Merrifield (2009) provides a useful polemic about the sad state of data analysis too frequently encountered in the school choice literature. Available data come from limited policy experiments with only modest amounts of choice and competition. The effects of very modest changes in school choice on school performance are, as one might expect, small and often not statistically different from zero. Those who study school reform initiatives mistakenly extrapolate from these null effects of small policy shifts to make sweeping predictions about the ineffectiveness of broader policy shifts aimed at expanding choice and competition through creation of new institutions governing the supply of educational services. In debates about how to interpret the existing evidential record and what our priorities for collecting more empirical information about the performance of different educational institutions should be, it is difficult to cleanly separate theoretical priors from empirical evidence. Virtually all empirical investigations rely on theoretical priors to motivate the design and measurement of particular effects. However, Merrifield's call for improved empiricism is worthwhile because it can accelerate the coevolution of researchers' theoretical priors and the accumulated evidential record. I suggest that the empirics of school choice would benefit greatly from separating the issues of state subsidization of education, regulatory mandates for achieving performance goals, and the institutions--private or public--that actually deliver educational services. Furthermore, a more rigorously empirical research agenda can advance school choice debates by squarely addressing six potential pitfalls based on economic theory that arise when attempting to move toward decentralized private provision of educational services. More detailed definitions of competition together with bold, new empirical evidence are clear priorities for school choice research.