This paper presents analyses of score reliability for a mathematics placement test developed by the Mathematical Association of America in 1984 for use with undergraduate college students. Subjects included all 589 students seeking admission to the college mathematics curricula at a private university in the South during the course of the study. Common misconceptions about reliability are also reviewed, such as the misconception that tests, as opposed to data, are reliable; and the idea that longer tests always yield more reliable data than do their shorter counterparts. The test was a revision of items previously reported. The revised version of the 50-item test included 2 subscales of 25 items each: an arithmetic scale, and an algebra scale. Reliability estimates and various item analysis statistics were reported. Revisions of the test yielded slightly better than expected alpha coefficients for the arithmetic and algebra subscales. Overall, the results suggest that the revised test can yield reasonably reliable data, and the test can be used to place students in an appropriate curriculum. One table contains the study data. (SLD)