Roller compacted concrete (RCC) stepped spillways are growing in popularity for providing overtopping protection for aging watershed dams with inadequate auxiliary spillway capacity and for the construction of new dams. Unobtainable land rights, topographic features, and land use changes caused by urbanization limit the ability to construct new dams or modify the dimensions of existing embankments and spillways. The advantages of stepped spillways are (1) they can be placed over the top of an existing embankment without causing significant changes to the dam or spillway dimensions, (2) they provide considerable energy dissipation in the chute, potentially reducing the size of the stilling basin, and (3) they permit shorter, more efficient, and feasible construction schedules than other design options. Currently, limited design guidelines are available in the literature for the design of stepped spillways constructed on flat slopes (&thetas; < 30°). Auxiliary spillways are designed to safely pass exceptionally large flood events to the downstream channel. In structural auxiliary spillways, spillway chute and stilling basin training walls are typically designed to prevent overtopping. However, the aspect of converging training walls increases the flow depth in the chute near the walls, and it changes the hydraulic parameters for designing the stilling basin. To assist with the design of converging stepped spillways having similar design parameters (i.e. chute slope, step height, etc.), a study utilizing a three-dimensional, 1:22 scale physical model was conducted to evaluate the flow characteristics in the spillway. This study is the first known attempt at developing generalized design criteria for converging stepped spillways having vertical training walls. Conclusions drawn from this study are that as the convergence of the training wall increases the flow depth near the wall also increases. A simplified control volume momentum analysis was used to predict the minimum training wall height necessary to prevent overtopping. The equation developed slightly under-predicted the results. This under-prediction may be a result of the assumptions made in the development of the prediction equation. Other design aids for determining training wall height were developed based on observations with the data. The results of the study will be discussed further herein.