Two assumed repeat orders gave opposite fMRI mean load history slopes in 28 adults; shared tests with 12 SEEG patients failed

We tested whether past memory load predicts a new measured response along a separately fitted load axis. Public studies supplied human SEEG band power and BOLD fMRI. Session 1 fitted each electrical reader. Three MRI adults fitted a fixed MRI reader. Tests used 42 later sessions from 12 patients, 16 other MRI adults, and 12 adults from a second MRI study. Current condition, order and recorded history were controlled. All 35 MRI behavior records matched public event sequences and probe durations. The neural repeat-to-time links remained unverified. Exact limits covered every exchange of two neural repeats within condition. Reversing every pair changed the sign of all four MRI group mean history slopes. Assignment limits included zero in 55 of 56 MRI person-and-phase cases. An audit found 17 invalid electrical history loads. Their removal gave Encode and Maintain means of -0.01353 and +0.00142, with 95% person Student intervals [-0.04360,+0.01655] and [-0.03898,+0.04182]. All four original and four corrected shared majority tests failed, with Holm p=1. Electrical encoding transport and second-study MRI maintenance transport were uncertain. Thus, no common neural history shift was confirmed. The contribution is a tested assignment audit, a public chronology check and a complete negative benchmark. Cohorts, source sites, filters, source selection, error assumptions and unknown repeat links limit scope. Code, permitted compact inputs, source hashes, readers, every result, figures and repeat instructions are supplied. Original MRI rows remain external.

working memory · trial history · SEEG · fMRI · repeat assignment · partial identification · reproducibility