This paper examines the barley and wheat breeding programmes of the Plant Breeding Institute (PBI), which was the most successful public plant breeding institute in the UK, until privatization in 1987. The PBI's shares in barley and wheat seed sales are explained, showing that the success with barley was largely a matter of serendipity, whereas the wheat programme followed a more normal pattern. For wheat, the causal chain, or recursive, model decomposes the well-documented link between research expenditures and increases in agricultural productivity into three stages. These are the effects of R&D expenditures on basic research output, measured by publications, the effect of publications and applied R&D expenditures on trial plot yields, and the diffusion of the trial plot technologies, which raises yields on farms. Applying the model to the PBI's wheat varieties allows estimation of the lag structures. In contrast to the results for aggregate agricultural research, for a single plant breeding programme alone there is a considerable lead time before there is any response, followed by a lag distribution only a few years long. The returns to the R&D investments are calculated from the causal chain model, from single equation estimates and by evaluating the yield advantage of the PBI varieties. All three approaches give consistent results, which show that the returns to barley and wheat alone were sufficient to support the entire PBI budget and still give rates of return to applied research of between 14 and 25%. The return to the basic science expenditures of the John Innes Institute has a lower bound of 17%, but must have been even higher than for the PBI if the other Institutes were taken into account. The paper concludes by commenting on the effects of the privatization of the PBI. © 1998 Elsevier Science B.V. All rights reserved.