Source
- Mashour, Roelfsema, Changeux & Dehaene (2020), Conscious Processing and the Global Neuronal Workspace Hypothesis, LIT-631, read in full in NOTE-510: Introduction, Central Tenets, Empirical Evidence, The Role of Recurrent Activity, The GNW and Other Theories, Table 1 and Figures 1–2. The PMC manuscript has no page numbers, so sections are cited by heading.
What was actually shown
The mechanism, as the review states it. “In the conscious state, a non-linear network ignition associated with recurrent processing amplifies and sustains a neural representation, allowing the corresponding information to be globally accessed by local processors” (LIT-631, abstract). Besides the specialised processors there is a second space of “GNW neurons” with long-range reciprocal axons, mainly large layer II/III pyramidal cells with a layer V contribution, forming a “high-efficiency cortical core” of prefrontal, parietal, cingulate and temporal hubs. Ignition is “the sudden, coherent, and exclusive activation of a subset of workspace neurons coding for the current conscious content, with the remainder of the workspace neurons being inhibited” (Central Tenets). The theory is “not a localizationist theory”: prefrontal cortex is a key hub because of its density of broadcasting neurons, but “not … the exclusive territory for conscious access”, and in tasks that need fine detail V1 neurons “become part of the GNW”.
The evidence for ignition. Across vision, audition and touch, the first ~200 ms can be preserved on unseen trials, and conscious perception correlates with later events, the P300 and late prefrontal and parietal activity; subliminal stimuli give a decaying wave instead (Empirical Evidence). The strongest single result is van Vugt et al. (2018): monkeys reported weak stimuli by saccade, and “both perceived and unperceived stimuli caused activity in V1 and V4, but only the perceived ones elicited a sudden, strong, and sustained activity in the frontal cortex akin to ignition”, with spontaneous prefrontal ignitions on false-alarm trials. “The SDT threshold corresponds to the threshold for ignition in the PFC”, and sensitivity to how well the signal propagates there. In mice, optogenetically silencing feedback from secondary motor to somatosensory cortex attenuates late amplification and prevents perception (Manita et al. 2015), one of the few causal results. Spiking simulations give either self-amplifying ignition or a decaying wave for the same input.
The step from access to experience. The review’s explanandum is access: “the term ‘consciousness’ in this review will be replaced by conscious access”. It takes the step to experience by adopting Dehaene and Naccache’s claim that “global availability of information (…) is what we subjectively experience as a conscious state” (Introduction), and the theory “brings into question the very distinction between phenomenal and access consciousness” (The GNW and Other Theories). Of the proposal that recurrence within sensory cortex suffices for phenomenal consciousness, it says that “short of an experimental method for producing conscious experience in the absence of conscious access, this proposal remains untested”. This account states the theory with that premise, because without it the theory is about access alone and has no rivals in the record. The premise is adopted, not shown: the evidence above is evidence about what reported or accessed contents have that unreported ones lack.
What could have come out otherwise: missed stimuli could have failed in V1 or V4 rather than en route to prefrontal cortex, or reported and missed stimuli could have differed by degree in prefrontal cortex rather than by an all-or-none event that also occurs on false alarms. In van Vugt et al.’s data, as reported, neither happened.
What this does not say
- It does not show that unignited contents are unexperienced. That is the adopted premise. The review’s data are contrasts between reported and unreported stimuli, and contrasts in level of consciousness.
- It does not settle the P3. An earlier negativity (the VAN, ~260 ms) survives when stimuli are task-irrelevant yet reported seen, without a P3. Whether such stimuli are “truly seen as opposed to being merely potentially visible but unattended” is called “(unresolved)”.
- It does not say prefrontal cortex is the seat. It is one hub of a distributed core, and the review places some reverberant loops through the thalamus and cerebellar nuclei.
- Its evidence is second-hand and largely correlational. It is a proponents’ review with no new data, and the authors say “it is difficult to adjudicate experimentally among these four related theories”; anaesthesia, for example, is consistent with all of them.
- It is not a claim about why access is felt. It identifies experience with availability and does not argue for the identification.
Rivals
Overflow (THEORY-103). That account holds, from partial-report studies as Block reports them (LIT-646, LIT-640), that subjects experience more items than the roughly four that working memory holds: the uncued items are experienced though unaccessed. This account says a content is experienced only when it ignites the workspace, and ignition is exclusive, so the uncued items, never broadcast, were not experienced: at most they were, in the review’s phrase about the Pitts stimuli, “merely potentially visible but unattended”. That is the refrigerator-light reading Block argues against. Both accounts accept the same capacity data and cannot both be right about whether the surplus is experienced. Neither yet corrects the other: the review does not discuss Sperling or overflow, and calls the alternative untested, while Block’s case for overflow is an inference to the best explanation.
A sensory, not prefrontal, basis (THEORY-098). That account holds
that the neural basis of perceptual experience does not include the
prefrontal machinery of access, and is recurrent activity in sensory
cortex; prefrontal involvement is causal at most. This account makes the
ignition of a long-range core with prefrontal hubs what makes a content
conscious. The two disagree on the same phenomenon and the same
evidence. The review cites Panagiotaropoulos et al. (2012), whose monkeys
passively watched binocular rivalry while prefrontal neurons encoded the
current percept, as “not confounded by working memory or effects related
to reporting”. Block (2019), LIT-637, a source of that account,
argues that this is just what a no-report design cannot show, since a
monkey not asked to report still thinks about what it sees, and he offers
inconspicuous rivalry, with no prefrontal difference by fMRI, as the
better test. The review was submitted before Block’s paper and does not
answer it. Each would be settled by the first kind of result in
promote_when.
How it bears on the record
- Report confounds (THEORY-094). That account holds that neither report nor no-report paradigms isolate the basis of experience, because thought about the percept persists without report. It applies to this account’s best prefrontal evidence: van Vugt et al.’s monkeys reported, and Panagiotaropoulos et al.’s did not report but were not shown to have stopped thinking about the percept. So the review’s claim that the rivalry data are free of confounds is exactly what that account denies. It does not say ignition is not the basis; it says the evidence offered cannot show it. No relation is declared, since that account is methodological.
- Vishne et al. (THEORY-102, LIT-644). Their intracranial data are consistent with the account but not support for it. Prefrontal and parietal content appeared as a burst at about 150–600 ms after a clearly seen image came on, later than in visual cortex, which is the onset ignition the workspace’s proponents had predicted for that paradigm. But the predicted ignition at offset was not seen, prefrontal content did not last while the image stayed on and stayed seen, and a stable code for the image’s content was carried for its whole duration in ventral temporal cortex. Every image was seen, so experience and access never came apart, and the met predictions were IIT’s as much as this account’s. The onset-only burst fits the review’s own statement that sustained prefrontal activity accompanies maintenance for a later response, and it sits uneasily with “sustains” if the experience of a steady image is continuous. That is why it is not a source.
- Homeostatic feeling (THEORY-093, filed alongside) declares this account a rival: it puts what makes a content conscious in its identification as the organism’s own by interoceptive feeling, and loss of consciousness in the brainstem, not in a cortical core.
- AI consciousness (THEORY-023). The review states the workspace as a computational system and closes hoping it can be implemented in a device, which is the functionalist premise that account says architectural indicators presuppose. Butlin et al.’s workspace indicators (LIT-056) abstract from this mechanism.
- IIT (LIT-445) is the review’s other named rival, a posterior “hot zone” sufficient for experience with prefrontal cortex “not essential”; no theory in the record states it, so no relation is declared.
Note of 2026-10-12: “sustains” against its proponents’ own prediction
THEORY-102 reports that the workspace’s proponents pre-registered transient prefrontal onset and offset representations with no sustained component for the COGITATE paradigm. Yet this account’s statement says ignition “amplifies, sustains and broadcasts” the content. The body above already says the onset-only burst “sits uneasily with ‘sustains’”. The tension is between this account’s wording and the theory’s own prediction, not between this account and THEORY-102, which reports the GNW predictions as met except at offset. Whether “sustains” should read as sustained for maintenance and report, not for the duration of the experience, is a question for a reading of LIT-631. It is not settled here, and the statement is unchanged. No relation is declared, since THEORY-102 is a finding that bears on two theories, not a rival account.
Note of 2026-10-14: the COGITATE results, read
COGITATE (LIT-973, read in NOTE-812; THEORY-245) counts against two clauses of this account as stated. Against “broadcasts that one content”: prefrontal cortex held the category of a clearly seen image but not its identity, and orientation at most weakly. Against ignition at a change of content: none of 655 prefrontal electrodes responded at offset, though the experience changed there. It also confirms the “sustains” tension of 2026-10-12: no sustained prefrontal content was found. None of this tests ignition without differential post-perceptual cognition, the result the promote_when asks for first, since every stimulus was seen. The promote_when is updated to say so. The status is unchanged.