We used a disaggregate approach to examine investment efficiency of wheat breeding research in India. India's total research effort comprizes 20 research programs spread across 50 experiment stations. A technology spillover matrix was constructed for both potential and actual spillovers. Spillovers and free-riding were dominant characteristics of technical change during the period studied. Although the aggregate rate of return to wheat improvement research in India was estimated to be 55%, eight programs were found to have earned a negative rate of return when spillins were taken into account. Research output is concentrated on a few strong programs. The two strongest programs generated 75% of all the technical change benefits, even though they claimed just 22% of research resources. These two programs include a significant degree of overlap, while on the other hand many farmers were not reached by any of the programs- 56 and 78% of rainfed and durum area, respectively, in 1990 was still sown with pre-1976 varieties.© 2001 Elsevier Science B.V. All rights reserved.