This paper introduces criterion-referenced tests (CRTs), compares them with norm-referenced tests (NRTs), discusses how they can be evaluated and revised, and presents a study of an actual class and textbook test evaluation using CRTs. NRTs have dominated testing methodology since the mid-1970s; an example is the Test of English as a Foreign Language (TOEFL). CRTs are much less well known; they determine the amount of material learned rather than spreading students out along a continuum of general ability. In foreign language learning, NRTs measure general language proficiency; CRTs measure specific objectives. NRTs are of little help in diagnosing students' strong and weak points, assessing achievement, or evaluating programs. CRTs, which can be designed and evaluated by using item analysis, serve much better in these areas. The test used in the study was designed by a teacher with many years experience in teaching English as a Second Language. Unfortunately, the test described and shown, using CRT with item analysis, is found to be ineffective. Specifically, the test lacked institutional goals, forcing reliance on the textbook, rather than course objectives, for test construction. (Contains 13 references.) (NAV)