Ground water discharge and recharge frequently have been estimated with hydrograph-separation techniques, but the critical assumptions of the techniques have not been investigated. The critical assumptions are that the hydraulic characteristics of the contributing aquifer (recession index) can be estimated from stream-discharge records; that periods of exclusively ground water discharge can be reliably identified; and that stream-discharge peaks approximate the magnitude and tinting of recharge events. The first assumption was tested by estimating the recession index from st earn-discharge hydrographs, ground water hydrographs, and hydraulic diffusivity estimates from aquifer tests in basins throughout the eastern United States and Montana. The recession index frequently could not be estimated reliably from stream-discharge records alone because many of the estimates of the recession index were greater than 1000 days. The ratio of stream discharge during baseflow periods was two to 36 times greater than the maximum expected range of ground water discharge at 12 of the 13 field sites. The identification of the ground water component of stream-discharge records was ambiguous because drainage from bank-storage, wetlands, surface water bodies, soils, and snowpacks frequently exceeded ground water discharge and also decreased exponentially during recession periods. The timing and magnitude of recharge events could not be ascertained from stream-discharge records at any of the sites investigated because recharge events were not directly correlated with stream peaks. When used alone, the recession-curve-displacement method and other hydrograph-separation techniques are poor tools for estimating ground water discharge or recharge because the major assumptions of the methods are commonly and grossly violated. Multiple, alternative methods of estimating ground water discharge and recharge should be used because of the uncertainty associated with any one technique.