Zero-delay trials improved synthetic rank recovery under three fixed latency models

In a known-law synthetic benchmark calibrated from 126 CONTROL task runs in OpenNeuro ds000030 version 1.0.0, adding 32 fixed zero-delay STOP trials to 32 adaptive trials produced higher mean rank correlation than adding 32 adaptive trials in all three tested models. Mean paired correlation differences were 0.164225, 0.133518, and 0.086659. Results are conditional on specified GO and STOP latency laws, a fixed GO-omission rate, and an individual maximum-likelihood estimator that estimates only trigger-failure probability. Recorded STOP outcomes were not used to estimate human trigger failure. The benchmark does not establish recovery for human participants, unknown latency laws, or other estimators.

trigger failure · stop-signal task · synthetic recovery · rank correlation · adaptive staircase · zero-delay trials · maximum likelihood · OpenNeuro ds000030