Computational models of the impacts of Comet Shoemaker-Levy 9 onto Jupiter may provide the best framework by which the observational data can be interpreted. Among the observations that have already been at least partially explained in a way that appears to be consistent with the impact models are: the sources and timings of multiple flashes observed from Earth, the temperatures and durations of the single flashes observed from the Galileo spacecraft, and the asymmetry of the plumes and ejecta patterns observed by the Hubble Space Telescope. Further modeling subsequent to the impacts has shown that (contrary to our pre-impact expectations) fireball trajectory data do not provide strong constraints on either fragment mass or maximum penetration depth. Instead, it is the cross-sectional area of the fragment (or swarm of sub-fragments) at the time of impact that determines the ejection velocity and trajectory of the fireball. The observation of seemingly consistent plume heights, coupled with this computational result, suggests that SL-9 fragments were loosely-bound ''rubble piles,'' possibly with widely varying masses, that in most cases dispersed to about the same diameter (2.0¿0.5 km) by the time they reached the Jovian atmosphere. After more data become available and correlated, and more simulations are performed, we expect that fragment size estimates will become more precise. ¿ American Geophysical Union 1995