Replication arises in a variety of forms in scientific research. Pseudoreplication involves treating multiple measurements of the system as independent replicates. Taking multiple measurements is a good thing; treating them as independent is not. Replication of treatments within a study is definitely a good thing; it gives us greater confidence in the results and also allows an estimate of its reliability. I argue that even more important is what I have termed meta-replication: the repetition of entire studies. Besides the value that it offers by expanding the scope of inference of the results, meta-replication provides a framework in which weak or unreplicated studies have