Probabilistic seismic hazard analysis (PSHA) provides the conceptual framework for estimating the likelihood that something of concern related to earthquake shaking will occur over a specified time period. Based on more than thirty years of research and development ( e.g. , Cornell, 1968; Algermissen et al. , 1982; SSHAC, 1997), PSHA has become a standard tool for combining information on earthquake occurrence, seismic radiation, and shaking response to produce hazard estimates, including the U.S. Geological Survey's national seismic hazard maps (Frankel et al. , 1996, 1997). PSHA methods, while now mature, continue to evolve as scientists improve the characterization of earthquake hazards and engineers develop new measures of seismic shaking for performance-based design. The Southern California Earthquake Center (SCEC), in collaboration with USGS, the California Geological Survey (CGS), and other partners, has undertaken a series of studies aimed at improving the regional application of PSHA methods. Phase I examined the implications of the 1992 Landers earthquake sequence for regional seismic hazards (WGCEP, 1992), Phase II developed a probabilistic earthquake forecast model (WGCEP, 1995), and Phase III assessed the wave-propagation and site effects that give rise to local variations in seismic shaking (see Field et al. [2000] for an overview). The Phase III study found that accounting for some site attributes ( i.e. , the 30-meter shear-wave velocity and basin depth) can lead to significant improvements in PSHA. It was also found that making such corrections does not significantly reduce the prediction uncertainty associated with empirical ground-motion relations. In fact, 3D waveform modeling presented in the same report (Olsen, 2000) implied that this residual uncertainty represents an intrinsic variability caused by complex propagation effects that are unique to each earthquake-rupture/site combination. The Phase III study therefore concluded that significant improvements in PSHA will require replacing the standard empirical-regression …