What is the best way of figuring out which variables can cause some outcome of interest? One prominent normative proposal is that learners should manipulate each candidate variable in isolation to avoid receiving confounding information. Here, we demonstrate that this strategy is not always the most efficient method for learning about a causal system. Using an optimal learner model, we show that when a causal system is sparse, that is, when the outcome of interest has few or even just one actual cause among the candidate variables, it is actually more efficient to test multiple variables at once. In a series of behavioral experiments, we then show that people are sensitive to causal sparsity when planning causal experiments.