Designing vehicles and roadside structures that are safer in side-impact collisions is an emerging area of concern in roadside safety research. Selecting impact conditions that are relevant to the way side-impact collisions occur in real world collisions is an important part of developing effective full-scale crash test procedures and evaluation criteria. If test impact conditions are unrealistically severe, improving the performance of roadside hardware for side-impacts may appear unfeasible. If test impact conditions are not demanding enough, good performance in full-scale crash tests may not be indicative of good performance in real-world crashes. The purpose of this paper is to examine the best available accident data to determine what the reasonable worst case test impact conditions are for side-impacts with roadside objects.