Due to incomplete and doubtful seismological information and internal contradiction in stan-dard design-analytical procedures the most efficient technique for estimating the actual seismic resis-tance of buildings consists in using experimental methods. One of them is the so-called resonance method, which is realized through the use of high-power vibration machine installed on the building or the ground around the building. The methods of physical modeling suggest:

1) to develop the model of experimental object;

2) to provide connection with the environment (to meet boundary conditions);

3) to model external effect.

It has been possible to provide the first two conditions in vibration testing but the third condition cannot be realized, in principle. A real earthquake cannot be modeled with the help of vibration machine because uniform harmonic loading differs sharply from the seismic effect in spectral composition. In this case there is a need to use a special procedure. It is based on transition from physical experiment to the methods of mathematical modeling. In this case instrumental or syn-thesized accelerograms of real earthquakes scaled as required, are taken as models of external effect. Calculated dynamic model of a building and its physical and mechanical characteristics recorded at high level of dynamic effect, commensurable with real seismic loads, are taken in accordance with the results of actual vibration testing. The experiments on the performance of vibration testing in seismic areas of the Irkutsk region are generalized in the paper. Assessments of actual seismic resistance of buildings of various types are given.