Replication and generalization in evolutionary ecology

Russell Bonduriansky 

Several authors have expressed concern about replicability in evolutionary ecology. But what does replicability mean in this field, and how can replication enable generalization? Evolutionary ecologists investigate probabilistic processes across a vast diversity of unique historical entities. I argue that the aims and subject matter of evolutionary ecology complicate inference and necessitate distinct concepts of replication and replicability grounded in biological diversity. I define narrow replicability as the potential to obtain consistent results within a species, and broad replicability as the potential to observe consistent patterns across species possessing relevant attributes. I argue that broad replicability is essential to progress in evolutionary ecology because broad replicability enables generalization. Furthermore, many outcomes in evolutionary ecology can result from more than one underlying mechanism, and tests of a single, simple prediction against a statistical null hypothesis can therefore rarely enable robust inference. Complex predictions that discriminate alternative biological models can enable “severe tests” and provide a better measure of replicability. Likewise, biological diversity and complexity often preclude quantitative predictions, necessitating qualitative tests of hypotheses and qualitative measures of replicability.