The 126 selected CONTROL event tables match across four ds000030 Git revisions

A fresh download and independent exact-fraction implementation reproduced the reported results for 126 selected CONTROL scanner runs in OpenNeuro ds000030. The same 126 event tables matched byte for byte across Git revisions 00001, 00002, 00016, and 1.0.0. All 509 source-file checksums matched the prior report, although the four dataset-description files differ. All 3,054 reference checks agreed, including 2,520 participant fields, with numerical tolerance 1e-8; the largest absolute difference was 4.8936170212765957446808510638297872340425531914894 ms. Mean recorded, planned, and hybrid SSRT estimates for the 126 participants were 225.175797235521, 197.838886174516, and 195.116273425997 ms. A behavioral screen retained the same 102 participants as the report. These fixed-record results do not establish when participants heard a sound, validate all race-model assumptions, or show biological or clinical effects. The report was known before this replication, and shared records can preserve shared errors. No imaging, other diagnosis groups, training runs, staircase history, or bootstrap intervals were analyzed. Exact values, checks, source hashes, code, and limitations are documented in the full text and packaged artifacts.

stop-signal task · event tables · computational replication · OpenNeuro · SSRT · exact fractions · reproducibility · version comparison