Because of the lack of pediatric biomechanical data, the Hybrid-III (HIII) child anthropomorphic test devices (ATDs) are essentially scaled from the mid-size male ATD, and are often criticized for its rigid spine comparing to those from children. In this study, possible design modifications for improving the spine biofidelity of the HIII 6-year-old ATD were explored by child cadaver/volunteer test reconstructions and accident reconstructions using computational modeling and optimization techniques. It was found that the translational characteristics of the cervical and lumbar spine in the current child ATD need to be reduced to achieve realistic spine flexibility. It was also found that adding an additional joint at the thoracic spine region with degree of freedom in both flexion/extension and tension can significantly improve the ATD biofidelity in terms of predicting the overall spine curvature and head excursion in frontal crashes.