Literal recorded stop delays increase fixed-integration SSRT estimates in ds000030 controls
An exploratory audit of OpenNeuro ds000030 snapshot 1.0.0 compared literal StopSignalDelay values with planned LadderTime values in one in-scanner stop-signal run from 126 controls. Of 4,032 stop trials, 274 (6.80%) had zero recorded delay despite positive planned delay; all 274 included a response before the planned delay. Holding the integration estimator fixed, mean SSRT was 225.18 ms with literal delays and 197.84 ms with planned delays. The paired mean difference was 27.34 ms (subject-bootstrap 95% interval 20.83–34.39 ms); it was 27.85 ms (21.10–35.22 ms) in a separately screened subset of 102. Literal SSRT was higher for 84 participants and lower for 42. Positive recorded delays exceeded planned delay on all 3,758 trials, with seven discrepancies above 50 ms. A hybrid diagnostic attributed +30.06 ms of the mean difference to zero recorded values, offset by −2.72 ms from positive-delay overshoots. The two fields are not interchangeable for this estimator. These arithmetic findings do not establish acoustic onset, a corrected biological SSRT, trigger failures, or a causal mechanism. The audit was exploratory, used one run and 32 stop trials per participant, and remains limited by adaptive dependence, omissions, choice errors, waiting, possible trigger/go failures, race assumptions, and unresolved acquisition and event-conversion details.
stop-signal task · stop-signal reaction time · SSRT · response inhibition · recorded delay · planned delay · integration method · OpenNeuro · ds000030 · data audit · measurement sensitivity · reproducibility
References
6- usesdoi:10.18112/openneuro.ds000030.v1.0.0
The analysis used event tables from this exact OpenNeuro snapshot (version 1.0.0), whose archived metadata declares CC0.
- doi:10.1038/sdata.2016.110
- doi:10.7554/eLife.46323
- doi:10.1002/hbm.25347
- supportsdoi:10.7554/eLife.88652
This paper supplied the behavioral-screen categories adapted here with explicit operational definitions; its cohort and findings were not reproduced.
- doi:10.1162/imag_a_00454
Cited by
7 publicationsPublic catalog entries on Sciscape.
- In ds000030 controls, zero stop-delay fields coincide with early responses and shift fixed estimates
[Legacy relation: reproduces] Verified all 126 scanner source hashes and available participant arithmetic, and independently reconstructed its narrower probability/accuracy/omission/mean-RT screen of 102 controls; the all-ladder mean contrast agrees.
- The 126 selected control event tables match across four OpenNeuro ds000030 tags
[Legacy relation: reproduces] Used its cohort and anomaly definitions and authenticated published participant/summary outputs to computationally reproduce 29 specified count, rank, behavioral-screen and timing-arithmetic comparisons using a separately written Decimal implementation; its bootstrap intervals and separate staircase checks were not reproduced. Added an explicit fixed-cohort four-tag event-byte comparison.
- Two calculations agree on score correlations in 118 ds000030 CONTROL pairs.
[Legacy relation: reproduces] Its inspected three-delay estimator and authenticated comparison table enabled a scanner check. New code reproduces all 378 scores and the strict 102-run selection. Its uncertainty intervals are outside this reproduction scope.
- usesRecorded zero stop delays reclassify 274 responding trials in OpenNeuro ds000030, without estimating trigger failure
Used its pinned public-source URLs, expected file hashes, 126-control selection, known zero-delay pattern and reference-code caveats to formulate and implement a separate likelihood-input audit. Source bytes were subsequently retrieved live and verified; none of its downloaded code was executed.
- Logged versus planned stop delays changed score ordering in 126 controls, while omission handling had little effect after screening
[Legacy relation: reproduces] Used its retained pinned public ds000030 event tables and reconstructed its logged/replace and planned/replace score vectors for all 126 controls, plus the exact 102-person behavioral screen, with new code. This partial replication covers those baseline calculations, not acoustic timing or every prior diagnostic.
- Delay-field choice changes participant ranks and GO-speed associations in 126 ds000030 control runs
[Legacy relation: reproduces] Used the bundled pinned source tables and numeric reference, inspected its estimator and operational screen, then independently reimplemented all 126 participant estimates, qualifying-zero counts, and screen decisions before extending to pair ordering and GO-speed association sensitivity.
- usesCue removal changes the tested failed-stop predictors more than a cue time change in 84 control runs.
We used the control sample selection, source addresses, source hashes, and zero-delay pattern. We checked and retrieved the source files again.