Abstract: Due to inherent uncertainties associated with laboratory testing, such as sampling disturbance and size effect, and the limitation that conventional in-situ large scale shear tests are generally carried out for soil masses of natural water content, researchers always use soil mass strength parameters determined by inversion methods instead of testing for geotechnical structural stability analyses and engineering designs. In this study, based on conventional in-situ large scale shear test apparatus, the authors developed two modes of apparatus, i.e., the vertical soaking mode and the lateral soaking mode, to accommodate saturated testing techniques. With the apparatus, seven in-situ large scale shear tests were carried out on saturated loess masses in Heifangtai, Gansu Province. The results show that the loess masses could be saturated by both modes. Compared with the lateral soaking technique, vertical soaking often causes more sample disturbance or even leads to failure although it takes less time to saturate. A comparison of parameters from in-situ large scale shear tests of the loess masses with natural water content shows that cohesion from in-situ large scale shear tests of the saturated loess masses decreased sharply from 44.65kPa to 17.35kPa, while internal friction angle decreased marginally from 14.18° to 11.95°. A comparison with parameters obtained through consolidated quick direct shear tests on the saturated loess shows that the increments of cohesion and the internal friction angle from in-situ large scale shear tests of the saturated loess masses were about 4kPa and 3° respectively.