Determinants and consequences of social structure in a neotropical primate, Cebus olivaceus

MISSING IMAGE

Material Information

Title:
Determinants and consequences of social structure in a neotropical primate, Cebus olivaceus
Physical Description:
xiv, 208 leaves : ill. ; 28 cm.
Language:
English
Creator:
O'Brien, Timothy G., 1955-
Publication Date:

Subjects

Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1990.
Bibliography:
Includes bibliographical references (leaves 197-207).
Statement of Responsibility:
by Timothy G. O'Brien.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001603283
notis - AHM7533
oclc - 23245267
System ID:
AA00003333:00001

Full Text











DETERMINANTS AND CONSEQUENCES OF SOCIAL STRUCTURE
IN A NEOTROPICAL PRIMATE, CEBUS OLIVACEUS


















By

TIMOTHY G. O'BRIEN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA
UNIVERSITY OF FLORIDA LIBRARIES
1990














ACKNOWLEDGEMENTS


I owe a huge debt of gratitude to a number of people

who have assisted and tolerated me over the years. First, I

would like to thank Sr. Tomas Blohm for his efforts on

behalf of conservation in Latin America and for allowing me

to conduct my field studies at Hato Masaguaral. I thank Dr.

John G. Robinson for everything. Dr. Robinson has guided me

throughout my work and has encouraged me every step of the

way. His prior research and background data gave me

opportunities that few biologists have for testing

hypotheses concerning social behavior, and his introduction

to the monkeys of Hato Masaguaral was a splendid experience.

I thank my doctoral committee, Dr. John F. Eisenberg,

Dr. Michael W. Collopy, Dr. Carmine A. Lanciani, and Dr.

Kenneth M Portier, for their encouragement throughout my

program and for their thoughtful reviews of this

dissertation. I thank Mr. John B. Thorbjarnarson and Sr.

Gustavo Hernandez for their their company and friendship at

Masaguaral. I would also like to thank Ms. Ximena Valderrama

for her assistance in the field and for her friendship

during my stay in Venezuela and since my return to Florida.









Many people have given me sound advice, clear opinions,

wild speculations and logistical support during the

development and completion of this project. They are not to

blame for the final product. I thank Dr. Thomas T.

Struhsaker, Ms. Lisa Leland, Dr. H. Franklin Percival, Dr.

Peter Feinsinger, Dr. Jay Whitehead, Dr. Susan K. Jacobson,

Mr. Jeffrey Hardesty, Mr. Clinton Moore, Ms. Mary Moore, Mr.

James P. Reid, Ms Kathryn Reid, Dr. Stuart D. Strahl, Dr.

Thomas O'Shea, Ms. Linda Cox and Sr. Carlos Martinez del

Rio.

I give my deepest thanks to my parents, Mr. Thomas H.

O'Brien and Mrs. Phyllis J. O'Brien, for their unending

patience, support, and understanding during the many years

of schooling and field work.

This research was supported by grants from The Tinker

Travel Fund, Sigma Xi, Friends of the National Zoo, National

Geographic Society grant 3584-87, National Science

Foundation grant BNS 8718377, and a fellowship from the

Organization of American States

Finally, my most heartfelt gratitude goes to Ms.

Margaret F. Kinnaird. She has read and commented on every

draft of this manuscript and has provided invaluable support

and insight during every phase of this study.


iii















TABLE OF CONTENTS



ACKNOWLEDGEMENTS .. ....... ii

LIST OF TABLES . . vi

LIST OF FIGURES . . .. ix

ABSTRACT . . xii

CHAPTER 1 INTRODUCTION. . .. 1

CHAPTER 2 ALLOGROOMING BEHAVIOR AMONG FEMALE WEDGE-
CAPPED CAPUCHIN MONKEYS: DESCRIPTION AND
DETERMINANTS OF AFFILIATIVE BEHAVIOR 5
Introduction . . 5
Study Area . . 12
Methods . . .. .. 12
Results ............... 21
Characteristics of Grooming Sessions .. 21
The Influence of a Newborn Infant 25
Reciprocity in Grooming .... 32
Characteristics of Females . 35
Grooming Relationships. ... 39
Seyfarth's Model . .. 49
Discussion . . 54

CHAPTER 3 THE ROLE OF AGGRESSION IN THE SOCIAL
STRUCTURE OF FEMALE WEDGE-CAPPED CAPUCHIN MONKEYS 62
Introduction . . 62
Methods .. . . 65
Nearest-Neighbor Distances . 70
Effects of Sociodemographic Variables 71
Results .. .......... 74
Association Among Females ... 74
Aggression Among Females ... 84
Individual Variability in Aggression 94
Discussion . . 97

CHAPTER 4 FEMALE-MALE SOCIAL INTERACTIONS IN WEDGE-
CAPPED CAPUCHIN MONKEYS: BENEFITS AND COSTS OF
GROUP LIVING . . .. 105
Introduction . . .. 105
Methods . . 109










Nearest-Neighbor Distances .
Distribution of Interactions .
Effects of Sociodemographic Variables .
Results . .
Distributions of Nearest-Neighbor Distances
and Social Behaviors . .
Variation in Female-Male Interactions .
Determinants of Social Interactions .
Discussion . . .. .
Male Quality and Social Interactions
Female Rank Affects Social Interactions With
Males . .
Comparison with Brown Capuchin Monkeys

CHAPTER 5 INFANT CARE AMONG FEMALE WEDGE-CAPPED
CAPUCHIN MONKEYS: EFFECTS OF AGE, RANK, AND


RELATEDNESS . .
Introduction . .
Methods . .
Results .
Poisson Regression Analysis. .
Residual Analysis .
Discussion . .
Timing of Infant Care .
Effect of Kinship .
Effect of Female Rank .
Effect of Age .. .
Allomaternal Nursing .
Maternal Care and the Evolution
Allomaternal Care .


of
. .


CHAPTER 6 CONCLUSION . .

REFERENCE LIST . .

BIOGRAPHICAL SKETCH . .


112
112
113
115

116
119
129
137
137

142
143


S 146
146
150
S 157
S 162
S 170
176
S 176
S 180
S 182
184
S 186


. 188

. 191


197

208



rr
rr


r
r


r
r















LIST OF TABLES


Table 2.1. Sample of lemurs, New World and Old World
primates indicating relationship between dominance
hierarchy and direction of grooming . 8

Table 2.2. Description of behaviors associated with
grooming sessions .. .. .. 17

Table 2.3. Identification and characteristics of female
wedge-capped capuchin monkeys involved in grooming
sessions . . .. .... 19

Table 2.4. ANOVA results for the effect of group size,
rank, and presence of infant on frequency of
involvement in grooming for adult females. ID is
the female identity. Infant refers to the presence
and age of infant . .. 27

Table 2.5. Spearman rank correlation matrix for
grooming characteristics of individual female
wedge-capped capuchin monkeys. See Table 2.3 for
definitions of variables . .. 40

Table 2.6. ANOVA results for component behaviors in
grooming relationships. ANOVA considered effects
of rank, age and relatedness (see text) and all 2-
way interactions . ... 48

Table 3.1. Distribution of nearest neighbor distances
within and between groups. Standard Deviations
(SD) within each group, and within and between
group variances are given. Distances are
classified for all behaviors and distances during
foraging only . .. 75

Table 3.2. Main effects ANOVA for AI. Mean squares are
calculated and tested using Type III sums of
squares. The number after each variable identifies
characteristics of female 1 and female 2. The
first female is higher ranking .. 81

Table 3.3. Involvement in fights and supplantation
during foraging for a subset of data in which
context was identified for each female. Data are









presented by % for aggressive interactions won by
each female and total involvement. Sample sizes
are given in parentheses ... 85

Table 3.4. Analysis of Deviance for Poisson regression
of sociodemographic variables on frequency of
coalitions, aggression, and supplantation among
female capuchin monkeys in a large group.The
number after each variable identifies the first or
second female of the dyad. First female is higher
ranking than second female. Percent variance is
the amount of variation accounted for by a given
variable .. . 93

Table 3.5. Main effects ANOVA examining the effects of
female identity as aggressor and victim on
supplantation and fights after the effects of
covariation between behaviors, age and rank have
been removed. Dependent variables are the
standardized residuals from the Poisson regression
and independent variables are female identities in
Main group. All F-tests use Type III mean square 96

Table 4.1. Observed distribution of 4 social
interactions (grooming, aggressive, defensive, and
displacement interactions) between female-female
and female-male dyads, and the expected
distributions of behaviors under an assumption of
random interactions weighted by potential number
of interactants, and random interaction weighted
by nearest neighbor distances. Number in
parentheses indicates the number of dyads included
in the observed distribution and the weighting
factor used in the expected distributions 120

Table 4.2. Distribution of female-male social
interactions in large and small groups, expressed
as number per dyad. Distributions consider
interactions with all male and interactions with
dominant male only . .... 128

Table 4.3. Results of Poisson Regression of influence
of sociodemographic variables on male-female
interactions in capuchin monkeys. Values are
percent of total variability accounted for by each
independent variable, total variability accounted
for by the model, and the significance of residual
variation. Dependent variables are frequencies of
displacements, female aggression toward males,
male aggression toward females and grooming
sessions ..... .. . 130


vii









Table 4.4. Number of residuals greater than the mean
standardized residual for female-male social
interactions in capuchin monkeys. Residuals are
classed by female and male rank for 4 behaviors .135

Table 5.1. Frequency of maternal and allomaternal care
by behavior, female age, and female relationship
with infant (mother, sibling, rank) for 2 young
juvenile, 4 old juvenile, 4 young adult and 11
adult females. Infant relationship with female is
indicated as infant of female (INF), sibling
(SIB), and rank of female relative to infants
mother (I>F, I number of potential female-infant dyads in each
class . . 158

Table 5.2. Summary of Poisson regression analysis of
the influence of sociodemographic variables on 4
maternal/allomaternal behaviors for the 1987 data.
A. dataset includes mother-infant interactions, B.
data excludes mother-infant interactions. Values
are the % variance accounted for by each variable,
total variance accounted for by the model, and
significance of residuals. . .163

Table 5.3. Summary of Poisson regression analysis of
the influence of sociodemographic variables on 2
maternal/allomaternal behaviors for the 1978 data.
A. dataset includes mother-infant interactions, B.
data excludes mother-infant interactions. Values
are the % variance accounted for by each variable,
total variance accounted for by the model, and
significance of the residuals ... 169

Table 5.4. Positive residuals for 1988 maternal and
allomaternal behaviors and for 1978 maternal
behaviors. Residuals are classed as between 1 and
2 standard deviations (SD) of the mean and as more
than 2 standard deviations of the mean 172


viii














LIST OF FIGURES


Figure 2.1. Rank and close kin (r=0.5) relationships
among females in 3 study groups. Ranks are
indicated by numbers and Rank declines in
counterclockwise direction. Mother-daughter (M-D)
and sibling (S-S) relationships are indicated by
brackets. In Main group, adult females are
positioned on inner ring and juvenile females are
positioned outside the ring . 15

Figure 2.2. Frequency and mean duration of grooming
sessions by month for female wedge-capped capuchin
monkeys . .. 24

Figure 2.3. Mean and SE of frequency and duration of
grooming sessions for female wedge-capped capuchin
monkeys in large and small groups. Females in
large groups are presented by ranks ...... 29

Figure 2.4. Interactions between membership in a large
versus small group, female rank, and presence of
infant as they affect frequency of grooming for
adult females . ... 31

Figure 2.5. Distribution of reciprocated and
unreciprocated grooming sessions as affected by
the presence of an infant and relative ages of
interactants .... ......... 34

Figure 2.6. Ratio of giving to receiving grooming (in
minutes) as a function of decreasing rank for the
3 study groups. Lines indicate trends in data for
3 groups . . .. .. .37

Figure 2.7. Grooming relationships as illustrated by
number of grooming sessions, average duration and
relative amounts of grooming by each member of a
dyad in White group. Ordering of females as in
Figure 2.1 . . 42

Figure 2.8. Grooming relationships as illustrated by
number of grooming sessions, average duration and
relative amounts of grooming by each member of a








dyad in Splinter group. Ordering of females as in
Figure 2.1 . .. 44

Figure 2.9. Grooming relationships as illustrated by
number of grooming sessions, average duration and
relative amounts of grooming by each member of a
dyad in Main group. Ordering of females as in
Figure 2.1 . . 46

Figure 2.10. Mean frequency of grooming sessions,
support during aggression and approaches between
females of different rank categories 51

Figure 3.1. Matrilineal and dominance relationships
between female wedge-capped capuchin monkeys in 3
study groups. Dominance rank is indicated below
each female's identity and relatedness is
indicated by brackets. M-D indicates mother-
daughter relationships and S-S indicates sibling
relationships . ... 67

Figure 3.2. Distribution of nearest neighbor distances
for all behaviors and while foraging for each
female . . .. 77

Figure 3.3. Distribution of Association Index (AI)
among 3 groups and among rank and relatedness
combinations in Main group .. 79

Figure 3.4. Maximum Association Index (AI) scores among
pairs of females in 3 groups. a) Maximum spanning
tree for AI scores. Boxes indicate relatedness of
r=0.5. b) Ranked maximum AI scores for all females
in study Females in boxes are from small groups.
AI=2.1 is used as middle reference line; scores
greater than 2.1 indicate at least 1 close
association . . 83

Figure 3.5. Distribution of aggression for a) 2 small
groups and b) 1 large group of wedge-capped
capuchin monkeys . ... 87

Figure 3.6. Distribution of Aggression Index (AgI)
among 3 groups and among rank and age combinations
of Main group . ... 90

Figure 4.1. Distribution of nearest neighbor distances
for female capuchin monkeys expressed as
proportion of observations at 3 distance classes.
Proportion among sexes is the distribution of all
observations. Proportion between sexes is the








distribution of female-female observations and
female-male observations calculated separately 118

Figure 4.2. Distribution of social interactions between
females and adult and subadult males by female
age. All frequencies are expressed as number of
interactions per dyad . .. 123

Figure 4.3. Indices of association and aggression
between females and high, middle and low ranking
males . . 125

Figure 4.4. Distribution of social interactions between
females and males by female and male rank. All
frequencies are expressed as number of
interactions per dyad . .. 132

Figure 5.1 Mean frequency, per dyad, of a) 3 maternal
behaviors and b) 3 allomaternal behaviors in 3
month time intervals. . .160

Figure 5.2 Percent of time infants spent nursing, being
carried by mother and independent of mother in the
first 9 months of life for 2 captive infants 179














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DETERMINANTS AND CONSEQUENCES OF SOCIAL STRUCTURE
IN A NEOTROPICAL PRIMATE, CEBUS OLIVACEUS

By

Timothy G. O'Brien

August 1990


Chairperson: Dr. John G. Robinson
Major Department: Forest Resources and Conservation
(Wildlife and Range Sciences)

I describe and compare social interactions of female

wedge-capped capuchin monkeys (Cebus olivaceus) with

infants, with males, and with other females. The purpose of

this study is to determine the relative importance of social

and demographic parameters to the evolution of female social

behavior by comparing patterns of affiliative and aggressive

behavior between females and different classes of

interactants. The general conclusion is that group

membership and individual differences among females exert

the most pervasive effects on social behavior and social

structure. Females in small groups minimize interactions

with other group members. Rates of association, grooming and

aggression were much lower in small groups for all classes

of interactants except infants. Female-female interactions


xii








were affected by rank, relatedness, and degree to which a

female withdrew from the social network of the group. Low-

ranking females withdrew to the periphery of the group to

avoid aggression. These females also withdrew from

affiliative interactions, contrary to expectation.

Reciprocal grooming sessions were concentrated among high-

ranking and adult females. Low-ranking females did not

reciprocate grooms, and adult females did not reciprocate

grooms by juveniles. Aggressive behaviors were directed by

high-ranking females toward lower ranking females.

Supplantations occurred primarily during foraging periods,

whereas fights occurred in nonforaging context as well.

Interactions with subadult and adult males were most

strongly affected by dominance ranks of males and females.

Dominant males were preferred grooming partners and females

associated more with this male. Females were aggressive

toward subadult males, especially higher-ranking subadult

males. Male aggression toward females targeted middle-

ranking and young-adult females. Allomaternal interactions

were more frequent, on average, in the large group. Female

rank and relatedness to infant were major determinants of

allomaternal carrying, association, and investigation, but

female age also influenced participation. Allomaternal

nursing appeared random. In large groups, increased

intragroup competition increased the need to defuse group

tension through assertion of dominance, affiliation and


xiii








cooperation. Females direct affiliative and cooperative

behaviors toward valuable group members, those that are

related or high-ranking. Females are aggressive toward group

members of low value, such as peripheral females and

subadult males.


xiv
















CHAPTER 1

INTRODUCTION

How social relationships develop and are maintained in

group-living species is a major question in the evolution of

sociality (Eisenberg 1966; Alexander 1974; Wilson 1975;

Wittenberger 1981; Trivers 1985). Social living may evolve

to maximize advantages of groups for individuals (resource

acquisition: Wrangham 1980, 1983; Koenig and Mumme 1987,

avoiding predation: Hamilton 1971; van Schaik 1983) or it

may be imposed as a consequence of habitat limitations (lack

of breeding sites: Woolfenden and Fitzpatrick 1984; Brown

1988), or demography (minimizing dispersal mortality: Brown

1988). Because group living imposes costs on individuals

such as competition for resources or mates, increased

probability of disease transmission or increased

conspicuousness to predators, group living can only be

expected when advantages exceed the detriments. Once groups

form, social behavior may evolve to enhance the original

advantages of group living or to regulate competitive

interactions within the group, as well as between groups

(Alexander 1974). Structure based on dominance hierarchies,

kinship, or reciprocity may serve to reduce aggression or










promote cooperation in a group when all animals recognize

the organizing principles. Social structure is mediated

through social behaviors that serve to form bonds or

establish relative status between individuals in a group.

The social organization of many species of primates

consists of permanent associations of females with males

competing for access to or control of these females

(Eisenberg et al. 1972; Greenwood 1980; Wrangham 1980).

These female-bonded groups are characterized by stable

relationships among individuals in definable classes of

social rank, relatedness, age, etc. (Crook 1970; Dittus

1977; Hinde 1983). The social relationships among group

members, as reflected in the type, frequency and quality of

behaviors, define the social structure of the group (Hinde

1979). Social interactions are influenced by a number of

demographic and structural effects. Survival, mortality, and

fecundity schedules affect group size, age-sex structure,

and patterns of kinship in a group (Keyfitz 1977; Altmann

and Altmann 1979). The presence or absence of certain age-

sex classes, and the number of individuals in a given class

can affect the expression and frequency social interactions.

Patterns of relatedness among group members may also

affect social interactions (Hamilton 1964; 1972) in many

species of primates (Sade 1965; Kurland 1977; Silk 1982;

Dunbar 1987; but see Jay 1963; Hrdy 1977). Individuals that

direct cooperative and affiliative behavior toward close kin










rather than unrelated animals are likely to increase

inclusive fitness because related animals are more likely to

share genes (Brown 1988). The degree to which kinship will

affect expression of social behavior will depend on the

cost:benefit ratio for the action and the degree of

relatedness between the interactants. For example, support

of a related individual during aggression may not be costly

for a high ranking female, but it might be very costly for a

low ranking female.

Relationships developed among certain individuals can

influence patterns of social interactions. Social

relationships developed among juvenile members of a group

may persist into adulthood and regulate adult social

interactions (Cheney 1977; Hausfater et al. 1982).

Competition for access to desirable social partners may

result because animals may vary in their abilities to secure

resources or to supply benefits to a partner (Seyfarth 1977;

Seyfarth et al. 1978). Competition for partners and the

relationships among partners should regulate social

interactions within the group. Seyfarth et al. (1978) have

demonstrated that, among several Old World primate species,

females exhibit grooming patterns that result from

competition for high-ranking partners.

This study investigates the determinants of social

structure among female wedge-capped capuchin monkeys, Cebus

olivaceus, Neotropical, arboreal primates. I consider the










relative importance to social structure of dominance rank,

relatedness, group size, and demographic characteristics of

individuals, such as age and reproductive value. These

sociodemographic variables have been proposed to explain the

evolution of social structure in primates. I address the

question of how much variation observed in social behavior

is attributable to particular parameters. I examine

affiliative and agonistic social interactions among females,

between females and males, and between females and infants.

For each class of interactants, I ask what parameters

consistently explain the suite of behaviors considered. I

attempt to assess the relative importance of a parameter by

determining how much of the variability in social behaviors

can be accounted for by that parameter. If a variable is

important in the evolution of social structure, it should be

reflected over a range of behaviors. Finally, I evaluate the

importance of individual differences among females in the

patterns of social behavior.
















CHAPTER 2

ALLOGROOMING BEHAVIOR AMONG FEMALE WEDGE-CAPPED CAPUCHIN
MONKEYS: DESCRIPTION AND DETERMINANTS OF AFFILIATIVE
BEHAVIOR.


Introduction

How social relationships develop and are maintained in

group-living species is a major question in the evolution of

sociality (Alexander 1974; Wilson 1975; Wittenberger 1981;

Trivers 1985). Social living may evolve to maximize

advantages of groups for individuals (resource acquisition,

Wrangham 1980, 1983; Koenig and Mumme 1987, avoiding

predation, van Schaik 1983; Hamilton 1972) or it may be

imposed as a consequence of habitat limitations (lack of

breeding sites, Woolfenden and Fitzpatrick 1984; Brown

1988), or demography (minimizing dispersal mortality, Brown

1988). Because group living imposes costs on individuals

such as competition for resources or mates, increased

probability of disease transmission or increased

conspicuousness to predators, group living can only be

expected when advantages exceed the detriments. Once groups

form, social behavior may evolve to enhance the original

advantages of group living or to regulate competitive










interactions within the group, as well as between groups

(Alexander 1974). Structure based on dominance hierarchies,

kinship, or reciprocity may serve to reduce aggression or

promote cooperation in a group when all animals recognize

the organizing principles. Social structure is mediated

through social behaviors that serve to form bonds or

establish relative status between individuals in a group.

Allogrooming behavior is a form of social communication

in which relationships are developed and maintained (Sparks

1967; Seyfarth 1977; Walters and Seyfarth 1986).

Allogrooming may serve as appeasement behavior by which

individuals reduce tension or redirect aggression (Marler

1965; Sparks 1969). In many birds and mammals (Eisenberg

1962, 1981; Harrison 1965; Sparks 1967; Gaston 1977; Kaufman

1983) allogrooming serves to maintain dominance

relationships; the dominant individual grooms. Wilson (1975)

interprets the solicitation to groom as a submissive posture

and Kaufman (1983) suggests that allopreening allows

individuals to avoid potentially aggressive face-to-face

encounters. Conclusions concerning grooming as a dominance

interaction are that grooming should occur among species

that are characterized by high levels of aggression and

dominance hierarchies, that grooming should be directed down

the hierarchy, that grooming should not be reciprocated, and

that subordinate animals should withdraw from or terminate

grooming. Long term consequences of grooming, such as










support during aggression, or correlation with other social

behaviors are not predicted by a dominance model.

Among primates, the direction of grooming often is

reversed; grooming is directed toward dominant animals

(Table 2.1). This appears to be the typical pattern among

female macaques and baboons (Family: Cercopithecidae). In

some New World primates, however, females direct grooming

down the dominance hierarchy (spider monkey, Ateles

geoffrovi and mantled howler monkey, Alouatta palliata).

Evidence exists to support the hypothesis that grooming

serves to maintain dominance relations through appeasement.

In Cebus albifrons, a reciprocal relationship between

grooming and aggression was interpreted as evidence of the

role of grooming in maintaining social structure (Berstein

1965), supporting Wilson's (1975) observation that in short

term interactions, aggression and grooming are inversely

related. Schino et al. (1988) provide behavioral evidence

that grooming reduces tension (as measured by a reduction in

displacement activities in Java macaques (Macaca

fascicularis) and Boccia (1987) provided preliminary

physiological evidence that receiving grooming reduced

tension as measured by heart rate monitoring. Appeasement,

however, is not sufficient to explain grooming patterns

other than dominance relations, nor does appeasement

contribute to understanding social interactions that are

correlated with grooming.













Table 2.1. Sample of lemurs, New World and Old World primates
indicating relationship between dominance hierarchy and direction
of grooming.


Female
dominance
Species hierarchy


Direction
of
grooming


Reference


Indri indri
Lemur catta

Alouatta seniculis
Alouatta palliatta
Ateles geoffrevi

Saimiri sciureus

Saquinus oedipus
Cebus olivaceus
Colobus quereza
Colobus badius
Presbytis entellus
Erythrocebus patas


Cercopithecus ascanius
Cercopithecus campelli
Cercopithecus mitis
Cercopithecus aethiops

Cercocebus galeritus

Macaca arctoides
Macaca fuscata
Macaca mulatta

Macaca radiata
Papio ursinus
Papio hamadryas
Papio cynocephalus
Theropithecus qelada
Gorilla gorilla

Pan troglodytes
Pan paniscus


yes
yes

yes
yes
yes

none

yes
yes
none
none
yes
?n


none
none
none
yes

yes

yes
yes
yes

yes
yes
yes
yes
yes
yes

yes
none


?
none

?d
down
down

none

?
down
none
none
none
none


none
none
none
up

up

up
up
up

up
up
up
up
up
none

?
none


Pollock 1979
Budnitz and
Dainis 1975
Saavedra 1984
Jones 1979
Eisenberg and
Keuhn 1966
Baldwin and
Baldwin 1981
Dawson 1978
this study
Oates 1977
Struhsaker 1975
Hrdy 1977
Hall 1965
Rowell and
Hartwell 1978
Cords 1986
Cords 1986
Cords 1986
Seyfarth 1980
Fairbanks 1980
Kinnaird pers.
comm.
Rhine 1972
Kurland 1977
Sade 1972
Seyfarth 1977
Silk 1982
Seyfarth 1977
Stammbach 1978
??
Kummer 1975
Stewart and
Harcourt 1986
Nishida 1979
Kuroda 1980










Allogrooming also may act as currency in reciprocal

relationships with long-term consequences (Seyfarth and

Cheney 1984; Walters and Seyfarth 1986). A theory based on

reciprocity incorporates a broader array of relationships

than a theory based only on appeasement and is especially

useful for long-lived social species such as primates.

Grooming relationships in primates may reflect dominance or

sexual relationships (male-female grooming), close genetic

relatedness (mother-offspring grooming), and individual

differences in attractiveness (Seyfarth et al. 1978).

Mothers with infants and high-ranking females, for example,

have been shown to be especially attractive in different

studies (Walters and Seyfarth 1986).

Seyfarth (1977) explored the causal roles of rank,

attractiveness and relatedness in a model that predicted

social-grooming networks among adult female primates.

Seyfarth observed that, among unrelated females,

attractiveness is directly related to rank, and that access

to attractive partners may be restricted by competition

(dominance relationships). He noted, furthermore, that in

many primate species, closely related females often occupy

adjacent ranks and that preference for grooming relatives is

reinforced by dominance and attractiveness. The interaction

between preference based on rank and preference for close

kin, may explain the common features of grooming networks

for a wide variety of species. His model predicts that










females are attractive as a direct function of their rank,

that females are selected to maximize interactions with

high-ranking females, and that competition for access will

result in compromise that determines the exact distribution

of grooming partners. Concerning the direction of grooming,

the model assumes that receiving grooming brings more

benefits than giving grooming. Lower-ranking females will

give more grooming because higher-ranking females are more

attractive. Grooming imbalances will be least when females

are of adjacent ranks, and competition among females will

affect the relative amounts of grooming given and received.

Support for the attractiveness of high rank and the

hypothesis that the majority of grooming is between adjacent

ranked females comes from many analyses of grooming among

macaques and savannah baboons (see Table 2.1). All tests

have involved ground-dwelling cercopithicine monkeys;

studies of colobines, forest guenons (also Cercopithecines)

and great apes do not support the pattern. In most of these

species, grooming does not follow rank patterns. Available

grooming data for New World monkeys also fail to support the

model. The New World examples are exceptional in that both

are arboreal, forest species and are characterized by female

dispersal (Crockett and Eisenberg 1986; Symington 1988).

These observations suggest that the attractiveness of high

rank and relatedness may be important only in species that










form female-bonded groups, or only in ground-dwelling

species.

Forest species, such as the wedge-capped capuchin

monkeys (Cebus olivaceus) are ideal subjects for testing

theories of social behavior and group living. The groups are

female-bonded, range in size from 5 to 40+, and have a

number of matrilines within groups (Robinson 1981, 1988a,

b). The groups have stable matrilineal dominance

hierarchies. Capuchin monkeys are highly social; aggressive

and affiliative interactions occur commonly. Capuchin

monkeys are arboreal, but spend most of the day within 20 m

of the ground and observation conditions are excellent.

Previous work has shown the benefits of group living in

relation to access to resources, vigilance for predators,

and reproductive success of males and females (Robinson

1981, 1986, 1988a, b; de Ruiter 1986; Srikosamatara 1987). A

rank-related cost of group living for females also has been

documented (O'Brien 1988).

This study addresses the questions of how female

capuchin monkeys maintain affiliative relationships and what

are the determining sociodemographic factors underlying

grooming relationships. I examine how females participate in

grooming, the structure of grooming sessions, and the effect

of age, rank and relatedness on grooming relationships. I

next test the Seyfarth model and discuss the relevance of

the model to non-cercopithicine primates.










Study Area

The wedge-capped capuchin monkey occurs in the llanos

of Venezuela, a 200,000 km2 region in the central lowlands.

Capuchins inhabit dry deciduous and gallery forests along

rivers and streams, and scrub forests (matas). Proximity to

permanent water is an important aspect of the wedge-capped

capuchin monkey's home range during the dry season (Robinson

1986) The study area is located on Fundo Pecuario

Masaguaral, a 3,000 ha cattle ranch maintained as a wildlife

management area since the 1940s. The ranch is located in the

state of Guarico in the central llanos, 50 km south of

Calabozo, 80 34'N, 670 35'W. My study site is a 300 ha

gallery/dry deciduous forest, and a 200+ ha area of broken

scrub forest. The climate is strongly seasonal with a 4-

month dry season (December-March), a 6-month wet season

(May-October) and 2 transitional months (April and

November). Vegetation characteristics have been described by

Troth (1979) and Robinson (1986).
Methods

I studied 7 females in 2 small groups and 12 females in

1 large group of wedge-capped capuchin monkeys. White group

ranged from 7 to 10, Splinter group ranged from 9 to 11, and

Main group ranged from 21 to 26 individuals over the course

of the study. One adult female migrated into White group

during the study, but stayed for only 3 months and rarely

interacted with the other females. I excluded this female










from the analysis. One Main group female disappeared near

the end of the study. This female was included in the

analysis. Dominance hierarchies and geneologies were known

for all groups (Fig. 2.1) and all individuals were

recognizable by facial differences, and by variation in

pelage, body size, and habit. The 3 groups yielded 3, 6, and

66 pairs of females dyadss) that could potentially interact.

I eliminated grooming among juveniles because there were

only 3 dyads present. Analyses of grooming relationships

among females therefore included 72 possible dyads.

I collected data for 15 months (December 1986 to March

1988, no data collected in June 1987). Daily behavioral

observations began between 0600 and 0630, and lasted for 12

hours. Observations consisted of a 20-minute focal sample on

an individual female, followed by a 20-minute scan sample of

the group, and a 20-minute opportunistic sample each hour

(Altmann 1974). Females were sampled systematically such

that each female was included in a focal sample in each of

three 4-hour time periods (morning, midday, and late

afternoon). During the first 12 months of sampling, I

recorded all behaviors that lasted at least 3 seconds during

a focal sample. For the last 3 months I recorded only social

interactions that were defined as affiliative or agonistic

interactions involving 2 or more individuals. During the

scan sample, the behavior and nearest neighbor of each group

member were noted.






























Figure 2.1. Rank and close kin (r=0.5) relationships among
females in 3 study groups. Ranks are indicated by numbers
and Rank declines in counterclockwise direction. Mother-
daughter (M-D) and sibling (S-S) relationships are indicated
by brackets. In Main group, adult females are positioned on
inner ring and juvenile females are positioned outside the
ring.















1


2 3
1______


2


White Group


Splinter Group


S S

2 iI
2 M I- ., 1


AL


S

S


4
I.


LI


Main Groul


D


M/1
12


11


D


M,

M D
5 i i 9


M 6 7 S

p D8 S


10


11


4


3


1











During opportunistic samples, I recorded all observed

social interactions involving females. I minimized

observation bias by moving constantly through the group

looking for social interactions. Because viewing conditions

were excellent, I probably missed very few social

interactions. I tested the frequency that individual females

were involved in grooming sessions (using Chi-square test;

Sokal and Rohlf 1981), the amount of grooming in a session

and duration of grooming sessions (two-sample T-tests; Sokal

and Rohlf 1981) in the focal and opportunistic samples.

There were no significant differences in the frequency of

grooming sessions, amount of grooming during a session, or

duration of a grooming session between the 2 samples.

Because the data sets were comparable, I combined data from

focal and opportunistic samples for analysis.

For each grooming session I recorded the identities of

interactants and the frequency and duration of each behavior

lasting more than 3 seconds (Table 2.2). Grooming sessions

were initiated when a female approached a soliciting or

stationary female and one party began grooming, or when 2

females already in close proximity began grooming. Grooming

sessions always included at least 1 grooming event;

solicitations without grooming interactions were not

included. Grooming sessions were terminated when one or both

animals moved away, when one animal failed to reciprocate a

groom or respond to a solicitation for more than 30 seconds,















Table 2.2. Description of behaviors associated with grooming
sessions.

Aqonistic Support. Coalition formed to assist in aggressive
encounters against other males or females.

Approach. Moving toward another individual. Approaches
usually result in a social interaction, either aggressive or
affiliative.

Solicit. An invitation to another animal to groom. Animal
lies down on its side with its back to target animal.
Alternatively, during grooming, animal may sit facing
grooming partner and raise arms and head upwards (sky-
point).

Groom. Manipulating fur and skin of another individual with
fingers, mouth and teeth to remove bits of dirt, dead skin,
ectoparasites, blood from wounds, etc.

Grooming Session. The time spent in close proximity during
which grooming occurred.

Termination. Cessation of grooming due to lack of response
to solicitation, failure to reciprocate on a groom, or one
or both individuals leaving the session.

Reciprocated Grooming. Both individuals groom in a single
session.










or when one or both animals engaged in 1 or more asocial

behaviors for more than 30 seconds but remained in close

proximity. A grooming session, therefore, included time

spent in close proximity as well as time spent grooming.

I considered 3 data sets: (1) the individual grooming

sessions between pairs of females (2) the summary of

grooming relationships between each pair of females and (3)

the summary of each female's participation in grooming

irrespective of partner. I use the individual grooming

sessions to describe the dynamics of grooming among females,

the fluctuation of grooming over time, the effect of a

female with a new infant on grooming sessions and to

evaluate reciprocity in grooming. The summary data for

grooming dyads to determine the effects of age, rank and

relatedness on grooming relationships, what aspects of

grooming are general and what factors might influence strong

grooming relationships. The data summarizing each female's

participation in grooming incorporate a set of counts,

proportions and ratios (Table 2.3) that measure a female's

participation in grooming sessions and distinguish how

females vary in their relationships with other females.

I considered duration of grooming session, time spent

grooming by each participant in a session, and the frequency

of grooming sessions as the basic data for the analysis of

individual grooming bouts and for the analysis of grooming

relationships. Grooming sessions were classified as











oN co o
ri 0 m
(, 8


00





00


>4
0)

0















0
.-I















0


g 0

















Feg
c0
0
0)
I



































4.4 F
4.4
0

0



5







o
ap

u





Fe

S)


41.
(0

aO









-H
U O
*H C


rn o U
000





000
8 ooC


0 V-1 V-4





000
V N %)
* oo


oo0 in

c4 rt



amwo
FCVI


H- L eH M
000


0000






0000


n 000N
000


r- Ln co

000




Scco
* .
0 0od


SONM V
rA0'.(


000
00 0
oo; 8


co n r-

0000
SlL- L r)










* 9 9
0000
lMt i .
in i ou o





* *


000 N


SN %0
C4 ; .. 1


CN N


COI P
00> 0



Go in 0
8e C;


0
n
o


M on Ln o m
IN t- p- in M





Cn in ** in qw
oo 8 C; 8

wamoo

amOQ\O
***** )~
ooooo


cre
* *
00





* *
00


COn Ln

SC4 ;
0om0


coco n

4NN (4


000


in o
in Ln in
000





000
oo; 8C

o 0 m
* ( M
000 ;


0% r-4 -)

0o0 r-


OO
0049


%\o0 Cn (NN I
00000


-I N H H N N r-44N r-
4 44 < t-34 44 t-) 44 -3


0) )
4.) 4-)
'- 41
.4.4


MM$4
4J+4JV4J.CCC CCCCCCC
0C1 C Cr. .9 C *-1 *- -r-i *i -> r4 -f- -r4l *- 4 -rrl -r- -r *
.C -#-i --I4-. 4 -HM it 0 0 M 0 10 90 10 (0 0 (0 0 O
::4 r-A r-4 SE
. -
c..o M ua


uMkgzwMOr o oZio pomgAr
M0H04(S4 WO440EEzmUAp404


&4
C vM
IA -f-1 44 0
o o >i
V0 00

10 $ 0 0* -4
*H0O M- V4
54 0 0V 0
00 4009


4 0 0004 0
r*- 0 0O -4
3 o ~ o) co
0 040 -4) 5.

114 0 *
(N *MQ
4 II. CtO 0) H
f.I 10 0)0
S0)4 4j >4
c kC

II 0 cn 4) 0)
WH r- 4J

S* C -Q
I ,-. .*H 0
C T -c1 *




S0) ) 0) -H4 0
0 II 4)- 0 --,
O 0 wO
V o ON W o V
C0 t)) 0 --4 4'4 WU)
44 --000 0


a- '4 0 t-M
O 4 ON -r4 p U
4J1 0 4-1 0 )
- I 0 ,4 0
o c: aa aUN 1
0 10. 0
II () 4 44


>cI go r 0



:3 O H Q..
H*-rrO 00

Ai 3>i :-4
V. 44 0 O
v0V 0 4.)

4 M-H 54
10 30) 0-0

t..o -.g4 V 0 9
0 -rH -rI $4 u = U

S1 r 0^ ) 00to

U 0 TO $4 4 M U
4)
c -f t* w>

0 4-0 (C 0%
r-I 0 0 0 *- -
U) 0*q U a 0
4*k0 W U)0 00
*V I M Lo r-q 0 0
0 II a Q) 0 14 M
040mm0tp
,;cam*se *0


in n.



oL


0

--. H











reciprocated or unreciprocated, depending on whether both

participants groomed. Sessions also were classified into 3

categories based on the presence and age of infants

belonging to at least 1 of the interactants (0 = no

infants, 1 = infant <1 month old, 2 = infant 1-2 months old.

Participants were classified by age (juvenile and adult) and

by rank (high, medium and low). Within a rank class, the

first individual identified was always the higher-ranking

individual of the dyad. In this way, rank relationships

within a rank class were preserved in the relative amounts

of grooming between participants. Degree of maternal

relatedness (r = 0.5, 0.125, 0) between participants was

noted and all values of r< 0.125 were set equal to 0. Group

membership was classified as large versus small.

From the basic information on grooming, a series of

ratios and proportions characterizing grooming relationships

can be calculated (as in Table 2.3). Asymmetry in a grooming

relationship is measured by the difference in amount of

grooming between 2 individuals and the ratio of grooming by

one individual to grooming by the other. Response to

solicitation is measured by the number of times that

grooming follows a solicitation divided by the number of

solicitations. Reciprocity by an interactant is measured as

the proportion of grooming sessions that an individual

reciprocates grooming. The proportion of sessions terminated

by one or the other interactant also was calculated. Females









21

that have long sessions, a high grooming ratio greater than

or equal to 1, high response to solicitation, high rates of

reciprocity and low termination rates are expected to be

good grooming partners and should have a high number of

grooming sessions. Similarly, grooming relationships that

are balanced in response to solicitation, grooming,

termination and reciprocity, should be stronger (more

grooming sessions) than relationships in which one partner

does all the grooming.

I used both parametric and nonparametric analyses. When

sample sizes were large, ANOVAs were complex, or when tests

of distributions were appropriate, I used parametric

analyses such as repeated measures ANOVA and Chi-Square

tests (Sokal and Rohlf 1981). Nonparametric Rank Sum tests,

Wilcoxon Sign Rank tests, and Kruskal-Wallis 1-way ANOVA

were used for testing differences in grooming attributes of

individual females and testing differences in patterns of

grooming as a function of relatedness (Hollander and Wolfe

1973). All analyses were performed on the SAS statistical

package (SAS 1985).

Results

Characteristics of Grooming Sessions

Grooming sessions between females occurred 384 times

during the course of the study. The average grooming session

lasted 2.51 minutes (SE=0.134) with the higher-ranking

female grooming an average of 1.14 minutes and the lower-











ranking female grooming for 0.71 minutes. Frequency of

grooming varied considerably by month (mean=25.80, SE=5.59;

Fig. 2.2) with minor peaks in February and a major peak in

April-May when infants appeared in Main group. Monthly mean

duration of grooming stayed relatively constant over time.

The largest deviations in durations occurred in months with

the least grooming (August, November, December) and are

probably the result of small sample size.

Several behaviors and behavioral sequences associated

with grooming sessions were noted. Of 173 approaches that

did not result in supplantation or aggression, 100

approaches were associated with grooming sessions. Other

sessions were initiated after animals were in proximity but

engaged in asocial behaviors or after an animal had been

following another. High-ranking animals initiated most

approaches (65%) and approached high-ranking animals more

often than low ranking animals. If a low-ranking animal

approached a higher-ranking animal, the higher-ranking

animal was usually related or of a similar rank. Very low-

ranking animals rarely approached high ranking animals.

Adults usually approached adults rather than juveniles.

Generally, relatedness was not an important factor in

approaches.

Approaches were often toward soliciting females (20

records) or toward a female that responded to the approach
































Figure 2.2. Frequency and mean duration of grooming sessions
by month for female wedge-capped capuchin monkeys.




















GROOMING DATA


I-


nn n


0-~ -


flrn rri


6
5z
40
2 0
0


DJ FMI4 AM J JAS N D J M


MONTH


90
80
5 70-
60-
50-
E 40-
30-
20-
10-
n


"\
~`I


m- -


B l E m


'








25

by soliciting (23 records). In 30 approaches the approaching

female then solicited a groom. The most common outcome of an

approach was immediate grooming (no solicitation or

solicitation lasting less than 3 seconds) of the female

being approached. Solicitations, in all contexts, resulted

in grooming only 44% of the time. The behavior was often

ignored, met with a solicitation, or the target moved away.

Ignoring a solicitation, moving or failing to

reciprocate grooming terminated the session. The lower

ranking member of the session was responsible for

termination in 218 of 373 sessions. In 47 of 72 dyads, the

lower-ranking female terminated more sessions than the

higher-ranking female. This pattern was consistent except

when medium-ranking females were grooming medium- or high-

ranking females. In these dyads, the higher-ranking animal

tended to terminate more often. Among adult females, the

lower-ranking female terminated more, but between adult and

juvenile females, both parties were equally likely to

terminate. Among unrelated females, the lower-ranking female

terminated more, whereas among closely related females

(r=0.5), both females were likely to terminate.

The Influence of a Newborn Infant

To determine the importance of the presence of young

infants to frequency and duration of grooming, I analyzed

the amount of grooming per month for 15 adult females of

different ranks in large and small groups as a function of











the presence of infants (none, 0-1 month old infant, 1-2

month old infant). Juveniles were excluded from the analysis

because they could not bear infants. I used repeated

measures ANOVA (Sokal and Rohlf 1981) that treated rank and

group as blocks and presence of infant as a treatment.

The duration of grooming was unaffected by group size,

female rank or the presence of an infant. Frequency of

grooming sessions was significantly higher in the large

group (4.1 vs 1.4 sessions/month; Table 2.4; Fig. 2.3).

High-ranking females were involved in grooming sessions

significantly more than middle- and low-ranking females (4.9

vs 2.6 and 1.7 sessions/month respectively). Females with

new infants were involved in significantly more grooming

sessions than females with 1-month old infants and females

with no infants were involved in the least amount of

grooming (16.0, 6.2 and 2.7 sessions, respectively).

Variability among females was substantial (interaction

terms, Table 2.4; Fig. 2.4). Females in small groups did not

respond strongly to rank differences, or to the presence of

infants. Grooming declined with rank for females with new

infants and females with no infants but not for females with

1-month old infants. However, 2 important data points in

this comparison have sample sizes of 1, and this result

should be treated as preliminary.
















Table 2.4. ANOVA results for the effect of group size, rank,
and presence of infant on frequency of involvement in
grooming for adult females. ID is the female identity.
Infant refers to the presence and age of infant.

Source d.f. S.S. M.S. F-test' P>F

Group 1 492.1 492.1 51.12 0.0001

Rank 2 553.3 276.6 28.73 0.0001

Group x Rank 2 211.6 105.8 10.99 0.0001

ID(Group x Rank) 9 86.6 9.6 0.99 0.4520

Infant 2 997.2 498.6 51.14 0.0001

Infant x Group 2 322.4 161.2 16.53 0.0001

Infant x Rank 4 451.2 112.8 11.57 0.0001

Error 179 1745.2 9.7

Total 201


1. F-tests for Group, Rank, and Group x Rank interaction use
ID(Group x Rank) as the error term. Other tests use M.S.
error.































Figure 2.3. Mean and SE of frequency and duration of
grooming sessions for female wedge-capped capuchin monkeys
in large and small groups. Females in large groups are
presented by ranks.























GROOMING DATA


100
90
80
70-
60-
50-
40-
30-
20
10
0
GROUP
RANK


~ZFL]FL]K1


MAIN


MAIN MAIN SPLN WHITE


HIGH MEDIUM LOW


z
I:
c
2E
C3<































Figure 2.4. Interactions between membership in a large
versus small group, female rank, and presence of infant as
they affect frequency of grooming for adult females.










GROUP BY RANK


n.H


HIGH


GROUP BY INFANT


,r l


NO INFANT
RANK BY INFANT


0-1 MONTH 1-2 MONTH


0-0 HIGH
e-* MEDIUM
A-- LOW


0-1 MONTH 1-2 MONTH


MEDIUM


LOW


NO INFANT











Reciprocity in Grooming

Unreciprocated grooming sessions were significantly

shorter than reciprocated sessions (1.60 vs 3.75

minutes;F=74.79, d.f.= 3, 380, P<0.0001) and the total

amount of grooming also was significantly less. There were

no significant differences in the quality of grooming as a

function of age, rank or relatedness of grooming partners

between reciprocated and unreciprocated grooming sessions.

Frequency of grooming did differ by age, however. Among

adult females, 47% of grooming sessions were reciprocating

sessions; only 33.6% of sessions involving adult and

juvenile females were reciprocating (Fig. 2.5). Females

reciprocate significantly more often when sessions are among

adults than when sessions are between adults and juveniles

(Chi-Square=6.475, d.f.=l, P=0.0109). In sessions with no

reciprocal grooming, the lower-ranking member of the pair

was the unreciprocating partner in 161 of 222 sessions.

Unreciprocated grooming was affected by the presence of

new infants (Fig. 2.5). Grooming sessions when no infants

were involved were as likely to be reciprocated as not. When

new infants were present however, grooming sessions were 3.3

times more likely to be unreciprocated, and 2.5 times as

likely to be unreciprocated when a 1-2 month old was present

(Chi-Square= 19.24, d.f.=2, P=0.0001). Usually, the mother

terminated these grooming sessions by moving away from the

groomer. Quite often, however, the groomer would follow the






























Figure 2.5. Distribution of reciprocated and unreciprocated
grooming sessions as affected by the presence of an infant
and relative ages of interactants.













150


GRO4
I-1


125
II

100

S 75

50-

25

0 rEi
NO INFANT 0-1 MO. 1-2
INFANT INF


150

125

z 100

O0
w 75

50

25


ADULT ADULT
ADULT JUVENILE


DMING SESSION
UNRECIPROCATED
RECIPROCATED


MO.
WNT










mother and another short, unreciprocated session would

occur. Occasionally, a persistent groomer might follow a

retreating mother 5 or more times before stopping. I

interpreted these grooming sessions as attempts by a female

to gain access to the infant by grooming the mother. The

mother, however, appeared to withdraw from grooming

interactions, especially when the infant was new.

Characteristics of Females

Grooming sessions were highly variable in structure,

frequency, and duration. Much of this variability was a

function of the participants (Table 2.3). High-ranking

females tended to approach, respond to solicitation, and

have a higher grooming ratio than low-ranking animals. Low-

ranking females failed to reciprocate grooming significantly

more often (50.7%; Kruskal-Wallis 1-way ANOVA, H=6.46, n=19,

P=0.0396) than high- (21.0%) or medium- (27.5%) ranking

females. Low-ranking females were less involved in grooming

sessions and terminated grooming more often than high-

ranking females but these difference were only marginally

significant (0.05< P <0.1). Adult females terminated

significantly more grooming sessions than juvenile females

(Rank Sum test; n=19, P=0.0404).

Group differences existed in the patterns of grooming.

Grooming ratios declined with rank in Main group and

Splinter group, but increased with rank among closely

related females of White group (Fig. 2.6; see below). Group






























Figure 2.6. Ratio of giving to receiving grooming (in
minutes) as a function of decreasing rank for the 3 study
groups. Lines indicate trends in data for 3 groups.






















4 O WHITE
A SPLNTER
SA O MAIN
0
0
o' 3
0 \
z
m 2



0 1



0a
1 3 5 7 9 11 13
DECREASING RANK








38

size differences were significant for frequency and duration

grooming. Females in large groups groomed more frequently

(large: 52.5 [SE = 10.43]; small: 19.0 [SE = 3.12]) but with

much greater variability than the females in small groups

(F=19.29, d.f.=6,11, P=0.0008; Rank Sum test; n=19,

P=0.0085). Females in small groups had significantly longer

grooming sessions than females in large groups although

variances in duration were similar for both categories

(small: 2.98 [SE = 1.93]; large: 2.41 [SE = 1.67]; Rank Sum

test; n=19, P=0.023).

Closer examination of the frequency and duration of

grooming data (Fig. 2.3) suggested that females in small

groups were similar to low-ranking large-group females in

the frequency of grooming but were similar to higher-ranking

large-group females in the duration of grooming. Analysis of

variance results confirmed that grooming frequencies were

significantly higher for medium- and high-ranking females of

large groups (69.25) than for low-ranking females of large

groups (19) and all females (19) of small groups (Kruskal-

Wallis 1-way ANOVA; H=12.37, P=0.0021). Average grooming

durations for small group females (2.98 minutes) were more

similar to medium- and high-ranking females of large groups

(2.69 minutes) than low-ranking (1.84 minutes) large-group

females (Kruskal-Wallis 1-way ANOVA; H=9.85, P=0.0073).

Grooming characteristics other than frequency and

duration were associated with different females, but these











characteristics varied considerably and not always

predictably (Table 2.3). Significant positive correlations

existed between the ratio of grooming given to grooming

received and response to solicitation, and between

proportion of terminations and proportion of unreciprocated

sessions (Table 2.5). Significant negative correlations

existed between approach and terminate, approach and

proportion of unreciprocated sessions, groom ratio and

response of partner to solicitation, and response to

solicitation and terminate. Females that are active in

grooming sessions are females that approach, respond to

solicitation, have a high grooming ratio, reciprocate

grooming, and tend not to terminate grooming sessions. These

characteristics tend to describe high- and middle-ranking

females better than low-ranking females.

Grooming Relationships

Of all female dyads that could potentially groom with

one another, 10 dyads failed to groom at all during the

entire study. An additional 32 dyads groomed less than 4

times during the study. I define a grooming relationship as

consisting of a minimum of 4 grooming sessions with total

grooming time of at least 1 minute. In this study, 56% of

the dyads failed to form grooming relationships. Failure to

form a relationship followed no obvious pattern. Among

closely related females, one third of the dyads fail to form

grooming relationships (Figs. 2.7-2.9). Three of the 4 low-

















Table 2.5. Spearman rank correlation matrix for grooming
characteristics of individual female wedge-capped capuchin
monkeys. See Table 2.3 for definitions of variables.


(6) (7)


NS
NS
NS
NS
-0.59**


0.59**
-0.72*** NS
-0.53* -0.48*
-0.71*** -0.52*


0.64**


(2)


(1)


NS
NS
NS
NS
NS
-0.64**
-0.57**


(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
*
**
***


P<0.05
P<0.01
P<0.001






























Figure 2.7. Grooming relationships as illustrated by number
of grooming sessions, average duration and relative amounts
of grooming by each member of a dyad in White group.
Ordering of females as in Figure 2.1.









Number of Grooming Sessions


/



4-5
6-10
11-15
>15


EEl


Average Session (min.)


1-2
2-3
3-4 =
>4 m


El


Grooming Up and Down
Up- m Down--1


Rank (min.)


5-10
10-15
15-20
> 20 mi






























Figure 2.8. Grooming relationships as illustrated by number
of grooming sessions, average duration and relative amounts
of grooming by each member of a dyad in Splinter group.
Ordering of females as in Figure 2.1.











Number of Grooming Sessions










/
4-5
6-10 -
11-15
> 15


Average Session (min.)











<1 min.
4-5
6-10
11-15 -
>15 m


Grooming Up and Down Rank (min.)
Up- m Down. =--











1-5
5-10 -
10-15
15-20-
>20






























Figure 2.9. Grooming relationships as illustrated by number
of grooming sessions, average duration and relative amounts
of grooming by each member of a dyad in Main group. Ordering
of females as in Figure 2.1.









Number of Grooming Sessions


Grooming Down Rank (min.) Grooming Up Rank (min.)

WH --- WH I

SAM \\ BA]O/ \B


1-5
5-10
10-15 -
15-20 -
20 m


Average Session (min.)


\ ML /


\rML,








47

ranking females in Main group failed to groom for more than

a minute over the study and thus have no identifiable

grooming relationships. Among juvenile females, one failed

to form grooming relationships, one formed relationships

only with closely related females, and two formed grooming

relationships with unrelated or distantly related females.

I first analyzed 72 dyads to determine the effects of

age, rank, and relatedness on grooming. Although the

analysis is dominated by the large group (60 dyads),

previous analysis has shown that large and small groups do

not differ in quality of grooming, only in frequency. Within

this data set there were 9 mother-daughter dyads, 4 sister

dyads, 2 dyads with relatedness values of 0.25 and 4 dyads

with relatedness values of 0.125. The 2 high-ranking

matrilines in Main group were related and this resulted in a

confounding of relatedness and rank. In order to separate

the effects of rank and relatedness, I ran each analysis

with all data included and then removed the 6 dyads that

were distantly related and ran the analysis again. I

considered an effect to be significant if it was significant

in both analyses.

I used ANOVA to determine the importance of rank,

relatedness, and age on 11 behaviors and ratios of grooming

(Table 2.6) directly related to grooming, or correlated with

grooming, to determine if sociodemographic variables

affected the grooming relationships. The results showed that



















8 5
0 z z


mo m
U) U)


2 2


MA
04


C
ft

*.4
CI









0
O1

0
c








.4 0
0 Q
s -







0a.



00
r .
o t
1 p
0

oC




O)
c c















04-4
Z 4
N0)






co

U)-
0no


o 0 r.-


U) 0
o
z


0
C-4
0
0





0




.*r

rO
.-I
*4

0 0
)0
04 p
3 O
tO (0


44c
4P 1 t U
C (1) >4 41) 4)
O OO r-I U -I -4


o o0o w H 4 4E-I
l v 0 0* r 0 *0
0 4 1 0 Q 0 0



eO 10 HO HO t00
P0 00 4 0 ) 0 O 4)
Zt so Zi Mi 0 -
O O' cO ;
CF a o o


S4
,-4
cO


r); -
4

0 U
*l-

W 0
C )o
0
040


2 M
z z


2 2









m i
2 2







2 2


44 4-4 4-)
00 40 40


u 0 p M p 4

W.,Q UR U

0 CcO CaO

4) 0 f00 0) 0
4-O ygO 5 41
04) I k 0 0 P 0


S z z az az


a az 2 2 2 2









OU) ) U) ) U) UO U)
2 2 2 2 2 2 2





r-4
0 ) U) ) ) U) U m
0 2 2 2 2 2 2


) i) E En rU W) UE


2 2 2 2 2a a


U) W
m 22











age, rank, and relatedness have little effect on grooming

relationships except for frequency of grooming. Rank was

important for support during aggression, approaches, and the

number of grooming sessions. High-ranking animals approach,

support and groom each other preferentially (Fig. 2.10).

Frequency of grooming was positively correlated with support

during aggression (r=0.617, P<0.0001) and with approach

(r=0.758, P<0.0001). Support during aggression was minimal

among medium- and low-ranking dyads. Age was a significant

factor only for approach; adult dyads approached more

frequently than adult-juvenile dyads. Closely related dyads

groomed twice as frequently as unrelated dyads. It appeared

that age, rank, and relatedness were not important

predictors of variables that measure the quality of the

grooming relationship among females. The reasons for this

lack of explanatory power were explored further.

Seyfarth's Model

The model of social grooming developed by Seyfarth

predicts that high-ranking animals will receive more

grooming than low-ranking animals (grooming is directed up

the hierarchy) and that the majority of grooming will occur

between adjacently ranked females. Figures 2.7, 2.8, and 2.9

show that females groom females of lower rank more often and

for longer duration than females of higher rank. A test of

the amount of grooming by each member of a dyad shows that

the higher-ranking female grooms significantly longer































Figure 2.10. Mean frequency of grooming sessions, support
during aggression and approaches between females of
different rank categories.




















7

20. --I AGON. SUPPORT
=8 APPROACH
16- GROOM SESSIONS

5 12

K 8 119 6
19

41 t 17


RANK 1 HIGH HIGH HIGH MEDIUM MEDIUM LOW
RANK 2 HIGH MEDIUM LOW MEDIUM LOW LOW










(median=0.62 minutes vs. 0.25 minutes, P=0.0047, Wilcoxon

Signed Rank Test). When high-ranking animals are involved in

a grooming interaction, for example, they groom an average

of 4.3 minutes for every minute of grooming they receive

from the lower-ranking partner. This pattern of grooming is

in contrast to what has been reported for cercopithecine

monkeys but consistent with results for some New World

primates (Table 2.1).

I tested the second hypothesis by comparing the

frequency of grooming sessions, length of grooming session,

and amount of grooming by each partner as a function of

occupying adjacent or more distant ranks. Animals occupying

adjacent ranks did not groom more frequently (Rank Sum Test,

P=0.140). Adjacent-ranked females did have longer grooming

sessions (Rank Sum Test, P=0.0034) and the higher-ranking

partner groomed longer (Rank Sum Test, P=0.0168) than

sessions involving more distantly ranked females. This

pattern is similar, in part, to results observed in

cercopithecine primates and suggests a special relationship

between adjacent ranked females.

Because the model does not discriminate between the

effects of rank and relatedness that might account for

grooming patterns among adjacent-ranked females, I made a

separate comparison of related (n=10 dyads) and unrelated

(n=6 dyads), adjacent-ranked females. I hypothesized that if

relatedness were an important factor determining patterns of









53

grooming among adjacent-ranked females, then related females

should groom longer and more often than unrelated females.

Rates of grooming and length of grooming session did not

vary as a function of rank. The frequency of grooming,

however, was marginally greater for related than for

unrelated females (5.9 vs. 3.3 sessions/dyad; P=0.065, Rank

Sum Test). As in the general patterns of grooming discussed

above, relatedness affects the frequency but not the

duration of grooming among adjacently ranked females.

The results of tests of kinship are somewhat surprising

because relatedness is often cited as a causal factor in

social relationships of female-bonded primate groups.

Figures 2.7, 2.8 and 2.9 show that closely related

individuals did not consistently form grooming

relationships. Of the mother-daughter dyads, 3 of 9 did not

form grooming relationships and 3 more groomed less than 10

times over the course of the study. Among sibling dyads, 1

of 4 did not form grooming relationships, and none groomed

more than 10 times. Strong asymmetries are evident in some

grooming relationships. Female MG groomed her sister (ML)

and mother (MO) 4-5 times as much as she received grooming

from either. On average, daughters groomed their mothers

twice as much as mothers groomed daughters, and older

sisters groomed younger sisters almost 3 times as much as

they received grooming.










There were no significant differences in frequency of

grooming or in duration of average grooming session between

mother-daughter dyads and between sibling dyads but sibling

dyads sessions were more variable in length than mother-

daughter (F=4.95, d.f.=3, 8, P=0.031). When sisters groomed

sisters, solicitation rates and grooming rates did not

differ significantly, but this may be a function of the

small sample size (n=4). Daughters groomed mothers

significantly longer than they received grooming (Wilcoxon's

Signed Rank Test; n=9, P=0.027). Comparisons in which a

female could groom either a mother or a sister (n=6) showed

that daughters groomed mothers preferentially to sisters

(Wilcoxon's Signed Rank Test; n=6, P=0.031). Daughters

averaged 6.5 more grooming sessions with their mothers than

with their sisters. Other aspects of grooming were similar

however; a female groomed mother and sister at approximately

the same rate, for approximately the same duration, and the

sessions were approximately the same length.

Discussion

Grooming relationships in wedge-capped capuchin monkeys

are differentiated by the frequency rather than the quality

of grooming. These social relationships are predicated on

rank and, secondarily, on relatedness. Age is an important

factor in determining reciprocity of grooming. High-ranking

females frequently engage in grooming sessions, are

responsive to solicitation, reciprocate grooming, and are










less likely to terminate grooming. Low-ranking females do

not engage in frequent grooming, are less responsive to

solicitation, are less likely to reciprocate grooming, and

are more likely to terminate the grooming sessions. Whereas

high-ranking females initiate and maintain grooming

relationships, low-ranking females withdraw from and

terminate grooming. Low-ranking females do not even form

grooming relationships with close kin.

A possible explanation for this pattern is that low-

ranking females are peripheralized early in life and never

develop affiliative relationships. Data suggest that

juvenile females have a different type of grooming

relationship than do adults. In this study, the 3 juvenile

females that had opportunity to interact with one another

did not. Interactions with adults, including mothers, often

were not reciprocated or the juvenile contributed most of

the grooming (Figs. 8-9). Adult females apparently do not

receive the same benefits from grooming with juveniles as

with other adults. A juvenile female that fails to establish

grooming relationships with peers, siblings, or adults may

be forced to the periphery of the group because, in the

absence of affiliative relationships, social contacts will

tend to be aggressive.

Because initiating grooming sessions involves

approaching or maintaining close proximity to another

female, grooming relationships are predicated on the








56

response of the target. A high-ranking female approaching a

low-ranking female elicits 3 possible responses: low-ranking

female moves, low-ranking female performs a submissive

behavior, or low-ranking female solicits groom. All three

responses can be interpreted as submissive behaviors that

minimize aggression between the pair. Approach by a low-

ranking female is difficult without possibility of

aggressive retaliation by the higher-ranking female: as a

result, low-ranking females seldom approach high-ranking

females. Consequently, high-ranking juvenile females are

able to initiate grooming with close relatives plus lower-

ranking females whereas low-ranking juvenile females can

initiate grooming only with relatives. A low-ranking female

therefore, is more constrained by the social structure of

status and kinship than a high-ranking female. In this

study, HA has grooming relationships with 5 females, whereas

ML has grooming relationships with mother and sister only

and PU does not have any grooming relationships. In Splinter

group, PE has a relationship with both lower-ranking

females.

Although a low-ranking female cannot initiate

affiliative relationships, she might respond to the

initiation if grooming is important to the maintenance of a

social bond. Females that move away from approaching females

are withdrawing from potential grooming sessions as well as

potential aggressive encounters. It is difficult for the










observer to impute motivation for moving away from an

approaching female except that the impetus to move is

greater than the impetus to hold one's ground and attempt to

thwart aggression through an appeasement gesture. A low-

ranking female may respond by inviting a grooming session.

If the low-ranking female were attempting to develop

relationships with high-ranking females, one would expect

her to actively participate in grooming; she would respond

to approaches with solicitations, reciprocating rates would

be high and termination rates low. That grooming ratios and

reciprocity rates are low, while termination rates are high

for low ranking females, indicates that these females are

tolerating grooms only as long as is necessary and are not

trying to maintain grooming sessions. Low-ranking females

treat grooming as an interaction to be avoided rather than

encouraged. Grooming interactions may reinforce bonds among

similarly ranked females, but they appear to serve more of

an appeasement function between high- and low-ranking

females.

Are low-ranking females peripheral? Robinson (1981)

showed that low-ranking females of Main group spent more

time at the edge of the group. Based on Robinson's Index of

Association (Robinson 1981), females in the high-ranking

matrilines have higher AI's than the females in low-ranking

matrilines. Low-ranking animals have no nearest neighbors

more frequently than high-ranking animals, suggesting that








58

they are peripheral. In Main group, low-ranking females have

no nearest neighbor 40% of the time whereas high-ranking

females have no nearest neighbor for 29% of observations and

middle-ranking females have no nearest neighbors for 35% of

observations.

Grooming as an appeasement behavior explains

directional grooming down the hierarchy, but does not

explain differential attractiveness of females with new

infants. These interactions, however, also follow rank.

Females in White and Splinter group do not show a strong

infant effect, nor do they show strong rank effects (Fig.

2.5). In Main group, grooming of mothers with infants was

strongly affected by rank: the high-ranking mother received

twice as much grooming as the medium- and low-ranking

mothers. Most of the interactions with new mothers are

unreciprocated grooming sessions with higher-ranking females

or kin (see O'Brien, Chapter 5). In many primate species

with matrilineal dominance hierarchies, mothers with infants

are the target of grooming by females attempting to gain

access to the infants. Mothers with infants often withdraw

from social interaction until the infant is relatively

independent.

How well do the grooming data from wedge-capped

capuchin monkeys fit the appeasement versus reciprocity

model of social grooming? If grooming is only a form of

appeasement, then low-ranking animals should receive rather








59

than give grooms, grooming should not be reciprocated, low-

ranking females should terminate, and there should be no

necessary correlation between grooming and other affiliative

behaviors. Inverse correlations might be expected between

aggression and grooming. The structure of grooming supports

the appeasement hypothesis in general. A strong correlation

exists between support during aggression and grooming, but

correlations between aggression and grooming are weak and

not significant. Although data from small groups are

equivocal, levels of aggression among females are generally

low and probability of relatedness in small groups may be

high. Such patterns suggest a reduced need for appeasement.

In general, White group females behave similarly to

matrilineal triads in Main group, and Splinter group females

behave like related versus unrelated females in Main Group.

Predictions of Seyfarth's model based on attractiveness

of rank are generally not upheld. Grooming is directed down

rank and low- ranking females avoid grooming rather than

competing to groom. Predictions of adjacent rank are also

not upheld. This is partially because it appears that

daughters groom their mothers preferentially to their

sisters, because daughters groom mothers less than mothers

groom daughters, and because low-ranking mothers and

daughters failed to form grooming relationships. There is,

however, a moderately strong correlation between frequency

of grooming and support during aggression which, is a










feature of models based on attractiveness of rank. This

support, furthermore, is concentrated among high-ranking

females that exhibit the most balanced, reciprocal grooming

patterns.

It appears that grooming serves more than 1 function.

The short-term benefit of grooming as appeasement serves to

defuse potentially aggressive situations between females

that otherwise do not have a relationship. Such females

react to approaches by soliciting a groom. If a groom is

given, or the female moves past, then the low-ranking female

has benefitted by not being subject to aggression, and by

not losing her position by supplantation. Among related and

high-ranking females, grooming may serve as appeasement but

it may also reinforce social bonds. That the 2 dominant

matrilines in Main group were related and reciprocated

grooming almost exclusively among themselves indicates that

these females possessed a tolerance for each other not

exhibited toward lower-ranking females. The interplay

between rank, attractiveness, and relatedness is consistent

with predictions; directions of grooming are contrary.

The model proposed by Seyfarth works well for ground-

dwelling cercopithecine monkeys but not for species that do

not exhibit strong female-dominance hierarchies, species

with female dispersal, and forest species. The highly

structured nature of sociality in ground-dwelling

cercopithecines may result from more intense competition for










resources relative to that of forest primates (Alexander

1974; McKenna 1979). The complex grooming patterns of

cercopithecines may be a refinement over the more primitive

appeasement pattern exhibited by birds, and most mammals.

That many Old World primates and lemurs do not exhibit

directional grooming (or high levels of aggression) suggests

that resource competition may not be a major factor in the

development of social behavior in these species (McKenna

1979). The primitive pattern of grooming in New World

primates suggests that while the social structure of 2

species appear very similar, the development of social

behavior within that structure remains flexible.
















CHAPTER 3

THE ROLE OF AGGRESSION IN THE SOCIAL STRUCTURE
OF FEMALE WEDGE-CAPPED CAPUCHIN MONKEYS.


Introduction

The development of social structure is mediated through

affiliative and aggressive behaviors among members of a

group (Alexander 1974). As group living evolves, social

behaviors develop to enhance the benefits of group living,

regulate competition and bond group members into tolerant

and even cooperating subunits. Affiliative behaviors include

cooperative and altruistic acts such as allogrooming, food

sharing, allomaternal care of offspring, alarm calling, and

coalitionary support during aggression. Aggressive behaviors

develop that maximize an individual's share of group

resources at the expense of others. Social structures based

on dominance hierarchies, kin relationships, or long term

social contact may enhance affiliative behaviors and

minimize aggression when group members recognize the

organizing principles.

Aggression among group-living females is a form of

social communication involved in the development and

maintenance of dominance relationships (Walters and Seyfarth










1986). Aggression is often an expression of competition

between individuals for food resources (Clutton-Brock and

Harvey 1976) although it also is observed in nonforaging

contexts (Bygott 1979; Johnson 1989). Because females

maximize reproductive success through the acquisition of

resources (Wrangham 1980, 1983; Janson 1984), distribution

and quality of resources affect spacing patterns of females,

influencing the opportunity for aggression. Fights that

appear to have no motivation may in fact be used to

establish social dominance between 2 individuals (Rowell

1974, Walters 1980). Recognition of a social structure may

reduce the need for females to establish priority of access

when resources are clumped (Alexander 1974). Supplantations,

in which a female relinquishes access to a resource in

response to the approach of a socially dominant female,

exemplify recognition of social structure. Low-ranking

females recognize the strength of dominant females and move

to avoid fights.

Matrilineal relatedness and age have been implicated as

organizing principles in social structures among female

primates (Wilson 1975; Chapais and Schulman 1980; Trivers

1985; Chapais 1983; Walters 1986). Juveniles that are

establishing social positions interact differently with

other juveniles than with adults and many of these

interactions involve aggression. Interactions among adults

also differ from interactions among juveniles. Relatedness










among females in a group serves to mitigate aggression

because closely related females have common interests

(shared genes) that promote affiliation over aggression. If

social dominance develops within families exhibiting low

levels of aggression, it is possible to develop social

structures with minimal aggression in groups of related

matrilines. Few studies of aggressive interactions among

females, however, have considered group differences, or

sociodemographic differences within groups simultaneously.

In this chapter, I examine the determinants of

aggression among females in 3 groups of wedge-capped

capuchin monkeys (Cebus olivaceus). I first establish that

female dyads have different levels of association that are

affected by group membership, social dominance, and

relatedness. I then show that fights and supplantations are

a function of the context in which they occur, social

structure, and identities of females in the dyad. The

results demonstrate that after the effects of group

membership, age, rank, and relatedness of interactants are

removed, a significant amount of variation is attributed to

individual variability. It is argued that aggression has a

social function that influences the spatial and social

structure of the group, but that social structure may be of

limited usefulness for predicting patterns of aggression.










Methods

I studied 7 females in 2 small groups and 12 females in

1 large group of wedge-capped capuchin monkeys (Fig. 3.1).

White group had 3 females and totalled 7 to 10 individuals

during the study, Splinter group had 4 females and totalled

9 to 11 individuals, and Main group ranged from 21 to 26

over the course of the study. Two females were present for

only part of the study. An adult female migrated into White

group during the study, but stayed for only 3 months and

rarely interacted with the other females. I excluded this

female from the analysis. One Main group female disappeared

near the end of the study. This female was included in the

analysis. Dominance hierarchies, based on direction of

supplantations and winners in agonistic encounters were

known for all groups (Robinson 1981, O'Brien unpubl. data).

The age of each female and the matrilineal geneologies were

known for all groups based on 10 years of annual censuses

(Robinson 1988; Fig. 3.1). All individuals were recognizable

by facial differences, and by variation in pelage, body size

and habit (Robinson 1981). The 3 groups yielded 75 pairs of

females dyadss) that could potentially interact.

I collected data for 15 months (December 1986 to March

1988, no data collected in June 1987). Daily behavioral

observations began between 0600 and 0630, and lasted for 12

hours. Observations consisted of a 20-minute focal sample on

an individual female, followed by a 20-minute scan sample of





























Figure 3.1. Matrilineal and dominance relationships between
female wedge-capped capuchin monkeys in 3 study groups.
Dominance rank is indicated below each female's identity and
relatedness is indicated by brackets. M-D indicates mother-
daughter relationships and S-S indicates sibling
relationships.










M




D/


2
I
S


1


/""\D


3


White Group


D/


2


Splinter Group


S S

I I
M7AZIA 1


AL


S

S


Main Gr


D


12

11


D


M,


I]DBU
DM D
5 1 9
MO MG
M< 6 7 S
oup D 8S


1


4


3


10


4










the group and a 20-minute opportunistic sample each hour

(Altmann 1974). Females were sampled systematically such

that each female was included in a focal sample in each of

three 4-hour time periods (morning, midday and late

afternoon). During the first 12 months of sampling, I

recorded all behaviors that lasted at least 3 seconds during

a focal sample. For the last 3 months I recorded only social

interactions that were defined as affiliative or aggressive

interactions involving two or more individuals. During the

scan sample, the behavior and nearest neighbor of each group

member were noted.

During opportunistic samples, I recorded all observed

social interactions involving females. I minimized

observation bias by moving constantly through the group

looking for social interactions. Because viewing conditions

were excellent, I probably missed very few social

interactions. There were no significant differences in the

frequency of fights and supplantations between focal and

opportunistic samples and I combined the data for analysis.

I analyzed 2 types of female-female aggressive social

behaviors:

1. Supplantation. Female approaches or looks at another,

causing other to move from an occupied position.

Supplantations were completely predictable by the relative

ranks of the interactants. The higher-ranking female always

supplanted the lower-ranking female.










2. Fight. Female stares at, vocalizes at, lunges at, or

chases other female. Occasionally grabbing or biting was

observed. Response of the victim included submissive

vocalization, defensive posture, or retreat.

Aggressive interactions could be ranked by level of

intensity. Supplantation, threats, and vocalizations were

relatively low intensity. Lunges and chases were more

intense and physical contact was considered the most intense

form of aggression because potential costs included injury.

Females often solicited for assistance but females were

never observed to form alliances against aggressors. Females

were observed forming coalitions against females (O'Brien in

prep.), but presence of a coalition partner did not reverse

the predicted outcome of the interaction. Females supported

winners in aggression, rather than losers.

For each aggression I recorded the identities of winner

and loser, the type of aggression, and, for fights, the

behaviors involved. I calculated an index of aggression

(AgI) for each dyad as the sum of supplantations and fights

divided by the frequency of nearest-neighbor distances less

than 10m. This measure adjusts the frequency of aggression

to variation in proximity of individuals (Robinson 1981).

Low scores result when females are tolerant of one another

whereas higher scores indicate either frequent fighting or

low rates of association. I analyzed the frequency of

supplantation and fighting and the aggression index for each










dyad for the effects of age, rank, relatedness, group

membership, and reproductive value. Age of interactant was

classified as juvenile, young adult, and adult. Young adult

females were at least 6 years old but were nulliparous at

the start of the study. Adults were multiparous at the start

of the study. No females changed age classification during

the study. Dominance rank of interactants was classified as

high, medium, and low. For a given dyad, female number 1 was

dominant to female number 2. Matrilineal relatedness was

classed as close (r=0.5), distant (r=0.125 to 0.25), and

unrelated (r<0.125). Reproductive values for females were

calculated for females in large and small groups using life

tables (Robinson 1988, O'Brien unpubl. data) and standard

formulae (Caughley 1976).

Nearest-Neighbor Distances

I analyzed the frequency that females were in

association with other group members during all behaviors

and during foraging periods to determine if females had

similar opportunities to interact with group members and

where differences in time spent alone occurred. Because

competition for food often results in aggression within and

between groups, I expected that females would spend more

time alone and less time in close proximity during foraging

than in general. Nearest neighbor distances estimated to the

nearest meter and divided into 0-2m, 3-6m, 7-9m and 10+m

classes. Distributions of close proximity (0-2m) and time










spent alone (10+m) were compared within and between groups

for all behaviors and for foraging alone using 1-way ANOVAs

on arcsin squareroot transformed proportions of nearest-

neighbor distances in each distance class (Sokal and Rohlf

1981). The frequency of nearest neighbors for each dyad

relative to the average for all other dyads was computed for

an association index (AI). The index increases from 0 as a

pair of females increase their frequency of association. AI

was analyzed for effects of group membership, age rank and

relatedness of interactants using analysis of variance.

Effects of Sociodemographic Variables

To assess the importance of different sociodemographic

variables to female aggression, I used a linear model

analysis of the frequency data. I chose a discrete

regression model (Poisson regression) that computes

parameter estimates using maximum likelihood methods

(McCullough and Nelder 1983). The analysis is similar to a

multidimensional contingency table analysis or log-linear

model analysis. The Poisson distribution is the basic

distribution for count data and Poisson regression assumes

this distribution for the error term. Frequency of a

behavior for a dyad was used as the dependent variable, the

other behavior was used as a covariate, and sociodemographic

characteristics of the interactants were used as independent

variables. A separate analysis was performed for each

behavior.










All sociodemographic variables that I considered,

except female reproductive value, were categorical variables

with discrete rather than continuous distributions.

Parameter estimates for categorical variables indicate the

magnitude of an effect that can be assessed for statistical

significance using T-tests. Reproductive value and

frequencies of other social interactions were included as

continuous variables and resulting parameters were tested

against the null hypothesis that the parameter was equal to

0 using T-tests.

A stepwise regression procedure was used in each

analysis. First an overall mean was fit to the data and then

each independent variable was added separately to the model.

A variable was considered for inclusion in a model if (1) it

resulted in a significant (P<0.05) improvement in the

percent of variance accounted for by the model, (2) resulted

in the smallest mean square error of variables considered,

and (3) accounted for at least 5% of the total variance in

the data. Variables that accounted for the most variance

were sequentially included in the model. Variables were

included until the addition of variables failed to satisfy

criteria for inclusion. This method allowed me to evaluate

the relative importance of independent variables affecting

aggression based on the percent of the total variance in the

data attributed to each variable. It also allowed me to

compare results of regression models for each behavior to










determine if both analyses shared important variables.

Shared variables indicate that an effect is an important

determinant of aggression rather than fights or

supplantations alone.

After the effects of independent variables in a model

are accounted for, a significant amount of variability may

remain in the residuals (observed error). This may indicate

that one or more variables have not been considered in the

model (detected as nonrandom patterns in a residual plot) or

that certain dyads are interacting more or less than the

model predicts (detected as outlier points on a residual

plot). I identified dyads with large residuals, defined as

more than 1 standard deviation (SD) above or below the mean

standardized residual (Neter and Wasserman 1974), and rated

them as moderate (between 1 and 2 SD) or strong (>2 SD). The

residuals are used to identify dyads that were unusually

aggressive or tolerant. Dyads with strong positive

residuals for both behaviors are composed of individuals

that have aggressive relationships (analogous to special

relationships: Seyfarth 1978b; Altmann 1980; Smuts 1985).

Residuals also were analyzed to determine the importance of

identity of interactants to aggression after other effects

were accounted for.










Results

Association Among Females

The distributions of nearest-neighbor distances for

females varied considerably within and between groups (Fig.

3.2; Table 3.1). Females in White group spent less time in

close proximity (1-way ANOVA; F=4.52, d.f.=2,16, P=0.028)

and more time without nearest neighbors (1-way ANOVA;

F=29.82, d.f.=2,16, P<0.0001) than females of Main group or

Splinter group. Females spent significantly less time in

close proximity when foraging than in general (paired T-

test; t=7.77, d.f.=18, P<0.0001) and significantly more time

without nearest neighbors (paired T-Test; t=-4.50, d.f.=18,

P=0.0003). Females in Splinter group spent more time in

close proximity to other group members when foraging than

the other females (1-way ANOVA; F=18.21, d.f.=2,16,

P=0.0001) and White group females spent more time without

nearest neighbors while foraging than other females (1-way

ANOVA; F=17.02, d.f.=2,16, P=0.0001). At 2-9 meter

distances, there were no differences in the distribution of

nearest neighbors among females.

The average AI for the 3 groups was 1.031 and there

were no significant differences in AI between groups (Fig.

3.3). AI ranged from 0.53 to 1.45 in White group, 0.43 to

1.58 in Splinter group, and 0.08 to 3.38 in Main group. An

ANOVA that considered the effects of dominance rank, age,

and coefficient of relatedness (r) of females on AI showed
















Table 3.1. Distribution of nearest neighbor distances within
and between groups. Standard Deviations (SD) within each
group, and within and between group variances are given.
Distances are classified for all behaviors and distances
during foraging only.

Nearest Neighbor Distances, All Behaviors
0-2m 3-6m 7-9m 10+m


White
Mean
SD


Splinter
Mean
SD

Main
Mean
SD

Variance
Between Groups
Within Groups


0.037
0.026


0.133
0.028


0.097
0.047


0.0081
0.0018


0.213
0.021


0.311
0.060


0.342
0.047


0.0198
0.0022


0.119
0.062


0.197
0.029


0.215
0.033


0.0110
0.0016


0.631
0.012


0.359
0.064


0.346
0.061


0.1004
0.0034


White
Mean
SD


Splinter
Mean
SD

Main
Mean
SD

Variance
Between Groups
Within Groups


Nearest Neighbor Distances, Foraging Only
0-2m >2-6m >6-9m >9-10+m


0.000
0.000


0.066
0.017


0.009
0.012


0.0054
0.0003


0.265
0.028


0.333
0.049


0.289
0.067


0.0045
0.0036


0.109
0.074


0.220
0.088


0.271
0.046


0.0321
0.0036


0.626
0.061


0.381
0.041


0.431
0.062


0.0583
0.0034































Figure 3.2. Distribution of nearest neighbor distances for
all behaviors and while foraging for each female.




















FEE= -01001\1-
-INES-M


1 0.8 0.6


0.4 0.2


0 0.2 0.4


0.6 0,8


All behaviors
-0-2m M 3-6m


Foraging only
- 7-9m = 10+m


PN
PE
RO
MA
FF
GE
IR

WH
HA
AM
AL
ON
MO
MG
ML
BU
CR
PF
PU


IMESSNEIII


mffilffiffilll\\\\\ W\\\\\ ME MEE,~\~


I


I


Flom
cum MEMILIMAI'm
LIMMIMMISSINEIN
EW,9rlllzA ME EM, 11110\11111
WIFEEEIZZM '11\1100\111

WE P, m a = ME a


-Nam4


MIUM \100019 Emmm


I


FEE= \110101IN \\\\\\-~,\\\\


I

































Figure 3.3. Distribution of Association Index (AI) among 3
groups and among rank and relatedness combinations in Main
group.
























WHITE SPLINTER MAIN


0


HIGH MEDIUM HIGH HIGH MEDIUM LOW
HIGH MEDIUM MEDIUM LOW LOW LOW
FEMALE DOMINANCE RANK COMBINATIONS


0.5
SIBUNG


0.25-.125


DEGREE OF MATERNAL RELATEDNESS (r)


4.00-


3.00-



Z 2.00-



1.00.



0.00


I.


3.00


2.00-




1.00.




0.00-


0.5
MOTHER


.











that most variation in AI was due to the dominance rank of

the higher-ranking female in the dyad and to r (Table 3.2).

Dyads composed of related individuals (r=0.5, AI=1.657;

r=0.125-0.25, AI=1.871) associated more than dyads composed

of unrelated females (r=0, AI=0.776). The top 25% of AI

scores were primarily (12 of 19) among related individuals.

Among closely related females, mothers and daughters

associated more often than siblings (Fig. 3.3). High-ranking

females associated more often than dyads in which a female

was low-ranking (AI=2.409 vs. 0.663; Fig. 3.3). Associations

with and between medium-ranking females were intermediate.

Low-ranking females had the lowest AI values of any dyads,

indicating that they were peripheral females.

The above analysis accounted for approximately half of

the variation in AI. Standardized residuals were randomly

distributed (Wilks-Shapiro statistic=0.96), indicating

important variables were not omitted. I attributed the

residual variability in the data to individual preference

among females. A maximum spanning tree analysis (MST; Morgan

et al. 1974; Fig. 3.4) using AI values shows which females

are closely linked. In White group, The mother (FF) is the

focal female in the group. In Splinter group, the juvenile

female (PE) was closely linked to the other adult females

but the adults were not closely associated with one another.

In Main group closely associated females were related and/or

high-ranking. Only females of the lowest-ranking matriline

















Table 3.2. Main effects ANOVA for AI. Mean squares are
calculated and tested using Type III sums of squares. The
number after each variable identifies characteristics of
female 1 and female 2. The first female is higher ranking.

Model d.f. Mean Square F P>F

Rank 1 2 1.115 4.425 0.016

Rank 2 2 0.352 1.397 0.254

Age 1 2 0.556 2.206 0.116

Age 2 2 0.201 0.798 0.458

r 2 1.236 4.905 0.010


Error 64


0.252































Figure 3.4. Maximum Association Index (AI) scores among
pairs of females in 3 groups. a) Maximum spanning tree for
AI scores. Boxes indicate relatedness of r=0.5. b) Ranked
maximum AI scores for all females in study. Females in boxes
are from small groups. AI=2.1 is used as middle reference
line; scores greater than 2.1 indicate at least 1 close
association.
















IGE ---FF -- IR


1.45


1.17


PN-PE --MA
1.8 1.58
1.10

RO


WH AM AL



ON HA
MO ML


2-
2 ------.----------- ----------- ----------------------------------------- --- --- -

MG PU PN PE MA
CR FF GE

1 -PF
BU


n











were not connected. Patterns of dominance rank and

relatedness were not predictable when juvenile females were

considered. Juveniles associated often among themselves (ML-

PU=1.65, HA-ML=1.74) and HA associated with lower-ranking

adults (CR and MG).

I next ranked females by their maximum AI scores to

show that group membership and dominance rank affected AI

values among females (Fig. 3.4). High- and middle-ranking

females in Main group form at least 1 close association

(AI>2.0), whereas low-ranking females of Main group and all

small-group females did not form close associations among

themselves. That small-group females exhibit AI patterns

characteristic of low-ranking females, irrespective of their

own rank and relatedness patterns, confounds generalizations

concerning rank and relatedness. Perhaps these factors are

meaningful only as group size increases.

Aggression Among Females

I observed 222 fights and 186 supplantations among

females. Main-group females were more aggressive, per dyad,

than small-group females (Fig. 3.5; Table 3.3). In small

groups, supplantations were more common than fights, and

foraging was the most common context for aggression. Females

in small groups also had much lower Aggression Index (AgI)

scores than females of Main group (Fig. 3.6), suggesting

greater tolerance for one another. Among Main group females,

in contrast, fights were more common than supplantations and
















Table 3.3. Involvement in fights and supplantation during
foraging for a subset of data in which context was identified for
each female. Data are presented by % for aggressive interactions
won by each female and total involvement. Sample sizes are given
in parentheses.


Fight
50 (24)
56 (9)
36 (11)
0 (7)
18 (17)
33 (3)
0 (1)
100 (1)
100 (2)
100 (3)
0 (0)
0 (0)


Wins
Supplantation
39 (36)
50 (10)
50 (12)
86 (14)
57 (7)
0 (0)
0 (0)
0 (0)
0 (0)
100 (1)
0 (0)
0 (0)


(3)
(1)
(0)
(0)

(0)
(0)
(0)


Fight
50 (24)
56 (9)
31 (13)
8 (13)
23 (22)
43 (14)
19 (16)
20 (10)
67 (3)
80 (10)
0 (2)
65 (17)


100
0
0
0

0
100
100


Total
Supplantation
39 (36)
50 (10)
43 (14)
56 (26)
47 (17)
70 (10)
71 (14)
60 (10)
75 (4)
62 (8)
0 (2)
38 (8)


100
100
0
0

100
0
100


Female
WH
HA
AM
AL
ON
MO
MG
ML
BU
CR
PF
PU


100
0
0
0

0
100
0


(1)
(0)
(0)
(0)

(0)
(1)
(0)


100
100
0
0

0
0
0
































Figure 3.5. Distribution of aggression for a) 2 small groups
and b) 1 large group of wedge-capped capuchin monkeys.