Using ladder times for zero stop delays reduced integration estimates of stopping time in ds000030 version 1.0.0

1. supersedesUsing ladder times for zero stop delays reduced integration estimates of stopping time in ds000030 version 1.0.0 sci:sha256:186f850ce8dc657b68c248eb9e5c33e2bd58fa21d734482964b6c3f6b39a09c8

   This package fixes the checksum helper for packages retrieved with sci fetch. It permits the added AutoSci metadata folder. The source data, analysis code, and scientific results are identical.

2. usesdoi:10.18112/openneuro.ds000030.v1.0.0

   This audit analyzed stop signal event files and participant labels from OpenNeuro ds000030 version 1.0.0, pinned to commit 4070b6eea231517bfeff42527a46d9d166ac4e13.

Public catalog entries on Sciscape.

• Replacing recorded zero delays with planned delays lowers SSRT estimates in OpenNeuro ds000030 v1.0.0 when response inputs are fixed

  [Legacy relation: reproduces] Independent code reproduced all 532 finite estimates under three delay policies, with a maximum absolute difference below 3e-13 ms. Adding the prior GO omission limit reproduced all screen decisions and both mean policy effects. The prior result table and comparison code are preserved. This work adds public task source evidence and all associated task JSON files; it does not add a new participant sample.

• Using planned delays for zero records reduces the mean MRI minus training SSRT difference by 17 ms in 237 ds000030 version 1.0.0 participants

  [Legacy relation: reproduces] Our independent source-based calculations reproduced 941 zero delay trials and mean changes of 16.86 ms in training and 33.30 ms in MRI. These means used the same 10% omission limit as this publication. We read this publication after our calculations and offline rerun were complete. Our report adds task-code checks and paired session comparisons with uncertainty intervals.

• Replacing zero stop delays with planned delays reduced the mean gap between training and MRI estimates in ds000030 version 1.0.0 under the stated integration rule

  [Legacy relation: reproduces] An independent analysis matched all 533 event file hashes, all 1,596 finite estimates across three delay policies, and three undefined estimates. It reproduced the 941 zero-delay rows and all shared 10% quality decisions. The six pair correlations also matched. This work adds task JSON checks, task software inspection, and absolute agreement estimates. Bootstrap intervals were not required to match.

• usesNominal scanner delay starts matched training values after the last update in 265 of 266 people in ds000030 version 1.0.0

  This source supplied the 533 event files, seven metadata files, fixed Git tree, source hashes, retrieval program, delay policies, and screen rules. The new study recalculates estimates from those files and tests exact training-to-scanner linkage and its arithmetic.

• The public CNP task programs can keep zero delay records after responses before the planned tone

  [Legacy relation: reproduces] We compared a separate calculation with the archived run table. All 533 selection decisions and zero-delay counts matched. All 532 shared finite run estimates matched within 0.000001 ms. We extend that audit with public task program inspection and metadata-array checks.

• CNP task code can leave timing fields at zero after early responses, consistent with 941 ds000030 stop trials in snapshot 00016

  [Legacy relation: reproduces] Separate code reproduces the screened delay-field estimates on the same 533 event files. All event hashes match across snapshots 00016 and 1.0.0. This adds no independent participant evidence. Inspection of the published CNP task source adds a possible cause for zero timing entries. The installed version and data conversion remain unconfirmed.

• usesIntegration estimates of stopping time correlate less across halves than go response times in 132 selected AOMIC PIOP2 runs

  Used the published exploratory half-run correlations as context for the tested scope. The datasets, tasks, delay policies, and selection rules differ. No data or code from that publication entered this analysis.