It is widely recognized that long-period surface waves generated by conversion of body waves at the boundaries of deep sedimentary basins make an important contribution to strong ground motion. The factors controlling the amplitude of such motion, however, are not widely understood. A study of pseudovelocity response spectra of strong-motion records from the Los Angeles Basin shows that late-arriving surface waves with group velocities of about 1 km/sec dominate the ground motion for periods of 3 sec and longer. The rate of amplitude decay for these waves is less than for the body waves and depends significantly on period, with smaller decay for longer periods. The amplitude can be modeled by the equation log y = f(M, RE) + c + bRB where y is the pseudovelocity response, f(M, RE) is an attenuation relation based on a general strong-motion data set, M is moment magnitude, RE is the distance from the source to the edge of the basin, RB is the distance from the edge of the basin to the recording site, and b and c are parameters fit to the data. The equation gives values larger by as much as a factor of 3 than given by the attenuation relationships based on general strong-motion data sets for the same source-site distance. It is clear that surface waves need to be taken into account in the design of long-period structures in deep sedimentary basins. The ground-motion levels specified by the earthquake provisions of current building codes, in California at least, accommodate the long-period ground motions from basin-edge-generated surface waves for periods of 5 sec and less and earthquakes with moment magnitudes of 7.5 or less located more than 20 km outside the basin. There may be problems at longer periods and for earthquakes located closer to the basin edge. The results of this study suggest that anelastic attenuation may need to be included in attempts to model long-period motion in deep sedimentary basins. To obtain better data on surface waves in the future, operators of strong-motion networks should take special care for the faithful recording of the long-period components of ground motion. It will also be necessary to insure that at least some selected recorders, once triggered, continue to operate for a time sufficient for the surface waves to traverse the basin. With velocities of about 1 km/sec, that time will be as long as 100 sec for a basin the size of the Los Angeles Basin.