A selected quarter of sites did not pass a shared face–house concentration test in 7 ECoG and 6 fMRI records

We tested whether a quarter of early selected sites supplies more than half of later face–house contrast products. Separate public human studies supplied seven ECoG patients and six fMRI participants. Each person had three sections for site selection and two later contrast estimates. We used 200 samples of 32 sites per person and selected eight sites in each sample. The fixed shared rule failed. Mean selected-minus-other score C was +0.591969 in ECoG, with a 95 percent person interval of +0.462319 to +0.708829. It was -0.028150 in fMRI, with an interval of -0.157839 to +0.101460. C was positive in seven of seven ECoG records and three of six fMRI records. One ECoG record had negative total later agreement. The joint sign-test p was 0.65625. Removing calibration by site standard deviation changed mean fMRI C to +0.2480, but that check did not rescue the primary claim. Ratios above one and negative remainders show why these products cannot be read as true signal fractions. Coordinate gaps limit the spatial check. The result does not establish a causal modality difference. The package preserves code, source terms and hashes, the local plan, every person result, negative checks, figures and repeat steps. A revised input loader reduced ECoG peak memory from about 900 to 223 MiB with an identical feature archive. Numerical audits and an exact analysis replay passed. No new positive cross-modal biological claim was confirmed.

human neuroscience · ECoG · fMRI · visual cortex · cross-validation · response concentration