Only 28 of 118 selected ds004789 people had variable scores under a recall error order reference that kept source counts fixed

We tested recall error order in OpenNeuro ds004789 version 3.1.0. We selected the lowest numbered FR1 session for each of 273 people. The list checks accepted 4,926 English lists with 12 words. We found 2,124 first error words from valid earlier lists. Adjacent eligible errors gave 353 pairs in 271 target lists from 118 people. A uniform order reference kept each target list's error words, source-list counts, and error positions fixed. Only 40 target lists from 28 people had scores that could change. The other 90 people had differences of zero. The mean excess with equal person weights was 1.44 percentage points. Its 95% person bootstrap interval was -0.43 to 3.67 points. The planned order draws gave an upper-tail probability of 0.0507. A later numerical enumeration check gave 0.049406. All four planned sensitivity checks are included. The package provides exact reference moments and an error-to-source ledger. An independent reader checked all error links and pair records. Fresh retrieval verified all 551 source files. Both replays reproduced every numerical result file. The result shows an information limit for this order reference in the selected sample. It does not identify a neural cause.

recall errors · free recall · conditional permutation · source-list order · bootstrap · OpenNeuro