Fodor takes propositional attitudes to be relations to tokens of an internal language (``mentalese'') that have content. If Jerry believes that P, Jerry has a token of mentalese in his belief structure that has the content that P. These tokens have causal properties as well as content; they are related in law-like ways in virtue of their causal properties. But the laws must (in the paradigm cases that are grist for the cognitivist's mill) make sense in terms of the contents of the tokens. It is the contents that are related in commonsense psychological principles, and it is the meshing of content and causal properties that makes it conceivable that cognitive psychology might work.
What exactly is the relation between the contents and causal properties of tokens? The first quotation above, from Propositional Attitudes, suggests a simple answer: tokens interact as they do in virtue of their content. But Fodor begins Methodological Solipsism with a quotation from Hume that states a problem for this view.
...to form the idea of an object and to form an idea simply is the same thing; the reference of the idea to an object being an extraneous denomination, of which in itself it bears no mark or character.
The fact that my idea of red is an idea of one color rather than another is an external denomination--a relation between the idea and a color--not something that can influence the way it interacts with other ideas. The view that emerges in Methodological Solipsism is not that tokens interact in virtue of their contents, but that both the causal facts and the content facts about a token are settled by its formal properties. The formal properties are, roughly, those that a processor scanning the tokens can detect and by which its actions can be systematically determined. If both the contents and the causal powers of a token depend on its form, the two kinds of properties might mesh in the ways necessary to have the content-based principles of cognitive psychology (or commonsense psychology) backed by causal laws relating formally individuated states. One imagines a species-wide causal role for a given type of token and a species-wide interpretive function that assigns contents to types. The two mesh so that, for example, within the human species, if a belief and a desire cause a volition, then the contents of the belief, desire, and volition have some sensible, rational connection with one another.
One imagines the interpretive function fixing the reference of a
certain formal type as the property of being red, once and for
all--the same for each member of a cognizing species. But it seems
that the reference of many of our ideas is not only external, but
circumstantial. The interpretive function and the formal properties
of a token do not fully determine the reference; particular
circumstances that vary among formally identical tokens, also must be
taken into account. Suppose, for example, that Jerry is looking at a cup.
The cup that he is looking at will be the referent of the mentalese phrases
that are analogues to ``the cup in front of me'' or ``that cup.''
Tokens of the same mentalese phrases, in the head of someone else,
looking at a different cup, will have different
referents.
How can causal and content properties of tokens mesh, if the content properties of a token depend both on its form and on such particular external circumstances, while the causal properties depend on form alone? If contents are sensitive to external circumstances, and so classify persons who are internally similar as different, and those who are internally different as similar, how can content-based principles of rationality mesh with causal laws?
If such circumstantially determined external references of phrases become part of the truth-conditions of the sentences in which they occur, then those truth conditions will also be infected with circumstantiality. It seems a causally coherent psychology should not individuate states in terms of such external denominations. But the contents that are ascribed to mental states by our practices of propositional attitude reporting are often based on just such circumstantially determined external denominations. The contents that a person believes and the people who are counted as believing the same thing depend in part on external facts about reference and hence on nonformal properties of their tokens. As a result, people with formally identical tokens may believe different things; people with formally differing tokens may believe the same thing.
Fodor would have liked the following response to this worry to work:
Transparent reports of attitudes do classify mental states in a circumstantial, nonformal way. But because this is so, we do not take transparent reports to tell us how the agent is thinking, and for this reason explanations using transparent reports are impotent. But opaque reports are explanatorily valuable, because they do tell us how agents think; they can do this because they don't classify mental states circumstantially, but in terms of contents that mesh with formal properties.
But in the remarks quoted above from Methodological
Solipsism, he indicates that this response is not quite
right. Even opaque reports rely on circumstantial
classification. In effect, he postulates a level of
fully opaque (nontransparent) content that works the
way he had hoped the opaque classification would. But he
does not provide very many details.
In this paper we sketch a theory, within a Fodorian conception of cognition, of how the formal properties and external circumstances of tokens relate to the contents of tokens, and of how commonsense explanations might work given this framework. We conclude that
Like Fodor, we believe cognitive psychology is possible, and that there are strong empirical arguments in favor, and only weak and avoidable arguments against, the thesis that a component of that psychology will be individualistic. We think of ourselves as providing a way of looking at content and content-based explanations that should be helpful to those, like Fodor and ourselves, who believe cognitive psychology can and should include such an individualistic component. There are nevertheless some points of disagreement between us and Fodor on the topics covered in this essay. We think transparent explanations can work well; Fodor does not. In our reconstruction, even fully opaque attitude reports will be somewhat circumstantial. We provide a completely noncircumstantial level of content, but claim that to understand the rationality of laws one needs to bring in the (admittedly rather modest) circumstance that the relevant tokens of mentalese belong to the same agent.
Given the amicable goal of this essay, perhaps we should indicate some
broad areas of disagreement before plunging in. Like Fodor, we
conceive of individualistic cognitive psychology as requiring
structured internal states that have both content and causal role.
And like him, we believe that the postulation of such states is
empirically reasonable and that the success and structure of our
ordinary psychological concepts provides important evidence for it.
We are inclined, however, to think that Fodor's talk of the language
of thought and mentalese may encourage the assumption that the
internal structures are more language-like than there is any reason to
believe. And we worry about whether the notion of formality
coalesces conditions from a number of different areas that have less
in common than Fodor thinks that they do. But
for the purposes of this paper we set aside these worries.
Other areas of disagreement concern questions of how the causal and
content properties of internal states relate to one another. An
important issue is the source of what we shall call the interpretive
function, which assigns contents to formally individuated mental
states. We sympathize with the aim and intent of approaches that see
content properties as ultimately reducible (in some weak sense) to
causal properties, with a role for information ``Wisconsin style.''
But we do not think that any such reduction to causal/informational
concepts can succeed. Our own view is a sort of naturalistic
functionalism that stresses functions and purposes--hence
the results of actions measured in terms of degree of success in
bringing about some specified condition--in the classification of
mental states. We are inclined to think, with Dennett, that the
distinction between merely attributed intentionality and real
intentionality can be understood within this framework, rather than
constituting a refutation of it. From this perspective, we try to
practice an approach that is an amalgam of the design-oriented
approach of work in artificial intelligence with H.P Grice's
conceptual creature construction. This approach seeks to understand intentionality through
understanding the reasons that intentionalistic theories are useful in
dealing with (or designing) various systems, beginning with very
subhuman systems. We think that cognitive psychology depends on
naturalistic psychology and that Fodor's arguments against the
possibility of the latter are not very convincing.
From both our perspective and Fodor's, it is necessary to have an account of how the contents and causal powers of the structured states of individuals intelligent beings can mesh in the way presupposed by common sense and by cognitive psychology. We try to provide such an account in this paper.
We shall look closely at an example that brings out the problems that concern us. Suppose Jerry wants to drink some decaffeinated coffee. He sees before him a brown cup, c; just a few seconds before, he had seen a brown cup being filled with what he believed was decaffeinated coffee. He moves his arm and hand in a complicated manner, grasping the cup, lifting it to his lips at an angle as he tilts his head a bit, and opens his mouth. Call this type of movement M for later reference. In Jerry's circumstances, this movement constitutes picking up the cup and bringing it to his lips. Gravity and his digestive system take over, and he gets what he wants. The explanation of Jerry's doings might go like this:
Why did Jerry decide to pick up the cup and bring it to his lips?He wanted to drink some decaffeinated coffee, and he believed that the cup in front of him was filled with decaffeinated coffee.
This example can be used to illustrate the problems engendered by the fact that common sense psychology classifies mental states in terms of external, circumstantial denominations. We explain Jerry's action by citing a belief and a desire. The belief and the desire make sense of the action. The action will promote the satisfaction of the desire if the belief is true. But the way the belief, desire, and action are individuated appears to depend on factors external to the formal properties of the mental tokens involved. One way to see this is by focusing on what we count as having the same belief. Construed transparently, the belief attributed to Jerry is individuated in terms of a certain cup c, rather than the way Jerry thinks of it. Someone would agree with Jerry-would have the same belief-who believed that c contained decaffeinated coffee, no matter how the other person thought of c. Yet only the ways of thinking, not the external reference, could conceivably be correlated with the formal properties of belief tokens.
Construed opaquely, things are a bit better. But the belief still
seems to be individuated in terms of Jerry. As Fodor points out, opaque attributions still allow some
slack as to how the agent is thinking. Someone else, who is looking
both at Jerry and at c but thinking ``The cup in front of him
contains decaffeinated coffee,'' would be said to believe the same
thing that Jerry does. But then it is at least possible that Jerry
could believe just what he does, opaquely construed, while being in a
different state. Perhaps, sitting in a Denny's,
gazing in a mirror while waiting impatiently for his cup of decaf,
Jerry sees a waitress fill a customer's cup from the pot with the
characteristic decaf indicator--the pot has an orange neck. He
thinks, ``Now the cup in front of him contains decaffeinated
coffee...I wonder when the cup in front of me will do so.'' In such a
circumstance, it would be misleading to report that Jerry believes the
cup in front of him contains decaffeinated coffee; but would it be
incorrect, would what the reporter said be false? It seems not, for as
we noted, someone else, who thinks ``The cup in front of that man
contains decaffeinated coffee'', with reference to Jerry, would be
counted as believing just what Jerry believes when Jerry thinks ``The
cup in front of me contains decaffeinated coffee.'' So it seems that opaque
attitude reports also do not focus exclusively on the sorts of
intrinsic (formal) properties needed by a causal theory of
mind.