Using firm data disaggregated by industry, we establish a set of regularities in the distribution of firm R&D intensities within manufacturing industries. We show how a simple probabilistic process, in which chance influences a key unobserved determinant of R&D and firm size conditions the returns to R&D, can account for these regularities and other features of the distributions. The model provides a unified, noncausal explanation of a series of long-observed relationships across mean R&D intensity, market concentration, and the coefficient of variation. It also offers a novel explanation for the inverse relationship between R&D productivity and firm size.