Cross-language mapping using neutral acoustic referents

Research in second language acquisition often relies on constructing a map between two phonological systems. Mapping studies typically use productions from a second language, requiring learners to provide labels in their first language. This study reports a method for determining cross-language maps using a synthetic array of fricative sounds. The fricatives consist of 588 stimuli generated with orthogonal variation of four synthesis parameters which determine the overall spectral coverage and spectral shape of the synthetic fricative noises. Native speakers of English and Japanese engaged with the stimuli in a self-directed search protocol, locating English /ʃ/ in very similar locations, but differing for /f//θ//h/ and /s/. Regression models predicting the probability of stimulus labels find a strong similarity between Japanese labels of English fricatives with Japanese C + V categories: /h/ = /ha/, /s/ = /sɯ/, and /ʃ/ = /ɕi/, with added effects of the native English category location. Two-category maps also are evident, e.g., /f/ and /h/ both map onto /ha/, and /ʃ/ maps onto /ɕi/ and /çi/. Overall, the synthetic reference technique revealed merged categories across the languages for learners with a small adjustment for learning the native English category.