Induced mitotic gene conversion studies on the CYC1 gene of yeast have shown that the actual base pair changes, the types of changes (base substitution, deletion or addition) and the distances between mutations all affect gene conversion yields. In crosses between mutations less than four bases apart, gene conversion rates are as low as back mutation rates. The same mutants crossed to alleles more than five bases away may recombine 50-fold more. In crosses between mutations five or more base pairs apart, recombination rates varying by up to ten-fold are observed when comparing mutations at the same codon sites. The actual mutations in crosses affect recombination rates at these distances. The data rules out models in which mutants are repaired independently. Models with large gaps at the initiation site are ruled out if the mutants are within the gap. Recombination models are favoured in which both mutations can interact at a distance to affect the probability of recombination; such interactions may reflect the geometry of the recombinational junctions. The specific interactions proposed are that the actual mutations, and residual mismatches arising on excision resynthesis, affect both the further migration of the recombinational junction, and the probability that excision-repair will detect and correct residual mismatches. Junction models in which interactions are expected include those composed of base tetraplexes. The data is interpreted in terms of these models. Meiotic recombination data is consistent with these models.