Theoretical perspectives on mating differentially emphasize whether (and why) romantic partner selection and maintenance processes derive from stable features of individuals (e.g., mate value, mate preferences, relationship aptitude) and their environments (e.g., social homogamy) rather than adventitious, dyad-specific, or unpredictable factors. The current article advances our understanding of this issue by assessing how peoples actual romantic partners vary on constructs commonly assessed in evolutionary psychology (Study 1), sociology (Study 2), and close relationships (Study 3). Specifically, we calculated the extent to which the past and present partners of a focal person (i.e., the person who dated all of the partners) cluster on various measures. Study 1 investigated consistency in the observable qualities of the romantic partners, revealing substantial evidence for clustering on coder-rated attributes like attractiveness and masculinity. Study 2 examined qualities self-reported by romantic partners themselves in a demographically diverse sample and found modest evidence for clustering on attributes such as IQ and educational aspirations; however, clustering in this study was largely due to demographic stratification. Study 3 explored target-specific ratings by partners about the focal person and found little evidence for clustering: The ability to elicit high romantic desirability/sexual satisfaction ratings from partners was not a stable individual difference. The variables that affect mating may differ considerably in the extent to which they serve as stable versus unpredictable factors; thus, the fields of evolutionary psychology, sociology, and close relationships may reveal distinct depictions of mating because the constructs and assessment strategies in each differ along this underappreciated dimension. (PsycINFO Database Record