AKASHICOMNI · CLAIM REVIEW TRIAL 01 · 26 SEPTEMBER 2026
One bounded claim from the Cogitate consciousness experiment, reviewed against the proposed AkashicOMNI claim record. This is an AI-assisted editorial draft awaiting independent human review.
THE EXACT CLAIM
In the Cogitate visual-perception experiment, preregistered multimodal results supported some predictions of integrated information theory (IIT) and global neuronal workspace theory (GNWT), while challenging specific predictions of both proposed neural implementations.
Scoped assessment: Supported in the published research account. Wider question: Unresolved. This trial does not determine which theory of consciousness is correct or why subjective experience exists.
SOURCE REGISTER
Nature 642, 133–142. DOI: 10.1038/s41586-025-08888-1. Accessible article abstract, main text, discussion and methods inspected. Supplementary files, recordings, analysis code and linked preregistration were not independently audited.
Scientific Data 13, article 779. DOI: 10.1038/s41597-026-07377-y. Article and access notes inspected; data not downloaded or reanalysed. Its 118 usable participants belong to the original Cogitate programme, not a new independent replication.
Both publications share dependence group COGITATE-EXP1. AkashicOMNI and its candidate specification provide review rules, not extra scientific confirmation.
DISCERN · DEEPEN
Across fMRI, MEG and intracranial EEG, the 2025 study reports information about conscious visual content in several regions. Some findings matched each theory’s predictions. The absence of sustained posterior synchronization challenged a tested IIT prediction; the general absence of ignition at stimulus offset challenged a tested GNWT prediction.
A missing signal can involve measurement or sampling limits. The paper discusses limited intracranial sampling, possible residual task engagement and the restricted range of clearly visible visual experiences studied. The results test proposed biological implementations more directly than the theories’ mathematical or philosophical cores.
Other live explanations include revising a theory’s proposed neural implementation, improving sensitivity to the predicted signal or testing a different account. These are possibilities for further study, not findings established by this record.
DEEPER DIVE · THREE PREDICTIONS
Researchers from both theory camps agreed in advance on predictions and interpretation criteria. The task presented clearly visible faces, objects, letters and false fonts for 0.5, 1 or 1.5 seconds, while varying whether a category was relevant to the participant’s target-detection task. The researchers then compared three distinct predictions across brain-recording methods.
IIT: posterior regions should carry the most informative content; adding prefrontal activity need not help decode it. GNWT: prefrontal cortex should represent conscious content through global broadcasting.
Observed: face versus object information appeared in both posterior and prefrontal regions. Face orientation was much more consistently decodable posteriorly; prefrontal orientation signals were absent in two modalities and weak in MEG, with possible signal leakage. Adding prefrontal signals did not improve critical time-resolved decoding. A category can be represented without every visual detail being represented there.
IIT: posterior activity should sustain information while the image is seen. GNWT: prefrontal ignition should briefly update the workspace at image onset and offset.
Observed: some posterior responses tracked the image duration, meeting a predefined IIT criterion, but sustained orientation information was not found. Prefrontal onset responses appeared; the predicted offset ignition was generally absent. The authors treat this as a substantial challenge to that GNWT implementation.
IIT: sustained communication within posterior visual regions. GNWT: long-range communication toward prefrontal regions.
Observed: content-selective frontal-to-visual connectivity appeared, while the predicted sustained posterior synchrony did not. Sparse intracranial coverage of early visual areas limits how strongly one can interpret the missing posterior signal.
Read the primary article and figures ↗ · See its preregistered prediction summary in Extended Data Figure 7. These are paraphrases of the authors’ analyses, not a fresh statistical reanalysis.
DESIGN AND ITS LIMITS
The consortium separated theory proponents from data acquisition and analysis, preregistered the contrasts, used three recording methods, and reserved two-thirds of qualifying fMRI and MEG data for a final test after an optimization phase. The paper reports 256 participants across the three modalities, including recruited participants later excluded from particular analyses.
The main study reports 120 recruited for fMRI, 102 for MEG and 34 for intracranial EEG. Its final fMRI analysis included 108 after quality exclusions, split between optimization and final-test samples. The later shared fMRI dataset describes 118 usable records under its release criteria. These are different analysis and sharing denominators from the same study, not an extra 118 participants or an independent replication.
The experiment compared how content is represented while people see clear stimuli; it did not compare conscious with unconscious perception, study every feature of lived experience, or compute IIT’s full integrated-information quantity. The authors themselves note that some neural signals could reflect stimulus processing and that results do not yield a definitive verdict on either theory. Recording methods have different spatial and temporal limits, and the project’s published interpretation remains open to adversarial response.
AkashicOMNI inference boundary: this is a strong test of specified neural predictions. It does not establish that consciousness is produced only by the brain, that it survives death, or that any symbolic geometry describes a physical dimension of awareness.
THE THIRTEEN ENTRIES
These are the public OMNI framework’s twelve analytical perspectives and accountable synthesis. They bring different review questions to the same research; they are not thirteen independent studies.
State both tested predictions before interpreting the result.
Keep the history and philosophical meanings of the theories in view.
Attend to what the participants experienced in this visual task.
Use challenged predictions to design a more discriminating experiment.
Enable responsible reuse of shared data; infer no clinical benefit.
Consider other explanations without exempting any from testing.
Separate measurements, model predictions and metaphysical claims.
Credit the investigators and communicate results proportionately.
Record predictions and preregistration before knowing outcomes.
Distinguish published findings from materials inspected here.
Ask for new samples and prespecified follow-up tests.
Keep neural, personal and cosmic scales distinct.
Preserve the bounded result, source links and revision history.
Optional DARK extension: neural activity is observed indirectly and reported experience is interpreted within this design. Limited access to a mechanism does not establish an otherwise untested explanation.
SYNTHESISE · REVISE
A strong next test would specify each theory’s required pattern, sensitivity and pass/fail criteria before collecting a new participant sample. Reanalysis of the released fMRI data could test further predictions but would remain a same-dataset analysis.
Revise this record if fuller analyses change a reported result, an independent replication narrows it, or a reviewer finds that the claim confuses a tested neural implementation with an entire theory. Preserve the previous version and log the disagreement.
This first trial shows that the candidate record can represent source dependence, support, contrary findings, OMNI perspectives and revision conditions. It does not establish that OMNI improves accuracy over a simpler four-question review. Independent human review, contrasting trials and release approval remain outstanding. BQ001 stays unresolved; accepted canonical edges remain zero. No evidence, rights or publication gate is promoted by this page.