2.1.1. Ontogenetic emergence of acoustic call morphology

The vocal repertoires of monkeys and apes encompass noise-like and harmonic components (Fig. 1A; De Waal Reference De Waal1988; Goodall Reference Goodall1986; Struhsaker Reference Struhsaker and Altmann1967; Winter et al. Reference Winter, Ploog and Latta1966). Vocal signals of both categories vary considerably across individuals, because age, body size, and stamina influence vocal tract shape and tissue characteristics, for example, the distance between the lips and the larynx (Fischer et al. Reference Fischer, Hammerschmidt, Cheney and Seyfarth2002; Reference Fischer, Kitchen, Seyfarth and Cheney2004; Fitch Reference Fitch1997; but see Rendall et al. Reference Rendall, Kollias, Ney and Lloyd2005). However, experiments based on acoustic deprivation of squirrel monkeys (Saimiri sciureus) and cross-fostering of macaques and lesser apes revealed that call structure does not appear to depend in any significant manner on species-typical auditory input (Brockelman & Schilling Reference Brockelman and Schilling1984; Geissmann Reference Geissmann1984; Hammerschmidt & Fischer Reference Hammerschmidt, Fischer, Oller and Griebel2008; Owren et al. Reference Owren, Dieter, Seyfarth and Cheney1992; Reference Owren, Dieter, Seyfarth and Cheney1993; Talmage-Riggs et al. Reference Talmage-Riggs, Winter, Ploog and Mayer1972; Winter et al. Reference Winter, Handley, Ploog and Schott1973). Thus, ontogenetic modifications of acoustic structure may simply reflect maturation of the vocal apparatus, including “motor-training” effects (Hammerschmidt & Fischer Reference Hammerschmidt, Fischer, Oller and Griebel2008; Pistorio et al. Reference Pistorio, Vintch and Wang2006), or the influence of hormones related to social status (Roush & Snowdon Reference Roush and Snowdon1994; Reference Roush and Snowdon1999). In contrast, comprehension and usage of acoustic signals show considerably more malleability than acoustic structure both in juvenile and adult animals (Owren et al. Reference Owren, Amoss and Rendall2011).

Figure 1A.

Acoustic communication in nonhuman primates: Call structure.

A. Spectrograms (left-hand section of each panel) and power spectra (right-hand section in each) of two common rhesus monkey vocalizations, that is, a “coo” (left panel) and a “grunt” (right panel). Gray level of the spectrograms codes for spectral energy. Coo calls (left panel) are characterized by a harmonic structure, encompassing a fundamental frequency (F0, the lowest and darkest band) and several harmonics (H₁ to H_n). Measures derived from the F0 contour provide robust criteria for a classification of periodic signals, for example, peak frequency (peakF; Hardus et al. Reference Hardus, Lameira, Singleton, Morrogh-Bernard, Knott, Ancrenaz, Utami Atmoko, Wich, Wich, Utami Atmoko, Setia and van Schaik2009a). Onset F0 seems to be highly predictive for the shape of the intonation contour, indicating the implementation of a “vocal plan” prior to movement initiation (Miller et al. Reference Miller, Beck, Meade and Wang2009a; Reference Miller, Eliades and Wang2009b). Grunts (right) represent short and noisy calls whose spectra include more energy in the lower frequency range and a rather flat energy distribution.

2.1.2. Spontaneous adult call plasticity: Convergence on and imitation of species-typical variants of vocal behavior

Despite innate acoustic call structures, the vocalizations of nonhuman primates may display some context-related variability in adulthood. For example, two populations of pygmy marmosets (Cebuella pygmaea) of a different geographic origin displayed convergent shifts of spectral and durational call parameters (Elowson & Snowdon Reference Elowson and Snowdon1994; see further examples in Snowdon & Elowson Reference Snowdon and Elowson1999 and Rukstalis et al. Reference Rukstalis, Fite and French2003). Humans may also match their speaking styles inadvertently during conversation (“speech accommodation theory”; Burgoon et al. Reference Burgoon, Floyd, Guerrero, Berger, Roloff and Roskos-Ewoldsen2010; see Masataka [Reference Masataka and Masataka2008a; Reference Masataka and Masataka2008b] for an example). Such accommodation effects could provide a basis for the changes in call morphology during social interactions in nonhuman primates (Fischer Reference Fischer and Ghazanfar2003; Mitani & Brandt Reference Mitani and Brandt1994; Mitani & Gros-Louis Reference Mitani and Gros-Louis1998; Sugiura Reference Sugiura1998). Subsequent reinforcement processes may give rise to “regional dialects” of primate species (Snowdon Reference Snowdon, Oller and Griebel2008). Rarely, even memory-based imitation capabilities have been observed in great apes: Thus, free-living chimpanzees were found to copy the distinctive intonational and rhythmic pattern of the pant hoots of other subjects – even after the animal providing the acoustic template had disappeared from the troop (Boesch & Boesch-Achermann Reference Boesch and Boesch-Achermann2000, pp. 234f ). Whatever the precise mechanisms of vocal convergence, these phenomena are indicative of the operation of a neuronal feedback loop between auditory perception and vocalization in nonhuman primates (see Brumm et al. Reference Brumm, Voss, Köllmer and Todt2004).

A male bonobo infant (“Kanzi”) reared in an enriched social environment spontaneously augmented his species-typical repertoire by four “novel” vocalizations (Hopkins & Savage-Rumbaugh Reference Hopkins and Savage-Rumbaugh1991). However, these newly acquired signals can be interpreted as scaled variants of a single intonation contour (Fig. 3 in Taglialatela et al. Reference Taglialatela, Savage-Rumbaugh and Baker2003). Since Pan paniscus has, to some degree, a graded rather than discrete call system (Bermejo & Omedes Reference Bermejo and Omedes1999; Clay & Zuberbühler Reference Clay and Zuberbühler2009), new behavior challenges could give rise to a differentiation of the available “vocal space” – indicating a potential to modulate call structures within the range of innate acoustic constraints rather than the ability to learn new vocal signals. An alternative interpretation is that hitherto un-deployed vocalizations were recruited under those conditions (Lemasson & Hausberger Reference Lemasson and Hausberger2004; Lemasson et al. Reference Lemasson, Hausberger and Zuberbühler2005).

2.1.3. Volitional initiation of vocal behavior and modulation of acoustic call structure

It has been a matter of debate for decades, in how far nonhuman primates are capable of volitional call initiation and modulation. A variety of behavioral studies seem to indicate both control over the timing of vocal output and the capacity to “decide” which acoustic signal to emit in a given context. First, at least two species of New World primates (tamarins, marmosets) discontinue acoustic communication during epochs of increased ambient noise in order to avoid signal interferences and, therefore, to increase call detection probability (Egnor et al. Reference Egnor, Wickelgren and Hauser2007; Roy et al. Reference Roy, Miller, Gottsch and Wang2011). In addition, callitrichid monkeys obey “conversational rules” and show response selectivity during vocal exchanges (Miller et al. Reference Miller, Beck, Meade and Wang2009a; Reference Miller, Eliades and Wang2009b; but see Rukstalis et al. Reference Rukstalis, Fite and French2003: independent F0 onset change). Such observations were assumed to indicate some degree of volitional control over call production. As an alternative interpretation, these changes in vocal timing or loudness could simply reflect threshold effects of audio-vocal integration mechanisms. Second, several nonhuman primates produce acoustically different alarm vocalizations in response to distinct predator species, suggesting volitional access to call type (e.g., Seyfarth et al. Reference Seyfarth, Cheney and Marler1980). Again, variation of motivational states could account for these findings. For example, the approach of an aerial predator could represent a much more threatening event than the presence of a snake. To some extent, even dynamic spectro-temporal features resembling the formant transients of the human acoustic speech signal (see below sect. 4.1.) appear to contribute to the differentiation of predator-specific alarm vocalizations (“leopard calls”) in Diana monkeys (Cercopithecus diana) (Riede & Zuberbühler Reference Riede and Zuberbühler2003a; Reference Riede and Zuberbühler2003b; see Lieberman [1968] for earlier data). Yet, computer models insinuate that larynx lowering makes a critical contribution to these changes (Riede et al. Reference Riede, Bronson, Hatzikirou and Zuberbühler2005; Reference Riede, Bronson, Hatzikirou and Zuberbühler2006; see critical comments in Lieberman Reference Lieberman2006b), thus, eliciting in a receiver the impression of a bigger-than-real body size of the sender (Fitch Reference Fitch2000b; Fitch & Reby Reference Fitch and Reby2001). Diana monkeys may have learned this manoeuver as a strategy to mob large predators, a behavior often observed in the wild (Zuberbühler & Jenny Reference Zuberbühler, Jenny, McGraw, Zuberbühler and Noe2007).

The question of whether nonhuman primates are able to decouple their vocalizations from accompanying motivational states and to use them in a goal-directed manner has been addressed in several operant-conditioning experiments (Aitken & Wilson Reference Aitken and Wilson1979; Coudé et al. Reference Coudé, Ferrari, Rodà, Maranesi, Borelli, Veroni, Monti, Rozzi and Fogassi2011; Hage et al. Reference Hage, Gavrilov and Nieder2013; Koda et al. Reference Koda, Oyakawa, Kato and Masataka2007; Sutton et al. Reference Sutton, Larson, Taylor and Lindeman1973; West & Larson Reference West and Larson1995). In most of these studies, nonhuman primates learned to utter a vocalization in response to a food reward (e.g., Coudé et al. Reference Coudé, Ferrari, Rodà, Maranesi, Borelli, Veroni, Monti, Rozzi and Fogassi2011; Koda et al. Reference Koda, Oyakawa, Kato and Masataka2007). Rather than demonstrating the ability to volitionally vocalize on command, these studies merely confirm, essentially, that nonhuman primates produce adequate, motivationally based behavioral reactions to hedonistic stimuli. A recent study found, however, that rhesus monkeys can be trained to produce different call types in response to arbitrary visual signals and that they are capable to switch between two distinct call types associated with different cues on a trial-to-trial basis (Hage et al. Reference Hage, Gavrilov and Nieder2013). These observations indicate that the animals are able – within some limits – to volitionally initiate vocalizations and, therefore, are capable to instrumentalize their vocal utterances in order to accomplish behavioral tasks successfully. Likewise, macaque monkeys may acquire control over loudness and duration of coo calls (Hage et al. Reference Hage, Gavrilov and Nieder2013; Larson et al. Reference Larson, Sutton, Taylor and Lindeman1973; Sutton et al. Reference Sutton, Larson, Taylor and Lindeman1973; Reference Sutton, Trachy and Lindeman1981; Trachy et al. Reference Trachy, Sutton and Lindeman1981). A more recent investigation even reported spontaneous differentiation of coo calls in Japanese macaques with respect to peak and offset of the F0 contour during operant tool-use training (Hihara et al. Reference Hihara, Yamada, Iriki and Okanoya2003). Such accomplishments may, however, be explained by the adjustment of respiratory functions and do not conclusively imply operant control over spectro-temporal call structure in nonhuman primates (Janik & Slater Reference Janik, Slater, Slater, Rosenblatt, Snowdon and Milinski1997; Reference Janik and Slater2000).

2.1.4. Observational acquisition of species-atypical sounds

Few instances of species-atypical vocalizations in nonhuman primates have been reported so far. Allegedly, the bonobo Kanzi, mentioned earlier, spontaneously acquired a few vocalizations resembling spoken words (Savage-Rumbaugh et al. Reference Savage-Rumbaugh, Fields and Spircu2004). Yet, systematic perceptual data substantiating these claims are not available. As further anecdotal evidence, Wich et al. (Reference Wich, Swartz, Hardus, Lameira, Stromberg and Shumaker2009) reported that a captive-born female orangutan (Pongo pygmaeus × Pongo abelii) began to produce human-like whistles at an age of about 12 years in the absence of any training. Furthermore, an idiosyncratic pant hoot variant (“Bronx cheer” – resembling a sound called “blowing raspberries”) spread throughout a colony of several tens of captive chimpanzees after it had been introduced by a male joining the colony (Hopkins et al. Reference Hopkins, Taglialatela and Leavens2007; Marshall et al. Reference Marshall, Wrangham and Arcadi1999; similar sounds have been observed in wild orangutans: Hardus et al. Reference Hardus, Lameira, Singleton, Morrogh-Bernard, Knott, Ancrenaz, Utami Atmoko, Wich, Wich, Utami Atmoko, Setia and van Schaik2009a; Reference Hardus, Lameira, van Schaik and Wich2009b; van Schaik et al. Reference van Schaik, Ancrenaz, Borgen, Galdikas, Knott, Singleton, Suzuki, Utami and Merrill2003; Reference van Schaik, van Noordwijk and Wich2006). Remarkably, these two acoustic displays, “raspberries” and whistles, do not engage laryngeal sound-production mechanisms, but reflect a linguo-labial trill (“raspberries”) or arise from oral air-stream resonances (whistles). Thus, the species-atypical acoustic signals in nonhuman primates observed to date spare glottal mechanisms of sound generation. Apparently, laryngeal motor activity cannot be decoupled volitionally from species-typical audiovisual displays (Knight Reference Knight, Dunbar, Knight and Power1999).

2.2.1. Brainstem mechanisms (PAG and pontine vocal pattern generator)

Since operant conditioning of the calls of nonhuman primates is technically challenging (Pierce Reference Pierce1985), analyses of the neurobiological control mechanisms engaged in phonatory functions relied predominantly on electrical brain stimulation. In squirrel monkeys (Saimiri sciureus) – the species studied most extensively so far (Gonzalez-Lima Reference Gonzalez-Lima and Brudzynski2010) – vocalizations could be elicited at many cerebral locations, extending from the forebrain to the lower brainstem. This network encompasses a variety of subcortical limbic structures such as the hypothalamus, septum, and amygdala (Fig. 1B; Brown Reference Brown1915; Jürgens Reference Jürgens2002b; Jürgens & Ploog Reference Jürgens and Ploog1970; Smith Reference Smith1945). In mammals, all components of this highly conserved “communicating brain” (Newman Reference Newman2003) appear to project to the periaqueductal grey (PAG) of the midbrain and the adjacent mesencephalic tegmentum (Gruber-Dujardin Reference Gruber-Dujardin and Brudzynski2010).Footnote ⁴ Based on the integration of input from motivation-controlling regions, sensory structures, motor areas, and arousal-related systems, the PAG seems to gate the vocal dimension of complex multi-modal emotional responses such as fear or aggression. The subsequent coordination of cranial nerve nuclei engaged in the innervation of vocal tract muscles depends on a network of brainstem structures, including, particularly, a vocal pattern generator bound to the ventrolateral pons (Hage Reference Hage and Brudzynski2010a; Reference Hage and Brudzynski2010b; Hage & Jürgens Reference Hage and Jürgens2006).

Figure 1B.

Acoustic Communication in nonhuman Primates: Cerebral Organization.

Cerebral “vocalization network” of the squirrel monkey (as a model of the primate-general “communication brain”). The solid lines represent the “vocal brainstem circuit” of the vocalization network and its modulatory cortical input (ACC), the dotted lines the strong connections of sensory cortical regions (AC, VC) and motivation-controlling limbic structures (Ac, Hy, Se, St) to this circuit.

Key: ACC = Anterior cingulate cortex; AC = Auditory cortex; Ac = Nucleus accumbens; Hy = Hypothalamus; LRF = Lateral reticular formation; NRA = Nucleus retroambigualis; PAG = periaqueductal gray; PB = brachium pontis; SC = superior colliculus; Se = Septum; St = Nucleus stria terminalis; VC = Visual cortex (Unpublished figure. See Jürgens Reference Jürgens2002b and Hage Reference Hage and Brudzynski2010a; Reference Hage and Brudzynski2010b for further details).

2.2.2. Mesiofrontal cortex and higher-order aspects of vocal behavior

Electrical stimulation studies revealed that both New and Old World monkeys possess a “cingulate vocalization region” within the anterior cingulate cortex (ACC), adjacent to the anterior pole of the corpus callosum (Jürgens Reference Jürgens2002b; Smith Reference Smith1945; Vogt & Barbas Reference Vogt, Barbas and Newman1988). Uni- and bilateral ACC ablation in macaques had, however, a minor and inconsistent impact on spontaneously uttered coo calls, but disrupted the vocalizations produced in response to an operant-conditioning task (Sutton et al. Reference Sutton, Larson and Lindeman1974; Trachy et al. Reference Trachy, Sutton and Lindeman1981). Furthermore, damage to preSMA – a cortical area neighboring the ACC in dorsal direction and located rostral to the supplementary motor area (SMA proper) – resulted in significantly prolonged response latencies (Sutton et al. Reference Sutton, Trachy and Lindeman1985). Comparable lesions in squirrel monkeys diminish the rate of spontaneous isolation peeps, but the acoustic structure of the produced calls remains undistorted (Kirzinger & Jürgens Reference Kirzinger and Jürgens1982). As a consequence, mesiofrontal cerebral structures appear to predominantly mediate calls driven by an animal's internal motivational milieu.

2.2.3. Ventrolateral frontal lobe and corticobulbar system

Both squirrel and rhesus monkeys possess a neocortical representation of internal and external laryngeal muscles in the ventrolateral part of premotor cortex, bordering areas associated with orofacial structures, namely, tongue, lips, and jaw (Fig. 1 in Hast et al. Reference Hast, Fischer, Wetzel and Thompson1974; Jürgens Reference Jürgens1974; Simonyan & Jürgens Reference Simonyan and Jürgens2002; Reference Simonyan and Jürgens2005). Furthermore, vocalization-selective neuronal activity may arise at the level of the premotor cortex in macaques that are trained to respond with coo calls to food rewards (Coudé et al. Reference Coudé, Ferrari, Rodà, Maranesi, Borelli, Veroni, Monti, Rozzi and Fogassi2011). Interestingly, premotor neural firing appears to occur only when the animals produce vocalizations in a specific learned context of food reward, but not under other conditions. Finally, a cytoarchitectonic homologue to Broca's area of our species has been found between the lower branch of the arcuate sulcus and the subcentral dimple just above the Sylvian fissure in Old World monkeys (Gil-da-Costa et al. Reference Gil-da-Costa, Martin, Lopes, Muňoz, Fritz and Braun2006; Petrides & Pandya Reference Petrides and Pandya2009; Petrides et al. Reference Petrides, Cadoret and Mackey2005) and chimpanzees (Sherwood et al. Reference Sherwood, Broadfield, Holloway, Gannon and Hof2003). Nevertheless, even bilateral damage to the ventrolateral aspects of the frontal lobes has no significant impact on the vocal behavior of monkeys (P. G. Aitken Reference Aitken1981; Jürgens et al. Reference Jürgens, Kirzinger and von Cramon1982; Myers Reference Myers, Harnad, Steklis and Lancaster1976; Sutton et al. Reference Sutton, Larson and Lindeman1974). Electrical stimulation of these areas in nonhuman primates also failed to elicit overt acoustic responses, apart from a few instances of “slight grunts” obtained from chimpanzees (Bailey et al. Reference Bailey, von Bonin and McCulloch1950, pp. 334f, 355f). Therefore, spontaneous call production, at least, does not critically depend on the integrity of the cortical larynx representation (Ghazanfar & Rendall Reference Ghazanfar and Rendall2008; Simonyan & Jürgens Reference Simonyan and Jürgens2005). Most likely, however, experimental lesions have not included the full extent or even the bulk of the Broca homologue of nonhuman primates as determined by recent cytoarchitectonic studies (Fig. 4 in Aitken Reference Aitken1981; Fig. 1 in Sutton et al. Reference Sutton, Larson and Lindeman1974). The role of this area in the control of vocal behavior in monkeys still remains to be clarified. Nonhuman primates appear endowed with a more elaborate cerebral organization of orofacial musculature as compared to the larynx, which, presumably, provides the basis for their relatively advanced orofacial imitation capabilities (Morecraft et al. Reference Morecraft, Louie, Herrick and Stilwell-Morecraft2001). As concerns the basal ganglia and the cerebellum, the lesion and stimulation studies available so far do not provide reliable evidence for a participation of these structures in the control of motor aspects of vocal behavior (Kirzinger Reference Kirzinger1985; Larson et al. Reference Larson, Sutton and Lindeman1978; Robinson Reference Robinson1967).

Prosimians and New World monkeys are endowed solely with polysynaptic corticobulbar projections to lower brain-stem motoneurons (Sherwood Reference Sherwood2005; Sherwood et al. Reference Sherwood, Hof, Holloway, Semendeferi, Gannon, Frahm and Zilles2005). By contrast, morphological and neurophysiological studies revealed direct connections of the precentral gyrus of Old World monkeys and chimpanzees to the cranial nerve nuclei engaged in the innervation of orofacial muscles (Jürgens & Alipour Reference Jürgens and Alipour2002; Kuypers Reference Kuypers1958b; Morecraft et al. Reference Morecraft, Louie, Herrick and Stilwell-Morecraft2001) which, together with the aforementioned more elaborate cortical representation of orofacial structures, may contribute to the enhanced facial-expressive capabilities of anthropoid primates (Sherwood et al. Reference Sherwood, Hof, Holloway, Semendeferi, Gannon, Frahm and Zilles2005). Most importantly, the direct connections between motor cortex and nucleus (nu.) ambiguus appear restricted, even in chimpanzees, to a few fibers targeting its most rostral component (Kuypers Reference Kuypers1958b), subserving the innervation of pharyngeal muscles via the ninth cranial nerve (Butler & Hodos Reference Butler and Hodos2005). By contrast, humans exhibit considerably more extensive monosynaptic cortical input to the motoneurons engaged in the innervation of the larynx – though still less dense than the projections to the facial and hypoglossal nuclei (Iwatsubo et al. Reference Iwatsubo, Kuzuhara, Kanemitsu, Shimada and Toyokura1990; Kuypers Reference Kuypers1958a). In addition, functional imaging data point to a primary motor representation of human internal laryngeal muscles adjacent to the lips of the homunculus and spatially separated from the frontal larynx region of New and Old World monkeys (Brown et al. Reference Brown, Ngan and Liotti2008; Reference Brown, Laird, Pfordresher, Thelen, Turkeltaub and Liotti2009; Bouchard et al. Reference Bouchard, Mesgarani, Johnson and Chang2013). As a consequence, thus, the monosynaptic elaboration of corticobulbar tracts during hominin evolution might have been associated with a refinement of vocal tract motor control at the cortical level (“Kuypers/Jürgens hypothesis”; Fitch et al. Reference Fitch, Huber and Bugnyar2010).Footnote ⁵

3.2.1. Anterior cingulate cortex (ACC)

There is some evidence that, similar to subhuman primates, the ACC is a mediator of emotional/motivational acoustic expression in humans as well (see sect. 2.2.2). A clinical example is frontal lobe epilepsy, a syndrome characterized by involuntary and stereotyped bursts of laughter (“gelastic seizures”; Wild et al. Reference Wild, Rodden, Grodd and Ruch2003) that lack any concomitant adequate emotions (Arroyo et al. Reference Arroyo, Lesser, Gordon, Uematsu, Hart, Schwerdt, Andreasson and Fisher1993; Chassagnon et al. Reference Chassagnon, Minotti, Kremer, Verceuil, Hoffmann, Benabid and Kahane2003; Iannetti et al. Reference Iannetti, Spalice, Raucci, Atzei and Cipriani1997; Iwasa et al. Reference Iwasa, Shibata, Mine, Koseki, Yasuda, Kasagi, Okada, Yabe, Kaneko and Nakajima2002). The cingulate gyrus appears to be the most commonly disrupted site based on lesion surveys of gelastic seizure patients (Kovac et al. Reference Kovac, Deppe, Mohammadi, Schiffbauer, Schwindt, Möddel, Dogan and Evers2009). This suggestion was further corroborated by a recent case study in which electrical stimulation of the right-hemisphere ACC rostral to the genu of the corpus callosum elicited uncontrollable, but natural-sounding laughter – in the absence of merriment (Sperli et al. Reference Sperli, Spinelli, Pollo and Seeck2006). Conceivably, a homologue of the vocalization center of nonhuman primates bound to rostral ACC may underlie stereotyped motor patterns associated with emotional vocalizations in humans.

Does the ACC participate in speaking as well? Based on an early PET study, “two distinct speech-related regions in the human anterior cingulate cortex” were proposed, the more anterior of which was considered to be homologous to the cingulate vocalization center of nonhuman primates (Paus et al. Reference Paus, Tomaiuolo, Otaky, MacDonald, Petrides, Atlas, Morris and Evans1996, p. 213). A recent and more focused functional imaging experiment by Loucks et al. (Reference Loucks, Poletto, Simonyan, Reynolds and Ludlow2007) failed to substantiate this claim. However, this investigation was based on rather artificial phonation tasks involving prolonged and repetitive vowel productions which do not allow for an evaluation of the specific role of the ACC in the mediation of emotional aspects of speaking. In another study, Schulz et al. (Reference Schulz, Varga, Jeffires, Ludlow and Braun2005) required participants to recount a story in a voiced and a whispered speaking mode and demonstrated enhanced hemodynamic activation during the voiced condition in a region homologous to the cingulate vocalization center, but much larger responses emerged in contiguous neocortical areas of medial prefrontal cortex. It remains unclear, however, how the observed activation differences between voiced and whispered utterances should be interpreted, since both of these phonation modes require specific laryngeal muscle activity. One investigation explicitly aimed at a further elucidation of the role of medial prefrontal cortex in motivational aspects of speech production by analyzing the covariation of induced emotive prosody with blood oxygen level dependent (BOLD) signal changes as measured by functional magnetic resonance imaging (fMRI; Barrett et al. Reference Barrett, Pike and Paus2004). Affect-related pitch variation was found to be associated with supracallosal rather than pregeniculate hemodynamic activation. However, the observed response modulation may have been related to changes in the induced emotional states rather than pitch control. On the whole, the available functional imaging data do not provide conclusive support for the hypothesis that the prosodic modulation of verbal utterances critically depends on the ACC.

The results of lesion studies are similarly inconclusive. Bilateral ACC damage due to cerebrovascular disorders or tumours has been reported to cause a syndrome of akinetic mutism (Brown Reference Brown and Newman1988; for a review, see Ackermann & Ziegler Reference Ackermann and Ziegler1995). Early case studies found the behavioral deficits to extend beyond verbal and nonverbal acoustic communication: Apparently vigilant subjects with normal muscle tone and deep tendon reflexes displayed diminished or abolished spontaneous body movements, delayed or absent reactions to external stimuli, and impaired autonomic functions (e.g., Barris & Schuman Reference Barris and Schuman1953). By contrast, bilateral surgical resection of the ACC (cingulectomy), performed most often in patients suffering from medically intractable pain or psychiatric diseases, failed to significantly compromise acoustic communication (Brotis et al. Reference Brotis, Kapsalaki, Paterakis, Smith and Fountas2009). The complex functional-neuroanatomic architecture of the anterior mesiofrontal cortex hampers, however, any straightforward interpretation of these clinical data. In monkeys, the cingulate sulcus encompasses two or even three distinct “cingulate motor areas” (CMAs), which project to the supplementary motor area (SMA), among other regions (Dum & Strick Reference Dum and Strick2002; Morecraft & van Hoesen Reference Morecraft and van Hoesen1992; Morecraft et al. Reference Morecraft, Louie, Herrick and Stilwell-Morecraft2001). Humans exhibit a similar compartmentalization of the medial wall of the frontal lobes (Fink et al. Reference Fink, Frackowiak, Pietrzyk and Passingham1997; Picard & Strick Reference Picard and Strick1996). A closer look at the aforementioned surgical data reveals that bilateral cingulectomy for treatment of psychiatric disorders, as a rule, did not encroach on caudal ACC (Le Beau Reference Le Beau1954; Whitty Reference Whitty1955; for a review, see Brotis et al. Reference Brotis, Kapsalaki, Paterakis, Smith and Fountas2009, p. 276). Thus, tissue removal restricted to rostral ACC components could explain the relatively minor effects of this surgical approach.Footnote ⁶ Conceivably, mesiofrontal akinetic mutism reflects bilateral damage to the caudal CMA and/or its efferent projections, rather than dysfunction of a “cingulate vocalization center” bound to rostral ACC. Instead, the anterior mesiofrontal cortex has been assumed to contribute to reward-dependent selection/inhibition of verbal responses in conflict situations rather than to motor aspects of speaking (Calzavara et al. Reference Calzavara, Mailly and Haber2007; Paus Reference Paus2001). This interpretation is compatible with the fact that psychiatric conditions bound to ACC pathology such as obsessive-compulsive disorder or Tourette syndrome cause, among other things, socially inappropriate vocal behavior (Müller-Vahl et al. Reference Müller-Vahl, Kaufmann, Grosskreutz, Dengler, Emrich and Peschel2009; Radua et al. Reference Radua, van den Heuvel, Surguladze and Mataix-Cols2010; Seeley Reference Seeley2008).

3.2.2. Supplementary motor area (SMA)

Damage to the SMA in the language-dominant hemisphere may give rise to diminished spontaneous speech production, characterized by delayed, brief, and dysfluent, but otherwise well-articulated verbal responses without any central-motor disorders of vocal tract muscles or impairments of other language functions such as speech comprehension or reading aloud (“transcortical motor aphasia”; for a review of the earlier literature, see Jonas Reference Jonas1981; Reference Jonas and Perecman1987; more recent case studies in Ackermann et al. Reference Ackermann, Hertrich, Ziegler, Bitzer and Bien1996 and Ziegler et al. Reference Ziegler, Kilian and Deger1997).Footnote ⁷ This constellation may arise from initial mutism via an intermediate stage of silent word mouthing (Rubens Reference Rubens1975) or whispered speaking (Jürgens & von Cramon Reference Jürgens and von Cramon1982; Masdeu et al. Reference Masdeu, Schoene and Funkenstein1978; Watson et al. Reference Watson, Fleet, Gonzalez-Rothi and Heilman1986). Based on these clinical observations, the SMA, apparently, supports the initiation (“starting mechanism”) and maintenance of vocal tract activities during speech production (Botez & Barbeau Reference Botez and Barbeau1971; Jonas Reference Jonas1981). Indeed, movement-related potentials preceding self-paced tongue protrusions and vocalizations were recorded over the SMA (Bereitschaftspotential; Ikeda et al. Reference Ikeda, Lüders, Burgess and Shibasaki1992). Calculation of the time course of BOLD signal changes during syllable repetition tasks, preceded by a warning stimulus, revealed an earlier peak of the SMA response relative to primary sensorimotor cortex (Brendel et al. Reference Brendel, Hertrich, Erb, Lindner, Riecker, Grodd and Ackermann2010). These data corroborate the suggestion – based on clinical data – of an engagement of the SMA in the preparation and initiation of verbal utterances, that is, pre-articulatory control processes.

4.2.1. The impact of pre- and perinatal striatal dysfunctions on spoken language

Insight into the potential contributions of the basal ganglia to human speech acquisition can be obtained from damage to these nuclei at a prelinguistic age. Distinct mutations of mitochondrial or nuclear DNA may give rise to infantile bilateral striatal necrosis, a constellation largely restricted to this basal ganglia component (Basel-Vanagaite et al. Reference Basel-Vanagaite, Muncher, Straussberg, Pasmanik-Chor, Yahav, Rainshtein, Walsh, Magal, Taub, Drasinover, Shalev, Attia, Rechavi, Simon and Shohat2006; De Meirleir et al. Reference De Meirleir, Seneca, Lissens, Schoentjes and Desprechins1995; Kim et al. Reference Kim, Ki and Park2010; Solano et al. Reference Solano, Roig, Vives-Bauza, Hernandez-Peña, Garcia-Arumi, Playan, Lopez-Perez, Andreu and Montoya2003; Thyagarajan et al. Reference Thyagarajan, Shanske, Vazquez-Memije, DeVivo and DiMauro1995). At least two variants, both of them point mutations of the mitochondrial ATPase 6 gene, were associated with impaired speech learning capabilities (De Meirleir et al. Reference De Meirleir, Seneca, Lissens, Schoentjes and Desprechins1995: “speech delayed for age”; Thyagarajan et al. Reference Thyagarajan, Shanske, Vazquez-Memije, DeVivo and DiMauro1995, case 1: “no useful language at age 3 years”). As a further clinical paradigm, birth asphyxia may predominantly impact the basal ganglia and the thalamus (eventually, in addition, the brainstem) under specific conditions such as uterine rupture or umbilical cord prolapse, while the cerebral cortex and the underlying white matter are less affected (Roland et al. Reference Roland, Poskitt, Rodriguez, Lupton and Hill1998). A clinical study found nine children out of a group of 17 subjects with this syndrome completely unable to produce any verbal utterances at the ages of 2 to 9 years (Krägeloh-Mann et al. Reference Krägeloh-Mann, Helber, Mader, Staudt, Wolff, Groenendaal and DeVries2002). Six further patients showed significantly compromised articulatory functions (“dysarthria”). Most importantly, five children had not mastered adequate articulate speech at the ages of 3 to 12 years, though lesions were confined to the putamen and ventro-lateral thalamus, sparing the caudate nucleus and the precentral gyrus.

Data from a severe developmental speech or language disorder of monogenic autosomal-dominant inheritance with full penetrance extending across several generations of a large family provide further evidence of a connection between the basal ganglia and ontogenetic speech acquisition (KE family; Hurst et al. Reference Hurst, Baraitser, Auger, Graham and Norell1990). At first considered a highly selective inability to acquire particular grammatical rules (Gopnik Reference Gopnik1990a; for more details, see Taylor Reference Taylor2009), extensive neuropsychological evaluations revealed a broader phenotype of psycholinguistic dysfunctions, including nonverbal aspects of intelligence (Vargha-Khadem & Passingham Reference Vargha-Khadem and Passingham1990; Vargha-Khadem et al. Reference Vargha-Khadem, Watkins, Alcock, Fletcher and Passingham1995; Watkins et al. Reference Watkins, Dronkers and Vargha-Khadem2002a). However, the most salient behavioral deficit in the afflicted individuals consists of pronounced abnormalities of speech articulation (“developmental verbal dyspraxia”) that render spoken language “of many of the affected members unintelligible to the naive listener” (Vargha-Khadem et al. Reference Vargha-Khadem, Watkins, Alcock, Fletcher and Passingham1995, p. 930; see also Fee Reference Fee1995; Shriberg et al. Reference Shriberg, Aram and Kwiatkowski1997). Furthermore, the speech disorder was found to compromise voluntary control of nonverbal vocal tract movements (Vargha-Khadem et al. Reference Vargha-Khadem, Gadian, Copp and Mishkin2005). More specifically, the phenotype includes a significant disruption of simultaneous or sequential sets of motor activities to command, in spite of a preserved motility of single vocal tract organs (Alcock et al. Reference Alcock, Passingham, Watkins and Vargha-Khadem2000a) and uncompromised reproduction of tones and melodies (Alcock et al. Reference Alcock, Passingham, Watkins and Vargha-Khadem2000b).

A heterozygous point mutation (G-to-A nucleotide transition) of the FOXP2 gene (located on chromosome 7; coding for a transcription factor) could be detected as the underlying cause of the behavioral disorder (for a review, see Fisher et al. Reference Fisher, Lai and Monaco2003).Footnote ⁸ Volumetric analyses of striatal nuclei revealed bilateral volume reduction in the afflicted family members, the extent of which was correlated with oral-motor impairments (Watkins et al. Reference Watkins, Vargha-Khadem, Ashburner, Passingham, Connelly, Friston, Frackowiak, Mishkin and Gadian2002b). Mice and humans share all but three amino acids in the FOXP2 protein, suggesting a high conservation of the respective gene across mammals (Enard et al. Reference Enard, Przeworski, Fisher, Lai, Wiebe, Kitano, Monaco and Pääbo2002; Zhang et al. Reference Zhang, Webb and Podlaha2002). Furthermore, two of the three substitutions must have emerged within our hominin ancestors after separation from the chimpanzee lineage. Since primates lacking the human FOXP2 variant cannot even imitate the simplest speech-like utterances, and since disruption of this gene in humans gives rise to severe articulatory deficits, it appears warranted to assume that the human variant of this gene locus represents a necessary prerequisite for the phylogenetic emergence of articulate speech. Most noteworthy, animal experimentation suggests that the human-specific copy of this gene is related to acoustic communication (Enard et al. Reference Enard, Gehre, Hammerschmidt, Hölter, Blass, Somel, Brückner, Schreiweis, Winter, Sohr, Becker, Wiebe, Nickel, Giger, Müller, Groszer, Adler, Aguilar, Bolle, Calzada-Wack, Dalke, Ehrhardt, Favor, Fuchs, Gailus-Durner, Hans, Hölzlwimmer, Javaheri, Kalaydjiev, Kallnik, Kling, Kunder, Mossbrugger, Naton, Racz, Rathkolb, Rozman, Schrewe, Busch, Graw, Ivandic, Klingenspor, Klopstock, Ollert, Quintanilla-Martinez, Schulz, Wolf, Wurst, Zimmer, Fisher, Morgenstern, Arendt, de Angelis, Fischer, Schwarz and Pääbo2009) and directly influences the dendritic architecture of the neurons embedded into cortico-basal ganglia–thalamo–cortical circuits (Reimers-Kipping et al. Reference Reimers-Kipping, Hevers, Pääbo and Enard2011, p. 82).

4.2.2. Motor aprosodia in Parkinson's disease

A loss of midbrain neurons within the substantia nigra pars compacta (SNc) represents the pathophysiological hallmark of Parkinson's disease (PD; idiopathic Parkinsonian syndrome), one of the most common neurodegenerative disorders (Evatt et al. Reference Evatt, DeLong, Vitek, Asbury, McKhann, McDonald and Goadsby2002; Wichmann & DeLong Reference Wichmann, DeLong, Koller and Melamed2007). This degenerative process results in a depletion of the neurotransmitter dopamine at the level of the striatum, rendering PD a model of dopaminergic dysfunction of the basal ganglia, characterized within the motor domain by akinesia (bradykinesia, hypokinesia), rigidity, tremor at rest, and postural instability (Jankovic Reference Jankovic2008; Marsden Reference Marsden1982). In advanced stages, functionally relevant morphological changes of striatal projection neurons may emerge (Deutch et al. Reference Deutch, Colbran and Winder2007; see Mallet et al. [Reference Mallet, Ballion, Le Moine and Gonon2006] for other nondopaminergic PD pathomechanisms). Recent studies suggest that the disease process develops first in extranigral brainstem regions such as the dorsal motor nucleus of the glossopharyngeal and vagal nerves (Braak et al. Reference Braak, Del Tredici, Rüb, de Vos, Jansen Steur and Braak2003). These initial lesions affect the autonomic-vegetative nervous system, but do not encroach on gray matter structures engaged in the control of vocal tract movements such as the nu. ambiguus.

A classical tenet of speech pathology assumes that Parkinsonian speech/voice abnormalities reflect specific motor dysfunctions of vocal tract structures, giving rise to slowed and undershooting articulatory movements (brady-/hypokinesia). From this perspective, the perceived speech abnormalities of Parkinson's patients have been lumped together into a syndrome termed “hypokinetic dysarthria” (Duffy Reference Duffy2005). Unlike in other cerebral disorders, systematic auditory-perceptual studies and acoustic measurements identified laryngeal signs such as monotonous pitch, reduced loudness, and breathy/harsh voice quality as the most salient abnormalities in PD (Logemann et al. Reference Logemann, Fisher, Boshes and Blonsky1978; Ho et al. Reference Ho, Bradshaw, Iansek and Alfredson1999a; Reference Ho, Iansek and Bradshaw1999b; Skodda et al. Reference Skodda, Rinsche and Schlegel2009; Reference Skodda, Grönheit and Schlegel2011).Footnote ⁹ Imprecise articulation appears, by contrast, to be bound to later stages of the disease. In line with these suggestions, attempts to document impaired orofacial movement execution, especially, hypometric (“undershooting”) gestures during speech production, yielded inconsistent results (Ackermann et al. Reference Ackermann, Hertrich, Daum, Scharf and Spieker1997a). Moreover, a retrospective study based on a large sample of postmortem-confirmed cases found that PD patients predominantly display “hypophonic/monotonous speech,” whereas atypical Parkinsonian disorders (APDs) such as multiple system atrophy or progressive supranuclear palsy result in “imprecise or slurred articulation” (Müller et al. Reference Müller, Wenning, Verny, McKee, Chaudhuri, Jellinger, Poewe and Litvan2001). As a consequence, Müller et al. assume the articulatory deficits of APD to reflect non-dopaminergic dysfunctions of brainstem or cerebellar structures.

Much like early PD, ischemic infarctions restricted to the putamen primarily give rise to hypophonia as the most salient speech motor disorder (Giroud et al. Reference Giroud, Lemesle, Madinier, Billiar and Dumas1997). In its extreme, a more or less complete loss of prosodic modulation of verbal utterances (“expressive or motor aprosodia”) has been observed following cerebrovascular damage to the basal ganglia (Cohen et al. Reference Cohen, Riccio and Flannery1994; Van Lancker Sidtis et al. Reference Van Lancker Sidtis, Pachana, Cummings and Sidtis2006).Footnote ¹⁰ These specific aspects of speech motor disorders in PD or after striatal infarctions suggest a unique role of the basal ganglia in supporting spoken language production in that the resulting dysarthria might primarily reflect a diminished impact of motivational, affective/emotional, and attitudinal states on the execution of speech movements, leading to diminished motor activity at the laryngeal rather than the supralaryngeal level. Similar to other motor domains, thus, the degree of speech deficits in PD appears sensitive to “the emotional state of the patient” (Jankovic Reference Jankovic2008), which, among other things, provides a physiological basis for motivation-related approaches to therapeutic regimens such as the Lee Silverman Voice Treatment (LSVT; Ramig et al. Reference Ramig, Fox and Sapir2004; Reference Ramig, Fox, Sapir, Koller and Melamed2007). This general loss of “motor drive” at the level of the speech motor system and the predominant disruption of emotive speech prosody suggest that the intrusion of emotional/affective tone into the volitional motor mechanisms of speaking depends on a dopaminergic striatal “limbic-motor interface” (Mogenson et al. Reference Mogenson, Jones and Yim1980).

4.3.1. Dopamine-dependent interactions between the limbic and motor loops of the basal ganglia during mature speech production

In mammals, nearly all cortical areas as well as several thalamic nuclei send excitatory, glutamatergic afferents to the striatum. This major input structure of the basal ganglia is assumed to segregate into the caudate-putamen complex, the ventral striatum with the nucleus accumbens as its major constituent, and the striatal elements of the olfactory tubercle (e.g., Voorn et al. Reference Voorn, Vanderschuren, Groenewegen, Robbins and Pennartz2004). Animal experimentation shows these basal ganglia subcomponents to be embedded into a series of parallel reentrant cortico-subcortico-cortical loops (Fig. 3A; Alexander et al. Reference Alexander, Crutcher, DeLong, Uylings, Eden, de Bruin, Corner and Feenstra1990; DeLong & Wichmann Reference DeLong and Wichmann2007; Nakano Reference Nakano2000). Several frontal zones, including primary motor cortex, SMA, and lateral premotor areas, target the putamen, which then projects back via basal ganglia output nuclei and thalamic relay stations to the respective areas of origin (motor circuit). By contrast, cognitive functions relate primarily to connections of prefrontal cortex with the caudate nucleus, and affective states to limbic components of the basal ganglia (ventral striatum). Functional imaging data obtained in humans are consistent with such an at least tripartite division of the basal ganglia (Postuma & Dagher Reference Postuma and Dagher2006) and point to a distinct representation of foot, hand, face, and eye movements within the motor circuit (Gerardin et al. Reference Gerardin, Lehéricy, Pochon, Tézenas du Montcel, Mangin, Poupon, Agid, Le Bihan and Marsault2003). Furthermore, the second basal ganglia output nucleus, the substantia nigra pars reticulata (SNr), projects to several hindbrain “motor centers,” for example, PAG, giving rise to several phylogenetically old subcortical basal ganglia–brainstem–thalamic circuits (McHaffie et al. Reference McHaffie, Stanford, Stein, Coizet and Redgrave2005). A brainstem loop traversing the PAG could participate in the recruitment of phylogenetically ancient vocal brainstem mechanisms during speech production (see sect. 3.1; Hikosaka Reference Hikosaka, Tepper, Abercrombie and Bolam2007).

The suggestion of parallel cortico-basal ganglia–thalamo–cortical circuits does not necessarily imply strict segregation of information flow. To the contrary, connectional links between these networks are assumed to be a basis for integrative data processing (Joel & Weiner Reference Joel and Weiner1994; Nambu Reference Nambu2011; Parent & Hazrati Reference Parent and Hazrati1995). More specifically, antero- and retrograde fiber tracking techniques reveal a cascade of spiraling striato-nigro-striatal circuits, extending from ventromedial (limbic) via central (cognitive-associative) to dorsolateral (motor) components of the striatum (Fig. 3A; e.g., Haber et al. Reference Haber, Fudge and McFarland2000; for reviews, see Haber Reference Haber, Steiner and Tseng2010a; Reference Haber, Iversen, Iversen, Dunnett and Björklund2010b). This dopamine-dependent “cascading interconnectivity” provides a platform for a cross-talk between the different basal ganglia loops and may, therefore, allow emotional/motivational states to impact behavioral responses, including the affective-prosodic shaping of the sound structure of verbal utterances.

The massive cortico- and thalamostriatal glutamatergic (excitatory) projections to the basal ganglia input structures target the GABAergic (inhibitory) medium-sized spiny projection neurons (MSN) of the striatum. MSNs comprise roughly 95% of all the striatal cellular elements. Upon leaving the striatum, the axons of these neurons connect via either the “direct pathway” or the “indirect pathway” to the output nuclei of the basal ganglia (Fig. 3B; Albin et al. Reference Albin, Young and Penney1989; for a recent review, see Gerfen & Surmeier Reference Gerfen and Surmeier2011; for critical comments, see, e.g., Graybiel Reference Graybiel2005; Nambu Reference Nambu2008). In addition, several classes of interneurons and dopaminergic projection neurons impact the MSNs. Dopamine has a modulatory effect on the responsiveness of these cells to glutamatergic input, depending on the receptor subtype involved (David et al. Reference David, Ansseau and Abraini2005; Surmeier et al. Reference Surmeier, Day, Gertler, Chan, Shen, Steiner and Tseng2010a; Reference Surmeier, Day, Gertler, Chan, Shen, Iversen, Iversen, Dunnett and Björklund2010b). Against this background, MSNs must be considered the most pivotal computational units of the basal ganglia that are “optimized for integrating multiple distinct inputs” (Kreitzer & Malenka Reference Kreitzer and Malenka2008), including dopamine-dependent motivation-related information, conveyed via ventromedial–dorsolateral striatal pathways to those neurons. It is well established that midbrain dopaminergic neurons have a pivotal role within the context of classical/Pavlovian and operant/instrumental conditioning tasks (e.g., Schultz Reference Schultz2006; Reference Schultz2010). More specifically, unexpected benefits in association with a stimulus give rise to stereotypic short-latency/short-duration activity bursts of dopaminergic neurons which inform the brain on novel reward opportunities. Whereas, indeed, such brief responses cannot easily account for the impact of a speaker's mood such as anger or joy upon spoken language, other behavioral challenges, for example, longer-lasting changes in motivational state such as “appetite, hunger, satiation, behavioral excitation, aggression, mood, fatigue, desperation,” are assumed to give rise to more prolonged striatal dopamine release (Schultz Reference Schultz2007, p. 207). Moreover, the midbrain dopaminergic system is sensitive to the motivational condition of an animal during instrumental conditioning tasks (“motivation to work for a reward”; Satoh et al. Reference Satoh, Nakai, Sato and Kimura2003).

The dopamine-dependent impact of motivation-related information on MSNs provides a molecular basis for the influence of a speaker's actual mood and actual emotions on the speech control mechanisms bound to the basal ganglia motor loop. Consequently, depletion of striatal dopamine should deprive vocal behavior from the “energetic activation” (Robbins Reference Robbins, Iversen, Iversen, Dunnett and Björklund2010) arising in the various cortical and subcortical limbic structures of the primate brain (see Fig. 1B). The different basic motivational states of our species – shared with other mammals – are bound to distinct cerebral networks (Panksepp Reference Panksepp1998; Reference Panksepp and Brudzynski2010). For example, the “rage/anger” and “fear/anxiety” systems involve the amygdala, which, in turn, targets the ventromedial striatum. On the other hand, the cortico-striatal motor loop is engaged in the control of movement execution, namely, the specification of velocity and range of orofacial and laryngeal muscles. The basal ganglia have an ideal strategic position to translate the various arousal-related mood states (joy or anger) into their respective acoustic signatures by means of a dopaminergic cascade of spiraling striato-nigro-striatal circuits – via adjustments of vocal tract innervation patterns (“psychobiological push effects of vocal affect expression”; Banse & Scherer Reference Banse and Scherer1996; Scherer et al. Reference Scherer, Johnstone, Klasmeyer, Davidson, Scherer and Goldsmith2009). In addition, spoken language may convey a speaker's attitude towards a person or topic (“attitudinal prosody”; Van Lancker Sidtis et al. Reference Van Lancker Sidtis, Pachana, Cummings and Sidtis2006). Such higher-order communicative functions of speech prosody involve a more extensive appraisal of the context of a conversation and may exploit learned stylistic (ritualized) acoustic models of vocal-expressive behavior (Scherer Reference Scherer1986; Scherer et al. Reference Scherer, Johnstone, Klasmeyer, Davidson, Scherer and Goldsmith2009). Besides subcortical limbic structures and orbitofrontal areas, ACC projects to the ventral striatum in monkeys (Haber et al. Reference Haber, Kunishio, Mizobuchi and Lynd-Balta1995; Kunishio & Haber Reference Kunishio and Haber1994; Öngür & Price Reference Öngür and Price2000). Since these mesiofrontal areas are assumed to operate as a platform of motivational-cognitive interactions subserving response evaluation (see above), the connections of ACC with the striatum, conceivably, engage in the implementation of attitudinal aspects of speech prosody (“sociolinguistic/sociocultural pull factors” as opposed to the “psychobiological push effects” referred to above; Banse & Scherer Reference Banse and Scherer1996; Scherer et al. Reference Scherer, Johnstone, Klasmeyer, Davidson, Scherer and Goldsmith2009). Thus, both the psychobiological push and the sociocultural pull effects, ultimately, may converge on the ventral striatum, which then, presumably, funnels this information into the basal ganglia motor loops.

4.3.2. Integration of laryngeal and supralaryngeal articulatory gestures into speech motor programs during speech acquisition

The basal ganglia are involved in the development of stimulus-response associations, for example, Pavlovian conditioning (Schultz Reference Schultz2006), and the acquisition of stimulus-driven behavioral routines, such as habit formation (Wickens et al. Reference Wickens, Horvitz, Costa and Killcross2007). Furthermore, striatal circuits are known to engage in motor skill refinement, another variant of procedural (nondeclarative) learning.Footnote ¹¹ For example, the basal ganglia input nuclei contribute to the development of “motor tricks” such as the control of a running wheel or the preservation of balance in rodents (Dang et al. Reference Dang, Yokoi, Yin, Lovinger, Wang and Li2006; Willuhn & Steiner Reference Willuhn and Steiner2008; Yin et al. Reference Yin, Mulcare, Hilário, Clouse, Holloway, Davis, Hansson, Lovinger and Costa2009). Neuroimaging investigations and clinico-neuropsychological studies suggest that the basal ganglia contribute to motor skill learning in humans as well, though existing data are still ambiguous (e.g., Badgaiyan et al. Reference Badgaiyan, Fischman and Alpert2007; Doya Reference Doya2000; Doyon & Benali Reference Doyon and Benali2005; Kawashima et al. Reference Kawashima, Ueki, Kato, Matsukawa, Mima, Hallett, Ito and Ojika2012; Packard & Knowlton Reference Packard and Knowlton2002; Wu & Hallett Reference Wu and Hallett2005). The clinical observations referred to suggest that bilateral pre-/perinatal damage to the cortico-striatal-thalamic circuits gives rise to severe expressive developmental speech disorders which must be distinguished from the hypokinetic dysarthria syndrome seen in adult-onset basal ganglia disorders. Conceivably, thus, the primary control functions of these nuclei change across different stages of motor skill acquisition. In particular, the basal ganglia may primarily participate in the training phase preceding skill consolidation and automatization: The “engrams” shaping habitual behavior and the “programs” steering skilled movements, thus, may get stored in cortical areas rather than the basal ganglia (for references, see Graybiel Reference Graybiel2008; Groenewegen Reference Groenewegen2003).

Yet, several functional imaging studies of upper-limb movement control failed to document a predominant contribution of the striatum to the early stages of motor sequence learning (Doyon & Benali Reference Doyon and Benali2005; Wu et al. Reference Wu, Kansaku and Hallett2004) or even revealed enhanced activation of the basal ganglia during overlearned task performance (Ungerleider et al. Reference Ungerleider, Doyon and Karni2002) and, therefore, do not support this model. As a caveat, these experimental investigations may not provide an appropriate approach to the understanding of the neural basis of speech motor learning. Spoken language represents an outstanding “motor feat” in that its ontogenetic development starts early after or even prior to birth and extends over more than a decade. During this period, the specific movement patterns of an individual's native idiom are exercised more extensively than any other comparable motor sequences. A case similar to articulate speech can at most be made with educated musicians or athletes who have experienced extensive motor practice from early on over many years. In these subject groups, extended motor learning is known to induce structural adaptations of gray and white matter regions related to the level of motor accomplishments (Bengtsson et al. Reference Bengtsson, Nagy, Skare, Forsman, Forssberg and Ullén2005; Gaser & Schlaug Reference Gaser and Schlaug2003). Such investigations into the mature neuroanatomic network of highly trained “motor experts” have revealed fronto-cortical and cerebellar regionsFootnote ¹² to be predominantly moulded by the effects of long-term motor learning with little or no evidence for any lasting changes at the level of the basal ganglia (e.g., Gaser & Schlaug Reference Gaser and Schlaug2003). Against this background, it might be conjectured that the basal ganglia engage primarily in early stages of speech acquisition but do not house the motor representations that ultimately convey the fast, error-resistant, and highly automated vocal tract movement patterns of adult speech. This may explain why pre-/perinatal dysfunctions of the basal ganglia have a disastrous impact on verbal communication and preclude the acquisition of speech motor skills.

How can the contribution of the basal ganglia to the assembly of vocal tract motor patterns during speech acquisition be delineated in neurophysiological terms? One important facet is that the laryngeal muscles should have gained a larger striatal representation in our species as compared to other primates. Humans are endowed with more extensive corticobulbar fiber systems, including monosynaptic connections, engaged in the control of glottal functions (see sect. 2.2.3 above; Iwatsubo et al. Reference Iwatsubo, Kuzuhara, Kanemitsu, Shimada and Toyokura1990; Kuypers Reference Kuypers1958a). Furthermore, functional imaging data point to a significant primary-motor representation of human internal laryngeal muscles, spatially separated from the frontal “larynx region” of New and Old World monkeys (Brown et al. Reference Brown, Ngan and Liotti2008; Reference Brown, Laird, Pfordresher, Thelen, Turkeltaub and Liotti2009). In contrast to other primates, therefore, a higher number of corticobulbar fibers target the nu. ambiguus. As a consequence, the laryngeal muscles should have a larger striatal representation in our species since the cortico-striatal fiber tracts consist, to a major extent, of axon collaterals of pyramidal tract neurons projecting to the spinal cord and the cranial nerve nuclei, including the nu. ambiguus (Gerfen & Bolam Reference Gerfen, Bolam, Steiner and Tseng2010; Reiner Reference Reiner, Steiner and Tseng2010). Apart from the nu. accumbens, electrical stimulation of striatal loci in monkeys, in fact, failed to elicit vocalizations. In the latter case, however, the observed vocalizations reflect, most presumably, evoked changes in the animals' internal motivational milieu rather than the excitation of motor pathways (Jürgens & Ploog Reference Jürgens and Ploog1970).

A more extensive striatal representation of laryngeal functions can be expected to enhance the coordination of these activities with the movements of supralaryngeal structures. Briefly, the dorsolateral striatum separates into two morphologically identical compartments of MSNs, which vary, however, in neurochemical markers and input/output connectivity (Graybiel Reference Graybiel1990; for recent reviews, see Gerfen Reference Gerfen, Iversen, Iversen, Dunnett and Björklund2010; Gerfen & Bolam Reference Gerfen, Bolam, Steiner and Tseng2010). While the so-called striosomes (patches) are interconnected with limbic structures, the matrisomes (matrix) participate predominantly in sensorimotor functions. This matrix component creates an intricate pattern of divergent/convergent information flow. For example, primary-motor and somatosensory cortical representations of the same body part are connected with the same matrisomes of the ipsilateral putamen (Flaherty & Graybiel Reference Flaherty and Graybiel1993). Conversely, the projections of a single cortical primary-motor or somatosensory area to the basal ganglia appear to “diverge to innervate a set of striatal matrisomes which in turn send outputs that reconverge on small, possibly homologous sites” in pallidal structures further downstream (Flaherty & Graybiel Reference Flaherty and Graybiel1994, p. 608). Apparently, such a temporary segregation and subsequent re-integration of cortico-striatal input facilitates “lateral interactions” between striatal modules and, thereby, enhances sensorimotor learning processes.

Similar to other body parts, it must be expected that the extensive larynx-related cortico-striatal fiber tracts of our species feed into a complex divergence/convergence network within the basal ganglia as well. These lateral interactions between matrisomes bound to the various vocal tract structures might provide the structural basis supporting the early stages of ontogenetic speech acquisition. More specifically, a larger striatal representation of laryngeal muscles – split up into a multitude of matrisomes – could provide a platform for the tight integration of vocal fold movements into the gestural architecture of vocal tract motor patterns (Fig. 2C).