The capacity for speech and language separates humans from other animals and is the cornerstone of our intellectual and creative abilities. This capacity evolved from rudimentary forms of communication in the ancestors of humans. By studying these mechanisms in animals that represent stages of phylogenetic development, we can gain insight into the neural control of human speech that is necessary for understanding many disorders of human communication.

Vocalization is an integral part of speech and is widespread in mammalian aural communication systems. The limbic system, a group of neural structures controlling motivation and emotion, also controls most mammalian vocalizations. Although there are little supporting empirical data, many human emotional vocalizations probably involve the limbic system. This discussion considers the limbic system and those neural mechanisms thought to be necessary for normal speech and language to occur.

The anterior cingulate gyrus (ACG), which lies on the mesial surface of the frontal cortex just above and anterior to the genu of the corpus callosum, is considered part of the limbic system (Fig. 1). Electrical stimulation of the ACG in monkeys elicits vocalization and autonomic responses (Jürgens, 1994). Monkeys become mute when the ACG is lesioned, and single neurons in the ACG become active with vocalization or in response to vocalizations from conspecifics (Sutton, Larson, and Lindeman, 1974; Müller-Preuss, 1988; West and Larson, 1995). Electrical stimulation of the ACG in humans may also result in oral movements or postural distortions representative of an “archaic” level of behavior (Brown, 1988). Damage to the ACG in humans results in akinetic mutism that is accompanied by open eyes, a fixed gaze, lack of limb movement, lack of apparent affect, and nonreactance to painful stimuli (Jürgens and von Cramon, 1982). These symptoms reflect a lack of drive to initiate vocalization and many other behaviors (Brown, 1988). During recovery, a patient's ability to communicate gradually returns, first as a whisper, then with vocalization. However, the vocalizations lack prosodic features and are characterized as expressionless (Jürgens and von Cramon, 1982). These observations support the view that the ACG controls motivation for primitive forms of behavior, including prelinguistic vocalization.

Figure 1..  

The lateral and mesial surface of the human brain.

The ACG has reciprocal connections with several cortical and subcortical sites, including a premotor area homologous with Broca's area, the superior temporal gyrus, the posterior cingulate gyrus, and the supplementary motor area (SMA) (Müller-Preuss, Newman, and Jürgens, 1980). Electrical stimulation of the SMA elicits vocalization, speech arrests, hesitation, distortions, and palilalic iterations (Brown, 1988). Damage to the SMA may result in mutism, poor initiation of speech, or repetitive utterances, and during recovery, patients often go through a period in which the production of propositional speech remains severely impaired but nonpropositional speech (e.g., counting) remains relatively unaffected (Brown, 1988). Studies of other motor systems in primates suggest that the SMA is involved in selection and initiation of a remembered motor act or the correct sequencing of motor acts (Picard and Strick, 1997). Speech and vocalization fall into these categories, as both are remembered and require proper sequencing. Output from the SMA to other vocalization motor areas is a subsequent stage in the execution of vocalization.

The ACG is also connected with the perisylvian cortex of the left hemisphere (Müller-Preuss, Newman, and Jürgens, 1980), an area important for speech and language. Damage to Broca's area may cause total or partial mutism, along with expressive aphasia or apraxia (Duffy, 1995). However, vocalization is less frequently affected than speech articulation or language (Duffy, 1995), and the effect is usually temporary. In some cases, aphonia may arise from widespread damage, and recovery following therapy may suggest a diffuse, motivational, or psychogenic etiology (Sapir and Aronson, 1987). Mutism seems to occur more frequently when the opercular region of the pre- and postcentral gyri is damaged bilaterally or when the damage extends deep into the cortex, affecting the insula and possibly the basal ganglia (Jürgens, Kirzinger, and von Cramon, 1982; Starkstein, Berthier, and Leiguarda, 1988; Duffy, 1995). Additional evidence linking the insula to vocalization comes from recent studies of apraxia in humans (Dronkers, 1996) and findings of increased blood flow in the insula during singing (Perry et al., 1999). Further research is necessary to determine whether the opercular cortex alone or deeper structures (e.g., insula) are important for vocalization. In specific cases, it is necessary to know whether mutism results from psychogenic or physiological mechanisms.

The perisylvian cortex may control vocalization by one or more pathways to the medulla. The perisylvian cortex is reciprocally connected to the ACG, which projects to midbrain mechanisms involved in vocalization. The perisylvian cortex also projects directly to the medulla, where motor neurons controlling laryngeal muscles are located (Kuypers, 1958). These neuroanatomical projections are supported by observations of a short time delay (13 ms) between stimulation of the cortex and excitation of laryngeal muscles (Ludlow and Lou, 1996). The perisylvian cortex also includes the right superior temporal gyrus and Heschl's gyrus, which are preferentially active for perception of complex tones, singing, perception of one's own voice, and perhaps control of the voice by self-monitoring auditory feedback (Perry et al., 1999; Belin et al., 2000).

The other widely studied limbic system structure known for its role in vocalization, is the midbrain periaqueductal gray (PAG) (Jürgens, 1994). Lesions of the PAG in humans and animals lead to mutism (Jürgens, 1994). Electrical and chemical stimulation of the PAG in many animal species elicits species-specific vocalizations (Jürgens, 1994). A variety of techniques have shown that PAG neurons, utilizing excitatory amino acid transmitters (glutamate), activate or suppress coordinated groups of oral, facial, respiratory, and laryngeal muscles for species-specific vocalization (Larson, 1991; Jürgens, 1994). The specific pattern of activation or suppression is determined by descending inputs from the ACG and limbic system, along with sensory feedback from the auditory, laryngeal and respiratory systems (Davis, Zhang, and Bandler, 1993; Ambalavanar et al., 1999). The resultant vocalizations convey the affective state of the organism. Although this system probably is responsible for emotional vocalizations in humans, it is unknown whether this pathway is involved in normal, nonemotive speech and language.

Neurons of the PAG project to several sites in the pons and medulla, one of which is the nucleus retroambiguus (NRA) (Holstege, 1989). The NRA in turn projects to the nucleus ambiguus (NA) and spinal cord motor neurons of the respiratory muscles. Lesions of the NRA eliminate vocalizations evoked by PAG stimulation (Shiba et al., 1997), and stimulation of the NRA elicits vocalization (Zhang, Bandler, and Davis, 1995). Thus, the NRA lies functionally between the PAG and motor neurons of laryngeal and respiratory muscles controlling vocalization and may play a role in coordinating these neuronal groups (Shiba et al., 1997; Luthe, Hausler, and Jürgens, 2000).

The PAG also projects to the parvocellular reticular formation, where neurons modulate their activity with temporal and acoustical variations in monkey calls, and lesions alter the acoustical structure of vocalizations (Luthe, Hausler, and Jürgens, 2000). These data suggest that the parvocellular reticular formation is important for the regulation of vocal quality and pitch.

Finally, the NA contains laryngeal motor neurons and is crucial to vocalization. Motor neurons in the NA control laryngeal muscles during vocalization, swallowing, and respiration (Yajima and Larson, 1993), and lesions of the NA abolish vocalizations elicited by PAG stimulation (Jürgens and Pratt, 1979). The NA receives projections either indirectly from the PAG, by way of the NRA (Holstege, 1989), or directly from the cerebral cortex (Kuypers, 1958). Sensory feedback for the reflexive control of laryngeal muscles flows through the superior and recurrent laryngeal nerves to the nucleus of the solitary tract and spinal nucleus of the trigeminal nerve (Tan and Lim, 1992).

In summary, vocalization is controlled by two pathways, one that is primitive and found in most animals, and one that is found only in humans and perhaps anthropoid apes (Fig. 2). The pathway found in all mammals extends from the ACG through the limbic system and midbrain PAG to medullary and spinal motor neurons, and seems to control most emotional vocalizations. Voluntary vocal control, found primarily in humans, aided by sensory feedback, may be exerted through a direct pathway from the motor cortex to the medulla. The tendency for vocalization and human speech to be strongly affected by emotions may suggest that all vocalizations rely at least in part on the ACG-PAG pathway. Details of how these two parallel pathways are integrated are unknown.

Figure 2..  

Block diagram and arrows indicating known connections between principal structures involved in vocalization. Structures inside the dashed box are involved in vocalization in other mammals as well as in humans. Structures outside the dashed box may be found only in humans and perhaps anthropoid apes. NA, nucleus ambiguous; NRA, nucleus retroambiguus; PAG, periaqueductal gray; RF, reticular formation.

See also vocal production system: evolution.