In this study we used strong-motion data recorded from 1957 to 1995 to derive a mutually consistent set of near-source horizontal and vertical ground-motion (attenuation) relations for peak ground acceleration and 5%-damped pseudoacceleration response spectra. The database consisted of up to 960 uncorrected accelerograms from 49 earthquakes and 443 processed accelerograms from 36 earthquakes of MW 4.7-7.7. All of the events were from seismically and tectonically active, shallow crustal regions located throughout the world. Some major findings of the study are (1) reverse- and thrust-faulting events have systematically higher amplitudes at short periods, consistent with their higher dynamic stress drop; (2) very firm soil and soft rock sites have similar amplitudes, distinctively different from amplitudes on firm soil and firm rock sites; (3) the greatest differences in horizontal ground motion among the four site categories occur at long periods on firm rock sites, which have significantly lower amplitudes due to an absence of sediment amplification, and at short periods on firm soil sites, which have relatively low amplitudes at large magnitudes and short distances due to nonlinear site effects; (4) vertical ground motion exhibits similar behavior to horizontal motion for firm rock sites at long periods but has relatively higher short-period amplitudes at short distances on firm soil sites due to a lack of nonlinear site effects, less anelastic attenuation, and phase conversions within the upper sediments. We used a relationship similar to that of Abrahamson and Silva (1997) to model hanging-wall effects but found these effects to be important only for the firmer site categories. The ground-motion relations do not include recordings from the 1999 MW >7 earthquakes in Taiwan and Turkey because there is still no consensus among strong-motion seismologists as to why these events had such low ground motion. If these near-source amplitudes are later found to be atypical, their inclusion could lead to unconservative engineering estimates of ground motion. The study is intended to be a limited update of the ground-motion relations previously developed by us in 1994 and 1997, with the explicit purpose of providing engineers and seismologists with a mutually consistent set of near-source ground-motion relations to use in seismic hazard studies. The U.S. Geological Survey and the California Geological Survey have selected the updated relation as one of several that they are using in their 2002 revision of the U.S. and California seismic hazard maps. Being a limited update, the study does not explicitly address such topics as peak ground velocity, sediment depth, rupture directivity effects, or the use of the 30-m velocity or related National Earthquake Hazard Reduction Program site classes. These are topics of ongoing research and will be addressed in a future update.