Choice and block adjustment reduced the mean linear feedback contrast by 9.05 percentage points in 78 ds004532 v1.2.0 Study 1 participants

We tested repeated choices in OpenNeuro ds004532 version 1.2.0. A transition links a choice to the next scheduled presentation of the same pair in the same training block. In 78 Study 1 participants and 17,046 valid transitions, the mean raw contrast between positive and negative feedback was 17.89 percentage points. A linear model with separate intercepts for previous stimulus and block gave a mean contrast of 8.85 points. The mean paired reduction was 9.05 points. Its 95% person bootstrap interval was 7.95 to 10.15 points. All eight planned checks gave a positive mean reduction. An exact covariance identity separates the raw score into within-stratum and between-stratum parts. Under independent choices with a fixed preference, a model example gives a raw contrast of 26.47 points and a zero contrast within previous choices. The human adjusted score remains an association. It does not identify a causal feedback effect or a neural mechanism. The package preserves all used source files, source hashes, code, person scores, all bootstrap draws, the proof, figures, and reproduction instructions. A separate parser and regression calculation passed. Two runs from fresh source files reproduced all ten main output files byte for byte.

probabilistic selection task · feedback learning · choice repetition · behavioral measurement · linear probability model · bootstrap