Prior knowledge is often strongly related to students' learning. In the present research, we explored the relationship between prior knowledge and the accuracy of students' predictive monitoring judgments (judgments of learning; JOLs) and postdictive monitoring judgments (confidence judgments). In four experiments, students completed prior knowledge tests over two domains (football and cooking), learned new items from those domains, and completed a test over those items. To ensure the learning was new, the items that students learned and were tested over consisted of made-up facts from each domain. In Experiments 1-3, students made JOLs predicting the likelihood that they would remember each item on the test. In Experiments 2 and 3, students also made confidence judgments in the accuracy of their response after answering each question on the final test. Prior knowledge was negatively related to the accuracy of students' JOLs, but unrelated to the accuracy of their confidence judgments. Exploratory analyses revealed that the relationship with JOLs was likely driven by high knowledge students giving relatively lower JOLs to items that they later remembered. Experiment 4 directly tested this possibility by having students make binary memory predictions about whether they would remember each item. Outcomes were consistent with the exploratory analyses. Thus, the negative relationship between prior knowledge and JOL accuracy appears to be driven by the greater tendency of high knowledge students, relative to low knowledge students, to remember items that they thought they would forget.