The current study investigated how metacognitive abilities such as Knowledge Confidence (subjective prior knowledge estimate) and Correctness Confidence (appraisal of the likelihood of closing the gap) relate to curiosity, and examined the roles of these metacognitive measures and curiosity in learning. Using a blurred picture paradigm in which participants viewed blurred pictures and provided their metacognitive and curiosity estimates, the current study identified distinct connections between Knowledge Confidence, Correctness Confidence and curiosity. Our finding suggests that when Knowledge Confidence is at low or medium levels, curiosity is particularly heightened. In contrast, Correctness Confidence linearly influences curiosity such that the higher the Correctness Confidence, the greater the curiosity. We also find that learning is best predicted by a learner's metacognitive appraisal of their knowledge gap, especially when they are on the verge of knowing, and this learning effect is independent of curiosity. These findings provide new evidence for the role of curiosity and metacognition in learning, highlighting limits of the effect of curiosity on learning.