The influential hypothesis that humans imitate from birth - and that this capacity is foundational to social cognition - is currently being challenged from several angles. Most prominently, the largest and most comprehensive longitudinal study of neonatal imitation to date failed to find evidence that neonates copied any of nine actions at any of four time points (Oostenbroek et al.,  Current Biology, 26, 1334-1338). The authors of an alternative and statistically liberal post-hoc analysis of these same data (Meltzoff et al.,  Developmental Science, 21, e12609), however, concluded that the infants actually did imitate one of the nine actions: tongue protrusion. In line with the original intentions of this longitudinal study, we here report on whether individual differences in neonatal "imitation" predict later-developing social cognitive behaviours. We measured a variety of social cognitive behaviours in a subset of the original sample of infants (N = 71) during the first 18 months: object-directed imitation, joint attention, synchronous imitation and mirror self-recognition. Results show that, even using the liberal operationalization, individual scores for neonatal "imitation" of tongue protrusion failed to predict any of the later-developing social cognitive behaviours. The average Spearman correlation was close to zero, mean rs = 0.027, 95% CI [-0.020, 0.075], with all Bonferroni adjusted p values > .999. These results run counter to Meltzoff et al.'s rebuttal, and to the existence of a "like me" mechanism in neonates that is foundational to human social cognition. © 2019 John Wiley & Sons Ltd.