EEG insights into predictive coding of temporal regularity in shape sequences

Abstract

The human brain extracts statistical regularities from sensory input as a foundational mechanism for anticipating future events. This process, known as statistical learning (SL), underpins predictive coding across diverse sensory modalities and stimulus types. However, the neural dynamics underlying predictive processing of statistical information remain insufficiently understood. To address this, we examined psychophysiological markers linked to the extraction and predictive application of temporal regularities. Electroencephalography signals were recorded from young adults (n = 30) as they performed a visual SL task. Following exposure to a continuous sequence of abstract shapes, participants completed a judgment task with both adjacent and nonadjacent dependencies, in which the final shape of each triplet was presented to either visual field. Behavioral findings revealed that participants responded with greater accuracy to target triplets than to foils, suggesting effective learning of predictive structures. Event-related potential analyses further showed that the final shape in foil triplets elicited a larger N300 component compared to targets irrespective of visual fields. This reflects neural adjustment to prediction errors in statistical structures. In addition, the beta oscillation discrepancy between targets and foils when responded correctly serves as an indication of the ability to predict visual inputs with regularities. The results demonstrate bilateral engagement in predictive processing of statistical regularities, with the N300 component and rhythmic entrainment specifically indexing the prediction and validation of learned temporal patterns. Our findings elucidate the neural dynamics in predictive processing, offering new insight into how the brain utilizes statistical regularities to anticipate and interpret sensory input.