Seminar SCALab 25/01/2024

Statistical learning in sequence processing and prediction

Statistical learning is a crucial cognitive mechanism particularly involved in language. It enables the automatic extraction of statistical regularities present in the environment, which in turn enables sequences to be learned. However, previous experiments have often focused on a single statistical regularity at a time and using mainly offline measures, thus neglecting to test the use of statistical regularities in sequence processing and prediction. In order to determine whether and how statistical regularities are involved in learning, sequence processing and prediction, a series of experiments was carried out to examine the effect of different types of statistical regularities (transitional probability and frequency) and their interactions in different types of sequences (linguistic or non-linguistic). The results showed that transitional probabilities are largely involved in prediction in all types of sequences. The current challenge is now to assess the generalizability of these findings to natural language processing.