Acoustic and perceptual correlates of comprehensibility in relation to suprasegmental features

Comprehensibility and other perception measures of non-native speaking ability are predicted by several objective suprasegmental features such as word stress, rate of speech, and pitch. This exploratory study discusses the relationship between comprehensibility and perceptual descriptors in relation to intonation. First, five Likert scales were created for the purpose of rating speaker intonation by asking listeners to describe several speakers. Then in a perception experiment, listeners were presented with audio clips of native and non-native speakers of English which contained either both segmental and suprasegmental information or only suprasegmental information (low-passed). The listeners rated the clips on comprehensibility, goodness of intonation, and the five perceptual scales. The relationship between these ratings and acoustic measurements of the clips was analysed through principal components analysis. A positive correlation was found between comprehensibility, good intonation, and fluid, confident, and natural production; these were also correlated with several speech rate, pause, and stress measures.