In all primate species examined so far the auditory cortex of consists of a “core” region, comprising three areas; a complex of surrounding “belt” areas, which are thought to provide more complex types of processing; and a poorly characterised “parabelt” region, which is also likely to consist of multiple areas. Although human homologues of the core, belt, and parabelt regions have been tentatively identified, the system remains best characterized in nonhuman primates. The primary auditory cortex (A1) remains the best understood component of this organization, although recent progress has also illuminated some of the functions of other core regions, including the rostral auditory area (R), the rostrotemporal auditory area (RT), and some belt areas (Bendor & Wang Reference Bendor and Wang2008; Petkov et al. Reference Petkov, Kayser, Augath and Logothetis2006; Rajan et al. Reference Rajan, Dubaj, Reser and Rosa2013).

It has been shown that trained animals can resolve temporal features of natural and synthetic human speech (Kuhl & Miller Reference Kuhl and Miller1975), and that populations of neurons in A1 of awake, untrained monkeys exhibit responses that are consistent with categorical perceptual boundaries. In particular, the voice onset time (VOT) parameter, which distinguishes pairs of spectrally similar phonemes in many languages, elicits a characteristic pattern of excitatory and inhibitory activity in A1 of awake monkeys, which is consistent with activity recorded in the auditory cortex of human subjects undergoing intracortical preoperative epilepsy monitoring (Steinschneider et al. Reference Steinschneider, Volkov, Fishman, Oya, Arezzo and Howard2005; Reference Steinschneider, Nourski and Fishman2013). Although the categorical nature of the VOT parameter has come under scrutiny (Toscano et al. Reference Toscano, McMurray, Dennhardt and Luck2010), it remains clear that the temporal features of human speech can be modelled across species – in short, the basic apparatus employed for processing of speech sound parameters is phylogenetically conserved. Therefore, it is feasible to address questions about the interaction between vocalization production and reception in animal models, and (carefully) extrapolate those results to the process of human speech. A particularly ripe area for investigation of the interactions between speech production and perception is in the realm of affective nonverbal content. It has been suggested that monkey vocalizations are akin to the nonverbal and automatic features of human vocalizations, such as laughter (Ross et al. Reference Ross, Owren and Zimmermann2010) and infant–mother interaction vocalizations (Whitham et al. Reference Whitham, Gerald and Maestripieri2007), and some monkey vocalizations have rhythmic similarities to human speech (Bergman Reference Bergman2013).

A key feature of the vocalization production model proposed by Ackermann et al. is developmental change in the role of striatal and cortico-striatal circuits in vocal skill learning. They suggest that in early life, the cortico-striatal circuits are critical for development of motor expertise, which is essential for normal speech production. Although some evidence suggests that there is limited developmental modification of monkey vocalizations (Owren et al. Reference Owren, Dieter, Seyfarth and Cheney1993; Winter et al. Reference Winter, Handley, Ploog and Schott1973), monkeys do exhibit maturational improvement of control over call features, a form of vocal “skill” (reviewed in Fedurek & Slocombe Reference Fedurek and Slocombe2011), and marmoset calls, in particular, undergo maturational change during progress toward adult communication (Pistorio et al. Reference Pistorio, Vintch and Wang2006). The motor programs that result from expert learning of speech in the Ackermann et al. model are ascribed to para- and subsylvian cortical areas, though it is unclear which areas in particular are implicated. This developmental trajectory leads to several testable hypotheses regarding the functional anatomy of auditory cortex, and of the rostral and ventral auditory and auditory associated areas of the temporal lobe in particular.

Anatomical tracing data have demonstrated that the rostral, most core area, RT, together with the belt area, RTL, have consistent monosynaptic connections with limbic areas in the rostral prefrontal, anterior cingulate, and temporal pole cortex, as well as subcortical limbic structures such as the lateral amygdala and ventral striatum (Reser et al. Reference Reser, Burman, Richardson, Spitzer and Rosa2009). Moreover, evidence from electrophysiological and imaging studies in monkeys indicates that areas in the anterior and lateral temporal lobe cortex exhibit response selectivity to sounds of increasing complexity (Kikuchi et al. Reference Kikuchi, Horwitz and Mishkin2010; Kusmierek et al. Reference Kusmierek, Ortiz and Rauschecker2012), including vocalizations, in preference to environmental sounds (Perrodin et al. Reference Perrodin, Kayser, Logothetis and Petkov2011). Selective responses to speech, and, to a lesser extent, nonverbal vocalizations including laughter and baby cries, have been obtained in recordings from likely homologous areas in the human anterior superior temporal lobe (Chan et al. Reference Chan, Dykstra, Jayaram, Leonard, Travis, Gygi, Baker, Eskandar, Hochberg, Halgren and Cash2014). It is reasonable to suspect that the anatomical and functional circuits formed by these auditory areas undergo developmental modification in parallel with the vocal output circuits described by Ackermann et al., given that we and other primates are expert listeners to conspecific vocalizations, in addition to being expert producers. This proposition could be tested in longitudinal studies of sub-adult animals (e.g., via implanted electrode arrays), and by tract tracing experiments involving rostral auditory areas at different developmental stages.

Another issue with the proposed model involves where and how the learned motor programs would be stored and encoded in “para- and subsylvian” cortical areas, and how this information could be accessed by the subcortical centers controlling laryngeal and pharyngeal movements. A notable feature of the connectional anatomy of primate auditory cortex is a paucity of projections to motor cortex. While in macaques parts of the parabelt and adjacent polysensory cortex send connections to putative homologues of Broca's area, which may be classified as a premotor area, there are few or no monosynaptic projections to the cingulate or subcortical output areas which feature prominently in Ackermann et al.'s proposal. Phonological and motor aspects of speech should be considered jointly, rather than as disparate perceptual and productive components (Ziegler et al. Reference Ziegler, Aichert and Staiger2012). Cortical microstimulation, as well as polysynaptic tract tracing using modified viruses make it feasible to map the connections from the auditory receptive areas to vocalization output pathways. We believe that further studies of descending cortical modulatory areas in the anterior cingulate will likely help understand the early development and evolution of language.