The methodology of using a visual estimation technique to assess dry matter yield of pastures was studied in eight experiments. Various aspects were examined, viz: the shape of the relationship between actual and visually assessed yield; whether observers could maintain consistency over a period of hours; the effects on consistency of prediction of the sequence of yield of pastures; whether a wide range in botanical composition could be accommodated; whether green and dry pasture could be assessed with equal precision. A linear relation, or occasionally a curvilinear relation, was obtained over a large range in yield. No consistent bias due to immediate previous values in a sequence was found. However, it was found that, without training, observers over-estimated the effects of height and under-estimated the effects of density in yield. In practice, trained observers gave much more accurate regressions; for green annual pastures at all growth stages, dry annual pastures, and irrigated lucerne, lovegrass and pangola grass, r2 values were greater than 0.70 on all but 11 per cent of 126 occasions.A wide range of botanical composition was able to be included in a single calibration, i.e. ryegrass, capeweed, sub clover and erodium. Recommendations are made for routine use of the technique in the field.