The effect of talker variability on word recognition under cochlear implant simulation

Abstract

Normal-hearing (NH) listeners have been shown to be less accurate and/or slower to recognize spoken words when the talker changes from trial to trial, compared to when the talker remains the same. Less is known about the effect of talker variability on cochlear implant (CI) users, who display a limitation in talker perception. As a first step, the current study investigated how limited talker information via CI simulation influences word recognition. In two experiments, NH listeners completed a word naming task in different talker conditions (single (ST), multiple female (MT-F), and multiple female/male (MT-M)) and simulation conditions (12, 8, and 4 vocoder channels). An effect of talker variability was observed, but only when talker differences were maximized. Listeners were less accurate, but not slower, at recognizing words under the MT-M condition, compared to ST and MT-F conditions. Talker variability effects did not vary by the simulation condition. These results are consistent with previous studies with NH listeners, showing similar performance in conditions where talker changes were not detected but talker differences were large. Taken together, these findings suggest that limitations in talker perception for NH listeners under CI simulation, and potentially CI users, alters the perceptual strategy for recognizing words under multiple-talker conditions.Normal-hearing (NH) listeners have been shown to be less accurate and/or slower to recognize spoken words when the talker changes from trial to trial, compared to when the talker remains the same. Less is known about the effect of talker variability on cochlear implant (CI) users, who display a limitation in talker perception. As a first step, the current study investigated how limited talker information via CI simulation influences word recognition. In two experiments, NH listeners completed a word naming task in different talker conditions (single (ST), multiple female (MT-F), and multiple female/male (MT-M)) and simulation conditions (12, 8, and 4 vocoder channels). An effect of talker variability was observed, but only when talker differences were maximized. Listeners were less accurate, but not slower, at recognizing words under the MT-M condition, compared to ST and MT-F conditions. Talker variability effects did not vary by the simulation condition. These results are consistent with previous studie...