Multivariate analysis of resting state activity and visually evoked response to natural stimuli to map the representation of information in intrinsic brain activity

This thesis investigates patterns in spontaneous brain activity within the framework of active inference, which posits that sentient behavior results from probabilistic inference guided by an internal model of the world. These internal models are built from statistical regularities in the environment and serve as "priors" that drive perception and behavior. The hypothesis tested is that these predictive coding priors are stored within the brain's spontaneous activity—activity that occurs independently of external stimuli and has been a growing focus in neuroscience. Chapter 1 introduces the idea that spontaneous brain activity serves as a generative mechanism for storing prior information critical for both online behavior and offline knowledge consolidation. Chapter 2 presents a cross-species review, showing that spontaneous activity mirrors task-evoked activity for natural stimuli, states, and behaviors across various brain regions, suggesting it encodes behaviorally relevant priors. Maintaining these representations is metabolically expensive and it serves in anticipating future needs. Chapter 3 tackles the challenge of identifying universal mechanisms of spontaneous activity across species by designing a human experiment that incorporates findings from human neuroimaging and animal electrophysiology. This chapter introduces a sequential learning paradigm using natural images, demonstrating that sequences can be learned implicitly, speeding up reaction times without explicit control. In Chapter 4, high-density electrophysiology is used to decode visual, motor, and sequence patterns during learning. Sequence information is encoded across brain regions like the visual and parietal cortex, and the hippocampus, with sequence representations emerging after visual category information is learned. In Chapter 5 a novel statistical method is used to detect reactivations during resting states, revealing that natural stimuli are more robustly reactivated than non-natural stimuli. Reactivations strengthen during learning and vary with behavioral performance. Finally, Chapter 6 explores replays of sequences during rest, showing that high-performing participants exhibit forward replays localized to visual and parietal regions occurring within high frequency ‘ripples’, while lower-performing participants show off-line reverse replays that occur within broader power modulation. The thesis concludes by highlighting the adaptive function of spontaneous brain activity, showing its role in encoding sequences and its relevance for prediction, learning, and behavior. It integrates insights from both human and animal studies, shedding light on how the brain uses spontaneous activity to interpret and respond to the world.