Keeping site labels improved repeated sound assignment after mean, size and global sign removal in six SEEG patients and 24 fMRI adults

We tested spatial response patterns with separate public human studies. SEEG supplied high-gamma responses. Two fMRI studies supplied BOLD estimates. Every sound response lost its spatial mean, size and global sign. A fixed reader kept site labels. A second reader used the same full distribution of values but removed site order. Each reader assigned two sounds within a source group to templates from another repeat. Development used two patients and four adults. A local code lock preceded tests in six other patients, 16 B2021 adults and eight NH2015 adults. Mean site-order gains were 15.27, 22.08 and 22.96 percentage points. Their 95% person intervals were [7.69,22.84], [19.65,24.52] and [19.18,26.74]. Every reserved person improved. The shared mean and majority test passed Holm correction across three claims, p=0.046875. Neither a gain from keeping relative signs nor the opposite prediction selected in development passed. All failures remain. Exact and simulated records show that these reader gains can occur with identical zero mean responses or one varying scalar plus a fixed site offset. Thus the finding concerns spatial layout in processed measurements. It does not identify a number of neural sources, a common neural cause or a causal effect. The package gives code, source hashes and terms, all results, figures, local plans, audits and repeat steps. Separate cohorts, source selection, fixed sounds, clinical sampling, unequal site counts and measurement filters limit wider transfer.

human neuroscience · SEEG · fMRI · sound identity · spatial patterns · relative signs · response normalization · repeatability