Various methods of evaluating phenotypic stability have been proposed; however, no single method can adequately describe cultivar performance. The objectives of this study were to integrate a number of methods of evaluating stability and to use this approach for cultivar selection. These objectives were considered in the context of the broad-based oilseed rape cultivar (Brassica napus spp. oleifera) evaluation system currently used in western Canada. Regression analysis was used to assess cultivar response to environments. Cluster analysis was used to assemble cultivars into groups with similar regression coefficients (b i ) and mean yield. Three parametric stability parameters, years within locations mean square (MS; Y/L), Shukla's stability variance (σ i (2) ), and Francis and Kannenberg's coefficient of variability (CV i ), were compared to determine which method would be most suitable for selection of oilseed rape cultivars from within clustered groups. Yield data from three cultivars and six breeding lines that had been tested for 2 years at 26 locations in the Western Canola Cooperative Test 'A' were used for all calculations. The cluster analysis was successful in identifying commercially acceptable breeding lines. The parameter MS i Y/L was considered to be more appropriate than either CV i or σ i (2) , because it measured only the unpredictable portion of the genotype x environment interaction and was independent of the other cultivars in the test. The use of cluster analysis to group entries with similar b i values and mean yields, followed by selection for stability within groups, was advocated.