Gain-controlled reconfiguration of long-range coupling explains visibility-dependent spatiotemporal neural coding dynamics
Numerous studies demonstrate the widespread cortical engagement in conscious and unconscious processing (1-5), yet the mechanistic principle that reorganizes spatiotemporal neural dynamics across different levels of conscious access remains unclear. We addressed this using a personalized, connectome-constrained large-scale biophysical model of whole-brain source-imaged electroencephalography (EEG) dynamics recorded from participants who reported the orientation of a backward-masked Gabor patch at four levels of visibility. The key control parameter in the model is the global input gain ({gamma}) to physiologically-grounded cell populations, defined operationally such that higher {gamma} denotes weaker effective global input, which reconfigures long-range couplings and thus modulates spatiotemporal neural dynamics. Fitted {gamma} maps showed higher gain in posterior regions and lower gain in frontal regions on visible trials, whereas the pattern was opposite on invisible trials - i.e., posterior regions received weaker and frontal regions stronger input under visibility, with the opposite under invisibility. This inversion of {gamma} map was observed in pyramidal, VIP, and PV populations, while supplementary pyramidal, SST, and PV time constant were invariant. The visibility-dependent reversal in gain topography mirrored the observed spatiotemporal neural coding dynamics - under visibility, early occipital-to-frontal dynamic coding switched to stable coding, whereas in invisible condition, frontal-dominant stable coding emerged early with attenuated dynamic coding. Simulations generated from the fitted parameters reproduced the observed spatiotemporal neural coding dynamics, supporting a gain-control axis - with pyramidal input efficacy as the principal readout and coordinated VIP/PV modulation - that selects between locally integrating and globally broadcasting regimes as visibility changes. Our results provide a bottom-up mechanistic account of differential perceptual processing depending on the degree of the stimulus awareness, offering insights into potentially resolving ongoing debates about theories of consciousness.