ACCURACY OF METACOMPREHENSION JUDGMENTS FOR QUESTIONS OF VARYING IMPORTANCE LEVELS

Students predicted future test performance and assessed past performan ce on questions tapping important sentences, questions tapping unimpor tant sentences, and higher order questions that did not explicitly tap specific sentences. In Experiment 1, predictions were accurate only f or higher order questions, but posttest confidence judgments were most accurate for the questions directly tapping text sentences. In Experi ment 2, student control of reading was removed, and predictions were a ccurate for both unimportant questions and higher order questions, alt hough predictions for unimportant questions may have been based on pri or knowledge. Both experiments show that students can predict their pe rformance on higher order test questions better than on questions tapp ing specific sentences in a text, suggesting that they use an overall understanding of the text in making such predictions. In addition, pos ttest confidence judgments were most consistently accurate for questio ns tapping the details in the text.