The hypothesis that children develop an understanding of causal mechanisms was tested across 3 experiments. In Experiment 1 (N = 48), preschoolers had to choose as efficacious either a cause that had worked in the past, but was now disconnected from its effect, or a cause that had failed to work previously, but was now connected. Four-year-olds chose the now-connected cause more often than 3-year-olds. Experiment 2 (N = 16) showed 4-year-olds responded appropriately to an irrelevant modification in the same causal system. Experiment 3 (N = 24) demonstrated when the mechanism was batteries rather than connection, 3-year-olds could properly distinguish between relevant and irrelevant modifications. Together, these data suggest that understanding of specific causal mechanisms develops at different ages.