Mean certainty profile transfer across four faces was positive in single neurons and EEG at seven morph levels; fMRI tests failed

We tested whether response profiles for physical expression certainty repeat across four face identities. Separate human cohorts supplied spike timestamps, scalp EEG and BOLD fMRI. The fixed panel had seven fear-happy morph levels. Development used three patients and four MRI adults. Reserved tests used 13 other patients and 15 other MRI adults. Mean all-level cross-identity scores were 0.09008 and 0.06890. Their 95% person t intervals were [0.01858,0.16157] and [-0.04924,0.18704]. The original shared all-level and intermediate-only rules failed, Holm p=0.231510 and 0.922157. A later plan fixed one new test before EEG scores in 23 separate adults. Mean EEG score was 0.16536 [0.05418,0.27654]. A conservative guard across five elementary means supported the selected firing/EEG relation, adjusted p=0.044414. This is separate EEG evidence for a candidate found in the patient data. Pooled EEG effects were known. It is not a new patient replication. Choice and response time controls and feature-mean removal kept positive electrical mean intervals as fixed descriptive checks. Intermediate-only checks did not support a broader shared claim. All failures and negative people remain. An exact one-axis example gives score one, so positive transfer does not identify distinct latent variables or a common cause. Source labels, raw values and score arithmetic passed stated audits. Sparse clinical sites, separate cohorts, fixed faces, source filters and unequal resolution limit scope. The package gives compact permitted inputs, source versions and terms, code, plans, all results, figures and repeat steps.

single-neuron · EEG · fMRI · facial emotion · expression certainty · cross-identity transfer · public data · replication