Many studies investigate the relationship between R&D expenditures as an input and patents as an intermediate product or output of a knowledge production function. We suggest that the productivity of research in patent production functions has been underestimated in the literature, as scholars typically use information about R&D, i.e. the sum of research expenditure and development expenditure, due to data availability. However, in most industries only (applied) research will lead to patentable knowledge, and development happens after the initial research phase that may have led to a patent. Instead of using data on R&D, we separate the knowledge creating process into 'R' and 'D'. This data stems from R&D surveys of Belgian firms. It turns out that only the 'R' part of R&D expenditure has a significant effect on patents and that development expenditure are insignificant. Thus previous literature relying on R&D expenditure suffers from a measurement error, such that the coefficient of R&D is biased towards zero, as R&D includes a large fraction of irrelevant expenditure, i.e. development expenditure, with respect to patenting.