There is a fundamental tension in cognitive development. Young children have severe limitations in planning, decision-making, executive function and attentional focus, roughly those abilities that involve prefrontal control. Yet young children are also prodigious learners, constructing everyday theories of the physical and psychological world with remarkable accuracy. I will suggest that children’s limitations in decision-making may actually be responsible in part for their superior learning. The argument is similar to that involving “explore/exploit” trade-offs in the course of reinforcement learning. The skills involved in swift efficient decision-making are in tension with those involved in constructing generally accurate models of the world — although those models are essential for forming the right decisions. I will describe several empirical studies showing that younger learners are better at inferring unusual or unlikely causal hypotheses than older learners, and will suggest that this reflects both the fact that they are less biased by prior knowledge and that they search hypothesis spaces more widely and creatively. The distinctive long immaturity of human children may reflect an evolutionary strategy in which a protected period allowing wide exploration and learning precedes the necessity for accurate decision-making.