Adults recognize that if event A predicts event B, intervening on A might generate B. Research suggests that although preschoolers draw this inference much like adults, toddlers do not (Bonawitz et. al, 2010). Here we look at whether toddlers’ failure is domain-general (i.e., they lack an adult-like concept of causality that integrates prediction, intervention, and agency) or domain-specific (i.e., toddlers have trouble recognizing some physical processes as causal but might succeed with psychological events). We showed toddlers (24 months) a block moving into a base, after which an effect occurred; we then gave children the block and asked them to generate the effect. Toddlers performed the intervention and predicted the outcome when the effect was psychological (a puppet laughing) but, replicating previous studies, not when the effect was physical (a toy activating). Experiment 2 showed that this was not due to the relative saliency of the psychological effect.