The human capacity for language is beyond doubt a very special ability and this thesis aims at shedding light on this ability by investigating the semantic basis for a theory of interpretation. Such a theory seeks to understand how a hearer understands what is meant by an utterance. Pragmatics plays an important role in accounting for a variety of topics relevant to communication – for instance, context of utterance, speaker’s communicative intentions, implicatures and inference, to mention but a few – but in this thesis it is the specific significance of semantics for a theory of interpretation that will be elaborated and defended. Semantics is fundamental, or so I shall argue, if we want to account for the special features of language that ground our human ability to convey complex thoughts and information across radically different contexts, a feature that is absolutely crucial for our rational nature.
I start out from a suggestion by Kent Bach that we can reserve a specific semantic role for a reconsidered version of Paul Grice’s notion of what is said. In the discussion I also defend his claim that saying does not entail meaning, something that in turn justifies his notion of what is said not to be identified with the proposition expressed or what is meant by an utterance. Yet, Bach’s use of ‘semantic context’ to determine indexical expressions are found wanting. I argue that since we have no good reason for letting contextual considerations into the semantic notion of what is said, given its explanatory purpose, the theoretical significance of what is said in addition to sentence meaning is minor.
In due course I defend a notion of semantics that concerns only the stabile features of language, namely the level of meaning that is the same for a sentence every time it is uttered and in every context of use. One of Bach’s main claims is that semantics is non-propositional and I defend that claim by showing how sentences that contain context-sensitive expressions often fail to express a proposition. I also take this a bit further and propose an even stronger argument against truth-conditional semantics, which is based on the observation that there are complete declarative sentences that nonetheless fail to semantically express a proposition. If I am right, we should give up truth-conditional semantics. Instead of trying to specify propositional meaning I suggest that we restrict our aim, as semanticist, to specify how the words and the grammar work together to restrict the available interpretations.