Although a considerable amount of studies have been carried out for many years related to response of building of long and huge configuration in plan, none of those studies gives theoretical explanation clearly on actual dynamic behaviors. In this paper, authors describe methdology of modelling of this type of building using an actual structure and present theoretical explanation on actual response behaviors they acquired through earthquake obsevation for past years. The modelling for computer simulation was based upon the following conditions : 1) Beams and columns are treated considering bending and shearing deformation. 2) Slabs and shear walls are replaced by macro type of finite elements formed into frameworks. 3) Soil is treated as half space and the values of static soil springs is corrected for layered soil condition. 4) Dynamic soil-strcture interaction effect is approximately taken into consideration using additional mass. 5) Influence between adjacent foundations and from adjacent buildings are assumed to be negligible. First total structure is subdivided into 10 structural blocks and stiffness matrices of them are to be calculated. As for the blocks representing tube structures enclosing staircases inside, the stiffness matrices can be obtained by 3-D frame analysis based upon assumption of rigid slab. On the other hand, as for the blocks representing main frameworks of the building, flexibility of slab is taken into consideration when analysed in order to obtain the stiffness matrices. Next objective restoring matrix concerning whole structure can be composed by means of connecting stiffness matrices of subdivided structural blocks by those of slabs. Here as for the soil springs sustainig superstructure, they are obtained by Tajimi's method modified by authors for extended application. In this method static soil springs are calculated, using fundamental functions related to displacement distribution in half space imposed a unit force on its surface. And the results are corrected approximately for layered soil condition based upon assumption that stress distribution in layerd strata is the same as one in half space. According to an earthquake observation being conducted by authors in Hatoyama Campus of Tokyo Denki University since 1983, ground motins can be thought to be spatially variant although some of them can be approximately regard as uniform ground shaking. Based upon the above observation and analitical results they have already published (Ref. 8 and 16), two types of exciting ground motions were adopted for the computer simulation. Namely in case of Eastern Offsore Ibaragi Pref. Earthquake of Feb. 12, 1986, traveling wave was assumed in time domain where the maximum acceleration of the motion appeared. Subsequently random phase variation of high frequency components due to difference between observation sites and unpredictable factors was assumed in the rest time domain. On the other hand in case of Kanto-Chubu Border Earthquake of Feb. 14, 1984 and Kujukuri Cost Boso Pen. of Dec. 17, 1987, uniform ground excitation was assumed. As a whole, theoretical results obtained through simulation were in good accordance with actual response behaviors. Difference between calculated behavior and actual one is about 8 % at the top of the structure and 22 % at the base on the average when observed amplitudes give maximum acclerations. According to comparison of simulation with observation, complicated base motion in transversal direction was able to be explained as response of foundation including sway under uniform excitation. Significant characterestics such as higher mode excitation with remarkable slab deformation caused by time delay which exists propagating ground motion and its rapid amplitude degradation due to a sort of kinematic interaction effect of spatially variant ground motion were also explaned.