The relationship between yield stability and populations containing various numbers and combinations of diverse homozygous and homogeneous lines was examined in an environment in which considerable variability occurs among and within growing seasons. Two groups (1,2), each containing 15 populations (4 pure lines grown singly and 11 multilines consisting of mechanical mixtures of all possible 2,3 and 4 way combinations of these 4 pure lines) were tested in each of 5 consecutive years. The pure lines in group 1 and 2 were selected on the basis of previous yield and yield variability respectively. In group 1, no significant differences were found among the 5 year means or the deviation mean squares of the 15 populations but highly significant differences among regression coefficients were present. The regression coefficients of the 4 pure lines differed considerably, indicating that this stability parameter was genetically influenced. The regression coefficients of the multilines tended towards unity regardless of the regression coefficients of the pure lines involved. In group 2, highly significant differences were found among the 5 year means, regression coefficients and deviation mean squares of the 15 populations. No consistent, predictable pattern was apparent between the mean and stability values of the pure lines and the multilines in which they were included. Differences between the 2 groups suggest that higher yield and greater stability result in the multilines if relatively high yielding pure lines are selected for inclusion. In general, the results indicated that multilines containing a number of diverse homozygous and homogeneous pure lines have satisfactory yields and enhanced yield stability. However, adequate testing of the pure lines and potential multilines over a broad range of environments is essential to determine desirable combinations.