A watershed is represented by a serial multistage system wherein each stage employs a generalized conceptual element. In the order that they 'process' rainfall, the stages are: initial abstractions, detention storage, translation, intermediate abstractions, channel storage, final abstractions, and outlet storage. The specific character of each stage is determined using data from an actual watershed and a hybrid computer. Data compiled from a sixteen acre tract on Atterbury Experimental Watershed near Tucson, Arizona are used. A braod spectrum of recorded conditions gives rise to problems typical of many involving rainfall-runoff relations on aridland watersheds. Infiltration becomes the important initial abstraction, and Horton's equation is modified to make capacity infiltration rate a function of time and precipitation intensity. The analysis relates depression losses to storm variables, allowing estimation of the starting time of runoff. Included is a provision for subsequent depression losses that may occur, for example, during multiple-peak flows. The translation model as developed permits shifting the computed hydrograph to coincide with actual runoff. Linear reservoirs are used in storage stages. Limitations of the approach are diminished by a suggested parameter matrix that increases reliability and predictions. Despite their limitations continued use of conceptual systems as a framework within which to discuss aridland watershed characteristics appears warranted. Comparative watershed modeling with analog, digital and hybrid computers is critically evaluated. (Author)