Belief Attributions

views updated


"Belief attributions" are uses of sentences of the form N believes that s (where N is a noun phrase, s a sentence). Their semantic and logical properties have been debated under the assumption that an account of "believes" will carry over to other propositional attitudes such as desire, knowledge, and fear. Most of the debate focuses on two issues: Does "believe" pick out a relation, and how do so-called de re and de dicto attributions differ?

Is "Believes" Relational?

The obvious hypothesis is that in

  1. Maggie believes that Twain lives.

"believes" has the semantic status of a transitive verb, picking out a relation between a believer and something (a proposition) provided by the verb's complement,

  • (2) that Twain lives.

Grammatical evidence suggests this: "believes" can be followed by names and demonstratives ("I believe Church's thesis," "she believes that") as well as expressions that behave like (nominal) variables ("whenever the pope says something I believe it").

Gottlob Frege and Bertrand Russell, whose work inspires most subsequent debate about belief attribution, agreed on the obvious hypothesis. Frege held that expressions embedded within "believes that" shift their reference to a way of thinking, or sense, of what they refer to unembedded. Russell held that no such semantic shift occurs; the proposition "that s " is determined by what s 's parts pick out when used unembedded.

Since "Twain" and "Clemens" refer to the same author, the Russellian approach seems committed to the identity of the propositions, that Twain lives and that Clemens does, and thus to (1)'s implying

  • (3) Maggie believes that Clemens lives.

Russell would avoid this by saying that "Twain" and "Clemens" typically function as truncated definite descriptions. This last suggestion is widely thought to have been discredited by Saul Kripke.

One problem Fregean views face is that sense is idiosyncratic: Different people associate with a name different ways of thinking of the referent. It is implausible that when I utter (1) I speak truly only if Maggie thinks of Twain as do I. But if (2) in (1) named Maggie's sense for "Twain lives," the argument "Maggie believes that Twain lives; Seth believes what Maggie does; so Seth believes that Twain lives" would be invalid.

Contemporary Russellians such as Nathan Salmon and Scott Soames hold that to believe a proposition involves grasping or representing it and its constituents; thus, belief is a three-place relation among a believer, a Russellian content, and a representation. Salmon and Soames nonetheless hold that (1) tells us only that Maggie believes ("under some representation") the Russellian proposition that Twain lives; the appearance that (1) and (3) may disagree in truth value results from mistaking a conversational or pragmatic implicature, about the representation under which a belief is held, for part of what a belief attribution, strictly speaking, says.

John Perry and Mark Crimmins have suggested that a belief attribution involves implicit reference to the Russellian's representations or modes of grasping: the complement of "believes" determines a Russellian proposition, but the verb has an "implicit argument place" for representations. A use of (1) makes a claim along the lines of "Maggie believes the Russellian proposition that Twain lives under representation r," with the representation referred to differing across occasions of use. A problem with this view is that it renders the argument mentioned two paragraphs above invalid.

Some think belief attributions implicitly quotational. The simplest version of such a view sees that s as a quotation name of s, "believes" naming a relation to sentence types. To this it may be objected that different uses of "Seth thinks I am sad" may have different truth values. Another view sees a "that" clause as picking out a fusion of linguistic items with their interpretationsfor example, the result of combining a sentence with the semantic values of its expressions.

Mark Richard's version of this view has that s pick out a fusion of the sentence s and its Russellian content. In belief attribution, such fusions are offered as "translations" of the believer's thoughts, where a thought is the result of combining a representation that realizes a belief with its Russellian content: (1) is true if the "that" clause provides a translation of a thought of Maggie's. Standards of translation shift from context to context: "Twain" may represent a representation of Maggie's in some but not all contexts. Thus, on this view, the truth of (1) does not demand that of (3).

Donald Davidson denies that (2) is a semantically significant part of (1). "Believes" is a predicate whose second argument is the demonstrative "that"; its referent is the ensuing utterance of "Twain lives." The overall force of (1) is roughly some belief state of Maggie's agrees in content with that utterance. (Davidson made such a proposal for "says" but clearly intended to generalize.) Yet more radical views deny that "believes" is a predicate. Arthur N. Prior took "believes" to combine with a name and sentence to form a more complex sentence; W. V. O. Quine has entertained the idea that "believes that Twain lives" is a predicate without semantically significant structure. A problem for Quine is to explain how infinitely many (semantically unstructured) belief predicates acquire their meanings; Prior thought little useful could be said on such issues.

De re and De Dicto

There seem to be two ways of interpreting such sentences as

  • (4) Sam believes that Melinda's husband is unmarried. Sam believes that some Frenchman is not French.

One interpretation attributes to Sam necessarily false beliefs; the other, suggested by

  • (4) Of Melinda's husband, Sam believes he is unmarried.

Of some Frenchman, Sam believes he is not French, does not. Note that (4) ascribes to Sam beliefs in some sense about particular individuals, while this is not true of the interpretation of (4).

The interpretations seem to correspond to different scopes that may be assigned to the quantifier phrases "Melinda's husband" and "some Frenchman." In a de re attribution, an expression functioning as a variable within the scope of "believes" is bound by a quantifier outside its scope (and the scopes of other verbs of propositional attitudes). Interpreting the sentences in (4) as in (4) is de re attribution: "he" and "she" are bound to "Melinda's husband" and "some women," which are not in the scope of "believes." An attribution that is not de re is de dicto. If we accept a relational account of "believes," we will say that a de dicto interpretation of "N believes that s " attributes to N a belief in the proposition expressed by s. (An attribution might also count as de re if it has a term anaphoric on a name outside of the attribution, as in the natural understanding of

  • (5) Twain was an author, but Seth believes that he was president.)

Not everyone would characterize the de rede dicto distinction as above. Quine held that it is impossible for a quantifier to bind a variable that occurs opaquelythat is, inside a construction, like "believes," which causes failures to substitutivity. If Quine were correct, some other account of the two understandings of (4) is needed. (Quine himself suggested that "believes" is ambiguous.) Quine's view is not widely shared. (See Kaplan, 1986, for discussion.)

The relations between de re and de dicto attributions are of interest in good part because de re attributions are anomalous on some views. A de re attribution identifies a belief in terms of the objects it is about, not in terms of how those objects are conceptualized. For a Russellian this is the norm: All there is to belief attribution is identifying the state of affairs believed to obtain. For a Fregean, (4) is at best an aberration, lacking information about sense, which belief attribution is supposed to convey. De re belief attributions provide a focus for the debates among Russellians, Fregeans, and others.

See also Causal or Conditional or Explanatory-Relation Accounts; Content, Mental; Davidson, Donald; Epistemology; Frege, Gottlob; Knowledge and Belief; Kripke, Saul; Prior, Arthur Norman; Quine, Willard Van Orman; Russell, Bertrand Arthur William.


Crimmins, M. Talk about Beliefs. Cambridge, MA: MIT Press, 1992.

Davidson, D. "On Saying That." In Inquiries into Truth and Interpretation. Oxford: Clarendon Press, 1984.

Frege, G. "Uber Sinn und Bedeutung." Zeitschrift fur Philosophie and Philosophische Kritik 100 (1892): 2550. Translated by P. Geach and M. Black as "On Sense and Reference," in Translations from the Philosophical Writings of Gottlob Frege, edited by Geach and Black. New York: Philosophical Library, 1952.

Higginbotham, J. "Belief and Logical Form." Mind and Language 6 (1991): 344369.

Kaplan, D. "Opacity." In The Philosophy of W. V. Quine, edited by L. Hahn and P. Schilpp. La Salle, IL: Open Court, 1986.

Kaplan, D. "Quantifying In." In Words and Objections, edited by D. Davidson and G. Harman. Dordrecht: Reidel, 1969.

Kripke, S. "A Puzzle about Belief." In Meaning and Use, edited by A. Margalit. Dordrecht, 1979. Also in Salmon and Soames, 1988.

Perry, J., and M. Crimmins. "The Prince and the Phone Booth." Journal of Philosophy 86 (1989): 685711.

Prior, A. N. Objects of Thought. Oxford: Clarendon Press, 1971.

Quine, W. V. O. "Quantifiers and Propositional Attitudes." Journal of Philosophy 53 (1956): 177187.

Quine, W. V. O. Word and Object. Cambridge, MA: MIT Press, 1960.

Richard, M. Propositional Attitudes. Cambridge, U.K.: Cambridge University Press, 1990.

Salmon, N. Frege's Puzzle. Cambridge, MA: MIT Press, 1986.

Salmon, N., and S. Soames, eds. Propositions and Attitudes. Oxford: Oxford University Press, 1988.

Mark Richard (1996)