Site amplification studies and building code provisions recognize that soil liquefaction can alter the characteristics of ground shaking at a site. However, guidance as to how the amplitudes of spectral accelerations are modified is lacking. To address this issue, a two-part study is undertaken. In the first part an empirical study of ground motions recorded at liquefaction sites is undertaken. Available recorded ground motions from shallow crustal earthquakes at sites that exhibited evidence of liquefaction are compiled. Analysis of spectral acceleration residuals of the recorded ground motions computed relative to Next Generation Attenuation (NGA) estimates reveal positive bias at longer periods, slight negative bias at intermediate periods, and slight positive bias at short periods. Trends with Vs30, NGA-estimated peak ground acceleration (PGA), and moment magnitude are also observed. A model is developed that removes the initially observed residual bias and reduces uncertainty. The proposed model can be used to adjust NGA-estimated acceleration response spectra to account for the effects of liquefaction on ground shaking. In the second part of this study a series of parametric 1-D site response analyses were performed to provide a much larger synthetic dataset and to study geotechnical parameter that are typically unavailable in the empirical data. An existing constitutive model was rigorously calibrated against a widely-used semi-empirical liquefaction triggering method. The calibrated model was used to perform 2988 site response analysis pairs : one with porewater pressure generation and the other without. The resulting surface spectra are compared in a way that is analogous to the empirical study. Liquefaction amplification factors from the site response analysis results exhibit similar trends with period compared to the empirical data, but their amplitudes are systematically lower. The differences might be attributable to the inability of the 1-D site response analyses to capture 2-D and 3-D effects such as surface waves and basin effects, or possibly shortcomings the in the 1-D model's ability to faithfully represent all salient aspects of 1-D wave propagation under liquefaction conditions. Factors that appear to influence liquefaction amplification the most include whether the input motion is pulse-like or not and the amplitude of ground shaking