In learning about the world children must not only make inferences based on minimal evidence, but must deal with conflicting evidence and question those initial inferences when they appear to be wrong. Four experiments (N = 144) found that young children were significantly more likely to revise their initial inferences when conflicting evidence was explicitly demonstrated for them. Four-and five-year-old children saw deterministic evidence about which objects had causal powers, and then saw counterevidence conflicting with that initial pattern. Critically, the conflicting evidence was either demonstrated communicatively and pedagogically, or produced in an intentional but nonpedagogical manner. Only when evidence was explicitly demonstrated for them did children revise their initial hypothesis and use a subtle clue to infer the correct rule. We thank Ariel Horn, Kendra Soh, and Erica Zippert for assistance with data collection and coding, and Andrei Cimpian for helpful feedback on a previous version of this manuscript.

No abstract