Vandermosten, M. 1, 2 , Wouters, J. 2 , Ghesquière, P. 1 & Golestani, N. 3
1 Parenting and Special Education Research Group, KU Leuven
2 Exp ORL, KU Leuven
3 Brain and Language Lab, University of Geneva
Statistical learning has been proposed as the main mechanism responsible for the developmental change in infants from allophonic to phonemic perception. It could be hypothesized that in dyslexia deficient statistical learning skills hamper the construction of well specified phoneme representations. In this behavioural study we investigate whether statistical learning plays a role in the construction of phoneme representations in a group of 42 normal reading 9-year olds. Children performed a categorical perception task of non-native sounds, i.e. a 7-step continuum of Hindu sounds, before and after statistical learning. During the learning phase, half of the participants were exposed to the Hindu-continuum having a unimodal distribution, whereas the other half (matched for age, gender and non-verbal IQ) were exposed to a bimodal distribution. Preliminary results indicate that children who passively listened to a bimodal distribution improved in categorizing the non-native sounds, whereas the children from the unimodal group did not (group * test moment: F(1,39)=5.74, p=.022). This shows the capacity of children to utilise implicit statistical regularities in a language to build up phoneme representations. However, this improvement observed in the bimodal group was not linked to reading, phonological awareness or native sound categorization.