Many studies have failed to replicate statistically significant published results. This talk addresses this issue within a general statistical framework that includes questionable research practices (QRPs). Analyses within this framework reveal that the base rate of true effects is the main factor determining the successful replication of significant results. For purely statistical reasons, replicability is low in research domains where true effects are rare. This point is underappreciated in current scientific discussions because the non-replicability of significant results is usually attributed to QRPs.