Although academic ability is the most important explanatory variable in studies of student learning, researchers control for it with a wide array and combinations of proxies. The authors investigated how the proxy choice affects estimates of undergraduate student learning by testing over 150 specifications of a single model, each including a different combination of 11 scholastic aptitude measures-high school grade point average (GPA) and rank and variants of college GPA and Scholastic Aptitude Test (SAT) scores. Proxy choices alone cause the magnitude of the estimated learning gains to vary by large and meaningful amounts, with increases ranging from a Cϩ to less than a BϪ or to a B. The authors found that collegiate GPA data offer the best proxy for students' individual propensities to learn economics-a result that runs counter to researchers' actual proxy choices. The results suggest that scholars should control for academic aptitude with college grades and either SAT scores or high school GPA or rank.