For 128 nonvoice sounds per study and fixed references, mean repeat area scores were positive in six SEEG patients and 24 fMRI adults

We tested whether two response directions repeat between people for 128 source nonvoice sounds per study. Separate public studies supplied human SEEG high-gamma amplitude and BOLD fMRI. Two patients and four MRI adults supplied fixed references. A local code lock preceded new scores in six other patients and 24 other MRI adults from two studies. One repeat set site scales and two paired directions. Other repeats gave a signed area score. Mean scores were 0.000034397, 0.00299704 and 0.00205993. Their pointwise 95% person Student limits were [0.000006421,0.000062372], [0.00241292,0.00358117] and [0.00127203,0.00284783]. All 30 target scores were positive. The fixed conditional mean-and-majority rule passed, p=0.025078. An added reference-and-target bootstrap crossed zero in SEEG. Neither single-patient reference passed the full rule. A 24-site electrical check was uncertain. An exact model had a rank-one true cross-map and independent repeat errors. Common reference noise made the person-only rule pass in half its sign-paired draws. This is an interpretation limit, not a human fit. The contribution is a checked tool, a conditional cross-modal benchmark and a reference-error limit. It does not establish two shared neural sources. Clinical sampling, source screens, different cohorts, shared schedules, remaining acoustics and signal filters limit scope. Code, source pins and terms, all aggregate outcomes, figures and repeat steps are supplied. Source neural arrays stay external.

human neuroscience · SEEG · fMRI · auditory representation · repeat area · reference uncertainty · rank constraints · public data