Altered auditory feedback induces coupled changes in formant frequencies during speech production

Speaking is one of the most complicated motor behaviours, involving a large number of articulatory muscles which can move independently to command precise changes in speech acoustics. Here, we used real-time manipulations of speech feedback to test whether the acoustics of speech production (e.g. the formants) reflect independently controlled articulatory movements or combinations of movements. During repetitive productions of \"head, bed, dead\", either the first (F1) or the second formant (F2) of vowels was shifted and fed back to participants. We then examined whether changes in production in response to these alterations occurred for only the perturbed formant or both formants. In Experiment 1, our results showed that participants who received increased F1 feedback significantly decreased their F1 productions in compensation, but also significantly increased the frequency of their F2 productions. The combined F1-F2 change moved the utterances closer to a known pattern of speech production (i.e. the vowel category \"hid, bid, did\"). In Experiment 2, we further showed that a downshift in frequency of F2 feedback also induced significant compensatory changes in both the perturbed (F2) and the unperturbed formant (F1) that were in opposite directions. Taken together, the results demonstrate that a shift in auditory feedback of a single formant drives combined changes in related formants. The results suggest that, although formants can be controlled independently, the speech motor system may favour a strategy in which changes in formant production are coupled to maintain speech production within specific regions of the vowel space corresponding to existing speech-sound categories.\n\nNew & NoteworthyFindings from previous studies examining responses to altered auditory feedback are inconsistent with respect to the changes speakers make to their production. Speakers can compensate by specifically altering their production to offset the acoustic error in feedback. Alternatively, they may compensate by changing their speech production more globally to produce a speech sound closer to an existing category in their repertoire. Our study shows support for the latter strategy.