Evaluating the relation between evidence and theory should be a central activity for science learners. Evaluation comprises both hypothetico-deductive analysis, where theory precedes evidence, and inductive synthesis, where theory emerges from evidence. There is mounting evidence that induction is an especially good way to help learners grasp the deep structure (i.e., underlying principles) of phenomena. However, compared to the clear falsification logic of hypothetico-deductive analysis, a major challenge for induction is structuring the process to be systematic and effective. To address this challenge, we draw on Sir Francis Bacon's original treatise on inductive science. In a pair of experiments, college students used a computer simulation to learn about Faraday's law. In the inductive conditions, students sought a general explanation for several cases organized according to Bacon's tenets. In contrast, other students used a more hypothetico-deductive approach of sequentially testing (and revising) their hypotheses using the simulation. The inductive activity led to superior learning of a target principle measured by in-task explanations and posttests of near transfer and mathematical understanding. The results provide two important pieces of information. The first is that inductive activities organized by Bacon's tenets help students find the deep structure of empirical phenomena. The second is that, without an inductive "push," students tend to treat instances separately and fail to search for their underlying commonalities.