With four fixed faces, a personal choice entropy term raised neural forecast error in 13 patients and 14 fMRI adults

We tested if personal fear/happy choices improve neural expression-response forecasts. Separate human cohorts supplied neuron firing, scalp EEG and BOLD fMRI. Reference choices came from 11 people. A local lock and a pre-score missing-cell amendment preceded new endpoint tests. The tests retained 13 patients, 19 EEG adults and 14 MRI adults. Models trained on three faces in one record half and tested the fourth face in the other half. All fixed positive shared rules failed, Holm p=1. Adding personal choice entropy to two physical expression terms raised error in all 13 patients and all 14 MRI adults. Mean costs were 2.072 and 4.048 points of centered test-profile energy. Their 95% person Student limits were [0.678,3.465] and [2.533,5.563]. EEG costs were mixed, with mean 2.031 [-0.209,4.270]. The cost direction was selected after the positive tests. A 72-test sign guard gave shared p=0.008789. It is exploratory, not independent confirmation. All personal readers had negative mean neural skill. A no-shared-profile simulation also gave a cost. Thus the result concerns this forecast method; it does not show absent neural uncertainty coding or a common cause. Separate cohorts, fixed faces, sparse clinical sites, source filters and measurement models limit scope. The package gives source pins and terms, all outcomes, compact permitted inputs, code, figures, audits and repeat steps.

human neuroscience · neural prediction · choice entropy · fMRI · single neurons · EEG · emotion ambiguity · cross-modal analysis · negative results