CONTEXT: Relaxation interventions are frequently used to promote symptom relief in palliative care settings, but little is known about the underlying mechanisms. OBJECTIVES: The present analysis aimed at examining the psychophysiological pathways of terminally ill patients\' cardiovascular response to a live music therapy vs. prerecorded mindfulness exercise. METHODS: Eighty-four patients of a palliative care unit were randomly assigned to either of the two interventions. Multilevel modeling was used to analyze trajectories of physiological change. Vagally mediated heart rate variability (VM-HRV) and blood volume pulse amplitude (BVP-A) served as indices of autonomic nervous system response. Participants\' gender, age, baseline scores, self-rated pain, and assignment to treatment were entered to the models as predictors. RESULTS: Both VM-HRV and BVP-A showed significant linear and quadratic trends over time, as well as substantial heterogeneity among individuals\' trajectories. Baseline scores, pain, and treatment significantly accounted for random variation in VM-HRV intercepts. BVP-A levels were significantly higher in women than in men. Moreover, assignment to treatment significantly accounted for differences in the linear slopes of peripheral blood flow. CONCLUSION: Higher levels of VM-HRV in the music therapy group highlight the importance of a therapeutic relationship for the effectiveness of relaxation interventions in end-of-life care settings. Music therapy caused significantly stronger reductions of vascular sympathetic tone and, therefore, may be indicated in the treatment of pain and stress-related symptoms in palliative care. Initial self-ratings of pain moderated patients\' physiological response and need to be taken into account in clinical practice and future theory building.