REM sleep as a dummy-model of the world: A theoretical framework.

Sleep is a complex neurophysiological process. A new integrative framework proposes that REM sleep is a state where the brain constructs a provisional dummy model of the external world. NREM sleep is a phase of synaptic and network reorganization with glymphatic clearance, updating neural circuits based on waking experience. REM sleep then provides an internal testing ground to simulate reality and refine these updates. The framework distinguishes phasic and tonic REM states and involves subcortical networks like the Papez circuit and claustrum. A central prediction is that prediction errors during REM drive selective synaptic consolidation in subsequent NREM via hippocampal sharp-wave ripple-mediated feedback. Awakening occurs when global prediction error falls below a threshold, and total sleep duration should be proportional to the complexity of novel waking experience.

Study at a glance

Abstract

Sleep remains a complex and only partially understood neurophysiological process, with current theories often fragmented across electrophysiological, metabolic, and phenomenological domains. In this paper, we propose an advanced and integrative framework that conceptualizes rapid eye movement (REM) sleep as a state in which the brain constructs a provisional dummy model of the external world. In this context, non-REM (NREM) sleep serves as a critical phase of synaptic and network reorganization, coupled with glymphatic clearance, during which neural circuits undergo structural updates in response to prior waking experience. REM sleep subsequently provides an internally generated testing ground, wherein the brain simulates reality - including the consequences of internally generated motor commands - to evaluate and refine these updates through iterative cycles. This framework further distinguishes between phasic and tonic REM states, highlighting the involvement of subcortical networks, including the Papez circuit and the claustrum, in orchestrating state transitions. The model incorporates cerebrovascular, metabolic, and glymphatic dynamics, thereby placing sleep within a multiscale systems conceptual framework. A central prediction is that prediction errors generated during REM simulation drive selective synaptic consolidation in the subsequent NREM episode via hippocampal sharp-wave ripple-mediated feedback. A further prediction is that awakening occurs when the global prediction error across the dummy-model network falls below a biological threshold, and that total sleep duration should be proportional to the complexity of novel experience acquired during prior wakefulness. By aligning electrophysiological signatures, single-unit spiking activity data, metabolic processes, and the phenomenology of dreaming, the model seeks to transcend the limitations of domain-specific theories.