In the past 15 or so years, the "evidence-based medicine" (EBM) framework has become increasingly institutionalized, facilitating its transfer across the globe. In the late 1990s, the basic principles of EBM began to have a marked influence in a number of non-clinical public policy arenas. Policy-makers working in these areas are now being urged to move away from developing policies according to political ideologies to a more legitimate approach based on "scientific fact," a process termed "evidence-based policy-making" (EBPM). The conceptual diffusion of EBM to non-clinical arenas has exposed epistemologically destabilizing views regarding the definition of "science," particularly as it relates to the demands of global versus national/sub-national policy-making. Using the maternal and neonatal subfield as an ethnographic case-study, this paper explores the effects of these divergences on EBPM in 5 developing countries (Bangladesh, Burkina Faso, Ghana, Malawi and Nepal). In doing so, our analysis aims to explain why EBPM has thus far had a limited impact in the area of context-specific programmatic policy-development and implementation at the national and sub-national levels. Results highlight that the political contexts in which EBPM is played out promote uniformity of methodological and policy approaches, despite the fact that disciplinary diversity is being called for repeatedly in the public health literature. Even in situations where national EBPM diverges from international priorities, national evidence-based policies are found to hold little weight in countering global policy interests, which some informants claim are themselves legitimated, rather than informed, by evidence. Informants also highlight the way interpretations of research findings are shaped by the broader political context within which donors set priorities and distribute limited resources - contexts that are driven by the need to provide generalisable research recommendations based on scientifically replicable methods. Added to this are clear rifts between senior and junior-level experts within countries that constrain national and sub-national research agendas from serving as tools for empowered knowledge production and problem-solving. We conclude by arguing for diverse forms of research that can more effectively address context-specific problems. While such diversity may render EBPM more conflict-ridden, debate is by no means an undesirable characteristic in any evolving system of knowledge, for it has the potential to foster critical insight and localized change.