For 54 sounds, donor panel gains were uncertain in six SEEG patients and positive in 24 fMRI adults

We tested whether neural donor responses can select 16 sounds that improve response-contrast forecasts for 54 other sounds. Public SEEG high-gamma activity and BOLD fMRI came from separate human studies. Development used two patients and four adults. A local code lock preceded tests in six other patients and 16 other adults. Reference panels had the same broad-class counts. Mean gains were 2.694% and 3.316% of test contrast energy. Their 95% person t intervals were [-0.308,5.697]% and [2.861,3.771]%. Five of six patients and all 16 adults improved. The fixed shared mean-and-sign rule failed, p=0.109375. The same BOLD donor model and panels gave mean gain 3.259% in eight adults from another study, interval [1.896,4.623]%, with all positive. A planned donor-label shuffle still gave positive mean panel gains in all three cohorts. Every shuffled model had negative skill against zero contrast prediction. A later exact zero-mean model gave positive gain through lower fitted-noise amplification, without sound signal or constant-response gains. Short-baseline subtraction removed the electrical benefit. A first order-term check was unstable; a later projection retained SEEG uncertainty. These checks do not repair the primary failure. Independent calculations checked all 30 reserved people within 4e-16. The package gives code, source pins and terms, all outcomes, fitted maps, figures, plans and repeat steps. Source selection, cohort differences, measurement filters, fixed sounds and small electrical sample limit scope. No common neural cause or positive primary cross-modal benefit is established.

SEEG · fMRI · auditory cortex · experimental design · donor responses · sound selection · response contrasts · negative result · measurement noise · reproducible research