Santh Prasad

Most contemporary theories of consciousness treat awareness, self-model, autobiographical encoding, and reportability as a single explanandum. Contemplative traditions that posit a witness (Ātmā) treat them as separable but rarely specify the separation in testable terms. This paper presents Coupled Witness-Manifold Theory (CWM), which formalises the distinction and makes it empirically constrainable. CWM adopts a dual-aspect substrate and defines an invariant, contentless subject-pole W — the fixed point remaining when content C, self-model S, memory-encoding M, and access/report A are subtracted. Neural dynamics are modelled not as producing W but as a coupling operator K that gates how much experiential structure is manifest. Global states — waking, NREM3, propofol, ketamine, psilocybin, meditative cessation — are represented as distinct configurations of K rather than as positions on a single unconsciousness axis. The heterogeneity of anaesthesia is treated as a constraint to be fitted, not an anomaly. The theory's load-bearing move is a non-circular formulation of masking. The traditional rescue claim — "awareness was present in dreamless sleep; you simply failed to encode it" — is unfalsifiable when the only evidence is report. CWM reformulates it as a dissociation between report ρ and an independent, report-free marker ψ (perturbational complexity, PCI): masking is the empirically checkable condition ρ = 0 while ψ > ψ*. Covert consciousness in behaviourally unresponsive patients establishes that this configuration is already actual, licensing the hypothesis for other states. Six studies are specified, each with the framework's prediction, the brain-bound rival's prediction, and explicit falsifying results. The flagship (Study 1) applies TMS-EEG through drug-free nirodha cessation in expert meditators: CWM permits in-range PCI across a reported phenomenal gap; standard accounts predict complexity collapses with report. Study 2 replicates this via serial awakening in NREM3, breaking the recall-failure-versus-true-absence circularity that report-based designs cannot escape. Study 3 tests whether a minimal-awareness index survives high-dose psilocybin after agency, ownership, and default-mode integrity have collapsed (S⊘, W≈). Study 4 extends no-report paradigms following the 2025 Cogitate result, in which conscious content was decodable posteriorly with prefrontal ignition frequently absent — precisely the awareness/access dissociation the framework requires. If ψ tracks ρ in every reported gap, the masking claim fails and CWM's core is refuted. Scope limits are stated rather than blurred. CWM makes no claim that W survives permanent substrate dissolution; the "presence-plane" is demoted to the state-space of possible K-configurations, not a place. Survival is marked explicitly undecidable, not affirmed. The end-of-life evidence is treated honestly: the dying-brain gamma surge appeared in two of four patients with epilepsy confounds and no reports, and AWARE-II's veridical-perception test was essentially null. The contribution is not that Ātmā is real. It is a precise specification of what would have to be true, which measurements would decide it, and which result would end the programme.

Santh Prasad is an independent researcher and machine learning infrastructure engineer. His curiosity runs in two directions at once: outward into large-scale multi-agent systems, and inward into what remains of a subject when self-model and memory are subtracted. He is the author of Coupled Witness-Manifold Theory and the architect of AIOS and CSES (Cognitive State Emulation System).