The objective of this study was to investigate the effects of age group, walking speed, and body composition on the accuracy of pedometer-determined step counts in children. Eighty-five participants (43 boys, 42 girls), ages 5-7 and 9-11 years, walked on a treadmill for two-minute bouts at speeds of 42, 66, and 90 m x min(-1) while wearing a spring-levered (Yamax SW-200) and a piezoelectric (New Lifestyles NL-2000) pedometer. The number of steps taken during each bout was also recorded using a hand counter Body mass index (BMI) was calculated from height and mass, and percentage of body fat (%BF) was determined using hand-to-foot bioelectrical impedance analysis. The tilt angle of the pedometer was assessed using a magnetic protractor. Both pedometers performed well at 66 and 90 m x min(-1), but undercounted steps by approximately 20% at 42 m x min(-1). Although age group, BMI, waist circumference, and %BF did not affect pedometer accuracy, children with large pedometer tilt angles (> or =10 degrees) showed significantly greater percent bias than those with small tilt angles (< 10 degrees). We suggest that the style of waistband on the child's clothing is a more important determinant of tilt angle and thus pedometer accuracy than body composition. Our results also indicate that the NL-2000 pedometer provides similar accuracy and better precision than the SW-200 pedometer, especially in children with large tilt angles. We conclude that fastening pedometers to a firm elastic belt may improve stability and reduce undercounting in young people.