Computer models are increasingly used in the design of crash protection systems, yet the experimental data available for benchmarking these models is limited. The goal of this study was to apportion the contributions of the rib cage, internal viscera and flesh to the mechanical response of the thorax by repeated dynamic loading of human cadavers. Three post-mortem human thoraces in three tissue states (intact, denuded and eviscerated) were subjected to dynamic anterior loading using four loading conditions (distributed, single diagonal belt, double belt and hub). The thoraces were significantly stiffer intact than eviscerated or denuded (paired t test p < 0.05). The distributed (598.7 N/cm) and double belt (529.2 N/cm) intact conditions were the stiffest, followed by the single belt (383.5 N/cm) and then the hub (170.0 N/cm). The degree to which the soft tissues influenced the response depended on the loading. For the hub and distributed conditions, the denuded thoraces retained 60% of the intact stiffness, and the eviscerated thoraces retained 30%. The single and double belt conditions were not as sensitive to tissue state (85% and 55%). The two loading conditions that did not engage the shoulder were highly sensitive to removing the soft tissues, while the two conditions that engaged the shoulder were less sensitive. This is important benchmarking data for computational thoracic models.