This study investigated heart rate variability (HRV) in healthy volunteers that were selected for extreme scores of trait anxiety (TA), during two opposite psychophysiological conditions of mental stress, and relaxation induced by autogenic training. R-R intervals, HF and LF powers, and LF/HF ratios were derived from short-term electrocardiographic recordings made during mental stress and relaxation by autogenic training, with respiratory rate and skin conductance being controlled for in all the analyses. The main finding was that high TA was associated with reduced R-R intervals and HF power across conditions. In comparison to mental stress, autogenic training increased HRV and facilitated the vagal control of the heart. There were no significant effects of TA or the psychophysiological conditions on LF power, or LF/HF ratio. These results support the view that TA, which is an important risk factor for anxiety disorders and predictor of cardiovascular morbidity and mortality, is associated with autonomic dysfunction that seems likely to play a pathogenetic role in the long term.