Abstract
The analysis of belief is concerned with what Russell calls 'prepositional attitudes' which are expressed by, e.g., 'Othello believes that Desdemona loves Gassio', or 'A knows that P,' 'A wishes that P', etc. The logical difficulties involved are many. Thus in 'A believes that P', what does 'that P' correspond to if it is false? It cannot be a fact. Since 'believes' is a two-place predicate, then it expresses a relation between two singular terms and 'that must be an entity. But philosophers disagree about whether 'that 'P is a proposition, a sentence, or a statement, and whether there is any difference between them. Other problems are posed by the rules of elementary logic - those of quantification, truth-functions, and identity. 'A believes that P' is obviously not truth-functional. It also transgresses against the laws of quantification and identity because even if has the same truth- value as 'P', the substitution of for in 'A believes that P' will affect the latter's truth-value. No two terms are synonymous and inter-substitutable because synonymy, analyticity, and prepositional identity are unclear concepts. These logical problems result from the application of logical rules which were originally devised for mathematical language. Natural language is quite different in that propositional- attitude and performative utterances are the norm and atomic sentences such as 'Socrates is mortal' are 'implicit' propositional-attitude and performative sentences with the prefix 'I know' that', 'I fear that', 'I state that', etc,, deleted or understood. Standard logical rules can no longer be applied to the analysis of implicit or explicit propositional-attitude or performative sentences. An alternative solution is to regard expressions of propositional attitudes and performatives as couched in meta-language - when they need not obey the rules of standard logic. Either solution will free the analysis from some of its traditional logical problems.