Older lures had higher mean false alarm rates than recent lures in 20 distinct ds004350 baseline two-back records

We compared recent and older lures in OpenNeuro ds004350 version 2.0.0. Two pairs of baseline event files were identical under different participant IDs. The correct file owners were unknown. We excluded all four affected IDs from both visits after the planned analysis. In the 20 retained baseline records, the mean false alarm rate was 10.64% for older lures and 4.88% for recent lures. The mean difference was 5.75 percentage points. The 95% person bootstrap interval was 3.18 to 8.29 points. The interval assumes independent retained records and fixed observed letter schedules. The lure-difference correlation between visits changed from 0.669 with 24 IDs to 0.310 with 20 IDs. The retained correlation interval was -0.202 to 0.767. Four assumed mappings with one owner per pair gave correlations of 0.417 to 0.756. This range is not a confidence interval. The source files cannot resolve the mapping. We preserve all event files, hashes, the local plan, both analyses, code, and figures. An independent parser checked every letter row. A fresh download and full replay matched all 17 result files and six figure files. The results describe behavior in the tested records. They do not establish a treatment effect or trait reliability.

working memory · two-back task · lure interference · false alarms · duplicate records · OpenNeuro