Modality-specific consolidation shapes long-term retention in statistical learning

Individuals routinely learn patterns from sights and sounds, yet it remains unclear how memories of those patterns compare across senses over time. Statistical learning, the implicit detection and extraction of patterns from continuous input, has been documented across sensory modalities, but most work has focused on immediate learning in unisensory contexts. This study, using a within-subject design, directly compared 24-hour consolidation trajectories of auditory, visual, and audiovisual statistical learning. Twenty-six young adults completed three statistical learning tasks across auditory, visual, and audiovisual modalities. In each task, participants were exposed to a continuous stream of stimuli with embedded triplet patterns. Learning of these patterns was tested immediately after exposure and again 24 h later. Across immediate and delayed tests, multisensory exposure yielded higher accuracy than unisensory exposure, demonstrating a robust multisensory-over-unisensory advantage. In contrast, only auditory statistical learning showed overnight gains, whereas visual and multisensory learning remained stable, suggesting a consolidation advantage for auditory input. Cross-task correlations, both at immediate test and for consolidation magnitudes, were weak, suggesting substantial modality-specific contributions to statistical learning. These findings provide the first evidence that multisensory input boosts long-term retention of learned patterns, and the consolidation of these patterns operates in a modality-specific manner.