An experiment is described in which the liveweight gains of cattle, grazing different pasture species, were measured when grazing sequences were arranged in the form of a Latin Square. The average coefficient of variation for the 6 “animal” Latin squares used was 28·5%, with a range of 15·1 to 42·6%. Significant improvements in liveweight gain were demonstrated arising from the inclusion ofStylosanthes gracilisin swards ofChloris gayanaandPanicum maximum.The problems of conducting and interpreting trials of this type are discussed in relation to the need for data on the economic value of pasture species in the tropics. Latin square grazing designs offer a useful means of measuring animal production from pasture, with the advantage of requiring relatively small numbers of animals in each trial, but trials should be repeated at different