A total of 20 full-scale frontal sled tests were conducted using the Hybrid III (HIII), THOR-M and post-mortem human surrogates (PMHSs) to evaluate the thoracic biofidelity of the HIII and THOR-M under various belted restraint conditions. Each surrogate was tested under three belted restraint conditions:​ knee bolster, knee bolster and steering wheel airbag, and knee bolster airbag and steering wheel airbag. In order to assess the relative biofidelity of each ATD, external thoracic deflections were quantitatively compared between the ATDs and PMHSs using an objective rating metric. The HIII had slightly higher biofidelity than the THOR-M for the external thoracic deflections. Specifically, the THOR-M lower chest was more compliant compared to the other surrogates. However, the THOR-M exhibited expansion of the lower chest opposite belt loading, which was also observed to some degree in the PMHSs. The efficacy of the current injury risk prediction instrumentation and criteria were also evaluated for each surrogate. The THOR-M and its proposed injury risk criteria predicted the injuries observed in the PMHS tests better than the HIII. The PMHS injury criteria over-predicted the amount of chest deflection necessary to produce a severe injury and, consequently, under-predicted injury risk. The results of this study indicate that further testing should be performed to evaluate the biofidelity of the THOR-M thorax under more conditions. Furthermore, current thoracic injury risk criteria, which were developed using censored data, may not be effective at predicting injuries for all restraints and experimental conditions.