Pooled repeats raised the energy–variation rank relation at 24 sites in six SEEG patients and 16 fMRI adults

We tested whether large sound-group response contrasts have large repeat variation. Separate public human studies supplied SEEG high-gamma amplitude and BOLD fMRI. Each person supplied 24 selected sites and three response estimates. One repeat set contrast energy; the other two set variation. We also tested a pooled rule that reused all repeats in both quantities. Development used two patients and four MRI adults. A local code lock preceded tests in six other patients and 16 other adults. Mean separated partial rank relations were 0.012 [-0.032, 0.057] and 0.007 [-0.019, 0.034]. Brackets give 95% person Student limits. The fixed shared positive-relation test failed, Holm p=0.28604. Mean pooled-minus-separated changes were 0.554 [0.454, 0.654] and 0.639 [0.587, 0.692]. All 22 changes were positive. The fixed shared change test passed, Holm p=0.03125. Another MRI acquisition gave a positive change in eight people. Source authors report separate MRI cohorts. Full sound and person refits test panel sensitivity. An exact zero-signal Gaussian model gives pooled correlation sqrt(2/3). Two other energy definitions give zero and -1/sqrt(3). Human energy-definition checks are exploratory. A variable-error model gives a positive separated relation with no true signal. We supply a model-only four-repeat design check. The result is a controlled estimator benchmark, not evidence for a common neural cause or unstable neural representations. All negative outcomes, code, new scalar statistics, source pins and terms, figures and repeat steps remain.

human neuroscience · SEEG · fMRI · auditory cortex · repeat variation · pattern energy · measurement error · cross-modal test · negative result · reproducibility