Voice familiarisation training improves speech intelligibility and reduces listening effort

Understanding speech among competing speech poses a substantial challenge. In these environments, familiar voices—including naturally-familiar (e.g., friends, partners) and lab-trained voices—are more intelligible than unfamiliar voices. Yet, whether familiar voices also require less effort to understand is currently unknown. We trained 20 participants to become familiar with three voices, then tested listening effort during a speech intelligibility task. During familiarisation and training, participants were exposed to three talkers for different lengths of time, either speaking 88, 166, or 478 sentences (‘Least Familiar’, ‘Moderately Familiar’, or ‘Most Familiar’ voice, respectively). During each trial of the speech intelligibility task, two competing sentences were presented at a target-to-masker ratio (TMR) of -6 or +3 dB. Participants reported target sentences that were spoken by trained or by novel, unfamiliar talkers. We assessed effort using self-reported ratings and physiologically, using pupil dilation. We found that self-report scores were more sensitive than pupil dilation to differences in TMR, with lower self-reported effort at +3 than -6 dB TMR. The two measures may also be differentially sensitive to the extent of training. We found lower self-reported effort for all three trained voices over unfamiliar voices, with no differences among the trained voices. Whereas, pupil dilation was only lower for the voice that had been trained for the longest. Thus, both self-report scores and pupil dilation showed advantages for the voice that was trained for the longest (~1 hour), but self-report scores additionally showed reduced effort even following relatively short durations of training (less than 10 minutes).