Cultural Comparison of Auditory Temporal Processing Based on Statistical Learning of the Brain

Abstract

This study investigates how children's songs (nursery rhymes) are acquired through statistical learning. Using a Hierarchical Bayesian Statistical Learning (HBSL) model that emulates the statistical learning system of the brain, we examined perception and generation processes via simulation experiments across English, Germany, Spain, Japan, and Korean songs. We explored how the model's characteristics evolved over 15 learning trials for each song. Furthermore, by utilizing the probability distributions of each model after 15 learning trials, we generated new songs through automated composition. The results suggest that regardless of culture, statistical learning gradually establishes a hierarchical structure of statistical knowledge over 15 learning trials. In terms of generation, statistical learning leads to a gradual increase in delta-band rhythm. These findings indicate that cultural differences may not significantly modulate the effects of statistical learning. Additionally, this study highlights the developmental origins of creativity and the importance of statistical learning during early development.