A second implementation reproduces 266 stop-signal runs, but delay conventions reverse participant orderings

A separately written implementation was compared with an archived AutoSci analysis of the same OpenNeuro ds000030 observations. It reproduced all 266 participant rows, 8,512 STOP rows, failed-screen flags, estimates under three delay conventions, rank statistics, and both specified participant-bootstrap intervals within the archived display-rounding tolerances. Among 241 runs passing limited behavioral screens, recorded versus planned delays reversed 5,700 of 28,920 participant pairs (19.71%; Spearman 0.756208; Kendall tau-b 0.605809). With at least 50 ms separation under both conventions, 405 of 11,886 pairs reversed (3.41%). Mean recorded-minus-planned estimate shift was 31.078 ms (95% participant-bootstrap interval 26.139–36.124 ms). The shift equals the mean planned-minus-recorded delay; this arithmetic does not establish which convention measures auditory onset or stopping latency. The analysis is a same-data computational reproduction, not an independent biological replication. Each run has only 32 STOP trials, and the convenience sample, selection, omissions, fast responses, trigger failures, and incomplete race-model checks limit biological interpretation. Cue delivery, acoustic onset, and repeated-measurement reliability were not established.

stop-signal task · stop-signal reaction time · computational reproduction · delay convention · participant ordering · OpenNeuro ds000030 · CNP · behavioral methods · bootstrap · reproducibility