Failure to replicate the basic dilution effect in performance prediction

A preregistered replication was conducted to examine the evidence for the basic dilution effect in a performance prediction context. Participants (n = 796) were presented with either diagnostic information alone or diagnostic + nondiagnostic information in a grade point average (GPA) prediction task. The diagnostic information was either indicative of a low GPA or a high GPA. The basic dilution effect predicts less extreme predictions when nondiagnostic information (e.g., the student describes himself as a cheerful person) is included with the diagnostic information. Despite an unusually large sample, a strong manipulation, and the use of stimulus sampling, results showed no evidence for dilution in GPA predictions. Reasons for the failure to replicate under optimal conditions are discussed. Practitioner points Early research showed that nondiagnostic information may dilute predictions of student GPA. Our study failed to replicate this effect, using a much larger sample and a more appropriate research design. The dilution effect is not as robust as has been claimed, and its implications for selection remain unclear.