Defence R&D continues to account for a very substantial share of many countries' research effort, often retaining or even increasing its role within the national innovation systems. Yet the quantitative analysis of defence research efforts and their impact is impaired by difficulties in defining defence R&D. The article studies these difficulties and focuses on the issue of identifying what constitutes, does not constitute, defence R&D. It finds that the OECD approach of defining defence R&D on the basis of the primary goals of the research is inadequate, particularly in the present context of the growing emphasis on dual-use technologies and research. It then analyses alternative approaches that could provide a more solid grounding for any systematic effort to collect international defence R&D data