Vocal ranks from repeat changes exceeded chance in six SEEG patients and two fMRI studies, but the fixed baseline rule failed

We tested whether directions of measured repeat changes rank vocal sounds above nonvocal sounds on sounds excluded from fitting. Separate public studies supplied human SEEG high-gamma activity and BOLD fMRI. Development used two patients and four MRI adults. A local code lock preceded new tests in six other patients, 16 other MRI adults, and eight adults from a second MRI study. Stable additive sound means cancel from two repeat contrasts. Training-only scales, spatial contrasts, unit normalization, and regularized quadratic class rules gave mean AUCs 0.52684, 0.59124, and 0.57281. Their 95% person Student intervals were [0.51231,0.54137], [0.56222,0.62026], and [0.54458,0.60105]. All 30 reserved people ranked above chance. But the full fixed rule also required a positive SEEG increase over pre-onset activity. Its interval was [-0.00035,0.04737], so the full rule failed; its equivalent gate value was 0.05233. A joint person-and-sound SEEG range crossed chance. The two electrical repeat contrasts gave mean AUCs 0.55044 and 0.50324. Pre-onset subtraction gave 0.49188 [0.47306,0.51070]. A common gain model gave positive ranks without class-specific random errors. An exact model gave AUC 0.9 with equal raw covariance in both classes. Thus angular ranks do not establish unequal raw covariance or intrinsic neural noise. This is a checked method and negative full-test benchmark. Source selection, clinical sampling, separate cohorts, acoustic differences, and unequal measurement filters limit scope. No shared stimulus-specific neural mechanism or causality was confirmed. Code, source pins and terms, all results, figures, and repeat steps are preserved.

human neuroscience · SEEG · fMRI · auditory cortex · repeat variation · vocal sounds · angular classification · negative result