Summary This paper shows how some simple 3D graphics tools can be combined to provide efficient soft-ware for visualizing and analyzing data obtained from reservoir simulators and geological simulations. The animation and interactive capabilities of the software quickly provide a deep understanding of the fluid-flow behavior and an accurate idea of the internal architecture of a reservoir. Introduction Reservoir geologists and engineers often deal with 3D data from simulation programs. These data may represent lithofacies programs. These data may represent lithofacies or petrophysical values in 3D space. Some data may also vary in time. Without an adaptable way to represent-these numbers, it is difficult to understand their underlying physical phenomena in depth. Without physical phenomena in depth. Without visualization tools one must imagine a complex 3D structure (e.g., the architecture of a reservoir) or try to understand a movement within this structure (e.g., the fluid flow). We describe a software environment that allows scientists to explore, manipulate, and visualize their data interactively and dynamically on a workstation. With this software, the user can view precomputed data from any angle, can sort through the volume of data, and can even see a film from the data, if these data are evolving in time. Our aim is not to make general-purpose graphics software for visualization of data from molecular chemistry, computational fluid dynamics, or oil reservoir engineering. Our concern is to provide reservoir-simulation, reservoir-geology, and basin-modeling scientists with attractive, standard, 3D-graphics tools that meet their specific needs. Examples of the specific needs encountered are described below. To handle lithofacies and petrophysical values, we introduce the distinction between discrete and continuous scalars; i.e., no interpolation between data is required for discrete scalars (interpretation between clay and sandstone makes no sense), but interpolation is authorized for continuous scalars. This implies that the calculations and the graphics representation of isosurfaces or isocontours will differ for continuous and discrete data. In 3D basin modeling, the simulation of the deposition of sedimentary bodies may need time-varying grids (e.g., the number of layers may depend on time). A specific grid management is then designed so that an animated visualization of the deposition process is available on the workstation process is available on the workstation screen. An interpolation in time is introduced for reservoir-simulation data to avoid having to store all simulation steps and to allow visualization of a smooth movement of fluid flow. Visualization of wells, cutting planes defined by two vertical wells, and successions of cutting planes defined by successions of interactively chosen wells are also among the specific needs encountered in the scientific fields for which our visualization software is designed. Data To Be Visualized Concept of Discrete and Continuous Data. We handle scalar data (lithofacies, pressure, saturation, etc.), not vector data. Data are computed on a grid that is described later. We distinguish two kinds of data: discrete and continuous. Discrete data are lithofacies values, and continuous data include such scalars as pressure, saturation, and porosity. Note that most general-purpose porosity. Note that most general-purpose visualization packages handle only continuous data, such as oil saturation, which may take an infinity of values within some authorized range (say between 0 and 1). Moreover, in such packages, it is implicitly assumed that the numerical values to be visualized are point samples of a continuous function point samples of a continuous function (continuous in space), which cannot be the case for lithofacies values. The set of lithofacies values the user deals with is finite, and there is no possible continuity. Lithofacies values actually are attributes rather than numerical values, although the simulation programs provide them to the visualization system as provide them to the visualization system as numerical values. Because discrete and continuous data are intrinsically different, their graphics representations do not have the same meaning. As a matter of fact, when continuous data are visualized with colors on a computer screen, each color represents a class of valuesnot just onebecause the number of available colors is finite. Even though the colored picture on the screen may look smooth, an infinite range of numerical values is approximated by a finite set of colors. On the contrary, the color representation for discrete values is not an approximation because there is an exact correspondence between a color and a value.