nucleation

LIT-973: Adversarial testing of global neuronal workspace and integrated information theories of consciousness

2 min read

Properties5
StatusActive — read 2026-10-14 (NOTE-812) in full; worth reading as the first preregistered adversarial test of two theories of consciousness, and for what it found: prefrontal cortex held a seen image's category but not its identity or (detectably) orientation, no prefrontal site responded at stimulus offset, and sustained content-specific gamma synchronization within posterior cortex, IIT's distinctive prediction, was absent. Read it knowing that every image was clearly seen, so nothing separates experience from stimulus processing or access, that it tests each theory's proposed implementation rather than its core, and that the proponents' replies are in the Supplementary Information, not read here. THEORY-245 and THEORY-296 hold what it shows.
Version2
Tagsconsciousness, neuroscience
Cited as “paper” byNOTE-812 — Cogitate Consortium — Adversarial testing of GNWT and IIT
Cited as “source” byTHEORY-245 — In a preregistered multimodal test with clearly seen images, prefrontal cortex carried a seen image's category but not its identity, and a face's orientation at most weakly, while posterior cortex carried category, face orientation and the identity of objects, letters and false fonts, though not of faces; no prefrontal site showed the preregistered onset-and-offset response, and one, found in an exploratory analysis, responded at offset earlier than predicted, THEORY-296 — In a preregistered multimodal test with clearly seen images, category-selective posterior cortex showed no sustained content-specific synchronization with early visual cortex, which iEEG sampled with 12 electrodes: the preregistered phase coupling was brief, low-frequency and mostly stimulus-evoked, and exploratory amplitude coupling, gamma included, was also brief

Graph View

Cogitate Consortium, Oscar Ferrante, Urszula Gorska-Klimowska, Simon Henin, Rony Hirschhorn, Aya Khalaf et al. (2025), Nature 642(8066), 133–142.

Key takeaways

From the reading, NOTE-812.

  • The design. The proponents of both theories preregistered divergent predictions and pass/fail criteria. Theory-neutral teams ran one paradigm in several laboratories: faces, objects, letters and false fonts, in three orientations and three durations, all clearly seen and mostly task-irrelevant.
  • Content. Category was decodable in posterior and prefrontal cortex. Face orientation and object identity were represented posteriorly and not detectably in prefrontal cortex in iEEG or fMRI, with Bayes factors for the null. Adding prefrontal sites never improved decoding.
  • Against GNWT. None of 655 prefrontal electrodes showed the predicted onset-and-offset ignition, and prefrontal category information appeared at onset only. The authors ask whether prefrontal cortex broadcasts “all conscious content … or only a subset”.
  • Against IIT. Content-specific synchrony between category-selective cortex and V1/V2 was early, brief, 2–25 Hz and mostly evoked, not the sustained gamma predicted. Orientation was not sustained in posterior cortex either. IIT passed its duration prediction for category.
  • What it tests. Biological implementations, not cores: “theorists could respond … by modifying the proposed biological implementation”. No verdict is given.

Standing in the record

Filed on 2026-10-12 for QUESTION-061 and QUESTION-048, and read on 2026-10-14 in NOTE-812. The filing called it “direct evidence on whether access is one all-or-none broadcast”. That was too strong. It shows that the prefrontal hub did not represent every dimension of a clearly seen image, which bears on QUESTION-061. But no dimension’s access was measured. THEORY-245 holds the prefrontal findings, and THEORY-296 the failed posterior-synchronization prediction.

It is the “COGITATE results” that THEORY-096’s and THEORY-102’s promote_when conditions name. For THEORY-102 it replicates the onset-only prefrontal pattern for category in a larger independent cohort, but finds orientation not sustained in posterior cortex. For THEORY-096 it counts against broadcast of the whole content and against ignition at every change of content. Those accounts are not edited here. For QUESTION-048 it is task-based no-report evidence with percept and stimulus never apart, so it stays within THEORY-094’s limit.

Backlinks