Long-term discharge records (17–81 yr) of 78 streams from across the continental United States were analyzed to develop a general quantitative characterization of streamflow variability and predictability. Based on (1) overall flow variability, (2) flood regime patterns, and (3) extent of intermittency, 11 summary statistics were derived from the entire record for each stream. Using a nonhierarchical clustering technique, nine stream types were identified: harsh intermittent, intermittent flashy, intermittent runoff, perennial flashy, perennial runoff, snowmelt, snow + rain, winter rain, and mesic groundwater. Stream groups separated primarily on combined measures of intermittency, flood frequency, flood predictability, and overall flow predictability, and they showed reasonable geographic affiliation. A conceptual model that incorporates the nine stream clusters in a hierarchical structure is presented. Also, the positions of the 78 streams in a continuous three-dimensional flow space illustrate the wide range of ecologically important hydrologic variability that can constrain ecological and evolutionary processes in streams. Long-term daily streamflow records are a rich source of information with which to evaluate temporal and spatial patterns of lotic environments across many physiographic and ecographic regions. Relative positions of streams in flow space provide a conceptual framework for evaluating a priori the relative importance of abiotic and biotic factors in regulating population and community processes and patterns.