We investigate the use of site factors derived from theoretical one- dimensional ground-response analyses as a means by which to estimate seismic site effects for earthquake ground-motion prediction. We review a set of theoretical site factors derived from equivalent-linear analyses for specific geologic categories in the Los Angeles area and the San Francisco Bay area. The theoretical site factors are a function of the peak horizontal acceleration for the reference rock site condition and the depth to firm material (taken as depth to the 1 km/sec shear-wave isosurface). We apply the site factors to predict the response spectral acceleration of ground motions at recording sites in a manner that mimics how they would typically be used in engineering practice: empirical ground-motion prediction equations are used to estimate a reference (rock) spectrum that is then modified by using the theoretical site factors. Analysis of residuals between the data and model predictions indicates that the theoretical site factors are able to capture the average effects of sediment nonlinearity. The analyses also show that short-period spectral accelerations are underpredicted and that long-period spectral accelerations are overpredicted. The underprediction at short periods may result from overestimation of a soft-to-firm reference rock correction factor that is needed to apply the site factors. The overprediction bias at long periods varies with depth, suggesting that the depth term in the site factors may not fully account for the effect of basin geometry at long periods.