Whether non-human animals have an ability to learn and process center embedding, a core property of human language syntax, is still debated. Artificial-grammar learning (AGL) has been used to compare humans and animals in the learning of center embedding. However, up until now, human participants have only included adults, and data on children, who are the key players of natural language acquisition, are lacking. We created a novel game-like experimental paradigm combining the go/no-go procedure often used in animal research with the stepwise learning methods found effective in human adults’ center-embedding learning. Here we report that some children succeeded in learning a semantics-free artificial grammar with center embedding (A2B2 grammar) in the auditory modality. Although their success rate was lower than adults’, the successful children looked as efficient learners as adults. Where children struggled, their memory capacity seemed to have limited their AGL performance.