Misperception patterns of American English consonants by Japanese listeners in reverberant and noisy environments

Abstract

Perceiving non-native sounds in adverse listening environments is challenging, even for proficient learners. An identification experiment in quiet, three reverberant (RT = 0.78 s, 1.12 s, 1.43 s) and multispeaker babble + reverberant environments was carried out to investigate the relation of listening environments and non-native language proficiency. This research aims to elucidate (1) Japanese and English listeners’ identification of intervocalic English consonants in quiet and adverse listening environments, and (2) the correlation of identification rates with the Japanese listeners’ TOEIC® scores, age of onset of English acquisition and length of residence in English-speaking countries. Results showed (1) English listeners achieved significantly higher overall identification rates compared to Japanese listeners on the subset of 11 consonants /f ʤ ʒ l ɹ s ʃ θ ð v z/ that are often difficult for Japanese to accurately identify, and (2) positive correlations between identification rates and English proficiency measured by means of TOEIC® scores and LOR. Analysis of misperception patterns revealed that an increased amount of background disturbance led proficient listeners’ misperception patterns to resemble those of the lower proficiency listeners, as in the example of native listeners misperceiving /g/ as /ð/ in RT = 1.12 s and highly proficient Japanese misperceiving /g/ as /ð/, /ɹ/, and /l/, and lower group as /ð/ and /l/. The degradation of identification accuracy as background disturbance increased shed light on how one's L1 and L2 proficiency influence speech perception.