0 have a great deal to learn about how to support collaboration among researchers. Al though tools thatsupport things likejoint authoring and brainstorming may apply to all disciplines, the real and perceived needs of disciplines will differ. 4) Data fusion. The mere existence of large amounts of on-line data will encour-age the analysis of phenomena observed by multiple sensors in multiple locationswhether the use of the data in that way was planned or not. This requires that the data not only be accessible and manipulable, but understood. Techniques need to be devel-oped for data description that facilitates use by disparate applications and enables inte-gration from heterogeneous sources in a form that leads to an understanding of the overall problem. 5) The modeler's workbench. Currently there is a gap between the development of a mathematical model and the creation of the computer program to simulate it. That discontinuity is apotential source of error and an opportunity for a divergence between the simulation and the original model. Languages and systems-a modeler's workbench if you will-to make this a single, seamless process would greatly am-plify the effectiveness of modelers. Each of the above is just a sample of the rich opportunities to better support scientific investigation. In fact, they are the rather obvious examples. The pace of the improvement of information technology makes it easier to see where the technology will be