At selected auditory sites, the top quarter by voice preference retained tuning in a source nonvocal panel in six SEEG patients and 24 fMRI adults

We tested whether sites with strong voice preference keep reliable response differences among source nonvocal sounds. Separate public human studies supplied SEEG high-gamma amplitude and fMRI BOLD estimates. One repeat selected the top quarter of auditory sites by voice AUC. Two other repeats measured nonvocal sound rank reliability. Development used two patients and four adults. A local plan and code lock preceded tests in six other patients, 16 B2021 adults and eight NH2015 adults. Mean retained reliability was 0.3715, 0.4593 and 0.3124. Pointwise 95% person intervals were [0.2930,0.4500], [0.4209,0.4976] and [0.2733,0.3515]. All 30 people had positive scores. The shared rule passed, with Holm p=0.046875 for three claims. The initial negative-association prediction failed. A positive direction fixed after development also failed, with Holm p=0.21875; one SEEG patient was negative. Controls retained nonvocal tuning after uniform spatial response removal and disjoint sound calibration. A small positive preonset score and discordant association controls remain. An exact normal model shows that unequal site noise can make observed preference and reliability rise together despite fixed latent preference. Thus the test does not identify a latent specialization tradeoff. Separate cohorts, source site screens, clinical sampling, broad sound labels and common schedules limit scope. We do not establish shared neurons or causality. The package preserves code, source pins and terms, local plans, all results, uncertainty draws, figures and repeat steps. Original neural arrays remain external.

auditory cortex · voice preference · nonvocal tuning · SEEG · fMRI · rank reliability · measurement precision · independent repeats