Analysts of agricultural research policy in the United States of America have claimed that the overall level of public investment in agricultural research is less than what would be socially optimal, that the present composition of public research investment is excessively myopic in that too little basic research is performed relative to the level of applied research, and that the allocation of research resources among commodities is inconsistent with economic efficiency. A non-linear optimal growth model of the U.S. economy was developed to analyse these propositions. Strong support was found for the claim that the overall level of investment has been inadequate. No support was found for the contention that basic research has been relatively underfunded compared to applied research. Weak support was found for the view that crop research has suffered from more acute underfunding than has livestock research.