CRASH3 plays an important role in the development of accident data files, and in addition it is often used for reconstructing specific accidents. Had it been intended from the outset for this latter purpose, its development may have progressed along slightly different lines. It is now appropriate to examine the use of CRASH3 in specific accidents, the use of available crash test data, and the nature of the likely CRASH3 user. Suitably accounted for, all these factors would have a significant influence on the design of accident reconstruction algorithms, including any contemplated revisions of CRASH3. Such considerations are the subject of this paper.