A diagnostic test in English as a second language should be a series of miniature tests on specific problems. Subscores in each area should be considered rather than a total score. The results should be used to probe mastery in an area rather than provide the means for comparing one student against another. The statistical reliability of the results does not necessarily depend on test length. The teacher should look at each item for each student rather than the score and should spend more time studying the analysis of each student's test. The criterion of the percent of correct decisions may be a more meaningful measure than ascertaining the traditional coefficients of reliability. Tables provide the statistical data under consideration. (VM)