This paper re-analyses three experiments for which results have previously been published. A simpler model for aggression, based only on the proportions of the component yields of the mixed plots, is adopted. This greatly simplifies the analysis, which requires only a straightforward application of least squares, and gives much more accurate estimates of competitive power. Omission of some combinations from the mixtures included in an experiment presents little extra difficulty.

Various secondary matters, some of which have implications for the design of experiments of this type, are discussed in the paper. Graphical plots which throw light on the effect of sole yield differences on competitive power are exhibited. A logarithmic transformation of the data, often used to equalize the variance, was found to lead to underestimation of enhancement.