In 124 ds000030 control runs, one removal can make 14 of 17 failed arrow groups pass the basic screen.

In the fixed sample of 124 control runs from OpenNeuro ds000030 version 1.0.0, the prior report identified 17 arrow groups with negative mean gaps; one record removal could make 16 gaps nonnegative. This calculation asks whether the same removal also passes the STOP response-probability rule (0.25–0.75, inclusive). The basic screen requires both a calculable nonnegative mean gap and this probability rule. Fourteen of the 17 groups have at least one separate STOP response removal that passes the basic screen. Across those groups, 36 removals qualify: 24 STOP response removals and 12 GO response removals. The pooled and opposite-arrow conditions exclude no further qualifying removals. The two mean-restoring counterexamples that fail the probability rule have post-removal STOP fractions of 2/15 and 3/15. Results are descriptive of the pinned observations and specified rules. Removals keep later recorded values fixed, do not represent behavior in a changed task, and are not independent samples or population probability estimates. Each arrow group has only 16 STOP trials; these observations do not establish a failure of the independent race model or overall task quality. Recorded decimal precision does not establish physical timing accuracy. No response should be discarded on these results alone.

stop-signal task · response probability · reaction time · record removal · OpenNeuro · reproducibility · exact arithmetic