This paper describes a study, now in progress, of new methods for representing the sampling variability of Mantel-Haenszel differential item functioning (DIF) results, based on the system for categorizing the severity of DIF that is now in place at the Educational Testing Service. The methods, which involve a Bayesian elaboration of procedures developed by Zwick, Thayer, and Wingersky (1994) are expected to yield more interpretable results in small samples than do conventional procedures. They may therefore be useful in the small-sample conditions that occur in adaptive testing. Two methods are suggested. One is for obtaining a probabilistic estimate of the true DIF category, and the other is for predicting the future observed category. These Bayesian methods have not yet been tested through simulation. Some potential limitations are discussed, subject to testing the accuracy of the probabilistic results and the sensitivity of results to the shape and parameters of the prior distribution. Two tables and three figures illustrate this exploration of techniques. (Contains 15 references.) (SLD)