Vol. 79 (2008) ► pp.21–29
Nederlandse Baby's Gebruiken Statistische Informatie om Spraakklanken te Leren Onderscheiden
Phonemes are perceived categorically and this perception is language-specific for adult listeners. Infants initially are "universal" listeners, capable of discriminating both native and non-native speech contrasts. This ability disappears in the first year of life. Maye et al. (Cognition (2002)) propose that statistical learning is responsible for this change to language-specific perception. They were the first to show that infants of 6 and 8 months old use statistical distribution of phonetic variation in learning to discriminate speech sounds. A replication of this experiment studied 10-11-month-old Dutch infants. They were exposed to either a bimodal or a unimodal frequency distribution of an 8-step speech sound continuum based on the Hindi voiced and voiceless retroflex plosives (/da/ en /ta/). The results show that only infants in the bimodal condition could discriminate the contrast, representing the speech sounds in two categories rather than one.
Article language: Dutch