Discovery learning is generally seen as a promising but demanding mode of learning that, in most cases, can only be successful if students are guided in the discovery process. The present article discusses a study on discovery learning with a computer simulation environment in the physics domain of collisions. In the learning environment, which is called Collision, students learned about collisions where two particles move in the same direction and interact via a conservative force in such a way that the total mechanical energy is conserved. In the experiment we conducted with Collision, we evaluated the effects of adding two different ways to guide students: model progression, in which the model is presented in separate parts; and assignments, small exercises that the student can choose to do. The effect of providing assignments and model progression was evaluated by comparing the learning behavior and learning results over three experimental conditions in which different versions of the simulation environment were presented: pure simulation, simulation plus assignments, and simulation plus model progression and assignments. Students' use of the environment was logged, their subjectively experienced workload was measured on-line, and their learning was assessed using a number of assessment procedures. Providing assignments with the simulation improved students' performance on one aspect of a so-called intuitive knowledge test. Providing the students with model progression did not have an effect. A subjective workload measure indicated that expanding the simulation with assignments and model progression did not raise the workload experienced by the students.