This study aimed at comparing the effectiveness of problem-based learning (PBL)(1) and traditional lecture-based instruction on elementary school students' academic achievement and performance skills in a science unit on genetics while controlling for their reasoning ability. For the specified purpose, two instructional methods were randomly assigned to intact classes of two different teachers. Each teacher had both PBL classes and traditional classes. Although students in PBL classes (n = 126) worked on ill-structured problems cooperatively with the guidance of the teacher, students in traditional classes (n = 91) received instruction based on teacher's explanations, discussions, and textbooks. Genetics Achievement Test was developed by researchers to measure the academic achievement and performance skills. Multivariate Analysis of Covariance results showed that the PBL students had higher academic achievement and performance skills scores (M = 11.44 and M = 2.67, respectively) when compared with those in traditional classes (M = 10.91 and M = 2.20, respectively). This indicated that the PBL students tend to better acquire scientific conceptions related to genetics and integrate and organize the knowledge. Moreover, it was found that the reasoning ability explained a significant portion of variance in the scores of academic achievement and performance skills.