Source
Cogitate Consortium, Ferrante, Gorska-Klimowska, Henin et al. (2025), Nature 642:133–142, LIT-973, read in NOTE-812: “Decoding of conscious content”, “Maintenance of conscious content”, Discussion, Methods, Extended Data Figs. 3–4 and 7.
What was actually shown
The design. Suprathreshold faces, objects, letters and false fonts, each in 20 identities and three orientations, shown for 0.5, 1.0 or 1.5 s. Participants searched for rare targets. Analyses used task-irrelevant stimuli, and orientation and duration were never task-relevant. iEEG (29–31 patients; 576–655 prefrontal and 583–657 posterior electrodes), MEG (n = 65) and fMRI (n = 73) on preregistered, theory-defined regions.
Category reaches prefrontal cortex. Face against object generalised across tasks in prefrontal ROIs: about 70% in iEEG for about 0.2–0.4 s, and about 60% in fMRI, in middle and inferior frontal cortex. Posterior cortex reached over 95% (iEEG) and about 75% (fMRI).
Orientation and identity do not.
- Orientation. Face orientation was decoded in posterior cortex in all three modalities. It was not decoded in prefrontal cortex in iEEG (BF01 = 5.11–8.65) or fMRI (34–55% of prefrontal voxels BF01 ≥ 3, 1–9% for the alternative). MEG gave a weak 35% against 33% chance, with possible leakage.
- Identity. Cross-temporal RSA found sustained object identity in posterior cortex and “identity information was absent for all categories” in prefrontal cortex.
- No added content. Adding prefrontal electrodes to posterior ones did not improve decoding of category or orientation (iEEG BF01 up to 1.91 × 10⁵).
No offset response. GNWT’s proponents predicted content-specific ignition 0.3–0.5 s after onset and after offset. Of 655 prefrontal electrodes, 99 responded at onset non-selectively and 24 selectively. “None … measured the temporal profile predicted by GNWT”, with BF01 > 3 for each. The same models found an onset–offset profile in 10 early visual electrodes. RSA found prefrontal category information at onset and not at offset, even when the category was task-relevant.
What could have come out otherwise: prefrontal cortex could have held orientation and identity alongside category, as it would if it broadcast the whole of a conscious content. It could have responded when the image went off, as it would if every change of conscious content ignited it. Both were preregistered as possible outcomes.
What this does not say
- It does not say the missing dimensions were unaccessed or unconscious. No one reported identity or orientation, and the paper assumes they were experienced because the stimulus had them. The test is of representation in a region, not of access by a consumer.
- It does not say prefrontal cortex lacks them in every signal. The signals were high gamma, MEG source power and BOLD. Single units, low-frequency activity and activity-silent states were not tested, and the authors note the lack of single units.
- It does not refute the global workspace. It challenges GNWT’s proposed implementation, prefrontal cortex as the hub. The authors note a theory can keep its computational core and revise the implementation. The paper’s question is whether prefrontal cortex broadcasts “all conscious content … or only a subset”.
- It does not say category coding in prefrontal cortex is conscious content. Category may stay partly task-related. The authors “cannot entirely rule out residual task engagement with respect to category”.
- The offset null is for content-specific onset-and-offset activity in these regions and bands. One inferior frontal sulcus electrode showed the pattern earlier than predicted, and MEG alpha results were inconclusive.
Connections
- QUESTION-061 asks whether a conscious content reaches every consumer at once or only some. This bears on it, as a fact about representation: for clearly seen images, the hub region receives the category and not the identity or orientation that posterior cortex holds. It does not meet the question’s own bar, a content shown absent from one consumer, with differential post-perceptual cognition removed.
- THEORY-096 states that ignition “amplifies, sustains and broadcasts that one content to the specialised processors”. This result counts against “that one content”, read as the whole content, and against ignition at every change of content. Its promote_when names the COGITATE results as needed.
- THEORY-102 (Vishne et al. 2023, LIT-644) found prefrontal content “a transient event at onset” for category. This replicates that in an independent, larger cohort, which the paper cites as agreeing. It extends it: identity and orientation never reach prefrontal cortex at all.
- THEORY-296, filed from the same reading, holds the IIT-side failure.