Articulatory Imaging Implicates Prediction During Spoken Language Comprehension

It has been suggested that the activation of speech-motor areas during speech comprehension may, in part, reflect the involvement of the speech production system in synthesising upcoming material at an articulatorily-specified level. In this study we seek to explore that suggestion through the use of articulatory imaging. We investigate whether, and how, predictions that emerge during speech comprehension influence articulatory realisations during picture-naming.

We elicited predictions by auditorily presenting high-cloze sentence-stems to participants (e.g., “When we want water we just turn on the…”). Participants named a picture immediately following each sentence-stem presentation. Pictures either matched (e.g., TAP) or mismatched (e.g., CAP) the high-cloze sentence-stem target. Throughout each trial participants’ speech-motor movements were recorded via dynamic ultrasound imaging. This allowed us to compare articulations in the match and mismatch conditions to each other and to a control condition (simple picture-naming). Articulations in the mismatch condition differed more from the control condition than did those in the match condition. This difference was reflected in a second analysis which showed greater frame-by-frame change in articulator positions for the mismatch compared to the match condition around 300-500 ms before the onset of the picture name. Our findings indicate that comprehension-elicited prediction influences speech-motor production, suggesting that the speech production system is implicated in the representation of such predictions.