Segregated Encoding of Auditory and Visual Information in EEG Oscillations

Extracting unisensory information from multisensory inputs is a fundamental function of a sensory system. However, little is known on how modality-specific stimulus information is encoded during multisensory processing. Electrophysiology recording approaches, with a temporal precision far greater than most neuroimaging techniques, allow investigations of how sensory inputs are encoded as neural activities in the temporal domain. Recently, information theory has demonstrated a powerful capability of capturing non-linear relationships, renowned for its robustness against challenges posed by non-Gaussian and discontinuous biomedical data. In this study, we recorded scalp EEGs in lightly anesthetized cats and adopted conditional mutual information (CMI) as a quantification to investigate how brains deploy the encodings of visual and auditory stimuli, respectively, while reacting to both. Our data demonstrated that the temporal dynamics of delta and low-gamma band neural oscillations showed higher trial-to-trial reproducibility associated with visual stimuli while those of alpha, the mid- and the high-gamma band oscillations showed higher reproducibility associated with auditory stimuli. Our findings suggested there exist modality-specific neural signatures for the encodings of auditory and visual inputs, which cannot be extracted with existing methods. The mutual information quantifications we report can be utilized as a way of segregating auditory and visual components for spectrally decomposed EEG signals in response to arrhythmic audiovisual inputs. • designed an audiovisual paradigm with independent sound and light sequences • EEG and mutual information analysis quantified reproducibility of sensory responses • stronger delta and low-gamma activities tuning to the visual activities • stronger alpha and high-gamma activities tuning to the auditory activities