Protected motion removal kept mean synthetic contrasts of 0.261 and 0.272 in 431 AOMIC working memory designs under a unit task residual RMS bound

We tested motion removal that keeps a task contrast unchanged. The motion terms were made orthogonal to the task and baseline terms. The inputs were 431 public AOMIC version 2.0.0 working memory designs: 207 PIOP1 and 224 PIOP2 runs. The target was change minus nochange. The main model used six motion terms. Each synthetic residual artifact had RMS equal to the residual unit-contrast task signal. The sharp maximum remaining contrast averaged 0.261 and 0.272 in the two cohorts. The artifact that reached each bound remained as the exact residual task shape. A joint fit gave zero contrast for these artifacts and one for a unit task signal. The bound squared also equaled one minus the base-to-joint variance ratio under independent errors. All seven fixed scenarios and 3,017 design tests are included. A reusable NumPy tool, proof, source bytes, fixed retrieval recipes, attaining signals, and separate checks are preserved. Offline and fresh source repeats matched all 17 output files exactly. These results test designs and specified artifact amplitudes. They do not measure brain activity or actual motion bias.

fMRI · motion artifacts · contrast estimation · regression · design audit · reproducibility