Sachiyo Kajikawa

Infant-directed speech supports phonetic category learning in English and Japanese.

Across the first year of life, infants show decreased sensitivity to phonetic differences not used in the native language [Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: evidence for perceptual reorganization during the first year of life. Infant Behaviour and Development, 7, 49-63]. In an artificial language learning manipulation, Maye, Werker, and Gerken [Maye, J., Werker, J. F., & Gerken, L. (2002). Infant sensitivity to distributional information can affect phonetic discrimination. Cognition, 82(3), B101-B111] found that infants change their speech sound categories as a function of the distributional properties of the input. For such a distributional learning mechanism to be functional, however, it is essential that the input speech contain distributional cues to support such perceptual learning. To test this, we recorded Japanese and English mothers teaching words to their infants. Acoustic analyses revealed language-specific differences in the distributions of the cues used by mothers (or cues present in the input) to distinguish the vowels. The robust availability of these cues in maternal speech adds support to the hypothesis that distributional learning is an important mechanism whereby infants establish native language phonetic categories.

Now you hear it, now you don't: Vowel devoicing in Japanese infant-directed speech

In this work, we examine a context in which a conflict arises between two roles that infant-directed speech (IDS) plays: making language structure salient and modeling the adult form of a language. Vowel devoicing in fluent adult Japanese creates violations of the canonical Japanese consonant-vowel word structure pattern by systematically devoicing particular vowels, yielding surface consonant clusters. We measured vowel devoicing rates in a corpus of infant- and adult-directed Japanese speech, for both read and spontaneous speech, and found that the mothers in our study preserve the fluent adult form of the language and mask underlying phonological structure by devoicing vowels in infant-directed speech at virtually the same rates as those for adult-directed speech. The results highlight the complex interrelationships among the modifications to adult speech that comprise infant-directed speech, and that form the input from which infants begin to build the eventual mature form of their native language.

Age-related changes in sensitivity to native phonotactics in Japanese infants

Japanese infants at the ages of 6, 12, and 18 months were tested on their ability to discriminate three nonsense words with different phonotactic status: canonical keetsu, noncanonical but possible keets, and noncanonical and impossible keet. The results showed that 12 and 18 months olds discriminate the keets/keetsu pair, but infants in all age groups fail to discriminate the keets/keet pair. Taken together with the findings in our previous study [Kajikawa et al., J. Acoust. Soc. Am. 120(4), 2278-2284 (2006)], these results suggest that Japanese infants develop the perceptual sensitivity for native phonotactics after 6 months of age, and that this sensitivity is limited to canonical patterns at this early developmental stage.

Vowel devoicing in Japanese infant‐ and adult‐directed speech

It is well known that parents make systematic changes in the way they speak to infants; they use higher pitch overall, more pronounced pitch contours, more extreme point vowels, and simplified morphology and syntax (Andruski and Kuhl, 1996; Fernald et al., 1989). Yet, they also preserve information crucial to the infants ability to acquire the phonology of the native language (e.g., phonemic length information, Werker et al., 2006). The question examined in this paper is whether information other than phonemic segmental information is also preserved, namely, information concerning the phonological process of vowel devoicing. Devoicing of high vowels between voiceless consonants and word‐finally after a voiceless consonant is a regular and well‐attested phonological process in Japanese (Shibatani, 1990). A corpus of speech by Japanese mothers addressed to their infants and addressed to another adult was examined, and the degree and frequency with which they apply vowel devoicing in each type of speech was ...

Perceptual development of phonotactic features in Japanese infants

Acceptable phonotactics differ among languages. Japanese does not allow consonant clusters except in special contexts and this phonotactic constraint has a strong effect on adults speech perception system. When two consonants follow one another in nonsense words, adult Japanese listeners hear illusory epenthetic vowels between the consonants. The current study is aimed at investigating the influence of language‐specific phonotactic rules on infants’ speech perception development. Six‐, 12‐, and 18‐month‐old infants were tested on their sensitivity to phonotactic changes in words using a habituation‐switch paradigm. The stimuli were three nonsense words: ‘‘keet (/ki:t/),’’ ‘‘keets (/ki:ts/),’’ and ‘‘keetsu (/ki:tsu/).’’ ‘‘Keetsu’’ perfectly follows Japanese phonotactic rules. ‘‘Keets’’ is also possible in devoicing contexts in fluent speech, but the acceptability of ‘‘keets’’ for adult native Japanese speakers is much less than ‘‘keetsu.’’ ‘‘Keet’’ is phonotactically impossible as a Japanese word. The resu...

Adult perceptions of phonotactic violations in Japanese

Adult Japanese speakers ‘‘hear’’ epenthetic vowels in productions of Japanese‐like words that violate the canonical CVCVCV form by containing internal consonant clusters (CVCCV) [Dupoux et al., J. Exp. Psychol. 25, 1568–1578 (1999)]. Given this finding, this research examined how Japanese adults rated the goodness of Japanese‐like words produced without a vowel in the final syllable (CVC), and words produced without vowels in the penultimate and final syllables (CVCC). Furthermore, in some of these contexts, voiceless vowels may appear in fluent, casual Japanese productions, especially in the Kanto dialect, and in some, such voiceless vowels may not appear. Results indicate that both Kanto and Kinki speakers rated CVC productions for contexts in which voiceless vowels are not allowed as the worst; they rated CVC and CVCC contexts in which voiceless vowel productions are allowed as better. In these latter contexts, the CVC words, which result from the loss of one, final, vowel, are judged to be better than...

Analysis on infant speech with longitudinal recordings

The utterances of five infants with their parents was recorded every month from their birth until five years old to investigate spoken language development from the viewpoint of acoustic characteristics. Recording time was at least one hour per month. An infant speech database is now being developed from the recordings. Each entry of the database contains a speech file of an utterance and its transcription with some tags such as speaker, utterance category, and clarity. Using a beta version of the database, preliminary analyses were conducted on utterance duration, speaking rate, interutterance interval, and utterance overlap. Some tendencies were observed. For example, utterance duration became longer for infants as a function of infant age, but it was almost constant for parents. On the other hand, the speaking rate was almost constant (about 4 mora/s) for infants but became faster (from 6 to 8 mora/s) for parents as a function of infant age. An interutterance interval of ‘‘Parent to Infant’’ correlated with that of ‘‘Infant to Parent,’’ even before infants spoke a word. Parent utterance often overlapped infant utterance only after infants spoke two‐word sentences. The implication of these tendencies for spoken language development is discussed.