What would break the theory

When the recursive meta-governance layer is disrupted, metacognitive accuracy (meta-d′) should degrade before basic perceptual sensitivity (d′) collapses. The system can still see — but it loses the capacity to know what it sees.

TMS or pharmacological disruption of dorsolateral and frontopolar prefrontal cortex during a perceptual decision task. Measure d′ (perceptual sensitivity) and meta-d′ (metacognitive accuracy) across increasing disruption levels.

Two divergent curves. Perceptual sensitivity (d′) holds nearly flat as disruption increases. Metacognitive accuracy (meta-d′) drops earlier and steeper. The dissociation is the signature.

If d′ and meta-d′ degrade at the same rate, or if meta-d′ remains intact while d′ collapses, the bounded-recursion hypothesis is in trouble. The claim is that meta-governance is a separable, prior layer — not just a downstream consequence of perception.

Disruption of temporal integration (via ketamine, certain seizure types, or experimental binding manipulations) should produce subjective fragmentation of the lived now — discrete episodic slices rather than continuous perspective — while moment-to-moment perception remains intact.

Pharmacological challenge with sub-anesthetic ketamine combined with first-person reports and continuous-flash suppression timing tasks. Measure subjective continuity of experience alongside objective temporal-binding window width.

Temporal binding window widens or fragments. Subjective reports show 'time slices' or discontinuity. Critically: this should occur even when single-moment perceptual accuracy is preserved. Continuity collapses without perception collapsing.

If temporal-binding disruption produces no subjective discontinuity — or if subjective continuity is preserved purely by intact moment-to-moment perception — RTC's claim that perspective requires diachronic reconstitution is weakened.

Emotionally and biologically salient stimuli should preferentially enter recursive stabilization — they become the contents that persist as conscious distinctions — independent of objective stimulus strength. Salience does not just amplify; it gates what the recursive architecture takes up.

Continuous-flash suppression or attentional-blink paradigm with stimuli matched on physical properties but varying in personal/emotional salience. Measure breakthrough rate, dwell time in conscious access, and post-experiment recall.

Salience-weighted stimuli cross conscious access threshold faster and persist longer, with the gap widening at low signal strength. The interaction term — salience × signal strength — is where RTC predicts the strongest effect.

If breakthrough and dwell are determined entirely by physical signal strength with no salience modulation, or if salience effects are uniform rather than gating, the claim that salience structures recursive uptake fails.

Systems with sophisticated signal processing but no recursive self-in-world model will fail behavioral and neural markers of perspective — even when their task performance is high. This is the predictive bridge from human consciousness to AI evaluation.

Battery of perspective-dependent tasks (rubber-hand illusion analogues, perspective-taking, source-monitoring, self-other distinction) administered to candidate systems: large language models, agentic models, memory-augmented systems, and embodied robotic agents with explicit self-world models. Compare against human baselines and against task-only performance metrics.

Strong dissociation between raw task performance and perspective-dependent measures. Systems without explicit self-world recursion fail perspective markers regardless of how well they perform on the underlying tasks. Performance and perspective decouple.

If systems without self-world recursion pass perspective markers, or if perspective markers reduce to task performance, the architectural claim that self-in-world modeling is necessary for perspective is undermined.

Conscious perspective requires recursion to operate within bounded ranges. Too little recursive depth produces reactivity without reflection. Too much recursion without governance produces unstable self-reference. Stability lives in a middle band — and this should be measurable as a U-shaped relationship between recursive depth and perspective integrity.

Cross-state and cross-population study: meditators (variable recursive depth), psychedelic states (high recursion), dissociative states (disrupted governance), depressive rumination (recursion without bounded termination), and waking baseline. Measure recursive depth (via prefrontal-DMN coupling and neurophenomenological report), governance markers, and perspective stability.

An inverted-U curve. Perspective stability peaks at intermediate recursive depth with intact governance. It falls off on the low end (reactive) and the high end (over-recursive). The signature is the curvature itself — not a monotonic relationship.

If perspective stability is monotonic with recursive depth (more is always better, or less is always better), the bounded-recursion claim — arguably RTC's central architectural commitment — is wrong. This is the prediction RTC most stakes its identity on.