Shoulder and lap belt scores have been previously quantified for anthropomorphic test devices on belt-positioning booster seats;​ however, they may not fully discriminate between good vs. poor dynamic outcomes. To determine the influence of initial belt fit and gap on dynamic outcomes, the ability of anthropomorphic test devices to represent realistic child belt fit and gap on belt-positioning booster seats must first be understood. This study compares posture, belt fit, and belt torso contact between the Hybrid III (HIII) 6- year-old, HIII 10-year-old, HIII 5th percentile Female, and the Large Omni Directional Child 10-year-old anthropomorphic test devices to a cohort of 50 children on ten belt-positioning booster seats and three seatbelt anchor locations. Novel belt fit metrics (e.g., gap size, gap length, and belt torso contact) and conventional shoulder and lap belt scores were quantified using a 3D coordinate measurement system. Overall, the anthropomorphic test devices had more inboard shoulder belt position, overestimated gap size and length, and underestimated belt torso contact compared to children;​ however, anthropomorphic test device posture, belt fit, and belt gap outcomes were moderately or strongly associated with child outcomes, suggesting that they can be used to represent realistic variation in initial belt fit and gap conditions.