As an effective method of studying soil anisotropy, hollow cylinder torsional shear (HCTS) tests have been performed extensively on sedimentary soil, thereby establishing the anisotropic behavior of sand and clay. However, little is known about the anisotropic behavior of granite residual soil (GRS) formed by weathering, partly because hollow cylindrical specimens of natural GRS have yet to be prepared successfully, hence the lack of HCTS tests performed on GRS. The unique geotechnical properties of GRS, including high intact strength, susceptibility to disturbance, and the minor fissures in soil, pose great challenges when trying to prepare natural specimens. This paper proposes a new method to address this issue, which involves preparing a solid cylindrical specimen, drilling an initial inner cavity and then enlarging it, and finely trimming the specimen to its ultimate dimensions. Two methods are proposed for enlarging the inner cavity, both of which work well. The reliability of the proposed method is confirmed through specimen quality as indicated by the limited void ratio change due to reconsolidation, test repeatability, and comparison of the results of HCTS tests with those of triaxial tests. The HCTS tests reveal the shear strength anisotropy of natural GRS. Although specified for residual soil derived from granite, the proposed method could also be used for other weathered materials.