Higher education relies on student ratings to evaluate faculty teaching, partly because the alternatives (expert peer appraisals or objective performance criteria) are costly or unavailable. Because student ratings are crucial not only to improving instruction, but also in making or breaking faculty careers, it is important to assure that they provide valid indications of instructional quality. Analyses of large data sets obtained at University of Washington show that student ratings are prone to artifacts that can produce occasional substantial underestimates of teaching ability for instructors who grade strictly (and overestimates for those who grade leniently). Some likely system impacts of this distortion of ratings are to nudge (1) instructors toward lenient grading, and (2) students toward nonchallenging courses. The bright side of this picture is that the usefulness of student ratings can be improved statistically. While it has been found that giving inflated grades produces inflated ratings and higher student workloads generally produce lower ratings, statistical adjustment of data, removing invalid variance, can derive more accurate ratings. The appendix contains the Instructional Assessment System form used for faculty evaluation at the University of Washington. (Contains 27 references.) (Author/JLS)