It has been known for some time that when a cylinder spans a wind tunnel the boundary layers on the walls of the tunnel may have important effects, especially on the pressure in the base region. By making the boundary layers thin by the use of end plates, it may be possible to approach a complete elimination of the effects of these layers on the central part of the span and obtain a flow that is very nearly two-dimensional. Most of the published pressure distributions for circular cylinders were obtained without end plates and this probably explains their inconsistency. Gerrard has described experiments in which the addition of end plates caused a substantial change of base pressure coefficient but he did not attempt to design an optimum end plate to give nearly two-dimensional flow.