Fixed covariance tests did not confirm vocal matching gains across six SEEG patients and two fMRI studies

We tested whether repeat correlations learned from other sounds improve vocal item matching. Separate human studies supplied SEEG high-gamma activity and fMRI BOLD estimates. Development used two patients and four MRI adults. A local code lock preceded tests in six other patients and 24 other MRI adults from two studies. The fixed metric used 128 source-defined nonvocal sounds for calibration and 37 vocal sounds for testing. It used sixteen sites per panel. Mean gains over diagonal distances were 1.274, 1.270 and 4.043 percentage points. All three mean estimates were positive. One SEEG patient lost accuracy. Both fixed-fit SEEG ranges included zero when people and target sounds varied. Both fixed cross-modal claims failed Holm correction, p=0.0625 and 0.0625. We preserve all controls and negative results. Independent arithmetic checked all 36 primary people, including development. Source audits matched all retained input values. Artificial examples show limits from changed target covariance, repeat gains and source selection. The result concerns processed repeat matching. It does not confirm a common neural decoding benefit or a cause. The package supplies code, source pins and terms, scores, figures, plans and repeat steps. Original neural arrays remain external.

human SEEG · fMRI · auditory cortex · repeat covariance · noise normalization · sound identification