Listeners pay attention to rhythmic cues when judging the nativeness of speech

Abstract

Native listeners are good at detecting whether speech is foreign-accented or not. Here, we tested the role of rhythmic cues from the amplitude envelope (ENV) in this process. In a binary forced-choice perception experiment, ten L1 Italian listeners listened to 32 stimuli, each containing two Italian utterances of identical lexical content under the following conditions: (a) Both utterances were produced by a non-native speaker and were manipulated with either the ENV of a native German speaker of Italian and Italian L1 speaker (at 10 and 30 frequency bands). (b) One utterance was produced by an L1 Italian, the other by an L2 speaker. For 10 and 30 frequency bands their ENVs were exchanged (speech chimeras). Listeners' task was to choose which of the utterances in a stimulus was the more native-like. In condition (a), listeners’ probability to choose the utterance with the L1 envelope was above chance at 10 bands (0.72) and increased with 30 bands (0.78). In condition (b) listeners’ probability to choose the utterance with the L1 ENV was 0.19 at 10 bands and 0.46 at 30 bands. We conclude that rhythmic cues in the speech ENV influence listeners' perception of nativeness.