This paper describes a new method for estimating pasture yield in which the yields of random quadrats are rated with respect to a set of reference quadrats preselected in the field to provide a scale which is available for reference throughout the sampling. Three alternative calibration procedures are described based either on the dry matter yields of the reference quadrats or on the yields of quadrats rated and cut either during or at the end of sampling. The second and third procedures are shown to have advantages over the first. When testing the method with four observers and two types of calibration scales, estimates of mean yield were generally within 4 per cent of the actual mean with a maximum difference of 8 per cent. Methods are given for calculating the number of samples required to obtain minimum variance for the estimate of mean yield for a given cost in terms cf time spent in the field.