The relationship between metamemorial accuracy and student test performance was assessed in a graduate educational psychology class. Metamemory is defined as information about the content of one's own memory. Adult students in two courses (N=67) were asked to select multiple choice answers in two midterms and a final examination and to rate their confidence in the answers on a five-point Likert scale. At the end of the test, students were asked for a prediction of the percentage of items they thought they answered correctly. They were then given ten replacement items and were allowed to replace any ten test items with any of these replacements. Those who had a high mean degree of confidence in their answers had higher examination scores than those with low confidence in their answers. There was a significant difference between good and poor scorers in their ability to predict what they did and did not know. Good students, as defined by scores on the final examination, appeared to have better metamemorial accuracy than the poor students. The use of replacement items allowed students to improve their scores significantly. (SLD)