Suppression (i.e., inhibiting one's emotional expression) has typically been associated with social and physiological costs. However, recent theorizing calls into question the inevitability of these costs. The present study takes a more nuanced approach and examines the social and physiological correlates of spontaneous (i.e., uninstructed) suppression when considering two potentially critical factors: the valence of the suppressed emotions (i.e., negative vs. positive) and the valence of the emotional context in which emotions are suppressed (i.e., negative conversation vs. positive conversation). Specifically, dating couples (N = 196 couples) completed both a negatively-valenced and a positively-valenced conversation in the laboratory while their autonomic-physiological responses were recorded. After each conversation, participants rated 1) the extent to which they had suppressed their negative and positive emotions, 2) the quality of the conversation, and 3) how connected they felt with their partner. We used Actor-Partner Interdependence Models to estimate actor effects (e.g., association of one's own suppression and one's own connectedness) and partner effects (e.g., association of one's partner's suppression and one's own connectedness). Suppression was associated with lower conversation quality and connectedness for the actors but largely not for the partners, regardless of the valence of the suppressed emotions and of the context, even when adjusting for felt emotion. Additionally, suppression was consistently not associated with physiological responses of actors or partners. Together, these findings suggest that, during emotional conversations with one's romantic partner, spontaneous (unlike instructed) suppression is associated with social but not physiological costs for the self but not one's partner.