Speech errors reveal phonotactic learning in the language production system

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Linked publications, trials & patents

Abstract

Project Summary/Abstract Long ago, Sigmund Freud said that speech errors reveal unconscious knowledge. Today, Freud's notion is most often called upon with regard to implicit linguistic knowledge. For example, speakers implicitly know the phonotactic patterns of their language, such as the fact that in English/h/ must be a syllable onset, and /ng/ must be a syllable coda. When an /h/ slips to another syllable, it will invariably occupy the onset slot, thus implicitly revealing the phonotactic constraint on its location. This project uses an experimental analogue to this effect to investigate how the language production system changes in response to experience. Language learning never stops. Each utterance changes the language production system so that it functions more effectively in that and similar circumstances. The failure to adapt can hinder language acquisition or delay recovery from acquired disorders. This project investigates such adaptation by studying how sensitivity to phonotactic constraints in production changes with experience. It does so through experiments and computational modeling. In the proposed experiments, participants recite strings of syllables that unbeknownst to them follow novel phonotactic constraints. For example, for a particular participant, whenever a syllable contains the consonant /f/, it appears only in onset position. The learning of the novel constraint is revealed in the participant's slips. Prior work has demonstrated that simple constraints, such as /f/ being restricted to onset position during the experiment, are quickly learned and influence the speaker's slips within minutes. More complex constraints appear to require a consolidation period (e.g. a day) before they come out in slips. Other constraints appear to be unlearnable. The central aim of the project is to confront these findings by developing and testing a theory of phonotactic learning that explains the adaptability of the production system. The proposed experiments test alternative accounts of learning and seek to show how this learning relates generally to language learning and use. The proposed computational work develops models of phonotactic learning that are capable of simulating the experimental findings.