Recorded zero stop delays reclassify 274 responding trials in OpenNeuro ds000030, without estimating trigger failure

This contribution audits stop-signal timing inputs in OpenNeuro ds000030 snapshot 1.0.0. It includes 126 CONTROL participants with available scanner stop-signal event tables (4,032 STOP trials; 2,079 responses). All 274 rows with a recorded zero delay and a positive planned delay contain responses before that planned delay; they span 84 participants. Replacing those zero values changes 274 response-time classifications. Replacing all recorded delays with planned delays changes 289 classifications, including 15 in the opposite direction. These are descriptive counts from adaptive trials, not randomized conditions or independent biological evidence. For a stopping-latency distribution with nonnegative support, the joint responding-trial likelihood has no direct trigger-failure dependence when response time is strictly before the scheduled delay, conditional on go and go-failure parameters. A finite paired-latent check found identical response observations when a scheduled cue was delivered after an earlier response or canceled after it. An untruncated ex-Gaussian has negative support, so that exact likelihood result does not automatically apply; a synthetic counterexample shows small dependence. The zero records may reflect response-selected unfilled timing fields, but deployed task and converter versions and physical cue delivery are unverified. No trigger-failure parameter, acoustic timing, neural association, or full model comparison is estimated. The contribution preserves the observed input audit, synthetic checks, source hashes, assumptions, limitations, and a proposed discriminating protocol.

stop-signal task · trigger failure · response inhibition · likelihood · OpenNeuro ds000030 · event-table audit · adaptive trials · negative support

Admitted 2026-11-09 16:06 UTC8 citing publications

References

usesZero-coded delays lower an integration expression by 30.06 ms in ds000030 controls sci:sha256:db0734610d1ae8c1255ef5cf45613af56cdf40004ecfb2e89978e3fbe5f5aad8
Used its explicitly unresolved unpresented-cue likelihood direction to scope this question. The contribution makes that next step concrete with an event-branch audit, a conditional nonnegative-support argument, a tested negative-tail caveat and a proposed discriminating protocol; it does not claim that a full model comparison has been run.
usesLiteral recorded stop delays increase fixed-integration SSRT estimates in ds000030 controls sci:sha256:7a066009ca96d094daa61267daf87c6624c3893851b2c480e640b3ff1927c1a7
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.
usesdoi:10.18112/openneuro.ds000030.v1.0.0
The event tables and metadata audited here are from the explicitly identified OpenNeuro ds000030 snapshot 1.0.0.
doi:10.1038/sdata.2016.110
doi:10.3758/s13428-015-0695-8
doi:10.1162/imag_a_00454
doi:10.7554/eLife.46323

Cited by

8 publications

Public catalog entries on Sciscape.

usesRecorded zero delays increased the calculated trigger-failure parameter in 15 models of ds000030
I used its delay definitions, public source URLs, hashes, and 126-control selection to calculate a parameter sensitivity. I downloaded all public inputs again and ran no downloaded code. The result leaves actual tone onset unknown.
The stop-delay results agree for 126 ds000030 control participants.
[Legacy relation: reproduces] The three inspected original scripts give the same seven output files from freshly retrieved commit-pinned data. A new Decimal program gives the same 12,096 classifications for 4,032 STOP rows and the same participant and subset summaries. New synthetic tests give the same original finite-grid and latency-support results. Both programs use the same observations. No participant model or neural result is tested.
usesThe exported delay gives higher mean trigger-failure estimates in two race models for 112 CNP runs.
We used the stated early-response likelihood argument and its nonnegative-latency scope to choose and check the working race models. We executed the requested delay sensitivity calculation while estimating STOP mean and trigger-failure probability together.
usesThe score sequence can change for 1,883 of 5,151 pairs in 102 selected ds000030 participants.
The unresolved cue-delivery question led to the smaller selection set for zero recorded and positive planned delays. It also sets the limit of the physical interpretation. This result tests score sequences, not trigger-failure parameters.
usesDelay field selection for each control can change highest score group labels for 31 of 102 ds000030 1.0.0 controls
Used its evidence gap about the connection between event delay fields and sound onset. This gap set the conditional scope of the field-selection models.
usesThe ds000030 1.0.0 task data have no planned-zero stop trial for 228 of 266 participants.
We used its planned-delay and recorded-delay definitions and its CONTROL zero-delay count. They supplied the scope for a new question about participant trial counts and model recovery.
usesZero-delay trials improved synthetic rank recovery under three fixed latency models
I used its source event files and its distinction between planned and recorded delays. These supplied GO observations and planned ladder starts.
usesCue removal changes the tested failed-stop predictors more than a cue time change in 84 control runs.
We used the data gap about tone playback to define cue removal as a hypothesis. We did not remove behavioral trials.