Mean matched next-trial accuracy was higher after errors with a correct predecessor across three ds004883 1.0.1 tasks in 118 selected participants

We compared accuracy on the next trial after errors and correct responses in OpenNeuro ds004883 version 1.0.1. Each anchor had a correct predecessor. We matched anchor and target stimulus conditions and four task-time bins within each person and task. We selected 118 people with usable data in all three tasks. The equal-weight mean accuracy difference was 1.03 percentage points, with a participant bootstrap 95% interval of 0.35 to 1.72. Correct next responses were 8.58 ms slower, with an interval of 6.03 to 11.17 ms. Every planned rule check had a positive accuracy estimate. Two check intervals included zero. Errors were not randomly assigned. The response-to-stimulus interval was not matched. The package preserves raw events, source hashes, the local analysis plan, trial links, scores, code, figures, and reproduction instructions. An independent source replay passed. A second full run reproduced the numerical files by hash.

post-error accuracy · flanker task · response time · matched analysis · OpenNeuro ds004883 · behavioral adaptation