A unit response signal produced negative stop contrasts in 209 AOMIC PIOP2 designs with the stated kernel, separate cue terms, and no response term

We tested whether a signal at recorded button response times can enter the successful-stop minus failed-stop signal contrast. We used 226 public AOMIC PIOP2 event schedules from OpenNeuro ds002790 version 2.0.0. Fixed trial rules retained 209 schedules. The base model had separate cue terms and no response term. A constructed unit common response signal gave a negative stop contrast in all 209 base designs. The median was -1.150. Adding a response term recovered zero contrast when its kernel matched the true signal. Under independent errors of equal variance, the median contrast standard error ratio was 1.113. An exact proof gave a median equal-error threshold of 0.396 for absolute response amplitude divided by frame noise standard deviation. All planned model checks retained negative base contrasts. A one-second shift in the true signal left error after a response term was added. The package includes all source events, checked retrieval hashes, design matrices, per-run scores, code, an audit command, the proof, and figures. All 685 fresh source checks passed. An independent solver passed. A second calculation reproduced 216 result files byte for byte. This result concerns constructed signals and specified models. It does not establish an effect in measured brain activity.

stop-signal task · fMRI · response regressors · linear models · contrast bias · design matrix · AOMIC PIOP2