We would like to clarify a couple of points raised by Casey Miller and Keivan Stassun in their criticism of the US Graduate Record Examination (GRE) for selection of graduate students (Nature 510, 303–304; 2014).

As documented in a meta-analysis by independent researchers, the GRE predicts more than just first-year graduate grades: it also predicts higher-degree attainment, time to complete, research productivity and citation counts (see N. R. Kuncel et al. Psychol. Bull. 127, 162–181; 2001).

And although we acknowledge that the GRE “reflects certain demographic characteristics of test-takers”, we reject any implication that this is a fault in the test: rather, test scores reflect the reality that more educational resources are available to students from wealthier families.

We agree that graduate admissions should be based on multiple sources of information, including standardized test scores, undergraduate grades, diligence and non-cognitive factors such as 'grit'.