Guard rows were almost eliminated from factorial fertilizer experiments on brussels sprouts, red beet, carrots and lettuce by arranging treatments systematically instead of at random. Edge effects within plots were small and largely cancelled each other out because of the stepwise progression of treatments. The possibility that natural fertility trends would lead to biased results was tested by comparing low-order and high-order treatment × replicate interactions, and it was found that such trends could be discounted. The economy of land and labour achieved permitted a thorough coverage of the response surface, with the result that far higher rates of application of nutrients than previously used were found worthwhile in some cases.