Estimated rates of return to research are distorted by problems of attributing the credit for particular research results, or for particular research-induced productivity increases, among research expenditures undertaken at different times, in different places, and by different agencies. A comprehensive assessment of the evidence from past economic evaluations of the returns to agricultural R&D indicates that studies generally report high rates of return, with enormous variation among studies, but that much of this evidence has been tainted by inadequate attention to attribution problems. This paper raises these concerns in a general way and illustrates their importance with reference to two particular types of attribution problem. First, we consider the problem of accounting for locational spillovers in attributing varietal-improvement technology among research performers, using US wheat varieties as an example. Second, we consider the temporal aspects of the attribution problem using the specification of research lags in econometric models to illustrate the problem of attributing aggregate productivity gains to research expenditures made at different times.