Using unique longitudinal quantitative and qualitative data, this paper examines the role that low learning plays in driving dropout in Ethiopia, India, Peru and Vietnam. Regression analysis using IRT-linked test scores and data on schooling attainment and dropout shows a strong, significant association with one standard deviation higher test scores associated with 50% lower odds of dropping out between the ages of 8 and 12, and a similar association between the ages of 12 and 15. Qualitative analysis indicates a direct relationship between low learning and dropout, with children and parents choosing to discontinue school when they realise how little is being learned. Qualitative findings also show that low learning interacts with and exacerbates more proximate causes of dropout, with low learning often contributing to choices of early marriage (for girls) and of leaving school to work (for both genders), with families making practical decisions about which options will best provide for children in the long run. Finally, learning, work and poverty often interact, as the need to work to help provide for the household reduces the opportunities to learn, and low learning tilts the opportunity cost of time in favour of working. These findings suggest that low learning may play a larger role in dropout decisions, by underlying and interacting with other causes, than has been typically recognised.