To test whether sister chromatid exchange (SCE) scores on human chromosomes have a uniform distribution, simulated SCE scores were generated and compared with observed scores using log-linear models. The analysis was performed at the level of the chromosome groups. Using this method we first tested whether the number of SCEs was distributed uniformly, i.e. proportional to the relative length of the chromosomes. Refinements of this hypothesis were made by considering a variable region around a first SCE to be inert for other SCEs and by making the occurrence of an SCE on a chromosome dependent on the occurrence of another SCE on the same chromosome. In further analyses it was tested whether the number of SCEs was proportional to the number of G bands on a chromosome, or to the DNA content of the chromosomes. None of the tested hypotheses fitted the observed data, establishing the non-uniform distribution of these events.