1. Similar changes in the body-size of Drosophila melanogaster have been achieved by different developmental pathways, especially either by altering the duration of the early exponential phase of larval growth or by influencing the growth rate in the phase which is independent of time.

2. Such changes have been effected by selecting in the same population for larger or smaller size or shorter development time on chemically defined media, deficient in alternative nutrients. Selection for larger size on media deficient in protein or choline does not involve correlated changes in the larval period, whereas selection on media deficient in RNA does. The evidence suggests that shortage of this nutrient may be uniquely favourable for promoting a correlated change between body-size and duration of the larval period.

3. Strains which differ in presence or absence of such correlation are characteristically different with respect to gene-environment interaction. In the former, the differences due to selection are generally more fully or completely expressed when the diet is changed whereas in the latter this is not so, and different, especially competitive condition, leads to a drastic reduction of the difference.

4. How far the expression of the differences due to selection are affected, when the diet is altered, is also influenced by how long selection has been carried out. In early generations, the difference is only or best expressed in the special conditions provided during selection, but later on the changes due to selection are either fully expressed or partly so, as noted above.

5. Many of the differences in gene-environment interaction between selected strains can be accounted for in terms of variation in the duration of the exponential phase. Thus two lines selected for small body-size on low RNA or low protein diets responded in different ways to the same nutritional change—one became relatively larger and took porportionately longer to develop, the other became relatively smaller and developed in a shorter time.

6. There is clear evidence from various tests in which the amino-acid composition of the diet has been altered, that the nutritional requirements in the two stages of growth are not identical and this is consistent with the evidence for considerable genetic independence as well.

7. It is proposed that the first stage of larval growth, which principally determines the duration of the larval period and may also influence body-size, is canalized. Genetic variation which can influence this stage is present in the population but contributes little to the phenotypic variation of adult size, except under special nutritional conditions as when ribonucleic acid is the sole limiting nutrient. But, at the same time, such canalization is dynamic in the sense that the absolute amount of growth which is completed in the first stage may vary with respect to diet and thereby lead to correlated variation in the duration of larval life and adult size. But individuals of an adapted population behave alike in this respect so that gene-environment interaction which leads to correlated variation in the two characters is of a very low order.

8. The canalized phase sets a limit to the potential growth in the later stage and thereby influences greatly the mean value about which such growth is equilibrated. This canalization plays a major role in the general stability of growth relations and body-size although this is normally concealed by the high level of phenotypic variation. This interpretation can account for a great variety of data and provides a rational guide to further analysis.