Animal Behaviour 81 (2011) 1223e1230

Contents lists available at ScienceDirect

Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav

When females trade grooming for grooming: testing partner control and partner choice models of cooperation in two primate species Cécile Fruteau a, *, Sylvain Lemoine b,1, Eléonore Hellard c, 2, Eric van Damme a, Ronald Noë d, 3 a

CentER, Tilburg University CERCOPAN (Centre for Education, Research and Conservation of Primates and Nature) c Laboratoire de Biométrie et Biologie Evolutive, UMR 5558, CNRS, Université Lyon1 d Ethologie Evolutive, DEPE, IPHR, UMR 7178, CNRS-UdS b

a r t i c l e i n f o Article history: Received 6 July 2010 Initial acceptance 11 August 2010 Final acceptance 2 March 2011 Available online 19 April 2011 MS. number: 10-00471R Keywords: biological market Cercocebus atys Chlorocebus aethiops grooming hierarchy parcelling raising the stakes time matching

We tested predictions following from the biological market paradigm using reciprocated grooming sessions among the adult females in a sooty mangabey, Cercocebus atys, group with 35 females (Ivory Coast) and in two groups of vervet monkeys, Chlorocebus aethiops (South Africa) with four and seven females, respectively. Closely ranked females often groomed frequently. The exchanges within such dyads were generally characterized by time matching, but the subordinates groomed for longer than their dominant partners. The reciprocal nature of over 90% of the grooming sessions allowed us to investigate ‘partner control’ strategies such as ‘parcelling’ and ‘raising the stakes’. Females of both species neither parcelled nor gradually invested more grooming in the course of sessions. Rather, the longer bouts of a grooming session were usually at the beginning of the session and the length of the first bout reliably predicted the length of the whole session for frequently grooming partners. Furthermore, we compared potential trust-building behaviour (or ‘strategies’) in frequent and infrequent grooming partners. We found that infrequent groomers of both species showed no signs of trust building and that the first bout they invested in a grooming session did not predict the session length. We conclude that each female has a good knowledge of her value as a grooming partner within each dyad and knows how much she has to invest to receive a satisfactory amount of grooming within the same session. Ó 2011 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

In biology, behaviours are called ‘altruistic’ if they result in a momentary decrease in fitness of the actor and increase in fitness of the recipient. Altruism has long been considered as paradoxical in evolutionary terms until Hamilton (1964) and Trivers (1971) proposed explanations for their occurrence in related (kin selection theory) and unrelated (reciprocal altruism theory) individuals, respectively. The iterated Prisoners’ Dilemma (IPD) has been the game-theoretical paradigm of choice for reciprocal altruism (RA) in spite of some crucial differences between the two concepts (Noë 1990). The essence is the same, however: both RA- and IPD-based models predict contingent strategies with relatively short memories, that is, the animals are expected to base their choice of action * Correspondence: C. Fruteau, CentER, Tilburg University, PO Box 90153, 5000 LE Tilburg, The Netherlands. E-mail address: [email protected] (C. Fruteau). 1 S. Lemoine is at CERCOPAN, 4 Ishie Lane, HEPO Box 826, Calabar, Cross River State, Nigeria. 2 E. Hellard is at the Laboratoire de Biométrie et Biologie Evolutive, UMR 5558, CNRS, Université Lyon1, 43 Bd du 11 Novembre 1918, 69622 Villeurbanne cedex, France. 3 R. Noë is at Ethologie Evolutive, DEPE, IPHC, UMR 7178, CNRS-UdS, 23 rue Becquerel, 67087 Strasbourg cedex 2, France.

on the behaviour shown by their partner during their last few interactions. It has proved difficult to demonstrate such contingent strategies in nature, however (e.g. Roberts 1998; Hammerstein 2003; Bshary & Bronstein 2004; Sachs et al. 2004; Bergmüller et al. 2007; Silk 2007; West et al. 2007a, b; but see Bshary et al. 2008 for a rare exception), while the results of experiments with captive primates have been mixed (e.g. Melis et al. 2008; Brosnan et al. 2009; Dufour et al. 2009). With the introduction of biological market theory (Noë & Hammerstein 1994, 1995) the emphasis switched from partner control to partner choice. According to this theory, the exchange rate of goods and services (‘commodities’), traded between partners to their mutual benefit, is determined by the law of supply and demand. The driving force is partner choice, which induces outbidding competition among the group of individuals from which the partner(s) are chosen. While many studies of mutualistic systems (reviewed in Bshary & Noë 2003; Sachs et al. 2004; Leimar & Hammerstein 2010) and intraspecific cooperation (reviewed in Kutsukake & Clutton-Brock 2008, 2010) have shown the importance of the biological market theory in predicting how cooperation operates, there is still some disagreement among primatologists (reviewed in Barrett & Henzi 2006; Schino & Aureli 2010).

0003-3472/$38.00 Ó 2011 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.anbehav.2011.03.008

1224

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

In primates, allogrooming (grooming, hereafter) has long been considered as an indicator of the long-term bonds between individuals (Hinde 1976; Schino et al. 2007; Dunbar 2010). More recently and in the framework of biological market theory, grooming has been considered as a low-cost currency in the sense that it can be used to compensate for imbalances in the trading of other commodities (Barrett et al. 1999, 2000). Hence, grooming is either reciprocated in kind by the partner (Barrett et al. 1999, 2000; Manson et al. 2004) or exchanged for another commodity, such as tolerance at food sources (Barrett et al. 1999, 2002), agonistic support (Hemelrijk & Ek 1991; Schino 2007), food (de Waal 1997), access to newborns (Henzi & Barrett 2002; Gumert 2007a; Fruteau et al. 2011) and mate compliance (Gumert 2007b). We investigated grooming sessions (hereafter called neutral sessions) during which grooming was not obviously and immediately exchanged for commodities such as food, infant and mate access in two free-ranging primate species, sooty mangabeys, Cercocebus atys, and vervet monkeys, Chlorocebus aethiops. While both species are mainly terrestrial and seasonal breeders they show substantial differences. Mangabeys range in the rainforest and form large groups. Females display exaggerated sexual swellings during their sexual cycles and few conflicts characterize their relationships outside the mating season. In contrast, vervets form small units of about 10 individuals and range in open areas such as savannah. Females have concealed ovulations and conflicts are frequent between group members. We defined a grooming session as a sequence of bouts exchanged between two partners. We used biological market theory (Noë & Hammerstein 1994, 1995) to predict how grooming bouts would be traded among females. We first checked whether supply/demand ratios or power asymmetries between partners influenced grooming bout lengths. The original function of grooming was undoubtedly the removal of ectoparasites and treatment of the fur (Hutchins & Barash 1976; Zamma 2002), but primates groom much more than seems necessary for these basic functions (e.g. Dunbar 2010). We therefore conjecture that the mechanisms that originally ensured that the animals liked to be groomed (e.g. the release of beta-endorphin, Keverne et al. 1989) became themselves rewarding in the course of evolution. Whatever the precise present-day function, we can safely assume that grooming has diminishing returns: ectoparasites are at some point removed and hormone titres at some point reach their maximum. In other words, in long grooming bouts the first minute of a bout is worth more than the last minute. However, we assume that grooming bouts of equal length given by a certain donor would be of equivalent value to a specific receiver. Power asymmetries such as rank distance between females should influence the grooming lengths, because dominants have additional commodities, such as tolerance at food patches, restraint in dyadic conflicts with the subordinate and agonistic support in conflicts, to trade that subordinates cannot offer. We therefore predicted that the subordinate of a dyad would generally groom more than the dominant, but that females of neighbouring rank would time-match their investments in term of grooming duration more than those of distant rank (Henzi et al. 2003). We expected to find a strong effect of the number of potential grooming partners for females. We therefore thought that this effect would be stronger in mangabeys than in vervets because of their larger groups. Biological market theory is not an alternative to partner control models and therefore does not exclude the use of ‘partner control’ strategies once cooperating dyads have formed. One partner control mechanism that would apply to grooming is parcelling (Connor 1995). Connor (1995) assumed that cooperating individuals are initially caught in an iterated Prisoner’s Dilemma. By delivering their goods and services in small packages they de facto change the payoff

matrix of each round of the game in such a way, however, that it is no longer a Prisoner’s Dilemma, and thus they reduce the risk of exploitation. The model applies especially well to grooming, since this service can be delivered in packages of almost any size. The model would thus predict that grooming bouts remain short within grooming sessions. We cannot predict, however, how short a bout should be to fulfil the requirements of the payoff matrix. We therefore simply predicted that grooming partners should take turns within grooming sessions and that the grooming bouts should be roughly equal in length within and between partners. Roberts & Sherratt (1998) developed the parcelling model further and proposed the ‘raising-the-stakes’ (RTS) strategy according to which animals can limit the risk of being exploited by starting with delivering small packages, but then increasing the portion delivered in each round, as long as the partner continues matching the investment. The dyadic structure of grooming sessions makes them likely interactions in which RTS could be used (Keller & Reeve 1998). RTS applies notably to the start of new cooperative relationships, but would also apply if the trust in partners goes back to (almost) zero between interactions during a series of repeated interactions. Barrett et al. (2000), after attempting to compare investments between grooming sessions chronologically dispersed over time in baboons, Papio ursinus, concluded that one has little chance of recording the real starting point of a relationship. We looked, therefore, for the use of RTS within grooming sessions but not between sessions. If RTS is used, we expected that partners would invest little at the beginning of the session and would gradually increase their investment if their partner at least matched the last bout of grooming given, but only in dyads in which trust needed to be built up during each grooming session. We therefore made a distinction between dyads grooming frequently and those grooming infrequently, assuming that trust building would still be necessary only in the grooming sessions of the latter. We predicted that frequent grooming partners would groom for longer at the start of the session than infrequent groomers. Finally, the first bout of a session may have the quality of a first bid and reflect the willingness of the individual to invest in this particular grooming session. Thus, we predicted that the first grooming bout within a session should reliably predict the total length of the session as long as grooming is not interrupted by external events. METHODS Research Areas, Subjects and Data Collection Sooty mangabeys We conducted the study in the Taï National Park, Ivory Coast between 1 November 2001 and 20 August 2002. The park is one of the last remaining blocks of West African primary forest and covers about 454 000 ha. The forest is classified as ‘tropical moist forest’ (Whitmore 1990), with a mean annual rainfall of 1875 mm, a mean annual temperature of 24
C (Taï Monkey Project data, 1991e1999) and a distinct dry season from December to March. Our group of mangabeys was well habituated to human observers before the study and we recognized all adult, subadult and infant members by facial features. Its home range covered about 7 km2 near the western border of the park. The group was not provisioned. During the study we observed 7e14 adult males, 35 adult females and about 70 juveniles and subadults. Seven infants were born between 10 December 2001 and 10 March 2002. One died on 2 February 2002. We focused the data collection on adult females. We used unidirectional ‘approach/retreat’ and ‘threat/retreat’ interactions to determine the female dominance hierarchy. It remained stable throughout the study period (linearity of the female rank order:

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

MatMan test: c241 ¼ 447:89, P < 0.0001, h ¼ 0.97, K ¼ 0.97). We used both ad libitum and focal sampling observation techniques (Altmann 1974) to collect data on grooming sessions occurring between all females. Grooming bouts were timed to the nearest 30 s. A bout was considered to have ended when either the direction of grooming changed or when there was a break of more than 30 s. We used 15 min focal sampling with at least 60 min between consecutive samples of the same individual and 3 min between samples of different individuals. However, for the analyses we also used the focal samples that were at least 9 min long (89 of 2272 samples) if they were truncated because the subject moved out of the observer’s sight. For each focal animal, we recorded each minute (instantaneous sampling, Altmann 1974) the presence/ absence of the female’s infant, the distance from its mother and the nearest adult female and adult male within 5 m. Social interactions were recorded continuously (detailed ethogram in Range & Noë 2002). Owing to limited visibility in the early evening, we opted for a sampling schedule from 0700 to 1600 hours. We collected a total of 568 h of focal samples for all of the 35 adult females (range 63e65 per female). All females were followed at least once every 3 days and we randomized each female’s sampling to account for the time of day. Ad libitum data were recorded all day long (even while doing focal sampling on a subject) as soon as a social interaction (aggression, grooming, mount, etc.) between two identified individuals was observed. Vervet monkeys We conducted the study in the Loskop Dam Nature Reserve, Mpumalanga province, South Africa. The Loskop reserve is characterized by a ‘bushveld’ (tall grasses, thick acacia bushes) type of habitat. The reserve covers approximately 25 000 ha, on average 1000 m above sea level. The area has dry and cold winters (temperatures below 5
C at night and 25
C during the day) from May to October and hot and humid summers (rainfalls of about 500 mm; temperatures 25e40
C) from November to April. Both study groups had home ranges of approximately 3 km2 each that were about 3 km apart. The home range of the Donga group followed narrow dry riverbeds and mainly contained tall trees, such as fig trees, while the home range of the Picnic group was situated in a plain essentially composed of tall grasses and acacia bushes. Two artificial lakes, the Loskop Dam and the Picnic Dam, respectively, provided water to the Donga and Picnic groups the whole year round. The Donga group did not have contact with tourists and was not provisioned outside the context of experiments (see Fruteau et al. 2009). The Picnic group was provisioned by tourists, almost exclusively on Sundays, and regularly ate from the dustbins of the picnic site. The group also obtained food rewards during experiments (see Fruteau et al. 2009). The Donga group was habituated to the presence of human observers at the beginning of the study (from May to mid-October 2004) and the Picnic group was habituated before the second field session (from February to July 2005). The Donga group had three to five adult males, seven adult females, one to two subadult males and one to two infants at a time. The female dominance hierarchy changed between the first and the second field period after the death of the beta female (linearity of the female rank order: MatMan test: first period: c223 ¼ 48, P ¼ 0.002, h ¼ 1, K ¼ 1; second and third periods: c220 ¼ 60:67, P < 0.0001, h ¼ 1, K ¼ 1). The Picnic group had two to three adult males, four adult females, one juvenile male and two to six infants at a time. The female hierarchy remained stable throughout the two field periods (linearity of the female rank order: MatMan test: c2undef ¼ undefined, P ¼ 0.373, h ¼ 1, K ¼ 1). The genetic relatedness between most members of the groups was known as a result of work done during a follow-up study (J. Hula, R. Pansini, R. Noë & M. Kruetzen, unpublished data), in which material for the genetic analysis was

1225

extracted from faecal samples of each individual. Kin coefficients were calculated based on 13 loci (D10S1483, D1S518, D11S928, D1S207, D3S3591, D1S244, D8S1106, D5S1466, D7S2446, D11S1902, D18S42, D15S108, D2S144) and using the pairwise relationship coefficients from Queller & Goodnight (1989) and the pairwise relationship coefficient r from Wang (2002), which gave similar results. Two individuals were considered unrelated when their coefficient of similarity was <0.1 or negative. During the first field period, we followed the Donga group on a regular basis from mid-October to mid-December 2004. Thereafter, we followed each group every second day during the second field period from September 2005 to the end of April 2006 and 2 days in a row every 4 days during the last field period from May 2006 to the end of September 2006. Observations were distributed throughout the day but most data were obtained during 0600e1300 and 1400e1800 hours. Data were collected by focal group sampling of the adult animals (Altmann 1974), that is, when all adults were visible simultaneously, or by ad libitum sampling when only one observer was in the field or when one adult animal was out of sight or missing from the group. We collected 605 and 422 h of group focal and 100 and 70 h of ad libitum sampling for the Donga and Picnic groups, respectively. Grooming bouts were timed to the nearest second. A bout was considered to have ended when either the direction of grooming changed or when there was a break of more than 20 s. A session was considered to have ended when partners moved apart or when there was a break of more than 20 s. Data Analysis and Statistics For this study, we excluded all grooming sessions involving juveniles, males, mothers of infants less than 3 months old and receptive females. Mangabey females display exaggerated sexual swellings, which give an indication of receptivity. Vervet females, however, have cryptic reproductive cycles so we excluded all grooming interactions occurring during the 15 days before a female was first seen mating with a male and 7 days after we recorded her last mating interaction. We also discarded the grooming sessions of vervets taking place during our experimental trials. Tests were performed using SPSS version 17.0 (SPSS Inc., Chicago, IL, U.S.A.) and R version 2.10.1 (R Core Development Team, Vienna, Austria). The alpha-level was set to 0.05. We first examined general descriptive statistics of grooming sessions for both species: daily budget, average grooming session length, time lag between a bout and its reciprocation. We used a two-tailed G test to compare the proportion of initiations performed by the lower-ranking females of a dyad with the expected value assuming no effect of status. We then tested whether females tended to time-match their exchanges (model 1) or were influenced by their partner’s status (model 2), using linear mixed-effects models. In model 1 we used the duration of grooming given by the initiators as the dependent variable. For each bout of grooming given by the initiator we used as fixed effects the group (mangabey, vervet Picnic group or vervet Donga group) and the duration of grooming given by the receivers. In model 2 we used the absolute difference of grooming time given by partners, defined as time given by receiver  time given by initiator as the dependent variable. For each absolute difference we used as fixed effects the group (mangabey, vervet Picnic group or vervet Donga group) and the rank distance between the partners. The rank distances were defined as: rank of receiver  rank of initiator. We inserted the identity of the grooming dyad in both models as a random effect on the intercept to prevent pseudoreplication. We also standardized the data set to compare the impact of each effect separately by subtracting the mean and dividing by the standard deviation for each data point. To highlight our most

1226

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

Table 1 Grooming pattern in mangabeys and vervets Groups

Mangabey

Vervet Donga

Vervet Picnic

Size No. of females No. of grooming sessions Daily budget (%) Average session length (s)

ca. 130 35 363 15.611.03 343129

ca. 15 7

ca. 11 4 323 15.301.08 323325

salient results, we performed post hoc simple linear regressions on the standardized data set. Finally, to test whether females used either parcelling or RTS, we used nonparametric two-way Friedman ANOVAs to test whether the durations of the grooming bouts fluctuated over the whole grooming session (1) irrespective of the groomer and (2) for each of the partners separately. Nonsignificant results would validate the parcelling model. In the case of significant results, we used post hoc sign tests to determine whether the durations increased or decreased (following the method proposed by Barrett et al. 2000). Following Barrett et al. (2000), we defined groomers as ‘frequent partners’ if each of them spent more than 5% of their total active þ passive grooming time grooming each other. We used two-tailed ManneWhitney U tests to determine whether frequent partners were closer in rank than infrequent ones. To analyse whether the first given bout could reliably predict the length of the whole session, we used linear mixed-effect models. We built different models for mangabeys and vervets, as we could include the available genetic relatedness for the latter only. For both groups we used the duration (s) of the rest of the session as the dependent variable. For each point we used as fixed effects the duration (s) of the first bout and the categorical variable ‘frequent groomer or not’. For vervets we also added the dichotomous effects group (Donga or Picnic) and relatedness (related or not). We inserted the identity of the initiator as a random effect on the intercept to prevent pseudoreplication in both models. We also standardized the data set to analyse the impact of each effect separately. To better visualize the combined effects of frequent/infrequent groomers and duration of the first bout on the duration of the session, we performed post hoc Spearman correlations. As linear mixed-effect models are robust to an ‘almost’ normal distribution, we visually assessed the normality of all models residuals using the R function quantileequantile plot (QeQ plot). RESULTS Pattern of Grooming Sessions The daily time budget and the average session length data are summarized in Table 1. In mangabeys, 355 sessions out of 363 (97.8%) showed grooming in both directions. The lower-ranking female initiated 199 of these 355 sessions, which is not a significant deviation from the null hypothesis of balanced initiative taking (two-tailed G test: G1 ¼ 2.374, P ¼ 0.193). A high proportion of grooming sessions in the vervets (292 of 323; 90.4%) also showed two-way grooming. However, the lowerranking females initiated 262 sessions, which was significantly more often than their higher-ranking partners (two-tailed G test: Donga: G1 ¼ 43.919, P < 0.001; Picnic: G1 ¼ 33.520, P < 0.001). Time Matching and Influence of Hierarchy The linear mixed-effect model 1 (summarized in Table 2) shows a significant effect of the time given by the initiators on the time given by the receivers in the three study groups. The vervet Donga

Table 2 Summary of the linear mixed-effect models for time matching and the influence of status in both mangabeys and vervets Effects Model 1: time matching SD (intercept: 0.376eresidual: 0.460) Time receiver Group mangabey Group vervet Picnic Time receiver*group mangabey Time receiver*group vervet Picnic Model 2: influence of status SD (intercept: 0.302eresidual: 0.253) Rank distance Group mangabey Group vervet Picnic Rank distance*group mangabey Rank distance*group vervet Picnic

Value

SE

df

t

P

0.090 0.990 0.268 0.308 0.771 0.210

0.083 0.030 0.093 0.190 0.055 0.050

503 1.091 0.275 503 33.434 <0.0001 177 2.886 0.004 177 2.621 0.007 503 14.071 <0.0001 503 4.180 <0.0001

0.122 0.183 0.132 0.908 0.079 0.174

0.065 0.009 0.072 0.170 0.009 0.062

503 503 177 177 503 503

1.866 0.063 19.914 <0.0001 2.839 0.008 5.355 <0.0001 8.390 <0.0001 2.815 0.005

Model 1: time initiator w time receiver*groups j random effect on the intercept: identity of dyad; Akaike’s information criterion: 1110.651. Model 2: time discrepancy w rank distance*groups j random effect on the intercept: identity of dyad; Akaike’s information criterion: 380.507. The vervet Donga group is included in the intercept, that is, results for ‘Group mangabeys’ and ‘Group vervet Picnic’ are for these groups compared to the vervet Donga group. Significant effects are in bold.

group is on the intercept, which means that both the mangabey group and the vervet Picnic group values need to be compared with the intercept. Hence, even though the three groups showed time matching of grooming bouts within sessions, the time matching was (1) weaker for the mangabey group as shown by the negative value of the combined effects ‘time receiver*mangabey group’ and (2) stronger for the vervet Picnic group (positive value of the combined effects ‘time receiver*Picnic group’). The simple linear regressions showed that the mangabeys matched the bout length of their partners to a low degree only (F ¼ 102.06, P < 0.001, r2 ¼ 0.21; Fig. 1a), while the bout length in vervets could be explained much 2 ¼ 0:89 and F ¼ 1146.669, better (F ¼ 2173.983, P < 0.001, rDonga 2 ¼ 0:94; Fig. 1b, c) by time matching. Hence, time P < 0.001, rPicnic matching did not seem to play a major role in mangabeys. The linear mixed-effect model 2 (summarized in Table 2) showed a significant effect of the rank distances on the time discrepancies in the three study groups. The vervet Donga group is again on the intercept, which means that each mangabey and vervet Picnic group value needs to be compared with the intercept. Hence, even though the results for the three groups showed that the lower-ranking female of a grooming dyad groomed her partner for longer than she was groomed back, the effect was weaker for both the mangabey and the vervet Picnic group as shown by the negative value of the combined effects ‘rank distance*group’. The simple linear regressions showed that rank distances explained up to 73% of the variation observed in time investments (F ¼ 1018.876, P < 0.001, r2 ¼ 0.73; Fig. 1d) in mangabeys and up to 89% and 87% in vervets (Donga group: F ¼ 2109.129, P < 0.001, r2 ¼ 0.89; Picnic group: F ¼ 2276.601, P < 0.001, r2 ¼ 0.87; Fig. 1e, f). Partner Control Strategies Friedman tests revealed that there were significant bout length differences over grooming sessions irrespective of the groomer’s identity (Table 3) and for each of the partners separately (Table 4) in both species. This is in contradiction with the parcelling model according to which bouts should have remained of similar length throughout the session. Furthermore, the only significant results of the post hoc sign tests suggest a decrease in the grooming bout lengths within the grooming sessions independent of the identity of the groomer (Table 3). They also showed a decrease in the grooming lengths for both partners across sessions and within

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

5 Time receiver

4 3 2 1 0 0

Time discrepancy

6 4

6

7 (b) 6 5 4 3 2 1 0

(a)

1

2

3

−1

4

(d)

2 0 −2 −4 −6 −8 −40 −30 −20 −10 0

10 20 30 40

1227

5

(c)

4 3 2 1 0 0

15 10 (e) 5 0 −5 −10 −15 −20 −25 −30 −35 −8 −6

1

2 3 4 5 Time initiator

6

7

−1

0

1

2

3

4

5

6

10 5 (f) 0 −5 −10 −15 −20 −4 −2 0 2 Rank distance

4

6

−25

−3

−2

−1

0

1

Figure 1. Linear regression plots and equations for time matching (a) in mangabeys, (b) in vervet Donga group, (c) in vervet Picnic group and for influence of status (d) in mangabeys, (e) in vervet Donga group, (f) in vervet Picnic group. Time discrepancies: time invested by the receiver  time invested by the initiator. Rank distances: rank of the receiver  rank of the initiator. Each data point is standardized, which means that the time axis does not have units and that some grooming times may appear to be negative. (a) Y ¼ 0.42 þ 0.46X. (b) Y ¼ 5.586E  8 þ 0.942X. (c) Y ¼ 6.999E  5 þ 0.961X. (d) Y ¼ 0.051 þ 0.111X. (e) Y ¼ 0.302 þ 3.012X. (f) Y ¼ 0.40 þ 5.233X.

shorter first bouts than infrequent partners (negative value of the effect ‘frequent or not’). The second level of interaction showed that for frequent partners, the length of the first bout could be used to predict the length of the rest of the grooming session (positive value of the combined effects ‘rest*frequent or not’). The post hoc Spearman correlations revealed that the initial bout of frequent groomers was significantly and positively correlated with the length of the rest of the session (rS2 ¼ 0:887, P < 0.0001; Fig. 2a). This contrasted with the lack of significance of the first bout initiated by infrequent groomers of both species (rS2 ¼ 0:009, P ¼ 0.938; Fig. 2b). In vervets, the linear mixed-effect model (summarized in Table 5) also showed a significant effect for the variable ‘frequenteinfrequent partner’. As infrequent partners were on the intercept, the model showed that frequent partners invested in shorter first bouts than infrequent partners (negative value of the effect ‘frequent or not’). The model also showed that the genetic relatedness did not significantly influence the way females invested in grooming and explained the data less well than the ‘frequent partner’ parameter. As in the mangabeys, the second level of interaction showed that for frequent partners, the length of the first bout could explain the rest of the grooming session (positive value

grooming sessions (Table 4). This contradicts the RTS model according to which females should have increased their investment in grooming time gradually in reaction to their partners’ investments. Thus, for both species and contrary to predictions, females neither kept the length of their own bouts constant (parcelling) nor increased their length (RTS) within sessions. Frequent and Infrequent Grooming Partners In both species and in spite of great differences in the number of potential partners, females frequently groomed from two to four other adult females (mangabeys: 3.79  1.09 frequent partners; vervets: Donga: 3.75  1.04 frequent partners; Picnic: 2.04  0.54 frequent partners). Frequent partners were significantly closer in rank than infrequent partners were (two-tailed ManneWhitney U test on means: mangabeys: U ¼ 4084.5, N1 ¼ 83, N2 ¼ 221, P < 0.01; vervets: U ¼ 41.5, N1 ¼ 13, N2 ¼ 14, P ¼ 0.014). In mangabeys, the linear mixed-effect model (summarized in Table 5) showed a significant effect of the parameter frequenteinfrequent partner. As infrequent partners were on the intercept, the model showed that frequent partners started with

Table 3 Parcelling and raising the stakes models: evolution of the length of grooming bouts across grooming sessions (Friedman ANOVA test) and within sessions (sign test) Group

Friedman ANOVA test c2

df

P

No. of bouts (no. of dyads)

Increase across session

Decrease across session

Sign test P

Mangabey

67.126 95.967 204.248 113.744 50.213 64.377 199.453 24.066

2 3 4 5 2 3 4 5

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

3 4 5 6 3 4 5 6

25 26 16 11 17 19 28 8

80 48 66 32 94 25 73 7

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.30 <0.0001 >0.999

Vervet combined probabilities

(105) (74) (82) (43) (111) (44) (101) (15)

Only the categories of grooming sessions that contained at least three bouts and five pairs are represented. For the vervet monkeys, the given probabilities are combined P probabilities for the two groups following the formula c2df ¼ 2 ln P with df ¼ 4 (Sokal & Rohlf 1995). All tests are two tailed.

1228

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

Table 4 Parcelling and raising the stakes strategies: development of grooming length for each partner’s contribution across sessions (Friedman ANOVA test) and within sessions (sign tests for initiators and receivers) Group

Mangabey

Vervet combined probability

Initiator Friedman ANOVA test c2

df

85.045 57.000 141.909 42.413 66.036 24.381 122.640 10.203

1 1 2 2 1 1 2 2

P

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.006

No. of bouts (no. of dyads)

(þ)

()

Initiator

3 (105) 4 (74) 5 (82) 6 (43) 3 (111) 4 (44) 5 (101) 6 (15)

19 11 9 0 15 3 10 3

86 63 73 43 96 41 91 12

Sign test P

Receiver Friedman ANOVA test c2

df

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.03

105.000 46.621 150.178 66.426 115.000 32.818 186.527 14.400

1 1 2 2 1 1 2 2

P

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.01

No. of bouts (no. of dyads) 3 (105) 4 (74) 5 (82) 6 (43) 3 (111) 4 (44) 5 (101) 6 (15)

(þ)

()

Receiver 18 6 11 0 16 1 12 1

87 68 71 43 95 43 89 14

Sign test P <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.01

Only categories of grooming sessions that contained at least three bouts and five pairs are represented. For the vervet monkeys, the given probabilities are combined P probabilities for the two groups following the formula c2df ¼ 2 ln P with df ¼ 4 (Sokal & Rohlf 1995). (þ): increase in grooming length; (): decrease in grooming length. All tests are two tailed.

of the combined effects ‘rest*frequent or not’). Finally, as the model showed that the two groups were not significantly different, we pooled them to perform the post hoc Spearman correlations. As for the mangabeys, the correlations revealed that the initial bout was significantly and positively correlated with the length of the rest of the session for frequent groomers (rS2 ¼ 0:920, P < 0.0001; Fig. 2c), while this was not significant for infrequent groomers (rS2 ¼ 0:066, P ¼ 0.492; Fig. 2d). DISCUSSION Both mangabey and vervet females allotted about 15% of their daily budget to grooming with an average session length above 5 min. This may seem surprising at first sight as mangabeys have many more potential grooming partners than vervets, but the fact that the number of grooming partners and total grooming time are almost independent of group size reflects a general pattern among the primates (reviewed in Dunbar 2010). A closer look at how females exchanged grooming revealed that most of them had preferred partners with which they spent a disproportionate

Table 5 Summary of the linear mixed-effect models for grooming time in mangabeys and vervets Effects Mangabeys SD (intercept: 0.632eresidual: 0.532) Rest Frequent or not Rest*frequent or not Vervets SD (intercept: 12.810eresidual: 49.638) Rest Group Donga Frequent or not Related or not Rest*frequent or not Rest*related or not Frequent or not*related or not

Value

SE

df

t

P

2.040

0.180

257

11.321

<0.0001

0.038 1.607 0.353

0.036 0.291 0.042

257 43 257

1.060 5.520 8.346

0.290 <0.0001 <0.0001

107.963

14.090

219

7.662

<0.0001

0.040 0.032 114.231 12.242 0.531 0.072 36.320

0.039 0.072 21.286 20.756 0.072 0.062 28.726

219 11 11 11 219 219 11

0.925 0.839 5.366 0.590 7.406 1.150 1.264

0.356 0.238 0.002 0.567 <0.0001 0.252 0.232

Mangabey model: first bout w rest*frequent or not j random effect on the intercept: identity of dyad; Akaike’s information criterion: 584.597. Vervet model: first bout w rest*group*frequent or not*related or not j random effect on the intercept: identity of dyad; Akaike’s information criterion: 2549.174. For both models infrequent groomers are included in the intercept, that is, results for ‘Frequent or not’ are for frequent groomers compared to infrequent ones. For the vervet model, both group Picnic and unrelated partners are included on the intercept, that is, results for ‘Group Donga’ and ‘Related or not’ are for the Donga group and related groomers compared to the Picnic group and unrelated groomers, respectively. Significant effects are in bold.

amount of their grooming time. On average, for both species, each female had about two to four frequent partners. These frequent partners were significantly closer in rank than infrequent ones and generally tended to time-match their grooming bouts within sessions. Indeed, when we used all grooming sessions occurring in each group, regression analyses revealed a stronger time-matching effect in small vervet groups than in the much larger mangabey group: in mangabeys frequent groomers were diluted among infrequent groomers, which led to a poor coefficient of regression (0.22), while in vervets, most females were frequent groomers. While we did not have the genetic relatedness of the mangabey females, Hula et al. (J. Hula, R. Pansini, R. Noë & M. Kruetzen, unpublished data) found that in the Donga group, closely ranked vervet females belonged to the same matriline, while in the Picnic group, none of the females were related. In the Donga group, however, kinship could not really predict which dyads were frequent or infrequent groomers. These results can be interpreted in opposite ways. (1) Females are close in rank as a result of mutual support. This is notably expected if they belong to the same matriline, but in principle could also hold for unrelated females. These long-term bonds are cemented by frequent grooming. (2) Females seek closely ranked partners because they prefer having minimal power asymmetries and more symmetrical grooming relationships as a result. The first causeeeffect chain is the one generally accepted in primatology. We cannot differentiate between the two interpretations in the mangabey case as we had insufficient information on the genetic structure of the group, but our results on vervets fit better with the second interpretation. Biological market theory correctly predicted the relationship between power and grooming asymmetries: regression analyses showed that females’ rank distances were correlated with their investment discrepancies. The larger the rank distance, the longer the lower-ranking member of a dyad had to groom relative to the amount of grooming received (Fig. 1). However, we would have expected the effect to be stronger in the large mangabey group than in the small vervet groups, as in mangabeys the power asymmetries in terms of absolute rank distance are larger (up to 34) than in vervets (up to 6 and 3, respectively). We can only offer as an explanation that a female’s hierarchical position does not necessarily say much about the value of tolerance and support she can provide. In fact, the value of tolerance may depend on the possibilities of monopolizing food patches and may even vary with the personalities (bold, shy, aggressive, tolerant, etc.) of the highranking individuals (e.g. Itoh 2002), while the value of support may vary according to the rate of conflicts occurring within the group. As for power differentials, they may depend on the steepness of the rank order in each species. In this sense, it is interesting that in mangabeys both dominant and subordinate partners initiated

Rest of the session length (min)

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

10 9 (a) 8 7 6 5 4 3 2 1 0

1229

12 (b)

10 8 6 4 2 0.5

1

1.5

2

2.5

3

3.5

4

4.5

0

1

2

3

4

5

6

Rest of the session length (s)

First bout length (min) 1000 900 800 700 600 500 400 300 200 100 0

1000 900 800 700 600 500 400 300 200 100

(c)

50

100

150

200

250

0 50 300 350 First bout length (s)

(d)

100

150

200

250

300

350

400

Figure 2. Spearman correlation plots for (a, b) mangabeys and (c, d) vervets. For each species, the correlations were done for (a, c) frequent and (b, d) infrequent grooming partners. For vervets, we plotted the two study groups together as the linear mixed-effect model considered they were statistically not different.

grooming sessions, suggesting a more egalitarian social system, while in vervets most of the interactions were started by the lowerranking member of the dyad. Our search for possible partner control strategies revealed that both sooty mangabeys and vervet monkeys failed to parcel their grooming bouts (in the sense of the parcelling model) or to raise the stakes either in answer to their partners’ grooming investment or to their own previous investment. Instead, our results showed that females invested in longer bouts at the beginning of the sessions, contrary to what one would expect if exploitation prevention or trust-building strategies were used. In species with female philopatry such as mangabeys and vervets, in which females have wellestablished relationships, trust building may not be apparent over short time frames such as single grooming sessions. We could not test trust building over longer time frames since it was impossible to determine the start of relationships (cf. Barrett & Henzi 2000). We therefore chose another approach by analysing the difference between frequent and infrequent grooming partners. Contrary to predictions, infrequent groomers, who may never reach the same level of trust as frequent groomers, did not invest in shorter initial grooming bouts than frequent groomers did. However, the first bout they gave at the beginning of a session could not predict the length of the session that followed. In contrast, the first bout predicted the length of the session for frequent groomers. Its length was directly correlated with the length of the rest of the grooming session and thus long first bouts predicted long sessions. This finding shows that females have a good knowledge of the quality of their relationships with others and may hint at how females choose between partners. Biological market theory predicts that animals base their preferences on past experiences with multiple partners that reach back deeper in the past than either parcelling or RTS (see Schino & Pellegrini 2009 for a discussion of attitudinal book keeping and Fruteau et al. 2009 for attitudinal partner choice). A relatively long grooming bout may also be a means to negotiate quickly the terms of the interaction by making a first bid. It would have been interesting to investigate this further by testing whether the length of the first bout also predicts whether the partner would reciprocate at all. Unfortunately, for both species, we had too few unreciprocated bouts to run the necessary logistic regressions.

To summarize, in both mangabeys and vervets, neutral dyadic grooming sessions exchanged between females showed that grooming was mainly traded for itself on a short time frame, which supports the notion of low cost positive social currency based on short term causal relationships introduced by Barrett et al. (1999). However, grooming was also strongly affected by dominance, partner choice and whether individuals were frequent or infrequent partners, which supports the idea of the knowledge of longterm relationships between partners (Seyfarth & Cheney 1984). We can conclude that grooming is an exchangeable commodity whose value can fluctuate according to both short-term and long-term relationships existing among all the partners comprising the market. Acknowledgments We thank the Ministère de la Recherche Scientifique, the Ministère de l’Agriculture et des Ressources Animales, the CSRS, the CRE and the Taï Monkey Project in Ivory Coast and the Mpumalanga Parks Board, UNISA and ABEERU colleagues, L. Brown, L. Barrett, S. P. Henzi and R. Bshary in South Africa for research permission and logistic support. We thank G. Gha, R. Peho and F. Range in Ivory Coast and S. Aubel, A. Barrett, A. Brotz, D. Carter, V. Dufour and E. van de Waal in South Africa for assistance in the field. We thank N. Poulin for statistical advice. Max-Planck-Institut fuer Verhaltensphysiologie, CNRS and NWO (Evolution and Behaviour 051-12-036) provided financial support. References Altmann, J. 1974. Observational study of behavior: sampling methods. Behaviour, 49, 227e267. Barrett, L. & Henzi, S. P. 2006. Monkeys, markets and minds: biological markets and primate sociality. In: Cooperation in Primates and Humans (Ed. by P. M. Kappeler & C. P. Van Schaik), pp. 209e232. Berlin: Springer. Barrett, L., Henzi, P., Weingrill, T., Lycett, J. E. & Hill, R. A. 1999. Market forces predict grooming reciprocity in female baboons. Proceedings of the Royal Society B, 266, 665e670. Barrett, L., Henzi, P., Weingrill, T., Lycett, J. E. & Hill, R. A. 2000. Female baboons do not raise the stakes but they give as good as they get. Animal Behaviour, 59, 763e770.

1230

C. Fruteau et al. / Animal Behaviour 81 (2011) 1223e1230

Barrett, L., Gaynor, D. & Henzi, P. 2002. A dynamic interaction between aggression and grooming reciprocity among female chacma baboons. Animal Behaviour, 63, 1047e1053. Bergmüller, R., Johnstone, R. A., Russell, A. F. & Bshary, R. 2007. Integrating cooperative breeding into theoretical concepts of cooperation. Behavioural Processes, 76, 61e72. Brosnan, S., Silk, J., Henrich, J., Mareno, M., Lambeth, S. & Schapiro, S. 2009. Chimpanzees (Pan troglodytes) do not develop contingent reciprocity in an experimental task. Animal Cognition, 12, 587e597. Bshary, R. & Bronstein, J. L. 2004. Game structures in mutualistic interactions: what can the evidence tell us about the kind of model we need? Advances in the Study of Behavior, 34, 59e101. Bshary, R. & Noë, R. 2003. Biological markets: the ubiquitous influence of partner choice on the dynamics of cleaner fisheclient reef fish interactions. In: Genetic and Cultural Evolution of Cooperation (Ed. by P. Hammerstein), pp. 167e184. Cambridge, Massachusetts: MIT Press. Bshary, R., Grutter, A. S., Willener, A. S. T. & Leimar, O. 2008. Pairs of cooperative cleaner fish provide better service quality than singletons. Nature, 455, 964e967. Connor, R. C. 1995. Impala allogrooming and the parcelling model of reciprocity. Animal Behaviour, 49, 528e530. Dufour, V., Pelé, M., Neumann, M., Thierry, B. & Call, J. 2009. Calculated reciprocity after all: computation behind token transfers in orangutans. Biology Letters, 5, 172e175. Dunbar, R. I. M. 2010. Brain and behaviour in primate evolution. In: Mind the Gap. Tracing the Origins of Human Universals (Ed. by P. M. Kappeler & J. B. Silk), pp. 315e330. Berlin: Springer. Fruteau, C., Voelkl, B., vanDamme, E. & Noë, R. 2009. Supply and demand determine the market value of food providers in wild vervet monkeys. Proceedings of the National Academy of Sciences, U.S.A., 106, 12007e12012. Fruteau, C., van de Waal, E., vanDamme, E. & Noë, R. 2011. Infant access and handling in sooty mangabeys and vervet monkeys. Animal Behaviour, 81, 153e161, doi:10.1016/j.anbehav.2010.09.028. Gumert, M. D. 2007a. Grooming and infant handling interchange in Macaca fascicularis: the relationship between infant supply and grooming payment. International Journal of Primatology, 28, 1059e1074. Gumert, M. D. 2007b. Payment for sex in a macaque mating market. Animal Behaviour, 74, 1655e1667. Hamilton, W. D. 1964. The genetical evolution of social behaviour. I & II. Journal of Theoretical Biology, 7, 1e52. Hammerstein, P. 2003. Why is reciprocity so rare in social animals? A protestant appeal. In: Genetic and Cultural Evolution of Cooperation (Ed. by P. Hammerstein), pp. 83e93. Cambridge, Massachusetts: MIT Press. Hemelrijk, C. K. & Ek, A. 1991. Reciprocity and interchange of grooming and ‘support’ in captive chimpanzees. Animal Behaviour, 41, 923e935. Henzi, S. P. & Barrett, L. 2002. Infants as a commodity in a baboon market. Animal Behaviour, 63, 1e7. Henzi, S. P., Barrett, L., Gaynor, D., Greeff, J., Weingrill, T. & Hill, R. A. 2003. Effect of resource competition on the long-term allocation of grooming by female baboons: evaluating Seyfarth’s model. Animal Behaviour, 66, 931e938. Hinde, R. A. 1976. Interactions, relationships and social structure. Man, 11, 1e17. Hutchins, M. & Barash, D. P. 1976. Grooming in primates: implications for its utilitarian function. Primates, 17, 145e150. Itoh, K. 2002. Personality research with non-human primates: theoretical formulation and methods. Primates, 43, 249e261. Keller, L. & Reeve, H. K. 1998. Familiarity breeds cooperation. Nature, 394, 121e122. Keverne, E. B., Martensz, N. & Tuite, B. 1989. Beta-endorphin concentrations in cerebrospinal fluid of monkeys are influenced by grooming relationships. Psychoneuroendocrinology, 14, 155e161.

Kutsukake, N. & Clutton-Brock, T. H. 2008. The number of subordinates moderates intrasexual competition among males in cooperatively breeding meerkats. Proceedings of the Royal Society B, 275, 209e216. Kutsukake, N. & Clutton-Brock, T. H. 2010. Grooming and the value of social relationships in cooperatively breeding meerkats. Animal Behaviour, 79, 271e279. Leimar, O. & Hammerstein, P. 2010. Cooperation for direct fitness benefits. Philosophical Transactions of the Royal Society B, 365, 2619e2626. Manson, J. H., Navarrete, C. D., Silk, J. B. & Perry, S. 2004. Time-matched grooming in female primates? New analyses from two species. Animal Behaviour, 67, 493e500. Melis, A. P., Hare, B. & Tomasello, M. 2008. Do chimpanzees reciprocate received favours? Animal Behaviour, 76, 951e962. Noë, R. 1990. A Veto game played by baboons: a challenge to the use of the Prisoner’s Dilemma as a paradigm for reciprocity and cooperation. Animal Behaviour, 39, 78e90. Noë, R. & Hammerstein, P. 1994. Biological markets: supply and demand determine the effect of partner choice in cooperation, mutualism and mating. Behavioral Ecology and Sociobiology, 35, 1e11. Noë, R. & Hammerstein, P. 1995. Biological markets. Trends in Ecology & Evolution, 10, 336e339. Queller, D. C. & Goodnight, K. F. 1989. Estimating relatedness using genetic markers. Evolution, 43, 258e275. Range, F. & Noë, R. 2002. Familiarity and dominance relations among female sooty mangabeys in the Taï National Park. American Journal of Primatology, 56, 137e153. Roberts, G. 1998. Competitive altruism: from reciprocity to the handicap principle. Proceedings of the Royal Society B, 265, 427e431. Roberts, G. & Sherratt, T. 1998. Development of co-operative relationships through increasing investment. Nature, 394, 175e179. Sachs, J. L., Mueller, U. G., Wilcox, T. P. & Bull, J. J. 2004. The evolution of cooperation. Quarterly Review of Biology, 79, 135e160. Schino, G. 2007. Grooming and agonistic support: a meta-analysis of primate reciprocal altruism. Behavioral Ecology, 18, 115e120. Schino, G. & Aureli, F. 2010. Primate reciprocity and its cognitive requirements. Evolutionary Anthropology, 19, 130e135. Schino, G. & Pellegrini, B. 2009. Grooming in mandrills and the time frame of reciprocal partner choice. American Journal of Primatology, 71, 883e888. Schino, G., Polizzi di Sorrentino, E. & Tiddi, B. 2007. Grooming and coalitions in Japanese macaques (Macaca fuscata): partner choice and the time frame of reciprocation. Journal of Comparative Psychology, 121, 181e188. Seyfarth, R. M. & Cheney, D. L. 1984. Grooming, alliances and reciprocal altruism in vervet monkeys. Nature, 308, 541e543. Silk, J. B. 2007. Social components of fitness in primate groups. Science, 317, 1347e1351. Sokal, R. R. & Rohlf, F. J. 1995. Biometry. 3rd edn. Basingstoke: W. H. Freeman. Trivers, R. L. 1971. The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35e57. de Waal, F. B. M. 1997. The chimpanzee’s service economy: food for grooming. Evolution and Human Behavior, 18, 375e386. Wang, X. 2002. An estimator for pairwise relatedness using molecular markers. Genetics, 160, 1203e1215. West, S. A., Griffin, A. S. & Gardner, A. 2007a. Social semantics: altruism, cooperation, mutualism, strong reciprocity and group selection. Journal of Evolutionary Biology, 20, 415e432. West, S. A., Griffin, A. S. & Gardner, A. 2007b. Evolutionary explanations for cooperation. Current Biology, 17, 61e672. Whitmore, T. C. 1990. An Introduction to Tropical Rain Forests. Oxford: Oxford Clarendon Press. Zamma, K. 2002. Grooming site preferences determined by lice infection among Japanese macaques in Arashiyama. Primates, 43, 41e49.