Testability & Evidence β logical structure, neuroscience methods, and what the data says
A systematic analysis of theories of consciousness β what gives rise to subjective experience β assessed along three dimensions:
The first large-scale, preregistered adversarial collaboration directly comparing Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT) β the two leading empirically-testable theories of consciousness.
Cogitate Consortium, Ferrante et al. (2025). Nature, 642, 133β142. DOI: 10.1038/s41586-025-08888-1
Our guiding principle: theories that make no testable predictions are not wrong β they are not in the game. We catalogue them, but our energy goes where evidence can reach.
Each theory is scored on a formal 5-dimension rubric assessing whether its conclusions can be tested with current neuroscience methods. Testability is about what could be tested, not what has been tested β a famous theory with many papers is not necessarily testable if its core constructs cannot be measured.
Click any theory to see its argument structure, scope line, and evidence assessment.
| Theory | Proponent | Category | Testability | Score |
|---|
This tab compiles the actual experimental and observational evidence that has been brought to bear on testable theories of consciousness.
Each theory's premises are tested against evidence from independent sources β papers that test the underlying physics, pharmacology, or neuroscience without necessarily referencing the consciousness theory. This avoids publication bias: nobody publishes "Orch-OR failed," but physicists routinely test quantum decoherence in biological systems.
π’ Supports π‘ Supports with caveats π΄ Refutes π Mixed βͺ Untestable π΅ Theoretical prediction
Before asking what creates consciousness, we need to ask what consciousness is. Theories of consciousness implicitly or explicitly adopt a definition. If we don't make that definition explicit, we risk talking past each other β and we miss a powerful form of critique: the overgeneration test.
Each theory's argument is reconstructed in Fitch-style natural deduction β the notation used in formal logic textbooks. Vertical scope bars on the left show which assumptions are active at each step. Nested bars mark sub-derivations that depend on additional auxiliary assumptions. Conclusions are written in abstract variables, with a symbol legend below each proof.
This makes visible which assumptions each conclusion depends on β and therefore which assumptions, if wrong, would collapse the argument. The scope line is the backbone of our testability assessment: we score each conclusion based on whether it can be tested with current neuroscience methods.
Each conclusion in a theory's scope line is scored on five dimensions (0β1). The theory-level score is a weighted mean (core conclusions weighted 2Γ, derived conclusions 1Γ). Citation count and publication volume do NOT enter the score β a famous theory with many papers is not necessarily testable if its core constructs cannot be measured.
| Dimension | Question | Score 1.0 | Score 0.0 |
|---|---|---|---|
| D1. Specificity | Does the conclusion predict something about a brain area, circuit, or measurable quantity? | Precise quantitative prediction | No empirical prediction |
| D2. Measurability | Can current neuroscience methods measure the predicted quantity? | Routinely measured (fMRI, EEG) | Not measurable with any existing tech |
| D3. Perturbability | Can we causally intervene on the mechanism (TMS, pharmacology, lesions)? | Selective perturbation in humans | Cannot perturb the core mechanism |
| D4. Falsifiability | Is there a clear disconfirming outcome, or can the theory absorb any result? | Clear falsification criterion | Any outcome can be reinterpreted |
| D5. Discriminability | Does the prediction differ from rival theories? | Unique to this theory | Shared by all theories |
Per-conclusion testability = mean(D1, D2, D3, D4, D5). Theory-level = weighted mean across conclusions (core 2Γ, derived 1Γ).
The rubric is grounded in what current neuroscience can actually do:
TMS Β· tACS/tDCS Β· Pharmacological (anesthesia, psychedelics) Β· Lesion studies Β· DBS Β· Optogenetics (animal)
fMRI Β· EEG Β· MEG Β· ECoG/iEEG Β· Single-unit Β· Calcium imaging (animal) Β· PET Β· fNIRS
Sleep stages Β· Anesthesia dose-response Β· Disorders of consciousness (coma, VS, MCS) Β· Psychedelic states Β· Dreaming
Binocular rivalry Β· Visual masking Β· Attentional blink Β· No-report paradigms Β· Change blindness
Every citation in this dashboard has been verified through Crossref, OpenAlex, and DataCite. Citations that could not be verified are flagged. No citation is presented as verified without checking.