Can a Good Theory Be Built Using Bad Ingredients?

Abstract

The replication crisis threatens to seriously impact theory development in the cognitive, behavioral, and social sciences. We canvas three desiderata of scientific theories (explanation, prediction, and unification) and argue that the extent to which failures of replication prove problematic depends on the primary purpose of a theory. If the aim is to explain how nature works, then accuracy—and thus replicability—of the findings on which the theory is built is essential. If the aim is to predict outcomes, then replicability of findings from which the predictive model or theory is built is only important as far as it affects the reliability and accuracy of the predictions. If the aim is to unify and organize disparate findings, then the replicability of findings plays a non-essential role. The result is that a multifaceted and nuanced perspective is required to assess the value of replicability and the need for replication studies. Specifying a theory’s purpose and background commitments should clarify the debate on replication and contribute to better theory development in the cognitive, behavioral, and social sciences.