Students' different standards may yield different kinds of bias, such as self-directed (higher than their past performance) bias and peer-directed (higher than their classmates) bias. Utilizing data obtained from a natural experiment where some students were able to see their grades prior to teacher evaluations, and to investigate possible sources of bias, we empirically analyzed the role of information (such as the actual grade students received in their current course and their previous grade point average), and the average grade of the course, on the student evaluation of teaching. Because bias is sensitive to the accuracy of grade information, the randomized data examined in this paper are a valuable source for estimating both self-directed and peer-directed bias. We identify the existence of the two kinds of biases and demonstrate that the influence of peer-directed bias tends to increase after the accurate information on the course grade is revealed.