The social relations model (SRM) is a useful tool for measuring relationship effects, defined as the unique perceptions or behaviors of 2 people. The sources of variance in SRM studies are persons (actors and partners), groups, and items; the relationship effect is defined as the actor–partner interaction. By removing variance because of persons and groups, a measure of a “pure” relationship effect is obtained. In this article, generalizability theory (G Theory) is applied to estimate the reliability of SRM components from round-robin data structures. Using G Theory, reliability formulas for actor, partner, group, and relationship are developed and interpretations for the reliability estimates are provided. The authors also discuss how these formulas can be used in both planning and interpreting results from relationship research.