De Brazza's monkeys (Cercopithecus neglectus), Campbell's monkeys (Cercopithecus campbelli) and red-capped mangabeys (Cercocebus torquatus) are closely-related (Cercopithecidae family, Cercopithecinae subfamily: Grubb et al., 2003) and they originate from comparable African tropical forests (Gautier-Hion et al., 1999; Matthews and Matthews, 2002; Buzzard, 2006). However, they differ strikingly when it comes to their social structure. De Brazza's monkey live in small family groups (including generally one but sometimes up to three adult females), average group size ranging from three to seven individuals in the literature (Quris, 1976; Gautier-Hion and Gautier, 1978; Brennan, 1985; Decker, 1995; Mugambi et al., 1997; Mwenja, 2006; King, 2008). Campbell's monkeys live in harem groups (including three to eight adult females) whose average size ranges from seven to 13 individuals (Harding, 1984; Galat and Galat-Luong, 1985; Buzzard, 2006; Ouattara et al., 2009b). Red-capped mangabeys live in relatively large multi-male multi-female groups that often split into smaller foraging groups whose average group size ranges from 19 to 21 (Mitani, 1989; Matthews and Matthews, 2002). In the three species, females are the “social core” of the group, whereas males act mostly as group protectors (Gautier-Hion and Gautier, 1978; Oswald and Lockard, 1980; Rowell, 1988; Lemasson et al., 2006; Ouattara et al., 2009b). The two guenons also differ from the mangabey species regarding social organization. Like most forest guenons, De Brazza's and Campbell's monkeys' social organization is based on rare physical interactions and a discrete hierarchy (Gautier-Hion and Gautier, 1978; Rowell, 1988; Lemasson et al., 2006). Conversely, the social organization of red-capped mangabeys, like most baboons and macaques, is based on relatively frequent peaceful and agonistic interactions and a strong hierarchy (Rowell, 1988; Dolado and Beltran, 2012).

Call recordings were conducted with five groups of De Brazza's monkeys, one group of Campbell's monkeys and three groups of red-capped mangabeys housed in indoor-outdoor enclosures enriched with either perches, ropes, shrubs or trees, at the Station Biologique de Paimpont in France, and at Howletts and Port Lympne Wild Animal Parks in the UK (Table 1—for more details on the housing conditions, see Bouchet et al., 2010, 2012b; Lemasson and Hausberger, 2011). Cumulatively, we collected data from twelve De Brazza's monkey, six Campbell's monkey, and seven red-capped mangabey adult females.

Subjects were provided with fruit, vegetables, and commercial monkey chow daily. Water was available ad-libitum. Animal care and research protocols used in this work complied with the current laws of France and the UK and were conducted under license from the Direction Départementale des Services Vétérinaires (DDSV n° 04672).

Observations were performed outdoors (to ensure high quality recordings suitable for acoustic analysis) between 09:00 and 18:00, outside of feeding times. Subjects were observed in a random order using the focal sampling method, sessions being distributed throughout the day for every subject. In De Brazza's monkeys and red-capped mangabeys, calls of focal (and occasionally of non-focal) individuals were recorded using a directional microphone (Sony ECM-672 for the females housed in Paimpont, Sennheiser MKH-70 for the females housed in Howletts and Port Lympne) connected to a digital stereo recorder (Sony DAT TCD-D100 for the De Brazza's monkeys housed in Paimpont, Marantz PMD-660 for the other groups). In guenons, females frequently produce soft sounds with their mouths closed, which made it difficult to identify the caller within a multi-female group, even in captivity. Thus, sound recordings in Campbell's monkeys were performed through a telemetric system composed of a transmitter (fixed on a leather harness on the monkey), a receiver, and a digital stereo recorder (TASCAM DA-P1) (for technical details, see Lemasson and Hausberger, 2011).

In total, 37 h50 of focal observations (3 h47 ± 1 h23 per individual, N = 10) were made on De Brazza's monkeys (plus 8 h15 ad-libitum on two adult females at Port Lympne, see Table 1), 92 h20 of telemetric recordings (15 h11 ± 4 h10 per individual, N = 6) were performed with Campbell's monkeys, and 34 h00 of focal observations (4 h50 ± 0 h16 per individual, N = 7) were conducted on red-capped mangabeys. We were able to extract from these recordings 1569 De Brazza's monkey calls (66–336 per individual), 1309 Campbell's monkey calls (99–466 per individual) and 3970 red-capped mangabey calls (163–939 per individual).

The overall level of structural intra-individual variability as well as the level of individual distinctiveness varied among the three functionally different call types (contact, threat and alarm; Figure 1) across the three species (Figure 2).

The level of intra-individual variation in the call acoustic properties did not differ significantly between call types within a species' repertoire (Friedman tests: De Brazza's monkeys: χ²₂ = 0, P = 1; Campbell's monkeys: χ²₂ = 2.80, P = 0.247; red-capped mangabeys: χ²₂ = 4.80, P = 0.091) (Figure 2). However, it significantly differed between species for the threat call; female red-capped mangabeys produced the most variable threat call pattern (Friedman test: χ²₂ = 7.60, P = 0.022; notably because red-capped mangabey threat calls were significantly more variable than those of Campbell's monkeys, post-hoc multiple comparisons: P < 0.05) (Figure 2). No inter-specific differences were observed for the two other call types (Friedman tests: alarm calls: χ²₂ = 3.60, P = 0.165; contact calls: χ²₂ = 4.80, P = 0.091) (Figure 2).

Interestingly, the level of inter-individual variation did not correlate with the level of intra-individual variation found in the various call types. In the three species, the potential for identity coding was the highest in contact calls (highest ΣHs values) (Figure 2). However, while red-capped mangabey threat calls displayed a higher potential for identity coding (higher ΣHs value) than their alarm calls, the two guenon species displayed the opposite pattern (Figure 2). Regarding the ability to encode individual identity of the various acoustic parameters, it appeared that the number of units per call was low-informative (Hs < 1 in all cases but Campbell's monkeys' alarm calls) while the frequency parameters (base and/or peak frequencies) were informative regarding identity in most cases (one or both frequency parameters had a Hs > 1 in each call type, except threat calls in De Brazza's monkeys and contact calls in red-capped mangabeys) (Figure 2). Conversely to frequency parameters, no systematic pattern of variability was found across species for temporal features (call and unit duration), except that unit duration was informative (Hs > 1) for contact calls in the three species (Figure 2).

Red-capped mangabeys and the two guenon species differed in their global vocal activity, the former calling more than twice as much as the two others (Kruskall-Wallis test: H₂ = 5.96, P = 0.051). The post-hoc one-tailed pairwise comparisons (chosen according to the “vocal grooming” hypothesis) revealed that call rates were significantly higher in red-capped mangabeys than Campbell's monkeys (P < 0.05), were higher in red-capped mangabeys than De Brazza's monkeys, although not significantly so (P < 0.075), and did not differ between the two guenon species (P > 0.15). However, we noted that one De Brazza's monkey female (102.5 calls/h) called ten times as much as the nine other conspecific females (X¯ ± SE: 15.3 ± 3.6 calls/h; range: 4.8–38.5). Therefore, we excluded this outlier (Grubb's test: Z = 2.673, N = 10) and reran the analyses (Figure 3). The global test result did not change (Kruskall-Wallis test: H₂ = 7.75, P = 0.021), but the post-hoc one-tailed pairwise comparisons revealed that while call rates remained significantly higher in red-capped mangabeys than Campbell's monkeys (P < 0.05), the difference between red-capped mangabeys and De Brazza's monkeys was now significant as well (P < 0.025). We still did not observe a difference between the two guenon species (P > 0.15).

In the three species, contact calls dominated the females' vocal repertoire, accounting for more than 60% of the total vocal production. Also, red-capped mangabeys produced more contact calls than the two guenon species (X¯ ± SE: red-capped mangabeys “Ro+”: 29.8 ± 9.2 calls/h; Campbell's monkeys “CH”: 8.6 ± 2.8 calls/h; De Brazza's monkeys “On”: 19.4 ± 9.5 calls/h; De Brazza's monkeys “On” without the outlier: 10.4 ± 3.2 calls/h). In addition, red-capped mangabeys uttered threat calls more than three times as much as the two guenon species (red-capped mangabeys “Un+(Uh)”: 7.7 ± 4.5 calls/h; Campbell's monkeys “RRC”: 0.2 ± 0.1 calls/h; De Brazza's monkeys “Wrr+”: 2.4 ± 0.8 calls/h; De Brazza's monkeys “Wrr+” without the outlier: 2.5 ± 0.9 calls/h).

When assessing the structural acoustic variability of each species' at the sound unit level, we identified 6 unit types in De Brazza's monkeys, 8 unit types in Campbell's monkeys, and 9 unit types in red-capped mangabeys (Figure 4). Those units were then concatenated into 9, 10 and 16 different “unit assembling patterns” respectively. Interestingly, the number of “combined” patterns was the lowest in De Brazza's monkeys (N = 1), intermediate in Campbell's monkeys (N = 3), and the highest in red-capped mangabeys (N = 6) (Figure 4).

The diversity index DI enabled us to estimate the structural complexity of each species' vocal repertoire in terms of “unit assembling pattern” types (Figure 4). As a result, it appeared that female De Brazza's monkeys displayed the less diverse repertoire (DI = 0.33), female red-capped mangabeys the most diverse (DI = 0.67), while female Campbell's monkeys showed an intermediate pattern (DI = 0.52). Indeed, the De Brazza's monkeys' repertoire was largely dominated by the contact call in its “single” form (80.6%), while that of Campbell's monkeys was dominated by the contact call in its “combined” pattern (60.8%). In contrast, the repertoire of red-capped mangabeys was dominated by the contact call in its “repeated” form (43.7%), followed by its “single” form (20.8%), but it was also largely composed of many different “unit assembling patterns” in various proportions (range: 0.5–8.3%) (Figure 4).

When performing intra-specific comparisons regarding pattern categories (Figure 5), we found that De Brazza's monkeys produced mostly “single” patterns (81.0%; Friedman tests: χ²₂ = 14.00, P = 0.001; notably because “single” patterns were more frequent than “combined” patterns, post-hoc multiple comparisons: P < 0.05), while red-capped mangabeys uttered mostly “repeated” patterns (56.4%; Friedman tests: χ²₂ = 11.14, P = 0.004; “repeated” were more frequent than both “single” and “combined” patterns, post-hoc multiple comparisons: P < 0.05). No significant preference was found for Campbell's monkeys (Friedman tests: χ²₂ = 4.00, P = 0.135) (Figure 5).

Furthermore, we found that “single” patterns were more frequently uttered by both De Brazza's monkeys and red-capped mangabeys than Campbell's monkeys (Kruskall-Wallis tests: H₂ = 7.48, P = 0.024). “Repeated” patterns were above all produced by red-capped mangabeys (H₂ = 13.02, P = 0.001). Lastly, “combined” patterns were more frequently uttered by both Campbell's monkeys and red-capped mangabeys (H₂ = 16.03, P = 0.001) (see Figure 5 for the results of the post-hoc multiple comparisons).

It has to be acknowledged that the acoustic variability evidenced in this study, both at the call and the repertoire levels, remains to be tested for its relevance to listeners. Nonetheless, many studies have shown that non-human primates are able to discriminate between individuals from their calls (e.g., rhesus macaques, Macaca mulatta: Rendall et al., 1996; olive baboons, Papio hamadryas anubis: Lemasson et al., 2008; chacma baboons, Papio hamadryas ursinus: Rendall et al., 2009), can perceive subtle differences in acoustic structure (e.g., Campbell's monkeys: Lemasson et al., 2005), and can distinguish between sequences of sound units (e.g., cotton-top tamarins, Saguinus oedipus: Hauser et al., 2001; Diana monkeys, Cercopithecus diana: Zuberbühler, 2002; white-handed gibbons, Hylobates lar: Clarke et al., 2006; putty-nosed monkeys, Cercopithecus nictitans: Arnold and Zuberbühler, 2008; Campbell's monkeys: Lemasson et al., 2011a). Such evidence allows us to hypothesize that the acoustic variability observed here is perceptible to monkeys; future research must, however, determine its actual salience to listeners.

In accordance with the “call social function” hypothesis (Snowdon et al., 1997; Griebel and Oller, 2008), we found the highest level of individual distinctiveness in contact calls, compared with alarm calls, in the three studied species. This is in agreement with the few studies in non-human primates and birds that have provided evidence of a higher level of individual distinctiveness in affiliative calls compared with distress, alarm, or courtship calls (Rendall et al., 1998, 2009; Charrier et al., 2001; Lemasson and Hausberger, 2011; Bouchet et al., 2012a). Interestingly, we provided further evidence of an effect of the social organization (in terms of strong vs. discrete hierarchy) on the level of acoustic variability of the threat call. In fact, we found a greater level of intra-individual structural variability in the threat calls of the “despotic” red-capped mangabeys than in those of the two “tolerant” guenon species. Moreover, we found that identity coding was stronger in female mangabey threat vs. alarm calls, whereas in guenons threat calls were the least individually distinctive. Consequently, we suggest that the call's relevance to the species' social organization (i.e., whether or not the call mediates intra-group interactions, which are essential to social functioning), rather than the type of message it conveys (e.g., contact vs. threat vs. alarm), accounts for its level of acoustic variability. Furthermore, we found that the degree of individual distinctiveness of a call type did not correlate with its level of intra-individual acoustic variability. Also, it seems that in the case of mangabey threat calls, the realized acoustic variability could serve to encode information not only about the caller's identity but also about the context of emission (e.g., rhesus macaques: Gouzoules et al., 1984; chimpanzees, Pan troglodytes: Slocombe and Zuberbühler, 2007).

Whereas contact calls and alarm calls displayed similar frequency patterns (tonal and noisy respectively) in the three species we studied, threat call structural patterns differed across species (tonal as their contact calls in red-capped mangabeys vs. repeated noisy pulses as their alarm calls in guenons). Owren and Rendall (2001), suggested a “structure-affective processing” relationship: (1) call types used by callers to directly influence the affect of listeners would be characterized by peculiar acoustic features well-designed for capturing their attention (e.g., sharp onset, high-amplitude noisiness, repeated energy pulses) but not suitable for conveying individual distinctiveness; (2) call types used to indirectly influence the affect of listeners would give clear cues to caller identity (e.g., tonal harmonically-rich calls) that listeners would associate, through a conditioning process, with past positive or negative interactions with the caller. Accordingly, we found that identity coding was the strongest in the tonal contact calls regardless of the species, and that female mangabey tonal threat calls were more individually distinctive than their noisy alarm calls, whereas in guenons the noisy threat calls were the least individually distinctive. Owren and Rendall (2001) notably illustrated their “structure-affective processing” hypothesis with calls produced by baboons and macaques, contact and threat calls having indirect affective effects, distress and alarm calls having direct affective effects. Interestingly, we found a similar pattern in red-capped mangabeys whose social organization is based on relatively frequent peaceful and agonistic interactions and on a strong hierarchy like most baboons and macaques (Rowell, 1988; Dolado and Beltran, 2012). In contrast, guenons' social organization is based on rare physical interactions and on a discrete hierarchy (Gautier-Hion and Gautier, 1978; Rowell, 1988; Lemasson et al., 2006), and they displayed a different pattern here, with threat calls having an acoustic structure more likely to have direct affective effects on listeners. Consequently, we suggest that the call's relevance to the species' social organization (i.e., its role in mediating interactions that are crucial according to the species-specific social needs) accounts for its acoustic structure and its associated level of individual distinctiveness.

Taken together, these results suggest that the species-specific social needs have exerted a selective pressure on call structure, favouring acoustic variability (e.g., individual distinctiveness or context-related acoustic variability) in some particular calls involved in critical aspects of intra-group social functioning. Also, the theory of an influence of the call social function on its level of acoustic variability can be extended to an influence on the coevolution between its acoustic structure and the associated communicative capacities.

In accordance with the “vocal grooming” hypothesis (Dunbar, 1998; Griebel and Oller, 2008), we found that female red-capped mangabeys called more than twice as much as female guenons. Also, in accordance with the “socially-driven repertoire complexity” hypothesis (Blumstein and Armitage, 1997; McComb and Semple, 2005; Freeberg, 2006; Freeberg and Harvey, 2008; Knotková et al., 2009; Gustison et al., 2012), we found that female mangabeys displayed the largest and most diverse vocal repertoire in terms of structural composition (greatest number of unit types and assembling patterns, highest diversity index: DI = 0.67), Campbell's monkeys showed an intermediate pattern (DI = 0.52), and De Brazza's monkeys produced the smallest and simplest repertoire (smallest number of unit types and assembling patterns, lowest diversity index: DI = 0.33). As red-capped mangabeys live in large multi-male multi-female groups, Campbell's monkeys in harem groups, and De Brazza's monkeys in small family units, our results suggest that the species' social structure account for the shaping of its vocal repertoire. While those evolutionary hypotheses have been, to date, tested separately in different taxa (e.g., primates: Dunbar, 1998, 2012; McComb and Semple, 2005; Gustison et al., 2012; rodents: Blumstein and Armitage, 1997; Knotková et al., 2009; Pollard and Blumstein, 2012; cetaceans: May-Collado et al., 2007; birds: Freeberg, 2006; Freeberg and Harvey, 2008; Krams et al., 2012), our study highlights, by using a multi-level approach (sound unit—assembling patterns—repertoire) of vocal variability in terms of both production and usage, that those hypotheses are complementary (Freeberg et al., 2012).

It is interesting to look more thoroughly at the way differential levels of vocal complexity are achieved across the three species. Parallel to the increase in social complexity between our three species, we found an increase in the number of unit types, in the number of assembling patterns, as well as in the number of “combined” patterns (i.e., concatenation of units of different types). One possibility of “creating” acoustic variability is to have non-fixed acoustic parameters (e.g., red-capped mangabey threat calls), but this has some limit in non-human primates as the control they have over their vocal apparatus is restricted (Lemasson, 2011). Another possibility of “creating” acoustic variability is to use syntactic-like sound combinations. The perceptual salience of those sound combinations remains to be tested for our subjects, nevertheless a growing number of recent studies provided evidence that those higher-level acoustic structures can be meaningful in a proto-syntactic-like way. Guenon males use suffixation (Campbell's monkeys: Ouattara et al., 2009a) and call sequences (putty-nosed monkeys: Arnold and Zuberbühler, 2008; Campbell's monkeys: Ouattara et al., 2009b) to increase the number of context-specific alarm messages delivered. Even in apes, it has been reported that white-handed gibbons concatenate the same set of notes into different songs (Clarke et al., 2006), chimpanzees combine calls (Crockford and Boesch, 2005), and bonobos (Pan paniscus) combine calls into sequences (Clay and Zuberbühler, 2011) in context-specific ways.

The ability of monkeys and apes to combine sound units or calls into more complex structures parallels, in some respects, the combinatorial ability pushed to extremes in human languages. The three closely-related forest species we studied here all used this possibility of syntactic-like sound combinations, but red-capped mangabeys who live in the most complex social groups are the ones who made the most of it. Also, combinatorial ability could have been enhanced in species whose social needs require individuals to increase the number of messages that they can deliver.

Whereas broad inter-species comparisons based on a review of the existing literature suggest some patterns (e.g., repertoire size correlates with group size and time spent grooming in non-human primates: McComb and Semple, 2005), detailed comparative studies on a restricted number of closely-related species allow to take a closer look at how sociality has driven the evolution of communication (Freeberg et al., 2012). In a recent study, vocal repertoire size was compared between two species living in large social groups: geladas who aggregate into a complex multi-level society composed of small reproductive units within which one male forms long-term bonds with several females, and chacma baboons who live in a single-level multi-male multi-female society where cross-sex relationships consist mainly of temporary consortships (Gustison et al., 2012). The authors identified a number of homologous call types common to both species, but they found that geladas displayed a larger vocal repertoire than chacma baboons. The non-homologous (derived) call types specific to geladas function in cross-sex bonding, and they were produced primarily by males who are responsible for the maintenance of those long-term social bonds. Two interesting conclusions emerge from this study. First, social bonding turns out to be a key factor, more critical than the size of the reproductive unit, in the emergence of socially-driven vocal complexity in geladas. Second, the influence of social factors on vocal complexity appears to be more sizeable in males, as a result of their social role within the gelada society.

In our case, we focused on females whose role in mediating intra-group social relationships is predominant in the three species we studied. Therefore, we expected the influence of social factors on vocal variability to be most remarkable in this sex class. Overall, we found that female mangabeys, who are involved in a more complex social network than female guenons, display higher degrees of vocal complexity at several levels (acoustic variability, repertoire size, calling rate). Nevertheless, we do not rule out the possibility of socially-driven vocal complexity in males. It is notable that sex differences in vocal repertoire, though sizeable in the three studied species, are more striking in guenons (Campbell's monkeys: Gautier, 1988; Lemasson and Hausberger, 2011; De Brazza's monkeys: Bouchet et al., 2012b) than in red-capped mangabeys (Bouchet et al., 2010). Thus, relatively speaking, male mangabeys tend to equal the level of vocal complexity displayed by females (e.g., calling rate, vocal repertoire size). However, it is impossible here to discriminate between the effects of group size (larger in mangabeys than in guenons) vs. degree of involvement within the social network (male mangabeys are more socially active than male guenons). In the future, comparative studies on a small number of carefully chosen species might help uncover the relative influence of distinct aspects of sociality (e.g., group size, social network complexity, strength of social bonds) in the evolution of communicative abilities, as well as the extent to which it affects individuals unevenly according to their role within the social group.

So far, we have focused our discussion on comparative studies which, like ours, highlighted a parallel between social and vocal complexity in line with the theory of a social-vocal coevolution of communicative abilities. Interestingly, a study on gestural complexity in three species of macaques (Macaca sp.) showed a relationship between the complexity of social organization and the size of the gestural repertoire as well as the level of variability in communicative patterns (Maestripieri, 2007). More recently, a study on facial mobility in 12 non-human primate species revealed a correlation between group size and the variety of facial movements a species can produce (Dobson, 2009). Another study on chemical complexity of glandular secretions in eight species of brown lemurs (Eulemur sp.) that differ in their social system provided evidence of a correlation between social complexity and olfactory complexity (Del Barco-Trillo et al., 2012). Lastly, a study on the auditory system of 20 non-human primate species revealed a correlation between social complexity and enhanced hearing sensitivity (Ramsier et al., 2012). This set of comparative studies suggests that the theory of a social-vocal coevolution of communicative abilities might also apply to other communicative modalities at both the production and perception levels (Freeberg et al., 2012).

Our study points out the advantage of addressing multiple facets of communicative complexity in a small number of closely-related but socially distinct species. It opens new perspectives for comparative research on the evolution of communication in animal species. The existing literature suggests that socially-driven vocal complexity is widespread in the animal kingdom (reviewed in Freeberg et al., 2012). Targeted detailed comparative studies would help uncover whether or not similar social pressures (e.g., group size, social bonding) have affected identical aspects of communication (e.g., calling rate, repertoire size), and led to comparable levels of vocal complexity in disparate taxonomic groups. This approach applied, more specifically, to representatives of the primate lineage would be of great interest to advance our understanding of the selective pressures that led to the emergence of a communicative system as complex as human language.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.