AbstractIn this paper a simple test is described, with which it is possible to check the assumption that in the cases of analysis of variance or covariance several interactions don't occur. This test is particularly useful also if some subclasses are empty and in the case of unequal subclass numbers in model I, model II and the mixed model.