The stress-buffering hypothesis (Cohen & McKay, 1984) proposes that social support attenuates cardiovascular reactivity (CVR) in times of stress. Past research which tested this hypothesis has obtained inconsistent findings. A dual effect model of social support and stress (Teoh & Hilmert, 2015) suggests that these inconsistent findings could be due to different effects of social support on CVR that depend on how engaged participants are during a stressful task. Specifically, this model proposes that when people are not engaged, social support encourages, increasing CVR relative to no support; and when people are engaged, social support comforts, attenuating CVR relative to no support. This study examined the dual effect model by empirically manipulating social support and task engagement while monitoring participants’ cardiovascular readings. We randomly assigned the participants (N = 121, all women) to give a speech on either a more engaging or a less engaging topic while receiving social support or no support from two evaluative female audience members. Before and after the speech, the participants completed several questionnaires that included measures of perceived stress and task engagement. Our results showed that, consistent with our prediction, socially supported participants responded to the task with greater CVR than nonsupported participants in the less engaging condition, indicating a social encouragement effect of social support. However, when the speech topic was more engaging, there was no significant effect of social support on CVR. Our findings show that task engagement moderates the effects of social support on CVR. The health implication of a CVR-elevating effect of social support is relatively unexplored and suggests that increased CVR to stress may be associated with positive health in certain situations.