Autonomic responses were measured while 45 adult women performed a standard experimental stress task in the laboratory with only the experimenter present and 2 weeks later at home in the presence of a female friend, pet dog, or neither. Results demonstrated that autonomic reactivity was moderated by the presence of a companion, the nature of whom was critical to the size and direction of the effect. Ss in the friend condition exhibited higher physiological reactivity and poorer performance than subjects in the control and pet conditions. Ss in the pet condition showed less physiological reactivity during stressful tasks than Ss in the other conditions. The results are interpreted in terms of the degree to which friends and pets are perceived as evaluative during stressful task performance. Physiological reactivity was consistent across the laboratory and field settings.