Five dimensions—wing length and width, thorax length, head width and tibia length—have been measured on samples of twelve populations of Drosophila subobscura taken from different parts of the species range, extending from Scotland to Israel. The populations had been started from thirty or more pairs of flies and maintained in the laboratory for eight to eleven generations. They were reared for measurement under standard conditions, so that any differences between them must be of genetic origin. The localities from which the populations came ranged over 25° of latitude and 15° F. in mean annual temperature.

The population means of the five dimensions all showed strong positive correlations with latitude and slightly lower negative correlations with mean annual temperature. There is, in consequence, a very uniform cline in the five dimensions, each increasing as we move northwards. This pattern differs from that found for essentially the same dimensions of D. robusta by Stalker & Carson (1947), where all the correlations with latitude are lower and those for head and thorax size are of opposite sign in the two species.

A partial correlation analysis and reference to selection experiments suggests that there are two distinct groups of genes involved in these clines: (1) a group of genes causing increase in relative wing and leg size, and responsible for the positive correlation of these dimensions with latitude in both species, and (2) a group of general size genes causing correlated changes in all dimensions, which have been selected in opposite directions in the two species, with the result that head and thorax size increase in subobscura but decrease in robusta as latitude increases. The ratio of wing length to thorax length has a high positive correlation with latitude in both species.

Further analysis of the data available on both species brings out the following points:

1. In subobscura there is considerable variation between localities remaining for all dimensions when the correlations with latitude are eliminated. This component of variance is much larger in proportion to the within-locality variance in subobscura than in robusta.

2. Comparison of the percentage regression coefficients of dimensions on latitude or temperature shows that individual dimensions change more rapidly in proportion to both geographical indices in subobscura, but the ratio of wing to thorax length ?changes more rapidly in robusta.

3. The ‘environmental’ regression of dimension on temperature of rearing in the laboratory, measured only for robusta, is very much higher than the corresponding genetic regression between populations on mean annual temperature for all the robusta dimensions and different in sign for head and thorax size. But this environ mental regression is uniformly only about twice as high as the corresponding genetic regression for each character in subobscura. Unexpectedly, the environmental and genetic regressions on temperature of the wing/thorax length ratio are equal in robusta.

4. D. subobscura is about 20% smaller in head and thorax size, and 26% less in wing width, than robusta, when both species are reared at 18° C.

5. The basic within-culture variances of the five dimensions are 50–100% greater in subobscura than in robusta for all dimensions. Variability in D. melanogaster is about the same as in D. robusta.

Some of the possible implications of these results are discussed.