Accurate seismic zonation for predicting hazards from earthquake shaking must take into account how local site conditions respond to incoming seismic waves as a function of frequency. Weak motion site response has been shown to be predictable from local geology, with greater amplification occurring at younger sedimentary sites, and the least at hard rock sites (Aki, 1993). However, for strong motion, nonlinear amplification can mean that site response must also be calibrated as a function of amplitude. Nonlinear amplification at sediment sites was inferred by Chin and Aki (1991) and Aki (1993) who report that for sedimentary stations within 50 km of the epicenter of the Loma Prieta earthquake, predicted accelerations based on weak motions significantly overestimated accelerations measured for strong motions (0.1 to 0.3g, Figure 3). As a result, Aki (1993) suggests that peak ground acceleration is site-independent in the epicentral region, i.e., that all sites are about equal.