Using the fault slip determined by Sieh (1978a), the great 1857 California earthquake is modeled as a propagating fault buried in a realistic medium consisting of a two-layer crust overlying the mantle. The computation is carried out by the discrete wavenumber representation method (Bouchon, 1979). The resulting motion is almost purely horizontal throughout central and southern California, and, at most of the locations, the direction of severe ground shaking is oriented NE-SW. In the Los Angeles area, the peak displacement is about 30 cm, and the duration of severe ground motion is about 75 sec. Since our model does not include the effect of incoherent rupture propagation, our estimate should be considered as a lower bound. The simulated seismic motions are presented in the form of three-dimensional snapshots as well as in the usual form of time series at selected locations. The resultant near-field seismograms show a great diversity in the characteristic of motion as pointed out for actual near-field accelerograms by Hudson (1976). Our simulated motions appear to be in harmony with the description of the effect of the 1857 earthquake by contemporary accounts compiled by Agnew and Sieh (1978).