Constraint-Dependent Perceptual Resolution

The COGITATE adversarial collaboration tested Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT) using multimodal neuroimaging data and found that key predictions of both theories were challenged. The consortium then released the dataset for independent reanalysis and called for new quantitative frameworks.

This paper tests whether the temporal dynamics of perceptual resolution are modulated by the observer’s task-dependent constraint state before content becomes reportable. Drawing on a pre-specified constraint-architecture model of perception (Jones, 2026), we derived three falsifiable predictions and tested them in the open COGITATE intracranial EEG dataset (34 analyzable patients).

Prediction A was supported: task-relevant non-target stimuli showed later high-gamma onset than the same stimulus category when task-irrelevant (25/34 patients in the predicted direction; Wilcoxon on subject means, p = 0.007; subject-level median +14.1 ms; bootstrap 95% CI on mean [+3.6, +19.8] ms). The effect was consistent across all four stimulus categories (faces, objects, letters, false fonts). Prediction B, concerning hysteresis across miniblock transitions, was not supported (p = 0.748). Prediction C, concerning task modulation of duration-tracking, showed a comparison-level effect but no subject-level effect, leaving the result inconclusive.

These findings identify a measurable task-dependent modulation of perceptual resolution timing in sensory cortical responses. More narrowly, they suggest that existing theories centered on integration or global broadcasting do not by themselves capture all of the temporal structure present in the COGITATE dataset. The results support further study of constraint-dependent temporal dynamics as a testable dimension in models of conscious perception.

**Keywords:** perceptual resolution, constraint architecture, task relevance, intracranial EEG, high-gamma activity, COGITATE, temporal dynamics, predictive processing