Ongoing studies in many species seek to understand the origins, architecture and consequences of phenotypic variation under normal and dysfunctional conditions, with the aim of identifying targets for intervention that can prevent, stabilize or reverse disease. Some suggest that only humans are appropriate for studying these questions and argue that candidate drug targets identified in mouse models are largely unreliable. Here, we review the vast evidence showing that mouse models continue to make fundamental contributions to our understanding of genetic principles, pathogenic mechanisms and therapeutic modalities. We propose a virtuous cycle in which the power of observational studies and natural experiments in humans are closely integrated with the rigour of true experiments in model organisms.