This proposal outlines a new stage in an ongoing research program being carried out jointly by the PI and Professor Malka Rappaport Hovav (Bar Ilan University, Israel). Its long-term aim is the development of a theory of what speakers of a language know about the verbs of their language that allows them to be used and understood properly. The PI's previous research, much of it NSF-supported, revealed that this knowledge is highly structured and that general principles could allow speakers to determine a verb's grammatical behavior from its meaning. This earlier research also led to the identification of grammatically-relevant elements of verb meaning, thus laying the groundwork for the development of a theory of the composition of verb meanings -- the goal of the proposed research. Such a theory must make explicit the exact components of verb meanings, the principles which govern the combination of these components into formal representations of verb meanings, and the principles that associate these representations with the appropriate syntactic structures. This project investigates the composition of verb meaning by combining previous research results with insights derived from new empirical studies of the syntax-semantics interface: explorations of regular verbal polysemy. In English certain verbs show a range of meanings (e.g., sweep can describe an activity as in Terry swept the floor , removing as in Terry swept the leaves off the walk , or putting as in Terry swept the leaves into the gutter ); this phenomenon is regular in that entire semantic classes of verbs show identical patterns of polysemy (e.g., the surface contact verbs rub and wipe pattern like sweep ). Regular verbal polysemy has been chosen as the research domain because preliminary studies show it to be a rich source of insights into verb meaning, yet extensive, systematic documentation of this phenomenon is still lacking. This project will fill this gap through several detailed corpus-bas ed case studies of verbal polysemy. These case studies will require documenting attested types of variations in verb meaning, uncovering regular patterns of verbal polysemy, and identifying constraints on possible verbal polysemy. The findings will then be used in the formulation of an explicit theory of the composition of verb meanings that can predict the range of meanings that a given verb can exhibit. The proposed investigations into verbal polysemy will provide a deeper understanding of the nature of verb meaning. This study is especially timely since in the recent linguistics literature verbal polysemy, although not well-understood, has repeatedly been argued to bear on the choice between two competing theoretical models of verb meaning; the proposed research, by illuminating the nature of the phonomenon itself, should contribute to the resolution of this controversy. Since polysemy has proved a vexing problem to lexicographers, computational linguists, psycholinguists, and researchers in child language acquisition, the results of this research should also prove valuable to researchers outside the theoretical linguistics community.