In this paper we assess the IMF approach to economic reform in developing countries. The impact of IMF program participation on economic growth has been evaluated empirically in a cross-country literature, with little evidence of IMF programs having been successful. This suggests that a fresh approach is in order. However, the cross-country approach is unlikely to provide a sound basis for drawing clear conclusions, so we review IMF programs from a different perspective, involving a broader literature on development strategy. In particular, it is widely accepted that a common characteristic of IMF programs is a high degree of policy rigidity. This is in contrast with studies which hold that unleashing an economy's growth potential hinges on a set of well-targeted policy interventions aimed at removing country-specific binding constraints. The process of locating constraints that bind involves growth diagnostics and policy trialing. This approach maintains that not all distortions are equally important and, by extension, not all policy reforms. From this point of view, IMF programs based on a list of standard conditionalities will not accomplish much. But policy trialing is more relevant to actors and entities with a broader, and more microeconomic, focus such as national policymakers and the World Bank. It is in choices among competing projects and programs that trial and error is most likely to be necessary. Nevertheless, reforms of the IMF such as the "streamlining initiative" should start from a good understanding of the reasons for adherence to policy orthodoxy. We discuss underlying institutional and organizational reasons for policy rigidity and consider some suggested reforms.