Acquisition of empirical Probability of Detection (POD) data involves use of controlled samples of material containing discontinuities of known characteristics, such as size, orientation, etc. For subsurface inspection, developing such a data base presents difficulties in obtaining or producing samples covering the full range of possible characteristics, and in adequately identifying those characteristics with a “referee” technique. In this paper, a method is described whereby estimates for the POD of subsurface discontinuities have been derived from a simple algorithm linking metallographic and ultrasonic measurements on a relatively small number of artificial and natural reflectors. Originally introduced solely to meet a need for mean POD, subsequent development has allowed adding confidence levels to the analysis; use of a closely related technique for providing probabilistic estimates of the size of individual discontinuities will also be described.