Delayed, Reduced and Redundant: Information Processing of Prediction Errors during Human Sleep

During sleep, the human brain transitions to a ‘sentinel processing mode’, enabling the continued processing of environmental stimuli despite the absence of consciousness. We employed advanced information-theoretic analyses, including mutual information (MI) and co-information (co-I), alongside event-related potential (ERP) and temporal generalization analyses (TGA), to characterize auditory prediction error processing across wakefulness and sleep. We hypothesized that a shared neural code would be present across sleep stages, with deeper sleep being associated with reduced information content and increased information redundancy. Twenty-nine participants (15 women) underwent an auditory ‘local-global’ oddball paradigm during wakefulness and an 8-hour sleep opportunity monitored via polysomnography. We focused on ‘local’ mismatch responses to a deviating fifth tone after four standards. ERP analyses showed that prediction error processing continued throughout all sleep stages (N1-N3, REM). Mutual information analyses revealed a substantial reduction in encoded prediction error information particularly during N3 and REM, although ERP amplitudes increased with deeper NREM sleep. We also observed delayed information encoding during sleep, and co-information analyses showed neural dynamics became increasingly redundant with increasing sleep depth. Temporal generalisation analyses revealed a largely shared neural code between N2 and N3 sleep, though it differed between wakefulness and sleep. We demonstrate how the neural code of the ‘sentinel processing mode’ changes from wake to light to deep sleep and REM, characterised by delayed processing, more redundant and less rich neural information in the human cortex as consciousness wanes. This altered stimulus processing reveals how neural information evolves with variations in consciousness across the night. Statement of Significance Even during sleep, the human brain remains responsive to its surroundings. Using an auditory stimulation paradigm, the study reveals how the neural code underlying this 'sentinel processing mode' changes from wakefulness to sleep and with increasing sleep depth. Using computational methods to precisely characterise information processing in the brain, we show that as sleep deepens, the brain encodes less information at increasing redundancy. These findings provide new insights that may help understand why we lose consciousness when falling asleep.