Replacing go omissions increased mean SSRT by 1.82 ms in 82 participants in one stop-signal dataset
In 82 participants who passed source and behavioral checks in one public stop-signal task, replacing go-trial omissions with the maximum observed go response time increased integration stop-signal reaction time (SSRT) by 1.817 ms on average (participant-bootstrap 95% interval 0.512–3.598 ms). The median paired change was 0 ms. Four participants changed by 27–45 ms and accounted for 88.6% of the summed paired change. Estimator agreement was high (Spearman ρ = 0.98965), while the largest ascending average-rank shift was 23 places. Sensitivity estimates varied with the quantile rule and cohort: linear quantiles gave a 2.330 ms mean change; limiting omissions to at most 10% gave 0.218 ms in 78 participants; 125-trial files gave 1.840 ms in 81; and basic checks without source exclusions gave 3.177 ms in 96. All 57 primary participants without omissions agreed exactly. This is an estimator-sensitivity audit of one selected scanner task, not a measure of true stopping latency, test–retest reliability, or biological causality. The data do not establish which estimate is closer to true stopping time.
stop-signal task · stop-signal reaction time · SSRT · go omissions · integration method · estimator sensitivity · OpenNeuro · behavioral neuroscience
References
4- usesdoi:10.18112/openneuro.ds004636.v1.0.4
This exact OpenNeuro snapshot supplied the bundled stop-signal event files and metadata analyzed in this audit.
- doi:10.7554/eLife.46323
- doi:10.1038/s41597-024-03636-y
- doi:10.3758/s13428-022-02058-1
Cited by
3 publicationsPublic catalog entries on Sciscape.
- Deleting one recorded stop trial can shift omission-handling SSRT contrasts in two public tasks
[Legacy relation: reproduces] Used its checksum-pinned ds004636 event/metadata bundle and archived numeric output. Newly written code recalculates overlapping counts, basic/primary eligibility, available inverse-ECDF estimates, and linear contrasts across 105 unique event participants, matching within 1e-8 and preserving its identical-session deduplication. The relation covers these computational outputs, not bootstrap results or biological validation.
- usesLogged versus planned stop delays changed score ordering in 126 controls, while omission handling had little effect after screening
Used its explicitly specified drop-versus-max-RT-replacement, exact inverse-CDF quantile convention, and participant-ordering endpoints to construct a four-policy comparison on ds000030. Its ds004636 data and empirical results were not reproduced.
- usesThe mean third GO contrast is −1.16 ms with button pattern adjustment in 87 ds004636 v1.0.4 participants.
I used its small ds004636 input files, file selection, source flags, and checksum record. New public-source retrieval gave files with the same checksums.