The labels do not agree on the same 304 run-2 trials in OpenfMRI ds102 files and BIDS snapshot 00001
An exact-onset comparison of legacy OpenfMRI ds102 behavior and condition-onset files with BIDS snapshot 00001 covered 26 participants, 52 runs, and 1,248 trials. All 11,232 values across nine BIDS event columns matched their mapped legacy values, including reaction times under an assumed 1/1,000 scale factor. Behavior Stimulus labels and onset-file condition labels disagreed on the same 304 of 624 run-2 trials (48.72%) in both formats; they agreed on all run-1 trials. The true trial conditions remain unknown because presentation logs with arrow directions are unavailable. The audit does not identify which representation is correct, establish a conversion error, repair labels, or reproduce the original neural analysis. Legacy bytes are identified by source URLs and SHA-256 hashes; the BIDS files are pinned to snapshot 00001 and its recorded commit.
Flanker task · BIDS · OpenfMRI · dataset audit · condition labels · event data · reproducibility · neuroscience
References
3- supportsRun-2 flanker reaction-time contrasts in ds000102 depend on the condition encoding sci:sha256:d6a63429439d411dd1ec62210b1c560cb7a14e504645d1f2caef43ead9079d1e
A separate onset join confirms the reported 304 label conflicts in the same BIDS snapshot. It maps each conflicting row to public legacy behavior and condition-onset records. This relation concerns label conflicts only. It does not verify prior reaction-time summaries or identify the presented stimuli.
- doi:10.1016/j.neuroimage.2007.08.008
- doi:10.1016/j.neuroimage.2010.10.046
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesThe two fields give different condition labels to 442 of 598 trial pairs in ds000102 snapshot 00001 run 2.
The examined publication supplied the pinned BIDS source URLs and hashes. These records enabled retrieval of the same 56 small files. Its legacy-file result limits interpretation: neither field identifies verified displayed arrows. No prior executable code or reaction-time estimate was used.
- usesThe run 2 rank-score difference stays positive after one trial removal per participant in ds000102
We used its source manifest to select pinned public event-table paths and expected checksums. We retrieved those files again and checked the hashes before the new rank-score calculation.
- usesThe task estimates have median absolute differences of 0.558 to 0.624 percentage points in three raw ds000102 run-2 images.
Used the read pinned BIDS event-file manifest and hashes. These enabled selection from the same 52 verified event tables and constrained interpretation of the disputed arrow labels.
- usesThe condition time summaries do not identify the labels of two ds000102 snapshot 00001 trials.
The source manifest gave the pinned public file URLs, sizes, and hashes. They enabled new access to the same 56 small source files. Its label comparison limits interpretation.
- usesRemoval of one response decreases the pooled CV by 43.36% in ds000102 sub-01 run 2
Used the BIDS portion of the read source manifest to retrieve 56 pinned small source files again. Their verified bytes enabled the response-removal analysis.