A rupture model for the 1989 Loma Prieta earthquake obtained from inversion of teleseismic body waves is used to calculate the mainshock, near-source, strong ground motions. The effectiveness of both synthetic and empirical (aftershock) Green's functions is evaluated in terms of velocity response spectra over the period range from 0.1 to 10 sec. A strictly deterministic approach is used in the simulation of near-source ground motion with both the synthetic and aftershock Green's functions. A deterministic method has advantages in that it is simple to apply and there is no stochastic element to introduce variability in the solution. Response spectral values are very sensitive to the choice of the mainshock source-time function. Using a 1.0-sec triangular time function obtained from the teleseismic study gives good estimates of response spectra from 1- to 2-sec period for aftershock Green's functions and from 1- to 10-sec period for synthetic Green's functions. The limited long-period response of the aftershock summations is due to the restricted bandwidth of the recording sensor. Using a 0.5-sec Kostrov time function extends the bandwidth of validity of the aftershock summations to 0.1- to 2-sec period and the synthetic Green's function summations to 0.3- to 10-sec period. The limited short-period bandwidth of the synthetic Green's function summations is due to the lower-frequency content (5 Hz) and absence of accurate high-frequency propagation effects in the synthetic seismograms. Although the teleseismic study could not distinguish between a 1.0-sec triangular time function and a 0.5-sec Kostrov time function, we feel justified in using the Kostrov function because it lies within our broader knowledge of earthquake source-time functions.