For 64 random tag sets on fixed category panels, mean rank accuracy across images was above 50% in 14 SEEG patients and five fMRI adults; the food rule failed

We tested whether visual response patterns rank food above nonfood when both test categories are absent from fitting. Separate public studies supplied human SEEG and BOLD fMRI. Three patients supplied development. A local lock preceded food tests in 14 other patients and all five MRI adults. Mean rank accuracies were 61.57% and 67.71%. Their 95% person Student limits were [48.48, 74.66]% and [49.45, 85.97]%. The fixed shared rule failed, p=0.0625. An added exploratory test used 64 fixed random category tag sets. Different image halves supported mean same-category tag accuracies of 54.63% and 55.10%. Their 95% person Student limits were [52.92, 56.33]% and [49.37, 60.82]%. Category-excluded means were 51.03% and 47.45%. A common category-matching permutation reference gave shared Holm p=0.001, conditional on observed panels and tag sets. A complete orthogonal-category example gave 100% same-category and 50% category-excluded accuracy for every tag assignment. It assigns no special food preference. The tag result supports measured category transfer between images in both modalities. It does not confirm food transfer to new categories, shared neural causes or causality. After eight visual terms were removed, food accuracies were 48.74% and 61.73%; this descriptive check did not give a shared positive pattern. Source selection on clinical images, sparse tissue coverage, different cohorts and filters, four MRI food kinds and five MRI people limit scope. Code, source pins and terms, all failures, scalar data, figures, exact checks and repeat steps are supplied. Source neural arrays remain external.

human SEEG · fMRI · food · category transfer · random tags · public data · negative result · neural decoding