These comments are mostly derived from my teaching experience at the University of Texas at Austin, which has a sophisticated and elaborate teacher evaluation system, and in the universities of London and York in England, which don't.

1. Students' evaluations of their teachers do not depend solely on the qualities of the teacher. If objective evaluations are to be obtained, a multivariate statistical analysis should be conducted, controlling for such factors as:

a. class size—the smaller the class, the better the evaluations.

b. expected grade—one of the strongest correlates of teacher evaluations proved to be the grade the student expects to receive—the higher the expected grade, the “better” the course. This too can be controlled for—by asking the students on the form what grade they expect to receive.

c. whether the class is required or optional—compulsory courses will obviously be less popular. One way to test for this is to ask students on the evaluation form to rate the class relative to their expectations of the class (above, below, or as expected).

2. A statistical study at Texas by J. Sidanius showed that student evaluations of teachers tend to be biased against women and minority teachers. Apart from the intrinsic worries this raises, such information, if true, could be used in court actions over denial of tenure for women/minorities where teacher evaluations played a role.

3. There is of course the general philosophical question of whether students are best able to assess whether they are learning anything from a given teacher.