Direction limits for voices failed a shared test in 20 iEEG patients and 24 fMRI adults; the response origin changed coverage at fixed RMS error

We tested whether acoustic forecast limits still cover fewer human voices after each forecast is given the observed response length. Separate public studies supplied signed intracranial local field potentials and two BOLD fMRI cohorts. Zero forecasts used a fixed score and remained in the test. Four patients and four adults supplied development. A local plan fixed the primary test before new scores. Tests used 20 other patients, 16 B2021 adults and eight NH2015 adults. Mean other-sound minus voice direction coverage was 7.76, 4.27, and 10.74 percentage points. Their 95% person t intervals were [3.42, 12.09], [-1.07, 9.61], and [2.52, 18.95]. The shared test for a positive deficit failed, p=0.40181. An added source origin check was fixed before confirmation. It kept the source baseline zero instead of subtracting the fitted site mean. RMS residuals stayed fixed. Mean direction deficits changed by -6.57, +54.23 and +49.77 points. These origin checks are descriptive. An exact example shows that subtracting a fitted mean can turn a pure size error into opposite directions. An unfitted noise model also changed direction coverage with one unchanged true direction. These checks do not explain the human cause. The contribution is a failed shared test of voice direction, a response origin benchmark and reproducible tools. Separate cohorts, different sound panels, clinical sampling, source selection and unequal measurement filters limit neural claims. The package gives code, licensed electrical inputs, source pins and terms, all results, figures, audits and repeat steps. Original MRI arrays remain external.

human auditory cortex · intracranial EEG · fMRI · voice · prediction limits · response direction · reference origin · negative result