Abstract
In this study, we examined how preverbal infants segment words from fluent speech without any explicit cues. We investigated the development of speech perception in terms of language generality (innate ability) and language specificity (learning from input) by comparing stress cues in English and accent cues in Japanese. We applied the Headturn Preference Procedure to test infants' segmentation ability. Our results suggest: (i) 7-month-old infants do not segment words by using the flat accent pattern (Type 0) as a cue, which most frequently appears in three-mora words; (ii) 7-month-old infants do not use the high-low-low accent pattern (Type 1) as a cue, where the initial syllable of a word is prominent; (iii) the intensity level and the vowel duration rather than the pitch of words, specifically those of the phonotactic pattern [a+u], might be segmentation cues for Japanese infants. The prominence of the initial syllable of a word is possibly a language-general cue for word segmentation.
