The research input-output relationship is quantified by use of a unique panel data set of research expenditures (1963-75) and aggregate publication output (1970-75) for each of the 48 contiguous U.S. state agricultural experiment stations. Year-to-year fluctuations in research expenditures showed little systematic influence on research output, while on-average or longer-run differences in research expenditures between states appear to influence research performance in a fairly systematic manner. Using citation performance to adjust publication output for differences in scientific quality increases the research expenditure output elasticity by around 25%, while also increasing the "mean" agricultural research gestation lag from 2.8 to 3.4 years. Residual state-specific effects, measuring relative research efficiencies, do not appear to be correlated with levels of research expenditures.