Why most of psychology is statistically unfalsifiable

Abstract

Low power in experimental psychology is an oft-discussed problem. We show in the context of the Replicability Project: Psychology (Open Science Collaboration, 2015) that sample sizes are so small in psychology that often one cannot detect even large differences between studies. High-powered replications cannot answer this problem, because the power to find differences in results from a previous study is limited by the sample size in the original study. This is not simply a problem with replications; cumulative science, which critically depends on assessing differences between results published in the literature, is practically impossible with typical sample sizes in experimental psychology. We diagnose misconceptions about power and suggest a solution to increase the resolution of published results