Abstract
Neonatal imitation is a cornerstone in many theoretical accounts of human development and social behavior, yet its existence has been debated for the past 40 years. To examine possible explanations for the inconsistent findings in this body of research, we conducted a multilevel meta-analysis synthesizing 336 effect sizes from 33 independent samples of human newborns, reported in 26 articles. The meta-analysis found significant evidence for neonatal imitation (d = 0.68, 95% CI = [0.39, 0.96], p < .001) but substantial heterogeneity between study estimates. This heterogeneity was not explained by any of 13 methodological moderators identified by previous reviews, but it was associated with researcher affiliation, test of moderators (QM) (15) = 57.09, p < .001. There are at least two possible explanations for these results: (a) Neonatal imitation exists and its detection varies as a function of uncaptured methodological factors common to a limited set of studies, and (2) neonatal imitation does not exist and the overall positive result is an artifact of high researcher degrees of freedom.