Chest Color and Social Status in Male Geladas (Theropithecus gelada)
- First Online:
- Cite this article as:
- Bergman, T.J., Ho, L. & Beehner, J.C. Int J Primatol (2009) 30: 791. doi:10.1007/s10764-009-9374-x
Conspicuous colored patches on animals often serve as sexually selected signals that advertise male quality. Such colored traits facilitate assessment of risks associated with a specific contest or benefits associated with a specific mate choice. Here, we investigate whether a colored patch of skin on the chests of male geladas (Theropithecus gelada) is a sexually selected signal. Specifically, we examine the relationship between color (redness), social status (a proxy for reproductive success), and age. We use observational data from known individuals from a population of wild geladas living in Ethiopia. We digitally quantified chest color using a previously-validated method for measuring color under field conditions. Results from this study are consistent with the hypothesis that redness is a quality signal in males. Baseline color correlates with status even when controlling for age. Indeed, males with redder chests were members of “better” groups: 1) leader males—the only males with reproductive access to females—had the reddest chests, and 2) within leader males, males with large units (>6 females) had redder chests than males with small units. At present, we are unable to address whether male chest color is directed at potential rivals or mates. Nevertheless, our data support the hypothesis that quality signals should prevail in large, fluid groups, where it is unlikely that individuals recognize all other group members. If individual recognition is limited in gelada society, this would favor the evolution of alternative means of assessment for making reproductive decisions.
Keywordschest patchcolorationprimatequality signalsexual selectionTheropithecus
Across a wide variety of taxa, sexual selection has produced dimorphic secondary sexual characteristics such as color, ornaments, vocalizations, enlarged canines, size dimorphism, and behavioral displays (Andersson 1994). In many cases, these characteristics communicate information about the signaler to a receiver that is either a member of the same sex (intrasexual competition) or a member of the opposite sex (mate choice). For example, males often use sexually selected signals to advertise their abilities in competitive encounters with other males. Because aggression can result in costly injuries to both winners and losers, individuals should display before fighting to resolve conflicts at the lowest possible cost (Bradbury and Vehrencamp 1998; Maynard Smith 1982; Vehrencamp 2000).
Researchers have proposed animal color as a possible signal advertising male quality (Gerald 2001). If the intensity of coloration is a reliable cue for condition, then colored traits could facilitate male assessment of risks associated with a specific contest or benefits associated with a specific mate choice. Across a wide variety of vertebrate taxa, researchers have demonstrated vivid colors to be honest signals of an animal’s quality (Maynard Smith and Harper 2003). However, among mammals, signaling with color is quite rare. Bright secondary sexual coloration occurs in only a subset of primate species, usually on the face or anogenital area (Dixson 1998, pp. 192–195). Among these, several authors have reported a relationship between the intensity of coloration and dominance rank. For example, the red face and sexual skin of mandrills (Mandrillus sphinx) correlated positively with androgen levels and rank (Setchell and Dixson 2001; Setchell et al.2008; Wickings and Dixson 1992). In male vervets (Chlorocebus aethiops), differences in blue scrotal coloration predicted eventual social status, (Gerald 2001) while in drills (Mandrillus leucophaeus) male groin color correlated with dominance rank (Marty et al., this issue). In addition, a handful of other studies have demonstrated that females prefer more brightly colored males in mandrills (Setchell 2005), rhesus macaques (Macaca mulatta: Waitt et al.2003), and brown lemurs (Eulemur fulvus: Cooper and Hosey 2003).
In contrast with species that rely on quality signals, many group-living species accumulate detailed knowledge about the competitive ability or attractiveness of others based on individual recognition and repeated interactions with the same individuals. Indeed, the vast majority of primate species assess conspecifics using individual recognition (Tomasello and Call 1997): a strategy that works quite well in small social groups that are relatively stable over time, e.g., Papio spp. (Cheney and Seyfarth 2007), vervets (Cheney and Seyfarth 1982), and rhesus macaques (Rendall et al.1996). Perhaps as a result, signaling in primates is the exception rather than the norm. However, individual recognition may have a reduced role in large primate groups where recognition of every member may not be possible. If individual recognition is limited in large primate societies, this might facilitate the evolution of alternative mechanisms of assessment, such as signals (Setchell and Kappeler 2003).
In gelada females, chest patch color changes with reproductive condition, growing brighter close to ovulation and as pregnancy progresses (Dunbar 1977; McCann 1995). However, in males, chest coloration has been hypothesized to reflect the frequency of involvement in agonistic encounters with other males (Dunbar 1984, pp. 180–181). Further, male chest color has been anecdotally observed to turn pale in color after a male is defeated by a rival (Dunbar 1984, p. 180). As such, a male’s chest color might serve as a quality signal to other geladas, broadcasting his fighting ability to males or his quality as a mate to females. In addition (or alternatively), we have shown elsewhere that male chest color maps onto crude age categories, with prime adult males (i.e., males 8–13 yr old) exhibiting significantly redder chest patches than with immature and older males (Bergman and Beehner 2008). However, as with many long-lived organisms (e.g., Setchell et al.2006), male age and status probably have a high correlation in geladas. In addition to age and status, 2 other factors might affect chest color. First, if chest color is a condition-dependent signal, then signal quality might vary seasonally with food availability (Nowicki et al.2002) or weather changes. Second, Dunbar (1984, p. 180) proposed that male chest color intensified with activity level. Thus, in addition to status and age, we include weather variables and activity level in our analyses.
Using cross-sectional and longitudinal analyses, we examine 3 yr of data from a wild population of geladas to test whether chest color is consistent with the sexually selected signal hypothesis. Based on the hypothesis, we predict that leader males will have redder chest patches than non-leader males, and that leader males with more females in their one-male unit will have redder chests than males with fewer females.
Gelada groups are organized into a multi-level society (Dunbar 1983; Dunbar and Dunbar 1975; Kawai et al.1983; Mori 1979b). The smallest level is the one-male unit (unit) consisting of a leader male, several adult females, their offspring, and possibly 1 or 2 follower males. Follower males are typically former unit leaders but can also be young males that have not yet acquired a unit (Mori 1979b). Unit females remain together after their leader male changes, and it is thought that females in the same unit are closely related (Dunbar 1983). Units and bachelor groups that share a common home range are called bands, and bands that temporarily merge to forage or sleep together are called herds (Dunbar and Dunbar, 1975; Mori, 1979b).
Reproductive access to females is largely mediated by competition to acquire and maintain control of a one-male unit (Dunbar and Dunbar 1975). Young males leave their natal unit before maturity and join an all-male group (bachelor group) where they remain for several years before acquiring their own reproductive unit (Dunbar and Dunbar 1975; Mori 1979a). Bachelor males that challenge and successfully defeat a unit leader immediately assume reproductive control over females in that unit (“takeover”; Dunbar and Dunbar 1975; Mori 1979c). Alternatively, bachelor males may also join a unit as a young, submissive, follower to establish a relationship with ≥1 young, pre-reproductive females in the unit. When the females reach maturity, the young follower male separates his female(s) from the parental one-male unit (Dunbar and Dunbar 1975). Although we have not yet observed this method of unit formation (Beehner and Bergman, unpub. data), such a mechanism presumably is a more gradual process. Important for this study, all female transfer between males that we observed occurred via unit takeovers.
We conducted all research in the Simien Mountains National Park of Ethiopia from January 2006 to December 2008 in the Sankaber area, where the gelada population totals ca. 1200 individuals across 4 bands. We have continuous observations on 2 of the bands since October 2005, and members of both bands are fully habituated to human observers on foot.
In all, we analyzed 968 photos from 102 known males from January 2006 through December 2008. We have continuously observed 42 of these males since January 2006 (focal males); they account for 858 (88.6%) of the photos collected. We photographed focal males once every 2–4 weeks. We sampled some subjects for shorter periods because we added them later in the study or they disappeared during the study. Nineteen of the focal males were leader males for ≥12 mo and comprise the main subjects for most of the analyses. Because the method of color quantification controls for variation in ambient light (Bergman and Beehner 2008), we took photographs across all times of the day and under different light conditions.
We quantified color in male chest patches using a method previously validated for use on geladas (Bergman and Beehner 2008). In brief, we took photographs of the gelada chest patch and a Gretag Macbeth (now X-rite) ColorChecker chart using a Nikon COOLPIX 8700 digital camera at the “fine” quality setting. We used the sequential method, photographing a male’s chest patch first, followed by the ColorChecker chart within 2 min, in the same position and orientation as the chest. We used manual settings for shutter speed and lens aperture, and the white balance was set to “daylight.” We purposely underexposed all photos (by 1–2 f-stops) to guard against “clipping.” We used the inCamera plug-in for Adobe Photoshop to create a color profile that adjusts the color in the photograph to the known color levels in each square of the ColorChecker chart. We then assigned and converted the corresponding photo of the male’s chest to this newly created color profile using the settings recommended by the inCamera plug-in. Finally, after this color correction, we selected the appropriate area of the photograph (Bergman and Beehner 2008) using the rectangular marquee tool (≥200 pixels) and recorded the RGB levels using the histogram palette, averaged over selected pixels. To measure “red,” we used the ratio of red to green (R/G ratio or chest color). If we could measure both sides of the patch (72.6% of photographs), we averaged the values for both sides. On average, the right and left patches of the same male differed by 6.0%. Four different people performed all color assessments and interindividual variation in measurements averaged 2.49%.
We recorded weather data daily at the Sankaber research station. We calculated the cumulative precipitation (mm, Rainfall), mean maximum temperature (˚C, Max Temp), and mean minimum temperature (˚C, Min Temp) for each month of the study.
At the time of each photograph, we categorized subjects as active or inactive. Active subjects had engaged in one of the following behaviors <20 min before we took the photo: 1) running vigorously for >50 m or 2) engaging in a physical fight with another male. We designated all other subjects as inactive.
Developmental stage for male geladas, estimated age ranges, and general physical characteristics describing each category
Estimated ages (years)
Adult body size in stature but not in weight. Complete canine eruption. Cape hair light in color, extending just past shoulders. Cheek tufts present but not extending below chin. Ears highly visible. Surrounding fur around chest patch gray-brown in color.
Heavier in appearance than young adults. Cape hair light in color and extending to elbows. Cheek tufts extend to chin or just below. Dorsoventral V of dark hair down center of crown (crown V) forming on head. Ears somewhat visible. Surrounding fur around chest patch white.
Heaviest stage for an adult male. Cape hair dark and extending past elbows. Cheek tufts extend well below chin. Prominent crown V on head. Ears not visible. Canines show signs of yellowing. Surrounding fur around chest patch white.
Same size as or slightly smaller than mid-prime males. Cape hair and cheek tufts begin to recede. Crown V beginning to fade. Ears not visible. Obvious canine yellowing and wear. Surrounding fur around chest patch gray.
Same size as young adult males. Collapsed look to back and shoulders. Cape receded back to shoulders and cheek tufts receded back to chin-line. Crown V no longer present. Ears somewhat visible to highly visible. Worn or missing canines. Hair dull in color and uneven in places. Surrounding fur around chest patch gray-brown.
Based on repeated behavioral observations, we categorized males as bachelors, leaders, old followers, or young followers. Bachelor males included all males that had not yet acquired a unit and were members of a bachelor group (N = 37). Leaders included all males in charge of a one-male unit (N = 42). Follower males included all adult males that were attached in some way to a unit but were not leader males. We split this group into old followers, which included only the former deposed leader(s) (N = 25) or young followers, which included young males that had attached themselves to a unit (N = 13).
Number of females
In addition to the broad categories of male status, we generated a more fine-tuned quantitative measure of male status for leader males by recording the number of adult females in each unit. We considered females adults when they began to exhibit sexual swellings, suggesting that they had reached sexual maturity.
We performed all statistical analyses using SPSS (17.0 for Mac); tests were 2-tailed, and the statistical threshold for analyses was p < 0.05. Chest color values did not deviate from a normal distribution (K-S test: p = 0.34), and therefore we used parametric statistical tests.
Does Chest Color Relate to Weather or Activity?
For each of the main 19 leader males, we investigated the relationship between weather variables and male chest color. If a male had >1 chest photo per month, we calculated a mean chest color for each male/month. We used a general linear mixed model (GLMM) to assess the effects of 3 continuous independent variables (Rainfall, Max Temp, Min Temp) on 1 continuous dependent variable (Chest Color). We accounted for multiple measures from the same subject by including each leader male as a random factor in the model, with their own random intercept and slope. Owing to the correlation between variables [Max Temp and Min Temp correlate positively (Pearson’s correlation: r = 0.57, p < 0.001) and Max Temp and Rain correlate negatively (Pearson’s correlation: r = −0.58, p < 0.001)], we fit separate models for each variable and used the Akaike’s Information Criterion (AIC) to compare the goodness of fit for different models.
We then examined whether male physical activity affected male chest color. For each focal male, we split all chest photos into active or nonactive photos and calculated a mean for each category. One leader male did not have any active photos and we removed him from the analysis. To ensure that each male served as his own control, we used a paired t-test to compare chest color for active versus nonactive photos.
Does Chest Color Relate to Age or Status?
Because gelada males did not appear to have dyadic interactions consistent with a dominance hierarchy, we approached this question in 2 ways. First, we used a cross-sectional analysis to determine the relative importance of age and status on chest color. For this analysis, our measure of male status was qualitative and consisted of the broad status categories of bachelor, leader, old follower, and young follower. We used a GLMM to assess the effects of 2 fixed factors (Age Category and Status Category) on 1 continuous dependent variable (Chest Color). Again, we accounted for multiple measures from the same subject by including each subject as a random factor in the model.
For both of these analyses we used the full set of photos from all males. Each male contributed only 1 mean chest color value per age/status category. If a male’s age or status category changed, that male received a color value for each new age or status category. For the quantitative analysis, we also calculated separate chest color means for units that changed size. As before, we accounted for multiple measures from the same subject by including each subject as a random factor in the models. We fit separate models for each variable and used the AIC to compare the goodness of fit for different models.
Does Chest Color Change Across Takeovers?
Finally, we examined male chest color across male takeovers to look at changes in redness across a very specific, relatively short period of time. We focused on the days surrounding takeovers that involved males for whom we had chest photos from before and after the takeover (N = 9). We classified these 9 males as winners, i.e., bachelor males that became leader males after the takeover (N = 3) or losers, i.e., leader males that became follower males after the takeover (N = 6). To explore changes in color visually, we pooled color values for the different takeovers over the following intervals: 100–51 days before, 50–1 day before, 0–10 days after, 11–50 days after, and 51–100 days after the takeover. At present, our sample size precluded statistical analyses. Therefore, our treatment of these data at this time is necessarily descriptive.
Does Chest Color Relate to Weather or Activity?
The model with the lowest AIC, i.e., the best fit, included only Min Temp as a covariate. However, this model was not significant (F(1,175) = 2.63, p = 0.11). Indeed, none of the weather variables in any of the models demonstrated an effect on male chest color (p > 0.10, all models). Therefore, we did not include weather variables in any further analyses.
Photos taken within 20 min of male activity were significantly redder than photos taken during nonactive periods (paired t-test: t = −4.31, df = 17, p < 0.001). Therefore, to examine differences in baseline chest color, we removed all active photos from subsequent analyses.
Does Chest Color Relate to Age or Status?
Does Chest Color Change Across Takeovers?
Our results are consistent with the hypothesis that chest color (specifically, redness) is a sexually selected signal in male geladas. First, despite evidence of some intraindividual plasticity in chest color resulting from increased activity, status category was the best predictor of baseline chest color. Further, when we included age in the model, status remained a better predictor of color. Specifically, males with redder chest patches had higher potential for reproductive success. Leader males—the only males with reproductive access to females—had the reddest chests; and within leader males, males with large one-male units, i.e., >6 females, had redder chests than males with small units.
One possible explanation for the relationship between chest color and status is that color reflects the quality of a male, either genetic quality or current condition, and that only high-quality males are able to maintain access to a unit (or a large unit). This explanation is consistent with a sexually selected signal hypothesis and indicates that chest redness is similar to other quality signals used in rival assessment or partner choice. For example, male red deer (Cervus elaphus) assess rival males based on acoustic properties of loud vocalizations, e.g., pitch and rate of calling, which are honest indicators of size and fighting ability (Clutton Brock and Albon 1979). A particularly relevant example is the red facial color of male mandrills. Male mandrills assess rivals based on red facial coloration (Setchell and Wickings 2005), and face color appears to be an honest signal of competitive ability (Setchell et al.2008). Similarly, gelada bachelor males may avoid leader males with relatively red chest patches, for a given number of females, if chest color advertises a male’s fighting ability. Signals tied to male quality can also be the basis of mate choice. For example, female stalk-eyed flies (Cyrtodiopsis spp.) prefer to mate with males with exaggerated eyespans (Wilkinson and Reillo 1994), and these males produce higher quality offspring (David et al.2000). Among primates, mandrills also provide one of the rare examples of female choice based on a quality signal; female mandrills prefer males with more brightly colored faces (Setchell 2005). Although we have yet to test this hypothesis, female geladas may prefer males with redder chest patches.
Alternatively, high status might result in redder chest coloration. For example, males with more unit females might have higher levels of overall activity, either physical activity or mating activity, and this could lead to higher levels of baseline redness. Consistent with this hypothesis, large units were taken over at higher rates than small units. At present, we are not certain whether this high rate of takeovers results because large units 1) are inherently more difficult to defend; 2) receive more takeover attempts; or 3) are challenged by higher quality males, but we suspect that all 3 possibilities may be true. As such, leaders of large units might have redder chests because takeover challenges, successful or not, involve high levels of activity. Although we removed all photos taken within 20 min of physical activity, it is possible that higher levels of overall activity maintain chest color at a higher baseline. Therefore, if chest redness is merely a byproduct of activity, as with the human face after exercise, and all males are equally capable of producing red chests, then it is unlikely that chest color is a sexually selected signal. However, if activity itself is a reliable indicator of condition (Riechert 1978) or aggressive intent (Searcy et al.2006), then chest color resulting from increased activity could still be a sexually selected signal. Moreover, if high levels of activity are reliably associated with a male’s condition or willingness to fight, bachelor males may avoid males that have particularly red chests for a given unit size because of the greater relative costs of attacking such males. At present, we are focusing our research efforts on unit size, activity levels, and baseline chest color to understand the causal relationship among these factors.
In contrast with other studies of rival assessment where males with the better signal win the challenge (Møller 1987), gelada chest color did not predict the outcome of intrasexual competitions; in other words, the redder male did not usually win the competition. Before a unit takeover, leader males (the eventual losers) were consistently redder than the challenging bachelor males (the eventual winners). Therefore, if gelada bachelor males are assessing leader males based on color, they are not comparing the leader male’s chest color to themselves, e.g., mutual assessment (Parker 1974). Instead, they might be comparing chest color among leader males and then challenging only relatively pale males for a given unit size. Under this hypothesis, leader males with more females in their unit would need to maintain redder chests than males with fewer females to avoid being targeted by bachelors. As such, chest color would be a unidirectional signal, with leader males signaling their condition to bachelors, but not the reverse. The red facial coloration of male mandrills exhibits a remarkably similar pattern; redness correlates with dominance rank, but the ascent to high rank often precedes the development of full red coloration, and the loss of rank precedes the loss of coloration (Setchell and Dixson 2001; Setchell et al.2008). Thus, for geladas and mandrills, maximum coloration is exhibited only by males in a position that needs to be defended, i.e., leader of a one-male unit or high rank.
Similarly, if chest color is the basis of female choice, females are not simply comparing chest color of their current leader to a would-be challenger, because such a comparison would always favor the leader male. Yet, chest color could still facilitate female choice if gelada females are able to assess the relative redness of their leader. For example, females could choose to defect from relatively pale leaders. Previous research on geladas has shown that females with close social bonds to other females can form coalitions against some males and successfully oust them (Dunbar and Dunbar 1975).
Similar to testosterone under the winner-loser effect, wherein winners of competitive encounters experience a rise in testosterone while losers experience a decline (Bernhardt et al.1998; Bernstein et al.1974; Mazur 1976; Mazur et al.1992), gelada males exhibited increased chest redness for winners and decreased redness for losers, suggesting a possible link between chest color and testosterone. Previous studies on captive male rhesus macaques have demonstrated that red skin color is mediated by testosterone, mainly via aromatization to estradiol (Rhodes et al.1997). Data from mandrills indicate that testosterone predicts short-term changes in dominance rank which, in turn, predict short-term changes in red facial coloration (Setchell et al.2008). Further, testosterone could provide a mechanism for keeping chest color “honest” (Zahavi 1975), because maintaining high levels of testosterone is costly (Buchanan et al.2001; Folstad and Karter 1992; Marler and Moore 1988). The mechanism underlying chest color is just one of several questions that arises from the results of this study.
This study also raises functional questions about the intended recipients of chest color: are they rival bachelors or female mates? Certainly, bachelors are making decisions about which leader males to challenge; thus it seems probable that a condition-dependent signal would benefit both leaders and bachelors to resolve conflicts at the lowest possible cost (Bradbury and Vehrencamp 1998; Maynard Smith 1982; Vehrencamp 2000). However, unusual among Old World monkeys, gelada society is relatively female-dominated, creating a strong potential for female mate choice, suggesting that leader males might signal their quality to females in their unit. Although at present we have no data to address this question, it seems unlikely that a leader male would need to advertise his quality to the individuals that are most familiar with him – mainly the females in his unit. Rather, theory suggests that the evolution of signals is favored when individuals frequently encounter others on which they do not have detailed information (Maynard Smith and Harper 2003; Setchell and Kappeler 2003).
Quality signals may be unusual among primates because most nonhuman primates accumulate detailed knowledge about other members of their group, and even neighboring groups, based on repeated interactions with those individuals. If, indeed, chest color is a quality signal, then we need to explain why geladas (and the other primate species with quality signals; see other articles in this issue) are the exception to the typical primate pattern. Despite the wealth of data on signals from other taxa, only very recently have quality signals been studied in primates (Fischer et al.2004; Gerald 2001; Gouzoules and Gouzoules 2002; Kitchen et al.2003; Setchell 2005; Setchell and Wickings 2005; Setchell et al.2008). Although speculative at this stage, there may be a relationship between geladas’ large, fluid groups and the evolution of sexually selected signals (Setchell and Kappeler 2003). In the largest social aggregate for geladas (>1000 individuals), it is unlikely that individual geladas recognize or have interacted with all other group members. Thus, it seems plausible that the unusually large groups of gelada society, where individuals need to assess others with whom they have no prior experience, has favored the evolution of alternative mechanisms of assessment for making reproductive decisions about unfamiliar individuals. For primate species with quality signals, one promising avenue for future research is to understand where individual recognition ends and where other means of assessment begin.
First and foremost, we thank James Higham for extending the invitation to us to participate in the symposium on primate color. We also thank the Ethiopian Wildlife Conservation Department, the Amhara National Regional State Parks Development and Protection Authority, and the wardens and staff of the Simien Mountains National Park for granting us the permission to conduct this research. We thank H. Gelaye and A. LeRoux for their help with data collection in the field, and S. Rosinus, K. Shaw, and K. Weldman for their help in analyzing photos. Finally, we thank 2 anonymous reviewers for their helpful comments on earlier drafts of the manuscript. Funding was provided by the Wildlife Conservation Society (SSF grant no. 67250), the National Geographic Society (NGS grant no. 8100-06), the L.S.B. Leakey Foundation, the National Science Foundation (BCS-0715179), and the University of Michigan. This research was approved by the University Committee on Use and Care of Animals (UCUCA) at the University of Michigan and adhered to the laws and guidelines of Ethiopia.