It is often said that experimental performances of a two-dimensional retarded cascade are not reproducible. In this paper, the technique of cascade testing to obtain the relaible data is presented. Using this method, it is possible to control the inlet and exit flow conditions, so that all the blades of a cascade are operating under the same flow conditions and the axial velocity ratios are equal to unity. Then, the allowable value of axial velocity ratios, which is necessary to ensure the two-dimensional flows, are evaluated. As the inlet angle increases, it becomes more difficult to get the two-dimensionality of flow, and the critical value of axial velocity into approaches unity. Some correcting methods for the cascade performances are discussed when the axial velocity ratio is not kept under the critical value. As the result, some cascade performances at extremely high inlet angle are obtained and their reliability is confirmed by a rotating by cascade test.