Data analysis in the computational sciences is typically made difficult by the large amount of information that must be correlated to make sense of measurements or simulations of many physical phenomena. As a result, visualization has become a critical tool in understanding these relationships among large volumes of data . However, visualization techniques are most effective if they are interrogative, that is, as long as they allow scientists to query and manipulate the visualized data at interactive rates. In the case of physical simulations, interrogative visualization capabilities are a key enabler for effective computational steering, allowing the scientist to control a simulation as it progresses through a tight coupling of the simulation and the visualization of the data it. These capabilities are thus dependent on the speed with which data analysis and display can be performed, scientists cannot control the production of their data nor can they query it until they have seen it displayed in an understandable way. Since the volume of data collected about many physical phenomena ever more sophisticated measuring instruments is growing much faster than the capabilities of analysis and display hardware, provision of meaningful interrogative visualizations of such datasets is an increasingly challenging. A similar situation exists with respect to the growth of data volumes produced by physical simulation programs, but in this case interrogative visualization capabilities are even more crucial if computational steering is to be done.