This meta-analysis examines the effects of learning with robots (r-learning) on young children and, on this basis, gives suggestions for using robotics in education for young children. A test of homogeneity was performed for 27 Korean studies done between 2008 and 2016 and a random effect model introduced to reveal the effect sizes. The overall effect size was medium to large at 0.72. After analyzing gaps in effect sizes with different categorical moderator variables, the authors found significant differences depending on platform type, activity type, dependent variables for r-learning effect, and age. The study also investigates relationships between effect size and continuous moderator variables such as treatment period and year of publication. The meta-regression model showed a significantly negative relationship between effect size and year of publication. Thus, effects of r-learning on a young child are generally beneficial, and r-learning also improves variables like a child's social nature, though conversely the effect on language development appears below average.