The goal of this study is to simulate the displacement waveforms of 8 sec-period Love waves observed at Tokyo and Yokohama stations in the Kanto sedimentary basin during the Izu-hanto-toho-oki Earthquake of 29 June 1980 (M0 = 7 × 1025 dyne · cm). The surface wave Gaussian beam method is applied for this purpose. On the basis of the 3-D seismic velocity and density structure of the Kanto basin and assuming that the earth medium is laterally homogeneous outside the Kanto basin, waveforms of Love waves are synthesized. The synthesized seismograms underestimate the observed peak amplitudes at Yokohama station. This is primarily because the station is located in the direction of the nodal plane of the Love-wave radiation. As indicated by Yamanaka et al. (1992), a Quaternary basin exists in the Sagami Bay between the source location and the Kanto basin. We include the Sagami basin in our model by the approximation of a circular low-velocity region. Excellent agreement between observed and synthesized waveforms was achieved not only for amplitude but also for phase for the early parts of the wave trains at both stations. We conclude that the low velocity Quaternary basin in the Sagami Bay acts like a lens to focus surface wave energy resulting in high amplitudes. The later arriving waves, in particular the long duration observed at Tokyo station, however, cannot be adequately explained by this method. One possible reason for the failure of simulating the later phases is that this method disregards the secondary Love waves converted from S-waves and/or surface waves at a laterally discontinuous boundary. Although the surface wave Gaussian beam method cannot adequately predict the duration of observed seismograms, it provides us with a satisfactory prediction of amplitudes and phases for early arrivals in laterally slowly-varying media at drastically lower computation costs and less memory requirements than does other methods.