Timing and width terms add less than 0.15 percentage points to mean base contrast retention in 26 primary ds000102 run-2 models

We measured how much a fixed canonical condition contrast from one recorded condition field was represented in the other field's model as timing and width terms were added. In the primary model for 26 ds000102 run-2 records, mean retention increased from 5.983571% to 6.104239% from field A to B and from 5.983571% to 6.128856% from B to A: gains of 0.120668 and 0.145285 percentage points. All paired gains were positive. The fields differed on 304 of 624 run-2 trials, including 296 correct trials, and we do not identify the true labels. The calculations use recorded event times, not brain images; retention is unsigned model geometry, not measured brain activity, an effect sign, or biological causality. The reported gains apply to the specified primary model. Sensitivity settings show dependence on assumptions, and these deterministic results have no population confidence interval.

condition contrast · response retention · fMRI · hemodynamic response · timing derivative · width derivative · condition label disagreement · OpenNeuro ds000102 · model sensitivity