0078 The Brain Fails to Pre-activate the Low-level Stimulus-features of an Expected Stimulus

Abstract Introduction It has been suggested that the human brain is involved in actively predicting upcoming sensory inputs based on previous experience. The extent to which such a mechanism operates during sleep remains elusive. Studies utilizing mismatch negativity (MMN) paradigms report disruption of hierarchical predictive coding in sleep, while others argue for limited but preserved detection of the violation of predictions. Here, we inquire, for the first time, whether the brain actually pre-activates the features of an expected stimulus during sleep. Methods In a passive listening nap design (2.5 hours nap opportunity), participants (N=34) listened to sequences of four different auditory tones varying in pitch, presented at a fixed rate of 3 Hz, while we collected simultaneous Electroencephalography (EEG) and Magnetoencephalography (MEG) data. By manipulating the transition probabilities of the tones, we created random and predictable tone sequences. We analyzed the MEG data using multi-level pattern analysis (MVPA) to decode low-level tone properties and to search for evidence of stimulus pre-activations in the predictable tone sequences, when predicting the next tone from the previous one was actually possible. We performed cluster-based permutation across time, time-generalization, as well as cross-state decoding to identify significant above-chance classification time points. Results The results indicate that subtle changes in pitch, reflecting low-level stimulus features, are decodable in N1 and N2 sleep. This is in line with previous studies showing preserved but attenuated cortical activations related to the processing of low-level stimulus properties during sleep. Neural codes between states appear to be similar, as we observed high cross-state classification accuracies. In wakefulness, evidence of neural codes of the expected stimuli was found in the prestimulus interval, but this was not observed in N1 and N2 sleep, suggesting disrupted predictive abilities even in light sleep Conclusion Processing of low-level stimulus features persists in sleep, but the detection of stimulus statistics appears to be disrupted. Support (if any) We would like to express our gratitude to the Doctoral College 'Imaging the Mind' for their financial support (FWF, Austrian Science Fund: W 1233-B)."