Recent comparative evidence suggests that anthropoid primates are the only vertebrates to exhibit a quantitative relationship between relative brain size and social group size. In this paper, I attempt to explain this pattern with regard to facial expressivity and social bonding. I hypothesize that facial motor control increases as a secondary consequence of neocortical expansion owing to cortical innervation of the facial motor nucleus. This is supported by new analyses demonstrating correlated evolution between relative neocortex size and relative facial nucleus size. I also hypothesize that increased facial motor control correlates with enhanced emotional expressivity, which provides the opportunity for individuals to better gauge the trustworthiness of group members. This is supported by previous evidence from human psychology, as well as new analyses demonstrating a positive relationship between allogrooming and facial nucleus volume. I suggest new approaches to the study of primate facial expressivity in light of these hypotheses.
Relative brain size is correlated with indices of sociality in a wide range of vertebrates [1–4]. However, recent evidence suggests that anthropoid primates (monkeys and apes, including humans) are fundamentally different from other vertebrates with regard to the relationship between relative brain size and social group size. Only anthropoids exhibit a quantitative relationship between these two variables, after controlling for confounding effects [5,6]. Conversely, in non-anthropoid vertebrate taxa, relative brain size tends to be larger in species characterized by monogamous, pair-bonded groups . This distinction might be owing to unique aspects of anthropoid social communication, in particular facial expressions of emotion.
Anthropoids are unique in their reliance on complex facial expressions . The classic example of this is the bared-teeth display (figure 1) [9,10]. This signal is often associated with submissive behaviours when used in agonistic contexts. However, the bared-teeth display can also be used to promote social bonding when used in peaceful contexts. For example, in common chimpanzees (Pan troglodytes), this signal is associated with increased rates of affinitive interaction between sender and receiver for several minutes after it is produced . Similar effects of the bared-teeth display have been observed in a variety of monkeys [12–14]. Another uniquely anthropoid facial expression, the lip-smack display (figure 1), is directly linked to the single most important form of social bonding in primates, allogrooming . Anthropoids use lip-smack displays to initiate and maintain grooming bouts with other group members . Thus, facial communication appears to be a unique component of social bonding in anthropoids.
Most anthropoid species exhibit a small number (less than 10) of facial display types, and repertoire sizes appear to differ very little between species . This is due in part to the overall similarity of the muscles of facial expression across primates [16–21]. In contrast to the apparent conservatism of facial display types in anthropoids, the degree of neural control over the muscles of facial expression appears to be much more variable between species [22–26], and is therefore amenable to comparative analysis.
A link between social bonding and facial expression is supported by the co-evolution of group size and facial motor control. Sherwood et al. were the first to examine this issue in primates . They used facial nucleus volume as a measure of facial motor control because the facial nucleus sends motoneurons from the brainstem to the muscles of facial expression via cranial nerve VII. A larger nucleus implies more neurons and presumably finer motor control. In order to take into account the confounding effects of allometry, Sherwood et al. examined facial nucleus volume in relation to the volume of the medulla, which is the part of the brainstem that houses the orofacial motor nuclei. When they examined the partial correlation between facial nucleus volume and social group size in a broad array of primates, they did not find a significant relationship. However, a subsequent reanalysis focusing on diurnal, group-living anthropoids revealed that the volume of the facial nucleus was positively correlated with group size in catarrhines, after controlling for medulla volume . This group-size effect was absent from platyrrhines (New World monkeys) (figure 2).
In a separate study, I examined the evolutionary relationship between social group size and a behavioural measure of facial motor control: observed facial mobility . To carry out this study, I applied the human facial action coding system [29–31] to video recordings of facial activity in zoo animals representing 12 non-human anthropoids. This approach allows for the enumeration of the number of visually distinct facial movements a species can produce . I found that species that live in larger groups tended to produce a wider range of facial movements, after controlling for body size. Thus, both neuroanatomical and behavioural evidence suggest that group size evolves in correlation with facial motor control, at least in catarrhines [27,28].
The functional significance of enhanced facial motor control can be understood in terms of emotional expressivity, or the frequency, intensity and/or accuracy with which individuals signal their emotional states or intentions. Research in human neuropsychology suggests that expressivity is directly related to facial motor control. For example, patients with Parkinson's disease (PD) tend to be less expressive overall and produce facial displays that are regarded as less intense and more difficult to recognize, compared with healthy people [33,34]. These difficulties are due in part to the facial motor deficits associated with PD. Similar patterns can be found among healthy individuals. For example, people with greater voluntary control over their muscles of facial expression tend to exhibit spontaneous displays that are more intense and easier to recognize than those with lesser degrees of facial motor control [35,36]. Thus, facial motor control appears to influence overall expressiveness, as well as the intensity and/or intelligibility of spontaneous displays of emotion.
In this paper, I explore the possibility that the covariation of group size and relative brain size in anthropoids is explained in part by the unique role of facial expressivity in facilitating social bonding. Towards this end, I will describe two interrelated hypotheses for future study and present preliminary tests of both hypotheses.
2. The cortico-facial complex hypothesis
Although the correlation between relative brain size and social group size is well established in anthropoids, the exact nature of this relationship is open to discussion [5,6,37–42]. One interpretation is that the relative size of the neocortex places an upper limit on the number of social relationships that an individual can establish and maintain . This hypothesis is supported by evidence of positive associations between relative neocortex size and a variety of social network parameters (e.g. grooming clique size) [43–45]. It is argued that the fundamental basis of this constraint lies in the superior socio-cognitive skills afforded by having relatively large neocortices . This view is supported by evidence of positive correlations between relative neocortex size and cognitively complex social behaviours (e.g. tactical deception) [1,46,47]. Moreover, species with relatively large neocortices appear to have higher general intelligence and superior executive brain functions [48,49], which is arguably a prerequisite for complex social networks.
An alternative explanation for the correlation between relative neocortex size and group size is that the former places a constraint on ‘the ability to recognize and interpret visual signals for identifying either individuals or their behaviour’ . This view is supported by two main comparative observations. First, species with relatively large neocortices tend to have relatively large visual brain areas, in particular primary visual cortices (V1) and lateral geniculate nuclei (LGNs) [41,50]. Second, species with relatively large visual brain areas tend to live in large social groups [39,41]. Moreover, V1 volume is correlated with facial nucleus volume in catarrhines, after controlling for the volume of the rest of the brain . Taken together, these findings highlight the potential importance of visual signalling, in particular, facial expressivity, in understanding neocortical expansion.
Indeed, detailed examinations of neural architecture support a link between face processing and neocortical expansion. Within the LGN, there are two main axonal pathways from the retina that are associated with separate parvocellular and magnocellular layers. Neuron density in the parvocellular, but not in the magnocellular, pathway of LGN is positively associated with neocortical expansion in primates . The parvocellular pathway projects to the ventral visual stream, which connects to multiple ‘patches’ of face-selective neurons in the temporal cortex [51–53]. These specialized face cells are part of the core system for face processing, which is responsible for detailed structural analyses of facial displays . The extended system incorporates several other brain regions that are involved in recognizing emotions and/or intentions . Face-selective patches of neurons are also found in the extended system, specifically the prefrontal cortex of the frontal lobe . This is important to note because frontal lobe size is argued to be the main constraint on social network size in anthropoids .
Perhaps the best argument for an evolutionary link between facial expressivity and neocortical expansion is the fact that at least five different areas of the motor cortex connect directly to the facial motor nucleus via descending axons in the corticobulbar tract . However, this point has received little attention with regard to anthropoid brain evolution until now. It is well established that brains evolve in a mosaic fashion [59–62]. In particular, regions that are connected through direct axonal pathways tend to expand (or contract) in correlation with each other, independently of other regions. For example, species with relatively large neocortices tend to have relatively large cerebellar cortices, after controlling for the size of the rest of the brain . This is owing to the fact that the cerebellum is structurally linked to the neocortex via axons passing through the pons. Whiting and Barton  refer to this distributed neural system as the ‘cortico-cerebellar complex’.
Given the substantial neocortical innervation of the facial nucleus , I tested the hypothesis that anthropoids possess a cortico-facial complex by examining the co-evolution of relative neocortex size and relative facial nucleus size. I used a phylogenetic generalized least-squares (GLS) approach [64–66] to generate multiple regressions predicting facial nucleus volume from two independent variables: medulla volume and neocortex ratio (i.e. neocortex volume/total brain volume – neocortex; see the electronic supplementary material for further details regarding the materials and methods). In support of the cortico-facial complex hypothesis, neocortex ratio was a significant predictor of facial nucleus volume, after controlling for medulla volume in catarrhines (b = 1.39, t10 = 3.35, p < 0.01), but not platyrrhines (b = −0.73, t9 = −1.17, p = 0.306). To determine whether or not the relationship between neocortex ratio and facial nucleus volume was owing to the correlation of both variables with V1, I recalculated the neocortex ratio for each species, excluding V1 from the numerator, and reran the regressions. The results were the same (figure 3); non-V1 neocortex ratio was a significant predictor of facial nucleus volume in catarrhines (b = 2.34, t7 = 4.05, p < 0.01), but not platyrrhines (b = −0.48, t8 = −1.70, p = 0.128).
These results support the existence of an evolutionarily coherent cortico-facial complex in Old World monkeys and apes. In other words, catarrhines that have relatively large neocortices, and presumably enhanced socio-cognitive skills, also have greater facial motor control, and hence expressivity. Thus, I hypothesize that selection for enhanced executive brain functions results in greater facial expressivity as a secondary consequence of neocortical expansion. The potential implications of this pattern for understanding the evolution of social bonding are explored in the following sections.
3. The trustworthy face hypothesis
Anthropoids appear to have a unique approach to social bonding. This is reflected in the dependence of monkeys and apes on allogrooming (hereafter, grooming) as a kind of social currency [15,44,67]. Although grooming has undoubted hygienic benefits, anthropoids spend much more time grooming than would be necessary for hygiene alone, up to 20 per cent of a given day in some species . The social value of grooming is supported by evidence of positive associations between grooming and agonistic support at the dyadic level within groups . In some monkey species, low-ranking individuals prefer to groom high-ranking individuals, presumably because individuals near the top of the hierarchy are more effective coalition partners [69–71]. Also, because grooming is pleasurable (i.e. releases endorphins) and reduces stress, it can be exchanged for itself in a reciprocal fashion . A broad view of the social function of grooming is that it provides a psycho-pharmacological mechanism that enables two individuals to build a bonded relationship based on trust [15,73].
The suggestion that anthropoid social bonds are based on ‘a psychological environment of trust’  is intriguing in light of recent evidence from human psychology that facial expressivity is a guide to trustworthiness. Boone and Buck define trustworthiness as the willingness to be conditionally cooperative . Several experimental studies have examined the relationship between facial expressivity and decision-making in the context of social dilemma games (e.g. the iterated Prisoner's Dilemma). The results of these studies suggest that highly expressive people are more likely to be cooperative and/or behave in an altruistic fashion than less expressive individuals [75,76]. Consequently, most people prefer to cooperate with highly expressive individuals because they regard high expressivity as an honest indicator of trustworthiness [77–79].
The relationship between facial expressivity and trustworthiness observed in humans might also be observed in non-human anthropoids. Indeed, this relationship might be essential to grooming. This is because, whether grooming is exchanged in a reciprocal fashion or used to buy favours from higher-ranking individuals, this type of cooperative interaction implies that each individual regards the other as trustworthy. Thus, if high expressivity is an honest indicator of trustworthiness in non-human anthropoids, then species with greater facial motor control, and presumably greater expressivity, should spend more time grooming than less expressive species. As a preliminary test of this hypothesis in catarrhines, I used a phylogenetic GLS approach to generate a multiple regression model predicting facial nucleus volume from grooming time while holding medulla volume constant (see the electronic supplementary material). In support of the trustworthy face hypothesis, grooming time was a significant predictor of facial nucleus volume independent of medulla volume (b = 0.21, t6 = 2.65, p < 0.05; figure 4).
This preliminary finding supports the idea that high facial expressivity promotes cooperative interactions, at least in catarrhines. It is important to note that as the group size increases, anthropoids tend to spend more time grooming relatively fewer individuals [44,67]. In other words, individuals become more selective with regard to the formation of grooming ‘cliques’ in larger groups. Thus, facial expressivity might play a greater role in partner choice during cooperative interactions in larger groups than smaller groups. The co-evolution of facial motor control and group size supports this idea [27,28].
4. Discussion and future directions
I have attempted to explain why anthropoids, and no other vertebrates, exhibit a quantitative relationship between relative brain size and social group size [5–7]. My approach has been to focus on two interrelated aspects of anthropoid neuroethology that have largely been overlooked until now: cortical innervation of the facial motor nucleus and facial cues of trustworthiness. The cortico-facial complex hypothesis posits that facial motor control, and hence expressivity, increases as a secondary consequence of neocortical expansion. The trustworthy face hypothesis suggests that high facial expressivity promotes cooperative interactions, especially in larger groups. Thus, I argue that facial expressivity is an important link between relative brain size and social group size, at least in catarrhines.
I did not find a significant correlation between neocortex ratio and facial nucleus volume in platyrrhines after controlling for medulla volume. This is consistent with previous evidence that facial nucleus volume does not evolve in correlation with social group size or V1 volume in New World monkeys . One potential explanation for this pattern is that the motor cortex does not directly innervate the facial nucleus in platyrrhines. Evidence supporting the existence of cortico-facial projections in New World monkeys is equivocal at best . Interestingly, there is good evidence of cortical innervation of the facial nucleus in rats . This suggests that the cortico-facial complex is not unique to catarrhine primates. Perhaps cortical innervation of the facial nucleus has evolved independently in species with specialized facial behaviours. Many rodents, for example, are known to exhibit a specialized pattern of facial movement referred to as ‘rhythmic whisking,’ which they use to explore the surrounding environment . Similarly, with regard to facial expressivity, catarrhines might be regarded as more specialized than platyrrhines, with few exceptions .
If facial expressivity helps explain the co-evolution of relative brain size and social group size in catarrhines, but not platyrrhines, then the latter should not exhibit a quantitative relationship between neocortex ratio and group size. Figure 4 suggests that this is the case. When group size is plotted against neocortex ratio in platyrrhines, there is a positive trend; species with larger neocortex ratios tend to live in larger social groups. However, this relationship is qualitative rather than quantitative. New World monkeys that live in groups of more than 10 individuals have relatively large neocortices, while the reverse is true of species that live in groups of less than 10 individuals. Within these two sets of species, the relationship between group size and neocortex ratio is decoupled (figure 5). These observations imply a fundamentally different approach to social bonding in platyrrhines versus catarrhines.
The cortico-facial complex hypothesis might help explain why orangutans (Pongo pygmaeus) are extreme outliers with regard to the relationship between facial motor control and social group size. Orangutans have the largest facial motor nuclei relative to medulla volume of any primate . Yet they are essentially solitary and spend less time grooming than any other anthropoid . Perhaps orangutans have relatively large neocortices for purely ecological, rather than social, reasons. Indeed, orangutans are the only non-human primates, other than chimpanzees, to regularly manufacture complex and geographically variable tool kits in the wild . This implies strong selection for enhanced executive brain functions, which might have resulted in the relative enlargement of the facial motor nucleus, and hence facial mobility, as a secondary consequence. The implication is that orangutan faces are capable of providing cues to trustworthiness. However, there is little need for such information in wild orangutan societies.
Most behavioural studies of primate facial expressions focus on the context-specific meaning(s) of facial displays [14,16,83,84]. The unit of analysis in these studies is usually the display itself, not the individual. This type of approach might not be suitable for behavioural studies of emotional expressivity, if we regard expressivity as a stable dispositional quality of an individual . In other words, facial expressivity might reflect a context-independent pattern of expression characteristic of an individual or species. As such, expressivity could be measured as overall expressiveness, i.e. the frequency of displays emitted across all contexts. Alternatively, it might be better to focus on individual differences in the visual appearance of any given display. For example, Bard et al.  have suggested that not all chimpanzee bared-teeth displays look the same, although they may be functionally equivalent. If bared-teeth display structure is consistent within and variable between individuals, then this variation might reflect individual differences in expressivity. Future studies of expressivity should focus on the individuality of facial displays, rather than the typical structure .
The trustworthy face hypothesis makes several predictions that can be tested at the behavioural level, given a method for measuring individual expressivity. First and foremost, the hypothesis predicts that highly expressive individuals are more likely to be cooperative than less expressive individuals. For example, in species where grooming is primarily exchanged for itself, expressivity should be positively correlated with reciprocity, after controlling for confounding effects such as kinship. Also, if facial expressivity is a guide to trustworthiness, then we might also expect to observe correlations between expressivity and individual measures of sociality derived from social network analysis . For example, highly expressive individuals might be more central to the group than less expressive individuals.
A potentially promising area of future research on facial expressivity relates to the link between serotonin and social bonding. Serotonin is a monoamine neurotransmitter biochemically derived from tryptophan. In a series of experiments on captive vervet monkeys (Chlorocebus aethiops), Raleigh and colleagues injected a select number of individuals with tryptophan [88–90]. They found that these individuals were more likely to engage in allogrooming after receiving tryptophan, presumably owing to enhanced serotonergic activity. Similar effects of serotonin on grooming have been observed in other monkey species . However, the mechanisms underlying these correlations remain obscure. It is interesting to note that very high densities of serotonin (5-HT) receptors are present in the facial motor nucleus . Intracellular studies of the facial nucleus demonstrate that the responsiveness of facial motoneurons to excitatory inputs is modulated in part by serotonin . Moreover, the modulatory effect of serotonin can be observed in patterns of facial movement, as in the case of the rhythmic whisking of rats . Taken together, these observations suggest the possibility that facial expressivity is the causal link between serotonin and grooming. It might be the case that individuals with enhanced serotonergic activity are more likely to engage in grooming because group members perceive them as more trustworthy owing to heightened expressivity. These predictions can be tested at the individual level, given data on facial expressivity and serotonergic activity and/or serotonin transporter polymorphisms.
Anthropoids are highly social, visually oriented and cognitively complex animals. But none of these attributes alone is unique to Anthropoidea. What is unique is the vast amount of information conveyed through the anthropoid face. Facial expressions convey the immediate emotional intentions of the sender in any given situation, but they might also indicate the long-term cooperative disposition of the sender. Thus, to fully understand the integration of anthropoid social, visual and cognitive adaptations, these traits need to be viewed through the lens of facial expressivity.
I would like to thank Todd Freeberg for inviting me to contribute to this special issue, as well as the associated symposium. I also thank Chet Sherwood and Lauren Brent for commenting on a draft of this manuscript. Last but not least, I am grateful to the efforts of three anonymous reviewers, who provided constructive criticisms that significantly enhanced my paper.
One contribution of 13 to a Theme Issue ‘The social network and communicative complexity in animals’.
- This journal is © 2012 The Royal Society