Directionality of ground motions is an important aspect to be considered in seismic design and is incorporated in the Pacific Earthquake Engineering Research (PEER) center’s Next Generation Attenuation Ground Motions (NGA) projects and ASCE7-10. However, the NGA models and ASCE7-10 are limited to ground motion response spectra. Many empirical relationships for strong ground motion durations have been developed, but few of these relationships have considered the directionality of strong ground motion durations. In this study, the variability of strong ground motion durations for all orientations is examined using a total of 3523 horizontal motion sets from the PEER ground motion database. As a measure of strong motion duration, significant durations are considered, which are defined as the time interval between 5% to 75% or 95% of Arias intensity. Also using the non-linear mixed-effects regression method, the empirical relationships relating orientation-independent strong motion durations to earthquake magnitude, source-to-site distance, and local site conditions are developed. The orientation-independent durations are defined as the median (50th percentile) and maximum (100th percentile) durations of a horizontal motion rotated over all orientations. In addition, the strong motion durations in the directions associated with the median and maximum spectral response over all orientations, respectively used in the NGA-West2 and ASCE7-10, are computed for the horizontal motions from the PEER ground motion database. These orientation-consistent durations are then compared with the predicted durations by the empirical relationships developed herein. This comparison is to show the appropriateness/consistency of using the empirical strong motion duration relationships proposed herein in conjunction with the median or maximum response spectra in developing seismic design ground motions.