LEC talk, 13th August, Molly Lewis

Tuesday 13 August, 11am

DSB 1.17

Multiple routes to solving the word mapping problem

Mapping a novel word to its referent is an under-constrained problem. To successfully learn words, children must therefore approach this problem with certain biases. In the literature, two classes of biases have broadly been proposed: in-the-moment pragmatic inferences and higher-order constraints. These two classes of biases have traditionally been viewed as competing accounts of word learning. I will discuss work from two lines of research that explores how biases at different timescales may jointly contribute to word learning. In the first part, I will present computational work suggesting that children’s bias to map a novel word to a novel object (often dubbed a ‘mutual exclusivity’ bias) could, in principle, result from both a pragmatic inference and a higher-order, one-to-one constraint on lexicons. In the second part, I will discuss recent work exploring the presence of a bias to map longer words to more complex meanings, and suggest how this bias too might be the result of both a pragmatic inference and a higher-order constraint.