Compositionality in Formal Semantics: Selected Papers (review)

Abstract

Reviewed by: Compositionality in formal semantics: Selected papers by Barbara H. Partee Anna Szabolcsi Compositionality in formal semantics: Selected papers. By Barbara H. Partee. Malden, MA: Blackwell, 2004. Pp. xi, 331. ISBN 1405109351. $47.95. The principle of compositionality says that the meaning of a complex expression is uniquely determined by the meanings of its constituents and how they are put together. The most important consequence for the linguist is that interpretation is an integral part of grammar. Useful as it may be to theorize about disembodied semantic phenomena, or the meanings of sentences as unanalyzed wholes, these are but preparatory activities. Eventually, we have to specify how sentence meanings are built, in much the same way that we specify how (pronounceable) forms are built. This much sounds obvious in 2004, but it was by no means so when Partee came to linguistics. Several strands of the systematic study of sentence semantics had just been initiated at the time, and P became the ‘founding mother’ (as the philosopher of language Robert Stalnaker puts it on the back cover) of the formal semantic strand inspired by Richard Montague’s work. The chapter entitled ‘Reflections of a formal semanticist’ is a recollection of this period. What is the status of compositionality, when it comes under closer scrutiny? One response has been that the principle is wrong, in view of the inevitable context dependence of some constructions, and as far as it works its significance is merely methodological, since the exact nature of meaning, the interpretations of individual lexical items, and the syntax involved are all determined in the process of making it work. Much of P’s oeuvre has been directed at taking up such challenges. Several chapters in the volume grapple with context dependence, and most chapters are concerned with designing general tools and particular analyses that do not sacrifice ‘significant empirical generalizations’ for compositionality—in fact, they often lead to significant new generalizations. Agreeing that compositionality is a theory-internal matter, P has strived to make sure that the theory is a good one. With two exceptions, the fifteen chapters are reprints of published papers, with some new notes. About half of them appeared in forums that are difficult to access and have exerted their [End Page 182] influence through photocopying or hearsay; one merit of this collection is that it makes them widely available. But thanks to its coherence and sustained intellectual excitement, the volume reads like a self-contained book. Five chapters deal with pronouns or binding in an extended sense. ‘Opacity, coreference, and pronouns’ starts with the problem of coreference without reference and offers a detailed comparison of personal and one pronominalization. Going against the then-standard logical treatment of tenses as sentential operators, ‘Some structural analogies between tenses and pronouns in English’ establishes a parallelism, by now classic, between the deictic, anaphoric, bound-variable, and ‘donkey’ style interpretations of pronouns and tenses. (‘Toward the logic of tense and aspect of English’, with Michael Bennett, makes a strong case for interval semantics.) ‘Bound variables and other anaphors’ draws a sharp line between the bound-variable and the pragmatic uses of pronouns and invokes Robin Cooper’s free pronoun strategy to deal with context dependency, a topic that comes up again later. ‘Anaphora and semantic structure’ (with Emmon Bach) is the first in a series of proposals in the literature to invoke an argument-identifying function originating in the lexical entry of the anaphor, as opposed to coindexing in syntax. Much like the paper on tense, ‘Binding implicit variables in quantified contexts’ points out that a host of expressions, local, enemy, foreign, arrive, opposite, and later among them, exhibit deictic, anaphoric, bound, and ‘donkey’ style dependencies. But P argues against explaining these latter by assuming hidden pronouns, because the patterns are not entirely pronoun-like and because they often involve heavy inferencing. Instead, she proposes to quantify over contexts in a rather holistic sense, and calls for a general theory that accounts for the full range of open-class and closed-class context-dependent elements, with pronouns as an extreme case. Although formal semanticists are supposed to be preoccupied with quantifiers, P has been...