Does near-criticality set the level of consciousness? An architectural prediction, and how to falsify it

Abstract

Proximity to a critical point is increasingly treated as a marker of the conscious brain: cortical dynamics sit closer to criticality in wakefulness than under anaesthesia, in seizures or in disorders of consciousness. Criticality characterises the regime of the substrate, however, and two systems poised at the same statistical edge can differ completely in what they compute. This Viewpoint argues that the level of consciousness is set by an architectural property, the organisation of the system's models of the world and of itself, and that this claim makes a prediction a single global dynamical variable cannot make. If the fidelity of the explicit world-model is indexed by first-order perceptual sensitivity (d′) and the fidelity of the explicit self-model by metacognitive sensitivity (meta-d′), a manipulation that perturbs the self-model while preserving the substrate should move the two orthogonally, whereas a scalar account, on which that variable sets the fidelity of every explicit model, predicts that they covary. The contrast separates architectural from scalar accounts and leaves higher-order theories as a near neighbour. Prefrontal lesions, prefrontal stimulation and sub-anaesthetic ketamine have each lowered meta-d′ without a detectable change in first-order performance, which supplies the building block; none of these designs let d′ vary freely, so orthogonality remains untested. The decisive test lets d′ vary within subjects, and covariation across such manipulations would refute the account.