The gametes produced in meiosis provide information on the frequency of recombination and also on the interdependence of recombination events, i.e. interference. Using F(2) individuals, it is not possible in all cases to derive the gametes, which have fused, and which provide the information about interference unequivocally when three or more segregating markers are considered simultaneously. Therefore, a method was developed to estimate the gametic frequencies using a maximum likelihood approach together with the expectation maximisation algorithm. This estimation procedure was applied to F(2) mapping data from rice (Oryza sativa L.) to carry out a genome-wide analysis of crossover interference. The distribution of the coefficient of coincidence in dependence on the recombination fraction revealed for all chromosomes increasing positive interference with decreasing interval size. For some chromosomes this mutual inhibition of recombination was not so strong in small intervals. The centromere had a significant effect on interference. The positive interference found in the chromosome arms were reduced significantly when the intervals considered spanned the centromere. Two chromosomes even demonstrated independent recombination and slightly negative interference for small intervals including the centromere. Different marker densities had no effect on the results. In general, interference depended on the frequency of recombination events in relation to the physical length. The strength of the centromere effect on interference seemed to depend on the strength of recombination suppression around the centromere.