Corpus-based L2 phonological data and semi-automatic perceptual analysis: the case of nasal vowels produced by beginner Japanese learners of French

Processing large amounts of non-native (L2) phonological data for acquisition-related research remains a challenging task, especially when acoustic analyses are not straightforward as is the case with nasal vowels. Within the InterPhonologie du Français Contemporain project (IPFC), we have developed a perceptual coding procedure and a piece of dedicated software aimed at providing an intermediate stage between fine-grained acoustic analyses and coarse-grained phonological categorization, such as ‘substitution’ or ‘deletion’, of nonnative productions. Our code allows us to examine the left and right phonological contexts of the segment under scrutiny and assess the nasality, quality and potential consonantal excrescences of the non-native nasal vowels. We have applied this procedure to Japanese data collected in a longitudinal study of French interphonology, focusing on the vowels /A~/, /O~/, /E~/ produced by 22 beginner university students in a wordlist repetition and reading task. Our study reveals an overall good production rate in terms of nasality for such beginner learners but also a lower rate of quality accuracy for the three vowels, as well as better performances in the repetition task. We discuss our results in light of current L2 learning theories and the phonetic-phonological characteristics of Japanese.