G. N. Atlin

In a small target region, it may be possible to exploit local adaptation to increase gains from selection. However, in a large region more extensive testing is usually possible, resulting in more precise estimation of genotype means. A correlated response model was adapted to determine if division of a large target region is likely to increase gains. Genotypic value in a large region and constituent subregions are considered correlated traits. Correlated response in a subregion to indirect selection across the undivided region, relative to direct response to selection within the subregion, is expressed as a function of heritability in the undivided region (H) and in the subregion (H_i ), and of the genotypic correlation between region and subregion means (r_G′). r_G′ depends on the magnitude of the genotype × subregion interaction (σ²_GS) relative to the genotypic variance (σ²_G). σ²_GS is the portion of the genotype × location interaction (σ²_GL) caused by local adaptation, rather than by random site‐to‐site variability in genotype means. Subdivision can increase heritability through the addition of σ²_GS to the numerator of H_i, but this may be offset by reduced replication across locations within the subregion. Modeling using variance estimates from several cereal programs indicated that, unless σ²_GL is large relative to σ²_G and at least 30% of σ²_GL is due to σ²_GS, subdivision is unlikely to increase response. These results help explain the success of breeding programs that test broadly.

Research funding for cooperators in the cereal cultivar testing system in Canada is declining, and cooperators are under pressure to reduce their testing effort. The impact of reducing the number of years, test sites, and replications on the standard error of cultivar means (SE), predicted response to selection (R), and the probability of accepting inferior cultivars in Maritime spring cereal registration trials was examined using variance component estimates. Reducing the number of replications increased SE less than reducing the number of sites or years of testing. Reducing replicates per site from four to two increased SE by no more than 0.02 t ha⁻¹ when the number of sites was three or more. Reducing the number of trial sites from five to three, however, increased SE by up to 0.06 t ha⁻¹ and reduced R by as much as 16%. The probability that a cultivar with a true yield 5% less than that of a standard would equal or exceed the standard in a 3-yr series of trials conducted at five sites ranged from 6% in 2-row barley to 21% in winter wheat. Cost reductions in the Maritime cereal registration trials should be achieved, where possible, by reducing within-site replication rather than reducing the number of trial sites. Some reduction in the risk of registering inferior cultivars can be achieved by comparing test cultivars with the mean of several standards, rather than with a single standard. Key words: Replication, genotype × environment interaction, selection response, acceptance probability, variance components, residual maximum likelihood