Ambridge, Bidgood, and Thomas (2020) conducted an elicitation-production task in which children with and without (high-functioning) autism described animations following priming with passive sentences. The authors report that children with autism were more likely than IQ-matched children without autism to commit reversal errors, for instance describing a scene in which the character Wendy surprised the character Bob by saying Wendy was surprised by Bob. We set out to test whether this effect replicated in a new sample of children with and without (high-functioning) autism (N = 26), and present a cumulative analysis in which data from the original study and the replication were pooled (N = 56). The main effect reported by Ambridge et al. (2020) replicated: While children with and without autism produced a similar number of passive responses in general, the responses of children with autism were significantly more likely to include reversal errors. Despite age- appropriate knowledge of constituent order in passive syntax, thematic role assignment is impaired among some children with high-functioning autism.