Two STOP models give the same predictions with different trigger-failure probabilities for 266 ds000030 GO-time samples.

For 266 runs of the ds000030 stop-signal task, I audited 96 GO and 32 STOP trials per run and compared exact response-time probability mass functions (PMFs) from two reference models. Model A has trigger-failure probability 1/10 and a 200 ms STOP latency when stopping starts. Model B has no trigger failure and assigns probability 9/10 to a 200 ms latency and 1/10 to a longer latency whose sum with any tested nonnegative cue delay exceeds every permitted GO time. Across 17,290 planned, recorded, and synthetic zero-delay cases, every PMF entry agreed exactly; two synthetic inputs outside the stated support conditions broke the equality. The recorded GO maxima ranged from 487.711227999171 to 1504.57972499999 ms (median 895.096297999771 ms), and 31 STOP responses in 27 runs exceeded the run-specific GO maximum. These are finite-sample support comparisons, not evidence of population GO limits or fits to participant data. The constructed point-mass models do not establish parameter recovery under restricted latency distributions. The next necessary study is parameter recovery for the adaptive 96-GO/32-STOP design with long STOP-latency tails and uncertain GO-time support. That study, including estimation and interval coverage, was not performed.

stop-signal task · trigger failure · parameter recovery · race model · response time · GO-time support · STOP latency · exact arithmetic · OpenNeuro · ds000030