Variations in ERP data quality across paradigms, participants, and scoring procedures

Abstract Although it is widely accepted that data quality for event‐related potential (ERP) components varies considerably across studies and across participants within a study, ERP data quality has not received much systematic analysis. The present study used a recently developed metric of ERP data quality— the standardized measurement error (SME)—to examine how data quality varies across different ERP paradigms, across individual participants, and across different procedures for quantifying amplitude and latency values. The EEG recordings were taken from the ERP CORE, which includes data from 40 neurotypical college students for seven widely studied ERP components: P3b, N170, mismatch negativity, N400, error‐related negativity, N2pc, and lateralized readiness potential. Large differences in data quality were observed across the different ERP components, and very large differences in data quality were observed across participants. Data quality also varied depending on the algorithm used to quantify the amplitude and especially the latency of a given ERP component. These results provide an initial set of benchmark values that can be used for comparison with previous and future ERP studies. They also provide useful information for predicting effect sizes and statistical power in future studies, even with different numbers of trials. More broadly, this study provides a general approach that could be used to determine which specific experimental designs, data collection procedures, and data processing algorithms lead to the best data quality.