This report gives a synopsis of the various tools and techniques used in selecting earthquake ground motion parameters for large dams. It presents 18 charts giving newly developed relations for acceleration, velocity, and duration versus site earthquake intensity for near- and far-field hard and soft sites and earthquakes having magnitudes above and below 7. The material for this report is based on procedures developed at the Waterways Experiment Station. Although these procedures are suggested primarily for large dams, they may also be applicable for other facilities. Because no standard procedure exists for selecting earthquake motions in engineering design of large dams, a number of precautions are presented to guide users. The selection of earthquake motions is dependent on which one of two types of engineering analyses are performed. A pseudostatic analysis uses a coefficient usually obtained from an appropriate contour map; whereas, a dynamic analysis uses either accelerograms assigned to a site or specified respunse spectra. Each type of analysis requires significantly different input motions. All selections of design motions must allow for the lack of representative strong motion records, especially near-field motions from earthquakes of magnitude 7 and greater, as well as an enormous spread in the available data. Limited data must be projected and its spread bracketed in order to fill in the gaps and to assure that there will be no surprises. Because each site may have differing special characteristics in its geology, seismic history, attenuation, recurrence, interpreted maximum events, etc., as integrated approach gives best results. Each part of the site investigation requires a number of decisions. In some cases, the decision to use a 'least ork' approach may be suitable, simply assuming the worst of several possibilities and testing for it. Because there are no standard procedures to follow, multiple approaches are useful. For example, peak motions at a site may be obtained from several methods that involve magnitude of earthquake, distance from source, and corresponding motions; or, alternately, peak motions may be assigned from other correlations based on earthquake intensity. Various interpretations exist to account for duration, recurrence, effects of site conditions, etc. Comparison of the various interpretations can be very useful. Probabilities can be assigned; however, they can present very serious problems unless appropriate care is taken when data are extrapolated beyond their data base. In making deterministic judgments, probabilistic data can provide useful guidance in estimating the uncertainties of the decision. The selection of a design ground motion for large dams is based in the end on subjective judgments which should depend, to an important extent, on the consequences of failure. Usually, use of a design value of ground motion representing a mean plus one standard deviation of possible variation in the mean of the data puts one in a conservative position. If failure presents no hazard to life, lower values of design ground motion may be justified, providing there are cost benefits and the risk is acceptable to the owner. Where a large hazard to life exists (i.e., a dam above an urbanized area) one may wish to use values of design ground motion that approximate the very worst case. The selection of a design ground motion must be appropriate for its particular set of circumstances.