In the past decade, research increasingly uncovers emerging metacognitive skills in young children by using child-friendly, creative, and non-verbal measures of metacognition. In the present study, the now often used non-verbal "opt-out" paradigm and classical verbal metacognitive judgments (judgments-of-learning and confidence judgments) were used in the context of a paired associates learning task. Prospectively (before the memory test) and retrospectively (after recognition) N = 138 children (N = 72 5-year-olds; N = 66 6-year-olds) evaluated their performance. Results revealed evidence for existing metacognitive skills in their relative sense (discriminating correct from incorrect responses in the metacognitive measures) but poor metacognitive accuracy in absolute terms. The non-verbal opt-out measures were not found to consistently and substantially facilitate especially 5-year-olds' metacognitive skills. Correlating non-verbal and verbal measures revealed shared but also distinct, yet to be explained variances calling for a strongly differentiated concept of children's early metacognition.