[PS-1.8] Extraction of fine-grained visual statistical information in infancy

Bertels, J. 1 , San Anton, E. 1, 2 , Bulf, H. 3, 4 & Destrebecqz, A. 1

1 ULBabyLab - Consciousness, Cognition, and Computation Group (CO3), Center for Research in Cognition and Neurosciences (CRCN), Université libre de Bruxelles, Brussels, Belgium
2 Fonds de la Recherche Scientifique - FNRS, Brussels, Belgium
3 Dipartimento di Psicologia, Università degli Studi di Milano - Bicocca, Milano, Italy
4 NeuroMI - Milan Center for Neuroscience, Milano, Italy

Infants' ability to detect statistical regularities between visual objects has been demonstrated in several studies (e.g., Kirkham et al., 2002). The extent to which infants learn the actual values of the transitional probabilities (TPs) between these objects nevertheless remains an open question. In order to obtain a fine-grained measure of infants' visual statistical learning, we examined in two experiments 8-month-old infants' ability to discriminate between high and medium values of TPs, on the one hand, and to differentiate between these two types of sequences and new sequences that involved null TPs, on the other hand. Results showed that infants discriminated between these three types of sequences, hence supporting that 8-month-olds extract fine-grained statistical information from a stream of visual stimuli. Statistical learning processes would thus not only involve chunking the stream in smaller units reflecting the associations between visual elements, but also learning the transitional probabilities between these elements.