When one meaning is not enough: Word learning in a polysemous world

Abstract

Words commonly refer to multiple meanings, a phenomenon known as polysemy: caps can appear on bottles, human heads, or mushrooms; green can refer to a color or an environmental stance; a sheet can be in a notebook or on a bed. Polysemy presents a challenge to many theories of meaning and learning, which posit individual meanings and competition between candidate referents. As a result, many open questions remain about how polysemous meanings are learned and represented. In this dissertation*, I address these fundamental issues from a number of angles, using behavioral, developmental, clinical, and computational methods. I first review existing approaches to word meaning and representation (Chapter 2). I then investigate how children and adults uncover and learn the rich, hidden conceptual structure of meanings using novel learning tasks (Chapter 3) and eye- tracking (Chapter 4). In Chapter 5 I investigate variation in generalization by considering a neurodiverse population. In Chapter 6, I develop and evaluate theories of word generalization in quantifiably-testable computational models. Finally, in Chapter 7 I outline a new approach to understanding word learning.