Coventional true score approaches to criterion-referenced tests are not applicable when tests are used for classification. Criterion-referenced testing is now used frequently for educational decision making. The assumption is made that these instruments are to be considered as tests designed to classify into "mastery" and "non-mastery" groups, rather than to measure. There is no precise methodology in use to establish cut-off scores beyond which mastery can be assumed to have occurred. Latent trait modeling of test performance can enable straightforward criterion-referenced interpretations of test scores. A future expansion in the use of testing to diagnose the learning problems of individual students is foreseen. Individual diagnostic testing requires criterion-referencing of test performance, and of each item response. Criterion-referencing with latent trait models will become more useful as individual diagnostic testing becomes more prevalent. (Author/DWH)