Publication Type: Individual Poster
Abstract: Background and Aims: Towards the end of the first year of life, infants show decreased sensitivity to phonetic differences not used in their native language (Werker & Tees, 1984) and increased sensitivity to the differences that are used (Polka, Colontonio, & Sundara, 2001). With an artificial language learning manipulation, Maye, Werker, and Gerken (2002) demonstrated that infants’ speech sound categories change as a function of the distributional properties of the input. In both studies, consonants were explored. The present study investigates distributional learning in vowels, specifically targeting vowel length. In Japanese every vowel has both a long and a short form (e.g., /kado/ ‘corner’ and /kaado/ ‘card’). In English vowel length does not distinguish different forms, but it does play an important role in speech processing, such as perception of voicing in obstruents. Our goal was to explore whether distributional learning could explain the learning of vowel length phonetic distinctions.

Method: Six-month-old Canadian-English infants were familiarized in a unimodal or bimodal manner to an 8-step continuum of short (88ms) to long (180ms) vowels embedded in a nonsense word ([tεki] - [tε:ki]). After that, a discrimination phase presented two types of test trials: Non-Alternating, where a single stimulus was repeated, and Alternating, where infants heard the alternation of the two endpoint stimuli of the continuum.

Key Results: Infants in both conditions (unimodal and bimodal) were able to discriminate between the endpoint tokens of the continuum. There was no interaction between conditions, demonstrating that the type of distribution heard by infants did not affect their performance. This result indicates that construction of vowel length phonetic distinctions does not seem to be principally guided by the distributional learning mechanism.

Conclusions: It is important to remember that in English vowel length also influences other aspects of phonetic processing. Moreover, because vowel length differences are temporal in nature, it could be that infants discriminated the endpoints by simply paying attention to the acoustical differences in duration. Therefore it is still unknown whether the lack of influence of the distributional vowel length information is due to its prominent temporal nature or its significance in other aspects of phonetic processing. Recent studies exploring vowel length in Dutch, Japanese and English infants (Dietrich et al., 2004; Mugitani et al., 2005) support the idea that vowel length perception develops in a different manner than other phonetic cues, and it may not be primarily guided by distributional learning.