When we start to look at more than 2 loci at a time, there is a complication due to the possibility of double recombinations, or double crossovers. In the drawing above, if there is a crossover between A and B AND one between B and C, they would cancel each other out. So if we could only score for A and C (not for B), we would conclude that the gamete was parental (when in fact there were two recombination events). Since we cannot observe double recombinants, the calculated recombination frequency is an underestimate. The Kosambi mapping function contains a correction for this.