This mixed-methods study was designed to compare the learning gains of seventh-grade students (N = 417) taught a 4-week conceptual change unit on density using either a student-centered, guided inquiry-based approach or a more direct, teacher-centered instructional strategy. Application of a multilevel model to data obtained from the administration of pre-/post-assessments demonstrated that while students in both conditions made statistically significant learning gains across the study period, average performance gains did not vary by instructional condition. A cross-level interaction between instructional method and students' prior mathematics performance and subsequent learning gains was identified however. Within the guided inquiry condition, students with the lowest initial math scores had the largest learning gains, while the reverse was true in the teacher-centered, direct instruction condition, where initially higher-performing students demonstrated the strongest learning gains. Results associated with an authentic task assessment and systematic interviews of student participants provided additional context for understanding the nature of the performance outcomes. Implications for middle school science instruction are discussed.