AOMIC PIOP2 version 2.0.0 marks 92.9% of available emotion matching task trials as incorrect

The audit uses all 431 available emotion matching task event files from AOMIC PIOP1 and PIOP2 at version 2.0.0. PIOP2 marks 9,901 of 10,656 trials as incorrect (92.9%). PIOP1 marks 588 of 10,032 trials as incorrect (5.9%). A rule that requires 90% correct labels keeps 1 of 222 people in PIOP2 and 172 of 209 people in PIOP1. All 10,344 face labels agree with the public response key and recorded response hand. The true accuracy remains unresolved because the original stimulus and response records are absent from the event files. The package preserves all source labels. It includes data, source versions, hashes, code, person and trial tables, figures, and reproduction commands. A separate program confirms the counts and response-key matches. A clean run gives the same hashes for all 11 result and figure files. This exploratory audit documents a difference that can affect data selection. It does not validate a label repair.

AOMIC · emotion matching · response accuracy · behavioral data · OpenNeuro · data audit · label validation