Forward directivity effects are known to cause pulselike ground motions at near-fault sites. We propose a comprehensive framework to incorporate the effects of near-fault pulselike ground motions in probabilistic seismic hazard analysis (PSHA) computations. Also proposed is a new method to classify ground motions as pulselike or non-pulselike by rotating the ground motion and identifying pulses in all orientations. We have used this method to identify 179 recordings in the Next Generation Attenuation (NGA) database (Chiou et al., 2008), where a pulselike ground motion is observed in at least one orientation. Information from these 179 recordings is used to fit several data-constrained models for predicting the probability of a pulselike ground motion occurring at a site, the orientations in which they are expected relative to the strike of the fault, the period of the pulselike feature, and the response spectrum amplification due to the presence of a pulselike feature in the ground motion. An algorithm describing how to use these new models in a modified PSHA computation is provided. The proposed framework is modular, which will allow for modification of one or more models as more knowledge is obtained in the future without changing other models or the overall framework. Finally, the new framework is compared with existing methods to account for similar effects in PSHA computation. Example applications are included to illustrate the use of the proposed framework, and implications for selection of ground motions for analysis of structures at near-fault sites are discussed.