A minimum voltage t test against zero rejected all sign patterns of 40 fixed ERP CORE waves within 110–230 ms

This study tested a minimum voltage score in 40 public ERP CORE N170 EEG records. It used face minus car waves at P7/P8 and a fixed 110–230 ms window. Separate fair sign reversals gave zero expected voltage at every time. The selected minimum had an exact null mean of -3.541 µV. A bound showed that a one-sided t test against zero rejected all 2^40 sign patterns at level 0.05. This bound concerns the fixed processed waves. A selection-aware sign test rejected in 4.86% of 20,000 null cohorts. Two score evaluations per person reproduced full selection for each sign pattern and an unstudentized mean. Test validity requires joint symmetry under separate person sign reversals. Marginal symmetry alone failed in a shared-sign example. Six fixed sensitivity checks gave the same complete naive rejection. The general peak-bias problem is known. The package provides its exact size in this sign model, a certificate, and a reusable test tool. It preserves code, source epochs, provenance, results, proof, and reproduction instructions.

event-related potentials · EEG · peak selection · sign test · randomization inference · symmetry · multiple comparisons · N170