Fixed spatial rank readers failed a shared load test in 12 SEEG patients and two fMRI groups

We tested whether the order of spatial response deviations predicts high versus low memory load in new recording blocks. Separate public studies supplied bipolar SEEG spectrum indices and fMRI response coefficients. Three people per modality supplied development. A local code lock preceded new scores in 12 patients, 16 Study 1 MRI adults and 11 Study 2 MRI adults. Readers used site ranks after training-reference subtraction. Electrical ranks were formed within each band. MRI ranks were formed across 33 regions. Mean test AUCs were 0.6844, 0.7083, 0.5568. AUC of 0.5 is chance. The fixed rank claim failed; its shared Holm p was 1. The added-value claim failed; its shared Holm p was 1. Both claims used person inference and block label references. A model with one input and unequal site loadings also gave strong rank prediction. Thus rank prediction does not identify multiple neural sources or a memory-storage mechanism. Clinical coverage, adaptive task history, source region selection, unequal tasks and measurement filters limit scope. The package gives code, permitted electrical inputs, source pins and terms, all scores and exclusions, figures and repeat steps. Exact author MRI rows stay external. No common neural cause was established.

human neuroscience · working memory · SEEG · fMRI · spatial ranks · cross-session prediction · reproducibility