Unreliable Yet Still Replicable: A Comment on LeBel and Paunonen (2011)

Abstract

Read online

LeBel and Paunonen (2011) highlight that despite their importance and popularity in both theoretical and applied research, many implicit measures continue to be plagued by a persistent and troublesome issue – low reliability. In their paper, they offer a conceptual analysis of the relationship between reliability, power and replicability, and then provide a series of recommendations for researchers interested in using implicit measures in an experimental setting. At the core of their account is the idea that reliability can be equated with statistical power, such that lower levels of reliability are associated with decreasing probabilities of detecting a statistically significant effect, given one exists in the population (p.573). They also take the additional step of equating reliability and replicability. In our commentary, we draw attention to the fact that there is no direct, fixed or one-to-one relation between reliability and power or replicability. In our commentary, we draw attention to the fact that there is no direct, fixed or one-to-one relation between reliability and power or replicability. More specifically, we argue that when adopting an experimental (rather than a correlational) approach, researchers strive to minimize inter-individual variation, which has a direct impact on sample based reliability estimates. We evaluate the strengths and weaknesses of the LeBel and Paunonen’s recommendations and refine them where appropriate.

Keywords

• replication
• experimental design
• Implicit Measures
• power
• Reliability