A fixed test with two donor maps did not confirm a shared sound prediction gain in six SEEG patients and 24 fMRI adults

We tested whether a short calibration can set weights for two donor sound maps and reduce error for other sounds in other repeats. Separate public studies supplied SEEG high-gamma responses and two fMRI cohorts. Two fixed people per cohort formed donor atlases. A local plan and code lock preceded all 30 target endpoints. Calibration and test sounds were disjoint, and target repeat roles were separate. Broad sound-group and additive sound effects were removed. Mean error gains were +0.1613%, +0.6434% and -0.1386% of target map energy in six SEEG patients, 18 B2021 adults and six NH2015 adults. Their 95% person t intervals were [-0.0406,+0.3631]%, [+0.3598,+0.9270]% and [-0.3986,+0.1214]%. Positive counts were 4/6, 18/18 and 1/6. The fixed shared test failed, p=0.890625. All negative scores and planned controls remain. Short baseline subtraction also removed the small electrical gain. An example had identical neural responses in all people and no noise. Different site samples gave useful weights for each person. The model was not fitted to human data. Thus a calibration gain need not identify individual neural codes. The contribution is a reproducible prediction benchmark, a failed cross-modal test and a control for site sampling. Source selection, fixed donors and sounds, small clinical sampling and unequal measurements limit scope. Code, source versions and terms, figures, all results and repeat steps are supplied. Original neural arrays remain external.

human neuroscience · SEEG · fMRI · sound geometry · calibration · cross-validation · negative result · measurement sampling