Four-part trial boundaries change the weighted response-time difference sign in 14 of 225 PIOP2 runs.

Across 225 calculable runs from OpenNeuro PIOP2, changing the three boundaries in a four-part trial partition changed the sign of the failed-STOP-weighted response-time difference in 14 runs among all 891 ordered partitions with part lengths of 20–30 trials. Fifteen of 19 previously reported pooled-to-equal-part sign changes persisted across every partition; four depended on the tested boundaries. With all STOP counts as weights, 19 signs changed. A prespecified screen retained 173 runs, and a narrower 23–27-trial family gave 8 failed-STOP-weighted and 11 all-STOP-weighted sign changes. The analysis used exact rational arithmetic and recorded event tables from 226 100-trial runs; one run had no failed STOP and no contrast. These are descriptive results for finite tested boundaries in one dataset. They establish no valid exclusion rule, stopping latency, neural inhibition, causal mechanism, or population uncertainty. The comparisons do not show which weighting is correct.

stop-signal task · response time · trial boundaries · weighting · sign changes · PIOP2 · sensitivity analysis · behavioral data