This paper introduces a new framework for measuring the extent of positive assortative mating (PAM) in the marriage market by relaxing the standard assumption of dichotomous sorting levels. Conventional PAM measures treat marriage sorting as a binary outcome—either perfectly matched or not—thereby failing to capture degrees of similarity between partner types. We propose a continuous measure of sorting based on trait similarity, where individuals are hypothesized to select mates according to the relative closeness of traits, which influences marital payoffs. Trait similarity is quantified using multidimensional attribute data and incorporated into a similarity-weighted matching matrix. We adapt conventional PAM indices—including the normalized trace, aggregate likelihood ratio, and perfect-random normalization—to this similarity-weighted framework. Applying our method to U.S. data on occupational, religious, and educational matching, we uncover patterns in assortative mating that are obscured under traditional approaches.