Consciousness at the Edge of Death: Dark Retreat, Near-Death Experience, and Generative Brain Dynamics

Abstract

Near-death experiences (NDEs) and advanced contemplative practices such as dark retreat are frequently reported to involve vivid inner imagery, encounters with non-ordinary agents, moral insight, and panoramic life reviews. Despite growing empirical interest, these experiences remain theoretically fragmented across neuroscience, psychology, and religious studies. In this paper, we develop a unified computational account of NDEs, dark-retreat phenomenology, and Tibetan Buddhist bardo (“between two, gap” often glossed as “transitional process”) models using the active inference framework.We argue that these experiences arise from lawful generative dynamics that emerge when sensory input is radically attenuated, action relevance collapses, and autobiographical self-models lose precision. Under these conditions, the brain enters a regime of offline inference in which deep priors governing identity, value, and meaning become available for revision. In extended darkness, previously unconscious processes may be brought into the light of consciousness, allowing the practitioner to relate to them rather than from them. Life review is interpreted not as episodic memory replay, but as hierarchical compression and reweighting of autobiographical information under conditions of severe constraint. Archetypal and deity-like figures are interpreted as agentive generative hypotheses that stabilize cognition when narrative selfhood can no longer organize experience, helping to explain their cross-cultural recurrence and exceptional salience.Dark retreat is examined as a trainable analogue of the dying process (a form of bardo yoga) that allows these dynamics to unfold gradually rather than catastrophically, on the practitioner’s terms rather than on death’s involuntary and non-negotiable terms. Integrating phenomenological reports, contemplative theory, and predictive-processing models, we show how moral salience, karmic imagery, and self-transcendent insight can emerge without invoking metaphysical assumptions. Taken together, this framework reframes experiences at the edge of death as principled reorganizations of generative self-models, offering a tractable scientific account of meaning-making under maximal uncertainty and self-evidencing across increasingly refined levels of self-modeling.