Dynamics of Female-Female Relationships in Wild Cebus capucinus: Data from Two Costa Rican Sites

Transcription

1 International Journal of Primatology, Vol. 20, No. 5, 1999 Dynamics of Female-Female Relationships in Wild Cebus capucinus: Data from Two Costa Rican Sites Joseph H. Manson, 1 Lisa M. Rose, 2 Susan Perry, 1 and Julie Gros-Louis 3 Received August 3, 1998; revised March 16, 1999; accepted April 5, 1999 Three questions about female-female relationships remain largely unexplored in female-philopatric platyrrhines. First, to what extent is female dominance status dependent on coalitional support? Second, how stable are female dominance hierarchies over multiyear periods? Third, what is the role of allogrooming in servicing long-term social relationships? We addressed these questions using data collected on Cebus capucinus at Lomas Barbudal and Santa Rosa, over a six-year period. Most female-female coalitions against females reinforced the existing dominance hierarchy, but such coalitions were 3-7 times more frequent at Lomas Barbudal than at Santa Rosa. The Lomas Barbudal group's female hierarchy was highly stable throughout the six years of observation, whereas all three of the intensively observed Santa Rosa groups experienced frequent dominance reversals accompanied by physical aggression. In the Lomas Barbudal group, grooming tended to be directed up the hierarchy, and more closely-ranked females groomed at higher rates than distantly-ranked females, whereas these patterns were not consistently found at Santa Rosa. At both sites, grooming was evenly balanced within 67% of female-female dyads, and mothers of young infants received more grooming than other females did. Females did not spend more of their time grooming each other in groups containing more females than in groups containing fewer females, but they distributed their grooming less evenly among their female groupmates in the largest observed group of females. 1 Department of Anthropology, University of California, Los Angeles, Los Angeles, California Department of Anthropology, Washington University, St. Louis, Missouri Department of Psychology, University of Pennsylvania, Philadelphia, Pennsylvania /99/ $16.00/ Plenum Publishing Corporation

2 680 Manson, Rose, Perry, and Gros-Louis Some of the intersite differences may be attributable to differences in the rate of female mortality or transfer or both. KEY WORDS: Cebus capucinus; coalitions; dominance; grooming; female social relationships. INTRODUCTION A widespread form of social organization among female primates, originally identified as female-bonded (Wrangham, 1980), includes female philopatry, nepotism, frequent coalitions and a linear dominance hierarchy. Other terms, e.g. resident-nepotistic (Sterck et al., 1997) have been coined for this suite of characteristics, and its ecological underpinnings and adaptive significance have been vigorously debated (van Schaik, 1989; Isbell, 1991; Isbell and van Vuren, 1996; Sterck et al., 1997). It is generally agreed, however, that females compete within female-bonded groups for access to resources, and as a result, dominant females often experience greater reproductive success than that of subordinants (Silk, 1993). Scholars have examined several questions pertinent to female-bonded social organization in considerable depth in cercopithecines, but hardly at all in the two platyrrhine taxa classified as resident-nepotistic: Saimiri sciureus (Mitchell et al., 1991) and Cebus (Janson, 1985, 1988; Robinson and Janson, 1987; Robinson, 1988a, 1988b; O'Brien, 1991, 1993; Fedigan, 1993; Perry, 1996a). These questions include: (1) To what extent is female dominance status dependent on coalitional support, as distinct from individual competitive ability? The role of polyadic interactions, especially coalitions between matrilineal kin, in establishing and maintaining primate female dominance relationships, most notably in rank acquisition by juveniles, is well documented for savanna baboons (Cheney, 1977; Walters, 1980; Pereira, 1989), macaques (Berman, 1980; Silk, 1982; Chapais, 1983; Datta, 1983; Netto and van Hooff, 1986; de Waal and Luttrell, 1988), vervets (Cheney, 1983; Hunte and Horrocks, 1987) and geladas (Dunbar, 1980). Chapais (1992, 1995), Walters (1980) and Cheney (1983) have argued that nepotistic loyalty is not the sole explanation for this process. Chapais (1992, 1995) suggested that high-ranking individuals, whether kin or not, usually form conservative alliances (both alliance partners are individually dominant to the target individual) to maintain their ranks against the challenge of potential bridging alliances (the target is intermediate in rank between the two alliance partners) and revolutionary alliances (both alliance partners are individually subordinate to the target). Bridging and revolutionary alliances are expected to be most likely in small

3 Female-Female Dynamics in Wild Cebus capucinus 681 groups because the cognitive demands imposed by such alliances will be smaller when fewer individuals are involved. (2) How stable are female dominance hierarchies over multiyear periods? This question is closely related to the first question. Given that (a) selection on females has produced a motivation to strive opportunistically for higher status, and (b) alliances are important in rank maintenance, smaller hierarchies should be less stable (Dunbar, 1988; Chapais, 1995). Dunbar (1988) proposed that provisioned macaque populations are characterized by highly stable matrilineal hierarchies because (a) dominant families grow at faster rates and thus acquire numerical advantages over small families and (b) large size buffers dominant matrilines against the effects of random fluctuations in matriline size caused by individual deaths. However, matrilineal revolutions occasionally occur even in large macaque and baboon groups (Koyama, 1970; Chance et al., 1977; Gouzoules, 1980; Samuels and Henrickson, 1983; Ehardt and Bernstein, 1986; Samuels et al., 1987). (3) What role does grooming play in the maintenance and servicing (sensu Dunbar, 1991) of female-female bonds, and how is this role manifest in grooming patterns as they are related to kinship, dominance rank, coalition formation and group size? Furthermore, how stable are dyadic grooming relationships over multiyear periods? Kin biases in grooming patterns have been widely reported for decades (Yamada, 1963; Sade, 1965; Cheney, 1992). Seyfarth (1977) proposed that adult female monkeys compete for the opportunity to interact with high-ranking females that are the most valuable coalition partners, and this competition produces a pattern in which high-rankers receive more grooming than they direct (vervets: Fairbanks, 1980; Seyfarth, 1980; bonnet macaques: Silk, 1982; rhesus macaques: Chapais, 1983; olive baboons: Sambrook et al., 1995), grooming rates are disproportionately high among females with adjacent ranks (chacma baboons: Seyfarth, 1976; vervets: Seyfarth, 1980), and grooming is exchanged for coalitional support (Seyfarth, 1980; contra Fairbanks, 1980; Silk, 1982). An additional prediction from Seyfarth's (1977) model is that grooming will be less reciprocal in dyads of more disparate rank since the subordinate of the dyad will be forced to accept a more unbalanced grooming relationship as the price of the dominant's protection (Silk et al. in prep). The predicted grooming patterns are modified by the attractiveness of young infants to females, which temporarily increases rates of grooming received by mothers of young infants (baboons: Rowell, 1968; Seyfarth, 1976; vervets: Seyfarth, 1980; Maestripieri, 1994). Dunbar (1991) developed further the view that group-living primates usually must service their mutually protective social relationships by means of allogrooming. He hypothesized that the amount of time available for grooming, after deducting time required for travelling, foraging and resting,

4 682 Manson, Rose, Perry, and Gros-Louis sets an upper limit on the sizes of coaltionary cliques, and this in turn may limit overall group size (Dunbar, 1992,1993). The degree to which females can distribute their grooming evenly among their female groupmates may also decrease with increasing group size, reducing group cohesion (Henzi et al., 1997; contra Silk et al. in prep). We address these three sets of questions using data from five groups of wild Cebus capucinus observed over 5-6 year periods at Santa Rosa National Park and Lomas Barbudal Biological Reserve, Guanacaste province, Costa Rica. Previous work at Santa Rosa revealed that males transfer between social groups, whereas most females remain and breed in their natal groups (Fedigan, 1993; Fedigan et al., 1996). At both sites, femalefemale social bonds are stronger than female-male or male-male bonds, and females form linear dominance hierarchies (Fedigan, 1993; Rose, 1994; Perry, 1996a). Polyadic agonistic interactions comprise a large proportion of all agonistic interactions. During a two-year study of one social group at Lomas Barbudal, 43/45 (96%) observed cases of female-female coalitions against females during focal observation were conservative (Perry, 1996a), and 38% of them consisted of the alpha and beta female. Among femalefemale dyads, grooming rate correlated significantly positively with coalition formation rate. Females with more similar dominance ranks groomed at significantly higher rates than females of more disparate rank, and there was a non-significant trend for grooming to be directed from subordinate to dominant females more than vice versa. Using data from multiple study sites affords an opportunity to discover ecological and demographic correlates, and possibly causes, of intraspecific variation in social relationships and social structure. In addition to the three sets of questions formulated above, we compare Lomas Barbudal and Santa Rosa with respect to female-female supplant rate a proposed indicator of feeding competition in baboons (Henzi et al., 1997) and adult female mortality rate. SUBJECTS AND METHODS Lomas Barbudal We collected data on members of a single social group (Abby's Group, henceforth LB Group) of white-faced capuchins in and just north of Lomas Barbudal Biological Reserve, Guanacaste, Costa Rica (center of study group's home range: 10 30' N, 85 22' W). The area consists largely of highly seasonal tropical dry deciduous forest (Frankie et al., 1988) traversed by several permanent streams. The social group was followed from dawn

5 Female-Female Dynamics in Wild Cebus capucinus 683 to dusk every day for up to 25 sequential days per month. Adults were easily identified by scars, hair color patterns, and facial contours. Here we present data collected during Feb.-June 1997, as well as new analyses of data collected during May 1991-May 1993 and presented by Perry (1996a). Over the course of the two study periods, the group contained 5-8 adult females, 3-4 adult males and immatures. We conducted brief censuses in Feb.-Mar. 1994, July-Aug and June-Aug Maternal kinship is known only for individuals born in 1990 or later. We collected data in the form of focal individual follows (Altmann, 1974) during which all social behaviors and the identities of the interactants were narrated onto microcassettes. In , we recorded the times of behavioral events only to within the min, except for onsets and ends of grooming bouts after 30 December 1991 (Perry, 1996a). We did not analyze grooming data collected before then. We recorded females' activities and neighbors at instantaneous samples every 2.5 min. In 1997, we recorded the times of all behavioural events to the sec. We recorded only one instantaneous sample per follow, at the end of each 10 min. Behavioral definitions, and methods of calculating interobserver reliability, are described by Perry (1996a). We determined dominance relationships on the basis of the direction of avoids and cowers in dyadic interactions. An avoid occurs when a monkey responds to the approach of another monkey by leaving or moving from its path while watching it. A cower is when a monkey crouches or leans away from another monkey in response to the its approach or aggressive display. A supplantation occurs when one monkey approaches another and takes its place as it moves away. Santa Rosa We collected data in Santa Rosa National Park, Guanacaste, Costa Rica (park headquarters: 10 50' N, 85 38' W) on four social groups: Los Valles (SR-LV), Cerco de Piedra (SR-CP), Sendero (SR-SE) and Nancite (SR-NC). The 108-km2 park lies within the larger Area de Conservacion Guanacaste, and consists of seasonal tropical dry forest in varying stages of regeneration with patches of old evergreen, riparian and mangrove forest (Janzen, 1983; Fedigan et al., 1996). The habitat is broadly similar to that of Lomas Barbudal, except that drinking water is restricted to a few permanent waterholes during the dry season (mid-december to mid-may), and the home range of one group (SR-NC) includes mangrove forest. All study subjects were individually identified and well habituated to human observers. We collected the focal individual data presented here during January- June 1991 (groups SR-LV, SR-CP and SR-SE), January-August 1995 (SR-

6 684 Manson, Rose, Perry, and Gros-Louis LV and SR-CP) and January-July 1996 (SR-LV, SR-CP and SR-NC). We collected additional demographic and ad libitum behavioral data during Jan-July 1997, and on SR-NC group during Group SR-SE became extinct in 1992, and group SR-NC was habituated as a new study group in Group SR-SE contained 3 adult females, 1 adult male and 4 immatures, and group SR-NC contained 4-5 adult females, 3 adult/subadult males and 8-9 immatures. Group SR-LV contained 4-6 adult females, 3-5 adult males and 6-12 immatures, and group SR-CP contained 2-4 adult females, 1-4 adult males, and 3-7 immatures. Data collection protocol and ethogram components resemble those used at Lomas Barbudal, except that data were collected by a single observer (Rose) and recorded manually in field notebooks. Typically, Rose followed groups from dawn until dusk on a rotating schedule: 10 consecutive days per group in 1991 (Rose, 1994) and 6 consecutive days per group in (Rose, 1997,1998). In 1991, she collected all data in the form of 10-min. focal samples (Altmann, 1974). The data represent equal proportions of 10-minute continuous focal samples and 30-min. focal interval samples. During focal interval samples, Rose recorded data on the minute, and recorded additional behaviors of interest (including all agonistic and coalitionary behaviors) on an all-occurrences basis, except that coalitions were not recorded systematically in Comparison of focal continuous and interval focal data indicated no significant difference in social time budgets, rates or durations of grooming or rates of agonism (Rose, 1998; submitted). As we anticipated a comparative study of social behavior at Lomas and Santa Rosa, data treatment and calculation of dyadic interaction rates followed the procedures established by Perry (1996a). Data Analysis We ascertained dominance relationships on the basis of the directions of avoidance and cowering within each female-female dyad. Two dyads in LB group in 1997 (MM-VV and TT-JJ) and one dyad in SR-LV group in 1997 (DK-FY) had undefined dominance relationships: dominance interactions occurred in both directions between MM and VV, and were not observed at all between TT and JJ or between DK and FY. We calculated dyadic interaction rates as the number of interactions (or number of seconds of grooming) divided by the number of hours during which either female was a focal subject. Because female social interactions, particularly grooming, were affected by the presence of young infants, in some analyses we classified each female as either having or not having a 0-90 day old infant during a particular observation period. During the

7 Female-Female Dynamics in Wild Cebus capucinus observations of LB group, only one infant (EE, to female TT) was born after we began recording onsets and offsets of grooming bouts to the second. We excluded from analysis all grooming bouts involving TT while EE was 0-90 days old, recalculated some of the grooming results reported by Perry (1996a), and classified all LB group females as not having an infant during the observation period. Santa Rosa group-periods SR-CP 1996 and SR-SE 1991 contained two and three females, respectively. Because of these small samples, we excluded these group-periods from analyses of within-group, between-dyad variation in measures of social relationships. We included these groupperiods in other analyses. Perry (1996a) used Hemelrijk's (1990) "MATSQUAR" program to determine whether female white-faced capuchins at Lomas Barbudal exchanged grooming for coalitional support. We used the same technique to address the same question in each Santa Rosa group-period. As a measure of grooming reciprocity within female-female dyads, we used a method developed by Boyd and Silk (in prep; Silk et al., in press). In a dyad composed of animals A and B, A grooms B for a bouts, and B grooms A for b bouts. The total number of bouts, a + b, is equal to n. The cumulative binomial probability of the smaller of a and b in a sample of n events gives the likelihood of obtaining the observed values of a and b if grooming is evenly balanced within the dyad. The reciprocity index, RI, is the ratio of the cumulative binomial probability of the smaller of a and b and the cumulative binomial probability of the larger of a and b. When these values are equal, the probabilities are the same, and the ratio is thus equal to 1. The reciprocity index ranges from 0, when interactions are highly skewed, to 1, when interactions are perfectly balanced within dyads. Unlike group-level measures of reciprocity in social relationships (Hemelrijk, 1990), this index produces a distinct value for each dyad. To determine whether there was any association between female group size, i.e., number of adult females in a group, and the amount of time spent by females grooming other females, we carried out the following analysis. For each of the ten group-periods of observation, we summed for each adult female the rates at which she groomed each other adult female. We then calculated the median of these summed grooming rates among all the females of that group-period. These values indicate the relative amounts of time that females of each group-period devoted to grooming their female groupmates (collectively). We plotted female group size against these values. To quantify how evenly females distributed their grooming among female potential recipients, we used the Shannon-Weiner diversity index H (see Cheney, 1992), which is calculated as:

8 686 Manson, Rose, Perry, and Gros-Louis where p i is the relative proportion of grooming directed toward the ith female. H is larger when females distribute their grooming among a larger proportion of their female groupmates. Because the maximum possible value of H increases with group size, we computed, for each group-period, the grooming diversity ratio (GDR) (Cheney, 1992), defined as the ratio between the observed value of H and the maximum possible value of H, H max. H max is calculated for a group of size n as: RESULTS At Lomas Barbudal, observations during yielded hr of focal follows of nine females (X ± SE = 80.4 ± 20.0 hr/female, range = hr/female). At Santa Rosa, observations during yielded 905 hours of focal follows on 19 females (X ± SE = 47.6 ± 6.2 hr/female, range = hr). Supplant/Avoid Rates At Santa Rosa, mean hourly supplant/avoid rates for female-female dyads (averaged across all dyads in each group-period) range from (Group SR-NC 1996) to 0.04 (Group SR-CP 1995). Across all femalefemale dyads, the mean supplant/avoid rate is Across all groupperiod means, the median rate is This rate is remarkably similar to female-female supplant/avoid rates observed at Lomas Barbudal in both the observation period (0.028) and the 1997 observation period (0.026). Stability of Group Membership and Dominance Relationships Table I compares the 1993 and 1997 dominance hierarchies in the Lomas Barbudal group. In 1992, adolescent DD overtook WW in the hierarchy (Perry, 1996a). All 10 dyadic dominance relationships between fully adult females remained unchanged between 1993 and Two females (MM and VV) that reached adulthood between the two study periods overtook TT between the two observation periods, and immigrant JJ held a low rank in Of the 8 females that were adults in 1991 or that

9 Female-Female Dynamics in Wild Cebus capucinus 687 Table I. Female dominance hierarchy in the Lomas Barbudal group during two observation periods Dominance rank AA SS NN DDa WWb TT 1997 AA SS NN DD MM or VVa VV or MMa TT or JJc JJ or TT afemale matured during this observation period or since the previous observation period. bfemale disappeared during this observation period or between this observation period and the following one. cfemale immigrated during this observation period or since the previous observation period. matured or immigrated between 1991 and 1997, only one (12.5%) disappeared. We observed one apparent attempt by an adult female to surpass a higher-ranked female in the dominance hierarchy. Between Feb. 1 and Feb. 6,1997, fourth-ranked DD exchanged facial threats with third-ranked NN on 13 separate occasions. Six of these conflicts were definitely initiated by DD, whereas only one was definitely initiated by NN. No conflicts appeared attributable to a dispute over food, and only one may have begun as a dispute over access to a social partner. Four of them included chasing, eight included vocal threats and/or screams, and during two of them NN chewed her own tail, a sign of great distress (Perry, 1997). On six of these occasions, alpha female AA joined DD in threatening NN, and on 10 occasions, beta female SS joined DD in threatening or chasing NN. However, on two occasions, SS switched from supporting DD against NN to supporting NN against DD during the course of a single conflict. The frequency of NN- DD aggressive interactions declined dramatically after Feb. 5. We observed no other monkey to support DD against NN after this date, and during conflicts on 19 March and 29 April, SS unequivocally supported NN against DD. In all subsequent DD-NN dyadic dominance interactions after Mar. 19, NN won. Table II shows female dominance hierarchies of the three Santa Rosa groups (SR-LV, SR-CP and SR-NC) for which relevant data were collected during at least two observation periods. Of the 12 females that were adults

10 688 Manson, Rose, Perry, and Gros-Louis Table II. Female dominance hierarchies during four observation periods Dominance rank a. SR-LV group BO GR SG BQ SOb BO CAa BQ KLa,c GRb BO CAb KL BQ BO KL BQ DLa DK or FYc FY or DKc b. SR-CP group LI BC CPb PAb BCb TUa,b LI SEa LI SE LI SE c. SR-NC group (Santa Rosa) CF MNb DE SPa TW DE SP CFb TW DE SP TW MMa afemale matured during this observation period or since the previous observation period. bfemale disappeared during this observation period or between this observation period and the following one. cfemale immigrated during this observation period or since the previous observation period. in 1991 or that matured or immigrated between 1991 and 1996 in the two groups (SR-CP and SR-LV) between 1991 and 1996, 7 (58.3%) had disappeared by The difference in proportion of adult females disappearing between Lomas Barbudal and the Santa Rosa is statistically significant (Fisher's Exact P = 0.05). In SR-NC group, which we observed only in ,2 of 5 (40%) females that were adults in 1995 had disappeared by Female TU in group SR-CP, and females CA and KL in group SR- LV, rose in rank as they reached maturity. In addition, all three Santa Rosa study groups experienced social upheavals among adult females. These apparent status struggles were the only contexts in which female-female aggression included physical fights and wounding. In group SR-NC, CF occupied the alpha position in 1992 and at the beginning of observations in Jan. 1995, but fell to third-ranking by Jan This change was preceded during a two-month period that included CF's increasing spatial peripheralization, the disappearance of her most frequent grooming partner, beta female MN, and the death of CF's five-month old son and her adoption of MN's four-month old son. Female SP, which had just given birth to her

11 Female-Female Dynamics in Wild Cebus capucinus 689 first infant and already overtaken TW in rank, formed frequent coalitions with DE against CF, and eventually they defeated CF in a decisive fight and chase, after which CF was subordinate to both SP and DE. In group SR-LV, alpha female SG's disappearance sometime between September and December 1995 touched off a struggle between beta female BO and gamma female CA (SG's daughter). Both females repeatedly attempted to enlist allies. After two physical fights during which CA was wounded (C. Goforth, personal communication; Rose, personal observation), CA appeared to accept subordinate rank. She disappeared between August and December In group SR-CP, LI and her adult daughter SE engaged in frequent aggression following alpha female BC's disappearance in January The decisive fight was not observed, but in early July, SE had a badly scratched face, after which she behaved submissively toward LI. During observations in 1997, SE was still subordinate to LI. Other dominance rank reversals at Santa Rosa were less well documented. LI was alpha female of SR-CP group in 1991, but was successfully challenged by BC, which became alpha female the following year and retained her status until she disappeared in January In SR-LV group, SG replaced BO as alpha female between 1993 and 1994 (C. Goforth, personal communication; C. Hall, personal communication). Grooming Relationships Lomas Barbudal After excluding grooming bouts involving female TT while her infant EE was 0-90 days old, we obtained similar results to Perry (1996a) with regard to associations among grooming rates, coalition formation, rank distance and relative rank during the observation period. Femalefemale dyads that groomed more formed more coalitions against females (N = 15, T = 0.52, P = 0.007), more closely ranked dyads groomed at higher rates (N= 15,r= 0.48, P = 0.01); and more grooming was directed up the dominance hierarchy than down the hierarchy (Wilcoxon Matched Pairs test, N = 15, T = 26.0, P = 0.05; X ± SE up = 14.2 ± 2.9 sec/hr, X ± SE down = 7.9 ± 1.9 sec/hr). The latter result had not been significant (Perry 1996a) before excluding observations of TT while EE was 0-90 days old. Five adult females were present in both the and 1997 data collection periods. Among the 10 dyads comprised by these females, grooming rate is positively correlated with 1997 grooming rate

12 690 Manson, Rose, Perry, and Gros-Louis (N = 10, T = 0.73, P = 0.003). Adding dyads containing the two females (MM and VV) that were juveniles in and adults in 1997 reduced the strength of this correlation, but it remained statistically significant (N = 21 dyads, T = 0.36, P= 0.02). Among adult females, the directionality of grooming within dyads, i.e., the percentage of dyadic grooming performed by each member of the pair, remained stable between the and 1997 observation periods, with a significant correlation between proportions of grooming in the same direction in each period (N = 9, T = 0.67, P = 0.01). As in , female-female grooming in 1997 occurred at higher rates between females of more similar dominance ranks, expressed as absolute rank difference (N = 28 dyads, T = -0.52, P = ). Within dyads, more grooming was directed up the dominance hierarchy than down the dominance hierarchy (N = 28 dyads, Wilcoxon T = 68.0, P = 0.03; X ± SE up-hierarchy = 14.2 ± 5.9 sec/hr, down-hierarchy = 5.9 ± 2.1 sec/hr). Perry (1996a) reported that female TT was groomed more (22.4 sec/ hour vs. 2.3 sec/hour, averaged over potential groomers) by females during the two months following the birth of her infant than during the two months preceding its birth. Table III shows that the three Lomas Barbudal females that gave birth in 1997 were also groomed more during postpartum focal observations (range = days among the three females) than prepartum focal observations (range = days). When observations of females with 0-3 month old infants were excluded from the analyses, we obtained qualitatively similar results, except that the tendency for grooming to be directed up the dominance hierarchy did not reach statistical significance (P = 0.13), nor did the correlation between directionality of grooming in and 1997 (P = 0.10). Table III. Mean rates of female grooming received (sec/hr) by females before and after giving birtha Group-period SR-LV 1991 SR-LV 1991 SR-CP 1991 SR-LV 1996 SR-LV 1996 LB 1992 LB 1997 LB 1997 LB 1997 Female GR SG PA KL BQ TT AA TT NN Grooming received, prepartum "No birth occurred during the study period for group SR-SE Grooming received, postpartum

13 Female-Female Dynamics in Wild Cebus capucinus 691 Santa Rosa Among all Santa Rosa adult dyads, female-female dyads engaged in more grooming than female-male dyads (Nf-f = 49, Xf-f = 36 ± 4.4 sec/hr; Nf-m = 109, Xf-m = 9 ± 1.4 sec/hr; t = 5.72, P < 0.001) or male-male dyads did (AU, = 37, Xm-m = 2 ± 0.6 sec/hr; t = 7.42, p < 0.001). Fig. 1 shows that this pattern was consistent across group-periods. Perry (1996a) reported similar differences across dyad types in grooming rates at Lomas Barbudal. The high female-female grooming rates in SR-LV group 1991, SR-CP group 1996, and especially SR-LV group 1996 are probably attributable to the presence of two or more young infants in those group-periods. Table III shows that Santa Rosa females, like Lomas Barbudal females, were groomed more during postpartum focal observations (range = 3-31 days among the five females) than prepartum focal observations (range = 5-38 days). One infant, PA's in SR-CP group 1991, lived only three days before being killed by an adult male following a male rank reversal (Rose, 1994). Combining data from Santa Rosa and Lomas Barbudal yields a significant difference between grooming received prepartum and postpar- Fig. 1. Grooming rates (X ± SE) of female-female (S), female-male ( ) and male-male ( ) dyads in 7 group-periods at Santa Rosa.

14 692 Manson, Rose, Perry, and Gros-Louis tum (N = 9, Wilcoxon T = 0.0, P = 0.008; X ± SE prepartum = 8.5 ± 2.1 sec/hr, postpartum = 56.6 ± 15.9 sec/hr). Unlike Lomas Barbudal, at Santa Rosa there was no tendency for closely ranked females to groom each other more than distantly ranked females. In three of six group-periods containing 3 females (SR-CP 1995, SR-LV 1996 and SR-NC 1996), correlations between rank distance and grooming rate are positive, i.e., more closely ranked dyads groomed less than more distantly ranked dyads, but nonsignificant (no P-value < 0.30: Kendall's tau tests). In group SR-LV in 1991 and 1995, there were nonsignificant trends for more closely ranked dyads to groom more than less closely ranked dyads. In group SR-CP in 1991, there was a trend, approaching statistical significance, for more closely ranked dyads to groom more than less closely ranked dyads (N = 6, T = -0.70, P = 0.06). Analyses of whether Santa Rosa females groomed preferentially up or down the female dominance hierarchy produced mixed results. Table IV shows that in two of six group-periods containing >3 females (SR-LV 1995 and SR-CP 1995), grooming within female-female dyads was directed significantly more frequently by the subordinate to the dominant than vice versa. In group SR-NC, there was a nonsignificant trend for grooming to be directed up the hierarchy, while in the other three group-periods, there was a nonsignificant trend for grooming to be directed down the hierarchy. Failure to find consistently significant effects of relative rank on direction of grooming may be partly a result of variation among group-periods in the number and ranks of females with small infants. Indeed, the two groups with significantly more up-hierarchy than down-hierarchy grooming (SR- LV 1995 and SR-CP 1995) were also the only two groups in which no female had a 0-3 month old infant. However, when we removed the effect Table IV. Dyadic dominance relationship and direction of female-female groominga Group-period SR-LV 1991 SR-CP 1991 SR-LV 1995 SR-CP 1995 SR-LV 1996 SR-NC 1996 LB 1997 N (dyads) Subordinate grooms dominant 26.4 ± ± ± ± ± ± ± 5.9 X ± Se Dominant grooms subordinate 41.0 ± ± ± ± ± ± ± 2.1 Wilcoxon T P agroup-years with <3 females (SR-91 and CP-96) a excluded from this analysis. For LB 91-93, see Perry (1996a).

15 Female-Female Dynamics in Wild Cebus capucinus 693 of infants in the other four group-periods by comparing up-hierarchy to down-hierarchy grooming rates in all dyads in which (a) both females or (b) neither female had a 0-3 month old infant, we found no effect of relative rank on grooming rate (N = 11, X ± SE up-hierarchy =21.5 ± 5.4 sec/ hr, X ± SE down-hierarchy = 21.0 ± 6.2 sec/hr, Wilcoxon T = 28.0, P = 0.66). Pooling these data points from different group-periods is justified because, as it happens, this set consisted of 11 unique dyads. There is no association, among group-periods, between direction of grooming (up- vs. down-hierarchy) and presence or absence of a relationship between rank distance and grooming rate. However, there are indications of a relationship between social stability and direction of grooming. In the only case (group SR-LV 1996) in which grooming was nearly significantly (P = 0.12) directed down the hierarchy more frequently than up the hierarchy, the two highest-ranking females (BO and CA) were engaged in a struggle for alpha rank, and the only dyad in which the subordinate groomed more than the dominant was also the group's only dyad that contained neither BO nor CA. Because too few female-female dyads were present in both the 1991 and 1995 observation periods, we cannot examine whether female-female grooming relationships remained stable over several years at Santa Rosa, as they did at Lomas Barbudal. Grooming Reciprocity Among all female-female dyads (counting each dyad separately in each observation period), number of bouts in each direction is positively correlated with grooming rate (sec/h) in that direction (N = 92 dyads; uphierarchy, r = 0.41, P < 0.001; down-hierarchy, r = 0.29, P = 0.006), confirming that bout frequency is a valid measure of grooming frequency. Of the 70 distinct female-females dyads (i.e. counting each dyad only once, regardless of how many years it appeared in our data set) represented by the Lomas Barbudal and Santa Rosa data combined, 4 never groomed during focal observation. Of the remaining 66, 44 (66.7%) did not show a statistically significant grooming imbalance in any year (i.e. the smaller of the two binomial probabilities was never <0.05). Of the cases of grooming imbalance, all but four were instances in which (a) a subordinate groomed a dominant significantly more than vice versa, or (b) a female without a 0-3 month-old infant groomed a mother of a 0-3 month old infant significantly more than vice versa, or both. There is no significant difference between the proportion of Lomas Barbudal dyads (12/29 = 34.5%) and the proportion of Santa Rosa dyads (10/37 = 27.0%) that ever deviated

16 694 Manson, Rose, Perry, and Gros-Louis significantly from grooming balance. Considering each dyad in each year as a separate data point (required because of dominance rank changes between years), there was no overall tendency for adjacently ranked dyads to have higher grooming reciprocity indices than more distantly ranked dyads (N adjacent = 39, N distant = 53, U = 896.5, P = 0.28). Indeed, RI was slightly higher in more distantly ranked dyads (X ± SE = 0.36 ± 0.05) than in adjacently ranked dyads (X ± SE = 0.28 ± 0.05). Restricting the analysis to dyads in which either both females or neither female had a 0-3- month old infant did not change this result. Examining each group-period separately revealed no significant difference in RI between adjacent and distantly ranked dyads in any group-period. Female Group Size and Female-Female Grooming Figure 2 shows that there is no relationship between female group size and amount of time spent by females grooming other females (N = 10 Fig. 2. Percentage of time spent grooming other females as a function of female cohort size in 10 group-periods at Santa Rosa and Lomas Barbudal.

17 Female-Female Dynamics in Wild Cebus capucinus 695 group-periods, T = 0.15, P = 0.54). Controlling for the proportion of females with a 0-3 month old infant did not improve the relationship between these two variables (partial T = 0.10, p > 0.50). Table V shows the grooming diversity ratio (GDR) as a function of number of females in a group-period, excluding SR-CP group in 1996, which contained only two females and for which the Shannon-Weiner diversity index, H, is necessarily zero. There was no tendency for females to distribute their grooming less evenly among their female groupmates as female group size increased from 3 to 6 (N = 8, T = -0.10, P = 0.87). However, the eight females of LB group in 1997 concentrated their grooming on a smaller number of females, producing a much lower GDR than in smaller groups, including the same group in , when it contained 6 females. Coalitions Lomas Barbudal In 1997, only four female-female coalitions against females occurred during focal observation. Pooling data from all females, this corresponds to a rate of coalitions per focal observation hour, compared with coalitions per hour during the observation period. In the 1997 data set, one of the four observed coalitions consisted of DD supporting SS against NN during DD's challenge of NN. The other three were conservative coalitions. Four distinct dyads formed these coalitions. The had significantly higher grooming rates than the 24 female-female dyads that did not form Table V. Female group size and grooming diversity ratio (GDR) by group-perioda Group-period Female group size GDR SR-SE 91 SR-LV 91 SR-CP 91 SR-LV 95 SR-CP 95 SR-LV 96 SR-NC 96 LB LB "Group SR-CP 96 is excluded from this analysis as only 2 adult females were present.

18 696 Manson, Rose, Perry, and Gros-Louis coalitions against females during focal observations (N 1 = 4, N 2 = 24, U = 8.0, P = 0.009). Santa Rosa A total of 124 agonistic interactions involving subjects of focal observation included obvious coalitions against one or more group members. Mean hourly rate of coalition formation is significantly higher for female-female dyads (X ± SE ± 0.010) than for male-female dyads (0.005 ± 0.002; Mann-Whitney U = 1055, P = 0.02) though not higher than for male-male dyads (0.005 ± 0.002; U = 414.5, P = 0.29). Only 7/70 (10%) of adult female-female coalitions during focal observation were formed against adult females. This corresponds to a rate of coalitions per focal observation hour. Of these cases, five were against a victim that ranked lower than both coalition partners, one was against a female of intermediate rank, and one was against a victim that outranked both coalition partners. The most common age-sex class of victims of female-female coalitions was adult males (33/70 = 47% of cases). Of the 33 coalitions, all but one included an alpha female, 13 (39%) consisted of an alpha and a beta female, and 18 (55%) were directed against an alpha male. In contrast to the Lomas Barbudal data set (Perry, 1996a), Santa Rosa females showed only a weak tendency to form coalitions more frequently with the females with which they groomed more frequently, and this tendency disappeared entirely when analysis was restricted to coalitions against adults. Among all female-female dyads (counting each dyad separately in different group-periods), grooming rate is significantly positively correlated with coalition formation rate (N = 34, T = 0.27, P = 0.04). However, K r tests (Hemelrijk, 1990) within each group-period revealed no tendency for grooming to be exchanged for coalitional support: results are significant in only one group-period, SR-NC 1996 (N = 6 dyads, K r = 11, P = 0.05). Considering only the seven observed female-female coalitions against females, we found no association between grooming rate and likelihood of forming these coalitions. We observed all seven coalitions during 1996 in SR-LV and SR-NC groups. Of the 12 female-female dyads in these two group-periods, the four dyads that formed at least one coalition against another female did not groom at higher rates than those that were not observed to form coalitions against other females (N 1 = 4, N 2 = 8, U = 16.0, P = 1.00). Nor was there any association between female-female grooming and formation of coalitions against adult males. However, there is a consistent trend for rate of resting in contact and nearest adult neighbor scores to be positively associated with female-female coalition rate. Among

19 Female-Female Dynamics in Wild Cebus capucinus 697 all female-female dyads, this correlation is significant (N = 34, r = 0.53, P = 0.002), and the same trend exists in all four group-periods with >1 female-female dyads in which we collected nearest adult neighbor data: SR-LV 1995, SR-LV 1996, SR-CP 1995, SR-NC DISCUSSION Our results address the three sets of questions posed for C. capucinus in the Introduction: (1) To what extent is female dominance status dependent on coalitional support, as distinct from individual competitive ability? Female-female conservative alliances (sensu Chapais, 1992) are implicated as an important, though not the only, element enabling dominant females to maintain their rank in our study groups. The frequency of female-female coalitions against females was considerably lower (by a factor of 3-7) at Santa Rosa than at Lomas Barbudal group, but most female-female coalitions against females at Santa Rosa during focal observation were conservative, and the mere possibility of such coalitions may have deterred dominance challenges (Chapais, 1995). Successful dominance challenges sometimes involved revolutionary alliances, e.g., in SR-NC group in 1996, whereas unsuccessful dominance challenges sometimes foundered on the challenger's failure to break up a conservative alliance, e.g., female DD's challenge of NN in LB group in Recent observations of LB group (J. Gros-Louis, unpublished data) suggest that alpha female AA's position was dependent on coalitional support. Her closest associate and coalition partner, beta female SS (Perry, 1996a), disappeared and presumably died sometime between August 1997 and January During January-May 1998, females NN, DD and MM formed several coalitions against AA, and AA no longer occupied a spatially or socially central position in the group. Because we observed no dyadic dominance interactions between AA and NN, DD or MM during this period, it is not entirely clear whether AA has lost her alpha rank. Important questions for future research are (a) the extent to which Cehus capucinus female-female coalitions are formed preferentially between close kin and (b) whether, as a result, kin tend to occupy similar dominance ranks. These are particularly intriguing questions in light of data from Cebus olivaceus (O'Brien and Robinson, 1993), which indicate that social inheritance of dominance rank does not occur as consistently as in female-philopatric cercopithecines, being restricted to the highestranking juveniles. At Lomas Barbudal, the only known adult female kin dyad is AA-MM. In 1997, AA was alpha female whereas MM ranked fifth or sixth out of eight females. Moreover, AA was never observed to support

20 698 Manson, Rose, Perry, and Gros-Louis MM against females that ranked between them, and in fact AA intervened against MM (then a 3-year-old) against other juvenile females on some occasions in 1993 (S. Perry, unpublished data). Our finding that femalefemale grooming at Lomas Barbudal occurred disproportionately often among closely ranked dyads could be interpreted as indirect evidence for a matrilineal dominance hierarchy. If this were the case, however, we would expect to observe low grooming rates between a minority of adjacently ranked females, specifically those dyads at the breaks between matrilines. During the observation period, three of the five adjacently ranked dyads were the three most frequent groomers whereas the other two ranked ninth and tenth in grooming rate (of 15 dyads), suggesting the existence of such grooming breaks. However, the adjacently ranked dyad that ranked ninth in grooming, NN-DD, increased its grooming rate by 50% in the 1997 observation period, when it ranked eighth of 28 dyads in grooming rate. Furthermore, in 1997, the lowest grooming rate among the seven adjacently ranked (rank distance 1 or 1.5) dyads was 16.4 sec/hr, more than double the median rate among all dyads, 6.9 sec/hr. At Santa Rosa, only two dyads (SG/CA and LI/SE) were known kin (mother-daughter pairs). In SR-LV group in 1995, the only group-period in which both females were present as adults, SG and CA ranked first and third, respectively, while in SR-CP group in 1995, LI and SE ranked third and fourth, respectively (Table II). In SR-LV group in 1995, SG and CA groomed at the highest rate among 15 female-female dyads. In SR-CP group in 1995, LI and SE had the second highest grooming rate among six dyads. Average genetic relatedness between females may be lower in capuchin groups than in macaque and savanna baboon groups, partly because female transfer may be more common than in the cercopithecine taxa. Of the five adult females recruited into SR-LV group in , three were immigrants (Table IIa) while only two were maturing natal females. In addition, another female, MY, transferred into SR-LV group in May 1991 (too late to appear as a focal subject in that group-period) and had disappeared by One of four females recruited into LB group during was an immigrant. These observations suggest that designating males as "the [only] dispersing sex" in Cebus capucinus may be a premature overgeneralization (Moore, 1993; Isbell and van Vuren, 1996). Individual competitive ability may also affect female dominance rank. The struggle for alpha rank in SR-LV group in 1996 included one-on-one physical fights between BO and CA. Females AA (in LB group in 1998) and CF (in SR-NC group in 1996) both appeared to be their respective group's oldest females, well past their physical primes, and this factor may have increased their vulnerability to dominance challenges. It is also possible that support, explicit or implicit, from alpha males influenced

21 Female-Female Dynamics in Wild Cebus capucinus 699 female dominance struggles. Although most attempts by females to recruit alpha male aid against other females were ignored in all group-periods, BO was supported in her second fight against CA by alpha male BU in SR-LV group in 1995, and alpha male NO supported LI against SE on two occasions in SR-CP group in (2) How stable are female dominance hierarchies over multiyear periods? White-faced capuchin female hierarchies can be highly stable over multiyear periods, as in our Lomas Barbudal study group. However, females sometimes exploited opportunities to improve their rank, and dominance struggles at Santa Rosa involved unusually intense female-female aggression. Events that triggered dominance struggles include disappearances of (a) an alpha female's principal female associate and (b) an alpha female herself, followed by dominance struggles among former subordinates. These disappearances apparently produced effects similar to those created by Chapais' (1988) experimental removals of female Japanese macaques from captive groups. In general, the relatively high frequency of female dominance reversals in Cebus capucinus is consistent with the small size, compared to some macaque and savanna baboon populations, of capuchin groups (Dunbar, 1988; Chapais, 1995). However, some vervet groups are small yet their female dominance hierarchies are highly stable (Isbell and Young, 1993; Isbell and Pruetz 1998). (3) What role does grooming play in the maintenance and servicing of female-female bonds? Our Lomas Barbudal data (including results presented by Perry, 1996a) are consistent with the view that females service valuable social relationships, including alliances, by allogrooming. Our findings indicate that (a) dyadic grooming relationships, like dominance relationships, were highly stable over a five-year period, (b) more closely ranked females groomed significantly more frequently in 1997, as they had during the observation period, and (c) there were trends in both observation periods (with varying significance depending on whether females with small infants were included in analyses) for subordinates to groom dominants more than vice versa. These findings provide some support for the view (Seyfarth, 1977) that females compete for access to higherranking grooming partners, which are more valuable allies. Conversely, we observed almost no case of overt competition, e.g. supplants, over grooming partners. In addition, Seyfarth's (1977) model implies that females exchange grooming for coalitional support, a pattern we observed at Lomas Barbudal during a period of social tension following a male rank reversal (Perry, 1996a), but which was not apparent in our Santa Rosa groups. Our Santa Rosa data paint a less straightforward picture than our Lomas Barbudal data. At Santa Rosa, as at Lomas Barbudal, female-female grooming relationships were stronger than female-male or male-male

22 700 Manson, Rose, Perry, and Gros-Louis grooming relationships. However, there is only a weak trend (observed in three of six group-periods, with statistical significance in only two of those) for subordinate females to groom dominants more than vice versa. Furthermore, there is no consistent relationship between female-female dyadic grooming rates and either (a) absolute dominance rank distance or (b) coalitional support, except in group SR-NC. Female-female grooming was evenly balanced within most dyads at both Lomas Barbudal and Santa Rosa, and was no more likely to be evenly balanced in adjacently-ranked dyads than in more distantly ranked dyads. Most deviations from balance are attributable to (a) subordinates' tendencies, in some group periods, to groom dominants more than vice versa and (b) the attractiveness of young infants' mothers as recipients of grooming. The latter result, which is also supported by within-female comparisons of prepartum to post-partum grooming received, is consistent with the prevalence of infant handling in Cebus capucinus (Perry, 1996a; Manson, 1999). In groups containing 3-6 females, our subjects distributed their grooming quite evenly among their female groupmates, relative to females of other female-philopatric primate populations. This inference can be illustrated by comparing our Table V to Cheney's (1992) Table 5, which indicates a median GDR of 0.71 among 12 nonbaboon populations of female-philopatric primates, and Table IV of Silk et al. (in press), which indicates a median GDR of 0.61 among seven savanna baboon groups. Because females very rarely participated in intergroup aggression either at Lomas Barbudal (Perry, 1996b) or Santa Rosa (Rose, 1994,1998), our study can be viewed as providing another data point supporting Cheney's (1992) refutation of the prediction (from the model of van Schaik, 1989) that female-female grooming is more egalitarian in primate groups that compete aggressively against other groups. We did not find that females spent more time grooming other females in groups containing more females than in groups containing fewer females, as would be expected if the cross-specific relationship reported by Dunbar (1991) held within Cebus capucinus. However, as in mountain baboons (Henzi et al., 1997), grooming diversity, as measured by GDR, decreased at the largest observed female group size (eight, in our sample). This finding supports Dunbar's (1991, 1992, 1993) hypothesis that primates in larger groups are forced to restrict their grooming to smaller coalitional cliques. Based on this view, group fission would be expected soon after female group size reached this number. As of June 1998,18 months after LB group was first observed containing eight females, it had not fissioned, though female cohort size had been reduced to seven with the disappearance of female SS.

23 Female-Female Dynamics in Wild Cebus capucinus 701 Intersite Differences A summary of differences in the dynamics of female-female dominance and grooming relationships between Lomas Barbudal and Santa Rosa is in Table VI. Based on the similar supplant rates, and the ecological similarity of the sites, there is no reason to infer any substantial difference in feeding competition, though future systematic study of food patch sizes and variation in food intake rates may reveal such differences. Intersite comparison of interindividual spacing patterns would also be valuable in this regard; unfortunately, we measured proximity using different methods at the two sites. Available data do point to one striking difference between the two sites during the study period: more females disappeared in the Santa Rosa groups than in the Lomas Barbudal group. Assuming that female whitefaced capuchins can adjust their social tactics adaptively in response to variation in the forces that regulate adult female mortality (or transfer) rates, e.g., predation and drought, high mortality rates in small groups would reduce the relative superiority, as alliance partners, of dominant females compared to subordinates. Thus, the two principal effects predicted by the Seyfarth (1977) model grooming is directed preferentially up the hierarchy, and more closely ranked dyads groom at higher rates would not be expected. Our failure to find associations between grooming and coalition formation in the Santa Rosa female-female dyads may be attributable to rapidly shifting alliance patterns and opportunistic coalitions during periods of social upheaval. Low levels of intra-group genetic relatedness among females may contribute to social instability. Extending the scope of comparison beyond Cebus capucinus to other capuchin species reveals findings consistent with some of our results, but it also increases the range of unexplained variation in the dynamics of female-female social relationships. Studies of wild Cebus olivaceus (O'Brien, 1993) and captive C. apella (Parr et al., 1997) indicate that femalefemale grooming tends to be directed down the dominance hierarchy. As Table VI. Summary of female-female social relationships at Lomas Barbudal and Santa Rosa Supplant rate Dominance relationships Female-female coalitions against females Relative rank and grooming Rank distance and grooming Association between grooming and coalitional support 0.027/hr Stable Common Lomas Barbudal Up-hierarchy > down-hierarchy Negatively correlated Positive 0.034/hr Unstable Rare Santa Rosa No consistent pattern No consistent pattern None