Human-robot interaction (HRI) spans many and diverse contexts, each varying in the preferred degree of social connection fostered therein. The design of robotic systems and selection of which robotic forms to use will thus benefit from an understanding of the factors by which social connection is enhanced or diminished. We address this fundamental problem through an exploratory study of role-taking, a core construct from structural social psychology. Through a laboratory experiment (N=86) in which participants interact with either human, humanoid robot, or non-humanoid robot partners, we examine variation in perceived role-taking accuracy, affection for the partner, and desire for continued interaction. Findings show that participants evaluate human interaction partners as more accurate role-takers than robots, but with variation between robotic types (humanoid vs non-humanoid) in relation to personality versus emotion. Sense of affection and desire for continued interaction are predicted by interpersonal factors rather than partner type or robotic form. We discuss possible reasons behind these preliminary findings and draw on them to formulate subsequent research questions. More broadly, we call for continued study at the intersection of HRI and structural social psychology, advancing both fields through the pairing.