The present study examined the relationship among learners' differences, behaviors in manipulating variables, and learning achievements in a simulation-based program that supports discovery learning in the subject of C-programming algorithm. Participants (n = 66) took the Group Embedded Figures Test, Action Control Scale, and Computer Self-Efficacy Scale to determine their field dependence, action-state orientation, and computer self-efficacy before interacting with the simulation program. Results from this study have shown that reading domain information supported learners to manipulate more simulation variables systematically. In addition, manipulating simulation variables in a systematic manner positively affects learning post-test achievement. In the context of individual differences, failure-related action control and computer self-efficacy were found to affect learners' behaviors in manipulating variables, while field dependence was shown to affect learning post-test achievement. Implications for both theoretical and instructional designs are discussed in terms of how various support types can accommodate the needs of different learners in simulation programs.