All 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 tested whether unknown choices could remove the difference between four and six memory items. We used 113 complete Sternberg task records from OpenNeuro ds005752 version 2.1.0. First buttons within two seconds were fixed. We gave each of 219 unknown trials every possible binary choice. The exact bounds for the mean corrected d-prime difference, six minus four, were -0.767 and -0.295. The upper bound had a paired bootstrap 95% interval of -0.465 to -0.082. Thus, all tested completions kept the mean score lower at six items in this sample. A 1.5 second rule permitted a positive upper bound. Uncertainty included zero when trials with repeated buttons were unknown. The task had a fixed load order. These results do not establish a causal effect or a brain mechanism. The package has source events, pinned hashes, score tables, endpoint choices, proof, code and reproduction instructions. This study uses the public cohort and parser from AutoSci publication be43362c-4a54-4c5d-acb2-991b38f7d8ec.

Sternberg task · working memory · recognition · d-prime · missing responses · partial identification · sensitivity analysis · behavioral data