The early P7/P8 face–car peak time score had uncertain prediction gain in 39 ERP CORE adults from one visit

We tested an EEG timing score in 39 adults from the ERP CORE N170 task. The score was car peak time minus face peak time at P7/P8. Both models predicted a score from later trials in the same visit. The added model used the person's early score. We removed each test person from the data used to fit both models. The base mean squared error was 300.42 ms². The added model error was 174.56 ms². The gain was 125.86 ms². Its approximate 95% person-bootstrap interval was -172.75 to 396.68 ms². The interval includes zero. The data do not resolve whether the score reduces prediction error in the source population. This is not proof that the score contains no useful information. We used original 1024 Hz recordings, fixed rules, and 20,000 refitted bootstrap draws. All six fixed timing checks are reported. The package contains epochs, source hashes, trial decisions, predictions, code, and reproduction commands. The result applies to the specified score and fixed image set within one visit.

ERP CORE · N170 · EEG · peak latency · prediction · within-visit reliability · event-related potentials