The seismic safety of dam structures is often evaluated using time history analyses conducted with a limited number of ground motions. The selection and scaling of the ground motions is usually the most effective factor determining the results of the safety assessment. The inherent variability in the ground motion as well as the difficulty of conducting the analyses for a large number of ground motions renders the selection as the most important factor in the analysis results. The guidelines for the nonlinear transient analyses of buildings, such as the one presented in ASCE/SEI-7-10, are well studied. For dams, however, it is not clear how the selection and scaling of the accelerograms should be conducted with the goals of a) consistency b) reliability and the c) practicality of the analyses. In this context, consistency implies obtaining consistent results for the same problem, reliability implies a reduction in the variability in the results while practicality implies the completion of the process with lesser effort. The selection and scaling of the ground motions for use in the nonlinear seismic analysis of the concrete gravity dams was investigated in this study with the aforementioned goals focused on the efficient prediction of the seismic demands on these structures. Three different concrete gravity dam monoliths were selected for this purpose, using 15 selected ground motions for the appropriate local site conditions. The material nonlinearity, dam-reservoir interaction and vertical component of ground motions were considered in the analyses. The engineering demand parameters were selected as the crest displacement, the maximum crest acceleration and the crack extent, a direct indicator of the damage on the monoliths. Nine different scaling methods were investigated. The effectiveness in the prediction of the mean demand and the corresponding dispersion levels were compared. The required number of motions to conduct effective analyses was determined. (c) 2017 The Authors. Published by Elsevier Ltd.