Presenter: Dr Péter Rácz.
When: 22nd May, 2020
Péter Rácz is a post doc in the Cognitive Development Center at the Cognitive Science Department of Central European University. His main interest is how cultural and conceptual systems are learned by the individual and transmitted in the community. He has worked on accent perception, peer pressure with robots, artificial language learning, cultural evolution, and infant development. You can find out more about Peter on his website, https://peterracz.wordpress.com/
Talk Title and Abstract:
Morphological convergence as on-line lexical analogy
Does convergence in word choice entail picking up specific words from each other (you say "dove", I say "dove"), is it driven by affix selection (you say "dived", I say "weaved", because you used the "-ed" suffix), or does it reflect the structure of the mental lexicon (you say "sprung", I say "swung", because "spring" and "swing" are similar)? To find out, we conducted an experiment in which participants played a word-matching game with a computer peer. Word choices were regular and irregular past tense forms of nonce words that resembled English verbs. The participant had to anticipate the peer's choice. We found that, on average, participants successfully converged to peer behaviour. We used lexical models of English morphological variation to model convergence. Modelling results suggested that convergence in this experimental setting was driven by rapid, instance-based analogy, rather than a shift in more abstract generalisations about the behaviour of English verbs