Theories on the evolution of recombination in regard to its ability to increase mean fitness require a consistent source of negative linkage disequilibrium among loci affecting fitness to show an advantage to recombination. Here we present evidence that, at least theoretically, genetic variation for recombination can spread in very large populations under a strictly multiplicative-fitness, deleterious-allele model. The model uses only Mendelian genetics in a multi-locus context to show that a dominant gene for recombination can spread when rare and resist invasion when common. In non-equilibrium populations driven by Muller's ratchet, the gene increases its probability of fixation by increasing the probability of being associated with the best individuals. This occurs at an optimal level of recombination. Its action results in both an immediate and a long-term advantage to recombination amongst the proto-meiotic organisms modelled. The genetic mechanism lends itself naturally to a model for the evolution of meiosis, where modern-day gametes are seen as derivative of ancient unicellular organisms.