It is well known that although relationship external stressors can harm couples, dyadic coping behavior can buffer the negative effects of stress. Thus far, however, less is known about how vocally encoded stress (i.e., f0) might affect the stress-coping process in couples during an interaction. Therefore, the goal of the current study was to compare two different stress hypotheses (i.e., paraverbal communication stress hypothesis and emotional resonance hypothesis). We observed 187 mixed-gender couples (N = 374 participants) interacting naturally after an experimental stress induction (Trier Social Stress Test), for which couples were randomly allocated into three groups (women stressed, men stressed, and both stressed). Results of a multi-group actor-partner interdependence mediation model (APIMeM) show that either the paraverbal communication stress hypothesis or the emotional resonance hypothesis could be confirmed, depending on whether the man, the woman, or both partners were stressed.