From cross-sectional data of 460 firms that responded to both the 1988 and the 1992 Dutch innovation surveys we have reexamined the causality direction between R&D and patents, using data on contemporaneous and four-year lagged patent applications and R&D expenditures. The two equations have been estimated jointly assuming a bivariate conditional distribution between the two variables, one being discrete and the other one continuous. We have experimented with different specifications of the count data for patent applications. We find that patents Granger-cause R&D in all specifications. One additional patent increases R&D four years later by 7.5%. The reverse causality from R&D to patents vanishes as soon as we depart in one way or another from the simple Poisson specification of patent counts.