The purpose of this study is to evaluate the hard tissue injury-predictive value of various thoracic injury criteria when the restraint conditions are varied. Ten right-front passenger human cadaver sled tests are presented, all of which were performed at 48 km/h with nominally identical sled deceleration pulses. Restraint conditions evaluated are 1) force-limiting belt and depowered airbag (4 tests), 2) non-depowered airbag with no torso belt (3 tests), and 3) standard belt and depowered airbag (3 tests). Externally measured chest compression is shown to correspond well with the presence of hard tissue injury, regardless of restraint condition, and rib fracture onset is found to occur at approximately 25% chest compression. Peak acceleration and the average spinal acceleration measured at the first and eighth or ninth thoracic vertebrae are shown to be unrelated to the presence of injury, though clear variations in peaks and time histories among restraint conditions can be seen. The maximum viscous criterion is found to correspond with injury, but only because it increases with the maximum chest compression. A simple analytical study is presented to elucidate the observed restraint condition dependence of rib fracture location and the restraint insensitivity of injurious maximum chest compression. Computed tomography images of a loaded torso are presented to show the load-distributing effect of the soft tissues superficial to the rib cage.