New software gives the same six sign changes in 225 PIOP2 task runs.

New standard-library Python software reproduced the recorded stop-signal task contrasts for OpenNeuro AOMIC-PIOP2 and UCLA CNP event files. For the primary four-part rule, 225 PIOP2 runs had all three contrasts; six changed sign from the pooled GO contrast to the all-STOP-weighted contrast, and 19 changed sign to the failed-STOP-weighted contrast. For CNP, no training runs and two scanner runs changed sign to the all-STOP-weighted contrast; none in training and two in scanning changed sign to the failed-STOP-weighted contrast. Across 2,271 records with all contrasts, the software checked the stated algebraic identity and target covariance identity. It compared 222,114 recorded fields or values and matched all signs, counts, flags, and missing values; all 120 bootstrap interval limits agreed within the specified tolerances. Six PIOP2 sign changes refer to the four-part rule and the full sample; the secondary screen and 2- and 8-part analyses gave different counts. The result concerns the specified files and method. It does not establish correct stimulus timing, a valid replacement for the pooled model check, SSRT bias, reliability, or biological causation. The CNP scanner delay uses training information, and part boundaries affect the PIOP2 result.

stop-signal task · response time · reproduction · software verification · bootstrap · OpenNeuro