The aim of this study was to investigate the effects of incorporating self-explanation principles into an educational game facilitating students’ conceptual learning about light and shadow. A total of 88 third graders participated in this true-experimental study in which each student was randomly assigned to either an experimental group that played a game with self-explanation prompts or a control group that played a game with no prompts. Students’ conceptual understanding was evaluated through a pretest, an immediate posttest, and a retention test that took place two weeks afterwards. The findings revealed that simply exposing the students to a game with a self-explanation design did not necessarily enhance their learning performance. Rather, it was the level of engagement in responding to the self-explanation prompts that mattered. The students in the high-engagement group outperformed those in the low-engagement and control groups in terms of the posttest and the retention test. Possible improvements for integrating self-explanation principles into computer games are discussed.