Like scientists, children and adults learn by asking informative questions and making informative interventions. How does this ability develop? We investigate how children (7- and 10-year-olds) and adults search for information to learn which kinds of objects share a causal property. We find an anticipated developmental improvement in information search efficiency. We also present a formal analysis that allows us to identify the strategic component responsible for children’s inefficiency. In our 20-questions-style task, children initially ask questions and make interventions no less efficiently than adults do, but continue to search for information past the point at which they have narrowed down their hypothesis space to a single option. In other words, the performance change from age seven to adulthood is due mainly to a change in implementing a “stopping rule”: when considering only the minimum number of queries participants would have needed to identify the correct hypothesis, age differences disappear.