Selective preservation of prediction-related signals in human sleep
Year: 2025
Type: article
Abstract: Imagine listening to a familiar song on the radio. As the melody unfolds, you often anticipate the following note or beat before it plays. This ability reflects the brain's capacity to extract statistical regularities from sensory input and predict future sensory events. It is considered automatic, requiring little to no conscious effort or attention. 1 , 2 , 3 , 4 , 5 But to what extent is this predictive ability maintained when cognitive resources are minimized, such as during sleep? Experimental findings from animal and human studies reveal a complex picture of predictive processing during sleep. 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 Although some forms of predictions persist—evidenced by differential brain responses to unexpected stimuli and rhythmic music 8 , 9 , 11 , 12 —neural markers of feedback processing linked to predictions are notably disrupted. 7 , 14 Here, we use multivariate pattern analysis (MVPA) to capture different facets of prediction-related signals, determining whether the brain preactivates the low-level features of expected stimuli and tracks statistical associations between stimuli. Using predictable and random tone sequences in a passive-listening paradigm, we recorded brain activity via electroencephalography (EEG) and magnetoencephalography (MEG) during wakefulness and sleep. We first show that subtle changes in tone features (e.g., tone frequency) elicit feature-specific responses in N1 and N2 sleep, though weaker and less sustained than in wakefulness. Critically, even during sleep, the brain preactivates feature-specific representations, but higher-order tracking of statistical associations between tones remains restricted to wakefulness. Altogether, our results suggest that feature-specific auditory processing is retained despite the fading of consciousness, while only some aspects of anticipatory predictive processing are preserved.
Source: Current Biology
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: hybrid
Estimated APC paid: $7,030
OpenAlex ID: W4409502677
