A registered report of preregistration practices in studies of electroencephalogram (EEG) and event-related potentials (ERPs): A first look at accessibility, adherence, transparency, and selection bias.

Common explanations for replication failures in neuroscience and psychophysiology include the exploitation of researcher degrees of freedom and ambiguous or inappropriate methodology, creating an environment in which flexibility during data processing and analysis could increase the probability of erroneous or irreplicable findings. The present registered report described preregistration practices in EEG/ERP studies, quantified adherence to preregistration, and estimated expected replication/discovery rates. Out of 506 preregistrations and 25 registered reports screened, 385 met eligibility. The EEG/ERP preregistrations resulted in 92 published manuscripts. For the preregistered studies, 57-99% included the minimal necessary methodological detail for replication. Adherence to preregistration in the 92 published studies averaged 60%. Exploratory analyses indicated that registered reports had the highest average adherence (92%), followed by articles explicitly mentioning preregistration (60%), and then by those not mentioning preregistration (39%). Only 16% of published studies fully adhered to preregistered plans or disclosed all deviations. Preregistered studies reported more methodological details (64% vs. 61%) and more frequently justified sample sizes and data exclusion than companion non-preregistered studies. A z-curve analysis indicated that selective reporting was likely present in published preregistered studies. Although preregistration can enhance transparency and reduce researcher bias in EEG/ERP research, current practices fall short. Ambiguity in preregistrations and inconsistent adherence undermine utility of preregistration. Moving forward, researchers should prioritize clarity and accessibility in preregistrations, and journals should implement policies to ensure the review of preregistration adherence.