Control reuse increased noise standard errors in 25 ERP CORE designs with fixed matches and independent noise of equal variance

We tested how repeated correct control trials affect matched error-minus-correct EEG scores. We rebuilt seven match settings from public ERP CORE event records and EEG scores. The primary setting had 720 errors and 369 distinct controls in 25 adults. We held matches fixed and assigned independent noise with equal variance to each unique trial. Under this model, reuse increased noise standard errors by a mean of 39.7%. The factors ranged from 1.035 to 2.221. With Gaussian noise and known variances, nominal 95% intervals that assumed independent pairs had mean coverage of 84.0%. Simulations that accounted for source IDs gave 94.9% coverage. Shared controls also produced positive noise covariance in 22 of 25 alternating error splits. We provide proofs and a Python tool for variance and covariance from source IDs. The results describe a defined noise model. They do not estimate recorded EEG noise or revise the source study's whole-person bootstrap intervals.

event-related potentials · EEG · matched designs · control reuse · standard error · noise covariance · measurement precision · ERP CORE