Data of uniformity trials on apple in farmers' gardens showed large localized and genetic variation, indicating that larger plots could be used with advantage. Formation of blocks on the basis of uniformity trials can be expected to increase block efficiency considerably. With only four to six years for experiments it was found better to use trunk girth as a calibrating variate for forming plots and blocks instead of losing two years for a uniformity trial. The minimum number of replications needed to obtain estimates of treatment effects with a given standard error has been calculated. There was a pronounced biennial swing in the yields as reflected by the intra-tree correlation.