Previous research has suggested that good students (those above the median) are better than poor students (below the median) at estimating their academic performance. To examine the relationship between pre-test predictions and post-test estimations of academic performance, 30 college students were asked to predict their scores on each of four tests in a psychology course. Following each test they were asked to estimate their performance, and make predictions about the performance of the class as a whole. Students also received feedback on prediction accuracy for the class. Before the third examination, they were given instructions for forming their predictions and estimations. Results showed that, while students generally overestimated their test scores, better students were closer to the target. When explicitly forewarned about overestimation errors, students were successful in lowering their estimations for the next test; when not forewarned, they were unsuccessful in lowering their estimated scores. Students above and below the median predicted class average with the same degree of precision. Results seem to be consistent with indications that good and poor students form differential assessments of their own test performance, but are similar in their assessments of the performance of others. (JAC)