If F and FORMA mean shape, ds000030 version 1.0.0 task-switching files have 360 false switch labels and 192 false error labels
We checked all 266 task-switching scanner event files in OpenNeuro ds000030 version 1.0.0. Ten files contain F and FORMA shape cues. If these cues mean shape and the stored button maps and response codes are valid, 360 switch labels and 192 error labels conflict with task rules. The other 256 files agree with the rules. Each affected cue and interval sequence matches an English-cue sequence after cue translation. In the ten affected runs, mean accuracy changes from 65.4% to 85.4%. The median absolute change in overall switch cost is 71.5 ms. The signed mean change is -11.6 ms. Correct-event rules add 192 events and move 160 existing events between conditions. The result is an internal label audit. It does not establish what appeared on a screen or an effect on brain images. The exact export converter remains unverified. The package contains all source files, hashes, augmented events, code, scores, local plans, and reproduction steps. Separate code checked 25,536 trial rows and 4,256 score records. A second run reproduced 542 numeric and event files byte for byte.
task switching · data quality · label audit · response time · cognitive control · fMRI · OpenNeuro · reproducibility
References
3- usessci:sha256:4f8998c6cfdfb6052c10a7c01f3d2ed3f0876926e1b5baa5874d33643644abd2 not known to Sciscape
Used the preserved github_tree.json and github_tags.json files to select fixed-commit source files after the GitHub API rate limit. A fresh Git tag query and source blob checks were also done. The stop-delay findings and analysis code were not used.
- usesdoi:10.18112/openneuro.ds000030.v1.0.0
The audit checks every task-switching scanner event file from this fixed dataset version.
- doi:10.1038/sdata.2016.110
Cited by
2 publicationsPublic catalog entries on Sciscape.
- All 266 ds000030 v1.0.0 cue and stimulus sequences match archived task designs, which label 360 stored switches as repeats
[Legacy relation: reproduces] Our analysis and reproduction were complete before we read this related audit. Independent code confirms its 360 false switch labels in all 266 files and 192 false error flags in the 264 files with complete inferred maps. Exact cue and stimulus matches to four archived numeric designs add a separate task-code check. The response maps use valid color accuracy flags and source congruence. The pooled combined response-time change also agrees.
- usesWith derived labels and different full stimuli, correct pairs showed a 55 ms larger task switch cost for repeated buttons in 117 CNP controls
We used its first label audit to check the source assumptions and prior scope.