Wanrooij, K. 1 , Boersma, P. 1 & van Zuijen, T. L. 2
1 Amsterdam Center for Language and Communication (ACLC), University of Amsterdam
2 Department of Child Development and Education, University of Amsterdam
An important mechanism for phonetic learning in the first year of life is 'distributional learning', i.e., learning by simply listening to the frequency distributions of the speech sounds in the environment. In the lab, where exposure to speech sound distributions typically lasts only a few minutes, distributional learning has been reported for infants of four months and older. The present study examined whether such fast distributional learning can also be demonstrated before this age. Two-to-three-month-old Dutch infants were presented with either a unimodal or a bimodal vowel distribution based on the English /?/~/æ/ contrast (as in 'bet' vs. 'bat'), for only twelve minutes. Subsequently, mismatch responses (MMRs) were measured in an oddball paradigm, where one half of the infants in each group heard a representative [?] as the standard and a representative [æ] as the deviant, and the other half heard the opposite pattern. The results disclosed a larger MMR for bimodally trained infants than for unimodally trained infants, thus extending an effect of distributional learning found in previous behavioral research to a younger age group and a new method (MMRs as measured from event-related potentials in the electroencephalogram). Moreover, the results revealed a robust interaction between the distribution (unimodal vs. bimodal) and the identity of the standard stimulus ([?] vs. [æ]), which provides direct evidence for an interplay between a previously reported asymmetry in vowel perception and distributional learning. These results were obtained when infants were awake, drowsy or in active sleep during the test; when infants were in quiet sleep there was no effect of distributional learning or perceptual asymmetry.