Methods of adjusting plots by their neighbours are discussed as a means of reducing experimental error independently of the blocking structure. Six cases are described, using from two to eight neighbouring trees with single and double covariance. The method was applied to trials of various perennial fruit crops from different parts of the world. Most trials have their experimental errors reduced, although there were exceptions. The technique was particularly successful with tea but unsuccessful with apples. Double covariance was superior to single covariance and generally the more trees used for adjustment the larger the reduction in error.