in our imaginary case, with average weights 10, 12, 14 grains respectively, and if the range of fluctuation of each of these pure lines is 12 grains, then our curve must be represented as made up of the three components A fluctuating between 4 and 16 with a mean of 10 B " " 6 " 18 " " 12 C " " 8 " 20 " " 14 [Illustration: FIG. 31. Curves to illustrate the conception of pure lines in a population.] as is shown in Fig. 31. A seed that weighs 12 grains may belong to any of these three strains. It may be an average seed of B, or a rather large seed of A, or a rather small seed of C. If it belongs to B its offspring will average 12 grains, if to A they will average 10 grains, and if to C they will average 14 grains. Seeds of similar weight may give a different result because they happen to be fluctuations of different pure lines. But within the pure line any seed, large or small, produces the average result for that line. Thus a seed of line C which weighs 20 grains will give practically the same result as one that weighs 10 grains. On this view we can understand why selection of the largest seed raises the average weight in the next generation. We are picking out more of C and less of A and B, and as this process is repeated the proportion of C gradually increases and we get the appearance of selection {163} acting on a continuously varying homogeneous material and producing a permanent effect. This is because the interval between the average weight of the different pure lines is small compared with the environmental fluctuations. None the less it is there, and the secret of separating and fixing any of these pure lines is again to breed from the individual separately. As soon as the pure line is separated further selection becomes superfluous.