Through global analysis and country-level case studies, this report analyzes the efficiency of investments in wheat improvement research at a disaggregated level and explores a range of options for restructuring research programs to enhance efficiency. Particular attention is devoted to evaluating "spillovers" (i.e., benefits which flow from one program to another) that result from large research systems with the potential to exploit economies of size and scope. The impacts of such spillovers on both research productivity and research strategy are explored, as are the relative roles of international agricultural research centers (IARCs) and national agricultural research systems (NARS) in generating technology. The homogeneity of agroecological environments across developing countries is clarified and a megaenviornment classification system is described. The authors address the question of whether research managers should suppress genotype-by-environment (GxE) interactions and develop widely adapted varieties or exploit GxE interactions and develop specifically adapted varieties. Econometric evidence is presented to suggest that broadly adapted varieties may be more robust and create greater spillover than has previously been reported. Research costs and intensities in developing countries are shown to be of the same magnitude or higher than in industrialized countries because the former have a large number of scientists per research program combined with a smaller mandate area for each program. Evidence suggests that many countries or regions within a country are investing more than is economically justifiable in wheat improvement research, either because of the small size of their mandate area or because they could capture research spillins at lower costs-or, more commonly, both. A cost benefit framework is used to assess the threshold levels of wheat production in a mandate region required to justify a breeding program rather than a testing program. The results indicate that many research programs could significantly increase their efficiency by reducing their research programs and screening varieties developed elsewhere. A case study from India reveals that investment inefficiencies at the sub-national level have frequently been underestimated and that large nationally mandated programs have a comparative advantage in generating successful technologies across wide areas. A case study in Australia clarifies the role of spillovers in industrialized countries. The report has important implications at the conceptual level in the methods used for research evaluation and at the policy level for decisions on crop improvement research.