Policy research on African agriculture is long on prescriptions for what needs to be done to spur agricultural growth but short on how such prescriptions might be implemented in practice. What explains this state of affairs? What might be done to correct it, and, most important, how? This paper addresses these questions via a comprehensive review and assessment of the literature on the role and impact of research in policy processes. Six major schools of thought are identified: the rational model; pragmatism under bounded rationality; innovation diffusion; knowledge management; impact assessment; and evidence-based-practice. The rational model—with its underlying metaphor of a “policy cycle” comprising problem definition and agenda setting, formal decision making, policy implementation, evaluation, and then back to problem definition and agenda setting, and so on—has been criticized as too simplistic and unrealistic. Yet it remains the dominant framework guiding attempts to bridge gaps between researchers and policy makers. Each of the other five schools relaxes certain assumptions embedded within the rational model—e.g., wholly rational policy makers, procedural certainty, well-defined research questions, well-defined user groups, well-defined channels of communication. In so doing, they achieve greater realism but at the cost of clarity and tractability. A unified portable framework representing all policy processes and capturing all possible choices and tradeoffs faced in bridging research, policy, and practice does not currently exist and is unlikely ever to emerge. Its absence is a logical outcome of the context-specificity and social embeddedness of knowledge. A fundamental shift in focus from a “researcher-as-disseminator” paradigm to a “practitioner-as-learner” paradigm is suggested by the literature, featuring contingent approaches that recognize and respond to context-specificity and social embeddedness. At bottom, the issue is how to promote “evidence-readiness” among inherently conservative and pragmatic policy makers and practitioners and “user-readiness” among inherently abstraction-oriented researchers.