Much recent research has been directed at intercropping. Although the planning and analysis of these experiments often raises complicated statistical questions, many statisticians have concentrated on sophisticated graphical and numerical procedures for summarizing results, without considering what system of intercropping, if any, is on balance advantageous to the farmer and how its advantage is to be measured. This paper surveys aspects of intercropping that need examination before farmers can be advised, such as the value of produce, and the effects of intercropping on yields and costs. Any synergism of productivity occurring directly or in the longer term is a central but not necessarily decisive aspect of the problem. A recently described experiment is used as an example. The main conclusion is that experiments on these topics need to be large and well-designed, with repetition in several years, and that interpretation must allow for alternative sets of numerical values expressing values and costs.