Over the past 10 years considerable empirical work has been reported on the stochastic description of rock mass fracturing, and on the statistical design of joint surveys. This has led to consistent conclusions on the distributional properties of such discontinuities, and is beginning to lead to improved survey designs. Two of the strongest conclusions appear to be the exponentiality of the distribution of spacings between discontinuities when measured by their intersections with sampling lines, and the lognormality of discontinuity trace lengths as observed in outcrops. Consistent conclusions on the form of orientation distributions appear more elusive. Sampling biases in joint surveys now seem more pervasive than was earlier thought. In addition to the well-known orientation bias in sampling from two-dimensional outcrops, proportional length bias in which larger discontinuities are sampled with increased probability, and censoring biases in which larger discontinuities are often only partially observed, complicate statistical inferences. These results are reviewed against a recent study involving some 15,000 data.