On the Reliability of Reliability Estimates in Infant Research

In this commentary on Byers-Heinlein, Bergmann, and Savalei’s (this issue) call to consider reliability in infancy measures, we note that the combination of highly unreliable measures and typically small sample sizes means that most reliability estimates based on single infant studies are, themselves, highly unreliable. To achieve reasonably precise reliability estimates requires either high population reliability or sample sizes several orders of magnitude larger than those typically seen in infancy research. Thus, we caution researchers to always report confidence intervals around reliability estimates, and to be circumspect when interpreting those estimates.