I will defend Keith Donellan’s view, that definite descriptions of the form “the F” (hereby, “DDs”) have two kinds of meanings, against the objection which deploys “Modified Ockam’s Razor” (hereby, “MOR”), that additional meanings should not be posited unnecessarily. My defence is this: if speakers understand definite descriptions as having both meanings, then application of MOR is not justified. On the other hand, the semantic ambiguity is justified by our intuitions about what is said when DDs are used.
The understanding of Donnellan’s view I defend is that expressions containing definite descriptions of the form “the F is G” can have two kinds of meanings: An attributive meaning where G is predicated of some unique F, and a referential meaning where the DD points out a salient person or thing which is G. To illustrate: Upon coming across Smith’s mangled dead body, you say that “Smith’s killer is insane.” You do not have a particular person in mind when you speak of Smith’s killer, in fact you do not even know that Smith was killed, but your statement presupposes[1] that some person satisfies the description of “Smith’s killer,” whom you have said to be insane. This is what Donnellan has in mind by the attributive meaning. In another scenario, you are at the trial of Jones who is charged with Smith’s murder. Jones behaves strangely on the stand and so, to the person sitting beside you, you whisper, “Smith’s killer is insane.” Unlike before, you do have a particular person in mind here, namely, Jones. Thus, you mean to refer by using the DD: you have said that Jones is insane. In the attributive case, truth assignment to your statement depends on there being a person who satisfies the DD; if Smith has no killer, it is neither true nor false. In the referential, that Jones satisfies the DD of “Smith’s killer” is not a truth condition; all that is relevant is whether you have spoken truly or falsely of Jones. These examples show that DDs of the form “the F” have both attributive and referential meanings, between which they are ambiguous.
The MOR objection argues that this conclusion should be rejected: additional meanings ought not be posited without necessity, and positing two meanings for DDs is supposedly unnecessary because the difference is better explained by Gricean pragmatic processing. Assuming a Gricean distinction between “what is said” and “what is meant,” the literal meaning of your statement stays the same in both cases (therefore avoiding ambiguity), and what you meant or implicated is retrieved by the hearer, in addition to the literal meaning, to make sense (or determine the truth conditions) of your statement in the given context. The literal meaning of the statement is quantificational: “Smith has a killer, at most he has one killer, and they are insane.” In the context of the referential use, your hearer determines via pragmatic processing (e.g., knowledge about how people tend to use DDs in conversation) that you are in fact talking about Jones and therefore also retrieves the implicated meaning that “Jones is insane.” Gricean pragmatic processes make sense of the difference between attributive and referential meanings by explaining the referential use in pragmatic terms and therefore, by MOR, Donnellan’s posited referential meaning should be dropped.
I think that application of MOR in this case is unjustified because the Gricean explanation does not plausibly reflect speaker intuitions about what DDs mean, whereas the ambiguity explanation does. First, the MOR alternative implies that an attributive expression is false when no entity satisfies the DD. However, in the attributive case, I don’t think it makes intuitive sense to charge you with having asserted a falsehood by “Smith’s killer is insane” if it turns out that Smith died of natural causes, but rather that you simply failed to assert anything meaningful at all. Donnellan would probably agree with this intuition.
Second, The MOR alternative suggests that, in the referential case your hearer processes an existential quantifier prior to the reference to Jones. I agree with Carston that it is likely that speakers only process meanings which are optimally relevant to the situation and therefore I find it implausible that, as the MOR alternative implies of the referential use, your hearer processes an implicated meaning as well as theliteral meaning. Likewise, it is just as implausible that the literal meaning of the expression should be anything other than “Jones is insane.” Given that Jones is salient to your audience, it would be epistemically useless to assert or process the assertion that “Smith has a killer and at most he has one” when referring to him, and it is strange to suggest that the proposition expressed by what is said is otiose in linguistic communication. Lastly, the MOR view has another absurd implication: Even if Jones is insane, if it turns out that he is not Smith’s killer, your statement was literally false– but I think most would say your statement was nonetheless true, regardless of the court’s verdict. So, the more obvious intuition is that “Smith’s killer” literally refers to Jones. If I am right in believing that this is how speakers and hearers understand the meaning of DDs, then MOR shouldn’t apply to the referential use as its meaning is literal. Therefore, we should accept that expressions which use definite descriptions are ambiguous between attributive and referential meanings.
[1] Donellan says he does not take a stance on whether speakers “presuppose or imply” (p33). I think by this he means that he does not take a stance on whether the difference between attributive/referential uses is due to ambiguity or implicature.
Read:
Ezcurdia, Maite, and Robert J. Stainton, eds. The semantics-pragmatics boundary in philosophy. Broadview Press, 2013.
Nadia Miller 