Tactile Abstract Rule Learning And Visuo-auditory Crossmodal Generalization:An Event-related Potential Study

Abstract rule learning, which means the ability to extract the abstract rule from the existing sensory sequences to predict the forthcoming stimulus and to generalize to the new instances, plays an important role in human cognition. However, previous studies only have investigated the fixed time sequence in auditory modality, not genuine abstract rules, and there is no underlying neural representations of abstract rule learning viewed from EEG in any other modalities or across modalities. Besides, neural mechanisms of abstract rule learning are controversial. One neural model called predictive coding model attracts more attention, which considers that the cortices learn an internal generative model of the environment and use this model to actively predict the incoming inputs, and the difference between predictions and inputs called the "predictive error". In this present study, we used local-global rule learning paradigm for subjects to learn the abstract rule in tactile modality and across visual and auditory modalities to investigate the neural correlates of abstract rule learning by recording EEG data.The behavior results showed that all subjects had learned abstract rule in tactile, visual and auditory modality respectively. Moreover, the ability of abstract rule learning could generalize across visual and auditory modalities. EEG results showed that local violation of abstract rule led to an early short-term mismatch response in tactile modality and across visual and auditory modalities, and the difference came from the parietal and and bilateral temporal lobe of the brain; whereas global violation of abstract rule led to a late and sustained response only in single modalities, and the generation of global effect is related with attention, and the sources of difference are widely distributed, including frontal, parietal, occipital lobe and bilateral temporal lobe of the brain. Particularly, when the last item of abstract rule was mismatched, the amplitude of P350 at frontocentral electrodes and P600 at posterior central electrodes became larger. When the the last item was omitted, we observed the response where the item should have appeared, which means the predictive error.These results consistently suggest that human brain is able to learn the abstract rule in different modalities and the ability could generalize across modalities, and also spontaneously predict and detect the violation of abstract rule. This study provides data support to universality of abstract rule learning in different modalities and predictive coding neural model.