Can listeners predict whether or not a stutter follows a stretch of fluent speech?

Abstract

Purpose

Neurophysiological studies report that people who stutter (PWS) exhibit enhanced motor preparation before they stutter. This motor preparation pattern raises the possibility of detecting upcoming stutter moments before they actually occur. This study examined whether these motor preparation differences are detectable by listeners in the corresponding acoustic signal, thereby allowing them to predict upcoming stuttering moments. If so, features in these acoustic patterns could potentially be employed by computational procedures to automate detection of upcoming stutters and to target auditory feedback alterations specifically on these locations.

Methods

Forty healthy normal-hearing participants (aged 18–30) listened to seemingly fluent speech extracts each of which was either followed by a fluent (control condition) or stuttered (experimental condition) moment after the fluent extract. Participants listened to each extract and rated the likelihood of the speaker stuttering on the next word on a scale of 1 (very unlikely) to 7 (very likely) as to whether they thought there was a subsequent stutter. Several measures were made on the speech extracts which were examined either as control requirements to ensure no differences between experimental and control material or as covariates to assess any effects they had on judgments between the two conditions.

Results

Listeners gave significantly higher stutter-likelihood ratings for speech originally followed by a stuttered moment although effects were small.

Conclusions

Naive listeners rated speech extracts that were subsequently followed by stuttered moments as more likely to be followed by a stutter than those that were followed by fluent words after the effects of significant covariates were excluded.