Abstract For many organisms, meiotic double crossing over is less frequent than expected on the assumption that exchanges occur at random with respect to each other. This "interference," which can be almost total for nearby intervals, diminishes as the intervals in which the double crossovers are scored are moved farther apart. Most models for interference have assumed, at least implicitly, that the intensity of interference depends inversely on the physical distance separating the intervals. However, several observations suggest that interference depends on genetic distance (Morgans) rather than physical distance (base pairs or micrometers). Accordingly, we devise a model in which interference is related directly to genetic distance. Its central feature is that recombinational intermediates (C's) have two fates--they can be resolved with crossing over (Cx) or without (Co). We suppose that C's are distributed at random with respect to each other (no interference); interference results from constraints on the resolution of C's. The basic constraint is that each pair of neighboring Cx's must have between them a certain number of Co's. The required number of intervening Co's for a given organism or chromosome is estimated from the fraction of gene conversions that are unaccompanied by crossover of flanking markers. The predictions of the model are compared with data from Drosophila and Neurospora.