In our previous study (Ohori and Hisada, 2006), we simulated the strong-motion records of the mainshock (MJ 5.4) of the 2001 Hyogo-ken Hokubu earthquake, Japan, on the basis of the empirical Green’s tensor spatial derivative (EGTD) estimated from data of 11 aftershocks (MJ 3.5–4.7). The agreement between the observed and calculated waveforms at the closest station in source distance was satisfactory over a long duration, and the amplitude was well reproduced. To further evaluate the EGTD estimation, it is important to compare the results of the EGTD method with those of a similar approach, the empirical Green’s function (EGF) method, which has been used widely in strong-motion simulation. In this study, we compare the strong-motion simulation results obtained using the EGTD method with those obtained using the EGF method. In the EGF simulation, we calculate the mainshock synthesis from each aftershock by correcting for the difference in the radiation pattern coefficients between the two events. The relative errors of synthesized waveforms against observation data are evaluated for the S-wave main portion, S-wave coda portion, and full waveform. The simulation results obtained using the EGF method indicate that most aftershocks reproduce well the observation data, especially for the S-wave main portion, but overestimate the amplitude when the sources are located near the node of the SH-wave or SV-wave radiation pattern. It seems that the EGTD method smooths the dependency of the EGF method on the aftershock selection and provides stable and reliable results in contrast to the EGF method, not only for the S-wave portion but also for the S-wave coda portion.