The widespread implementation of air bags has increased the incidence of upper extremity injuries in the automotive crash environment. The first step in reducing these injuries is to determine upper extremity injury criteria. The small female was used to establish the criteria in order to provide a conservative estimate for the general driving population. The forearm, humerus, elbow, and wrist injury criteria were determined by utilizing a three phase methodology. First, computer simulations were performed to determine the upper extremity loading patterns and the significance of experimental boundary conditions under frontal and side air bag impact. Second, the upper extremities of whole cadaveric specimens were impacted with both frontal and side air bags to confirm the computer model’s predictions, and to determine injury patterns and loading rates. The third phase involved controlled impact loading of disarticulated cadaveric upper extremities such that the applied load was directly quantified. Mass scaling and statistical analysis of the component test data were used to develop injury risk functions for the fifth percentile female. The criteria were established with a 50% risk of injury at 128 Nm bending of the humerus, 58 Nm bending of the forearm, 1832 N axial loading of the elbow, and 1700 N axial loading of the wrist. A modified dummy upper extremity was designed for the evaluation of frontal and side air bag interactions. It is recommended that the injury criteria be implemented directly in the dummy upper extremity. This recommendation was justified by comparing observed injury rates to those predicted by the dummy upper extremity in matched air bag tests, and through a detailed beam analysis of the differences between the dummy and human upper extremity.