The commercial value of basic knowledge depends on the arrival of follow-up developments mostly from outside the boundaries of the inventing firm. Private returns would depend on the extent the inventing firm internalizes these follow-up developments. Such internalization is less likely to occur as knowledge becomes more general. This motivates the historical concern of insufficient private incentive for basic research. The present paper develops a novel empirical methodology of identifying unique patterns of knowledge flows (based on patent citations), which provide information about whether 'spilled' knowledge is reabsorbed by its inventor. Using comprehensive data on the largest 500 inventing firms in the US the classical problem of underinvestment in basic research is confirmed: spillovers of more general knowledge (and in this respect, more basic) are less likely to feed back to the inventing firm. This translates to lower private returns, as indicated by the effect of the R&D stock of the firm on its market value.