Four experiments were conducted to assess the performance of ranked set sampling relative to random sampling for the estimation of herbage mass and clover content in grazed swards. The expected theoretical efficiencies were not observed due to the method of selection of quadrats and the nature of the distribution of herbage mass. Nevertheless there should be worthwhile improvement in precision over random sampling, provided that the quadrats within a set are as well spaced as possible, allowing, if necessary, for visual comparison.