Research focusing on automotive rollovers has garnered a great deal of attention in recent years. Substantial effort has been directed toward the evaluation of rollover resistance. Issues related to crashworthiness, such as roof strength and restraint performance, have also received a great deal of attention. Much less research effort has been directed toward a more detailed study of the rollover dynamics from point-of-trip to point-of-rest.The reconstruction of rollover crashes often requires a thorough examination of the events taking place between point-of-trip and point-of-rest. Increasing demands are placed on reconstructionists to provide greater levels of detail regarding the roll sequence. Examples include, but are not limited to, roll rates at the quarter-roll level, CG trajectory (horizontal and vertical), roll angle at impact, and ground contact velocity.Often the detail that can be provided in a rollover reconstruction is limited by a lack of physical evidence. However, there are many cases where the physical evidence - both on the vehicle and the ground - allows for detailed re-positioning of the vehicle during the roll sequence. Using such a high level of detail would lead one to believe that the reconstruction of the rollover dynamics will be an accurate representation of the actual event. But, just how accurate will it be?This paper seeks to examine the commonly used constant deceleration method of rollover reconstruction in light of detailed analyses of actual rollover tests. By comparing methods used in the field to detailed experimental observations this paper will provide an understanding of the accuracy and limitations of those methods. The underlying data that provide the foundation for this article are derived from two high-speed soft-surface dolly rollover tests of sport utility vehicles.