Yield data of single crosses, three‐way crosses and inbreds of Sorghum bicolor (L.) Moench, obtained in East Africa, are analyzed by previously described techniques on both an arithmetic and a log scale to obtain estimates of parameters describing yield stability over environments. Comparison of regression coefficients of yield on environmental index show that hybrids are more stable in terms of this parameter, but that there is no difference between three‐way and single crosses. In terms of deviations from regression, weak evidence is presented that indicates that three‐way crosses may be more stable than single crosses. This is interpreted as being due to population buffering. However, one single cross showed particularly low deviations, indicating that this may be a heritable characteristic. The analyses on the arithmetic scale and on the log scale are compared, with the conclusion that analysis on the log scale is preferable if varieties differ markedly in yield.