UC San Diego

That a theory fits data is meaningful only if it was plausible that the theory would not fit. Roberts and Pashler (2000) knew of no enduring theories initially supported by good fits alone (good fits, that is, where it was not clear that the theory could have plausibly failed to fit). Rodgers and Rowe (in press) claim to provide six examples. Their three non-psychological examples (Kepler et al.) are instances of what we consider good practice: How the theory constrained outcomes was clear, so it was easy to see that data might plausibly have contradicted it. Their three psychological examples are flawed in various ways. It remains possible that no examples exist.