The impact of semantic expectation on the acceptance of a voice imitation

Abstract

Previous research has shown that listeners’ ability to detect high quality voice imitation results in 67% to 100% misidentification (Schlichting & Sullivan, 1997). There is little research on which features of speech samples may affect listeners’ discrimination of imitated voices. This paper considers whether the semantic content of a message impacts upon listeners’ readiness to accept an imitation of a well-known voice and assign the voice to the imitated speaker. Two voice imitations of a well-known Swedish politician, spoken by the same imitator were used. One imitation was a political speech, the other instruction on how to bake a cake. The imitations were presented to two groups of listeners, one imitation to each group. The listeners were asked to familiarise themselves with the speaker’s voice before they undertook Yes-No tests, in which they had to indicate whether the voice they heard was the one they had been asked to remember and recognise. The results indicated that the listeners’ expectation of the topic of the message impacted on their acceptance of the voice imitation as that of the person being imitated.