People generally make inaccurate judgments of their knowledge. This is demonstrated by low correlations between learning judgments and learning outcomes in metacomprehension research (i.e., low accuracy). Three experiments investigate how the framing and wording of a metacognitive judgment affects accuracy. Experiments 1 and 2 extend metacomprehension research by systematically varying whether judgment prompts ask for a judgment of learning or understanding, measuring both retention and transfer performance, and using a computer-based multimedia lesson. In two experiments, college students viewed a lesson on how solar cells work and were asked to indicate at five points in the lesson either how well they remember (JOL) the material or how well they understand the material (JOU). Although the pattern of results differed in the two experiments, the combined results yield large correlations between retention and transfer with judgments of understanding (r = .62 and r = .59, respectively), and medium correlations with judgments of remembering (r = .39 and r = .38, respectively). Experiment 3 investigated the effect of framing knowledge judgments for four categories of judgments. The judgment prompts either asked about one's amount of knowledge, one's confidence in knowledge, one's ability to answer questions about their knowledge, or one's perceived difficulty in obtaining knowledge. The former three judgments all significantly predicted retention and transfer performance, but the judgment of difficulty did not predict transfer. These results show the benefits of including judgments of understanding and transfer tests to studies that examine metacomprehension of expository materials, and the importance of choosing appropriate wording for judgment prompts.