Achievement goals and self-efficacy are both found to be related to students' achievement-related cognition, affect, and behavior. But recent research evidence suggests that self-efficacy may mediate relationships between achievement goals and achievement behaviors (Cumming & Hall, 2004). Such mediation, however, has not been examined within the 2 x 2 achievement goal model (mastery-approach [MAp], mastery-avoidance [MAv], performance-approach [PAp], and performance-avoidance [PAv]) in the domain of physical education. Therefore, the purpose of this study was to explore the mediating effect of self-efficacy on the relationships between achievement goals and fitness testing scores among secondary physical education students. Participants were 276 (115 boys, 161 girls) students in grades 6 - 9 from a suburban public school. They completed questionnaires assessing their achievement goals and self-efficacy toward fitness testing, and took a FITNESSGRAM test (consisting of BMI, PACER, curl-up, trunk-lift, and push-up scores) at the end of the school year. The test scores were converted to T-scores and served as dependent variables. Structural equation modeling analysis using Amos 5.0 was conducted with self-efficacy as a mediating variable and achievement goals as independent variables. Fit statistics indicated an adequate fit between the model and the data (χ2 (22, N = 276) = 55.6, p > .05; CFI = .92, TLI = .93, IFI = .93, RMSEA =. 06). The results revealed MAp (b = .48, p < .01), PAp (b = .20, p < .01), and MAv (b = -.16, p < .01) significantly influenced self-efficacy and accounted for 37% of the variance for self-efficacy. Self-efficacy exhibited direct effect on BMI (b = -.17, p < .01), PACER (b = .22, p < .01), curl-up (b = .33, p < .01), and push-up (b = .24, p < .01). Additionally, MAv and PAp both directly and indirectly influenced PACER and push-up. They also indirectly influenced BMI and curl-up through self-efficacy. Finally, MAp had indirect effects on these four tests scores through self-efficacy. All the direct and indirect influences of the achievement goals accounted for 3%, 19%, 11%, and 18% of the variance for BMI, PACER, curl-up, and push-up respectively. Overall, the results support the mediating function of self-efficacy between achievement goals and fitness testing scores. They also revealed students' achievement goals only accounted for part of the testing scores, suggesting that other important motivational factors such as interest and task values should be taken into account in future research.