A clip correction reduced mean area estimates for three sound groups in 20 iEEG patients and 16 fMRI adults; fixed tests failed

We tested the geometry of mean responses to speech, other human voices and animal sounds in separate public human studies. A local development split supplied four people per modality. Fixed tests used 20 other iEEG patients and 16 other fMRI adults. We derived and audited a signed area estimate that corrects finite clip sampling under independent clips, finite second moments and fixed calibration. Ordinary area was positive in every reserved person. Mean corrected scores were 0.024457 and -0.046826; their 95% person intervals were [-0.002351, 0.051265] and [-0.085407, -0.008245]. The shared positive-area test failed, Holm p=0.989686. A development-led test that correction removes over half mean ordinary area also failed, Holm p=0.31575. Descriptive mean decreases were 0.037960 and 0.247368. That summary was chosen after reserved electrical results. It is not a third confirmation. Independent covariance calculations, exhaustive half splits, simulations and source checks support the tool and arithmetic. A nonlinear scalar counterexample shows why area cannot count neural sources. All negative and discordant results remain. The package gives code, permitted inputs, source versions and terms, plans, scores, draws, figures, audits and repeat steps. Different clips, clinical coverage, cohort composition, measurement filters, source selection and uncertain clip independence limit scope. We establish neither shared positive neural area nor a common cause.

human neuroscience · intracranial EEG · fMRI · sound groups · response geometry · sampling bias · signed area · negative result · reproducibility