Abstract Speech production and recognition studies have revealed cross-linguistic variation in the grain size of phonological processing, particularly in the relative importance of individual segments and segment strings (analytical processing) and entire word forms (holistic processing). This variation has been ascribed to orthography (alphabets highlight individual phonemes) and/or syllabary size (the fewer the syllable types, the easier to process them as wholes). In the present study, we attempt to deconfound the independent effects of these correlated language-level variables by testing a varied language sample: Cantonese, English, Hakka, Japanese, Northern and Southern Vietnamese, Polish, Southern Min, and Taiwan Mandarin. In each, native speakers made binary auditory wordlikeness judgments for randomly generated minimal nonwords. We analyzed the judgments as a function of neighborhood density (suggesting the processing of whole test items), and of phonotactic probability and initial-segment frequency (suggesting the processing of segments and bigrams), as modulated by orthographic type and syllabary size. As expected, phoneme-based orthographies tended to be associated with stronger phonotactic probability effects, while smaller syllabaries were associated with stronger neighborhood density effects. This new evidence, gathered in an unusually broad typological survey and using a consistent task, may help contribute to the development of a truly universal model of phonological processing.
Chen et al. (Thu,) studied this question.