Summary: | Reduced abilities in talker identification are observed when listeners are presented with the voices of familiar talkers while in an unfamiliar emotional tone. Despite the acoustic variations caused by different emotional states, listeners tend to demonstrate extraordinary abilities in matching the voices to their corresponding talkers, which suggests their perceptual constancies of voice-identity correspondence. However, the distinctive acoustic-perceptual correlates contributing to the formation of perceptual constancy have been rarely studied from the aspect of within-talker variability (i.e., how do listeners know they are hearing the same talkers when many of the key acoustic features of their voices are inconsistent across different contexts?). This study investigated the influence of variation in the emotional tone of voice on listeners’ abilities for talker identification. We explicitly trained our participants with five voices in an emotional state and tested their generalization abilities in talker identification by presenting them with the same talkers’ voices in the trained emotional state and the other untrained emotional state. Our results showed that listeners were more accurate when they were presented with trained emotional states than untrained emotional states. The improved accuracy supported that listeners benefitted from their early experience with the voices across multiple contexts to form their constant perceptual representations of voice identities. We also observed a significant correlation between the difference of mean fundamental frequency (f0) and accuracy, suggesting the differences of mean f0 tended to be a distinctive parameter to quantify the invariant features in the vocal signals.
|