A method is proposed to assess the importance of differential item functioning (DIF) by estimating the largest possible fraction of the population in which DIF does not occur, or equivalently, the smallest possible portion of the population in which DIF may occur. The approach is based on latent class (C. C. Clogg, 1981) or mixture concepts, and was proposed by T. Rudas, C. C. Clogg, and B. G. Lindsay (1994) in the context of assessing the fit of an arbitrary model to a contingency table. Application of this procedure produces an estimate of the minimum proportion of the population that would have to be removed to make the rest of the population free from DIF, as well as information about the portion of the population that is the source of DIF. Simple methods for maximum likelihood estimation are described. Numerical results are presented for a simulated data set and actual data from the 1993 Advanced Placement Physics examination. (Contains 3 tables and 27 references.) (SLD)