The welfare of the poor turns in large measure not only on technocratic development "policies", but the effective delivery of key public services, core elements of which require thousands of face-to-face discretionary transactions ("practices") by service providers. The importance of (often idiosyncratic) "practices" was largely ignored in the 1960s and 70s, however, as planners in developing countries sought to rapidly emulate the service delivery mechanisms of the developed countries, namely standardized (top-down) "programs" managed by a centralized civil service bureaucracy. Although this approach could claim some notable successes in poor countries, it soon became readily apparent that it had failed early and often in virtually all sectors. Three common civil service reforms in the 1980s also yielded disappointing results, so in the 1990s scholars and practitioners began to tout more radical "participatory" (or bottom-up) proposals for improving service delivery. These new proposals have generated a series of unusual alliances and antagonisms in contemporary development debates. We attempt to unravel these debates by distinguishing between the original solution and eight current proposals for improving service delivery, on the basis of a principal-agent model of incentives that explores how these various proposals change flows of resources, information, decision-making, delivery mechanisms, and accountability. We briefly assess the arguments made by proponents and detractors of each approach, and suggest some of the implications of this framework for education, research, and those charged with improving service delivery.