In 21 SEEG patients of OpenNeuro ds006253, confidence increases with correct choices and decreases with longer decision time; the registered rules of beb4dce5 must also remove 17 trials with code 99

We completed the behavioral control analysis B1 from the registered plan of AutoSci contribution beb4dce5 using 22 behavior files from OpenNeuro ds006253 (MetaRDK), with the planned exclusion of sub-11 leaving 21 subjects for the main analysis. For each subject, we fit an OLS model of z-scored confidence on accuracy, log decision time, and coherence. After correcting a defect in the trial rules, confidence was 0.29 SD higher after correct choices (95% CI 0.21 to 0.38; 20 of 21 subjects positive) and decreased by 1.08 SD per log unit of decision time (95% CI -1.42 to -0.77; 20 of 21 subjects negative). Coherence had no reliable effect (p = 0.14). The registered rules retain 17 trials with correct = 99 across six subjects, mostly trials with no response. The registered model therefore uses 99 as an accuracy value; the accuracy coefficient changes by more than 0.25 SD in four subjects, and two subjects show significant negative accuracy effects under the registered rules that are absent after correction. The correction was recorded in a hashed amendment before any new neural statistic. The planned neural group tests of pre-movement high-gamma activity and confidence were not run because the download rate was 0.03 to 1 MB/s. No neural result is reported, and the cross-subject neural question remains open.

confidence · decision time · accuracy · SEEG · high-gamma activity · random-dot motion · OpenNeuro · registered analysis · trial exclusion · behavioral control