A brain theory becomes more interesting when it risks being wrong. Saying that consciousness involves information or networks leaves many possibilities open. Predicting where a particular content should be represented, how long its representation should last and what should happen when it disappears gives an experiment something to test.
Neuroscience investigates the mechanisms associated with conscious conditions and particular experiences. It has competing explanations, rather than a single settled account. Two prominent approaches—global neuronal workspace theory and integrated information theory—make different claims about what matters. A published comparison exposes both the value of those claims and the limits of their support.
The first article examines what counts as explaining experience. Here the question is narrower: how do proposed brain mechanisms become predictions, and how should their results change our judgment?
A workspace makes information widely usable
The global neuronal workspace hypothesis proposes that a representation becomes consciously accessible through amplification and recurrent interaction that make it available across specialized systems. Its ignition is a network event, not a flame or a tiny observer inside the skull. The proposed network is distributed; a central role for prefrontal areas does not mean consciousness lives exclusively in one spot. The proponents explain the model in their 2020 review.
An original hypothetical example helps clarify availability. A person sees a changed departure time. They can use the information to revise a plan, tell a companion and remember what happened. The workspace proposal asks about the mechanism supporting that coordinated availability. Listing the three abilities alone does not identify the network; the biological account must connect them to operations that can be investigated.
This is also why finding a response in one region would be insufficient. The hypothesis concerns interactions and access. A local signal might reflect input to the process, part of the process or an effect of preparing an answer. The experimental task must help separate those roles.
Integration concerns a system's causal structure
Integrated information theory, or IIT, approaches the problem from properties its proponents take to be essential to experience. IIT 4.0 proposes that experience corresponds to an intrinsic, irreducible cause–effect structure: a system must make a difference to itself in a way that cannot be reduced to independent pieces. Its formal quantities depend on the defined causal system and state. They are not a generic score for the number of connections, the amount of data stored or how busy a recording looks. Read the theory's postulates and explanatory identity.
Consider an original hypothetical collection of independent devices. Each produces an elaborate record, but none influences the others. Adding their outputs produces more data. It does not demonstrate the kind of integrated causal structure the theory proposes. The example clarifies a distinction; it is no calculation of consciousness in those devices.
A claim that IIT predicts consciousness therefore needs the relevant version of the formalism, a justified system boundary and an account of how the causal model was obtained. A number labeled complexity cannot silently take the place of that work. Neither can the existence of a mathematical expression establish that its interpretation as consciousness is correct.
An actual comparison placed bets in advance
The COGITATE consortium brought proponents of the two theories together with theory-neutral researchers and specified predictions before collecting the decisive observations. Its 2025 Nature paper reports 256 participants viewing clearly visible images for varying durations, measured across functional MRI, magnetoencephalography and intracranial EEG. The comparison addressed representations of content, their persistence and interareal connectivity.
Important results challenged both accounts. The authors identify absent sustained posterior synchronization as a challenge for IIT and a general lack of ignition at stimulus offset as a challenge for the workspace account. They also report limited representation of some conscious dimensions in prefrontal cortex. The study tested proposed biological implementations, with methodological limitations; it did not directly settle every mathematical or computational commitment of either theory. The original paper gives the predictions, results and interpretations.
This is a stronger evidential situation than placing two impressive brain images beside two descriptions. The accounts faced common conditions and declared expectations. A reader can ask which observation was expected, which occurred and how much of the theory depends on the failed expectation.
Separate a failed prediction from a complete verdict
Suppose, in an original hypothetical study, a theory predicts a response when a visible object disappears. The response is absent. Three different questions now arise: was the predicted event genuinely necessary to the account, could the instrument detect it, and did the experimental situation instantiate the condition the prediction concerned?
If the study cannot detect the event even under a suitable comparison condition, the result may be inconclusive. If detection is adequate and the event was a central requirement, the failure is more damaging. A theorist can revise the proposed mechanism, but the revision needs to explain the failure and make a new commitment. Simply describing every outcome as compatible would remove the comparison's purpose.
Changing a biological implementation while retaining an underlying idea can be reasonable. What matters is keeping the change visible. An account that now places a mechanism elsewhere has a different empirical obligation from the one originally tested. New data should test it, rather than having the old data repeatedly serve as both inspiration and confirmation.
The reverse mistake is to claim a whole research program has been disproved by one missing response. That conclusion requires showing that the failure reaches a commitment the program cannot change without losing its explanatory identity. Often the defensible conclusion is more precise: this prediction, under these conditions, did not fare well.
Ask whether the signal answers the intended question
A study can find a reliable relation and still leave its meaning open. Imagine a hypothetical sound task in which every heard tone requires a button press. A neural difference between heard and unheard trials might relate to hearing, deciding, preparing the movement or remembering the instruction. The design has bundled them together.
A stronger inquiry changes reporting demands, examines what stays stable and tests alternatives. Even then, evidence from one task does not automatically establish a universal detector for every experience or every kind of system. The target might be visual perception under a particular set of conditions. Transfer to dreaming or artificial systems would require further argument and evidence.
This gives a reader a manageable way to assess a claim. Name the experienced content, the proposed mechanism and the predicted difference. Identify an alternative that could produce the observation. Then inspect the control that distinguishes them—or acknowledge that the available result cannot.
The result of this approach is a more exact account of progress. Neuroscience can constrain where and how an explanation works while the larger philosophical disagreement remains open. A theory earns trust by explaining a phenomenon and surviving discriminating tests, including results its proponents would rather not have found.
Loading comments…