Fixed sound tests did not confirm shared repeat alignment in six SEEG patients and 24 fMRI adults; mean removal raised the score in a sparse model

We tested alignment between a sound's two-repeat difference and its response on a third repeat. Separate public human studies supplied SEEG high-gamma amplitude and two BOLD fMRI cohorts. Calibration used other sounds. The score removed diagonal site products and compared matched sounds with different sounds in the same source group. Development used two patients and four adults. A local plan and code lock preceded tests in six other patients, 16 B2021 adults and eight NH2015 adults. Both raw and training-fitted gain procedures were required. Raw mean scores were 0.002320, 0.000695 and 0.001541. Their 95% person t intervals were [-0.001554,0.006194], [-0.000741,0.002130] and [-0.001995,0.005077]. The shared test failed, p=0.636719. Five of six primary intervals included zero. All fixed controls remain. Spatial mean removal gave positive mean intervals in all three cohorts, but the B2021 majority rule failed. An exact model explains an interpretation limit. Three independent sources had six signed-axis sounds and independent continuous positive gains. With known population centering, spatial mean removal raised the diagonal-free matched-minus-wrong score from zero to 2/5. This is an unfitted model result, not a human cause. Simulations also showed small positive scores from estimated gain correction under additive errors. The study supplies a checked assay, a failed cross-modal test and exact processing controls. Clinical sampling, source selection, shared schedules, separate cohorts and unequal sensing volumes limit neural claims. Code, source pins and terms, all results, figures and repeat steps are supplied. Original neural arrays remain external.

human neuroscience · SEEG · fMRI · auditory cortex · repeat variability · spatial mean removal · measurement model · negative result · independent cohorts