Most methods for assessment of in situ seismic soil liquefaction potential require evaluation of the earthquake-induced cyclic shear stress ratio (CSR). Estimates of the in situ CSR can be developed directly, using dynamic response analyses, but it is common in 'simplified' analysis methods to develop estimates of the in situ CSR using empirical relationships. Unfortunately, the most widely used existing empirical relationships are based on limited response analyses and do not take full advantage of current knowledge of factors affecting this response problem. As a result, they are both biased and unnecessarily imprecise. This paper presents the results of a relatively comprehensive suite of site response studies (2153 site response analyses), performed using carefully selected suites of site conditions and input time histories, to provide an improved basis for development of estimates of in situ CSR using the r(d)-approach. The resulting empirical correlations, developed using the Bayesian updating method, provide a much improved basis for simplified empirical evaluation of CSR as a function of (1) depth; (2) earthquake magnitude; (3) intensity of shaking; and (4) site stiffness. (C) 2003 Elsevier Ltd. All rights reserved.