This paper presents an empirical and theoretical evaluation of a step-point transect procedure for estimating the relative and absolute abundance of kangaroos in sheep-grazed rangeland paddocks. The method assumes that the proportion of kangaroos in the total (sheep + kangaroo) population can be estimated from their proportional representation in the dung. The actual population of kangaroos can then be estimated if the population of sheep is known. Proportional representation in the dung is estimated by the probability that dung of a given species will lie closest to a 'random' point. For operational simplicity, sample points are not strictly random but are located systematically along walked transects. The procedure was applied in seven paddocks spread throughout the Western Division of NSW and south-west Queensland where actual kangaroo and sheep populations were determined by ground survey and from station records, respectively. Dung was also collected from sample transects to allow comparison of step-point population estimates with corresponding estimates based on dung weight or pellet counts. A theoretical estimator of the kangaroo population based on step-point data was also derived and compared with actual populations estimated from ground surveys. Results indicate that the step-point transect procedure should produce acceptable estimates of both relative and absolute kangaroo populations in sheep-grazed rangeland paddocks. The appropriate equations, and minimum sample sizes, are provided. The step-point transect technique has the practical advantage that it avoids the tedious sampling procedures inherent in count- or weight-based methods. Further, in contrast to these methods, it may be expected to become both more reliable and less time consuming as the density of dung increases, thus providing the greatest operational advantage under circumstances where reliable estimation is most required.