UCL Psychology and Language Sciences


Phonotactics vs Stress in Word-Learning during Speech Segmentation

Funded: British Academy (2010-2012)
Research Team: Andrew Nevins, Katrin Skoruppa, Adam Gillard
This project focuses on the use of conflicting cues in speech segmentation, focusing on a strong and salient cue, word-stress, versus phonotactic constraints on sound sequences. Segmenting the words from running speech is a difficult problem in first and second language acquisition, because there are not reliable silences between words (and moreover, some words have silences within them). While research with infants and adults has shown that both word-stress and phonotactic constraints (e.g. the fact that an English word cannot end with a short/lax vowel such as "ih") are used in speech segmentation, it is not known which type of cue is more reliable, salient, or preferred. The research will involve experiments that put the English preference for trochaic stress (strong-weak) in conflict with phonotactic sequences that favor a parsing of the speech stream into the weak-stress pattern. The methodology will involve artificial grammar experiments, and the results will have implications for first and second language acquisition and for linguistic typology.