Fixed tests did not confirm shared confidence correlations within errors for readers trained on correct choices in 14 EEG and 20 fMRI adults

We tested whether a neural reader learned from correct choices has a positive correlation with confidence within errors in other runs. Separate public human studies supplied EEG and BOLD fMRI. Four people per study were assigned to development. A local code lock preceded reserved outcomes. EEG used cleaned components at 200 to 450 ms after the response. MRI used 100 coarse spatial cells and trial estimates with task, motion and rating motor controls. All readers used only correct training trials. The tested score was a within-person partial correlation after stated behavior controls. Reserved means, after the Fisher transform and its inverse, were EEG +0.048 [-0.115, +0.209] and MRI +0.052 [-0.027, +0.130]. Brackets give 95% person Student limits. Positive counts were 9/14 and 13/20. The recovered-row shared p was 0.266581; its Holm p across two source-row scopes was 0.533162. The fixed shared test failed. A source trial audit and fixed minimum counts limit the tested sample. EEG source processing used all runs. Tasks, rating scales, timing and spatial sampling differ. Report, movement, vascular and unmeasured evidence effects remain possible. This test does not identify a unique confidence mechanism or establish causality. The package gives code, source versions and terms, compact trial inputs, all scores and exclusions, figures, baseline checks and repeat steps.

confidence · metacognition · EEG · fMRI · cross-run prediction · error awareness · public data