Two miss scoring rules reverse 16.4% of strict score pairs in 190 PIOP1 working memory records

We compared two rules for source miss labels in AOMIC working memory data. We used OpenNeuro ds002785 and ds002790, version 2.0.0. The primary sample required eight scored answers in each active condition. All-trial accuracy used every active trial. Response-only accuracy excluded source misses. The rules reversed 2,762 of 16,885 strict pairs in 190 PIOP1 records: 16.36% [13.55%, 19.39%]. In 221 PIOP2 records, 179 of 23,285 strict pairs reversed: 0.77% [0.18%, 1.68%]. A strict pair had no tie under either rule. Brackets give 95% intervals from 10,000 participant bootstrap draws. All 623 known PIOP2 miss hands matched the next-stimulus rule in the public converter. Selected original statements produced such hands without a response. Different response time rules limit cohort comparisons. The archive has 432 event files, source hashes, code, full results, figures, and offline instructions. Independent checks passed. A clean rerun reproduced all 33 result files and four figure files exactly. The results concern recorded scores. We did not estimate memory capacity.

working memory · accuracy scoring · source miss labels · AOMIC · OpenNeuro · participant ranking · bootstrap intervals · response time · data curation