One Theory, All the Phenomena
Every major consciousness phenomenon — anesthesia, psychedelics, split-brain, blindsight, and more — mapped against the Five leading theories. Only one covers them all.
The field of consciousness science has a fragmentation problem. Integrated Information Theory explains integration. Global Workspace Theory explains access. Predictive Processing explains perception. Higher-Order Thought theories explain metacognition. Each handles its home turf — and struggles everywhere else.
The Four-Model Theory proposes a single architecture that accounts for all of them.
Four model kinds (implicit/explicit × world/self) under self-referential closure, with free modelling. Consciousness is what the closed loop produces. Disrupt any component, and phenomenology changes exactly as neuroscience observes.
If you’re new to FMT, start with the introduction. This page is the evidence map.
The comparison
Eleven consciousness phenomena. Five theories. One question: which framework covers the landscape?
| Phenomenon | FMT | IIT | GWT | PP | HOT |
|---|---|---|---|---|---|
| Anesthesia (propofol) | ✓ | ~ | ~ | ~ | ~ |
| Language under narcosis | ✓ | ✗ | ~ | ✓ | ~ |
| Sleep / wake criticality | ✓ | ~ | ~ | ~ | ✗ |
| Psychedelics / ego dissolution | ✓ | ✗ | ✗ | ~ | ✗ |
| Split-brain | ✓ | ~ | ✗ | ✗ | ~ |
| Anosognosia | ✓ | ~ | ~ | ~ | ~ |
| Blindsight | ✓ | ~ | ✓ | ~ | ✓ |
| Lucid dreaming | ✓ | ~ | ✗ | ✗ | ✗ |
| DID (dissociative identity) | ✓ | ~ | ✗ | ✗ | ~ |
| Animal consciousness | ✓ | ✗ | ✗ | ✗ | ✗ |
| Developmental emergence | ✓ | ✗ | ✗ | ✗ | ✗ |
What the architecture accounts for
Anesthesia
Propofol breaks the explicit self-model’s closure — the loop that integrates — while leaving lower-level models intact. The brain continues processing without anyone home.
This isn’t metaphor. Katlowitz et al. (Nature, 2026) showed hippocampal neurons performing grammar parsing, word prediction, and semantic discrimination under full anesthesia, with zero conscious recall. The processing didn’t stop. The integration did.
IIT predicts phi reduction, which captures the direction but not the dissociation between preserved processing and absent consciousness. GWT says the workspace goes offline — true, but why does language processing continue? FMT answers: because the world model keeps running. Only the self-model’s closure is lost.
Psychedelics and ego dissolution
Psilocybin pushes the brain toward a supercritical dynamical regime. As the dynamics move past the critical band, the explicit self-model destabilizes. Integration fragments. Ego boundaries dissolve.
This is not “more consciousness.” It’s fragmented consciousness — and the phenomenology matches precisely.
IIT faces a problem here: entropy increases under psychedelics, which should increase phi. But consciousness doesn’t expand — it breaks apart. GWT predicts workspace flooding should enhance consciousness. It doesn’t. FMT predicts exactly what happens: disrupted integration produces disrupted experience.
Split-brain
Sever the corpus callosum and two independent self-model closures become possible. Two conscious streams sharing one substrate.
For FMT, this is a prediction, not a puzzle. If consciousness requires integration across four models, and the integrator is physically divided, two independent consciousnesses follow directly. GWT struggles: one workspace, or two? PP continues predicting in both hemispheres — but which one is “conscious”?
Blindsight
V1 lesion destroys visual consciousness while leaving visual behavior intact. The patient dodges obstacles they cannot see.
FMT: visual processing continues at the world-model level without self-model integration. Behavior without experience. Both GWT (no broadcast) and HOT (no higher-order representation) handle this one — it’s the theories’ strongest shared success.
Anosognosia
Stroke damages the body model. The explicit self-model is built from a corrupted self-representation. The patient genuinely doesn’t know they’re paralyzed — they’re not in denial, they’re not lying. Their conscious experience literally doesn’t contain the deficit.
The architecture explains pathology.
Lucid dreaming
REM sleep suppresses the explicit self-model. Ordinary dreaming follows — experience without self-awareness. Then, occasionally, the self-model partially reactivates. The dreamer becomes aware within the dream.
GWT predicts no conscious awareness when the workspace is offline. HOT predicts no higher-order awareness during sleep. FMT predicts exactly the gradient we observe: partial integration yields partial awareness.
Animal consciousness
Not binary. Graded by architectural completeness. Mammals with all four models: conscious. Octopuses with distributed but integrated models: likely conscious. Insects possibly lacking an explicit self-model: a principled, testable boundary rather than a gut feeling.
Every other theory struggles here. IIT assigns phi to grid structures. GWT can’t define a workspace criterion. PP says all organisms predict. HOT asks whether animals have higher-order thoughts — an unanswerable question.
Developmental emergence
Consciousness emerges as the infant brain assembles the four-model architecture — a continuous process, not a sharp threshold. The model space is continuous; the architecture assembles gradually.
No competing theory offers a specific developmental account.
A note on honesty
The matrix above is not a claim that competing theories are useless. IIT’s insights on integration, GWT’s account of access, PP’s Bayesian framework, and HOT’s emphasis on metacognition all contribute genuine understanding. Several phenomena — anesthesia, blindsight, anosognosia — receive partial treatment from multiple theories.
The claim is narrower: no single competing framework provides mechanistic accounts across the entire landscape. FMT does, through one architecture and one dynamical requirement. That’s the test of a unifying theory.
Read the framework
- Full preprint: Four-Model Theory of Consciousness (Zenodo, open access)
- Intelligence model: Recursive Intelligence Model (RIM) (Zenodo, open access)
- English edition: Amazon US
- German edition: Amazon DE
Matthias Gruber is an independent consciousness researcher and R&D AI Transformation Manager. The Four-Model Theory was first published in 2015.