1957ReseaRch W hen testing selection candidates over multiple environments, uncertainty in the estimates of genotype values increases with the magnitude of G E. This increases the difficulty of identifying superior genotypes and compromises genetic progress from selection (e.g., Annicchiarico, 2002;Bos and Caligari, 2008; DeLacy et al., 1996a,b). A better understanding of GE effects within a MET testing regime allows a reevaluation of resource allocation and selection strategy in a breeding program. The type and extent of G E is of particular interest to plant breeders as the characterization of environments will help, in part, to define selection strategies. For example, measures of quantitative G E (heterogeneity of variance or the scale ABSTRACT Differences in trait responses of genotypes across environments, or genotype environment interactions (G E), hinder the progress of genetic improvement. Characterization of these effects helps to determine breeding strategies and improve resource allocation in a cultivar development program. This study used historical multienvironment trial (MET) data (34 trials in five locations) for the analysis of marketable yield of advanced selections in a New Zealand potato (Solanum tuberosum L.) breeding program. A factor analytic (FA) model was used for the analysis of these MET data. Contrasts based on the environmental loadings were observed between the program's main trial locations in the North Island (pukekohe) and the South Island (Lincoln), indicating that these locations optimized differentiation between genotypes in terms of G E effects. Genetic correlation estimates between trial environments were mostly moderately high (>0.5) to high (>0.8) and ranged from zero to positive with a maximum coefficient of 0.97, suggesting that quantitative (rescaling) rather than qualitative (crossover) G E effects were of greater importance. A number of newly developed varieties were shown to have higher genetic yield potential than older and established commercial cultivars but did not necessarily show better yield stability over the locations tested.