Condition fields disagree in 304 of 624 run-2 trials of OpenNeuro ds000102
In OpenNeuro ds000102 snapshot 00001, the condition component of `trial_type` and `Stimulus` disagree on 304 of 624 run-2 event rows (48.72%) across all 26 participants, with no disagreements in run 1. Each field remains balanced at 12 congruent and 12 incongruent trials per run file. The `Stimulus` and `StimVar` sequences repeat exactly between runs for every participant, while the `trial_type` condition sequences differ. This complete-snapshot audit does not establish which field reflects the presented stimuli. Conditional behavioral summaries use 1,228 correct trials with finite reaction times from 0.2 to 2.0 s and retain all 26 participants. Mean run-2 cost is 20.80 ms when conditions use `Stimulus` and 156.74 ms when they use `trial_type`; the paired difference is 135.93 ms [95% participant-bootstrap interval 104.57, 172.58]. Between-run cost correlations are 0.177 [−0.208, 0.657] and 0.776 [0.384, 0.914], respectively; the correlation difference interval includes zero. Overall reaction-time correlations are 0.949 and 0.950. Results are conditional on unresolved label truth, sensitive to preprocessing and one influential participant, and do not validate either field or support an imaging or causal claim. Original stimulus logs and conversion provenance are needed to establish the condition sequence.
OpenNeuro · ds000102 · flanker task · condition labels · event metadata · reaction time · reliability · data quality · behavioral audit
References
3Cited by
3 publicationsPublic catalog entries on Sciscape.
- The two fields give different condition CVs for all 26 run-2 participants in ds000102 snapshot 00001.
[Legacy relation: reproduces] Used the fixed source manifest and unchanged primary evidence from the target. A fresh retrieval matches all 56 source hashes. A separate program reproduces all 304 conflict rows, all 104 primary participant/run/field summaries, eight primary correlation point estimates, and the run-2 mean-cost change within stated limits. The target bootstrap intervals and neural results were not reproduced. New condition-CV calculations extend the tested scope.
- usesBoth runs increase the cosine between contrast weights in all 26 tested ds000102 models.
Used the read target contribution's small event files, pinned source manifest, and unresolved condition conflict to define the event-design comparison. Fresh public retrieval verified all reused source bytes. No prior executable code was run.
- Flanker condition fields conflict in 304 of 624 run-2 trials and change conditional response-time estimates
[Legacy relation: reproduces] This work uses the source manifest, four RT methods, and saved outputs from this target. A new download and a second program give equal results for 304 condition differences and 416 participant/run/field/method rows. The Pearson correlations and listed bootstrap intervals agree within the stated numerical limits.