Publication Date


Journal or Book Title

Proceedings of the Second Tbilisi Symposium on Language, Logic, and Computation


In this paper we discuss the analysis of expressions such as John’s team, John’s brother, John’s favorite movie, Mary’s favorite chair, Mary’s former mansion. Before introducing the concrete problems, we briefly describe our theoretical perspective. Our theoretical concern is the integration of formal semantics and lexical semantics, especially but not exclusively in the traditions of Montague Grammar and the Moscow School (Apresjan (1994), Mel’èuk (1982), Paducheva (1996)), respectively. We have proposed (Borschev and Partee (in press)) to modify the Moscow school approach and represent lexical information in the form of sets of meaning postulates, which may or may not exhaust the meaning of the given lexical item. We believe this use of meaning postulates is consistent with actual Moscow school practice, and it makes it possible to integrate lexical semantics with the compositional “semantics of syntax” given by formal semantics. If the formal semantic interpretation of a sentence is given as a formula of intensional logic in which lexical items are primitives, and lexical semantics as a set of meaning postulates for these lexical items, then their integration can be seen as the drawing of entailments from these sources. This approach is in principle extendable to the integration of semantic interpretation with contextual and other information as well. So we semantically represent a sentence or a text as a theory consisting of different sorts of formulas, i.e. different sorts of axioms and their entailments. By “theory” here, we do not mean the metalevel linguistic theory, but the set of axioms from various sources plus the consequences that can be drawn from these axioms, which together constitute the interpretation of such a sentence in a given context. Such a theory (see Borschev 1996) characterizes the class of all models that are consistent with the content of the given text, or of the text together with certain aspects of its context, if the theory includes axioms representing contextual information. The most general structure (features and constraints) of such models have to represent what the Moscow School calls “naivnaja kartina mira” ‘the naive picture of the world’, and what formal semanticists, following Bach (1986) and Link (1983), call Natural Language Metaphysics or Ontology. This general scheme, particularly the principles governing interactions among axioms from different sources, has to be investigated with concrete linguistic material. On our modeltheoretic perspective, all of the “axioms” from all of the different sources jointly constrain the possible models, and their joint effects may account for phenomena ranging from ambiguity reduction to meaning-shift phenomena such as “coercion”. On this view, cooccurrence restrictions reflect the sometimes incompatible demands that different elements may make on the interpretation of the whole. Ambiguities are decreased when not all of the possible variants provide a consistent (or sufficiently plausible) interpretation. Inconsistency, which should in principle always result in “anomaly” judgments, may lead instead to type shifting or other meaning shifts, the complexities of which are one of the main concerns of this paper. It will probably turn out that the mechanism of axiom interaction is rather complicated, and may include modifications (shifts) in some axioms in the context of the others. We do not pretend to have an articulated view of the nature of all the different sorts of axioms that may play a role in the “theory” of a text, but we will illustrate some of the possibilities for a few of them.



Included in

Linguistics Commons