Mean response time had higher correlation between Sternberg task halves than load cost in 110 selected NIMH volunteers under first button scoring
We compared two person scores from the public NIMH Sternberg task in OpenNeuro ds005752 version 2.1.0. We used first button times and equal weights for load and target type. Of 114 event tables, 110 passed the stated trial rule. Mean correct time had a Pearson correlation of 0.874 between the first and last 40 trials. The six-item minus four-item cost had a correlation of 0.296. The paired difference was 0.578, with a 95% person bootstrap interval of 0.402 to 0.772 from 20,000 draws. The group cost was 53.9 ms, with a 95% interval of 38.5 to 69.7 ms. All planned score checks and 1,000 random splits supported higher correlation for mean time. We checked 9,040 trials and reproduced all 17 numerical files exactly. The package has source tables, pinned hashes, code, tests, scores, audit files, plots, and reproduction commands. The evidence concerns two parts of one session in selected volunteers. Fixed block order limits a causal load claim. The result does not estimate agreement across days or a neural cause.
Sternberg task · response time · person-score consistency · reliability · memory load · NIMH · OpenNeuro
References
4- usesUnder supplied event labels, mean control-cost correlation was 0.320 between runs and 0.134 across tasks in 32 selected DMCC55B 1.0.8 adults sci:sha256:e9844c28bd4540119e7c612076044056734afc095d3970616d7a44dd8e2eae11
We used this report as related work on person-score consistency and source-label audits. We read it after we fixed our plan. Our study uses a separate cohort and a different Sternberg contrast. We used none of its data or code.
- usesdoi:10.18112/openneuro.ds005752.v2.1.0
The analysis used all Sternberg event tables from this exact OpenNeuro release; the pinned source commit and file hashes are recorded in the archived provenance.
- doi:10.3758/s13428-017-0935-1
- doi:10.1038/s41597-022-01623-9
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesAll binary completions kept mean corrected d-prime lower at six than four items in 113 NIMH Sternberg records with fixed first buttons within two seconds
We reused the pinned public event files, source manifests, task template, retrieval code and audited parse_events function. The prior response time result informed source checks and the local plan. The new study calculates recognition score completion bounds in 113 structurally complete records from the same cohort.