Steve Perrin argues that animal models of human disease should be better characterized to avoid false-positive results (Nature 507, 423–425; 2014). Poor design, execution and reporting of studies are pervasive contributors as well.

For example, randomization and blinding in animal studies is pitifully infrequent (see, for instance, J. P. Ioannidis et al. Lancet 383, 166–175; 2014). Randomly allocating animals to experimental and control groups makes the experimental groups as similar as possible in other respects. Blinding promotes comparable handling and measurement by experimenters, and publicly preregistering animal-study protocols and the outcomes to be measured would identify and reduce bias in reporting results.

Creating largely homogeneous experiments aids reproducibility and boosts statistical power, but has a cost of generalizability: the few drugs that have translated successfully from animals are effective across a broad range of circumstances (see, for example, E. S. Sena et al. J. Cereb. Blood Flow Metab. 30, 1905–1913; 2010).

Funders and ethics committees need to ensure that study designs include these features, and journals should make them criteria for publication if they have not done so already.