Mahesh Srinivasan — Associate Professor, University of California, Berkeley

Mar 29, 2023, 12:00 pm1:00 pm


Event Description

Re-imagining how children learn the meanings of words

Learning a language requires children to develop exquisitely complex and subtle intuitions about the meanings of words, based on limited data and minimal explicit instruction. An overarching goal of my research is to understand how children solve this puzzle, and what this reveals about the nature of linguistic and cognitive development more broadly.

In this talk I will explain how my work motivates a re-thinking of two general, widely-accepted solutions for how children learn words. The first and main part of the talk will interrogate one classic solution—the idea that children are guided by certain assumptions about the meanings of words that greatly simplify the learning problem—in particular, the assumption that a new word will have only a single meaning, corresponding to a single taxonomic category. I will argue that this theory does not make sense of why most common words across languages are polysemous (i.e., they express multiple distinct but related meanings), because it predicts that children should struggle to learn these words. Instead, I will present evidence suggesting that, rather than impeding learning, polysemy helps children overcome some of the challenges inherent to learning new words, by allowing children to use their knowledge of one meaning of a word to make inferences about other, related meanings for that word. These findings challenge prevailing theory and highlight children’s creativity in expecting words to be used flexibly, across families of related meanings.

Finally, in the last part of my talk, I will re-visit another long-standing idea about how children learn words: the notion that, in direct conversations with children, caregivers provide labels that follow their children’s focus of attention, as well as other important clues to the meanings of new words (e.g., eye gaze, gestures). An extension of this idea is that children should be able to learn little from speech that they overhear, which often lacks these features and requires attending to interactions among others. Challenging this prediction, I will present evidence of early word knowledge—including knowledge that could only have been learned from overhearing—among infants from a Mayan community in Southern Mexico who are rarely spoken to directly. These results highlight the active role that children can play in their own language development, and open the possibility that children from different communities may deploy different strategies to break into language, with their respective strategies adapted to the specific contexts in which they are developing.

Taken together, these two lines of work show how theories of language learning can be re-imagined by broadening the scope of research to capture the complexity of the knowledge that children have to acquire, and the diversity of contexts in which they develop this knowledge.

Department of Psychology
Event Series