From the general discussion of the effect of mass selection on the genotypic array of the next generation we have seen that under certain conditions mass selection can lead to homozygosity, but does not do so necessarily. Outcrossing, mutations and disadvantages of certain genotypes may prevent reaching equilibrium condition with complete homozygosity even in the simplest case of one locus with two alleles. It depends on the magnitude of these factors and the degree of heterozygosity of the variety how close to genetic uniformity we ultimately can come. The advancing of one generation has only a small effect, which becomes less as the frequency of one of the genotypes becomes less. Mass selection can reduce segregation in a variety, but only in ideal situations and only in small steps. Obviously the most effective way to promote genetic uniformity is to begin with non-segregating material obtained through careful inbreeding and within-family selection. Then, if this is available, mass selection is hardly necessary and seed collection should be done so as to prevent a return to a heterozygous condition. When only segregating populations are available, some changes can be expected if the selection intensity is low, but they will not be great and may not be noticed until after several generations. Returning to the original question of the stability of the tobacco variety, the conclusion can be drawn that when large numbers of plants from a field are selected as seedplants the changes in the next generation as a whole will be small, regardless of the selection procedure used. When 10-30 % of all plants are allowed to produce seed we cannot expect important changes. Also, when the environmental variations are as great as in shade tobacco, the possibilities of selecting against certain genotypes for a number of characteristics simultaneously becomes virtually impossible. Concerning whether or not mass selection can lead to improvements in the variety, it was explained that the selection intensity and the heritability of a character determine the selection response. Perhaps this response can be predicted in the case of one character, but it becomes difficult, if not impossible, to forecast the results when many characters are involved at the same time. Reference was also made to a selection index which has been effective in livestock breeding. Even if an index could be determined for shade tobacco, its use is not necessarily effective, as Kempthorne (1957) pointed out. The genotype-environment interaction makes questionable the choice of certain fields over others. Without experiments, such questions cannot be answered. It is likely that mass selection with low selection pressure changes a variety very little. Again, in a field where up to 30 % of the plants are selected for seed, the pressure cannot be very great and we should not expect great changes. Only when a few plants are carefully selected for certain characters should progress become noticeable, as is the experience of single plant selection in plant breeding. We conclude that mass selection for seed and plant breeding to improve a variety should not be confused with each other. Each has its own aims and methods, which are not interchangeable. Where uncertainty exists about seed, progeny tests usually are made to compare a seed with the parent seed. Bolsunov (1959) has described a number of such procedures for tobacco, though the more elaborate of them appear impractical.