Beliefs, desires, perceptions, and other mental states and events are said to possess content. We attribute such states and events with sentences such as
(1) Arabella believes that the cat is crying.
(1) contains a propositional attitude verb ("believes") and a sentence complement ("the cat is crying"). The verb specifies a type of mental state (belief), and the complement sentence indicates the content of the state. On most accounts this content is the proposition expressed by that sentence. Propositions have been variously conceived as abstract entities composed of modes of presentation, sets of possible worlds, sets of synonymous sentences, and structured entities containing individuals and properties. All these accounts agree that propositions determine truth conditions. Some mental states and events (e.g., desiring to visit Paris) seem to have contents that are not propositions. However, for most of the current discussion, contents will be identified with propositions and contentful mental states with mental states that possess truth conditions.
Both natural-language sentences and mental states possess contents. The relation between content properties of the two items is controversial. Some philosophers think that natural-language expressions derive their contents from mental states, while others hold that, at least in some cases, the dependency goes the other way. In any case, it is plausible that there are mental states whose contents cannot be expressed or cannot be completely expressed by sentences of English (or other natural languages). For example, the full propositional content of a person watching the sun set is only partially captured by an attribution such as "A sees that the sun is setting." Also, some of the states posited by cognitive psychology and the mental states of animals plausibly have contents that fail to correspond to any contents expressible in English.
Content apparently endows mental states with a number of remarkable features. First, they or their constituents refer to extramental reality. When a person perceives that the sun is setting her perception refers to and thus puts her into contact with the sun. Second, they seem to be essentially normative. For example, a person ought to believe that the sun is setting only if the sun is setting, and if she believes that the sun is setting she ought not believe that the sun is not setting. Third, they apparently cause other mental states and behavior in virtue of their contents. For example, Arabella's belief that the cat is crying causes her to feed it. Fourth, a person can apparently know the contents of her own thought a priori and with an authority available only to her.
It is difficult to see how anything can exemplify all these features. The problem is especially difficult for philosophers who endorse naturalism, the view that all genuine properties are constituted by or realized by properties that are mentioned in true theories of the natural sciences. Content properties are prima facie so different from physical and biological properties as to raise the question of whether they are natural properties.
Hilary Putnam (1975) and Tyler Burge (1979) described thought experiments that have been taken to have important consequences for the nature of mental contents. Putnam imagined two thinkers, Oscar and twin-Oscar, who are identical with respect to their intrinsic neurophysiological properties but whose environments differ. Specifically. Oscar shares our environment, but twin-Oscar lives on twin-earth where the abundant substance that quenches thirst, fills the twin-earth oceans, and so forth is not H2O but XYZ. H2O's and XYZ's superficial properties are identical, and the two substances are indistinguishable without chemical analysis. Putnam claims that, while Oscar's sentence "Water is wet" and the thought he expresses with it are about H2O, the same sentence in twin-Oscar's language and the thought he expresses with it are about XYZ. The two thoughts differ in their propositional contents, since one is true if and only if (iff) H2O is wet and the other iff XYZ is wet. Putnam supports these conclusions with the intuition that, were Oscar and twin-Oscar to learn that the substances each refers to with the word water differ in their chemical natures, they would agree that their utterances of "Water is wet" possessed different truth conditions.
Putnam's thought experiment has been taken to establish the truth of content externalism, the thesis that the individuation conditions of mental content are partially external to the thinker. The point generalizes to other mental states whose contents are the same as the contents of sentences containing natural-kind terms such as water. Burge described further thought experiments that he thinks show that practically all thoughts expressible in natural language are externally individuated, and others have argued that all mental states that express extramental truth conditions are externally individuated (LePore and Loewer 1986).
Some philosophers (Fodor 1987, Loar 1987) react to content externalism by granting that mental states possess externally individuated contents but adding that they also possess narrow contents that are not externally individuated. Oscar's and twin-Oscar's beliefs possess the same narrow content. Philosophers sympathetic to narrow contents raise a number of considerations. One is the Cartesian intuition that thinkers in the same intrinsic state have the same mental lives. It seems essential to our conception of a mental life that it possess content, so there must be some kind of content that such thinkers share. The other consideration is that the causal powers of Oscar's and twin-Oscar's mental states seem to be, in an important way, the same. Jerry Fodor (1987) claims that if these causal powers involve the states' contents, then that content must be narrow.
Whether or not these considerations are persuasive, it has proved difficult to formulate a satisfactory notion of narrow content. If natural-language sentences express only externally individuated contents, then we do not attribute narrow content with sentences such as (1). While identity of intrinsic neurophysiological states is sufficient for identity of narrow content, it is not a plausible necessary condition. To adopt it as such would make it enormously unlikely that two people have ever shared the same narrow content state and impossible for a state to maintain its content in the course of reasoning. While some proposals for necessary and sufficient conditions for identity of narrow content have been forthcoming (Fodor 1987), there is little agreement concerning whether they are correct or, for that matter, whether a notion of narrow content is even needed.
Externalism seems to be in tension with our having a priori knowledge of the contents of our thoughts (Boghossian 1989). If the content of the thought (e.g., that water is wet) is individuated in part by external factors, then it seems that a person could know that she is thinking this thought only if she knows that those external factors obtain, and thus it is implausible that such knowledge is a priori. One response to this is to grant that we have a priori knowledge only of narrow contents. But a number of philosophers (Burge 1988, Warfield 1994) have responded that the tension is only apparent. Burge claims that judgments of the form "I am now thinking that water is wet" are self-verifying, since one cannot make the judgment without thinking the thought that the judgment is about. If this is correct, then externalism and a priori knowledge of content are not always incompatible. But such self-verifying thoughts seem to be a very special case of the thoughts whose contents we seem able to know a priori. It is likely that little progress concerning the epistemology of content can be made without an account of the nature of contentful mental states.
The dominant view in the philosophy of mind is that contentful mental states are functionally individuated internal states. Some philosophers (Dretske 1981, Fodor 1987) posit that these states are partially constituted by mental representations that are the bearers of propositional content. Mental representations are conceived of as picturelike (mental images), maps, or linguistic expressions. One view (Fodor 1979) is that mental representations are expressions in a language of thought, Mentalese. On this account thinking that the cat is crying involves tokening a Mentalese sentence with the content that the cat is crying. The thought inherits its content from the semantic properties of its constituent sentence, which in turn obtains its content from the semantic properties of its constituent expressions. Fodor identifies concepts with Mentalese expressions. So, for example, possessing the concept cat is being able to token a Mentalese expression that refers to cats. Some philosophers (Peacocke 1986) have argued that the contents of perceptual states are nonconceptual. If so, then the contents of these states are not borne by Mentalese expressions.
The nature of the bearers of mental content is best seen as an empirical issue. Fodor (1987) cites the fact that thought is productive and systematic as support for the language-of-thought hypothesis. Productivity is the capacity to produce complicated thoughts by combining simpler thoughts, and systematicity involves being able to think thoughts that are systematically related to each other, as are the thoughts that Bill loves Newt and that Newt loves Bill. Fodor argues that the language-of-thought hypothesis provides the best explanation of these phenomena, since languages are productive and systematic. Further, cognitive scientists have constructed theories of cognitive processes, language comprehension (Pinker 1994), perception (Marr 1982), and so forth that involve subpersonal contentful mental representations. For example, on one such theory understanding a natural-language sentence involves tokening a representation of its grammatical structure. These representations are not accessible to consciousness and have contents that are not usually available as the contents of a person's beliefs.
There have been various attempts to specify conditions in virtue of which mental states or mental representations possess their contents. Some of these are attempts to naturalize content properties. Following are brief descriptions of the main proposals.
According to interpretationist theories (ITs; Davidson 1984, Lewis 1974) our practices of interpreting one another partially constitute the contents of mental states. On Donald Davidson's approach interpretation is constrained by principles of rationality and charity. These principles say, roughly, that a person's mental states are generally rational and her beliefs are generally true. According to Davidson the evidential base for an assignment of contentful mental states to a person consists of her dispositions to hold true sentences under various conditions. She believes that p (desires that p, etc.) iff assignments of content to her sentences and to her mental states that systematize these holding true dispositions and that conform to the principles of charity and rationality assign to her the belief that p (desire that p, etc.).
On ITs, content properties are holistic, since whether or not a person exemplifies a particular contentful mental state depends on what other mental states she exemplifies and on their relations to each other and to environmental conditions. Davidson's IT is externalist, since a state's content is partially determined by relations to environmental conditions. But his account does not provide a naturalistic account of content, since it explains content in terms that presuppose content: holding true, rationality, truth. The primary difficulty with extant ITs is their vagueness. No one has formulated the principles of rationality and charity with sufficient clarity to permit an evaluation of proposed ITs.
According to conceptual role semantics (CRS), the content of a mental representation (or mental state) is determined by the inferential relations among representations and causal relations between representations and extramental events (Block 1986, Loar 1981, Sellars 1963). In this respect CRS is similar to IT. The difference is that, whereas ITs employ holistic principles of interpretations (rationality and charity), CRS attempts to spell out inferential patterns associated with particular concepts. CRS seems plausible for the logical connectives. For example, if a thinker is disposed to infer the representation A #B from A and B and vice versa, then # is the thinker's conjunction concept. Some philosophers (Peacocke 1992) have attempted to formulate conditions that are necessary and/or sufficient for possessing certain predicate concepts. It appears that any such account is committed to a substantial analytic-synthetic distinction, since it will hold that certain inferences involving a concept are necessary to having the concept (Fodor and LePore 1992). Willard Van Orman Quine's arguments (1960) that there are no analytic inferences poses an important problem for CRS.
Another approach is informational semantics (Dretske 1981, Stalnaker 1984). These theories are supposed to provide naturalizations of content; that is, they specify naturalistic properties that are claimed to be sufficient for possessing content. Informational theories claim that the content of a belief is constituted by the information the belief state carries under certain conditions. A state S carries the information that a property P is instantiated just in case the occurrence of S is caused by and nomically implies the instantiation of P. Informational theories have difficulty accounting for the possibility of error, since if a belief state has the content that p it carries the information that p. To solve this problem Fred Dretske proposed that the content of a belief is the information that it carries during what he calls "the learning period." A different suggestion (Stalnaker 1984) is to identify belief content with the information the belief state carries under epistemically optimal conditions. Barry Loewer (1987) has argued that these accounts are not successful as naturalizations, since they appeal to notions—learning, epistemic optimality—that themselves presuppose semantic notions.
Fodor has developed a sophisticated variant of informational theories that applies to the reference of Mentalese predicates. On this account, asymmetric dependency theory (ADT), a Mentalese predicate C refers to, for example, the property of being a cow if it is a law that cows cause C s, and any other causal relation between something other than cows and C s depends on this law but not vice versa. That is, if the other causal relations were to fail, it would still be a law that cows cause cows, but if the law were to fail, so would the other causal relation.
ADT is an atomistic account of content in that, contrary to CRS and ITs, it implies that the property of possessing a particular reference is metaphysically independent of inferential connections among thoughts and, indeed, independent of the existence of any other items with content. Whether or not one sees this as an advantage will depend on how one views the analytic-synthetic distinction. Obviously, ADT makes heavy use of metaphysical notions that are less than perspicacious, so one may wonder about its naturalistic credentials. It has also been argued (Boghossian 1991) that it is equivalent to an optimal-conditions account and is subject to the objections that show that account not to be a naturalization.
Teleological theories of content ground the contents of mental states in biological functions. The biological functions of a system in an organism are those of its features that increased the organism's fitness. Teleological accounts are quite elaborate, but the basic idea (Millikan 1984, Papineau 1992) is that there are desire-producing and belief-producing biological systems with certain biological functions. The desire-producing system has the function of producing states that tend to bring about certain effects. The effect associated with a particular desire is its content. The belief-producing systems have the function of producing states that tend to be tokened when certain states of affairs obtain. The state of affairs thus associated with a belief is its content.
Teleological accounts are appealing, since they are naturalistic, assign biological significance to contentful states, and seem to supply them with a kind of normativity. But various serious objections have been raised to teleological theories of content (Fodor 1992). The most serious is that it is doubtful that teleological considerations are sufficient to assign determinate contents to mental states. A desire state will typically tend to bring about a number of different advantageous effects. Natural selection does not select any one of these effects as the content of the desire. Similarly, natural selection will not single out one of the states of affairs a belief state will typically be associated with as its unique content.
Whether or not content properties can be naturalized is an open question. Some consider it a very important question, since they think that if content properties cannot be naturalized then they are unsuitable to appear in scientific theories or, even worse, that they do not exist or are uninstantiated (Stich 1983). The unsuitability of content properties for science would be a blow to the emerging cognitive sciences. But the nonexistence of content properties would be devastating to the way we think about ourselves and others, since these ways are permeated with attributions of contentful states. In fact, it has been argued (Boghossian 1990) that the thesis that there are no content properties is incoherent. Fortunately, no dire consequences strictly follow from the failure of naturalization. It may be that content properties are natural but not naturalizable (McGinn 1991). It is possible that, while content properties are natural, connections between them and properties that occur in the natural sciences are too unsystematic or too complicated for us to discern. But whether or not this is so is also an open question.
Following Gareth Evans's discussion (1982), there has been growing interest in the proposals that there is a distinction between conceptual and nonconceptual content and that the latter plays a significant role in perception (and perhaps imagination) and in subdoxastic (and so unavailable to consciousness) cognitive processing. Exactly what this distinction amounts to, whether there is nonconceptual content, and what its explanatory and epistemological roles may be are all controversial matters.
Beliefs and other propositional attitudes involve relations to thoughts (or propositions), and concepts are constituents of thoughts. It follows that for someone to have the belief, for example, that the Supreme Court is about to convene, he must have the concepts supreme court, about to convene, and so on. A widely held necessary condition for concept possession is that one has the concept C only if one can think an array of thoughts involving C. This is similar to the systematicity that Jerry Fodor (1987/1990) appealed to support his claim that mental representations involved in thought are languagelike. In fact, he and others (who do not necessarily share his views about meaning) think of concepts as words in the mental language deployed in thinking and in propositional attitudes. Evans observed that it is plausible that there are mental states whose contents are not conceptually articulated in this languagelike way. Visual perception seems to involve such states and processes. When one is looking at, for example, a sunset over a distant mountain range, one's perception seems much richer than what can be expressed in thought. There are particular colors and shapes represented in the perception that one is not able to represent in thought. Further, there do not seem to be components of visual representations that one can combine in the systematic ways in which concepts can be combined. In addition to perceptual states, the mental states of animals and the subdoxastic mental representations of humans posited by cognitive scientists are also said to have nonconceptual contents.
On some accounts of mental content, it is not clear that there can be nonconceptual content. For example, accounts like Donald Davidson's (1984), in which there is an intimate relationship between mental contents and the contents of public-language expressions and in which rationality constraints play a role in content determination, seem to preclude there being contents that cannot be expressed in public language. John McDowell (1994), who advocates such an account, has argued against the existence of nonconceptual content. Specifically, he thinks it is essential to mental states with content that they enter into rational and justifying relations with one another and claims that this requires that their contents be articulated conceptually. Famously, views like these resist attributing contentful mental states to animals and to subdoxastic mental processes, since animals and subdoxastic mental processes cannot harbor concepts.
Philosophers who think of mental content in terms of information (examples are Evans, Fodor, and Fred Dretske) can allow for states with nonconceptual content since non-conceptual representations can possess informational content. Dretske thinks that nonconceptual content is more basic than and prior to conceptual content, and that the latter is in some way derived from the former. Since information makes a division of possible worlds into those in which the information is correct and those in which it is not an informational state can stand in semantic relations of entailment and incompatibility with conceptual representations. On the other hand, it is not obvious how nonconceptualized information states can be involved in inference and reasoning.
There are a number of issues that advocates of nonconceptual content need to address. One is whether the distinction between conceptual and nonconceptual content is really a distinction between kinds of content or a distinction between different ways of representing content. Of course, this depends on what one takes content to be. As noted, states with nonconceptual content, like those with conceptual content divide possible worlds into those that are, and those that are not, in conformity with the content. However, conceptual content is often thought of as involving structure that reflects conceptual composition. The question is whether this structure is better understood as a species of content or belongs to the representation that represents the content. Those who think of contents solely in possible-world terms, such as Robert Stalnaker (1998), will see structure as belonging to the representation.
Another issue is that the alleged nonconceptual content of a perception seems to be informationally much more rich than the content of a thought. At the same time, it seems to be finer-grained in that there are distinctions that can be made in perception that we do not and perhaps cannot represent conceptually. It is not clear how these two features fit together. One idea is to think of nonconceptual content as analogous to pictorial or maplike content (Peacocke 2001). If this is correct, it raises the question of whether the pictorial structure belongs to the content or to the representation. Also, as mentioned above, there are issues concerning the epistemological role of nonconceptual content. Can a nonconceptual perceptual state justify a perceptual belief that it causes? Finally, it is not clear whether the contents involved in perception, the mental states of animals, and subdoxastic states and processes are all the same kind of nonconceptual contents. Indeed, theorists who appeal to subdoxastic states and processes often posit sentencelike representations as involved in mental computations. Their contents are nonconceptual only in that they are not available to thought and propositional attitudes.
See also Belief; Concept; Davidson, Donald; Internalism versus Externalism; Knowledge, A Priori; Language of Thought; Naturalism; Philosophy of Mind; Propositional Attitudes: Issues in the Philosophy of Mind and Psychology; Propositional Attitudes: Issues in Semantics; Putnam, Hilary; Quine, Willard Van Orman.
Block, N. "Advertisement for a Semantics for Psychology." In Midwest Studies in Philosophy. Vol. 10. Minneapolis: University of Minnesota Press, 1986.
Boghossian, P. "Content and Self-Knowledge." Philosophical Topics 17 (1989): 5–26.
Boghossian, P. "Naturalizing Content." In Meaning and Mind, edited by B. Loewer and G. Rey. Oxford, 1991.
Boghossian, P. "The Status of Content." Philosophical Review 99 (1990): 157–184.
Burge, T. "Individualism and Self-knowledge." Journal of Philosophy 85 (1988): 644–658.
Burge, T. "Individualism and the Mental." In Midwest Studies in Philosophy. Vol. 4. Minneapolis: University of Minnesota Press, 1979.
Davidson, D. Inquiries into Truth and Interpretation. Oxford: Clarendon Press, 1984.
Dretske, F. Knowledge and the Flow of Information. Cambridge, MA: MIT Press, 1981.
Evans, Gareth. The Varieties of Reference. Oxford, U.K.: Oxford University Press, 1982.
Fodor, J. The Language of Thought. Cambridge, MA, 1979.
Fodor, J. Psychosemantics. Cambridge, MA: MIT Press, 1987.
Fodor, J. A Theory of Content and Other Essays. Cambridge, MA: MIT Press, 1990.
Fodor, J., and E. LePore. Holism: A Shopper's Guide. Oxford: Blackwell, 1992.
Grice, P. "Meaning." Philosophical Review 66 (1957): 377–388.
LePore, E., and B. Loewer. "Solipsistic Semantics." Midwest Studies in Philosophy (1986).
Lewis, D. Convention. Cambridge, MA: Harvard University Press, 1969.
Lewis, D. "Radical Interpretation." Synthese 23 (1974): 331–344.
Loar, B. Mind and Meaning. Cambridge, U.K.: Cambridge University Press, 1981.
Loar, B. "Social Content and Psychological Content." In Contents of Thought, edited by R. Grimm and D. Merrill. Tucson: University of Arizona Press, 1986.
Loewer, B. "From Information to Intentionality." Synthese 70 (1987): 287–317.
Marr, D. Vision. San Francisco: Freeman, 1982.
McDowell, John. Mind and World. Cambridge, MA: Harvard University Press, 1994.
McGinn, C. The Limits of Philosophy. Oxford, 1991.
Millikan, R. Language, Thought, and Other Biological Categories. Cambridge, MA: MIT Press, 1984.
Papineau, D. Naturalism. Oxford, 1992.
Peacocke, C. "Does Perception Have a Non-conceptual Content." Journal of Philosophy 98 (2001): 239–264.
Peacocke, C. A Study of Concepts. Cambridge, MA: MIT Press, 1992.
Peacocke, C. Thoughts: An Essay on Content. New York: Blackwell, 1986.
Pinker, S. The Language Instinct. New York: Morrow, 1994.
Putnam, H. "The Meaning of Meaning." In his Mind, Language, and Reality. Cambridge, U.K.: Cambridge University Press, 1975.
Quine, W. V. O. Word and Object. Cambridge, MA: MIT Press, 1960.
Schiffer, S. Meaning. Oxford: Clarendon Press, 1972.
Sellars, W. Science, Perception, and Reality. New York: Humanities Press, 1963.
Stalnaker, R. Inquiry. Cambridge, MA: MIT Press, 1984.
Stalnaker, Robert. "What Might Non-conceptual Content Be?" In Concepts: Philosophical Issues, edited by Enrique Villanueva. Atascadero, CA: Ridgeview, 1998.
Stich, S. From Folk Psychology to Cognitive Science. Cambridge, MA: MIT Press, 1983.
Warfield, F. "Externalism and Self-Knowledge." Analysis (1994).
Barry Loewer (1996, 2005)