Abstract: Languages describe “who is doing what to whom” by distinguishing the event roles of agent (doer) and patient (undergoer), but it is debated whether they result from nonlinguistic representations that may already exist in preverbal infants and nonhuman animals. The phenomenon of causal perception, where the subsequent movements of two objects A and B evoke the impression of A launching B, is a simple depiction of an agent-patient relation. The seminal study by Leslie and Keeble from 1987 proposed that infants of 6 months old may be able to attribute agent and patient roles to such causal displays, after they demonstrated the infants’ dishabituation upon seeing a launching event that was reversed. They introduced the idea that a role reversal had taken place upon reversing the direction of the launching event (launcher becoming launchee), but not in a noncausal temporal gap event where the agent and patient roles were not present. The present study tested this hypothesis in three different populations: 6-month-old human infants, human adults, and Guinea baboons (Papio papio). For the human infants, we applied a habituation-dishabituation design, and for the human adults and baboons, a conditional match-to-sample task. We were unable to replicate the findings of Leslie and Keeble in human infants. Similarly, we did not find evidence for an effect specific to reversing launching events in human adults and baboons. Inconsistent results across different studies call into question the role reversal paradigm for Michottean launches to study event role attribution.