To investigate earthquake responses of structures with basements affected by soil deposits, centrifuge tests were performed using an in-flight earthquake simulator. The test specimen was composed of a single-degree-of-freedom structure model, a basement and sub-soil deposits in a centrifuge container. The test parameters were the dynamic period of the structure model, boundary conditions of the basement, existence of soil deposits, centrifugal acceleration level, and type and level of input earthquake accelerations. When soil deposits did not exist, the earthquake responses of the structures with fixed basement were significantly greater than those of the structure without basement. Also, the earthquake responses of the structures with the fixed basement surrounded by soil deposits were amplified, but the amplifications were smaller than those of the structures without basement. The earthquake responses of the structures with the half-embedded basement in the soil deposits were greater than those estimated by the fixed base model using the measured free-field ground motion. The test showed that the basement and the soil deposit should be simultaneously considered in the numerical analysis model, and the stiffness of the half-embedded was not effective.