Defensible Descriptions

Abstract

Donnellan’s (1966) referential/attributive distinction has been widely discussed in recent years, and various aspects of his proposal have come under scrutiny at different times. One of his most striking claims concerns misdescriptions; he suggests that a speaker can use the man drinking martini to refer to a man drinking tonic water, and say something true even if there is no individual at all who actually satisfies the description. If this were so, then the truth conditions for sentences containing definite descriptions would have to differ radically from those proposed by Russell. However, Donnellan’s arguments for this position have been effectively countered by Kripke (1977), and I do not want to dwell on the topic here.¹ In order to keep this parameter of Donnellan’s distinction fixed throughout the ensuing discussion, therefore, I shall make the following two assumptions:

• (a) let λxϕ(the ζ) represent a sentence of English containing a definite description the ζ in an extensional position.² Then the truth conditions of λxϕ(the ζ) are correctly stated by Russell’s theory; i.e. it is true at a world w iff there is exactly one object u which satisfies ζ in w and λxϕ is true of u in w.

• (b) The RAD (referential/attributive distinction) is to be accounted for at the level of pragmatics, not (truth-conditional) semantics.

The initial stages of the research which led to this paper were supported by the Netherlands Organization for the Advancement of Pure Research. I am grateful to Hans Kamp for helpful criticisms, and to Frank Heny, for his editorial encouragement and patience. My thanks also to Jane Blackett for typing the final draft.