The Reflective Functioning Scale (RFS) was developed to assess individual differences in the ability to mentalize attachment relationships. The RFS assesses mentalization from transcripts of the Adult Attachment Interview (AAI). A global score is given by trained coders on an 11-point scale ranging from antireflective to exceptionally reflective. Coding procedures rely on a distinction of demand and permit questions during the AAI. Demand questions directly probe for reflective functioning (RF), whereas permit questions do not. Coding focuses on detecting qualitative markers of RF and qualitative markers of absent RF, respectively. Despite its relevant empirical contributions in clinical research, several psychometric properties of the RFS are still unclear. In this article, we present data on the reliability and internal structure of the RFS based on a combined sample of 196 subjects. We were able to show that (a) the global score can be assessed with good interrater reliability, is relatively stable across time, and is significantly reduced in persons with mental disorders; (b) demand questions are based on a single latent factor; (c) demand questions do not differ in terms of difficulty; (d) all demand questions but 1 are incrementally predictive of the global score; (e) 5 permit questions contribute to the global score over and above demand questions; and (f) the number of qualitative markers of RF is also predictive of the global score. Our results have important conceptual and methodological implications for future studies using the RFS. (Contains 4 tables and 1 figure.)