Unclassified failure pairs gave the largest mean clustering term in 171 ds004883 participants under uniform cell allocation with fixed counts and five time bins

We split a corrected score for adjacent trials into four terms for each failure type in public flanker event records. The source was OpenNeuro ds004883 version 1.0.1. All 171 people with three task files were included. The reference used uniform label allocation within fixed condition, segment, five time-bin cells, and category counts. The mean binary lag score was -1.363 score points. Its descriptive 95% person interval was -2.067 to -0.697. The two-sided conditional shuffle p value was 0.00010. Negative scores denote excess clustering. The unclassified failure pair term was -1.228 points and was the largest observed mean clustering term. This category contains intervals with absent or multiple responses. The recorded-error pair term was 0.303 points, with an interval from -0.200 to 0.798. That interval included zero. The four terms sum exactly to the binary score. A check that used only single recorded responses did not reject the reference (p=0.5505). We also rebuilt an earlier 126-person score to numerical precision. The result does not identify a causal error effect or the reason for absent responses. The package preserves a categorical tool, proofs, exact checks, source files and hashes, all results and draws, figures, and complete repeat instructions.

flanker task · response errors · correct-response yield · adjacent-trial clustering · conditional permutation test · categorical decomposition · OpenNeuro ds004883