R&D is slow magic. It takes many years before research investments begin to affect productivity, but then they can affect productivity for a long time. Many economists get this wrong. Here we revisit the conceptual foundations for R&D lag models used to represent the temporal links between research investments and impact, review prevalent practice, and document and discuss a range of evidence on R&D lags in agriculture and other industries. Our theory and evidence consistently support the use of longer lags with a different overall lag profile than is typically imposed in studies of industrial R&D and government compilations of R&D knowledge stocks. Many studies systematically fail to recognize the many years of investment and effort typically required to create a new technology and bring it to market, and the subsequent years as the technology is diffused and adopted. Consequential distortions in the measures and economic understanding are implied.