This study aims to understand individual differences between children in metacognitive monitoring and control processes and the developmental trajectories of metacognition over one year. Three indicators of procedural metacognition were used: monitoring accuracy (discrimination of confidence judgments between correct and incorrect test responses), effective restudy selections, and accuracy of response maintenance/withdrawal decisions. These indicators were measured for two tasks (text comprehension and Kanji memory) at two measurement points one year apart. Participants were 151 second graders (M age 7.61 years) and 176 fourth graders (M age 9.62 years). With latent profile analyses, distinct metacognition profiles were found for both grade levels at both measurement points. Children showed heterogeneity in the proficiency of metacognition but also in the extent to which metacognitive skills were generalizable across the two tasks. For second-grade children, being low at metacognition at the first measurement point was not associated with extra risks for low metacognition one year later. However, for fourth graders, children with low metacognitive skills appeared likely to stay low in metacognition over time and particularly showed ineffective restudy decisions. This indicates that they seemed at risk for a longer-term metacognitive deficiency. Findings may improve understanding of the heterogeneity of metacognition and support distinguishing typical from at-risk metacognitive development.