Accent affects voice identification

Voice identification across and between utterances are distinguishable tasks. To test how accent affects voice identification, we used voice-cloning to isolate and extract speech patterns from English-language recordings of people from the UK, Poland, and China. These speech patterns were applied to a different set of cloned voices. The result is a set of voices that have the same or different identity or accent and speak the same or different phrases. Participant accuracy in distinguishing voices was tested using varied voice pairs. Different accents applied to the same identity and the same accent to different identities increased errors by ~10% in telling whether people were the “same” or “different”. Using different sentences diminished error rates only when different identities had the same accent. Collectively, these findings reveal that expectations about accent use affect our ability to identify people as same or different, in addition to familiarity and other language-related cognitive processes. Our research supports the view suggesting that voice identification is a complex task that humans employ several different cognitive strategies to solve, depending on the context.