In The Case for Referential Descriptions Michael Devitt presents his case for the claim that definite descriptions like, for instance, ‘the author of Naming and Necessity’, are semantically ambiguous between referential and attributive uses. The alternative to this is the Russellian view which treats definite descriptions as semantically equivelent to quantifier phrases (‘there is someone who is the author of Naming and Necessity’) and to explain the referential uses as some kind of Gricean implicature. So, one view has ‘the author of namming and necessity’ as having two meanings (just like ‘bank’) whereas the other just gives it one meaning.
Kripke gives an argument against the referential thesis in his paper “Speaker Reference and Semantic Reference” that, as far as I know, Devitt never replies to, in the following passage
there is no reason to suppose that in making an indirect discourse report on what someone else has said I myself must have similar intentions, or be engaged in the same kind of speech act; in fact it is clear that I am not. If I say ‘Jones says the police are around the corner,’ Jones may have meant it as a warning but I need not say it as a warning. If the referential-attributive distinction is neither syntactic nor semantic, there is no reason, without further argument, to suppose that my usage, in indirect discourse, should match the man on whom I report, as referential or attributive. The case is quite different for a genuine semantic ambiguity. If Jones says, ‘I have never been to a bank,’ and I report this, saying, ‘Jones denied that he was ever at a bank,’ the sense I give to ‘bank’ must match Jones’ if my report is to be accurate. (p83 SR)
I think that argument goes like this. If I say “The author of On Denoting” was a lady’s man” and I am using the description referentially and then you report that to someone else saying ‘Richard said that the author of On Denoting was a lady’s man’ you do not have to be using the description referentially (or be giving it that sense) and vice versa. In fact you may not even be able to use it referentially, as you may not know who the author of On Denoting is. If the referential use was a pragmatic phenomenon then this is what we expect as that matches speech acts in general. If the referential use were a semantic phenomenon then this would be ruled out because in genuine semantic ambiguities we do have to report the words with the same sense. Thus since descriptions act more like pragmatic phenomena there is no ambiguity. Notice that we can also adapt this argument to the case involving names. I may say “Kripke is the greatest living philosopher” and you may go and say “Richard said that Kripke is the greatest living philosopher” you may not even be in a position to use the name referentially. If in this case you think you do get a belief about Kripke (because it is a famous name), then consider if I say “Jenny McArthur is a good cook” you then say ‘Does Jenny know how to make pheasant under glass?’ you may do so even though you do not know who Jenny is. So this is a reason to think that names are not ambiguous as between referential and attributive.