A week in the life of the human brain reveals stable states punctuated by chaotic-like transitions.

Critical real-world neurocognitive processes, such as settling into a conversation and neurophysiological fluctuations, vary over minutes-to-days in real-world environments. We harnessed simultaneous multi-electrode intracranial and video recordings in twenty people during a week of unconstrained, spontaneous behavior. Using dynamical deep learning algorithms, we found neurodynamics linked to circadian rhythm, heart rate, and multiple aspects of behavior (socializing, watching a screen, sleep depth, etc.). Transitioning between behaviors was associated with bursts of rapid, chaotic neural exploration that stabilized into new states. Despite this chaos, large-scale dynamics anchored to a stabilizing center manifold associated with neurophysiological and conscious states, with a central attractor involving default mode network activation. When perturbed by sleep deprivation, neural transitions were more chaotic and dynamics around the central attractor were suppressed, suggesting diminished neurodynamic control due to lack of sleep. These findings highlight how the brain chaotically transitions around a stabilizing equilibrium to balance dynamic exploration and stable equilibria during real-world behavior.