Human imaging studies implicate the insular cortex in processing complex sounds and vocal communication signals such as speech. In addition, lesions of the insula often manifest as deficits in sound or speech recognition (auditory agnosia) and speech production. While models of acoustic perception assign an important role to the insula, little is known about the underlying neuronal substrate. Studying a vocal primate, we identified a predominantly auditory region in the caudal insula and therein discovered a neural representation of conspecific communication sounds. When probed with natural sounds, insula neurons exhibited higher response selectivity than neurons in auditory cortex, and in contrast to these, responded preferentially to conspecific vocalizations. Importantly, insula neurons not only preferred conspecific vocalizations over a wide range of environmental sounds and other animal vocalizations, but also over acoustically manipulated versions of these, demonstrating that this preference for vocalizations arises both from spectral and temporal features of the sounds. In addition, individual insula neurons responded highly selectively to only a few vocalizations and allowed the decoding of sound identity from single-trial responses. These findings characterize the caudal insula as a selectively responding auditory region, possibly part of a processing stream involved in the representation of communication sounds. Importantly, our results provide a neural counterpart for the human imaging and lesion findings and uncover a basis for a supposed role of the insula in processing vocal communication sounds such as speech.