With the advent of the United Nations Education First initiative, and considering the continued efforts to focus on the quality of education in low-income countries, there has been a renewed interest in the improvement of learning (as distinct from school attendance) in poor and marginalized populations (Wagner, Murphy, and de Korne, 2012).1 There is a large and diverse empirical research base in the area of human learning. Yet much of the available research is substantially limited by boundary constraints of various kinds. Most prominent among them is the limited ability to generalize from findings in one population context to other distinct population contexts. Similarly, research methods may vary greatly between one set of studies and another, making it difficult to discern whether the findings vary due to the methods or to other factors. These are classic problems in the social sciences, and inevitably lead to substantive trade-offs in how policy development takes place in education.