Fixed donor sound models lowered test error for selected iEEG means in six patients and auditory BOLD in 20 adults with 120 s target fits

We tested whether fixed donor sound models help when a new person supplies 120 s for model fitting. Separate public studies supplied selected intracranial EEG (iEEG) contact means and early auditory fMRI region means. Three iEEG patients set a donor weight of 0.75. Six other patients and 20 fMRI adults tested the fixed rule. The fMRI donor set had five other adults. Mean later error fell by 1.495% in iEEG, with a 95% person interval of 0.601% to 2.487%, and by 13.598% in fMRI, with an interval of 11.615% to 15.676%. All test people improved. The rule also beat equal attenuation in both measures. Both planned shared Holm p values were 0.03125. The iEEG gain was uncertain with 300 s fitting and in the last test half. Pure donor prediction beat the fixed mixture in 19 of 20 fMRI adults. A clock audit led to one fMRI timing repair before confirmation outputs were read; all initial outputs are retained. The result supports a simple prediction tool in these settings. It does not establish a shared neural cause or a two-minute recording workflow. Donors had longer training records, and source selection and normalization used more target data. Fixed stories, clinical coverage, broad fMRI pooling and full-run processing limit scope. The package preserves code, compact inputs, source terms and hashes, plans, all results, figures, audits and repeat steps.

human neuroscience · intracranial EEG · fMRI · auditory cortex · transfer learning · ridge regression · data efficiency · natural speech · reproducibility