This paper discusses what economists want experimenters to do, and why, and criticises existing methods from the economists' viewpoint. The subject has received very little serious attention outside of North America and practically none, as far as I am aware, in Australia. The exception is a valuable critical survey, by Pearse, of the Department of Agriculture's experimental work in a large part of Western Australia. We are concerned in this paper only with certain types of agricultural experiments; namely (i) experiments which investigate physical input-output relationships (such as fertiliser, feeding and stocking rates) and (ii) where the technical data alone does not suffice to indicate an optimum.