Replication, a Hallmark of Good Science: Unraveling the Factors That Predict Replication Success

Considerable discussion in recent years has focused on failures to replicate findings in the psychological literature.In a Monte Carlo simulation of the research process, we examined several characteristics of research studies that might predict replication success (i.e., when both studies show similar effect sizes), and when estimated effect sizes reflect true effects.In our simulation, a successful replication was most likely when the initial findings had already been replicated once by the original author and when measurement reliability was high.As expected, greater replication success was also associated with narrow confidence intervals around effect-size estimates.However, sample sizes (i.e., those typically found in experimental psychological research) contributed relatively little to replication success.The estimates of true effect sizes were more accurate, aligning closely with the values specified in the simulation, under the same conditions associated with replication success.We discuss our findings in terms of how changes in research practices might produce more reliable psychological research.