Stimulation of the round window (RW) has gained increasing clinical importance. Clinical, as well as human temporal bone and in-vivo animal studies show considerable variability. The influence of RW stimulation on the cochlea remains unclear. We designed a human temporal-bone study with controlled direct mechanical stimulation of the RW membrane to identify conditions for successful RW stimulation. Eight human temporal bones were stimulated on the RW by piezoelectric stack actuators with cylindrical aluminium rods of diameter 0.5 mm and with either flat or 30° inclined top surface. Using a dedicated two-stage positioning protocol for the actuator, we achieved highly reproducible measurements of the stimulus vibration at the RW and of the resultant vibration of the stapes footplate. The reverse transmission, characterized by the displacement ratio of the stapes-footplate relative to the actuator tip on the RW membrane, yielded an average displacement ratio of 0.089 up to 1.2 kHz when the actuator was coupled without angular misalignment to the RW membrane. The results suggest that 90-μm pretension of the RW membrane is essential for optimum and reproducible RW stimulation. The displacements are shown to be roughly consistent with the equal-volume displacement hypothesis under specific assumptions about the displacement mode of the RW membrane. It is further suggested that the large inter-patient variability in the effectiveness of RW stimulation might be due primarily to the success of coupling, rather than to the variability of functionally relevant anatomical parameters.