Types of learning with a strong emphasis on the responsibility of the learner (such as discovery learning) are gaining popularity over traditional forms of (expository) instruction. Discovery learning distinguishes itself by the central role of learning processes such as hypothesis generation (induction), experiment design, and data interpretation. Expository instruction pays more attention to directly ‘exposing’ definitions and equations to learners. In the current study, students worked with either a simulation (discovery learning) or a hypertext learning environment (expository instruction) with the same domain content. Each of the environments contained a large number of assignments. The study followed a pre-test, post-test design. To measure the knowledge acquired a definitional knowledge test, an intuitive knowledge test (where both correctness and speed of answering are aspects that are measured) and a test in which relations needed to be explained were administered. It was predicted that the hypertext group would outperform the simulation group on the definitional knowledge test and it was expected that the simulation group would perform better on the intuitive knowledge test. Results showed that both the interaction with the simulation and with the hypertext resulted in substantial learning gains. It was found that the hypertext group performed better on the definitional knowledge test. On the intuitive knowledge test the hypertext scored better than the simulation group on the correctness of the items but not on the time needed to answer items. On the explanation test there was no difference between the two groups. An analysis of interaction processes as recorded in the logfiles indicates that the differences between both environments in their actual usage were less distinctive than expected. In the simulation group many students followed the assignments given and did not engage in self-guided discovery. Since the assignments were rather directive, this resulted in ‘discovery behaviour’ that focused on generating outcomes; outcomes that were also, and more directly presented in the hypertext environment. For research and practice, this implies that simulations are to be considered only when clear benefits of discovery are expected, and only with complex domains, sufficient learning time and freedom for students in the assignments to engage in discovery.