Reactivity to psychological stressors has been hypothesized to be related to the development of cardiovascular disease. Because mental arithmetic (MA) has been shown to produce significant increases in heart rate and blood pressure, it is one of the most commonly utilized laboratory psychological stressors. However, the use of MA to assess hemodynamic reactivity raises two issues: 1) increases in heart rate can be produced by vagal withdrawal, sympathetic activation, or a combination of the two, and these mechanisms may differ in their pathogenic implications; and 2) in most MA studies, subjects are instructed to perform the task aloud, thus raising the possibility that speaking may interfere with respiratory patterns which in turn can influence hemodynamic outcomes. To address these two issues, we studied heart rate responses of 10 subjects to 2 different versions of MA and a control condition in which vocalization of answers was manipulated. Heart rate (HR), heart rate variability (HRV) in the low frequency (LB) and respiratory frequency (RSA) bands, and respiratory rate were measured. Results indicate that, although the two task conditions produced similar heart rate increases, RSA decreased only in the nonspeaking condition. Overall, the findings suggest that HR changes during MA are attributable to vagal withdrawal but that vocalization of answers during the task interferes with analysis of HRV, thus obscuring the mechanisms responsible for these changes.