functionalism seems tailor-made to go along with psychologism about action. Psychologistic views require a theory of mental states on which they can intelligibly be the causes of certain actions, and a functionalist theory which identifies beliefs with inner states which have certain second-order properties seems to meet this requirement.
But a possible problem arises from the fact that the second-order properties of internal states correlated with beliefs by the functionalist theory will typically include relations to‘behavioral output’ of the agent. One question is then whether this output is restricted to intentional action, or not. If it is, then this is a kind of circularity.
To make matters more concrete, we can build up to a kind of sample functionalist theory of belief, by beginning with the plausible idea that the contents of inner states are determined, in some way or other, by their causes. Two problems with the simple causal theory: (i) the possibility of false beliefs, (ii) the need for behavioral constraints.
Two ways of revising the simple causal theory to respond to these problems: (i) the restriction to ideal conditions, (ii) the addition of a clause requiring certain dispositions to action.
Roughly: for an agent to believe p is for that agent to be in some internal state that, were she in optimal conditions, she would believe p, and for her to be disposed to act in ways which would satisfy her desires in a world in which p (and her other beliefs) were true.
The question of whether only dispositions to intentional actions should be allowed to satisfy the requirement on the agent’s dispositions.
Hornsby, in “Physicalist Thinking and Conceptions of Behaviour”, also focuses, beginning in §3, on the question of the notion of behavior put to use by the functionalist.
Hornsby claims that the notion of behavior used in functionalist theories differs from our ordinary conception of the behavior caused by beliefs and desires in two ways: (1) it does not involve things beyond the agent’s body (as does, e.g., a description of an agent as skating on the pond), and (2) it makes reference only to the movements of the bodies of agents, not to the actions of those agents. (100)
The motivations for (1) and (2) are introduced (pp. 100-101) via two traditional arguments against the behaviorist idea that having a certain belief is a matter of having certain dispositions to action.
Argument 1. Belief, e.g., cannot be analyzed in terms of the having of certain behavioral dispositions, since the having of such dispositions is a matter not only of having the relevant belief, but also of having certain other mental states. E.g., my belief that there are cookies in the drawer may usually be accompanied by the disposition to open the drawer and eat what is inside; but I will not have this disposition if I also believe that the cookies in the drawer are poisoned, or if I have a strong desire to lose weight and believe that devouring cookies will frustrate this goal, etc. This indicates that there is no set of behavioral dispositions stably associated with any particular belief.
Argument 2. Beliefs do not lead to action even with desires and other mental states; we also need to stipulate that the belief in question is true. Suppose that I desire some cookies and have the cookie-belief described above. These will not cause me to eat cookies (typically) unless the belief is true, and there are in fact cookies in the drawer.
Hornsby claims that functionalism is usually introduced as a response to the first argument; but it also owes us some response to the second argument. Hornsby suggests two ways in which a functionalist might respond to this second argument.
Functionalist response 1: restriction to behavior not dependent on the truth of beliefs (§4)
This first response embodies restrictions (1) and (2) on the notion of behavior employed in functionalist theories, discussed above. If we make the movement relevant to the above example something like moving my arms in such and such ways and directing my visual attention to such and such a point relative to the location of my body rather than opening the drawer and eating some cookies, then it may seem that we have responded to Argument 2: we have come up with a kind of movement which is not hostage to the truth of the beliefs of the agent.
To understand Hornsby’s criticism of this first response, we will have to get clear on an important distinction between two kinds of functionalist theory: commonsense functionalism and psychofunctionalism. Commonsense functionalism claims that the second-order properties (‘functional roles’) which make certain internal states mental states of particular kinds can be extracted from commonsense platitudes about the connections between perception, various mental states, and action; psychofunctionalism claims that the relevant second-order properties can only be revealed by theoretical psychology, and are not present, even implicitly, in our ordinary conception of the relevant mental states.
Hornsby’s main criticism of this first response is that it is incompatible with commonsense functionalism, since our commonsense platitudes about mental states do not connect them with actions under this kind of austere description.
So this forces the functionalist to psychofunctionalism. But Hornsby thinks that psychofunctionalism is implausible. (See pp. 106-7.) Whether or not Hornsby’s argument here is convincing (footnote 17 does not make it sound very convincing), I think that we can give an argument against psychofunctionalism, on at least one construal of what a functionalist theory does.
Functionalist response 2: restriction to world-conditioned behavior
Suppose we instead describe the action caused by the belief-desire pair above as doing something which, if there is a drawer there with cookies in it and ...will result in my eating cookies. This may seem to be better than the austere description in that it has a closer connection to our commonsense platitudes about beliefs, desires, and actions.
Hornsby argues against this on pp. 107-108, though I’m not clear on exactly what the argument is supposed to be.