Non-segmental correlates of foreign accent: Perception of accentedness in filtered speech.

Abstract

This study investigated foreign accent detection in filtered speech samples of German bilingual speakers of English. The investigation was conducted with a perceptual experiment and acoustic analysis of the production data. A low-pass audio filter was applied to speech samples elicited by an English sentence-reading task from German and Scottish speakers. The filtering process eliminated most of the segmental information from the speech signal, leaving non-segmental cues available to the listeners, such as intonation, pitch variations, word durations and speaking rate. Two groups of native Scottish speakers were recruited so that 12 listeners evaluated the filtered stimuli and 12 the original, unfiltered material. Both groups used a 7-point scale, ranging from 1=definitely spoken by a non-native speaker of English and 7=definitely spoken by a native speaker of English. The results indicate that listeners were able to identify foreign-accented speech at above chance level from the filtered speech signal. The listeners' background in phonetics, or lack thereof, did not affect the ratings, i.e.: both trained and untrained participants were able to make the distinction between native and non-native speech. However, the inexperienced listeners used a wider range of scores in their judgments. The exploratory acoustical analysis of the filtered sentences revealed that the lowest-rated German speaker employed a slower speaking rate and narrower pitch span than the highest-rated German and Scottish speakers. While no firm conclusions were drawn about the properties of the filtered stimuli that cued the listeners to a non-native accent, this study serves as evidence of the significance of non-segmental factors in foreign accent perception, adding to a body of literature aimed at influencing second language teaching.