A fixed three-group area test failed across six SEEG patients and 24 fMRI adults despite mostly positive scores

We tested whether three broad auditory group means follow one spatial response axis. Separate public studies supplied SEEG high-gamma responses and fMRI BOLD estimates. A signed cross-repeat area score used nonvoice, nonspeech human voice, and speech-plus-song means. One excluded repeat set site scales. Development used two patients and four adults. We locked the code after development. Tests used six other patients, 16 B2021 adults, and eight NH2015 adults. Mean scores were 0.141272, 0.128790, and 0.035806. Their pointwise 95% person t intervals were [-0.079566,0.362110], [0.075305,0.182275], and [0.019155,0.052457]. Positive counts were 6/6, 15/16, and 8/8. The fixed shared mean and sign rule failed, p=0.080505. Both MRI studies passed their own rules, but the SEEG mean was uncertain. These results do not confirm a shared one-axis rejection. Simulations showed that independent repeats alone are insufficient. Asymmetric group errors gave positive area from a true response line. A nonlinear readout of one scalar also gave a measured plane. Thus positive scores would not count neural generators or establish a cause. Code, source pins, terms, plans, results, figures, and repeat steps are supplied. Independent audits checked source values and numerical results. Original neural arrays remain external. Source selection, clinical sampling, separate cohorts, group acoustics, unequal sensing volumes, filters, and structured error limit scope.

human neuroscience · SEEG · fMRI · voice · group means · representational geometry · crossvalidation · error covariance