Page 392 - Encyclopedia of Philosophy of Language

Formal Semantics
Seuren P A M 1988 Presupposition and negation. Journal of Semantics 6: 175-226
Steinberg D D, Jakobovits L A (eds.) 1971 Semantics. An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology. Cambridge University Press, Cambridge
Strawson P F 1950 On referring. Mind 59: 320-44
Strawson P F 1952 Introduction to Logical Theory. Methuen, London
Weijters A 1985 Presuppositional prepositional calculi. Appendix to Seuren P A M 1985
Wilson D 1975 Presuppositions and Non-Truth-Conditional Semantics. Academic Press, London
Semantics versus Syntax: Perspectives on Natural Language Content R. M. Kempson
The period between 1972 and the early 1990s was a period of an assumed dichotomy between syntactic and semantic forms of explanation. However, at the end of this period, the sharpness of the division is being questioned as problems of interpretation emerge that need structural solutions. As will be demon- strated, the resolution of this dichotomy involves an accompanying shift in assumptions about the language faculty itself.
In 1972, the battlelines between syntactic and sem- antic investigations into properties definitive of natu- ral language were drawn up by Lewis (1972): semantics was announced to be the articulation of truth-theoretic content, and representational approaches to content were dismissed as 'markerese' with the comment that 'one mightjust as well translate into Latin.' These antimarkerese arguments were addressed against Katz's theory of semantic markers (cf. Katz 1972), but were taken as applying to any characterization of meaning which advocated mean- ing representations intermediate between syntactic explications of structure and the semantic objects which constitute the interpretation of that structure. Chomsky responded to this challenge with the retort (articulated in most detail in Chomsky 1986) that sem- antics was not part of any natural language grammar, hence afortiori not definitive of the language faculty. These two positions became ideological stances, the Lewis approach to natural language understanding developed by those working in the Montague para- digm (Thomason 1974; Dowty 1979; Dowty, et al. 1981; Chierchia and McConnell-Ginet 1990), while the Chomskian concept of natural language developed into the 'government and binding paradigm' (Chom- sky 1982; May 1985; Chomsky 1986).
Aphenomenonwhichposesbothsidesofthedivide with problems of equal severity (though recognized only within semantics, cf. Partee 1984b) is the phenomenon of context dependence. Theinformation conveyed by natural language expressions varies from
context to context, and the process whereby we as hearers establish such values involves processes of general reasoning. The simplest examples involve pro- nouns indexically used. Example (1) can be an assertion about Tom, Dick or Harry, depending on who is being talked about:
370
He is sick.
(1)
By model-theoretic criteria, the sentence is ambigu- ous, having different interpretations as the referent varies. The phenomenon of multiple ambiguity is by no means restricted to such indexical uses. There is an array of different kinds of interpretation assigned to pronominal expressions, labeled variously as bound- variable pronouns, discourse coreference (Reinhart 1983, 1986), E-type pronouns (Evans 1980; Heim 1982), donkey-type pronouns (Kamp 1981; Heim 1982), and lazy pronouns (Karttunen 1968; Cooper 1979). All share the property that the interpretation of the pronoun is determined by some form of linkage with an antecedent, but the type of linkage varies, as does the type of model-theoretic content):
John came in. He was sick. (2) Every student worries that she is going to fail. (3) Joan worries that she's going to fail. (4)
Only a few students entered the exam, but they (5) were confident they would pass.
Every student who entered for an exam, passed it. (6)
Every student who puts her cheque in the (7) building society is more sensible than the student
who puts it in her current account.
With model-theoretic assumptions underpinning the concept of linguistic content, the full set of pro- nominal uses is nonunitary. Such assumptions thus fail to provide a semantic basis for characterizing the information conveyed by a pronoun qua pronoun (there is a voluminous literature on the degree to