If the units are sufficiently small it is practical to measure the crop on the whole area. Steiner [1987] recommends plots of 30 m 2 for on-farm cereal and arable legume variety trials. In such trials the high level of researcher control means that variation in crop yields is minimised and sufficient precision can be obtained from small plots. It is relatively easy to harvest and weigh the crop from such plots under carefully supervised conditions. However, if economic analysis, including labour input, is contemplated, then plots of up to 1000 m 2 are needed [McIntire and Fussell, 1989].