Calendar of events

Title: Can 4-year-olds’ language learning be driven by prediction errors?

Abstract: rediction errors are a fundamental component of influential accounts of learning and memory. However, the role of prediction errors in language learning is unclear. To clarify their role in English 4-year-old’s developing comprehension of sentence structure, we focus on a particular structure that they are known to struggle with - double object datives (DOs). These are sentences in which the recipient of a transfer action (e.g., giving) precedes the theme (e.g., the given object), e.g., “Marshall will give the elephant [recipient] an apple [theme]”. Recently, Gambi & Messenger (2023) found tentative evidence that 4-year-olds exposed to DO sentences encouraging generation of stronger incorrect linguistic predictions showed greater improvement in comprehension of DOs (measured by “acting out” the sentence) than those exposed to DO sentences that did not encourage strong predictions. This provided indirect evidence of prediction error-based learning. In this talk, I will present data from two replications (one on-going) of Gambi & Messenger (2023) which seek to provide both further indirect evidence of prediction-error based learning, and more direct measures of prediction-errors from eyetracking data, as well as introducing a planned follow-up.

After completing my PhD at Sheffield looking at intentional communication in infants, I have been a postdoc in York, Reading, UCL and am now at Warwick. My current postdoctoral research focuses on prediction-error based language learning in 4 year olds and the role of prediction-error and curiosity on memory/learning in children and adults. Additionally, my research focuses on deception in children and multimodal communication (vocalisations, gestures, facial movements) in both humans and chimpanzees.