jTRACE modeling of L2 Mandarin learners’ spoken word recognition at two time points in learning

Abstract

This study used the TRACE model of spoken word recognition to simulate adult second language (L2) learners’ spoken word recognition at two time points in learning. The pre-existing architecture of jTRACE with the TRACE-T phonology was used to simulate spoken Mandarin word recognition by adult L2 learners at week 1 and week 15 of structured elementary classroom learning. A modified lexicon with reduced tonal information was used to capture recognition during week 1 of learning. Partially restored tonal information was used to capture the change observed at week 15. jTRACE simulations were validated by comparing the results to eye fixation data taken at week 1 and week 15. The eye-tracking task consisted of viewing four Mandarin words written in pinyin while one of the words was presented auditorily. Roughly half of the trials contained words that were segmentally and tonally contrastive (e.g., gān, chá, pǐ, xiàn ). The remaining trials contained a target and competitor that were segmentally identical but tonally contrastive (e.g., mā , má, pěn, gòng ). Proportion of looks to the target were calculated and compared to the jTRACE simulations using multiple linear regression. The results showed evidence of activation and recognition, thereby corroborating our modeling approach.