Introduction

Speech is a highly skilled behavior, which requires the synchronized movement of the respiratory, laryngeal, and orofacial muscles1. In contrast to most primate calls, which are thought to be innate and genetically determined2, speech is acquired through vocal learning and imitation of the sounds of conspecifics. In the brain, a phylogenetically older system for innate vocalizations and a more recent one for speech in humans have been described1,3. However, little is known concerning reorganization of the primate brain to accommodate the neural substrates underlying a learned, flexible vocal communication system.

Attempts to answer this question have proven challenging not only because fossil evidence relevant to understanding extinct hominin communication skills are scarce and difficult to interpret, but also because studies of nonhuman primate calls have shown only limited evidence of vocal learning2. As more evidence of primate communication accumulates, however, it is becoming apparent that vocal production is more flexible than it had been previously appreciated4.

For instance, it has been reported that, in captivity, the calls of marmosets caged together eventually developed convergence of their acoustical properties5. Similarly, a recent study has claimed that the acoustic structure of referential food grunts of two distinct populations of chimpanzees converged over the course of three years as a result of living together in the same zoo6, though some have questioned the interpretation of these findings7. Individual cases of great apes imitating sounds produced by a human (e.g., whistling), and controlling vocal folds movements to produce idiosyncratic communicative sounds (i.e., wookie), have also been reported8,9. Instances of vocal flexibility in nonhuman primates have also been documented in the wild. For example, distinct features in the calls of geographically contiguous populations of chimpanzees have been deemed to function as vocal “social markers” for differentiating neighboring communities10. There is also evidence that wild chimpanzees use alarm calls or hoots to inform ignorant conspecifics about the presence of a snake, suggesting the these sounds are produced intentionally and voluntarily11,12,13. Taken together, these studies suggest that some forms of learned vocal signals, possibly precursors to the extreme vocal flexibility of human speech, may be present in chimpanzees and related ape species. However, the putative neural bases of vocal, flexible communication in the nonhuman primate brain remain largely unstudied14,15.

The current study examined differences in grey matter distribution in a population of captive chimpanzees varying in their capacity to voluntary produce sounds (e.g. extended grunts, raspberries) to attract the attention of an otherwise inattentive human experimenter (attention-getting sounds — AG). Previous studies on these animals have shown that, in the presence of food, chimpanzees are more likely to produce AG sounds when a human experimenter is facing away from them compared to conditions in which the experimenter is oriented towards them16. Additional studies have confirmed that the likelihood of producing an AG sound when a human and food are near the focal subject is higher compared to conditions in which either only the food or human is present17. Thus, it appears that the production of AG sounds is under voluntary control, and is selectively used to capture the attention of an otherwise inattentive audience18.

Crucially, considerable individual differences in the use of AG sounds have been reported, with data showing that about 50% of the sample reliably produce these sounds while the remaining do not. While the reasons underlying these individual differences are unclear, previous research has shown that AG sounds may be socially learned from the mother, as higher rates of concordance in their production have been found between mother-offspring diads compared to chimpanzees raised in a different social group. Other factors, including age and housing conditions, do not appear to influence the production and use of AG sounds19,20.

Drawing on these intrinsic individual differences in communicative capacities, a voxel-based morphometry (VBM) analysis was conducted21,22,23 to assess variation in grey matter distribution in the chimpanzee brain as a function of vocal capacities. VBM is a structural neuroimaging technique that makes use of non-invasive magnetic resonance images (MRI) collected in vivo, and provides a whole-brain, unbiased, spatial analysis of localized differences in grey matter distribution between groups of interest. For this study we preferred a whole-brain approach to a ROI-based approach because of the limited amount of information on regions involved in the production of this atypical form of communication in chimpanzees14,15.

Results

An independent sample t-test assessing differences in grey matter distribution as a function of vocal phenotype in the VBM analysis revealed that chimpanzees that produce attention-getting sounds (AG+) had greater grey matter density than chimpanzees that failed to exhibit this behavior (AG−) in the right ventrolateral prefrontal cortex (VLPFC), bordering the anterior bank of the precentral gyrus, and the left dorsal premotor cortex (dPMC) (uncorrected p < 0.005) (Table 1 and Fig. 1).

Table 1 VBM results AG+ > AG− including the coordinates of reported clusters and t statistics at p-value = 0.005 (uncorrected).
Full size table
Figure 1
figure 1

Coronal and sagittal views of the clusters showing greater grey matter density in AG+ relative to AG− chimpanzees.

(A) ventrolateral prefrontal cortex; (B) dorsal premotor cortex. Clusters are superimposed onto the standard chimpanzee template.

Full size image

The VLPFC of chimpanzees has been cytoarchitecturally mapped previously, demonstrating the localization of areas 44 and 45 on the inferior frontal gyrus24,25. The location of the ventrolateral prefrontal cluster in the current analysis was compared with that of area 44, as defined by a probabilistic map of this area in chimpanzees24. The cluster identified by the VBM fell within the most posterior border of area 44 mapped from a sample of 12 individuals by Schenker et al.24. Because area boundaries are known to be subject to individual variability, a more restricted map including only the portion of this region shared by at least 50% of individuals was then used. Under this more restrictive criterion, the cluster identified by the VBM fell outside the boundaries of area 44, in a location adjacent to the border with ventral premotor areas along the anterior bank of the precentral gyrus, which have been shown to elicit movements of the lips, mouth and vocal cords in chimpanzees26. The second cluster was located in the dorsal precentral gyrus, in the caudal part of the dorsal premotor cortex (dPMC).

Discussion

The current study provided a whole-brain, unbiased approach to examine differences in grey matter distribution in chimpanzees that vary in their voluntarily production of attention-getting (AG) sounds in order to obtain the attention of an otherwise inattentive human experimenter. While AG sounds in this population of chimpanzees have been extensively described at the behavioral level17,27,28, less is known regarding their underlying neural correlates14,15,29. A previous PET functional neuroimaging study examining a small subset of AG+ and AG− individuals identified a number of cortical and subcortical regions that showed increased activation in association with the production of vocal and manual communicative gestures relative to a control task (i.e., grasping). These areas included the left inferior frontal gyrus, left pre- and postcentral gyrus, right parietal and cerebellar areas, and the left striatum (caudate nucleus and putamen)15.

A follow-up study expanding the sample size to four individuals confirmed a key role of the homologue of Broca’s area in chimpanzee communication, by showing that activation of this area was greater for the production of AG sounds and manual communicative gestures than a resting state condition in AG+ but not AG− chimpanzees14. At the structural level, differences in grey matter proportions have been found between chimpanzees that perform differently at a joint attention task in the anterior cingulate cortex30, an area that is involved in the voluntary control of vocalizations1,31, and is part of the attention system that may underlie the socio-cognitive component of these communicative gestures32. Consistent with previous findings, the VBM analysis performed in the current study revealed differences in grey matter distribution between AG+ and AG− in the ventral prefrontal region and premotor areas.

Given the little information currently available on the mechanisms underlying individual differences in chimpanzee vocal capacities, the possibility that motivational factors may influence the production and use of AG sounds cannot be entirely ruled out. Nonetheless, by showing structural differences, our results indicate that variation in the neural organization may indeed distinguish between AG+ and AG− chimpanzees. Moreover, as most MRI scans were collected within 5 years after initial vocal phenotype assessment, the persistence of these links between AG production and brain morphology are especially striking considering the possibility of age-related and other plastic changes in neural structure over time.

A cluster showing increased grey matter density in AG+ relative to AG− was found in the right VLPFC. Within this region, both area 44 (pars opercularis of Broca’s area) and ventral premotor cortex are key regions in the control of orofacial movements and speech processes. In humans, lesions to area 44 and the surrounding regions in the inferior frontal gyrus have been associated with forms of aphasia characterized by lack of speech fluency and agrammatism33,34,35. Functional neuroimaging studies have described activation of the inferior frontal gyrus in a broad variety of language tasks including word retrieval, articulation, and semantic and syntactic processes36,37,38.

Electrophysiological stimulations of the homologue of this area in the ventral part of the premotor cortex of macaque monkeys elicited movements of the jaw and lips, but not vocalizations39. Accordingly, earlier studies had ruled out a role for Broca’s area homologue in nonhuman primate vocalizations40. As suggested by recent studies, however, this area may be involved in the planning of voluntary, goal-directed, conditioned vocalizations in rhesus macaques41. In support to this view, functional studies of communicative gestures in chimpanzees indicated that the inferior frontal gyrus was associated with voluntary forms of communication, including AG sounds14,15. Functional, structural, and lesion studies have claimed that the inferior frontal gyrus, and in particular area 44 in the left hemisphere, regulates language-production functions. The cluster identified by the VBM, however, was located in the posterior part of the right VLPFC. In fact, although motor control of speech is typically lateralized in the left hemisphere42, sensorimotor control for the execution of speech and non-speech orofacial gestures has been described bilaterally43. Furthermore, although a linguistic role of Broca’s area has been emphasized, studies have also indicated a more general function of this area in regulating processing of hierarchical, motor sequences of behavior, regardless of their linguistic nature. Crucially, tasks involving hierarchical processing of learned motor sequences have been shown to activate both Broca’s area and its homologue in the right hemisphere44.

It should also be noted that in association with language functions, the right hemisphere is involved in emotional, non-propositional, pragmatic45, as well as prosodic (i.e., intonational) aspects of speech46. Since AG sounds are produced in social contexts that require some understanding of others’ intentions17,30, it is possible that a rightward increase in grey matter in AG+ relative to AG− chimpanzees reflects intrinsic differences in socio-cognitive communicative processes. Indeed, a rightward functional activation in association with the perception of conspecific calls has been described in a number of areas, including the right inferior frontal gyrus in chimpanzees29.

The second cluster observed in the AG+ > AG− contrast was located in the left dorsal precentral gyrus, in the caudal part of the dorsal premotor cortex (dPMC). The dPMC is extensively connected with the primary motor areas47,48, and plays a role in motor control in relation to the preparation and planning of movements in humans as well as nonhuman primates49,50,51. In particular, this region appears to regulate visuo-motor processes guided by arbitrary cues, a capacity that may underlie the evolution of greater flexibility over motor control by “permitting the choice of an action based on the prevailing context” rather than on spatial constraints52. Because arbitrary mapping is a key property of language (e.g., sound-meaning), it has been suggested that control over learned guided actions may have provided one motor preadaptation for language52.

Accordingly, a recently proposed model of language processing posits that dorsal premotor areas are part of a dorsal system of language production that provides a phonological-articulatory interface53, and potentially, an integration area for vocal learning. Interestingly, a recent study measuring activity-induced cFos expression in marmosets found that stronger labeling of the dorsal prefrontal cortex was associated with call production, while expression in the ventral premotor areas was associated with auditory perception of conspecifics calls54.

While a role for dPMC in the evolution of motor control for language is plausible, reasons underlying purported differences in grey matter distribution of this area between AG+ and AG− chimpanzees should be further assessed. Indeed, the cluster identified by the VBM was located slightly dorsal to the “knob”, or hand region of the precentral gyrus55. As such, this location is compatible with an area adjacent to the motor somatotopic representation of the upper limbs26,56. Numerous lesion, electrophysiological, and functional neuroimaging studies in macaque monkeys as well as humans have shown that the dorsal premotor cortex is a key region underlying motor control of the arm, including movements directed towards a target, such as grasping and reaching51,57. This may be relevant because in combination with vocal signals like AG sounds, chimpanzees have been shown to use a number of communicative manual gestures, such as extending their arm in a food begging gesture, and pointing with either the open hand or the finger to attract the attention of the experimenter19,58.

Furthermore, although most chimpanzees engage in manual gestures to some extent, those producing AG sounds appear to more often throw objects to attract the attention of the experimenter59. Accordingly, a study comparing the neuroanatomical and cognitive correlates of throwing capacity in chimpanzees indicated that individuals that throw objects are characterized by greater white matter to grey matter volume in the hand ‘knob’ region compared to those that engage less in this behavior59. Crucially, “throwers” also showed increased socio-communicative skills as compared to “non-throwers”, while no differences in measures of physical cognition distinguished the two groups59.

Thus, it is possible that a trend towards grey matter differences in the dorsal premotor cortex of AG+ and AG− chimpanzees reflects intrinsic motor capacities for gestural movements of the arm and hand that are co-produced with AG sounds, such as manual gestures, throwing, clapping, and using objects (e.g., making noises) in order to attract the attention of the experimenter. A left lateralization of the results is also consistent with previous reports of population level right-handedness of chimpanzees for gestural communication and throwing59,60. Whether and how vocal and gestural capacities interact to produce greater flexibility for communication in chimpanzees, however, remains to be ascertained.

Our findings revealed localized increases in grey matter in prefrontal and premotor areas, thus suggesting that intrinsic differences in the capacity to communicate flexibly may be associated with variation in the plasticity of areas of the cerebral cortex supporting motor control of orofacial and gestural movements.

Comparative studies have shown that relative to Old World monkeys, the great apes have increased neuropil space for neuronal integration in the orofacial representation of the primary motor cortex, possibly as a result of selection for increased control of the orofacial communicative gestures61. Indeed, the great apes produce many facial expressions in communicative contexts, which may be associated with vocalizations, but also often lack in vocal content (e.g., pouting)62.

If and to what extent human language capacities started in a vocal or gestural modality is a long-standing, and hotly debated question63,64. By showing that flexible communication in chimpanzees involves morphological differences in cortical regions important for both vocal/orofacial movements as well as manual gestures, our results encourage further study of the co-evolution of multi-modal communication and the brain. How brain reorganization in human evolution led to the acoustic-vocal modality of speech as the preferred, while not only, form of communication of our species, remains to be fully understood.

Methods

Sample

All the chimpanzees examined in the current study were housed at the Yerkes National Primate Research Center (YNRC) in Atlanta, GA, USA, in accordance with institutional guidelines. Most of these individuals had taken part in previous behavioral studies on tool-use, handedness, and communication (e.g. refs 14,20,28,65 and 66).

The current study included sixty-nine chimpanzees, of which 34 were classified as individuals that did not produce attention getting sounds (AG−) and 35 as individuals that produced attention-getting sounds (AG+) (see below for a description of their vocal assessment). The AG− group was comprised of 7 males, and 27 females, while the AG+ group consisted of 15 males, and 20 females. All individuals were adults, and no significant differences were found between AG groups as function of age (t1,67 = 1.31, p = 0.19).

Behavioral assessment of vocal phenotype

Chimpanzees were grouped according to their vocal phenotype, where AG+ indicated individuals that produce attention-getting sounds, and AG− denoted individuals who did NOT reliably show this behavior. The procedure for vocal assessment has been described elsewhere17,20,28. Briefly, each chimpanzee received 6, 30-sec test trials. During each test trial, food was placed outside of the subjects’ home enclosure and a human experimenter would approach the subject. On three trials, the experimenter was oriented away from the chimpanzees while on the remaining 3 trials the human was oriented toward the subject. All the communicative behaviors directed toward the experimenter in each condition were recorded. AG sounds were defined as audible sounds that are distinct from the naturally occurring chimpanzee calls (e.g., food calls, alarms28), and included “raspberries”, extended grunts and other idiosyncratic sounds. If the chimpanzees produced at least one AG sound during any of the 6 trials, they were classified as AG+ while all others were classified as AG−.

MRI scanning procedure and preprocessing

Magnetic Resonance Image (MRI) data acquisition was performed at the imaging facilities of the YNPRC following institutional guidelines. The procedure has been described in detail elsewhere (e.g. ref. 67). Briefly, each individual was first immobilized with a telazol injection (2–6 mg/kg), and then anesthetized with propofol (10 mg/kg/h). The total time, including scan acquisition (40–50 mins) and the transport to and from the MRI facility, was about 2 hours. A 3.0 Tesla scanner (Siemens Trio; Siemens Medical Solutions Inc., Malvern, PA, USA) was used. After the scanning was concluded, the chimpanzees were housed in a separate cage for 6–12 hours until while they recovered from the anesthesia.

MRI scans were preprocessed using FSL software and included skull stripping and image normalization using the BET function (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/BET). MRIs of the individuals examined in the current study have been previously employed in studies investigating structural features of the chimpanzee brain (e.g. ref. 67), and their correlations with known behavioral traits such as handedness (e.g. refs 30 and 59).

VBM analysis

Whole-brain comparisons between AG− and AG+ chimpanzees were performed on MRI structural data with the VBM tool available in FSL software (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM). An optimized VBM protocol22 adapted to the chimpanzee brain was used. Structural images were first brain-extracted, and grey matter-segmented. All native grey matter images were then non-linearly registered to a chimpanzee standard template68, and modulated to correct for local expansion (or contraction) due to the non-linear component of the spatial transformation (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM). The modulated grey matter images were then smoothed with an isotropic Gaussian kernel with a sigma of 2 mm, FHWM = ~4.7 mm. To assess group-level differences between AG+ and AG− chimpanzees, voxel-wise GLM was applied using permutation-based non-parametric testing, with a TFCE cluster-based method (fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM). Results from the comparisons between AG+ and AG− chimpanzees were thresholded at p = 0.005 (uncorrected), and only clusters with t > 2.00 and a size >20 voxels were reported. In the analyses, sex, age and overall brain volume were included as covariates, to rule out effects of inter-individual differences due to potentially confounding variables. The data of this study will be made available at chimpanzeebrain.org.

Additional Information

How to cite this article: Bianchi, S. et al. Neocortical grey matter distribution underlying voluntary, flexible vocalizations in chimpanzees. Sci. Rep. 6, 34733; doi: 10.1038/srep34733 (2016).