Objective: The purpose of this study was to determine the relationships between the body composition characteristics, body mass index (BMI), sum of skinfolds (SSF), % body fat (%BF), fat-free mass (FFM) and waist-to-hip ratio (WHR), and nine physical fitness items in undernourished rural primary school children in Ellisras, South Africa. Design: A cross-sectional study. Setting: The study consisted of 462 boys and 393 girls who were aged 7-14 y. Measurements: Five body composition measures were assessed: BMI, SSF, %BF, FFM and WHR. Nine physical fitness test items were assessed: standing long jump, bent arm hang, sit-ups, 10 × 5 m shuttle run, 50 m sprint, 1600 m run, flamingo balance, sit and reach, plate tapping. Results: BMI was highly correlated with FFM (r=0.7, P<0.001). In line with findings from Western countries, regression coefficients (B) showed that children with higher BMI or SSF performed worse in bent arm hang (girls, B= -0.84, P<0.001, and B= -0.06, P=0.02, respectively) and in 1600 m run (B=6.68, P<0.001). BMI was significantly associated with flamingo balance(B=0.26, P=0.04). WHR was positively associated with bent arm hang (B=9.37, P=0.03), and inversely with sit and reach (B= -7.48, P=0.01). In contrast, significant relationships were found between BMI and standing long jump (B=0.74, P=0.04), sit and reach (B=0.51, P<0.001), flamingo balance (B=0.26, P=0.04) and plate tapping (B= -19, P=0.01). SSF was significantly associated with sit and reach (B=0.04, P=0.03). Significant inverse associations were found between FFM and bent arm hang (girls, B= -0.06, P=0.05), 1600 m run (girls, B= -2.33, P=0.003) and 50 m run (boys, B= -0.11, P=0.006). FFM was significantly associated with standing long jump (boys, B=0.99, P<0.001; girls, B=0.73, P<0.001), flamingo balance (B=0.17, P<0.001), and with sit and reach (boys, B=0.59, P=0.03). Conclusion: In the present study in undernourished children, body composition was significantly related to physical fitness, but not always in the expected direction. It is therefore important to note that in this population, BMI should not be interpreted as a measure of fatness/overweight, but rather as an indicator of muscle mass. © 2005 Nature Publishing Group. All rights reserved.