Children can track someone else's learning before their own
The clearest picture comes from experiments where children were taught a new fact, or watched a puppet being taught, and were then asked whether they (or the puppet) had known the fact before or had just learned it. Four- and five-year-olds performed at chance when judging their own learning, and in one experiment six- and seven-year-olds did too [2]. The same children, however, were usually accurate when judging whether the puppet had just learned something [2].
That split matters because it rules out a simple memory failure. The children could hold onto the information and reason about learning in general; what they struggled with was applying that reasoning to themselves. The authors frame this as a difficulty with the conceptual understanding of learning and with cognitive monitoring, not with remembering the fact itself [2].
Children were not uniformly bad at temporal judgments. When asked when they or a puppet had performed a physical action, or whether a fact had been learned by seeing a picture versus being told, they often answered correctly [2]. So the weakness is specific: judging the moment a piece of knowledge entered their own head is harder than judging the source or the timing of an action.
Overconfidence is the default, and pushing children to pre-judge can backfire
When young children rate how confident they are in an answer, their monitoring is often overoptimistic—they are more sure than their accuracy warrants [4]. This is a different flavor of monitoring failure from the timing judgments above: here the child is not confused about when they learned something, but about how well they know it.
A natural fix would be to make children evaluate their confidence before committing to an answer, so they weigh all the options. Kindergarteners and second graders were given exactly that pre-monitoring step in a paired-associates learning task. Contrary to the researchers' expectation, the instruction increased overconfidence rather than reducing it [4]. The authors discuss this against possible memory–metamemory interactions, confirmation bias, and methodological issues—so the backfire is an observed result whose explanation is still open, not a settled mechanism [4].
The practical implication is that children's monitoring is not simply a matter of insufficient effort. Telling a young child to think carefully before judging their own certainty did not make their judgments more accurate in this study [4].
What develops with age, and where the adult comparison is thin
The developmental story is that self-monitoring improves as children get older, but the direct child-versus-adult comparison is the weakest part of the available evidence. The timing-of-learning experiments included children up to seven years old and no adults [2], and the pre-monitoring study covered kindergarteners and second graders only [4]. One review of feedback and monitoring discusses metacognitive feedback for both adults and young learners, but its abstract does not report a head-to-head accuracy comparison [3].
Two other studies compare children and adults on adjacent skills rather than on monitoring itself. A self-prioritization study found that 9- to 22-year-olds all showed a robust bias toward self-related information over mother, friend, or stranger associations, and a bias toward mother over friend and stranger—stable across older children, teenagers, and young adults [1]. That tells us something about how children process self-relevant material, not about how accurately they judge their own learning. Separately, in a live-event recall study, conceptual thinking ability predicted children's accurate free recall better than age did, though this concerns memory accuracy rather than monitoring [5].
So the honest answer to how age affects monitoring accuracy is directional but not finely calibrated: young children are markedly inaccurate at judging their own learning, some of that inaccuracy persists at least to age seven, and the studies available do not yet pin down the adult endpoint or the shape of the curve in between [2][4].
Sources used in this answer
Self-Prioritization Effect in Children and Adults
A self-prioritization bias toward self over mother, friend, and stranger associations was robust and stable across older children (9–13), teenagers (14–17), and young adults (18–22).
Young children’s monitoring of learning by self or other
Four- to seven-year-olds answered at chance when judging whether they had just learned a fact versus known it before, but often judged a puppet's learning accurately and handled some action-timing and source judgments correctly.
Using feedback to support children when monitoring and controlling their learning
The paper examines feedback and metacognitive support for children's monitoring accuracy and discusses metacognitive feedback for both adults and young learners.
Motivating children to (pre)monitor: positive effects on monitoring accuracy?
Instructing kindergarteners and second graders to pre-monitor their confidence before answering increased rather than decreased their overconfidence.
Recall Accuracy in Children: Age vs. Conceptual Thinking
Conceptual thinking ability predicted children's accurate free recall and some inaccurate responses better than age did, in a study of 7- to 14-year-olds and adults.