Conflicting ds000102 congruency fields yield different run-2 interference estimates, but neither encoding is verified
The 26-participant NYU Slow Flanker event tables contain conflicting congruency labels in 304 of 624 run-2 trials (48.7%); run 1 has no conflicts. Both fields agree on correctness, but the public tables do not establish which label reflects the stimuli presented. Using the same correct reaction times, mean run-2 interference is 24.93 ms under `Stimulus` and 160.07 ms under `trial_type`, a paired difference of 135.15 ms (participant bootstrap 95% interval 104.65–170.40). Between-run raw-interference correlations are 0.399 and 0.833, but the paired interval for their difference (−0.055–0.641) includes zero. The work reproduces the counts and untrimmed contrasts from an earlier analysis of byte-identical event files and adds conditional correlations, ICCs, and sensitivity analyses. Neither encoding is adjudicated. The original task logs and conversion ground truth are absent; the same 26 participants, only two consecutive runs, and one repository snapshot do not establish long-term reliability or independent biological replication. No neural result is claimed.
Flanker task · reaction time · congruency labels · data quality · reproducibility · within-session reliability
References
4- Congruency fields disagree on 304 run-2 trials in ds000102 snapshot 00001 sci:sha256:f90e75435f7bf8adef6363a04414db283c4ca12780311676b25d22b288251dd5
[Legacy relation: reproduces] Separate calculations on byte-identical source event tables match all 304 conflicts, 104 untrimmed participant/run/encoding rows, seven aggregate estimates, and the between-run sequence audit. Extends the evidence with conditional correlations, ICCs, paired uncertainty, and additional sensitivities; no independent cohort or label ground truth is claimed.
- doi:10.1016/j.neuroimage.2010.10.046
- doi:10.3758/s13428-017-0935-1
- doi:10.3389/fnimg.2022.953215
Cited by
5 publicationsPublic catalog entries on Sciscape.
- The between-run RT correlation has both signs in the ds000102 snapshot 00001 label model.
[Legacy relation: reproduces] Our new code and public source download give the same 104 primary participant/run/field score rows and four Pearson correlation values. This check does not cover the other filters, intervals, or neural analyses.
- The ds000102 Pearson correlations and bootstrap intervals agree for six score specifications.
[Legacy relation: reproduces] New code and fresh retrieval of the same observations reproduce all 624 score rows, Pearson correlations, ICC values, and bootstrap endpoints under six score specifications within the stated tolerances. Four Spearman values differ because rational arithmetic preserves a source-decimal tie. The strict discrepancy report is retained. No new cohort or correct condition encoding is established.
- uses10 of 26 ds000102 run-2 contrast weight cosines are less than zero with six motion columns.
Its inspected event-label result supplied the two conflicting encodings and the question for this motion test. Its response-field limit ruled out a response-button direction check.
- usesThe contrast weight cosine is less than zero in 9 of 26 primary ds000102 run-2 models.
Its inspected source manifest supplied the version-pinned event URLs and expected checksums. Its label conflict supplied the two encodings tested in the new fMRI design calculation.
- The run 2 decrease in mean response time is larger in the first half of the tested ds000102 task.
[Legacy relation: reproduces] The read manifest supplied 55 pinned source URLs and expected checksums. Separate calculations reproduce all 52 overall participant/run means and the 304-conflict count on the same source observations. The largest mean error is 2.2737367544323206e-13 ms. This does not reproduce the prior confidence intervals, condition reliability, or neural analysis.