Investigating the nature of cross-language perceptual comparisons: Evidence from production

Abstract

In cross-language speech perception research, a metric of perceived phonetic similarity between native and non-native sounds both predicts and describes non-native listeners’ perceptual accuracy [Guion et al., J. Acoust. Soc. Am. 107, 2711–2724 (2000)]. By hypothesis, judgments of perceived phonetic similarity reflect the match between a non-native sound and the sound category used to process it. Whereas measures of perceptual accuracy often mask the nature of the perceptual categories used to process non-native sounds, the acoustics of non-native segment articulations may unambiguously indicate whether non-native sounds are processed in terms of perceptually similar native or newly established non-native sound categories. This study examines the relationship between judgments of phonetic similarity between Korean sounds and six English vowels in bVd and bVt words and productions of the same vowels by 40 Korean listeners who differed in age and amount of exposure to English. A metric of perceived phonetic similarity between English and Korean vowels was correlated with results of acoustic comparisons between English stimulus tokens and the listeners’ productions, revealing that a metric of perceived phonetic similarity also predicts production accuracy and that the nature of the sound categories used to process non-native sounds determines cross-language perceptual relationships.