At fixed windows, a selective variance reduction test failed across six ECoG patients and five fMRI participants

We tested whether image viewing reduces repeat variation more along a learned face versus house direction than in other directions. Separate public human studies supplied surface ECoG power and BOLD fMRI. One person from each study supplied development. This selected spatial mean removal and a reduction prediction. A local plan and code lock preceded new tests in six other ECoG patients and five other MRI participants. Electrical residuals removed exact image means; MRI residuals removed category block means. Mean bounded reduction scores were -0.0860 and +0.3755. Their 95% person Student intervals were [-0.2426,+0.0705] and [+0.1132,+0.6377]. The shared rule required positive reduction and category reading in each study. It failed, p=0.90625. All five MRI scores were positive, but earlier and later MRI window mean intervals included zero. Category AUC was 0.9671 in ECoG and 1 in each MRI test panel. Zero-variance Student intervals for MRI AUC are flagged. Added conservative bounded intervals remain wide. All negative people and controls remain. An independent covariance calculation reproduced all eleven main records within 4.7e-14. All thirteen feature archives rebuilt exactly, and the main table and summary repeated. An exact model added after the human test changed the score sign when averaging changed from 0.3 to 5 s. It was not fitted to human data. Different cohorts, sparse clinical sampling, source selection, missing MRI image labels and temporal filters limit scope. No positive neural generalization or common cause was established. The package supplies code, provenance, terms, results, figures and repeat instructions. Original neural arrays remain external.

human neuroscience · ECoG · fMRI · visual categories · neural variability · reproducibility