Information Structure and Discourse Semantics

Information structure has been assumed to be an autonomous part of grammar for more than 100 years. It is represented as a partition of the sentence into two units, e.g., focus background, topic comment, focus topic, focus presupposition etc.1 Even though information structure is sometimes understood as textual or discourse structure, its organizational level is still defined with respect to the sentence or the proposition that is expressed by the sentence. Information structure is often described as effecting felicity conditions, rather than truth conditions. However, focus-sensitive particles interact with information structure exhibiting truth-conditional or presuppositional effects. This interaction is called association with focus and stands in the center of interest of semantic theories of information structure. These theories also assume that focus indicates a partition of the sentence into two informational units. The focus particle is translated into an operator that takes these informational units as its arguments. Like for other functional words, the semantics of the operator is defined as its contribution to the propositional content of the whole sentence. Again, informational effects are restricted to the propositional level, while effects on the discourse level are not taken into considerations. In the beginning of the 80s, a new family of semantic theories was established. They were designed to describe discourse semantics, i.e. effects that extend the sentence boundaries. The central concern of discourse semantics are cross-sentential anaphora. The main insight of the new theories is that anaphoric relations do not relate linguistic expressions to their denotations, but they relate so-called discourse referents to each other. Discourse referents "live" on an intermediate representational level that mediates between the linguistic expression and its denotation.2 Thus discourse representation theories (cf. Kamp 1981, Heim 1982) give a model of how we organize our discourse. The representation of the discourse is incrementally built up: Each sentence contributes its representation to the representation of the previous discourse. Different conditions govern this process, e.g., discourse referents must be linked, subdomains are created that correspond to the scope of operators, additional representations are inserted by accommodation if the previous discourse does not license certain conditions etc. The development of discourse representation theories led to new ways of explaining other discourse