The Semantics and Pragmatics of Polarity

(1 17) :k.lcd~~i gave anyone nothing. (1 18) John gave tiotlii~~g to anyotle. (1 19) *John gave anything to no one. While I have no accoutit of this property of tleed to offer. I believe it may provide a little bit of additional motivation lor a view o l n w d as somewhat exceptional. For some thirty years negative polarity items (NPIs) have provided crucial evidence for litig~iistic theory. But the various accounts of NPIs have not yet attained explanatory adequacy .The goal of this paper is toderive the distribution of polarity items (and in particular of different types of polarity items) from their semantic structure and independently motivated pragmatic principles. Section 1 provides an overview of existing theories of NPIs and their problems. In 5 2, 1 outline my explanation of the distribution of so-called weak polarity items, and in 5 3.1 discuss the semantic nature and distribution of strong polarity items. Section 4 offers a comparison of weak and strong NPIs. Section 5 discusses a wider range of polarity items. Section 6 is devoted to so-called "double licensing", and $ 7 to certain locality effects. In $ 8, 1 discuss NPIs in questions. There is an ongoing debate between syntacticians and semanticists about the proper explanation of the distribution of NPIs. Klima [21] may be seen as the earliest proponent o f a syntactic theory. According to him, NPIs must be "in construction with", or in more recent terms, be c-commanded by, a trigger. Triggers are either an overt negation or an "affective element', e.g., a verb like s/11~/1;Â¥ised Stanford. I wish to express my gratitude to these organizations and tor the financial Parts of the content of this paper were presentcil at talks at the University of Massachusetts ;it Aniticrst in November 1993. and at the conference SALT4 ("Sein;intics iititl Linguistic Theory") at tlie University ofRochester in May 0 9 4 ; I am pratcl'ul for the comments tliiil I received From the audiences there. In particular. I wisli to thank Gene Rohrhaugh and Jack tlockscm;~ for helpful suggestions concerning the con~cnt and the prescntatioi~ ol this paper.