Public logs restore 5,473 EEG choice codes after unique trial matches in ds004917 version 1.0.1
We tested whether public task records can restore choice information in OpenNeuro ds004917 version 1.0.1. We audited 7,185 decision events in 38 fMRI records and 24 EEG records. In 23 EEG records, 5,474 choice fields contain reaction time values instead of side codes. We linked each complete event sequence to source task rows by probabilities, ambiguity, reaction time, and trial order. Choice and reward did not select these matches. Unique matches and event log responses restore 5,473 invalid choices. One trial has two opposite button codes at the same time and remains unavailable. All 1,711 valid original side codes agree with the source tables. The resulting dataset has 7,184 verified choices, normalized reward fields, original values, and source row links. An independent program confirms the matches and recovered choices. A descriptive behavior check finds fewer choices for the higher visible probability under ambiguity in both cohorts. A separate schema check finds that the supplied hidden division values fail the stated probability equation on all ambiguity trials. We preserve that field without correction. The package includes source tables, log extracts without personal fields, source hashes, code, figures, and reproduction instructions. The result applies to this source snapshot. It does not test the source paper's causal brain claims.
ambiguity · decision-making · EEG · fMRI · data audit · choice recovery · trial matching · OpenNeuro
References
2- usesdoi:10.18112/openneuro.ds004917.v1.0.1
The audited OpenNeuro event tables come from ds004917 version 1.0.1.
- doi:10.1371/journal.pbio.3002452
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesThe last choice and its win interaction gave no clear gain in fixed choice models of 38 ds004917 fMRI records
We used the verified trial ledger, restored EEG choices, reward fractions, and matched source lines from this audit. These inputs enabled the fixed test of adjacent choices for people outside the fitting data. We also used its source snapshot, manifests, and log removal code to check sources and preserve a retrieval recipe. We ran its complete alignment verifier again. The new prediction models and results are in this publication.
- usesA complement rule with an edge limit reconstructs 7,185 stored visible probabilities in ds004917 version 1.0.1 within 0.0001
This study reported that the stated probability equation failed on all 3,595 ambiguity decisions. We used that result as a starting point. We verified the failure from original files, tested an empirical complement and edge rule, and measured reconstruction errors. We did not use recovered choices.
- usesA fixed model with equal hidden probability assignment gives no clear ambiguity prediction gain in 15 EEG test participants from ds004917 version 1.0.1
We use its 7,184 verified choices, source records, source hashes, and independent source checks to fit and test the two fixed choice models. We preserve the complete source package and reproduce its choice recovery.
- usesAdding four outcome features increased observed mean log loss in fixed logistic models for 24 EEG records from ds004917 version 1.0.1 under equal division of the hidden area
The input audit supplies verified choices, source row links, source files, and source hashes.
- usesA fixed model with equal hidden probability assignment gives no clear ambiguity prediction gain in 15 EEG test participants from ds004917 version 1.0.1
We use its 7,184 verified choices, source records, source hashes, and independent source checks to fit and test the two fixed choice models. We preserve the complete source package and reproduce its choice recovery.