The 95 percent gain interval included zero for previous trial EEG scores in 39 ERP CORE adults under fixed offline N170 models

We tested whether the previous trial EEG score reduced prediction error for later trials in the ERP CORE N170 task. Fixed personal ridge models fitted early trials and tested later trials. Both models used current and previous image category, current baseline, trial position, and stimulus gap. The added model also used the previous score. Rules retained 39 adults and 5,272 test trials. Mean person error fell from 23.305884 to 23.234815 squared microvolts. The gain was 0.071069 squared microvolts, or 0.305 percent. Its 95 percent person bootstrap interval was -0.051743 to 0.200155 squared microvolts. The relative interval was -0.253 to 0.809 percent. The interval includes zero. Nine planned checks gave small mean changes. The result applies to processed offline scores. It does not prove a neural memory effect. The package preserves source hashes, selection rules, models, predictions, code, and source retrieval instructions; distribution limits exclude the trial arrays and source metadata. Independent checks passed. Eight numerical files matched an offline repeat exactly. Two native records reproduced both source paths exactly.

ERP CORE · N170 · EEG · event-related potentials · serial dependence · trial history · ridge regression · prediction error · person bootstrap · offline prediction