Unreliability as a threat to understanding psychopathology: The cautionary tale of attentional bias.

The use of unreliable measures constitutes a threat to our understanding of psychopathology, because advancement of science using both behavioral and biologically oriented measures can only be certain if such measurements are reliable. Two pillars of the National Institute of Mental Health's portfolio-the Research Domain Criteria (RDoC) initiative for psychopathology and the target engagement initiative in clinical trials-cannot succeed without measures that possess the high reliability necessary for tests involving mediation and selection based on individual differences. We focus on the historical lack of reliability of attentional bias measures as an illustration of how reliability can pose a threat to our understanding. Our own data replicate previous findings of poor reliability for traditionally used scores, which suggests a serious problem with the ability to test theories regarding attentional bias. This lack of reliability may also suggest problems with the assumption (in both theory and the formula for the scores) that attentional bias is consistent and stable across time. In contrast, measures accounting for attention as a dynamic process in time show good reliability in our data. The field is sorely in need of research reporting findings and reliability for attentional bias scores using multiple methods, including those focusing on dynamic processes over time. We urge researchers to test and report reliability of all measures, considering findings of low reliability not just as a nuisance but as an opportunity to modify and improve upon the underlying theory. Full assessment of reliability of measures will maximize the possibility that RDoC (and psychological science more generally) will succeed. (PsycINFO Database Record