Articulatory evidence for feedback and competition in speech production

Abstract

We report an experimental investigation of slips of the tongue using a Word Order Competition (WOC) paradigm in which context (entirely non-lexical, mixed) and competitor (whether a possible phoneme substitution would result in a word or not) were crossed. Our primary analysis uses electropalatographic (EPG) records to measure articulatory variation, and reveals that the articulation of onset phonemes is affected by two factors. First, onsets with real word competitors are articulated more similarly to the competitor onset than when the competitor would result in a non-word. Second, onsets produced in a non-lexical context vary more from the intended onset than when the context contains real words. We propose an account for these findings that incorporates feedback between phonological and lexical representations in a cascading model of speech production, and argue that measuring articulatory variation can improve our understanding of the cognitive processes involved in speech production.

Acknowledgements

Portions of this work were presented at the 2006 CUNY and 2006 AMLaP conferences. This work was partially funded by an NRSA Kirschstein Fellowship Award (F31 DC07282) from the National Institute on Deafness and Communication Disorders (NIH/NIDCD). The authors wish to thank Susannah Moat for helpful discussions during the writing of this paper, and Robin Hellier for help with data collection and transcription.

Notes

¹For comparison with previous studies we also analysed the untransformed error percentages. This analysis yielded a comparable pattern of results, with only the effect of Competitor reaching significance, F ₁(1, 45) = 8.53, p=.005, MSE=.0006; F ₂(1, 94) = 9.48, p=.003, MSE=.014.