Intentionality - naturalization of
Daniel Kelly, Kelby Mason, Dennis Whitcomb
Department of Philosophy
Rutgers, the State University of New Jersey, USA

Some mental states are about things. For instance, the belief that the cat is white is about the cat. States that are about things are intentional, that is, they have content. The precise nature of intentional states is a matter of dispute.What makes some states, but not others, intentional? Of those states that are intentional, what makes them about what they are about as opposed to something else, i.e. what gives them their specific content?

To naturalize intentionality is to give these questions scientifically legitimate answers. Philosophers consider this project important because they want to know where intentional states fit within our rich scientific picture of the world, which is rooted in other areas of scientific enquiry such as biology and physics. Many think that, if intentional states do not fit in this picture, then in some sense they do not really exist.

To understand the debates about intentionality, three distinctions are important. The first is between narrow and wide theories of content. Narrow theories explain the content of intentional states via conditions that are internal to the mind; wide theories do not. For example, some theories say that part of what makes the belief that the cat is white about the cat is the fact that it is caused by the cat. Since the causal relation between the cat and the belief extends beyond the boundaries of the mind, these theories are wide. Narrow theories deny the relevance of any such mind-external conditions.

Our second distinction is between two persistent problems. Attempts to naturalize intentionality are often unable to account for mismatches between the way the world is and the way intentional states represent it to be. This is the problem of error. They are also often unable to capture the fine differences between related but distinct contents. This is the problem of fine-grainedness.

Our third distinction is between mental states and sentences of natural language. Both of these can be intentional, or about things. We are concerned mainly with the former, but it is worth noting that the project of naturalizing intentionality can cover both mental states and natural language. Terminologically, the issues are usually described in terms of meaning and semantics in discussions about language, content and representation in discussions about mental states.

Description of the theory
One earlier project that overlapped with the naturalization of intentionality was the causal theory of reference. After Russell and Frege, the dominant theory of names in the early twentieth century was that they were abbreviations of much longer descriptions, e.g. “Napoleon” meant “the French emperor who did such-and-such…”. Against this, Kripke and Putnam [4, 9] claimed that the meaning of proper names and category names is set by causal and historical connections between the names (“Napoleon”, “water”) and what they refer to (Napoleon, water). This emphasis on causal relations gave encouragement to other philosophers that a general causal/scientific account could be given for language, mental states and any other intentional states. More fully developed and explicitly naturalistic theories of intentionality followed; they can be split into three major types.

The causal theory of reference had the most direct influence on informational/causal theories of intentionality, which elaborate the causal relations necessary for content in general. Informational theories, closely associated with Dretske [1], analyse the content of a state in terms of the information it carries: a particular mental state is about hammers, for instance, because it carries information about hammers. Information is then defined non-intentionally, the basic idea being that x carries information about y just in case there is some reliable generalization that, whenever x obtains, so does y (other things being equal). Smoke carries information about fire; tree-rings carry information about the tree’s age; hammer-thoughts carry information about hammers.

This theory has trouble with the problem of error. Suppose that most people reliably thought “there’s a frog” whenever they saw a toad, as well as a frog. The thought would carry the information that a frog or toad is present, but that would not be its content; thinking frog thoughts when only toads were about would be a mistake.

Fodor’s [3] causal theory aims to solve the problem of error by restricting the causal relations that are relevant to content. Even if in general people have frog-thoughts in the presence of toads, that is irrelevant to the content of the thoughts, because this generalization asymmetrically depends on the generalization that people have frog-thoughts when they see frogs. It is not clear, however, whether the notion of asymmetric dependence can itself be specified naturalistically.

Teleosemantic theories like Millikan’s [8] place informational accounts in an evolutionary context. According to such theories, the content of a state is given by its biological function: frog-thoughts have their content because their biological function is to carry information about frogs. Biological function is then defined in terms of natural selection: just as the heart was selected to pump blood, frog-thoughts were selected to carry information about frogs.

Teleosemantic theories readily handle the problem of error. Even if everyone now made mistakes with their frog-thoughts, and even if our ancestors did too, the thoughts would still be about frogs and not toads provided they were selected to carry information about frogs specifically. Where such theories might have trouble is in individuating finely-grained contents. For instance, suppose that in our evolutionary history we only came into contact with one species of frog. Then what makes our frog-thoughts about frogs in general, rather than about that particular species? It is not clear that teleosemantic theories can answer such concerns.

Another family of attempts to naturalize intentionality falls under the heading of conceptual role semantics (CRS). According to this sort of theory, advanced by e.g. Field [2], the content of a given mental state is determined by that state’s conceptual role. Both wide and narrow versions of CRS have been advocated. Narrow versions take the conceptual role of a given mental state to be determined by its patterns of interaction with other mental states. Wide versions take the conceptual role of a mental state to be determined by its patterns of interaction with other mental states as well as features of the external world. For instance, part of the conceptual role of the belief that the cat is white, and thus part of what makes that belief about the cat, may be the fact that that belief tends to come to mind when one is looking at the cat. Since the cat is part of the external world, versions of CRS that subscribe to such views are wide rather than narrow.

There are many problems with CRS. The most important of these is that no one has yet been able to adequately explain how to pick out a specific pattern of interaction that is the conceptual role for the belief that the cat is white, as opposed to a similar but distinct pattern of interaction that is the conceptual role for the belief that the cat is hungry, etc. Without such an explanation CRS has almost no resources for specifying content, and thus faces a particularly acute form of the problem of fine-grainedness.

In addition to the objections raised to each of these particular approaches to the naturalization problem, some philosophers have viewed the entire project as problematic. Such philosophers have been influenced by Quine’s sophisticated arguments concerned with language [10], which supported two related doctrines: the indeterminacy of translation and the inscrutability of reference. Each of these questioned whether there is any fact of the matter about certain alleged features of meaning, such as what the correct translation of an utterance is, or what a particular term refers to. The totality of physical facts fail to fix the relevant semantic facts, Quine argued, and concluded from this that meaning facts have no place in a naturalistic worldview, and so semantic and intentional terms deserve no role in a serious science.

When behaviourism, which self-consciously refused to deal in intentional and other mentalistic terminology, was replaced by cognitivism as the dominant paradigm in psychology, the shift raised problems for Quine’s conclusions. Psychology began producing robust theories that were couched in intentional terms, and which were therefore committed, at least on the face of it, to the existence of intentional states. For many philosophers, this only made the need to naturalize intentionality more urgent, and their subsequent efforts to produce an adequate account have lead to the various theories described above.

Still, the project itself has continued to have its critics, who take the proliferation of approaches, together with the outstanding lack of success (or even consensus), as indirect evidence for their own sceptical position. More specifically, Stich and Laurence [11], suggest that a fundamental confusion is generated by ambiguity in the notion of naturalization itself. They claim that although fear of intentional irrealism – the doctrine that no intentional properties exist whatsoever – has motivated the project of naturalization, the connection between irrealism and naturalization is much less clear than most naturalizers tacitly assume. Once they have surveyed all of the currently available accounts of ‘naturalization’, Stich and Laurence argue that, however it is construed, failure of the naturalization project does not lead to such an unacceptable irrealism.

This issue of what naturalization amounts to emerges less explicitly in many other discussions as well, where ‘naturalize’ is used to mean something like ‘legitimize’. For some, the pressing question is whether a science should be committed to contentful states and intentional properties, as contemporary cognitive science is widely held to be. The answer lies in whether or not those intentional properties can be naturalized, and thus legitimized, according to some further, more stringent criterion. Others hold that if a successful natural science is committed to the existence of intentional properties, then this fact suffices to naturalize those properties. The search for some further method of legitimization is thus taken to be misconceived.

Other concerns about the viability of naturalizing intentional properties stem from an alleged relation between intentional and normative properties. Most notably, Kripke [5], under the influence of Wittgenstein, starts by considering the activity of rule following, and generates arguments that intentional properties are essentially normative. He suggests that since normative properties cannot be naturalized, neither can the intentional properties that essentially contain them.

In sum, the only thing beyond controversy is that, despite much work and debate, none of these issues have been definitely settled yet. Loewer [6] sees a pattern in the failures of the particular approaches, however. He notes that the accounts that successfully deal with the problems of error and fine-grainedness smuggle in and rely, tacitly or otherwise, on non-natural notions, while the accounts that employ only natural notions are unable to give satisfactory answers to the problems of error and fine-grainedness. It is possible that giving a fully adequate account that both solves the problems of error and fine-grainedness and uses only naturalistic notions is simply too complicated, given our formidable but limited cognitive resources. Perhaps, that is, we are cognitively closed with respect to the issue. Though some have endorsed this position in other areas of philosophy [7], urging that time and resources would be better spent addressing more tractable problems, it strikes others as defeatist, and methodologically unhelpful. For them, the project of naturalizing intentionality continues.


1. Dretske F (1981) Knowledge and the Flow of Information. Clarendon Press, Oxford
2. Field, H (2001) Truth and the Absence of Fact. Clarendon Press, Oxford
3. Fodor J A (1990) A Theory of Content and Other Essays. MIT Press, Cambridge, Mass.
4. Kripke S (1972) Naming and Necessity. Harvard University Press, Cambridge, Mass.
5. Kripke S (1982) Wittgenstein on Rules and Private Language. Harvard University Press, Cambridge, Mass.
6. Loewer B (1997) A Guide to Naturalizing Semantics. In: Hale B, Wright C (eds) A companion to the philosophy of language. Blackwell, Oxford, p 108-126
7. McGinn C (1993) Problems in Philosophy: The Limits of Inquiry. Blackwell, Oxford
8. Millikan R (1984) Language, Thought, and Other Biological Categories. MIT Press, Cambridge, Mass.
9. Putnam H (1975) The Meaning of 'Meaning'. In: Gunderson K (ed) Language, Mind and Knowledge. University of Minnesota Press, Minneapolis, MN, p 131-93
10. Quine W V O (1960) Word and Object. MIT Press, Cambridge, Mass.
11. Stich S P & Laurence S (1994) Intentionality and Naturalism. In: French P A, Uehling, T E, Jr. (eds) Midwest Studies in Philosophy, v. 19, Naturalism, University of Notre Dame Press, p 159-182

Comments, suggestions, clever derision and unqualified flattery all welcome at