The Changing Character of the Mental Lexicon:
An Information-based Account of Early Word Learning

Jesse Snedecker
University of Pennsylvania
Michael Brent
Washington University
Lila Gleitman
University of Pennsylvania

Excerpt from the Introduction

In the present paper, we investigate word learning by focusing experimental attention on a typical property of the novice lexicon; namely, its concreteness (Gentner, 1978; Smith, Jones, and Landau, 1996). Early vocabularies are concrete in the sense that they usually label the kinds of things - the dogs and the spoons and the scooters - that populate our perceptual world at its "middlesized" grain. Crosslinguistically, labels for such things tend to be nouns and so, artifactually, early vocabularies are "noun dominant." As we will try to demonstrate, a concrete and noun-dominant vocabulary will arise from the information structure of the input that is available to novice language learners, above and beyond any further constraints that may be imposed by their conceptual biases or limitations. To distinguish between information-based accounts and accounts invoking conceptual change, we hold conceptual sophistication constant by studying word learning in adults, and then we manipulate the information made available to these adult learners (for an earlier experimental series using these methods, Gillette, Gleitman, Gleitman, and Lederer, 1999).

Specifically, the experiments are directed to the following questions: If adults are asked to guess the meanings of words from observing their extralinguistic contexts, will they succeed on items as a function of their concreteness (hence, mimicking the "noun dominance" effect in young children)? Our second question pertains to the source of the noun dominance effect: Can we demonstrate that our subjects' word-learning procedure, and its implicitly perceived successes and failures, can support growth of a bias to acquire nouns in preference to items from other lexical classes? Third, if - by experimental artifice - we implant this bias to acquire nouns in our adult subjects, will they improve at learning words from their real-world contingencies? Before presenting the experiments, we want to provide some further detail about early child vocabularies and the kinds of accounts that have been given to explain them.

(Snedecker, Brent, & Gleitman, Submitted)

Return to LSRG home