The Hybrid III 6 year‐old anthropomorphic test device has limited pelvis and abdomen biofidelity, limiting its ability to discriminate good and poor restraint performance in dynamic events. The objective of this study is to determine whether the Hybrid III 6 year‐old with modified pelvis and gel abdomen provides better discrimination of restraint performance than the standard 6 year‐old in a series of sled tests with highback, backless, and no booster conditions. Sixteen paired sled tests were conducted in a vehicle buck subjected to a 56 km/h crash pulse simulating a frontal impact. Injury and kinematic measures were calculated and the population means were compared between the standard and modified 6 year‐old using paired sample t‐tests. Overall kinematics of the standard and modified 6 year‐old were similar for measures such as head acceleration, chest acceleration and compression, and head excursion. The modified 6 year‐old had 257% larger abdominal penetrations, 19% smaller anterior superior iliac spine loads, 5% larger knee excursions, 70% larger knee – head excursions, and 25% smaller torso rotations, all potential indicators of the propensity to submarine. However, these submarining indicators did not necessarily correspond to those test conditions more likely to result in submarining in real‐world data, such as initial belt placement on the abdomen.