Prototype tests were conducted during 11-14 August 1986 at Taylorsville Dam, Kentucky, to evaluate the performance of the project's selective withdrawal system. Taylorsville Dam is located on the Salt River in north-central Kentucky 50 miles above the confluence with the Ohio River. The existing project consists of a rock-filled dam, uncontrolled spillway, and controlled outlet works. Reservoir releases are regulated by a gated intake tower consisting of two flood-control intakes at the base of the structure and two wet wells with five 6- by 6-ft water-quality intakes in each wet well. All flows pass through two separate 5.5- by 14.75-ft rectangular passages transitioning into a single 11.5- by 14.75-ft oblong conduit. The primary purpose of the prototype measurement program was to obtain prototype information on the performance of the selective withdrawal system. The data are used to determine the reservoir withdrawal zone characteristics, intake tower blending characteristics, the occurrence of density blockage, the degree of mixing of different water qualities, and the amount of dissolved oxygen uptake. Results were also used to compare prototype performance with both physical and numerical model predictions. The basic measurements included dissolved oxygen and temperature profiles in the reservoir, at locations within the outlet works, and at one station in the downstream channel. Two-directional intake velocity profiles were measured for each port and total discharge was measured in the downstream channel. The water-quality data show that the Taylorsville water-quality intake structure can function effectively for release temperature control by selective withdrawal and that the system effectively reaerates flow through the structure. The results also show that the one-dimensional-numerical model SELECT can predict release dissolved oxygen concentration and temperature for Taylorsville. Density blockage of the upper ports due to reservoir density stratification occurs during the multilevel tests at flows less than 100 cfs. The hydraulic measurements revealed basically uniform flow distributions within each intake. The prototype data for submerged orifice flow fall slightly below the physical model data. Discharge coefficients for the total water-quality-system were computed to be less than those determined in the model.