The observation of a stressed individual can trigger a stress response in a passive observer. Little is known about the mechanisms of this so-termed empathic stress, including the observer's empathic involvement with the stressful situation. In 108 opposite-sex stranger dyads, we expected to increase the observer's empathic involvement with a stressed target performing a standardized laboratory stressor (Trier Social Stress Test, TSST; Kirschbaum et al., 1993) by exposing observers themselves to the TSST one week earlier. Conversely, we intended to decrease empathic involvement by granting observers a powerful position over the targets (by asking them to evaluate the targets' TSST performance and allegedly decide on their financial compensation). A control group without any manipulation was also included. In the preregistered data analysis, two types of empathic stress were investigated: vicarious stress, which evolves irrespective of the target's stress response, and stress resonance, which is proportional to the target's stress response. Irrespective of manipulation, observers exhibited vicarious stress in subjective and high-frequency heart rate variability (HF-HRV), and synchronized with the targets' stress reactivity in cortisol release. Prior TSST experience unexpectedly decreased observers' self-reported empathy and vicarious cortisol stress reactivity. The power manipulation, conversely, led to stronger observer vicarious stress in overall heart rate and HF-HRV reactivity. Based on Wondra and Ellsworth's (2015) appraisal theory, we propose that, due to their prior stressor exposure, observers habituated to said stressor, and consequently changed their evaluation of the target's stressful situation. In contrast, observers in the powerful position may have felt responsible for the targets, triggering a stronger vicarious stressful experience.