The expected efficiencies of some methods of selection of components for inter-genotypic mixtures

The gains in yields of mixtures expected to follow various methods of selection of components from populations of randomly constituted mixtures are formulated in terms of a statistical model and a model of first-order inter-plant competition. The types of selection investigated include that among whole mixtures and among groups or individual components on a within-mixture or unrestricted basis, or on the basis of the yields of sets of mixtures to which the group or individual is common. In all cases the sizes of mixture used for selection and for measurement of gain may differ. While evaluation of components in groups or whole mixtures allows selection for component interactions, gains are lower overall because of the reduction in variance caused by grouping. Gains due to interaction are lost if the components are pooled after selection, as in a population improvement programme. Individual selection carries some risk of negative gains, but these are reduced if assessment is made on an unrestricted rather than within-mixture basis. When second and higher order competitive interactions are absent, monoculture assessment is expected to be an efficient means of selection of components for binary and tertiary mixtures.