Three-component (throughfall, soil water, groundwater) hydrograph separations at peak flow were performed on 10 storms over a 2-year period in a small forested catchment in north-central Maryland using an iterative and an exact solution. Seven pairs of tracers (deuterium and oxygen 18, deuterium and chloride, deuterium and sodium, deuterium and silica, chloride and silica, chloride and sodium, and sodium and silica) were used for three-component hydrograph separation for each storm at peak flow to determine whether or not the assumptions of hydrograph separation routinely can be met, to assess the adequacy of some commonly used tracers, to identify patterns in hydrograph-separation results, and to develop conceptual models for the patterns observed. Results of the three-component separations were not always physically meaningful, suggesting that assumptions of hydrograph separation had been violated. Uncertainties in solutions to equations for hydrograph separations were large, partly as a result of violations of assumptions used in deriving the separation equations and partly as a result of improper identification of chemical compositions of end-members. Results of three-component separations using commonly used tracers were widely variable. Consistent patterns in the amount of subsurface water contributing to peak flow (45--100%) were observed, no matter which separation method or combination of tracers was used. A general conceptual model for the sequence of contributions from the three end-members could be developed for 9 of the 10 storms. Overall results indicated that hydrochemical and hydrometric measurements need to be coupled in order to perform meaningful hydrograph separations. ¿ 1998 American Geophysical Union