The Effects of Population Structure on Responses to Artificial Selection: An Investigation of the Shifting Balance Theory

Abstract

Sewall Wright's Shifting Balance Theory, which postulates that evolution will be most rapid in populations subdivided into numerous small, semi-isolated demes, was evaluated by comparing responses to selection for increased adult bodyweight of 'D. melanoccaster' in three population models. Two were models previously evaluated (sub-lining with crossing of selected sub-lines at intervals, and a circular stepping-stone model), but which had not shown any advantage in subdivision. The third was a new model of Wright's recommended structure, and included excess diffusion from demes with higher phenotypic means to those with lower means every generation. Responses in these models were compared with those obtained by simple mass selection in a single large population. As reported in previous studies, no clear advantages in response were obtained in any of the subdivided models. In one replicate of the new "Wrightian" model however, the pattern of responses suggested the presence of a major non-additive effect producing extremely heavy flies. This effect spread throughout the system of semi-isolated demes comprising this treatment in a manner similar to that described by Wright for the operation of the Shifting Balance Process. The genetic basis of this effect was investigated by offspring-parent regressions with the effect present and absent, by crosses with unselected flies to produce F₁ and F₂ generations, and by attempting to map the gene(s) underlying the effect by chromosomal substitution techniques. However, no clear description of the effect was obtained. In addition to the selection programme, electrophoretic surveys of the experimental populations were conducted. These provided information on levels and partitioning of allozymic variation between and within demes/population units. The description of genetic structuring provided by this data was similar to that based on partitioning the phenotypic variance in bodyweight. Results obtained suggested that models used to evaluate subdivided populations both here and in previous studies, do not produce sufficient genetic differentiation to support inter-deme selection, at least on a simple additive basis. Finally, the relevance of these results to wider understanding of the Shifting Balance Theory is discussed. It is concluded that further evaluation of the Theory should be based upon computer simulation. This approach could be used to define necessary conditions for the operations of the Shifting Balance process, and thus provide a firmer basis for both experimental designs and recommendations regarding structuring of domestic and wild populations.