Wang et al. (Eos, 18 November 2003, pp. 501, 506, 508) provide an interesting introduction to the heart of probabilistic seismic hazard analysis (PSHA). However, the criticisms of PSHA recited in the second half of the article do not tell the whole story.
In particular, the limitations listed by Wang et al. apply equally to deterministic seismic hazard analysis, the only real alternative to PSHA. Deterministic and probabilistic seismic hazard analyses employ earthquake hazard information in a somewhat different manner to estimate earthquake ground motion (shaking) hazard at a site [Reiter, 1990]. Both approaches model in the same manner the distribution of earthquake sources and the decay (attenuation) of earthquake ground motion with increasing distance from an earthquake. In a deterministic analysis, a particular damaging earthquake is chosen from among the possible damaging earthquakes that can affect a site (usually the earthquake expected to be the most damaging). Then the distribution of probable ground motions, represented as a median ground motion and its lognormal standard deviation, from one or several attenuation relations is used to select a ground motion hazard at the site.