Abstract

In this thesis, I use an artificial selection experiment and multivariate simulation modelling to study some basic questions concerning the evolution of sexual size dimorphism (SSD). Fecundity selection is often suggested as the main causal factor underlying the prevalence of female-biased SSD, but this assumption has not been empirically tested. I selected female Drosophila melanogaster for increased or decreased fecundity for 20 generations, and measured the effect on SSD in three morphological traits. SSD generally increased with selection for increased fecundity, but showed no consistent trend with selection for decreased fecundity. These results support the general hypothesis that SSD can evolve rapidly in response to selection for increased fecundity. SSD can evolve for a number of reasons. The two main causes are thought to be sexual selection on males, and natural selection favouring different trait optima in the two sexes. Lande (1980a,b) has produced analytical models that can be used for predicting the change in SSD through either of these mechanisms. Although these models are often cited, they have never been adequately tested, either empirically or through simulation modelling. They rely on a large number of simplifying assumptions, and their robustness to violations of these assumptions is largely unknown. In this thesis, I present results from stochastic simulation models designed to test the effects of mutational variance assumptions, finite populations, and finite numbers of loci on the robustness of the analytical models predictions. The quality of the predictions depends on the nature of allelic distributions in the original population. If allelic effects are approximately normally distributed, the predictions can be very accurate. If, as is likely, allelic effects have a leptokurtic distribution, Lande's equations underestimate the rate of response and correlated response, and overestimate the time required for the trait means to reach their equilibrium values. Predictions for the magnitude of SSD at equilibrium can be very accurate for weak sexual selection. However, with stronger sexual selection the total response is greater than predicted. The results suggest that genetic correlations constrain both the short-term and long-term evolution of SSD less than predicted by the Lande model.