Nine mean RT checks change after recorded-zero row removal in selected ds000030 version 1.0.0 runs. Each reference mean is less than one.

In 122 selected scanner runs and 121 selected training runs from OpenNeuro ds000030 version 1.0.0, removing recorded-zero STOP rows changed five and four initially satisfactory mean reaction-time checks, respectively. We compared these removals with every same-size subset of STOP responses in each run, while keeping GO responses, other STOP rows, task sequence, and run selection fixed. The reference mean counts were 0.34462587206395257 for scanner runs and 0.6203463203463203 for training runs. Across nine changed runs, reference fractions ranged from 1/286 to 984/1820; the recorded-zero rows had the minimum possible removed RT sum in only 43 of 170 selected runs with such rows. This is an exact, descriptive removal reference, not a biological sampling model or a significance test. The analysis is exploratory, uses the same data for selection and calculation, and does not reproduce stopping-time scores or establish physical STOP cue timing. The task adapts delays, the scanner and training runs share people, and recorded-zero labels can depend on response timing.

stop-signal task · reaction time · recorded-zero rows · subset enumeration · removal reference · ds000030 · reproducibility