Using an artificial gene network model, we have shown that recombination imposes selection for robustness to mutation, and that negative epistasis can evolve as a by-product of this selection (1). These results suggested that sexual reproduction may select for conditions that favor its own maintenance. The trends in mean fitness of sexual and asexual populations reached after different periods of evolution in isolation provide tentative support for this hypothesis. However, a long-term advantage of sex may not be sufficient to prevent mutations that reduce recombination from spreading within a sexual population. MacCarthy & Bergman (2) found that low-recombination alleles consistently outcompeted high-recombination alleles in this model. Here we show that the ability of a sexual population to resist invasion by a low-recombination allele increases with the period of time the population is allowed to evolve before the invasion assays are conducted. This effect is largely driven by a gradual reduction in recombination load over thousands of generations. These observations provide the first direct evidence that sexual reproduction does indeed select for conditions that favor its own maintenance. These data also suggest that genetic and ecological factors that reduce recombination load may play an important role in the evolution of sex. We have tested three types of nonrandom mating. Consistent with our prediction we have found that assortative mating and selective mating based on gene expression phenotype, and population structure, all reduce recombination load and increase the ability of a sexual population to resist invasion by asexual mutants. Furthermore, assortative and selective mating decrease phenotypic diversity, whereas local dispersal and competition in population structure slow down the spread of asexual mutants facilitating their mutational deterioration. All these effects confer a short-term advantage to sexual reproduction, thus suggesting that certain deviations from random mating favor the maintenance of sex. (1) Azevedo et al. 2006. Nature 440: 87-90. (2) MacCarthy & Bergman. 2007. PNAS 104: 12801-12806.