A study was conducted to develop a behaviorally anchored rating scale for the evaluation of college teaching that could be used in all academic areas and that would reduce leniency, central tendency, and other rater biases. The Smith-Kendall procedure was modified by having undergraduate subjects generate behavioral examples for specified levels of effectiveness. A Likert-type scale was included to provide a frame of reference to aid judges in the generation of mid-range behaviors. For nine dimensions of teaching effectiveness, 243 critical behaviors were generated, 165 of which were eliminated because they did not meet criteria for retention as anchors. The seven scales that resulted encompassed dimensions of teacher effectiveness that students were best able to rate reliably. It is claimed that the high degree of consensus among judges, as well as the criteria for item retention, ensure that the final anchors are unambiguous, but generalizable to any classroom situation. (SW)