Abstract:

Statistical learning is a core cognitive ability that enables humans to extract regularities from the environment. However, the characteristics and mechanisms underlying its cross-modal transfer remain theoretically contentious. This study systematically investigated the cross-modal transfer mechanisms of statistical learning across the visual and auditory modalities through four experiments. Experiment 1 employed a classic statistical learning paradigm to validate individuals’ visual statistical learning ability using animal pictures as stimuli. In Experiment 2, following unimodal visual learning, participants were tested with either animal pictures or animal sounds. Results revealed no significant difference in statistical learning performance between the two test conditions, thus confirming cross-modal transfer from vision to audition. Experiment 3 presented synchronized audio-visual streams consisting of animal pictures and meaningless syllables; the results indicated that the vision-to-audition transfer effect remained stable irrespective of whether the auditory stream contained statistical regularities. Experiment 4 further presented synchronized audio-visual streams of animal sounds and meaningless shapes, which demonstrated that robust cross-modal transfer also occurs from audition to vision. Collectively, these findings verify that statistical learning exhibits robust bidirectional cross-modal transfer capability, thereby providing empirical support for the theoretical perspective that statistical learning is modality-general. The cognitive underpinnings of this capacity may reside in a hierarchical parallel system for representing statistical structures.

Key words: statistical learning, animal pictures, animal sounds, multimodal, cross-modal transfer