Sex allocation theory has assumed that hermaphroditic species exhibit strong genetically based trade-offs between investment in male and female function. The potential effects of mating system on the evolution of this genetic covariance, however, have not been explored. We have challenged the assumption of a ubiquitous trade-off between male and female investment by arguing that in highly self-fertilizing species, stabilizing natural selection should favor highly efficient ratios of male to female gametes. In flowering plants, the result of such selection would be similar pollen:ovule (P:O) ratios across selfing genotypes, precluding a negative genetic correlation (r_g) between pollen and ovule production per flower. Moreover, if selfing genotypes with similar P:O ratios differ in total gametic investment per flower, a positive r_g between pollen and ovule production would be observed. In outcrossers, by contrast, male- and female-biased flowers and genotypes may have equal fitness and coexist at evolutionary equilibrium. In the absence of strong stabilizing selection on the P:O ratio, selection on this trait will be relaxed, resulting in independence or resource-based trade-offs between male and female investment. To test this prediction, we conducted artificial selection on pollen and ovule production per flower in two sister species with contrasting mating systems. The predominantly self-fertilizing species (Clarkia exilis) consistently exhibited a significant positive r_g between pollen and ovule production while the outcrossing species (C. unguiculata) exhibited either a trade-off or independence between these traits. Clarkia exilis also exhibited much more highly canalized gender expression than C. unguiculata. Selection on pollen and ovule production resulted in little correlated change in the P:O ratio in the selfing exilis, while dramatic changes in the P:O ratio were observed in unguiculata. To test the common prediction that floral attractiveness should be positively genetically correlated with investment in male function, we examined the response of petal area to selection on pollen and ovule production and found that petal area was not consistently genetically correlated with gender expression in either species. Our results suggest that the joint evolutionary trajectory of primary sexual traits in hermaphroditic species will be affected by their mating systems; this should be taken into account in future theoretical and comparative empirical investigations.