The objective of this study was to estimate the marginal productivity in coffee research carried out by the University of Puerto Rico Agricultural Experiment Station. Time series data from 1970/71 to 1985/86 were used in the analysis. A multiple regression analysis was conducted with coffee production as the dependent variable and number of "cuerdas" harvested, fertilizer, labor, rainfall and research investment as the independent variables. Fertilizer and rainfall variables were lagged one year and research investment eleven years. Three types of equations were applied to the data: lineal, logarithmic and semi-logarithmic. Results showed that the logarithmic equation (Cobb-Douglas) was the most statistically significant. Using the logarithmic equation with three independent variables, we estimated the research investment marginal productivity at 1.71. This means that an investment of $1.00 in research produces 1.71 cwt coffee. The labor marginal productivity was estimated at 6.69 and rainfall at 1943.43. The results were calculated on the basis of the geometric mean.