Exploring the Developmental Feedback Loop: Word Learning in Neural Networks and Toddlers

Abstract

Early word learning may be supported by a developmental feedback loop: the kind of words a child learns early on support the development of attentional biases, which in turn facilitate further word learning. In neural network simulations and a longitudinal study of toddlers we investigated how the emergence of an attentional bias to shape in word learning impacts vocabulary growth with respect to different kinds of words. If this relationship is causal, we should see that the emergence of a shape bias leads to an increase in the rate of learning of shape-based words relative to other kinds of words. The networks supported this prediction, showing an acceleration of shape- compared to material-based word learning. However, in toddlers, shape- and material-based words were learned similarly around the shape bias emergence. Implications are discussed for the developmental feedback loop account and causal relationships between attentional bias development and vocabulary growth.


Back to Table of Contents