<p>Previous research has demonstrated that anchoring information biases the prediction of current learning states such as judgments of learning (JOLs). In an actual educational setting, anchoring information is not always consistent with the actual learning experience. Despite this, there has been no evaluation of the effect of experience-inconsistent anchors on JOLs. This study examines the effect of experience-inconsistent anchors on JOLs and their accuracy through a comparison with the no anchor and experience-consistent anchor conditions. In the experiments, the participants in the experience-inconsistent anchor condition were informed that the task was extremely difficult and approximately 10% of individuals could remember the item when making JOLs for the easy items (i.e., strongly or weakly related pairs), and they were informed that the task was extremely easy and approximately 90% of individuals could remember the item when making JOLs for difficult items (i.e., unrelated word pairs). In contrast, participants in the experience-consistent anchor condition were informed that the task was easy for easy items and difficult for difficult items. In the no anchor condition, participants were not given any information. The results demonstrated that the difference in the JOLs between the easy and difficult items in the experience-inconsistent anchor condition was smaller than that in the no anchor and experience-consistent anchor conditions. Notably, the experience-inconsistent anchor impaired the JOL accuracy when the difference in difficulty between the easy and difficult items was clearer. These findings suggest that experience-inconsistent anchors can reduce the accuracy of learners’ metacognitive monitoring by reducing their reliance on predictive cues for JOLs such as item relatedness.</p>