Literal versus zero-filled delays reversed 18% of participant pairs in ds000030 stop-signal runs

An independent implementation reproduced 20 participant-level output fields for 266 OpenNeuro ds000030 snapshot 1.0.0 stop-signal runs and compared participant ordering under literal recorded versus zero-filled stop-signal delays. The two expressions reversed 6,345 of 35,245 participant pairs (18.0026%; Spearman correlation 0.783209). After the original behavioral screen, 5,720 of 29,646 pairs reversed (19.2943%; correlation 0.762129; 244 runs). The largest-expression set overlapped for 55 of 67 participants in the full cohort and 47 of 61 in the screened cohort. Participant bootstrap intervals describe resampling variation, not true stopping ability. These are arithmetic sensitivity results from the same observations as the reproduced analysis, not an independent biological sample or a refutation. The recorded zeros, planned delays, and event fields do not establish actual tone onset or a corrected individual estimate. Only 32 stop trials were available per run; race-model validity, trigger failures, and downstream neural or clinical associations were not tested.

stop-signal task · stop-signal reaction time · response inhibition · participant ranking · measurement sensitivity · OpenNeuro · reproducibility