This paper discusses research conducted in the spring of 1991 that measured the relationship of reading subtest scores to teacher ratings of students' reading abilities. Sixty-eight advanced-level students in an intensive English program took an institutional version of the Test of English as a Foreign Language (TOEFL) and a specimen reading module of the International English Language Testing System battery. The students' reading abilities were assessed by their teachers, using a scale devised by the American Council on the Teaching of Foreign Languages. Reliability estimates were obtained and correlations were run. The research tested the hypothesis that the results of a reading test for academic purposes, based on current theories of test design and construction, would correlate better with teacher observations than would results of a more traditional test (the TOEFL). Data analysis indicated instead that each of the tests correlated moderately well with teacher observations. Although somewhat different patterns of correlations occurred with graduate versus undergraduate students, and with natural sciences majors versus arts and social sciences majors, most differences were not statistically significant. (KM)