Using hierarchical statistical learning models to model individual statistical learning

Abstract

Statistical learning is essential for individuals to discover structure in the sensory environment, especially during communication via speech or music. Individual differences in statistical learning abilities have been proposed to account for differences in various cognitive functions and development, including developmental disorders such as dyslexia. In this study, we used a Hierarchical Bayesian Statistical Learning (HBSL) model to model individual learning trajectories as recorded using electroencephalography (EEG) while adults with and without dyslexia listened to structured tone sequences. Although we did not find a significant group difference, our results showed a close correspondence of between the model simulations and the real EEG data and novel sequences generated based on individual models were highly similar to the original stimulus sequence. This provides a proof of concept for future research and suggests that the HBSL model accurately represented the statistical sequence structure in a similar way as did human listeners.