This study examines the possibility that social support operates as a moderator of cardiovascular reactivity in women.Two models by which social support may operate were examined: the direct effects and buffering models. Twenty-six subjects were exposed to four conditions while playing a video game: two levels of stress (low, high) and two levels of social support (alone, together). Blood pressure and heart rate were monitored continuously. Ratings of stress were obtained for each condition. Results indicated that the support manipulation produced significant main effects for diastolic blood pressure and stress ratings, with lower diastolic blood pressure and ratings observed in the "together" condition, and that the interaction between support and stress produced lower reactivity for the cardiovascular measures in the high stress (but not the low stress) condition. No interaction was found for the stress ratings. We conclude that the results provide support for both the buffering and direct effects models. Implications concerning the (within-subjects) design of the study and the stress ratings are discussed.