This thesis will employ a certain amount of linguistic terminology, which I will clarify before continuing. Here I define several key terms:
Implicit in these definitions are various assumptions. Most important is the assumption that there is a distinction between linguistic and world knowledge, as other assumptions follow on from this. This point will be discussed in detail in Chapter 2; the fundamental motivation for this distinction derives from evidence of idiosyncratic uses and interpretations of words which could not follow from general reasoning on world knowledge and a desire to isolate language-specific knowledge from general knowledge which does not vary among speakers of different languages. The implication of this point for the lexicon and lexical semantics is that words in the lexicon do not necessarily correspond directly to concepts, that the structure of the lexicon (i.e. relationships between lexical items) will not necessarily reflect the structure of the world knowledge associated with the denotations of the words in the lexicon, and that the information in lexical entries is restricted to information which has direct relevance for truth-conditional meaning and syntax. The last point means that there is justification for postulating that a certain piece of knowledge is lexicalised only when a particular linguistic phenomenon or a particular interpretation of something cannot be explained purely in terms of world knowledge. The full spectrum of knowledge associated with the concept referred to by a word should not be present in the lexicon; only conventionalised linguistic information should appear there. This assumption will have implications in particular for my analysis in Chapter 5.
An additional assumption which follows from the first is that pragmatic reasoning works with the result of linguistic processing -- that is, it has access to the logical form which results from syntactic and semantic combination but not to information encoded in the lexicon or syntactic constraints. This means that the interaction between linguistic knowledge and world knowledge occurs through manipulation of a logical form, and that any linguistic information which is relevant for this interaction must somehow be represented in this logical form. Any syntactic or lexical semantic constraints on the interpretation of a sentence must be taken into consideration during the construction of a logical form. This assumption will inform the analyses utilised in Chapters 3 and 4, in which particular syntactic combinations of elements in a sentence will affect the interpretation of that sentence.
Much of the dicussion of discourse constraints and pragmatics in this thesis comes from an intuitive, data-driven perspective. What formal elements I introduce derive largely from work by Asher asher:93a,asher:93b, Asher and Lascarides asher_lasc:95,asher_lasc:95b, and Lascarides and Copestake lasc_copestake:95, but formal models of discourse coherence are still in the early stages of development and do not yet fully reflect the complexity of discourse-level reasoning. I have tried to harness the intuitions which underpin these models in my discussions, in order to argue for consideration of the influence of discourse context on interpretation. My goal has not been to identify weaknesses in these models or to develop an alternative model; I aim solely to suggest how a model of discourse coherence might interact with the results of syntactic and lexical semantic processing to explain difficult or apparently idiosyncratic data.