Although strategic management of R&D portfolios is common practice in private sector R&D, government R&D management tends to be more discrete and ad hoc, focusing on generating maximum output from individual projects. Often, there is no clear notion of the desired public sector output. Whereas private sector R&D evaluation is generally straightforward, with the function of R&D being measured in terms of a company's internal return on investment, the benefits of public-sponsored R&D tend to be more diffuse with respect to both type and impact. Drawing from evidence developed in twenty-four case studies of R&D projects sponsored by the Department of Energy's Office of Basic Energy Sciences, one of the primary sources of funds for basic research in US universities and national laboratories, we examine implications of R&D project impacts for a sort of 'portfolio management' of government-sponsored R&D. Although the meaning of 'portfolio' is not the same in public sector R&D management, it is nonetheless possible to think strategically about projects and their cumulative impact. Indeed, it is clear that many government R&D managers already do so. In this study, we contrast two types of 'portfolio': (1) R&D output portfolios (focusing on one type of scientific output, such as, for example, fundamental knowledge or technology development); and (2) a balanced portfolio that considers both R&D outputs and 'scientific and technical human capital', the capacity created by R&D projects. The case-study evidence shows different approaches to achieving each portfolio type. An analysis of the relation of project attributes to output types focuses on aspects of projects potentially under the control of strategic public managers, including magnitude of funding, degree and type of management oversight, and interorganizational and intraorganizational linkages. Each of these variables affects the type of project outputs obtained. From the results, we suggest that a balanced approach to government R&D portfolio management is appropriate for many government agencies. That is, government managers may wish to consider the extent to which their projects both produce traditional outputs such as articles and patents, as well as provide contributions to scientific and technical human capital.