The purpose of this study was to examine the impacts of embedding collaboration into a game with a self-explanation design for supporting the acquisition of light and shadow concepts. The participants were 184 fourth graders who were randomly assigned to three conditions: a solitary mode of the game with self-explanation, a collaborative mode with self-explanation, or the control condition of a single-user game without integrating self-explanation. Students' conceptual understanding was measured through an immediate posttest and a retention test with a three-week delay. Further, students' engagement in answering the prompts was also investigated. The findings showed that having students collaboratively play science-based games with a self-explanation design embedded was not sufficient to help them learn the science concepts. Rather, it was the level of engagement in responding to the self-explanation prompts that mattered.