Additive region removal left a positive person match gap between two movies from one visit in ten ds006192 people

We tested person matching between two movies from one visit in all ten numbered people from ds006192 version 1.0.0. The primary analysis used standard reconstruction, 100 cortical regions, and Fisher correlation edges. An exact least-squares fit removed all edge effects of the form a_i+a_j. The residual true-pair mean was 0.637018. The other-pair mean was 0.489614. The gap was 0.147404. Only the true pairing reached this gap among all 3,628,800 pairings. The exact one-sided p value was 2.75573192e-7 under exchangeable person labels. Nearest matches were correct for 7/10 people from Bao to Rat and 6/10 in reverse. Removal did not improve raw-edge counts. All five planned settings and all ten person omissions retained a positive gap. An independent solver verified the fit, scores, ranks, and exact counts. Archived-source and fresh-download repeats matched every numerical output byte. The package preserves 215 source files, fixed URLs and hashes, code, proof, results, all primary null scores, and instructions. Motion was correlated across movies. A common-factor counterexample shows that additive removal does not remove all regional causes. The result does not establish a neural cause or stability across days.

functional connectivity · person matching · fMRI · regional effects · permutation test · OpenNeuro