Successful social interaction relies on multimodal integration of non-verbal emotional signals. The neural correlates of this function, along with those underlying the processing of human faces and voices, have been linked to the superior temporal sulcus (STS) in previous neuroimaging studies. Yet, recently it has been demonstrated that this structure consists of several anatomically defined sections, including a trunk section as well as two separate terminal branches, and exhibits a pronounced spatial variability across subjects. Using functional magnetic resonance imaging (fMRI), we demonstrated that the neural representations of the audiovisual integration of non-verbal emotional signals, voice sensitivity and face sensitivity are located in different parts of the STS with maximum voice sensitivity in the trunk section and maximum face sensitivity in the posterior terminal ascending branch. The audiovisual integration area for emotional signals is located at the bifurcation of the STS at an overlap of voice- and face-sensitive regions. In summary, our findings evidence a functional subdivision of the STS into modules subserving the processing of different aspects of social communication, here exemplified in human voices and faces and audiovisual integration of emotional signals from these sources and suggest a possible interaction of the underlying voice- and face-sensitive neuronal populations during the formation of the audiovisual emotional percept.