According to past surveys of pile-supported RC buildings, severe building damage was not reported so often although recorded seismic motions around the buildings were considerably larger than their seismic design motion. An effect of soil-pile-structure interaction is one of most possible reasons for this earthquake-damage reduction. However, actual shaking records at those buildings have been rarely obtained. Therefore, it is difficult to demonstrate actual phenomena with actual input motion and building records, and past research works were usually performed by numerical analyses under reasonably assumed conditions.<br> In light of this, a dynamic experiment of a pile-supported RC structure was conducted using E-Defense shaking table to obtain series of data, which are expected to help understanding its failure process. The structure model was three-story RC frame supported by PC piles. The model was built in a RC container filled with dry sand, and the PC piles were embedded in the sand. The model was built in a scale of 1:2.5. Long spanned two steel beams were installed on footings of the RC structure to prevent the model from falling in case when significant damage of the piles induced considerable loss of their axial load capacity. An acceleration wave of input motion was evaluated based on an attenuation relationship of response spectra.<br> Prior to the shaking table test, a static loading test of a footing supported by four PC piles in dry sand was conducted to make a plan of the dynamic test model. From this static loading test of the footing, the strength and deformation capacity of the soil-pile interaction system were obtained. Dynamic response analyses were conducted using a frame model with a nonlinear sway spring assumed by this static test. Based on the results from the above analyses, an excitation plan of shaking table was scheduled from 10% to 300% or 400% of the designed acceleration motion.<br> The shaking table tests were started from 20% level showing linear response characteristics and finished at the level of 300% with fatal failure of the pile heads. In the case of 200% input, no sign of structural failure such as subsidence or tilting of the frame appeared. On the other hand, strains of PC steel bars in the PC piles exceeded yielding strain, and compression strain of their surface concrete also exceeded failure strain. This means a field investigation only on damage of superstructures after earthquake might miss a finding of pile damage.<br> In the case of 300% input, the strength deterioration and then the loss of axial load capacity of the piles were recorded at the pile head displacement expected by the above dynamic response analyses. Thus, it was shown that the preliminary analyses of the dynamic tests had given reasonable behaviors of the test model. In this input, the pile heads finally reached compression failure, and such damage corresponded to the actual failure. On the other hand, damage of the superstructure were slight. This result suggests the possibility that the damage of the superstructure was suppressed by nonlinear behaviors of the soil-pile system.<br> Through the shaking table tests, series of data were obtained from small input tests where both a structure and piles were undamaged and large input tests where pile heads were damaged like actual pile damage in past large earthquakes.