This paper develops a new theoretical and empirical approach to studying historical trends in social inequalities in mortality. The author describes how within group inequality in mortality (e.g. IQRo) can be used to identify the timing and pace of epidemiological transitions, and she relates differences between population groups in IQRo to theory on social inequalities in mortality. Using data on mortality for U.S. white and non-white men and women over the period 1900-2002 and decomposing the patterns of change in IQRo, she finds that mortality declines have been later, slower, and have involved more irregular age-composition among non-whites than whites. As a result, racial differences in IQRo have dynamically diverged and converged while there has been a century of convergence in eo. These findings show that IQRo provides information about mortality differences that are concealed by eo and they provide insight into the theoretical trends in mortality associated with economic development, diffusion of innovations, institutional changes, and social stratification.