A new measure of association for ordinal variables is proposed. The new measure of association, named the empirical polychoric correlation coefficient, builds upon the theoretical framework of the polychoric correlation coefficient, but relaxes its fundamental assumption so that an underlying continuous joint distribution is only assumed to exist, not to be of any specific distributional family. The empirical polychoriccorrelation has good properties in terms of statistical robustness and asymptotics, and is easy to compute by hand. Moreover, a simulation study indicates that the new measure of association is more stable, in terms of standard deviation, than conventional polychoric correlation coefficients.