The study by Southgate, V., Senju, A., and Csibra, G. (Southgate et al., 2007) has been widely cited as evidence for the ability of false-belief attribution in young children. Recent replication attempts of this paradigm have yielded mixed results: several studies were unable to replicate the original finding, raising doubts about the suitability of the paradigm to assess non-verbal action prediction and Theory of Mind. In a preregistered collaborative study including two of the original authors, we tested 160 24- to 26-month-olds across two locations following the original stimuli, procedure, and analyses as closely as possible. We found no evidence for action anticipation: only about half of the infants correctly anticipated the protagonist’s actions when action prediction did not require taking into account the agent’s beliefs. In addition, even those who appeared to anticipate failed to do so when a false belief was involved. These findings indicate that the paradigm of Southgate et al. (2007) cannot reliably elicit anticipatory action prediction and is unsuitable for testing false belief understanding in 2-year-olds.