In order to delineate efficient grazing systems it is important to periodically assess the herbage mass at the farm level. However, the knowledge concerning both, the number of samplings that should be performed and the quadrat (size and format) that should be used in these evaluations remains incomplete. In this context, three experiments were delineated to evaluate the effect of t he number of samplings (varying from two to nine) and the dimension:format ratio of the quadrat (.25, 1.0, 2.0, and 3.0 m 2 for quadrate and rectangular formats) on the coefficient of variation in herbage mass determinations performed on tussock-forming pastures. A randomized com plete block design, following a factorial arrangement with three or four replicates, according to the experiment, was adopted. The in teraction between the number of samplings and the dimension:format ratio of the quadrat was not significant. The overall analysis of the main factors indicated that at least four samplings with a 1 m 2-quadrat (1 m x 1 m) should be performed in order to stabilize the coefficient of variation and provide a reliable estimate of the herbage mass.