One aim of criterion-referenced testing is to classify an examinee without reference to a norm group; therefore, any statements about the dependability of such classification ought to be group-independent also. A population-independent index is proposed in terms of the probability of incorrect classification near the cutoff true score. The compound binomial model leads to the conclusion that a criterion-referenced test is more reliable if the item difficulties are unequal. The proposed index also has the following properties: (1) A test with a higher value of r will result in a smaller loss due to misclassification; (2) It is easily generalized to a case of more than two classes; and (3) It has no theoretical upper limit, but increases with test length. (Author/CTM)