The Walk · 12 of 14

THE TABLE

What would we expect if a theory were true, and what would put it at risk?

Five living theories. Equal frames. A 2025 adversarial test of two of them.

GNWT

Global Neuronal Workspace

Why do some contents become available to the whole mind, while others remain local?

Notes
Claims
A percept becomes conscious when it is broadcast in a global workspace, with a characteristic ignition involving prefrontal and parietal networks.
If true
Content-specific signatures in prefrontal cortex. Brief ignition at relevant updates, including offset in the Cogitate design. Long-range broadcasting rather than purely local maintenance.
Supports
Converging work on inattention, masking, and late all-or-none reports. Inferior frontal content information appeared in the 2025 collaboration.
Challenges
The 2025 Cogitate test found a general lack of ignition at stimulus offset and limited representation of some conscious dimensions in prefrontal cortex.
Leaves open
Whether prefrontal involvement is necessary for all conscious contents, or only for some. How silent storage between ignitions should be understood.

IIT

Integrated Information Theory

Why is there experience at all, and why does it have this structure?

Notes
Claims
Consciousness is identical with a system's integrated information. The physical substrate is expected in a posterior cortical hot zone whose cause-effect structure specifies experience.
If true
Information about conscious content in posterior cortex, maintained across stimulus duration. Connectivity within that posterior complex specifying the experience.
Supports
Posterior content information and duration-tracking responses in occipital and temporal cortex in the 2025 collaboration. Clinical interest in perturbational complexity as a related measurement program.
Challenges
The 2025 test found a lack of sustained synchronization within posterior cortex, challenging the claim that network connectivity there specifies consciousness. Some contents present in experience were not sustained as predicted.
Leaves open
How to measure integrated information in a living brain. Whether identity claims can be tested, or only anatomical predictions derived from them.

HOT

Higher-Order approaches

What is the difference between merely seeing and being aware that one is seeing?

Notes
Claims
A first-order state becomes conscious when it is the target of a suitable higher-order representation, such as a thought or perception that one is in that state.
If true
Dissociations between first-order processing and awareness. Prefrontal or higher-order involvement in whether a state is conscious, not only in reporting it.
Supports
Philosophical fit with some introspective and metacognitive data. Overlap with debates about empty higher-order states and blindsight.
Challenges
Disagreement about whether prefrontal damage abolishes experience or only report. Some 2025 challenges to prefrontal necessity also press related higher-order predictions.
Leaves open
What kind of higher-order state is required. Whether the relevant awareness is itself conscious, and how to avoid regress.

RPT

Recurrent Processing

Can there be rich experience without, or before, cognitive access?

Notes
Claims
Local recurrent processing in sensory cortex can suffice for phenomenal experience. Widespread broadcasting is needed for report and access, not for experience as such.
If true
Experience tracking posterior recurrent activity even when prefrontal ignition is absent. Dissociations between local recurrence and global access.
Supports
Arguments from masking, feedforward sweep versus recurrence, and some posterior findings that do not require a prefrontal ignition.
Challenges
Difficulty proving experience in the absence of report. Shared anatomical predictions with IIT can be challenged by the same posterior connectivity results.
Leaves open
How to study unreported experience without smuggling access back in. Whether local recurrence is sufficient or only necessary.

Predictive

Predictive and active inference

Why is experience structured, world-presenting, and bound up with action?

Notes
Claims
The brain is a prediction engine. Conscious contents are those inferred states that win as the system's current best model of the causes of its sensory data, often at particular hierarchical levels.
If true
Contents that track precision-weighted prediction error. Changes in experience when priors or sensory precision are altered, as in some illusions and psychedelic models.
Supports
Explanatory reach across perception, action, and some psychiatric phenomena. Fit with a broader family of Bayesian brain models.
Challenges
The framework is wide. It can be hard to say what would count as a failure. It is not yet a single, equally specified theory of why some inferences are experienced.
Leaves open
Which level of inference is conscious. How this family relates to workspace, recurrence, and integrated information without absorbing them by metaphor.

Hear

Five living theories. One table.

Same frame. Same questions.

What does it claim? What is it trying to explain?

What would we expect if it were true? What supports it?

What challenges it? What does it leave open?

In 2025 a theory-neutral consortium published an adversarial test of Integrated Information Theory and Global Neuronal Workspace Theory.

Human participants saw images while brains were measured with fMRI, MEG, and intracranial EEG. Some predictions of each theory found support.

Important claims of both were substantially challenged. IIT was pressed by a lack of sustained posterior synchronization.

GNWT was pressed by a lack of offset ignition and limited prefrontal representation of some contents. This is not a double funeral.

It is what it looks like when theories are designed to be made uncomfortable. Science should not only accumulate evidence that favors a preferred view.

Current theory

Equal dignity

The five families at the table are not the only theories in the literature. They are major, currently argued programs.

Expansion belongs in depth, not in a larger frame.

Contemporary empirical result

Put the theories at risk

Cogitate Consortium, Nature, 30 April 2025. Preregistered divergent predictions.

Mixed results. The honest public sentence is: both theories were challenged in important ways, and neither was simply confirmed or simply destroyed.

Sources
  1. Cogitate Consortium (Ferrante et al.). “Adversarial testing of global neuronal workspace and integrated information theories of consciousness.” Nature 642. 2025.

    Preregistered adversarial collaboration. Some predictions of both IIT and GNWT were supported; important claims of both were substantially challenged.

    Neither theory was simply disproved.

    Open source