What we mean as what we said or would have said

We usually mean what we say, but sometimes we do not. When I ironically utter ‘What lovely weather’ on a rainy day, or mistakenly utter ‘Jim is a barn door’ instead of ‘Jim is a darn bore’, I say one thing and mean another. However, although utterances like these are not uncommon, they are greatly overshadowed by the volume of humdrum utterances of ‘There is wine in the fridge’ or ‘I really like nachos’ where we mean what we say. And since we usually mean what we say, the following simple thesis will usually be correct. 

(N) S meant that p by uttering e iff S said that p by uttering e.

The simplicity of (N) raises the question of whether it can be refined in order to also cover situations where we do not mean what we say. In this paper we defend a modal version of (N): an analysis where speaker meaning in most situations is identified with what is actually said, and in other situations with what the speaker would have said in certain counterfactual situations. The analysis constitutes a radical, but welcome, break with Gricean orthodoxy, where linguistic meaning, rather than speaker meaning, is ultimately used to explain other semantic notions.