This article examines how the empirical Green's function seismogram synthesis method might perform as a tool for earthquake hazard analysis by “predicting” strong ground motion from the 18 October 1989 Loma Prieta earthquake. A suite of accelerograms are synthesized for five sites 40 to 80 km from the nearest fault rupture from 25 different source models that might have been considered prior to the occurrence of the Loma Prieta earthquake. Three-component, full wavetrain ground accelerations are synthesized over the frequency range of 0.5 to 25.0 Hz with a sparse distribution of source events for empirical Green's functions and fairly simple source models. The suite of source models includes a range of rupture histories and geometries, asperity models, and slip models that reflect the uncertainties both in predicting the actual rupture history and in knowledge of modeling rupture processes. The suite of synthesized accelerograms includes a good match to observed time series amplitudes, frequency content, energy distribution, and some waveforms and demonstrates that the character of actual time series can be captured with a range of simple source models. The average levels of peak acceleration, pseudo-velocity response, and Fourier amplitude spectra provide a good prediction to the observed values at four of the five sites. The log-normal standard deviations provide bounds on the predictions that are specific to the source and recording site.