Repeat variance profiles agreed on average between people in SEEG and one fMRI study; a second fMRI transfer test failed

We tested whether natural sounds have repeat variance profiles that agree between people after control for source classes and measured mean response strength. Separate public studies supplied human SEEG high-gamma activity and BOLD fMRI. A local plan fixed the endpoint before development. Fixed references used two SEEG patients and four fMRI adults. Tests in six other patients and 16 other fMRI adults gave mean adjusted correlations of 0.1513 and 0.0983. Their pointwise 95% person t intervals were [0.0969,0.2057] and [0.0585,0.1382]. The fixed shared mean, sign, and label-reference rule passed, p=0.015625. The same MRI reference failed transfer to eight adults from a second study: mean -0.0192, interval [-0.0860,0.0476]. We retain all negative results. A source audit found the same three sound orders in all eight SEEG records. A deterministic trend with presentation rank gave agreement of one with no sound effect or random error. Simple schedule controls did not remove the measured agreement. Thus the example limits interpretation but does not identify the human cause. The score concerns relative disagreement of processed estimates. Electrical task repeats and MRI scan estimates have different precision and filters. Source selection, fixed sounds, small reference groups, clinical sampling, and unequal cohorts limit scope. No intrinsic neural noise mechanism or causal result is established. The package gives code, derived summary inputs, all results, source pins and terms, figures, local plans, failure logs, and repeat instructions.

SEEG · fMRI · auditory cortex · repeat variance · measurement uncertainty · stimulus order · replication · human neuroscience