Cue derivatives reduced maximum absolute false stop coefficients on a -0.5 to +0.5 s grid and raised standard errors with a fixed kernel and independent noise in 209 PIOP2 designs

We tested cue timing error in 209 public PIOP2 stop-task schedules. Both models had separate canonical cue columns and two stop columns. One model also had raw cue temporal derivatives. Each constructed cue group had amplitude one and an independent response shift from -0.5 to +0.5 s in 0.01 s steps. The true stop response was zero. Adding cue derivatives reduced the maximum absolute false coefficient in all 209 schedules for all three stop contrasts. For successful stops, the median paired false coefficient ratio was 0.08694. The median paired standard error ratio was 1.94463. Standard errors assumed independent errors with equal fixed variance. Five fixed sensitivity checks are included. Exact finite-grid extrema gave shift assignments for every run and target. A fresh download checked all 683 source hashes. Separate SVD fits checked all 2,508 primary extreme signals. A full fresh-input repeat matched all 220 main numeric files. The package includes code, event tables, metadata, proofs, designs, weights, traces, results, and reproduction commands. These are fixed-kernel design calculations. They do not show a measured neural effect, give continuous timing bounds, or establish a preferred model. The new contribution is the stop-task audit and tool. The linear model and finite-grid methods are established.

fMRI · stop-signal task · cue timing · temporal derivatives · design audit · linear models · AOMIC PIOP2