Knowledge of the presence and magnitude of genotype ✕ environment (GE) interactions is important to plant breeders in making decisions regarding the development and evaluation of new cultivars. The forage yields of three cultivars and 21 experimental stains (genotypes) of bermudagrass, Cynodon dactylon (L.) Pers., were determined for 3 yr (1981–1983) at five sites. These data were used to estimate and compare methods of analysis of GE interactions and yield stability in this perennial crop. First harvest (second and third years only), regrowth, and seasonal total dry matter forage yield data were analyzed by: (i) a standard multifactor analysis of variance with locations, years, and genotypes as the main effects, (ii) linear regression of individual genotype means on the mean of all genotypes in each environment, and (iii) an orthogonal contrast analysis that partitioned the variation over environments for each genotype into sources due to years. Genotypes, locations, years, all first order interactions, and the genotype ✕ location ✕ year interaction were significant sources of variation for all yield measurements. Partitioning the GE interaction sum of squares into heterogeneity among regressions and a remainder revealed significant mean squares for both sources. Regression of genotype means on the environmental means resulted in coefficients ranging from 0.61 to 1.38,0.77 to 1.20, and 0.80 to 1.20 for first harvest, regrowth, and seasonal total yields, respectively. Mean squares for deviation from regression for these respective variables were significant for 41, 33, and 46 % of the genotypes. Genotypes differed in their response to years and locations. As stand age increased, yields of some genotypes increased while others decreased, but the response was not consistent from location to location. Use of standard analysis of variance, regression statistics, and partitioning of the year variation into linear and nonlinear components were all appropriate and helpful in characterizing performance of individual genotypes.