Driven in part by the need for better information about fluvial systems for the purpose of nonmarine reservoir evaluation and development, much valuable work is now being conducted on modern rivers and their deposits, aided by such techniques as ground-penetrating radar. However, studies of modern and recent systems cannot address the question of the long-term preservability of the present-day deposits. Only studies of the rock record itself can explore this issue. Two separate studies of ancient fluvial systems illustrate some of the problems. A study of the Hawkesbury Sandstone (Triassic, Sydney Basin, Australia), highlighted the difficulty in interpreting the dimensions of large sand bodies from comparisons with a modern analog, even when very large outcrops are available. A seismic time-slice study of Pliocene-Pleistocene fluvial systems in the Gulf of Thailand revealed major changes in channel size and fluvial style over short vertical intervals. Braided and meandering systems (meander-belt widths 4 to > 10 krn [2.5 to > 6 mi]) are separated by a few tens of meters of section, or less, and are interbedded with the deposits of much smaller rivers, showing straight, meandering, and anastomosed patterns. Incised valleys and underfit streams are also present. These variations can be interpreted in terms of a sequence model, but they indicate the problems that could arise from the use of a single suite of dimensional variables as input into numerical reservoir heterogeneity and flow models. Most numerical simulation models make use of sets of equations relating such parameters as channel width, depth, and sinuosity, but most such equations are generalized across the whole spectrum of fluvial styles and can be conditioned to the reality of individual reservoirs only with difficulty. The application of the principles of sequence stratigraphy to fluvial deposits is rendered difficult by the complex response to allogenic forcing that characterizes fluvial systems. Episodes of aggradation and degradation that may be used to define sequences, and their bounding unconformities in the stratigraphic record may be the result of the complex interplay of several allogenic mechanisms governing varying stream power and sediment supply, mechanisms that may be operating at different time scales and may be out of phase with each other. In developing practical solutions for reservoir development, numerical modeling and simulation may provide generalized starting points for the analysis. History matching commonly demonstrates inaccuracies in many initial models. Further progress may be made by direct study of the reservoir itself, using three-dimensional (3-D) seismic and surveillance techniques. There is a continuing role for the study of ancient analogs as providing a realistic database on the long-term preservation styles of fluvial reservoir deposits.