Expanding Scales of Influence: Behavioral, Physiological, and Reproductive Implications of

Relative Power within Social Groups

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By

Isaac Y. Ligocki

Graduate Program in Evolution, Ecology and Organismal Biology

The Ohio State University

2015

Dissertation Committee:

Ian M. Hamilton, Advisor

Elizabeth A. Marschall

Steven W. Rissing

J. Andrew Roberts

Copyright by

Isaac Young Ligocki

2015

Abstract

Individual behavior and the social interactions of individuals in groups are related to one another.

Individuals may respond dynamically to interactions with other group members, and their behavior may influence that of those they interact with. I define such interactions as direct interactions, and the resulting effects as direct effects. Individual behavior may also influence and be influenced by the behavior and interactions of group members with whom they do not directly interact. I define such interactions as third party interactions, and the resulting effects as third party effects. In cooperatively breeding groups, in which some group members forego or limit their own reproduction and provide care for the offspring of other group members (alloparental care), third party interactions are predicted to have potentially dramatic effects on group characteristics because group members of different sex or social status may be in conflict regarding the extent of reproductive skew or expectation of alloparental care. The primary aim of the studies in this dissertation was to investigate the impact of third party interactions in the cooperatively breeding fish Neolamprologus pulcher through lab- and field-based studies as well as a game theoretical model. In chapter 2, I determined that the relative size of the dominant male and female fish influenced their direct interactions with one another as well as their relative response towards conspecific and heterospecific intruders on their territory. In chapter 3, I found that dominant fish are in conflict regarding the presence and role of subordinate male and female group members. In chapter 4, I concluded that direct agonistic interactions did not influence subordinate female cortisol levels, but one measure of agonism between the dominant pair was associated with elevated cortisol levels in subordinate female fish. In chapter

ii

5, I investigated how members of naturally formed groups respond to territorial intrusions

(determined to be perceived as potential joiners to the group) depending on their social status within the group and whether a vacancy existed in the group which the intruding fish could fill.

Size-matched subordinate fish and dominant females were most aggressive towards intruders.

Dominant and subordinate fish also shifted their response in opposite ways after the removal of a large subordinate group member, suggesting dominants and subordinates are not in agreement regarding group size or the addition of potential joiners to the group. In chapter 6, I developed game theoretical models to predict when subordinate females should attempt to reproduce within groups, and examined how the order of decisions in games influences their outcome. Collectively, this research builds on previous work on conflict within groups by providing evidence that individual group members do not always agree on the role and presence of other group members.

I extend this body of research by showing that this lack of consensus leads to third party effects within groups which may have implications for group size and membership, reproductive skew in groups, and the extent to which subordinates participate in alloparental care.

iii

Acknowledgments

First and foremost I owe earnest thanks to my advisor Dr. Ian Hamilton for his continual support and guidance. His example as a researcher and advisor has been one of my most formative experiences while having the privilege of being a student in his lab. I am also especially grateful to my committee members Dr. Elizabeth Marschall, Dr. Steven Rissing, and Dr. Andrew Roberts for their commitment to my development as a scientist. I also owe sincere thanks to numerous collaborators who I’ve had the pleasure of working with on numerous projects including Dr. Sigal

Balshine, Dr. Ryan Earley, Jennifer Hellmann, Dr. Constance O’Connor, Dr. Adam Reddon, and

Susan Marsh-Rollo. I am indebted to the members of the Animal Behavioral and Ecological

Complex Systems (ABECS) Lab, as well as numerous other graduate students, faculty, and staff of EEOB for their guidance, advice, and support for me as a graduate student, colleague, and friend. I am also grateful for the pedagogical support and educational opportunities I been given through the Center for Life Sciences Education and the University Center for the Advancement of

Teaching at OSU. Lastly, I am forever grateful for my family; who always fostered my interest in science and have continually supported and encouraged me in both my academic and non- academic pursuits. My Ph.D. research was supported by the Metro Fellowship, the OSU Graduate

Teaching Fellowship, OSU Alumni Grants for Graduate Research and Scholarship (AGGRS), the

OSU Council of Graduate Students (Ray Travel Award), the OSU Fish Systematics Endowment,

SciFund Challenge, and the Department of Evolution, Ecology, and Organismal Biology.

iv

Vita

2002 ...... Dublin Coffman H.S., Dublin, OH

2006 ...... B.A. Biology, Wittenberg University

2006-2009 ...... Teacher, Dunedin H.S., Dunedin, FL

2009 to present ...... Graduate Teaching Associate, Dept. Evol.,

Ecol., Org. Biol., The Ohio State University

Publications

O’Connor CM, Reddon AR, Ligocki IY, Hellmann JK, Garvy KA, Marsh-Rollo S, Hamilton IM, Balshine S. Accepted. Motivation but not body size influences territorial contest dynamics in a wild cichlid fish. Animal Behaviour.

Hellmann JK, O’Connor CM, Ligocki IY, Farmer TM, Arnold TJ, Reddon AR, Garvy KA, Marsh-Rollo SE, Balshine S, Hamilton IM. In Press. Evidence for alternative male morphs in a Tanganyikan cichlid fish. Journal of Zoology.

Garvy KA, Hellmann JK, Ligocki IY, Reddon AR, Marsh-Rollo SE, Hamilton IM, Balshine S, O’Connor CM. 2014. Sex and social status affect territorial defence in a cooperatively breeding cichlid fish, Neolamprologus savoryi. Hydrobiologia. doi: 10.1007/s10750-014-1899-0.

O’Connor CM, Reddon AR, Hellmann JK, Ligocki IY, Marsh-Rollo SE, Hamilton IM, Balshine S. 2014. A comparative study of immune response and social system in Lamprologine cichlid fishes. Naturwissenschaften. 101:839-849.

Hamilton IM, Ligocki IY. 2012. The extended personality: indirect effects of behavioural syndromes on the behaviour of others in a group-living cichlid. Animal Behaviour. 84:659-664.

Fields of Study

Major Field: Evolution, Ecology and Organismal Biology

v

Table of Contents

Abstract ...... ii

Acknowledgments ...... iv

Vita ...... v

Publications...... v

Fields of Study ...... v

Table of Contents...... vi

List of Tables ...... ix

List of Figures ...... x

Chapter 1: Introduction...... 1

Chapter 2: The influences of size and the social environment on territory defense in a social cichlid fish ...... 6

Introduction: ...... 7

Methods:...... 9

Results: ...... 14

vi

Discussion: ...... 18

Chapter 3: Third party effects in a cooperatively breeding cichlid: effects of subordinate sex and relative size ...... 22

Introduction: ...... 23

Methods:...... 26

Results: ...... 29

Discussion: ...... 41

Chapter 4: Variation in glucocorticoid levels in relation to direct and third-party interactions in a social cichlid fish...... 45

Introduction: ...... 46

Methods:...... 49

Results: ...... 55

Discussion: ...... 60

Conclusions:...... 63

Chapter 5: Social status influences responses to unfamiliar conspecifics in a cooperatively breeding fish ...... 65

Introduction: ...... 66

Methods:...... 68

Results: ...... 75

vii

Discussion: ...... 76

Chapter 6: Strategic decision making in a three-player sequential game: order effects influence the outcome in cooperatively breeding groups...... 80

Introduction: ...... 81

The Model: ...... 83

Results: ...... 92

Discussion: ...... 95

Chapter 7: Conclusion ...... 100

References...... 104

Appendix A: N. pulcher Ethogram (Chapter 4)...... 131

Appendix B: Principal Component Loadings (Chapter 4) ...... 133

viii

List of Tables

Table 1: Fixed effects in final LMMs with square-root transformed counts of aggressive acts towards intruders by parental males (a) and parental females (b) towards intruders. 15

Table 2: Loadings of behaviors in Principal Component Factors 1 and 2 for each dyadic interaction in Experiment 1...... 29

Table 3: Output of GLMMs with PC scores as dependent variables...... 30

Table 4: Loadings of behaviors in Principal Component Factors 1 and 2 for each dyadic interaction in Experiment 2...... 36

Table 5: Output of GLMMs with PC scores as dependent variables...... 37

Table 6: PC Factors for pairwise interactions amongst group members...... 56

Table 7: Output of Linear Mixed Models which considered log-transformed subordinate female cortisol levels as the dependent variable...... 59

Table 8: Parameters and boundaries of parameter space explored in models...... 83

Table 9: Fitness outcomes for each of the eight potential outcomes of the game...... 86

Table 10: Sequence in which individuals made decisions in each of the six games...... 89

ix

List of Figures

Figure 1: Plot of solution of LMM with square-root transformed resident female aggression towards intruders as the dependent variable...... 16

Figure 2: Plot of counts of aggressive acts by the parental male towards the intruder versus counts of aggression by the parental female towards the parental male...... 17

Figure 3: Plot of counts of aggressive acts by the parental female towards the intruder versus counts of aggression by the parental female towards the parental male...... 17

Figure 4: Residual of E1 DMDF Factor 1 versus Dominant Female SL (mm)...... 32

Figure 5: Best fit surface of E1 DMSM Factor 1 versus Dominant male and dominant female SL (mm)...... 33

Figure 6: Residual of E1 DFSM Factor 1 versus dominant male (a) and subordinate male

(b) standard length...... 34

Figure 7: Residual of E2 DMSF Factor 1 versus subordinate female standard length.. .. 39

Figure 8: Best fit surface of E2 DFSF Factor 2 versus dominant male and dominant female standard length...... 39

Figure 9: Relationship between subordinate female cortisol and one measure of agonistic interactions (restrained aggression and a lack of courtship) between the dominant male and dominant female (DMDF Factor 2, see Table 7c)...... 57

x

Figure 10: Relationship between subordinate female cortisol and a measure of affiliative behavior (joining and bumping) by the subordinate female towards the dominant female

(DFSF Factor 2, see Table 7b)...... 58

Figure 11: Boxplot of counts of displays (white) and attacks (gray) towards the stimulus fish in control groups (n=22)...... 72

Figure 12: Boxplot of aggressive acts towards the stimulus fish; a) attacks by dominant females, b) displays by dominant females, and c) displays by dominant males in control groups (n=22) and removal treatments (n=21)...... 73

Figure 13: Boxplot of a) displays and b) attacks towards the stimulus fish by non-focal subordinate group members in the removal treatment and control groups...... 74

Figure 14: Extensive form depiction of a sequential game in which subordinate females make the first decision, followed by the dominant female, and then the dominant male. 84

Figure 15: Outcome of the six sequential games plotted against relatedness (r) and the costs of policing (CPOL)...... 90

Figure 16: Outcome of the six sequential games plotted against relatedness (r) and the costs of policing (CPAT)...... 91

xi

Chapter 1: Introduction

Variation in the extent of sociality within and between species is hypothesized to reflect the net sum of the costs and benefits of social living. These benefits may lie in protection from predation when in a group (1-5). Additionally, hunting or foraging may be more effective when done collaboratively with other individuals (6, 7). Sociality may also allow for greater access to mates or reproductive opportunities (8). Social behavior is also associated with costs for individuals. In some species, increasing group size may be associated with increased conspicuousness to predators (9) or more rapid exhaustion of limited resources (10). Some individuals in groups may actually receive fewer reproductive opportunities than they otherwise would (11), and the offspring of some individuals may be competitively disadvantaged in a social setting (12). The net costs and benefits associated with sociality often varies depending on an individual’s status within the group, but individuals also are able to influence the fitness of other group members and thus shape the emergent social environment.

Individuals living in groups are components of the social environment within that group, and their presence and behavior is a cause and consequence of the behavior of other group members. Individuals and their behavior can have direct effects on those with whom the individual has direct interactions (13-15), but also may correlate with aspects of the social environment with which they are not directly involved (16, 17). For instance, Flack et al. (18,19) found evidence of third-party interactions in pigtailed macaques (Macaca nemestrina), and after removing key third-parties from the group the social environment shifted and the group’s structure destabilized.

1

The costs and benefits associated with a group’s social environment are not equal for all members; what is good for one individual may be suboptimal for another. If some group members are older, larger, or better-connected socially, they may be better able to influence the social environment to reflect their own interests (“power” – 20, 21).

Cooperatively breeding groups provide an ideal opportunity to examine how power influences the outcome of conflict between group members with contrasting optimal social environments. These groups are characterized by the delayed dispersal of offspring from their natal groups, high reproductive skew favoring more-dominant group members, and alloparental care provided by non- or less-reproductive subordinate group members. Cooperatively breeding systems are taxonomically widespread, found in arthropods (22), birds (23), mammals (24), and fish (25, 26). Subordinates in these groups presumably would accrue greater fitness gains through direct reproduction, whereas dominants are expected to benefit from subordinates foregoing reproduction to provide alloparental care. Dominants may face reproductive competition from same-sex subordinates and resist their entry into the group, whereas dominants may welcome opposite-sex subordinates as potential mates (27, 28). Subordinates and dominants alike may engage in inter- and intragroup conflict with individuals who might challenge their status in the group (27-29).

The aim of this dissertation research is to investigate whether and how differences between group member’s sex, status, size, and relative power influence the 1) direct interactions within groups and towards non-group members, the 2) indirect, third-party interactions within groups, and 3) their influence on the emergent social environment of the cooperatively breeding cichlid fish Neolamprologus pulcher.

In chapter two, I describe an experiment in which I tested whether asymmetries in body size (a determinant of dominance and the ability to win fights, and thus a likely influence on

2 power) between dominant males and females influences their participation in territorial defense against conspecific and heterospecific intruders. In this study I found that dominant males did not adjust their investment in territorial defense depending on their own or their mate’s size. In contrast, dominant females did adjust their aggression towards intruders depending on their own size as well as her size relative to that of her mate, such that smaller dominant females, and small dominant females paired with small dominant males were more aggressive towards intruders of all types. Lastly, in this study I found that both male and female dominants behaved aggressively towards mates who were less active in defending the territory against staged intrusions. This suggests that females, but not males adjust their defensive behavior depending on their social environment, and that both sexes may use punishment to enforce participation in parental behaviors.

In chapter three, I investigated how the direct and indirect interactions of dominant male, dominant female, and either male or female subordinates were influenced by the individual and relative sizes of group members. I found that subordinate males faced greater aggression from dominant males which were much larger than their mates, and that subordinate females faced greater aggression from dominant females when dominant females were similar in size to dominant males. These results suggest that dominants differ in their optimal social environment, and their dyadic interactions with same-sex subordinates are indirectly influenced by the relative size of dominant fish.

In chapter four, I examined correlations between subordinate female cortisol levels and direct and indirect interactions within groups consisting of a dominant breeding pair and a subordinate female. Cortisol is a glucocorticoid steroid hormone associated with the stress response of vertebrates (30). Subordinate female cortisol levels were positively correlated with direct non-aggressive interactions with the dominant female. Interestingly, I also found that

3 subordinate female cortisol levels were positively correlated with agonism between dominant males and females. Therefore, both direct and indirect interactions influence the stress response of subordinate females in groups. I suggest that escalated conflict between dominants may be indicative of instability within the group, and that this instability is a stressor for subordinate females in groups.

In chapter five, I exposed wild, naturally formed groups of N. pulcher to unfamiliar conspecific intruders in order to simulate a visitation preceding a group joining event. In one treatment, groups remained intact. In the other, a large subordinate of the same sex and roughly the same size as the unfamiliar conspecific had been removed. After a removal, dominant fish

(who were not at risk of being usurped) were less aggressive towards the intruder, while smaller subordinates (who were at risk of being usurped) behaved more aggressively. Taken together, these results suggest dominants benefit from replacing lost subordinates while smaller subordinates do not.

In chapter six, I describe a three player sequential game theoretical model I developed to determine when subordinate females in groups should attempt to reproduce, when dominant females should police subordinate reproduction and attempt to destroy subordinate clutches, and when dominant males should defend subordinate female clutches which he fathered. This model was structured as a sequential game (individuals make decisions in a particular order), and all six possible sequences in which the three players could make decisions were simulated. The results of this model suggest that when subordinate females make decisions relative to dominant females influences the outcome in nearly all circumstances, whereas when dominant males decide whether to patrol only influences the outcome of the model when patrolling is extremely cheap and subordinate female clutches are large. In all scenarios in which the order of decisions

4 influenced the outcome of the game, individual payoffs were higher when they played earlier in the game.

5

Chapter 2: The influences of size and the social environment on territory defense in a social

cichlid fish

Abstract: In social groups, individuals often share some common interest in the successful completion of particular tasks or activities, such as care of offspring or territory defense. Because parental care is costly, parents may benefit from reducing their investment in parental care, especially if their mates compensate for their reduction in care. In the present study, we examined the relationship between the size of parents and the relative size of parents on each parent’s participation in territory defense. We exposed breeding pairs of the cooperatively breeding cichlid

Neolamprologus pulcher to staged territory intrusions by conspecific and heterospecific fish, varying the relative size of the parents (resident fish). Male resident fish did not vary in their defensive response depending on the type of intruder or their size relative to their mate. Female resident fish performance of defensive behaviors towards all intruder types depended on her size relative to that of her mate. Small females paired with large males performed the most defensive behaviors while large females paired with large males performed the fewest. Neither parent compensated for mates who participated less in defensive behavior, but both male and female resident fish who performed fewer defensive behaviors received more frequent aggressive behaviors from their mate. These findings suggest that female, but not male parents adjust their participation in defensive behavior depending on their social environment, and that both sexes may use punishment to enforce participation in some parental behaviors.

6

Introduction:

In with parental care, the provision of care is presumed to enhance the survival and success of offspring and thus lead to fitness gains for both parents (31-34). Both parents benefit from increased investment in care, even if only one parent actually provides it. Parental behaviors are not without costs, however; behaviors such as brood defense are energetically costly (35). Additionally, parental care may be associated with increased predation risk for parents (36, 37), and increased parental investment may limit a parent’s ability to invest in additional offspring (38). Biparental care is presumed to be favored when fitness gains are greater for both parents to stay rather than for each to move on in search of additional reproductive opportunities, leaving the offspring in the care of the other parent (34, 39, 40). Thus, in biparental systems, decisions regarding participation in parental care depend on a the net costs or benefits parents face if they perform parental care, as well as how much parental care can be expected from their mate.

One parent may primarily perform particular tasks, and individuals may specialize in a subset of behaviors depending on their sex, age, or social status (41, 42). In biparental systems, there is often variation in the investment in parental care between males and females (39, 43).

Sex-specific specialization has been documented in a number of biparental species, and in cichlid fishes is generally characterized by females primarily defending against heterospecific predators and providing direct care to young; while males primarily defend against conspecific intruders

(29, 45, 46). These patterns may result because of sex-specific differences in the costs and benefits of the current brood (39, 47, 48).

However, variation in apparent differences in parental investment between males and females may result from differences in the net costs and benefits associated with performing particular parental roles irrespective of sex, because these costs and benefits scale with variables

7 such as size, condition, or experience, each of which may covary with sex. For example, larger individuals may be better able to defend offspring from predators. Itzkowitz et al. (49) varied the relative size of breeding pairs of convict , Amatitlania nigrofasciata, and found that while males were consistently the primary defenders against conspecific intruders, larger individuals of both sexes displayed more defensive behavior than smaller individuals of the same sex.

In addition, variation in parental investment in biparental systems may result from interactions among parents. Reciprocal cooperation requires that parents invest only if the other parent does so as well (50-52). Under negative reciprocity, an individual’s investment in joint defense is enforced by punishment from its partner (53, 54). Both positive and negative reciprocity can lead to a cooperative outcome in which both parent’s investment in defense is high. Under compensation, parents invest more if their partners invest less. Generally, full compensation is not expected in biparental systems (55), however, partial compensation is frequent (56). At the extreme, compensation may allow an individual to directly or indirectly coerce its mate into performing costly parental behaviors more frequently, thereby reducing its own parental care costs (34, 57, 58). Coercion in this sense could occur either through the threat of punishment (53), or result from refusing to participate in care, leaving the other mate to compensate (59).

Here, we examined the effects of an individual’s sex and size, as well as its partner’s size and behavior on offspring defense by exposing long-established and successfully breeding pairs of the cooperatively breeding cichlid fish Neolamprologus pulcher to the fry-specialist predator

Altolamprologus compressiceps (61, referred to hereafter as heterospecific intruders) as well as both male and female subordinate-sized N. pulcher (referred to hereafter as male and female conspecific intruders, respectively). If variation in parental investment reflects only differences in the costs and benefits of the current brood depending on the sex of the parent, we predict that

8 defense will exclusively or mainly fall to one sex, and not vary depending on relative size of males and females (although it may vary depending on absolute size if costs and benefits of the current vs. future broods depend on size or age). If variation in investment in defense arises from variation in ability to perform care, we expect that participation in defense of offspring increases with size, as larger individuals experience lower costs and higher chances of winning during fights. However, size may also influence interactions among parents. In N. pulcher, like many social fish, size is a reliable predictor of fighting ability and indicator of dominance (26, 49, 61-

63). Individuals that are large relative to their mates may be better able to punish or coerce mates

(they have more “power” - 20, 21). If so, then we expect that relatively large fish will perform less offspring defense. We also tested for correlations between aggression toward mates and intruders for each parent, and between parents as we predict positive correlations in each parent’s participation in defense under positive reciprocity and negative correlations in each parent’s participation in defense under compensation. Additionally, we predict positive correlations between each parent’s defense and aggression received from its mate if it is coerced into defending; we predict a negative correlation between a parent’s defense and aggression received from its mate if it is punished for not defending.

Methods:

Study Species:

Neolamprologus pulcher is a cooperatively breeding cichlid native to in

East Africa, which has become a model system for examining the evolution of cooperative behavior (26). N. pulcher form groups consisting of a breeding pair and an average of 7 (64) to 9

(65) subordinate helpers of both sexes. N. pulcher has a sex-specific dominance hierarchy, in which dominance is correlated with relative length (26). Dominant breeders and subordinate 9 helpers perform a number of behaviors within the group to maintain the territory. These include direct care of eggs and fry, territory maintenance (digging, moving, and cleaning sand), and defense against both conspecific and heterospecific intruders, including a number of predatory species. Groups in our study only included a dominant breeding pair and their sexually immature offspring (<25 mm). We refer to the members of the focal breeding pair as resident males and females.

At the time of the experiment, pairs varied in relative size. Because dominance is size- based in this species (26, 62, 66), relative size is likely to be a determinant and consequence of power. Thus, we expect the resolution of conflict to reflect relative sizes of the resident male and female. While males are invariably larger than females in this species, we assume that females are better able to resist male coercion or punishment (or to coerce or punish males) when they are only slightly smaller than males.

Experimental Design:

15 breeding pairs of N. pulcher were created in the lab over one year prior to the beginning of experimentation. In that time all groups had produced at least one clutch of eggs and raised offspring beyond the fry stage (~two weeks of age, when fry are free swimming and foraging for food). All individuals were either wild caught from the Kipili (Tanzania) region of

Lake Tanganyika, or were F1 individuals from wild stock collected from the same region. These groups were maintained in 113.6 L (30.5 cm X 61cm X 61cm height) with a sand substrate of an average depth of 30 mm. Each contained a halved clay flowerpot which served as a breeding substrate. Fish received daily feeding ad libitum of either dry TetraCichlid

Cichlid Flakes (Tetra Holding (US) Inc. 3001 Commerce St. Blacksburg, VA 24060) food (5x week) or Frozen (Hikari Sales U.S.A., Inc. 2804 McCone Ave., Hayward, CA 94545) and (San Francisco Bay Brand, Inc. 8239 Enterprise Drive, Newark, CA 94560)

10

(2x week). All tanks were kept on an illumination cycle of 12 L : 12 D, and conditions in these tanks were maintained to reflect those in Lake Tanganyika (Mean Temperature = 25.4 + 0.7 ° C, pH = 7.8-8.4).

Resident male and female standard length (SL: measured from the tip of the nose to the end of the caudal peduncle) varied, as did the ratio of resident male to female standard length in groups. We used variation in both absolute and relative size to better distinguish between differences resulting from variation in individual size, differences resulting from variation in the size of an individuals’ mate, and differences resulting from variation in relative size. In natural groups, the dominant male is typically the largest group member, the dominant female is the second largest (26); we maintained this in our pairs. At the time of this experiment, resident males ranged in standard length from 53 to 70 mm (mean±SE =60.13+4.82 mm) and resident females ranged in standard length from 50 to 58 mm (mean±SE = 53.67+2.23 mm).

Over the course of 3 weeks, pairs were presented with each of three territorial intruders in a random sequence: 1) A. compressiceps; (F1s from Kipili, Tanzania: mean SL: 32.6+2.746 mm) as well as 2) male and 3) female N. pulcher (mean unfamiliar male SL = 46 +1.51 mm; mean unfamiliar female SL = 47 +2.68 mm). These intruders were presented in a 1.5 L plastic jar placed approximately 20 cm from the breeding shelter. This jar was sealed to eliminate the risk of contradictory chemical cues influencing the response of focal fish. Similar “fish-in-a-jar” experiments have been performed in this and other fish species (29, 67-70). The observed response of individual N. pulcher to jarred intruders is consistent with how fish naturally respond to unfamiliar conspecifics (29, pers. obs.). In an unrelated study performed later on these same pairs, we found that individuals did not respond aggressively towards empty jars, and performed significantly fewer aggressive acts towards model predators than the live predators used in this study (unpublished data). We selected small A. compressiceps specimens, such that they were not

11 a threat to either member of the pair, and that either dominant should have been capable of defending against it, but were large enough that they would have been able to consume juvenile

N. pulcher less than 10 mm.

All behavioral observations were performed by the same observer and occurred between

6-2-12 and 28-2-12 and between the hours 13:00 and 15:00. Each observation lasted 15 minutes.

Behaviors were identified based on the behavioral descriptions of Coeckelberghs (71), Kalas (72), and Taborsky (73); and recorded using the program JWatcher 0.9 (74). Counts of all behaviors performed by either member of the dominant pair were recorded during the observation. No group was observed more than one time per week, and the order of observations was randomized.

At least 6 hours prior to all behavioral observations, any juveniles larger than 10 mm SL, but less than 25 mm SL, were confined to a net cage in the upper corner of the focal aquarium, as they are known to participate in many of the relevant behaviors (29, 41). No juveniles > 25 mm SL were present in any tanks. In tanks without juveniles present, 6 hours prior to observation a dip net was swirled in tanks for one minute in groups to simulate the disturbance of capturing juvenile fish.

Some, but not all pairs had fry under 10 mm in length at the time of observations. These individuals remained free in the aquarium, but were noted and this information was included as a binomial factor in subsequent analysis (see below).

Statistical Analysis:

For each observation, we counted aggressive behaviors [slow approaches, fin raises, opercula spreading, fast approaches, head-down displays, and ramming] by each parental fish towards the intruder. Similarly, we counted each individual’s aggressive displays towards its mate. Counts of aggression towards intruders were square-root transformed to achieve normality which was tested using a Shapiro-Wilk test.

12

We tested whether resident fish of one sex more frequently performed defensive behaviors towards intruders using a general linear mixed model (LMM) using the MIXED procedure in SPSS (IBM SPSS 22.0). We used square-root transformed counts of aggression by resident fish towards intruders as the dependent variable in this model; resident sex, intruder type, the interaction of resident sex and intruder type, whether fry were present, and whether groups contained larger juveniles were all fixed effects. Individual and group identifiers were also included in this model as random effects.

We tested whether residents differed in their response to intruders depending on intruder type, depending on their own or their mate’s size, or their size relative to that of their mate, using a similar LMM. Again counts of aggressive behaviors towards intruders by each resident were the dependent variables. Intruder type, resident male standard length, resident female standard length, and their interactions were included as fixed effects in the full models, as was whether fry were present, and whether groups contained larger juveniles, as well as the interaction of intruder type with each of these. Pair identity was included in this model as a random effect. Final models presented for male and female aggression towards the intruders were those with the lowest AICc values, relative to the full model, after removals of interactions and effects.

To determine whether participation in defensive behavior towards intruders was related to rates of aggression between residents, we performed of the following Spearman’s rank-order correlations: To determine whether participation in defensive behavior towards intruders was related to their rate of aggression towards their mates, we tested for correlations between counts of aggressive acts by parents towards the intruder and counts of aggressive behavior towards their mate. We tested for correlations between rates of aggression by each resident towards intruders and rates of aggression received from that parents’ mate. We performed an additional Spearman’s rank-order correlation which tested for correlations between parental male and parental female

13 aggression towards intruders. For the Spearman’s rank-order correlations, we controlled for multiple comparisons using a false discovery rate (FDR) procedure (75, 76).

Ethical Note:

This experiment and all organisms used in it were approved under IACUC protocol 20080095-

R1. A length of PVC tubing was placed in each tank as a sanctuary for fish facing excessive aggression from their tank mates. While it did not occur during this experiment, if a fish remained in the PVC tube for several days or showed signs of physical damage (missing scales or frayed fins) we were prepared to isolate the aggressor for 24 h in a net cage (15 x 13 x 15 cm), which has been shown to temper aggressive interactions in this species (77). If aggression continued after reintroduction the pair would have been permanently dissolved and excluded from the study.

While our heterospecific and conspecific intruders were physically separated from intruders and thus faced no risk of physical injury, we assume the procedure was at least somewhat stressful.

As such, we used a particular intruder at most once per week in order to minimize any effect of possible stress.

Results:

Do parents of one sex perform more defensive behaviors?

Resident fish did not differ in their overall response to intruders depending on sex

(resident sex: F1,16.997=0.495, p=0.491), nor did resident fish differ in response to specific intruders differently depending on their sex (intruder type*resident sex: F2,54.054=0.81, p=0.922).

Resident male’s response to intruders was not influenced by the presence of fry or if the group had juveniles present prior to testing (Table 1a). There was a marginally non-significant pattern of resident females behaving less aggressively towards intruders in groups from which juveniles had been present prior to testing regardless of the type of intruder (Table 1b). 14

a. Male agg. towards intruder ΔAICc=0 Fixed Effect F df P Intercept 25.76 1,9 0.001 Male SL 1.905 1,9 0.201 Female SL 0.995 1,9 0.344 MSL*FSL 0.413 1,9 0.536 Intruder type 1.907 2,20 0.175 MSL*Intruder 1.161 2,20 0.333 FSL*Intruder 0.101 2,20 0.904 Fry present 0.052 1,9 0.825 Juveniles > 10 mm 0.718 1,9 0.419 Int. type*Fry pres. 0.164 2,20 0.850 Int. type*Juv. rem. 0.833 2,20 0.449

b . Female agg. towards intruder ΔAICc=0 Fixed Effect F df P Intercept 90.17 1,9 <0.001 Male SL 1.008 1,9 0.342 Female SL 0.568 1,9 0.470 MSL*FSL 7.153 1,9 0.025 Intruder type 0.887 2,20 0.428 MSL*Intruder 0.293 2,20 0.749 FSL*Intruder 1.516 2,20 0.244 Fry present 0.040 1,9 0.847 Juveniles > 10 mm 4.930 1,9 0.054 Int. type*Fry pres. 0.984 2,20 0.391 Int. type*Juv. 0.392 2,20 0.681

Table 1: Fixed effects in final LMMs with square-root transformed counts of aggressive acts towards intruders by parental males (a) and parental females (b) towards intruders. ΔAICc values calculated in contrast to full model (see “statistical analysis” in methods section). “Fry present” and “Juveniles removed” were binominal fixed effects describing whether fry (offspring <10 mm in length) were present, or whether juveniles (offspring >10 mm) had been temporarily isolated from the group, respectively. Fixed effects in bold were considered significant at p=0.05.

15

Do parents differ in their response to intruders depending on their size, their mates’ size, or their relative size?

Resident males did not vary in counts of aggression towards intruders depending on their own size, their mates’ size, or the interaction of their sizes (Table 1a). Resident female aggression towards intruders was predicted by their own size, their mate’s size, and the interaction of the two

(Table 1b, Figure 1) such that regardless of the intruder type, resident females paired with small males responded equally aggressively towards intruders regardless of their own size; small resident females paired with large mates behaved most aggressively towards intruders, and large resident females with large mates behaved the least aggressively towards intruders. Resident males and females did not differ significantly in the counts of aggressive acts towards A. compressiceps and either type of conspecific intruder (Table 1).

Figure 1: Plot of solution of LMM with square-root transformed resident female aggression towards intruders as the dependent variable. Omitted portions of the plot represent regions which were not represented by the pairs in our study.

16

Figure 2: Plot of counts of aggressive acts by the parental male towards the intruder versus counts of aggression by the parental female towards the parental male. Parental females were more aggressive towards parental males that performed fewer aggressive acts towards intruders.

Figure 3: Plot of counts of aggressive acts by the parental female towards the intruder versus counts of aggression by the parental female towards the parental male. Parental males were more aggressive towards parental females that performed fewer aggressive acts towards the intruders.

17

Do rates of aggression towards intruders correlate with rates of aggression towards mates?

Rate of aggression by resident males towards intruders was negatively correlated with rate of aggression by the resident female towards the resident male (Spearman’s rank-order correlation: -0.505, p<0.001; Figure 2). Rate of aggression by resident males toward intruders was not significantly correlated with rate of aggression by the resident male towards the resident female after accounting for multiple comparisons (Spearman’s rank-order correlation: -0.320, p=0.032). Rates of aggression by the resident female towards intruders were negatively correlated with rates of aggression by the resident male towards the resident female (Spearman’s rank-order correlation -0.651, p<0.001; Figure 3), but not with rates of aggression by the resident female towards the resident male after controlling for multiple comparisons (Spearman’s rank-order correlation: -0.336, p=0.024). Rates of aggression towards intruders by the resident male and female were not significantly correlated with one another (Spearman’s rank-order correlation:

0.282, p=0.61), nor was there a significant correlation between rates of aggression by each resident towards its mate after controlling for multiple comparisons (Spearman’s rank-order correlation: 0.359, p=0.015).

Discussion:

Investment in parental care is assumed to be positively related to offspring survival (39,

78). In our study, individuals of both sexes which participated less in territory defense faced higher rates of aggression from their mates. This suggests that mates that do not participate in defense receive punishment from partners (e.g., 53). We did not find evidence that coercion allowed one mate to reduce its own investment in defense by increasing its partner’s investment, as there was not a significant correlation between partners’ aggression toward intruders. It may be that punishment, or the threat of punishment, enforces cooperation in joint territorial defense (79), although punishment may serve other functions than motivating ‘lazy’ partners, (e.g., 80). 18

Male and female N. pulcher differed in whether the characteristics of mates influenced aggressive responses to territory intrusions. Variation in defense by resident male N. pulcher towards intruders was not explained by any effects in our models. However, females did respond to the size of their mate. Female resident size did not influence rates of defense towards intruders when they were paired with small males, but it did when paired with large mates. Resident females that were much smaller than their mates (small females paired with large males) performed more defensive behaviors towards intruders, which is consistent with our predictions under punishment or coercion. Large females paired with large males performed the fewest defensive behaviors. Larger females may be better able to avoid or withstand punishment from males for performing less defensive behavior.

Interestingly, resident females in groups with juveniles were also marginally less aggressive towards intruders of all types. One interpretation of this pattern is that resident females in groups with juveniles expected them to participate in defense. In this species, dominant fish in large groups do have lower workloads (64), however no such pattern was observed in our resident male fish. Collectively, resident females’ defensive behavior towards all intruders varied in response to characteristics of their social setting, while that of resident males did not. In N. pulcher, dispersal is male-biased (81, 82) and males may have other reproductive opportunities outside the group (83, 84) and thus females may have more at stake in current groups.

Resident N. pulcher in our study did not differ in their response to heterospecific, conspecific male and conspecific female intruders. While N. pulcher do occasionally eat eggs laid by conspecifics (85, 86), subordinate group members often migrate from other groups as

“helpers,” and unfamiliar conspecifics may be perceived as joiners to the group (26, 87, 88). If potentially joining conspecifics are perceived as reproductive competitors (28, 89, 90), residents’ defensive response to conspecific intruders may be influenced by the costs and benefits that they

19 may impose on adult group members as reproductive competitors, mates, and potential providers of alloparental care. Opposite-sex dominants may benefit from additional future fitness gains through accepting these potential joiners, while same-sex dominants may face additional costs of reproductive competition and lost alloparental care. This conflict may lead to aggression by dominants towards same-sex intruders, and affiliative or even defensive behaviors by dominants on behalf of opposite-sex intruders. In taxa as diverse as false clown anemonefish (Amphiprion ocellaris) and spotted hyena (Crocuta crocuta), group members have been found to be most aggressive towards individuals of their same sex or status (91, 92). In this species, Desjardins et al. (29) found that socially dominant males in wild groups were more aggressive towards dominant-sized male intruders than dominant-sized female intruders. In their study, dominant females responded equally aggressively to male and female dominant-sized intruders (29).

Similarly, Mitchell et al. (27) found that dominant males behaved more aggressively towards subordinate male group members during non-reproductive periods. It may be that parental fish in our study did not perceive the relatively small unfamiliar conspecifics in our study as a threat to their status or reproductive success, however if this is the case it is unclear why these conspecifics still faced such high rates of aggression. The jar used in our study was not porous as containers were in Desjardins et al. (29) and Ligocki et al. (see Chapter 5); if chemical cues are important for sex determination, our fish may not have been able to determine the sex of the conspecific.

Failure to detect chemical cues may also explain why resident fish were not highly defensive against the heterospecific fry predator, particularly when fry were present in the territory at the time of presentation.

Reproductive males and females in cooperatively breeding groups have a shared interest in defending their territory and offspring. While our results suggest that both sexes prefer their mates to perform defensive behaviors, there was no evidence that residents could coerce their

20 mates to increase defense, thereby allowing themselves to perform less defense. Rather, we suggest that the threat of punishment may be used to enforce participation in defense. Our study shows that females’ defense against intruders was influenced by several aspects of their social environment, while males’ defense was not. In many biparental species, including N. pulcher, parents of a particular sex often predominantly or exclusively perform particular parental tasks

(29, 39, 45). We suggest that apparent sex-specific or sex-biased behavior may in fact result from interactions between parents. In systems with limited inter-pair variation in the relative power of parents, the emergent behavioral patterns may be mistaken for sex-specific behaviors. Further, our results indicate that individual behavior can be an emergent effect of the social environment in which the individual lives.

21

Chapter 3: Third party effects in a cooperatively breeding cichlid: effects of subordinate sex

and relative size

Abstract: In cooperatively breeding groups, socially subordinate individuals limit their own reproduction and provide alloparental care to the offspring of dominant group members. If some subordinate reproduction occurs, dominants of the same sex as a reproductive subordinate face a trade-off between the benefits of gaining alloparental care and the costs of reproductive competition. Dominants of the other sex do not experience such costs, leading to conflict between male and female dominants over the presence and behavior of subordinates. In two experiments, we examined how dominants’ interactions with subordinates, as well as interactions between dominants, influenced and were influenced by the relative size of group members in the cooperatively breeding cichlid fish, Neolamprologus pulcher. Size is a reliable determinant of winning in aggressive conflict in this species, so we predicted that as dominant males increase in size relative to dominant females, patterns of aggression between dominants as well as patterns of agonism between dominants and subordinates would be more likely to reflect the interests of dominant males. As predicted, subordinate males faced greater aggression from dominant males when dominant males were much larger than dominant females suggesting dominants are in conflict regarding the role of subordinate males in groups. Similarly, subordinate females faced greater aggression when dominant males and females were similar in size. Conflict between dominants was influenced by relative size in groups containing subordinate males, but not when groups contained subordinate females. This study provides evidence that conflict within groups

22 reflects the characteristics of individuals in conflict, but also results from characteristics of other group members.

Introduction:

For group-living animals, grouping is presumed to provide some advantage over solitary living for some or all group members. Potential advantages could be rooted in a reduced risk of predation (3, 93-95), increased foraging efficiency (6, 96, 97), or for certain social systems, alloparental care (23-25). However, individual interests may not necessarily align. For example, in groups with queues for dominance, lower ranking individuals may face costs from joiners to the group who would challenge or usurp them in the hierarchy, while higher ranking individuals do not (92). In groups in which members compete for mating opportunities, individuals may not agree on the division of reproduction in the group; more dominant individuals may be better able to monopolize breeding opportunities resulting in high reproductive skew (85, 98, 99).

Cooperatively breeding systems provide an opportunity to examine the impact of both conflicting and common interests in animal groups. In such systems, groups may consist of individuals of varying relatedness that provide alloparental care to the offspring of other group members (100, 101). Cooperative breeding is also characterized by high reproductive skew, although skew varies within (98, 102) and between systems (103, 104). Where reproductive skew is incomplete, the presence and role of subordinate group members may lead to different costs and benefits for dominant males and females (28, 77, 105). For both dominants, the presence of a

“helpful” subordinate may lead to fitness benefits from increased offspring survival and lower parental workload (106, 107). For the opposite-sex dominant, this subordinate is also a potential mate, as reproductive subordinates frequently mate with dominants (108). However, for the same- sex dominant, this subordinate is a potential reproductive competitor. This may result in conflict

23 between dominants regarding the role and presence of subordinate group members. Hamilton and

Ligocki (77) found that aggressive conflict between the dominant male and female in the cooperatively breeding cichlid Neolamprologus pulcher was negatively correlated with the helpfulness of subordinate males. The outcome of conflict between the dominant breeding pair may feed back to influence the likelihood a subordinate will be permitted to remain in or reproduce in the group (105).

Where conflict between group members exists, some group members may be better able to influence the size and characteristics of the group (i.e., they have more “power,” 20, 21). For instance, in meerkats (Suricata suricatta), pregnant dominant females are able to temporarily evict subordinate females from groups to prevent them from also producing offspring which would compete with their own (109). If group members’ interests are in conflict, we expect the characteristics of the group will better reflect the interests of these powerful individuals. In the case of emergent sexual conflict among dominants arising from the presence and behavior of subordinates, the relative power of dominants may influence not only their direct interactions, but also direct and indirect interactions with subordinates. For example, in a group with a powerful dominant male, subordinate males that avoid interactions with the dominant female may also avoid aggression from the dominant male.

The cooperatively breeding cichlid fish Neolamprologus pulcher has become a model system for investigating questions related to conflict, cooperation, and sociality in vertebrate animal groups in part due to its relative ease of care and the ability to manipulate groups in the lab (26). In natural N. pulcher groups, dominant males are consistently the largest group member, and typically the dominant female is the second largest group member (26). Both male and female subordinate N. pulcher reproduce in groups (102, 108), and the risk of subordinate reproduction should impose different costs on same-sex and opposite-sex dominants. Indeed,

24

Mitchell et al. (27) found that dominants behaved more aggressively towards subordinates of the same sex, and dominant males behaved more affiliatively towards subordinate females than subordinate males. They suggested dominant males may be seeking extra pair reproduction or preventing subordinate male cuckholdery (27).

A substantial body of literature exists describing the relationship between size, dominance, and conflict in N. pulcher (25, 28, 61, 62, 66, 110; for a review see 26), as well as how subordinates respond to conflict with more-dominant individuals (15, 95, 111). In this and other cichlid species, the relative size difference between parents has substantial influence on parental investment and roles (49, 112, see Chapter 2). For N. pulcher of the same sex, both the occurrence and the outcome of conflict are closely related to relative size (15, 62, 66), so that relative size is an important determinant of power within groups. In pairwise contests, opponent size significantly influenced the giving up point of contestants, and when contestants’ body size was within 5% of each other the conflict involved more physical aggression and a winner could not be predicted a priori (62). Hamilton et al. (15) found that large subordinate males were more submissive and less affiliative towards small dominant males, and these subordinate males generally stayed farther away from the breeding shelter. Additionally, Heg and Hamilton (85) found that female subordinates close in size to dominant females are more likely to reproduce in groups.

To test whether relative power within breeding pairs influences social interactions between the dominant pair and between dominant and potentially reproductive subordinates, we used groups composed of a dominant breeding pair and an unrelated male (Experiment 1) or female (Experiment 2) subordinate. We hypothesize that with increasing dominant pair size ratio

(dominant male size: dominant female size), males will be better able to influence within-group interactions, and females less able to do so. Thus, the outcome of sexual conflict between the

25 dominant pair will be more likely to favor the interests of the male when dominant pair size ratio is large. Therefore, we predict that as the breeding pair size ratio increases, subordinate males will have more frequent agonistic interactions (aggressive and/or submissive) with dominant males and less-frequent affiliative interactions with dominant females. Similarly, we predict subordinate females will have more affiliative interactions with dominant males and fewer agonistic interactions with dominant females when breeding pair size ratio is large.

Methods:

Study organisms and housing conditions

All fish in Experiment 1 and 2 were either wild caught or F1 offspring of wild fish collected from the Kipili region of Lake Tanganyika (Tanzania). Fish were housed in 113.6 L aquaria (30.5 cm x 61 cm x 61 cm). Aquarium conditions were maintained to reflect those of

Lake Tanganyika (Mean Temperature = 25.2° C, pH = 7.8-8.4) and were maintained on a

12h:12h light cycle. Fish were fed ad libitum a diet of cichlid flake food (Tetracichlid Cichlid

Flakes – Tetra Holding Inc.) 5x per week and a mix of frozen brine shrimp, Artemia spp. (San

Francisco Bay Brand, Inc.) and Daphnia (Hikari Sales U.S.A., Inc.) 2x per week.

Experiment 1: Subordinate Males

Eight groups consisting of a dominant pair and a smaller subordinate male were introduced into aquaria. All group members were unrelated to one another. Prior to introduction, all three fish were measured for standard length (hereafter SL) and mass, and given an identifying dorsal fin clip. In order to reduce aggression amongst group members, all fish were initially housed in individual net cages. The subordinate male was introduced first, followed by the dominant female the following day, and the dominant male several hours after the dominant

26 female. This method has been found effective at reducing the risk of the subordinate fish being immediately evicted (95).

The day the dominant fish were released was considered day 1 of a 30 day experimental period. During this period, we took two-15 minute behavioral observations of each group. At the end of this period, all fish were again measured, dorsal fin clips were renewed, and individuals were moved into new tanks containing unfamiliar individuals to form new groups. Introductions into new groups occurred as before. At the end of a second 30 day experimental period, fish were again measured, and fin clips renewed. Fish were again moved into new unfamiliar groups and the entire procedure was repeated for a third time.

Experiment 2: Subordinate Females

In a similar fashion to Experiment 1, we formed unrelated groups consisting of a dominant pair and a subordinate female. Fish were released from net cages as in Experiment 1, with the subordinate female first, followed by the dominant female, then the dominant male.

Groups were left together for 30 days, during which time three-15 minute behavioral observations took place of each group. After this 30 day period, fish were again measured and dorsal fin clips renewed. Fish were moved into new groups for a second 30 day period during which another three-15 minute observations took place.

Differences between the experimental protocols emerged from what we learned during

Experiment 1. It became clear that including a third behavioral observation during each experimental period was both feasible and beneficial to the study. Additionally, we determined that two rather than three experimental periods were sufficient to ensure subordinate fish were members of groups containing dominant pairs varying in relative size.

27

Behavioral Observations:

All behavioral observations took place between 1000 and 1500 h and were documented using the program JWatcher (74). Behaviors were classified based on the behavioral descriptions of Coeckelberghs (71), Kalas (72), and Taborsky (73). Some behaviors either did not occur or only occurred rarely in a small subset of groups (<25%) and were thus omitted from subsequent statistical analysis.

Statistical Analysis:

Counts of all behaviors within each dyad in groups from both experiments were square root transformed to achieve normality; after transformation all behaviors that were normally distributed were summarized using Principal Components Analysis with a Varimax rotation and

Kaiser Normalization. Loadings greater than 0.5 were considered particularly informative for interpretation of PC factors (Hamilton et al. 2005). For each dyad, the first two components

(those that explained the greatest proportion of overall variation) were used in separate general linear mixed models (GLMM) using the MIXED procedure in SPSS (SPSS version 22.0). Scores for each of these components were dependent variables in the models, while the standard lengths of each of the three group members (SL±SD: Experiment 1: DM=59.8±3.4 mm, DF=55.5±2.4 mm, SM=39.8±3.3 mm; Experiment 2: DM=62.4±5.0 mm, DF=55.1±1.4 mm, SF=38.3±2.7 mm), the interaction of the dominants’ standard length, and the interaction of subordinate standard length and that of the same-sex dominant were included as covariates. Each of the interactions was removed from the model if 1) the model was unable to converge with it included, or 2) if the exclusion of an interaction reduced the AICc value of the model by > 2. The identities of each of the three group members were included as random effects in each model.

28

Although observations in both studies were performed by the same observer, and aquarium conditions were maintained in the same manner, these experiments occurred at different times in association with other studies (Experiment 1 see (77); Experiment 2 see (113), Chapter

4). Because of this, we do not make direct statistical comparisons between groups containing subordinate males and those containing subordinate females aside from an independent samples t- test intended to confirm that individuals were similarly sized in each experiment.

Results:

Comparison of study fish in Experiment 1 and 2:

Dominant males, dominant females, and subordinates did not differ significantly in size between experiments (Dominant males: t24.098=-1.831, p=0.08; Dominant females: t37.305=0.52, p=0.606; Subordinate fish: t36.305=1.535, p=0.133).

i. Dominant Male - Dominant Female ii. Dominant Male - Subordinate Male iii. Dominant Fem. - Subordinate Male PC 1 PC 2 PC 1 PC 2 PC 1 PC 2 DF → DM Bump 0.553 0.182 DM → SM Slow app. 0.699 -0.441 DF → SM Slow app. 0.917 -0.056 Slow app. -0.531 -0.018 Join -0.190 0.429 Join -0.307 0.437 Join 0.315 0.408 Fast app. 0.451 0.061 Fast app. 0.034 0.545 Oper. sprd. 0.052 0.787 Fin raise 0.303 0.406 Ram -0.138 0.393 Ram -0.154 0.138 SM → DM Avoid 0.781 -0.445 Fin raise 0.842 -0.115 Fin raise 0.156 0.643 Bump 0.486 0.700 SM → DF Bump 0.040 0.777 Tail quiv. 0.908 -0.181 Join -0.185 0.618 Join 0.208 0.734 DM → DF Slow app. 0.831 -0.287 Fast app. 0.685 -0.132 Hook 0.911 -0.062 Join -0.004 -0.619 Fin raise 0.845 0.205 Fin raise 0.867 0.017 Oper. sprd. 0.413 0.336 Tail quiv. -0.037 0.726 Tail quiv. 0.687 0.477 Fast app. 0.295 -0.578 Ram 0.926 0.030 Fin raise -0.125 0.423

Table 2: Loadings of behaviors in Principal Component Factors 1 and 2 for each dyadic interaction in Experiment 1. Behaviors with loadings greater than 0.5 were considered particularly informative in our interpretation of PC Factors.

29

Dominant male - dominant female E1 DMDF Factor 1 ΔAICc=-12.142 Fixed effect F d.f. Sig. DM sl 2.684 1,14.382 0.123 DF sl 13.508 1,19.597 0.002 SM sl 0.441 1,18.696 0.515

E1 DMDF Factor 2 ΔAICc=-11.067 Fixed effect F d.f. Sig. DM sl 0.626 1,16.495 0.44 DF sl 0.009 1,19.262 0.925 SM sl 1.646 1,19.971 0.214

Dominant male - subordinate male E1 DMSM Factor 1 ΔAICc=-5.561 Fixed effect F d.f. Sig. DM sl 5.185 1,18.653 0.035 DF sl 4.798 1,18.787 0.041 SM sl 2.768 1,18.989 0.113 DM sl * DF sl 5.089 1,18.735 0.036

E1 DMSM Factor 2 ΔAICc=-10.665 Fixed effect F d.f. Sig. DM sl 0.217 1,6.516 0.657 DF sl 0.098 1,18.743 0.758 SM sl 1.335 1,16.494 0.264

Dominant female - subordinate male E1 DFSM Factor 1 ΔAICc=0 Fixed effect F d.f. Sig. DM sl 9.233 1,5.501 0.025 DF sl 5.225 1,8.150 0.051 SM sl 55.421 1,5.137 0.001

E1 DFSM Factor 2 ΔAICc=-10.846 Fixed effect F d.f. Sig. DM sl 0.269 1,6.948 0.62 DF sl 3.029 1,7.324 0.123 SM sl 0.013 1,14.848 0.91

Table 3: Output of GLMMs with PC scores as dependent variables. Individual standard lengths, the interaction of dominant standard lengths, and the interactions of the dominant and subordinate male’s standard length were included as covariates. ΔAICc scores were calculated in relation to the most complete model that achieved convergence.

30

Experiment 1 - Groups with Subordinate males: a. Dominant male-dominant female interactions

The first two principal components described, respectively, 26.03% and 18.29% of variation in dominant male-dominant female interactions after rotation. The first of these (E1 DMDF Factor

1, Table 2) had high positive loadings on slow approaches and rams by the dominant male towards the dominant female, dominant female tail quivers, and dominant female “bumping” of dominant males, and high negative loadings on dominant female slow approaches towards the dominant male. Slow approaches and rams are aggressive behaviors, tail quivering is a submissive signal, and bumping may be a submissive or affiliative behavior, although its function is somewhat unclear (15). We interpreted this factor as a measure of dominant male aggression and dominant female submission. The second factor (E1 DMDF Factor 2, Table 2) had high positive loadings on opercular spreading and fin raises (aggressive displays) by the dominant female towards the dominant male and high negative loadings on joining (affiliative) or fast approach (aggressive display) by the dominant male towards the dominant female. We interpreted this factor as a measure of agonism within the dominant pair.

In our GLMM, E1 DMDF Factor 1 was significantly predicted by dominant female standard length (Table 3, Figure 4). In other words, larger dominant females faced less aggression from dominant males, and less frequently “bumped” or performed tail quivers towards dominant males. We did not find a significant effect of any of the standard lengths or their interactions on E1 DMDF Factor 2 (Table 3).

31

Figure 4: Residual of E1 DMDF Factor 1 versus Dominant Female SL (mm). This PC factor was heavily loaded on aggressive behaviors by the dominant male towards the dominant female, and submissive behaviors by the dominant female towards the dominant male.

b. Dominant male-subordinate male interactions

The first two PC factors associated with dominant male-subordinate male interactions

explained 28.86% and 22.02% of variation, respectively. The first of these factors (E1 DMSM

Factor 1, Table 2) had high positive loadings on slow approaches by the dominant male towards

the subordinate male and false charges and fin raises (aggressive displays) as well as avoidance

behavior by the subordinate male towards the dominant male. We interpreted this factor as a

measure of mutual restrained aggression and subordinate male avoidance behavior. The second

PC factor (E1 DMSM Factor 2, Table 2) had high positive loadings on joining and tail quivering

(submissive) behaviors by the subordinate male towards the dominant male, we interpreted this

factor as a measure of submission and affiliativeness by the subordinate male towards the

dominant male.

32

Figure 5: Best fit surface of E1 DMSM Factor 1 versus Dominant male and dominant female SL (mm). This PC factor was heavily loaded on mutual aggression between the dominant and subordinate male, as well as the subordinate male avoiding the dominant male.

In our GLMMs, the first of these factors (E1 DMSM Factor 1) was significantly predicted by dominant male and dominant female standard length, as well as their interaction

(Table 3). Dominant and subordinate males behaved most agonistically when large dominant males were in groups with relatively small dominant females. Dominant males and subordinate males behaved less agonistically in groups in which the dominants were closer in size (either because dominant females were larger or because dominant males were smaller; Table 3, Figure

5). This factor was not significantly influenced by subordinate male size. We did not find a significant effect of any of the standard lengths or their interactions on E1 DMSM Factor 2 (Table

3).

33

Figure 6: Residual of E1 DFSM Factor 1 versus dominant male (a) and subordinate male (b) standard length. This PC factor was heavily loaded on mutual restrained aggression and subordinate male submissive behavior.

34 c. Dominant female – subordinate male interactions

The two largest PC factors describing interactions between dominant female and subordinate males explained 37.64% and 20.33% of the variation in behavior, respectively. The first PC factor (E1 DFSM Factor 1, Table 2) had high positive loadings on slow approaches and fin raises by the dominant female towards the subordinate male, as well as fin raises, “hook” displays and tail quivering by the subordinate male towards the dominant female. Hook displays are considered a submissive display. We interpreted this factor as a measure of mutual restrained aggression and subordinate male submissive displays towards the dominant female. The second

PC factor (E1 DFSM Factor 2) had high positive loadings on false charges by the dominant female towards the subordinate male; and bumping and joining by the subordinate male towards the dominant female. We interpreted this factor as a measure of dominant female aggression and subordinate male affiliative and submissive behavior. E1 DFSM Factor 1 was significantly predicted by dominant male standard length and subordinate male standard length such that subordinate males and dominant females performed fewer restrained aggressive behaviors when dominant males or subordinate males were relatively large (Table 3, Figure 6). We did not find any significant effect of individual standard length or their interactions on E1 DFSM Factor 2

(Table 3).

35

i. Dominant Male - Dominant Female ii. Dominant Male - Subordinate Female iii. Dominant Fem. - Subordinate Fem. PC 1 PC 2 PC 1 PC 2 PC 1 PC 2 DF → DM Slow app. -0.385 0.673 DM → SF Slow app. 0.795 0.074 DF → SF Slow app. 0.765 -0.150 Fast app. -0.501 -0.399 Fast app. 0.449 0.570 Fast app. 0.262 -0.726 Ram 0.335 -0.012 Ram 0.068 0.850 Ram 0.036 -0.878 Fin raise 0.221 0.878 Fin raise 0.319 -0.734 Fin raise 0.828 0.101 Oper. sprd. 0.835 0.215 Join 0.361 0.051 Join -0.206 0.016 Avoid -0.040 -0.183 SF → DM Bump 0.564 0.544 SF → DF Fin raise 0.861 0.030 Bump -0.314 0.411 Tail quiv. 0.732 0.019 Bump 0.082 0.519 Tail quiv. 0.769 0.191 Zig zag 0.268 0.759 Tail quiv. 0.837 0.123 Join -0.710 0.142 Join 0.776 0.094 Zig zag 0.531 0.360 DM → DF Slow app. 0.737 0.299 Join 0.257 0.706 Fast app. 0.699 -0.294 Ram 0.843 0.171 Soft touch -0.201 -0.599 Fin raise 0.214 0.758 Oper. sprd. 0.652 -0.068 Join 0.129 -0.507

Table 4: Loadings of behaviors in Principal Component Factors 1 and 2 for each dyadic interaction in Experiment 2. Behaviors with loadings greater than 0.5 were considered particularly informative in our interpretation of PC Factors.

Experiment 2 - Groups with Subordinate females: d. Dominant male-dominant female interactions

The two largest PC factors describing variation in interactions between dominant males and dominant females explained 29.53% and 19.31% of the variation between groups, respectively. The first of these PC factors (E2 DMDF Factor 1, Table 4) had high positive loadings on slow approaches, false charges, and rams by the dominant male towards the dominant female, as well as frequent opercular spreading and tail quivering, and high negative loadings on

36

Dominant male - dominant female E2 DMDF Factor 1 ΔAICc=0 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 1.607 1,1.833 0.342 0.045 0.034 DF sl 0.639 1,6.491 0.452 -0.102 0.176 SF sl 2.893 1,2.768 0.195 -0.121 0.048

E2 DMDF Factor 2 ΔAICc=0 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 0.16 1,3.264 0.714 0.011 0.027 DF sl 4.055 1,5.276 0.097 -0.223 0.111 SF sl 0.187 1,7.954 0.677 0.04 0.091

Dominant male - subordinate female E2 DMSF Factor 1 ΔAICc=-5.159 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 0.263 1,12 0.617 -0.025 0.049 DF sl 3.547 1,12 0.084 -0.351 0.186 SF sl 2.89 1,12 0.115 -0.158 0.093

E2 DMSF Factor 2 ΔAICc=-4.233 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 0.542 1,9.501 0.479 0.039 0.053 DF sl 1.281 1,7.268 0.294 -0.238 0.21 SF sl 1.23 1,9.908 0.294 -0.112 0.101

Dominant female - subordinate female E2 DFSF Factor 1 ΔAICc=-3.432 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 2.035 1,5.064 0.212 -0.04 0.028 DF sl 2.076 1,5.638 0.203 0.234 0.162 SF sl 11.408 1,3.795 0.03 0.234 0.069

E2 DFSF Factor 2 ΔAICc=0 Fixed effect F d.f. Sig. Est. Fixed S.E. DM sl 15.192 1,5.336 0.01 -3.69 0.947 DF sl 17.63 1,5.351 0.007 -4.26 1.02 SF sl 4.097 1,9.903 0.071 0.198 0.098 DM sl * DF sl 15.533 1,5.341 0.01 0.067 0.017

Table 5: Output of GLMMs with PC scores as dependent variables. Individual standard lengths, the interaction of dominants’ standard lengths, and the interactions of the dominant and subordinate females’ standard length. ΔAICc scores were calculated in relation to the most complete model that achieved convergence.

37 false charges and joining by the dominant female towards the dominant male. We interpreted this factor to as a measure of the degree of agonism within the dominant pair. The second PC factor

(E2 DMDF Factor 2, Table 4) had high positive loadings on fin raises, and high negative loadings on joining and soft touches (courtship) by the dominant male towards the dominant female, as well as high positive loadings on fin raises and slow approaches by the dominant female towards the dominant male. We interpreted this factor as a measure of mutual agonism and a lack of courtship and joining behavior by the dominant male towards the dominant female. We did not find a significant effect of individual standard length or their interactions on either of these factors (Table 5).

e. Dominant male – subordinate female interactions

The two largest PC factors describing the interactions of the dominant male and subordinate female explained 28.86% and 27.51% of the variation between groups, respectively.

The first of these PC factors (E2 DMSF Factor 1, Table 4) had high positive loadings on slow approaches by the dominant male towards the subordinate female, as well as frequent bumping, tail quivering and joining by the subordinate female towards the dominant male. We interpreted this factor as a measure of restrained aggression by the dominant male and submissive and affiliative behaviors by the subordinate female. The second PC factor (E2 DMSF Factor 2, Table

4) had heavy positive loadings on false charges and rams, and high negative loadings on fin raises by the dominant male towards the subordinate female as well as high positive loadings on bumping and “zig zag” behaviors by the subordinate female towards the dominant male. We interpreted this factor as a measure of aggression by the dominant male and submissive behaviors by the subordinate female. Neither of these factors was significantly affected by variation in individual standard length or their interactions (Table 5).

38

Figure 7: PC scores for E2 DFSF Factor 1 versus subordinate female standard length. This PC factor was heavily loaded on dominant female aggression towards subordinate females and subordinate female submissive and affiliative behavior towards dominant females.

Figure 8: Best fit surface of E2 DFSF Factor 2 versus dominant male and dominant female standard length. This PC factor was heavily loaded on subordinate female affiliative behavior towards the dominant female and a lack of aggression by the dominant female towards the subordinate female. 39 f. Dominant female – subordinate female interactions

The two largest PC factors describing behavioral patterns between the dominant and subordinate females explained 31.82% and 22.43% of variation between groups, respectively.

The first of these PC factors (E2 DFSF Factor 1, Table 4) had high positive loadings on slow approaches and fin raises by the dominant female towards the subordinate female as well as fin raises, tail quivers, and zig-zagging by the subordinate female towards the dominant female. We interpreted this factor as a measure of mutual restrained aggression and subordinate female submissive behavior. The second PC factor (E2 DFSF Factor 2, Table 4) had high negative loadings on false charges and rams by the dominant female towards the subordinate female, as well as high positive loadings on bumping and joining behaviors by the subordinate female towards the dominant female. We interpreted this factor as a measure of lack of aggression by the dominant female and frequent affiliative and submissive behavior by the subordinate female. E2

DFSF Factor 1 was significantly predicted by subordinate female standard length such that larger subordinate females faced greater aggression from dominant females, and behaved more submissively towards them (Table 5, Figure 7). E2 DFSF Factor 2 was significantly predicted by dominant male and dominant female standard length, as well as their interaction (Table 5).

Subordinate females faced less aggression from and were more affiliative with large dominant females when these dominant females were much smaller than their mates. Subordinate females faced less aggression from and were more affiliative with small dominant females when those dominant females were paired with relatively small dominant males (Table 5, Figure 8).

Subordinate female standard length did not have a significant effect on this factor.

40

Discussion:

In groups containing subordinate males (Experiment 1), larger dominant females faced less aggression from and were less submissive towards dominant males, supporting the assumption that size differences within the dominant breeding pair indeed influence conflict between these individuals. We detected no such pattern in groups containing subordinate females

(Experiment 2). We suggest that male and female subordinates impose different costs and benefits on dominants depending on the sex of the subordinate, and that this asymmetry in the net benefits for dominants of subordinate presence is greater in the case of subordinate males.

Subordinate males that provide alloparental care provide benefits for both dominants, but impose clear costs on dominant males if reproductively active.

Subordinate male reproduction is common in groups (27, 28, 108), and represents an immediate reduction in the reproductive output of the dominant male. In Experiment 1, dominant and subordinate males behaved most agonistically when dominant males were much larger than dominant females. When dominants were similar in size however, subordinate males faced less aggression from dominant males. In Experiment 2, dominant females also varied in their rates of aggression with subordinate females, but in contrast to the conflict between dominant and subordinate males in Experiment 1, dominant female aggression towards subordinate females was negatively associated with submissive (and also affiliative) behavior. In this case, subordinate females either performed frequent submissive and affiliative behaviors towards dominant females, or faced higher rates of aggression from them. Large dominant females paired with relatively small dominant males (i.e. groups in which dominants were similar in size) behaved aggressively towards subordinate females. In contrast, large dominant females paired with relatively large males and small dominant females paired with relatively small dominant males

(i.e. groups in which dominant males were larger than dominant females) were less aggressive

41 towards subordinates and received more submissive and affiliative displays from subordinate females. It is unclear why small dominant females paired with large males were more aggressive towards subordinate females. Larger dominant males are more likely to be polygynous (84), and thus small dominant females paired with large dominant males might perceive subordinate females as a higher risk in the group. Taken together, we suggest that dominants are in conflict regarding the role of and presence of subordinates; same-sex dominants face potential costs from subordinates as reproductive competitors, while opposite-sex dominants may acquire reproductive benefits from these subordinates.

Interactions between dominants and subordinates were also directly influenced by the size of participants. Dominant male - subordinate male interactions were less agonistic when dominant males were large, and dominant female - subordinate male interactions were more agonistic when subordinate males were large. In groups containing subordinate females

(Experiment 2), relatively large subordinate females faced greater aggression from and were more submissive towards dominant males. In teleost fish, female size and fecundity are positively associated with one another (114, 115), so larger subordinate females capable of producing larger clutches might benefit more than a small subordinate female from attempting to reproduce. While larger male subordinates may not be more fecund, they may have a better chance of successfully accessing reproductive females. Dominant females only benefit from subordinate males participating in reproduction if it increases their fitness. This could occur through increased alloparental care or genetic benefits associated with the additional mate (116, 117), but net gains may be lost if dominant males destroy clutches they didn’t fertilize or evict subordinate males who attempt to reproduce (118).

Agonistic interactions between subordinates and dominants of either sex may enforce alloparental care in this species (73, 119). Also, agonistic interactions between subordinates and

42 same sex dominants may suppress subordinate reproduction, either through direct control of subordinate reproduction or by pre-emptively dissuading subordinates from doing so (restraint models – for a review see 99), and larger, more powerful dominants are presumably better-able to do so. Young et al. (109) found that pregnant dominant female meerkats (Suricata suricatta) behaved aggressively towards some subordinate female group members, eventually driving them from the group. This temporary eviction was associated with elevated glucocorticoid levels which rendered these subordinate females infertile and thus suppressed these individual’s reproductive potential. A similar mechanism has also been suggested in Alpine Marmots, Marmota marmota

(120). Previous studies found no relationship between dominant aggression and subordinate cortisol levels in N. pulcher (113, 121, see Chapter 4), so reproductive suppression through elevated stress is not likely in this case.

Individuals living in cooperatively breeding groups face different costs and benefits associated with other group members as well as their behavior. The composition of the group itself may influence how group members assess these costs and benefits, and in turn how they interact within groups. For instance, while dominant males may acquire fitness gains if additional females reproduce in the group, if reproducing subordinate females cease alloparental care or their presence leads to escalated conflict with dominant females, it may be in dominant males’ best interests to prevent subordinate female reproduction. These emergent properties are often overlooked in studies focused on dyadic interactions or when the characteristics of other group members are ignored, however they may have significant impacts on behavioral patterns in groups.

Through two experiments we found evidence that subordinate males and subordinate females influence conflict between dominant group members in different ways. These results support previous findings (28) that subordinates impose different costs and benefits on dominants

43 depending on their sex. Further, our findings show that the relative power of influence of dominants influences the outcome of their conflict regarding subordinate group members. Recent work suggests the importance of third parties in understanding dyadic interactions (18, 19, 122), in applying this framework our study shows that third-party effects can also influence individual behavior and interactions thus shaping the emergent social landscape in animal groups.

44

Chapter 4: Variation in glucocorticoid levels in relation to direct and third-party

interactions in a social cichlid fish.

Abstract: Social status and within-group conflict are related to glucocorticoid levels in a number of social animal species. In complex animal societies, direct interactions between group members can influence the behavior and glucocorticoid levels of individuals involved. Recently, it has become apparent that third-party group members can influence dyadic interactions, and vice versa. Thus, glucocorticoid levels may vary depending on interactions of other members of the social group. Using the social cichlid fish Neolamprologus pulcher as a model system, we examined the relationship between levels of the glucocorticoid hormone cortisol in subordinate females and 1) direct interactions with dominant group members, as well as 2) dyadic interactions between the dominant male and female, which do not directly involve the subordinate female.

Subordinate females that frequently engaged in non-aggressive interactions with dominant females had lower cortisol levels. There was no relationship between subordinate female cortisol and agonistic interactions between the subordinate female and either dominant. Subordinate females had higher cortisol levels when in groups in which the dominant breeding pair behaved agonistically towards each other and performed fewer courtship behaviors. In the case of subordinate females, variation in cortisol levels is associated with their own affiliative behavior, but also can be explained by social context, particularly interactions between dominant members of the group.

45

Introduction:

The effect of sociality and dominance status on individual glucocorticoid levels has received a great deal of attention in recent years. Glucocorticoids (hereafter GC) mediate an individual’s physiological response to stressors that they perceive or encounter (123). In the short term, the release of GCs redirects energy to processes essential for escaping a stressor. Examples of such shifts include increasing the available glucose in circulation (124, 125) and upregulating circulatory system function while redirecting blood away from nonessential organs (123).

However, chronically elevated GC levels can impact organisms well past the conclusion of exposure to the stressor, and lead to reduced investment in growth and reproduction (123). Thus,

GC levels can provide information about the stress level experienced by an animal, as well as potential lasting costs associated with exposure to a stressor.

Social species may face additional stressors associated with group living. Individuals in groups generally live in closer proximity and share resources, which may lead to frequent and potentially stressful interactions with other group members. Repeated agonistic interactions over time can act as a stressor and elevate individual cortisol levels (109). While social living may impose “stressors” on individuals, such social systems also provide opportunities to participate in non-aggressive interactions (hereafter “affiliative” interactions) as well as submissive behaviors which may reduce or prevent agonistic interactions, and these behaviors may be inversely related to individual cortisol levels (121).

In addition to the effects of the frequency and nature of social interactions, organisms living in hierarchical groups also face potential stressors related to social status and stability.

Elevated GC levels have been suggested as a cause of subordinate reproductive suppression (109,

126, 127), as well as a cost of dominance (98, 128). However, across animal groups there is great variation in the relationship between GC levels and social status (129; see reviews 130, 131). In

46 some systems subordinate group members have higher basal GC levels than more dominant group members (such as house mice, Mus musculus - 132), whereas in others dominant group members have higher basal GC levels than more subordinate group members (such as Florida

Scrub Jays Aphelocoma coerulescens - 127). Yet other social vertebrates show no apparent differences in GC levels depending on social rank (such as grey wolves, Canis lupus - 133).

Abbott et al. (129) suggested that this wide variation in relative GC levels is related to the number of stressors individuals face, and the amount of social support available, both of which may vary among taxa and correlate with status in some systems. Similarly, Goymann & Wingfield (131) suggested that patterns of status-based differences in cortisol levels reflect status-based differences in “allostatic loads”, the summed burden on individuals due to physiological, environmental, and social factors, (134).

Social species exist in communication networks in which individuals can acquire information about other individuals through signals not necessarily intended for them (135). In

Siamese fighting fish, Betta splendens, and green swordtail fish, Xiphophorus helleri, individuals eavesdropped on dyadic conflicts and used acquired information to inform future decisions (136,

137). Contestants also adjust their aggressive actions depending on the presence and sex of eavesdroppers (138). Individuals within communication networks thus have access to a great deal of information about group dynamics and stability. The behavior of individuals in hierarchical groups influences and is influenced by both their own direct interactions with other group members and by indirect interactions among others in the group (18, 19, 77).

In spite of a certain degree of common interest for group members, group members may experience conflict with one another. In cooperatively breeding systems with incomplete reproductive skew, the presence of subordinate group members may lead to conflict between the dominant breeders. While subordinate group members may provide alloparental care, which

47 benefits both parents, subordinate attempts to reproduce within the group may impose a cost on the same-sex dominant, and a potential benefit to the opposite-sex dominant. Variation in subordinate behavior is causally linked to variation in conflict between dominant males and dominant females in a cooperatively breeding fish (77). The outcome of this conflict between dominants may feed back directly and indirectly on subordinate group members’ behavior and fitness, and these subordinates may respond to prevent or reduce negative effects. Thus, we expect that GC levels of subordinate individuals in cooperatively breeding systems will reflect the sum of stressors related to the frequency of direct interactions, as well as the indirect effects of behavioral interactions among other group members.

Neolamprologus pulcher is a cooperatively breeding cichlid endemic to Lake Tanganyika in East Africa (25). This species has become a model system for the study of cooperative and other social behavior (see review - 26). N. pulcher groups are characterized by a dominant breeding pair and a number of subordinates that form a size-based dominance hierarchy (25, 61,

66, 90). Reproductive skew is high, but monopolization of reproduction by dominant group members is not complete. Extra-pair reproduction has been documented in N. pulcher for both male (89) and female (90) subordinate group members, as well as by non-group members (108,

139). In N. pulcher, levels of the GC hormone cortisol are higher in dominant individuals than in subordinate individuals (140, 141), and are higher in subordinate group members than size- matched individuals maintained in non-breeding groups (142). Further, Bender et al. (142) found that subordinate male submissiveness to dominant males is inversely correlated with cortisol, testosterone, and 11-ketotestosterone levels demonstrating that subordinate GC levels are influenced not only by social status but also individual behavior in N. pulcher.

We examine the relationship between cortisol levels in subordinate female N. pulcher and the behavioral interactions among subordinate females, dominant females, and dominant males.

48

Specifically, we examined the relationship between subordinate female cortisol and affiliative and agonistic (aggression, and associated submissive and avoidance behaviors) interactions that directly involve subordinate females, as well as affiliative and agonistic interactions that occur between the dominant female and dominant male. We predict that affiliative behavior between the subordinate female and either dominant will be inversely correlated with subordinate female

GC levels. Affiliative behavior may indicate a lack of conflict within the dyad, or prevent future agonistic interactions. We predict that agonistic interactions between the subordinate female and the dominant male or dominant female will be positively correlated with subordinate female GC levels as such interactions are likely stressful to the subordinate female. Finally, we predict that agonistic interactions within the dominant breeding pair will be inversely correlated with subordinate female cortisol levels. We predict this for two reasons. First, dominants occupied in conflict with their mates may have less time to engage in suppression of, or other conflict with, the subordinate female. In this case we would expect subordinate female cortisol levels to be both negatively correlated with dominant conflict and positively correlated with direct agonistic interactions between the subordinate female and one or both dominants. Second, subordinate females with lower cortisol levels may be more likely to reproduce and thus create greater conflict between the dominant breeders.

Methods:

Fish were either wild caught or F1 fish that originated in the Kipili, Tanzania region

(7°26’02”S, 30°35’59”E) of Lake Tanganyika. Groups of three fish (dominant male, dominant female, subordinate female) were maintained in 113.6 L aquariums (30.5 cm X 61cm X 61cm) with a sand substrate of an average depth of 30 mm. Each aquarium contained two halved clay flowerpots that served as a breeding substrate as well as a submersible heater. Each aquarium had 49 a biweekly water change (20%), appropriate water chemistry was maintained using commercially available Tanganyika Buffer (Seachem Laboratories, 1000 Seachem Dr., Madison GA 30650).

Water temperature was checked daily, and water chemistry was checked biweekly the day before the scheduled water change. Tank conditions were maintained to reflect those in Lake

Tanganyika (mean temperature ± SD = 24.9±1.6° C, pH = 7.8-8.4). Fish received daily feeding ad libitum of either dry TetraCichlid Cichlid Flakes (Tetra Holding (US) Inc. 3001 Commerce St.

Blacksburg, VA 24060) food (once daily on 5 days per week) or frozen daphnia (Hikari Sales

U.S.A., Inc. 2804 McCone Ave., Hayward, CA 94545) and brine shrimp (San Francisco Bay

Brand, Inc. 8239 Enterprise Drive, Newark, CA 94560) (once daily on remaining 2 days per week). All tanks were kept on a 12h L : 12h D illumination cycle.

Prior to introducing fish into the experimental tanks, all fish were measured for standard length (SL, mm) and mass (g), and given a dorsal fin clip in order to easily distinguish each from tank mates. Eight groups were created, each consisting of a dominant male (mean SL ± SD=63.4

± 4.98 mm), a dominant female (mean SL ± SD=55.19 ± 1.38 mm), as well as a smaller female fish (mean SL ± SD=38.3 ± 2.7 mm). All fish within a group were unrelated. In order to reduce conflict between the fish upon introduction, the subordinate female fish was released first and the dominant pair was released 24 hours later. Groups then were observed regularly in the following days to ensure that there was no excessive aggression.

Groups were left together for 30 days (Series 1), during which time three 15 minute behavioral observations took place (observation 1: days 4-8; observation 2: days 11-15; observation 3: days 18-20 , and a hormone sample was collected (days 29-30; see below). All reproductive events were documented and eggs were removed when discovered. On day 30, all fish again were measured for SL and mass, and their dorsal fin clips were renewed. Fish were then shuffled randomly among groups such that all individuals were in groups with unfamiliar,

50 unrelated fish and released in the same fashion as before (Series 2) for a total of N=16 groups.

Behavior observations, documentation of reproductive events, and hormone analysis were collected as before.

Behavioral Observations

During each observation all behaviors performed by each group member were recorded using the program JWatcher 1.0 (74). Observed agonistic behaviors included overt aggression

(rams and chases), restrained aggression (displays: fin raises, puffed throat, head down displays, slow approaches), other agonistic behaviors recorded included “avoiding” aggressors and

“blocking” aggressors from attacking another group member. Submissive behaviors (tail quivers), affiliative behaviors (joining and “bumping”), as well as courtship behaviors (soft touch) were also recorded. See Appendix A for detailed descriptions of each specific behavior. All observations were performed by IYL between 1000 and 1500 h. No observations took place within 3 days of spawning.

Immediately before or after each behavioral observation we took photographs of each aquarium every minute for 15 minutes. The camera was oriented ~1.5 m in front of each aquarium such that the frame included the side with the greatest surface area. From these photographs we calculated the mean distance (considering the X and Y, but not the Z axis) between each group member using the software MICAM (version 1.6, http://science4all.nl). The distance between individuals is relevant in understanding both behavioral interactions and GC levels, because avoidance may allow stressed individuals to reduce the frequency of agonistic interactions. Additionally, reducing the frequency of agonistic interactions through avoidance may serve to reduce social stress.

51

Hormone Measurements

In the final two days of each series, we took one measurement of cortisol levels for each fish using a water-borne hormone collection technique. Water-borne hormone analysis has been shown to reliably measure circulating hormone levels and is less invasive than other methods such as caudal venipuncture often used in similarly-sized fish (142-146, see reviews by 147, 148).

Fish were placed in 300 mL of distilled water in a 600 mL beaker within a Styrofoam cooler for one hour. Opaque dividers were placed between beakers to eliminate the risk of visual cues confounding results. Once fish were returned to their tanks, the holding water in each beaker was run through Whatman filter paper (Grade 1) and drawn through a Hypersep C18 column (Thermo

Scientific, Inc.) using a vacuum manifold. These columns were frozen and then shipped to

Tuscaloosa, AL, USA for hormone extraction. Columns were thawed at room temperature, and 2 mL distilled water were passed over the columns using a vacuum manifold to purge salts. The free cortisol fraction was extracted from the column with 2 x 2 ml ethyl acetate elutions into 13 x

100 mm borosilicate vials. Ethyl acetate was dried under a gentle stream of nitrogen gas at 37°C in a water bath, leaving a hormone residue that was resuspended immediately in 800 μl of 5%

EtOH:95% EIA buffer provided with Cayman Chemicals, Inc. kit (40 μl 100% EtOH followed by

1 min vortexing and addition of 760 μl EIA buffer followed by 20 min vortexing).

Assays were validated for N. pulcher by assessing parallelism of a serial dilution curve derived from pooled N. pulcher samples with the standard curve. A 3 ml N. pulcher pool of water-borne hormones was obtained by combining 100 μl from 30 resuspended experimental samples. The pool was serially diluted 1:1 to 1:128. The slopes of the pooled serial dilution curve and the standard curve were parallel [slope comparisons, Zar (149), p. 355: t12=0.04, P=0.97].

Samples were run in duplicate on three 96-well plates and a pool was included in duplicate at the

52 beginning and end of each plate. Intra-assay coefficients of variation were 13.34%, 5.31%, and

3.85%; the inter-assay coefficient of variation was 12.08%.

In order to reduce the risk of stress related to the procedure itself confounding our results, we performed three habituations under identical conditions to those during the actual analysis

(146). These habituations occurred once a week and occurred at least 4 days before the next behavioral observation took place for that group. All water-borne hormone collections occurred between 1230 and 1900 h and not within 3 days of spawning.

Statistical Analysis:

Square-root transformed frequencies of behaviors for each pair (subordinate female- dominant female, subordinate female-dominant male, dominant female-dominant male) were summarized using three separate Principal Components Analyses. We separated behaviors based on the actor and recipient in each dyad (e.g., the frequency of dominant males “ramming” a dominant female was categorized separately from the frequency of dominant females “ramming” a dominant male). All frequent behaviors that were normally distributed after transformation were included in the PCA (see Appendix 2 for list of behaviors included). Factors with eigenvalues greater than 1.0 were rotated using Varimax rotation and Kaiser normalization. Table 1 shows all factors with eigenvalues greater than 1.0 for each Principal Components Analysis. To interpret these factors, we considered loadings of greater than +0.5 or -0.5 in the rotated component matrix to be particularly relevant following Hamilton et al. (15).

Factors that were interpreted as measures of agonistic behavior within each dyad under consideration, as well as affiliative behaviors by the subordinate female toward each dominant

(which may prevent aggression from the dominant or reflect a lack of conflict with the dominant) were each included as fixed effects in a linear mixed model (LMM), with log-transformed

53 subordinate female cortisol levels as the dependent variable. One cortisol value was an extreme outlier (>2 times the next-greatest value), this sample was omitted from our analysis. Cortisol levels were measured in pg/sample; to account for variation in cortisol levels based on subordinate female mass, we considered subordinate female mass as a fixed effect in each model.

In order to account for any variation in dyadic behaviors based on the proximity of individuals to each other, we also considered mean distance between the focal dyad of interest in each model as a fixed effect. Additionally, we considered whether reproduction had occurred in each group as a binomial fixed effect. Finally, we considered the interaction of each principal component factor and mean distance between the focal dyad. To account for individual effects on cortisol levels in models comparing subordinate female behavior and cortisol levels, we considered subordinate female identity as a random effect in all models. Similarly, to account for individual consistency in behavioral interactions across series, we also considered the identity of the other interactant(s) in the dyadic pair of interest as a random effect in each model. If models failed to converge, we removed variables, starting with the interaction between the PC factor and mean distance; final models are the most complete model that achieved convergence.

Full Model: yijk = intercept + Presence of eggs + Sub. fem. mass + PC Factor + Mean dist. + PC Factor*Mean

ɸ ɸ ɸ dist. + Sub. fem. Identity + Other identity 1 + Other identity 2†

ɸ Denotes random factors.

† This factor was only relevant to models which considered dominant male – dominant female interactions.

54

Over the course of the experiment, each subordinate female was placed in two different groups. To determine whether subordinate female identity was relevant in understanding variation in cortisol levels (that is, were individual cortisol levels between treatments more similar than would be expected by chance), we performed a likelihood ratio test between models that included log-transformed subordinate female cortisol levels as the dependent variable and either i) subordinate female mass as a fixed effect, or ii) subordinate female mass as fixed effect and subordinate female identity as a random effect.

Results:

Principal Components Analysis:

We extracted 3 factors with eigenvalues greater than 1.0 for the interactions between the dominant male and subordinate female, 3 factors with eigenvalues greater than 1.0 for the interaction between the dominant female and subordinate female, and 6 factors with eigenvalues greater than 1.0 for the interaction between the dominant male and dominant female. Rotated factors are presented on Table 6 (see Appendix B for specific loadings of behaviors on each factor). For each set of pairwise interactions, one or more factor was extracted that we interpret as a measure of agonistic interactions (Table 6). The factor that explained the greatest proportion of variation for each interaction between the subordinate female and each dominant fish [Dominant

Male Subordinate Female Factor 1 (DMSF Factor 1) and Dominant Female Subordinate Female

Factor 1 (DFSF Factor 1), respectively] was characterized by heavy loadings of aggression on the part of the dominant fish and submissive behavior on the part of the subordinate female. The second factor for the dominant female subordinate female interaction (DFSF Factor 2) was heavily loaded with joining and “bumping” behaviors; we interpret this factor as a measure of

55 affiliativeness within this dyad. We interpret the first factor for the dominant male – dominant female dyad (DMDF Factor 1) as a measure of aggression by the dominant male and submissive behavior by the dominant female. The second factor for the dominant male dominant female

a) Dominant Male - Subordinate Female Factors

Eigenvalue Var. Exp. Description Dominant male aggression, subordinate female DMSF Factor 1 2.640 37.710 affiliative and submissive behavior DMSF Factor 2 1.225 17.503 Dominant male joining DMSF Factor 3 1.175 16.790 Dominant male fin raise (display) b) Dominant Female - Subordinate Female Factors

Eigenvalue Var. Exp. Description Dominant female aggression, subordinate female DFSF Factor 1 2.897 36.218 fin raise (display) and tail quiver (submissive) DFSF Factor 2 1.557 19.460 Subordinate female affiliative behavior Dominant female fast approaches (display) and DFSF Factor 3 1.178 14.721 joining behavior c) Dominant Male - Dominant Female Factors

Eigenvalue Var. Exp. Description Dominant male overt and restrained aggression, Dominant female opercular spreading (display), tail DMDF Factor 1 3.931 23.122 quivering (submissive), and aversion to joining (affiliative) Mutual restrained aggression, lack of "soft DMDF Factor 2 3.054 17.963 touch/bite" (courtship) by dominant male Dominant female slow approaches (display) and DMDF Factor 3 2.572 15.132 "bumps," dominant male unlikely to perform fast approaches (display) DMDF Factor 4 2.092 12.307 Dominant male "blocks," joining behavior DMDF Factor 5 1.460 8.586 Dominant female avoiding dominant male DMDF Factor 6 1.451 8.536 Dominant female ram (attack) dominant male

Table 6: Principal Component Factors for pairwise interactions amongst group members. Factors considered in subsequent LMMs are in bold.

56

interaction (DMDF Factor 2) was characterized by heavy loadings of mutual restrained aggression, as well as a lack of “soft touches,” which are associated with courtship and often precede breeding (150). The fifth factor for the dominant male dominant female interaction

(DMDF Factor 5) was characterized by heavy loadings of the dominant female avoiding the dominant male. In sum, six PC factors characterized patterns that aligned with behavioral patterns we predicted to correlate with subordinate female cortisol levels. Each of these factors was included in a linear mixed model. These factors were DMSF Factor 1, DFSF Factor 1, DFSF

Factor 2, DMDF Factor 1, DMDF Factor 2, and DMDF Factor 5 (see Table 6 for a description, and Appendix B for the specific loadings of particular behaviors for each of these factors).

Figure 9: Relationship between subordinate female cortisol and one measure of agonistic interactions (restrained aggression and a lack of courtship) between the dominant male and dominant female (DMDF Factor 2, see Table 7c). Plotted values are subordinate female cortisol levels (pg/gram) in relation to DMDF Factor 2.

57

Figure 10: Relationship between subordinate female cortisol and a measure of affiliative behavior (joining and bumping) by the subordinate female towards the dominant female (DFSF Factor 2, see Table 7b). Plotted values are subordinate female cortisol levels (pg/gram) in relation to factor scores for DFSF Factor 2.

What was the range of subordinate female cortisol levels, and do individual subordinate females have similar cortisol levels across series?

Subordinate female cortisol levels ranged from 151.79-3080.79 pg/gram body mass

(mean ± SE: 817.3 ±239.5); note that linear mixed models considered log-transformed cortisol levels (pg/sample) as the dependent variable, and included subordinate female mass as a fixed effect. Individuals were found to have significant consistency in cortisol levels across series

2 (maximum likelihood test: X 1=7.04, p<0.01).

58

a) Dominant Male - Subordinate Female: Agonism Numerator Denominator D.F. D.F. F p Subordinate Female Mass 1 6.938 0.836 0.391 DM-SF mean distance 1 7.973 0.087 0.775 DMSF Factor 1 1 10.924 0.428 0.527

b) Dominant Female - Subordinate Female: Agonism Numerator Denominator D.F. D.F. F p Subordinate Female Mass 1 9.296 0.920 0.362 DF-SF mean distance 1 7.103 0.170 0.693 DFSF Factor 1 1 8.540 0.129 0.729

c) Dominant Male - Dominant Female: Agonism Numerator Denominator D.F. D.F. F p i Subordinate Female Mass 1 8.779 0.814 0.391 DM-DF mean distance 1 9.530 0.240 0.635 DMDF Factor 1 1 8.954 2.785 0.130

ii Subordiante Female mass 1 11.621 2.467 0.143 DM-DF mean distance 1 5.500 3.021 0.137 DMDF Factor 2 1 5.169 6.822 0.046

iii Subordinate Female mass 1 9.114 0.793 0.396 DM-DF mean distance 1 7.595 1.195 0.308 DMDF Factor 5 1 7.182 0.784 0.405

d) Dominant Female - Subordinate Female: Joining and "bumping" Numerator Denominator D.F. D.F. F p Subordinate Female Mass 1 7.364 0.014 0.910 DF-SF mean distance 1 12.000 12.914 0.004 DFSF Factor 1 1 9.046 0.303 0.595

Table 7: Output of Linear Mixed Models which considered log-transformed subordinate female cortisol levels as the dependent variable. Model selection was based on relative AICc values. Significant effects are in bold and italics. “Eggs” was a binomial fixed effect which described whether reproduction had occurred within the group.

59

Is there a relationship between dyadic interactions and subordinate female cortisol?

We did not find a significant effect of agonistic interactions between subordinate females and either dominant males or dominant females (DMSF Factor 1, DFSF Factor 1) on subordinate female cortisol levels (Table 7a,b). We found a significant, positive effect of DMDF Factor 2

(restrained aggression and lack of courtship behaviors within the dominant breeding pair) on subordinate female cortisol levels (Table 7c, Figure 9). This effect was amplified when the interaction of the principal component scores and mean distance between the dominant breeding pair was included; subordinate female cortisol was highest when the dominant pair engaged in frequent mutual restrained aggression and were also in close proximity to one another.

Subordinate females that displayed more joining and “bumping” behaviors towards dominant females also had significantly lower cortisol levels (Table 7d, Figure 10). In both of these models, subordinate females in groups in which reproduction occurred had lower cortisol levels (Table

7c,d; respectively).

Discussion:

We found an association between one measure of conflict between the dominant pair;

(DMDF factor 2) and subordinate female cortisol; when the dominant pair engaged in more frequent conflict subordinate females had higher cortisol levels. This relationship was magnified when individuals of the dominant pairs were also physically closer to one another. DMDF factor

2 included heavy loadings of restrained aggression by both dominants towards their mates as well as relatively infrequent “soft touches,” which are associated with courtship behavior (150). High scores for this factor may not simply indicate conflict within the pair, but also a low likelihood of reproduction occurring within the pair.

Reproduction in groups may be indicative of some level of social stability which could explain lower cortisol levels in subordinate females from these groups. We found evidence that 60 when reproduction had occurred in groups, subordinate female cortisol levels were lower; however we cannot say with certainty whether dominant or subordinate females produced these eggs. Reproduction by subordinate females is rare in the lab and the field, but it does occur (85,

90, 108). Reproduction occurred infrequently in groups in this study, and none of the principal components describing interactions between the dominant male and subordinate female indicated patterns of courtship behavior within this dyad. Knowing the maternity of eggs in more groups would be important to determine whether the presence of eggs is indicative of group stability, or conversely, is associated with increased conflict.

Relatively few studies have investigated indirect effects of fellow group members on an individuals’ endocrine state. Alberts et al. (151) found that an individual could influence the hormonal state of other group members without direct agonistic interactions with that individual.

Specifically, they described elevated cortisol levels in a group of yellow baboons (Papio cynocephalus) after a particularly aggressive male joined the group. While this shift was most pronounced in individuals that had experienced direct agonistic interactions with the joiner, an increase in cortisol levels occurred in all group members. In a metaanalysis of the relationship between dominance and stress, Goymann & Wingfield (131) associated differences in allostatic load with status-specific variation in cortisol levels. In their framework, escalated conflict between dominants, which would presumably increase their allostatic load, might be associated with subordinate individuals having lower cortisol levels relative to dominants. While our analysis did not compare cortisol levels of individuals of different status, we did find that subordinates in groups with more agonistic interactions between dominants had higher cortisol levels than subordinates in groups with less dominant aggression.

Subordinate females behaved differently depending on whether they received aggression from the dominant male or female. In DMSF Factor 1, aggression by the dominant male was

61 associated with affiliative and submissive behaviors by the subordinate female. We found no instances of aggression by the subordinate female towards the dominant male. Dominant female aggression in DFSF Factor 1 was associated with fin raises by the subordinate female, which we classified as a form of restrained aggression. These differences in subordinate response to dominant aggression suggests that competition over reproduction may exist between the females; subordinate females behaved aggressively towards relatively aggressive potential reproductive competitors (dominant females), but behaved submissively towards relatively aggressive potential mates (dominant males).

Regardless of the differences in how subordinate females behaved in agonistic interactions with dominant males and dominant females, we found no evidence that these behaviors were associated with subordinate female cortisol levels. Similar results were found by

Bender et al. (121) for subordinate males. They suggested variation in subordinate male cortisol has less to do with aggression but rather variation in submissiveness, which may reduce future aggression. While we did not find any patterns related to subordinate female submissiveness, affiliative and bumping behaviors by the subordinate female towards the dominant female were associated with lower subordinate female cortisol levels. These behaviors may indicate a lack of conflict with the dominant female. Bumping in particular may be either (or both) affiliative or conciliatory (15). If conciliatory, bumping may allow subordinate females to maintain social connections within the group while avoiding agonistic interactions or risking eviction.

Mileva et al. (141) found no relationship between subordinate female cortisol levels and subordinate female affiliative behavior towards group members in N. pulcher. We suggest a resolution to this apparent conflict may lie in discriminating between the individuals with which a focal fish interacts. The groups in Mileva et al.’s (141) study included additional subordinates, and affiliative (and other) behaviors were summed for all recipients. Subordinate individuals

62 often interact differently with group mates depending on sex and relative rank (33, 66), and rates of affiliative behavior by the subordinate female towards the dominant female may not be accurately reflected by summed affiliative behavior towards all group members. Mileva et al.

(141) also described conflicting findings between their results and those of Bender et al. (121).

Mileva et al. (141) found cortisol levels to be positively correlated with the frequency of submissive behaviors in subordinate males, while Bender et al. (121) found the opposite, and suggested the discrepancy between studies might result from acute stress caused by the water- borne hormone analysis procedure. We find this unlikely as an explanation for our differing results, as we implemented a habituation procedure that has been shown to dissipate any acute stress associated with the procedure (146). Given the value of this species in the study of sociality and cooperation, and the fact that the small size of N. pulcher currently prohibits performing repeated measures of plasma cortisol without sacrificing or at best severely harming subjects, a direct comparison of these methods in this species would be valuable not only to provide insights regarding apparent conflict in already published results but also to allow for more consistent experimental design in future studies.

Conclusions:

Subordinate female cortisol levels were positively related to the frequency of agonism between the dominant breeders in their group, and inversely related to the frequency of affiliative behaviors performed by the subordinate female towards the dominant female. Subordinate female cortisol levels were not associated with the frequency of direct agonistic interactions with either the dominant male or female. These results suggest that interactions between dominant individuals are related to basal cortisol levels of subordinate female group members. This study highlights the importance of direct interactions with other group members as well as indirect effects associated with interactions among other group members in understanding the causes and 63 consequences of variation in GC levels of highly social animals. Ecologists have long understood the relevance of indirect effects in ecological communities (152, 153); more recently the relevance of indirect effects in social systems also has come to light (18, 19, 122, 154, 155). This study compliments this body of literature by emphasizing the importance of considering indirect influences on individuals in other settings – the behavior, physiology, and life history strategies of organisms are often better understood when the larger social context is considered.

64

Chapter 5: Social status influences responses to unfamiliar conspecifics in a cooperatively

breeding fish

Abstract: Individuals may visit a social group with the intention of eventually joining the group or to exploit the group’s productivity (via cuckholdery or cannibalism) without permanently joining. For members of the social group, the costs and benefits of such visits will depend on both the visitor’s intentions and the group member’s status within the group. In this study we experimentally tested how visitors are perceived, and how social status influenced the responses of the group. Using wild groups of the cichlid fish Neolamprologus pulcher, we compared group member responses to unfamiliar conspecifics in control groups and in groups that had been experimentally manipulated by temporarily removing a subordinate of the same size and sex as the visitor. In control groups, high-ranking females and subordinates of the same size and sex as the stimulus fish responded most aggressively. High-ranking fish were less aggressive towards stimulus fish in removal groups than in control groups, while low-ranking subordinates were more aggressive in the removal treatment. These results suggest that current group members perceived stimulus fish to be potential group joiners, and that high-ranking group members may benefit from the replacement of lost group members, while low-ranking group members benefit from a shortened breeding queue.

65

Introduction:

Social groups that defend territories against conspecifics may nevertheless receive visits from non-group members (26, 156-158). The reaction to these visitors can be variable among and within groups (88, 92, 159) and this variation in the response to visitors may reflect differences in the goals of visitors as well as the costs and benefits inflicted by the visitor on different group members. For example, an individual may visit a social group in order to begin the process of joining that group, facilitating entry into this social unit sometime in the future

(158). In contrast, individuals may visit a social group to mate or to exploit the group by parasitizing reproduction or cannibalizing young (160-163). Such behaviour may impose costs upon the visited group, but certain members of the group may face lower costs than others and some members may even benefit from these actions. Differences in net costliness of visitation for different group members can lead to differences in the reaction to visitors. In this study, we investigate how group members of a cooperatively breeding cichlid fish respond to simulated visitors, and whether a group member’s social status influenced its response to these visitors.

There are multiple ways in which visitors could inflict differential costs across individual group members. For example, a dominant reproductive male in the group could lose parentage to a male visitor who manages to spawn with his partner, but would not lose parentage to a female visitor (he may even gain additional reproductive success if he can mate with her - 131, 162).

Therefore, males would be expected to defend more against male visitors and less against females

(29). If visitors exploit the productivity of groups in other ways such as stealing food (163) or cannibalizing vulnerable group members (160, 161), then the response to visitors should depend on the value of the lost resource to each group member. For instance, if offspring are consumed, then close relatives (e.g., both parents and full siblings) will face greater costs than other, less related group members.

66

In many group-living animals, individuals disperse from their natal group into a new group, and visits to prospective groups may precede this dispersal event (25, 33, 101, 158, 159,

164-169). If visitation precedes joining the group now or in the future, then the response of resident group members to visitors will depend on both the current costs and benefits of the visit and the future fitness effects of joining. When an individual joins, the group will increase in size, which may carry both benefits and costs on current group members (3, 170). For example, being part of a larger group may offer greater protection from predation through dilution, predator confusion, defensive efficiency, or increased vigilance (3, 25, 101, 155, 171, 172). Living in a larger group, however, may carry costs including increased conspicuousness to predators, greater competition for resources and reproduction, and higher parasite or disease transmission (171,

173). If a social hierarchy exists then higher ranked individuals may stand to gain from potential joiners, while a joiner might push lower ranked individuals further down the social ladder (106,

174-177). Therefore, the behavioural response to a potential joiner is expected to depend on both the size of the current group and on each individual’s risk of being usurped from its current social position.

Using wild groups of the cooperatively breeding cichlid fish Neolamprologus pulcher, we examined how group members of different social and reproductive status respond to staged visitation events. N. pulcher is a useful system to examine how group members perceive visiting individuals because extra-pair and extra-group reproduction, egg and fry cannibalism, and inter- group dispersal all occur (25, 102, 165, 178-180). Dominance is strongly size based in this species (62, 66, 181). In this study, we presented an unfamiliar conspecific (hereafter “stimulus fish”) of the same sex and similar size to the largest subordinate in the group and recorded the rates of aggression that this stimulus fish received from each group member. In half the social groups, we temporarily removed the largest subordinate group member (hereafter “removal

67 treatment”) and so reduced the group size before presenting the stimulus fish. The remaining groups maintained their full membership and served as controls.

If the stimulus fish were perceived as attempting to parasitize reproduction, then we predicted that dominant males would be more aggressive towards male stimulus fish than towards female stimulus fish. As the costs associated with such visits are not expected to change with group size, we predicted no differences in aggression between the removal and control treatments.

If instead stimulus fish were perceived as egg or fry predators, then we predicted that both dominant individuals (males and females who stand to lose direct fitness) would behave most aggressively towards the stimulus fish. Expected relatedness of subordinates to current group offspring is lower than that of dominants to current group offspring, and relatedness further declines with the age of the subordinate in this species (82). Under this hypothesis we predicted that group members would not differ in aggression towards stimulus fish between treatments, as previous work has shown that N. pulcher do not compensate for lost subordinates (106).

Finally, we predicted that if stimulus fish were perceived as potential future joiners, then group members at risk of rank usurpation would respond the most vigorously and aggressively towards the stimulus fish (92, 182). After the removal of a high-ranking subordinate, we predicted that dominant fish would reduce their rates of aggression towards stimulus fish because dominant fish may stand to gain from replacing the removed group member (106). However, we predicted that the remaining subordinates who are of lower rank would increase their rate of aggression towards the stimulus fish in defence of their recent promotion in rank.

Methods:

Study Species:

68

Neolamprologus pulcher is endemic to Lake Tanganyika in East Africa, where it forms social groups that typically consist of a single dominant breeding pair and on average 7-9 subordinates of both sexes (range: 1-20, - 64, 65). Dominance in N. pulcher is strongly associated with body size (62); the dominant male is the largest group member, while the second largest is typically the dominant female (26). Subordinate group members form a size-based dominance hierarchy (33, 66). Subordinates maintain and defend the group territory and may provide alloparental care (25, 183). The presence of subordinates has a net positive effect on the reproductive success of dominants (106). Groups with more subordinates can raise more young, and larger groups of N. pulcher hold higher quality territories (64). Territories containing large groups are more likely to continue to support social groups across years (65), suggesting a positive effect of group size on group persistence over time.

Experimental Protocol:

This experiment was conducted in the spring of 2013. Using SCUBA we located n=43 N. pulcher social groups between 11 and 14 meters in depth offshore of Kasakalwe Point (8°46’S,

31°4’E) on the southern shore of Lake Tanganyika, near Mplungu, Zambia. In each social group we identified and captured the largest subordinate (hereafter the focal subordinate, there were 22 groups where the largest subordinate was male: mean standard length [SL] ± S.E.: 41.8±3.2 mm and 21 groups where the largest subordinate was female: mean SL ± S.E.: 44.0±2.1 mm). Sex was determined by examination of the genital papilla (33, 184). For each of these focal fish we measured standard length from the tip of the snout to the end of the caudal peduncle (185).

We randomly assigned half of the social groups in this study (n=21) to the removal treatment in which the focal subordinate fish was placed inside a mesh bag within a closed minnow bucket and moved at least 3 meters away from the focal social group’s territory. This

69 distance ensured that group members would not interact with the removed subordinate as N. pulcher rarely move far from their territory boundaries (186). In the n=22 control groups, we immediately released the captured focal subordinate back into its group following measurement and sexing. We then caught a sex and size matched stimulus fish (within 2 mm SL) to each focal subordinate fish from a territory at least 15 meters away. This stimulus fish was placed in a transparent perforated plastic 2.75 L container (~14 cm x 14 cm x 14 cm) that had eight small holes to allow water flow thus providing group members with visual and chemical cues from the stimulus fish.

Within ten minutes of capturing the focal subordinate, the container containing the stimulus fish was placed within the groups’ territory approximately 20 cm from the center of the territory. We gave the group and stimulus fish one minute to habituate to the disturbance of placing the container in their territory and then began a 10-minute observation on all group members. We recorded counts of aggressive displays and overt aggressive attacks directed towards stimulus fish by the dominant pair, by the focal subordinate fish (in control groups only), and by any other smaller subordinates in each social group. Aggressive displays consisted of head down postures, fin raises, and puffed throats, and did not involve the aggressor contacting the presentation container. Overt aggressive attacks consisted of chases, rams, and bites in which the individual made physical contact with the presentation container. These behaviours are clearly defined and have been described in a number of ethograms developed for this species (71, 72,

140, 150, 187, 188). “Fish in a jar” protocols have previously been used in fish behavioural studies (67-69), including in studies on N. pulcher (29, 70, 119, 165, 186-189), and elicit naturalistic responses from the animals exposed to this stimulus. After the 10-minute observation period, the stimulus fish was returned to its capture site. Also, focal subordinate fish from the

70 removal treatment groups were returned to their groups. All presentations occurred between

10:00 and 17:00.

The frequency of aggression produced by each group member towards the stimulus fish could be interpreted as the amount of defense against an intruder (29, 70), or as the inverse of the willingness of that group member to accept the stimulus fish into the group now or in the future

(190). The number of subordinates (individuals greater than 1.5 cm SL) varied between social groups (range: 1-13). However, the number of smaller subordinates per group did not differ between treatments (independent samples t-test: t=0.104, df: 41.9, P=0.92). Nonetheless, to account for any variation among social groups related to group size, we included the number of subordinates in the group as a factor in subsequent statistical models (see below). We did not track each subordinate fish smaller than the stimulus fish individually, so the counts of aggressive behaviours performed by all non-focal subordinates within a group were pooled. Anecdotally we observed that attacks and displays were primarily performed by the largest of these smaller subordinates. Desjardins et al. (191) found that N. pulcher vary in their activity through the course of the day. Because our observations occurred in both the morning and afternoon, we controlled for this variation by including this information as a binomial effect in our statistical analyses (see below).

We first tested whether the sex or status of individuals in the control groups was related to their aggressive responses towards stimulus fish using generalized linear mixed models with the attack and display counts by individual group members (and a pooled value for the non-focal subordinates) as the dependent variable. This model included individual status, stimulus fish sex, their interaction, group size, and time of day as fixed effects. Group identity was treated as a random effect. Counts of aggressive acts were overdispersed so a negative binomial distribution with a log-link function was used.

71

Second, we investigated whether the removal of a high-ranking subordinate influenced the status-specific response of dominant males, dominant females, and smaller subordinates. We included individual counts of aggressive attacks and displays (respectively) towards stimulus fish as the dependent variable in each of these generalized linear models. Treatment (removal versus control groups), the sex of the stimulus fish, their interaction, as well as the time of day were included as fixed factors, and as in the previous model, the number of subordinates in the group was included as a covariate. As before, these models used a negative binomial distribution with a log link-function. All statistical analysis was performed using SPSS 22.0.

Figure 11: Boxplot of counts of displays (white) and attacks (gray) towards the stimulus fish in control groups (n=22). Values for non-focal subordinates represent the sum of all subordinate group members smaller than the focal individual. Letters A and B indicate significantly different counts of displays, letters C and D indicate significantly different counts of attacks. Data presented as boxplots show the median and quartiles, as well as minimum and maximum values excluding >1.5 times the interquartile range (marked with circles; diamonds indicate values >3 times the interquartile range).

72

Figure 12: Boxplot of aggressive acts towards the stimulus fish; a) attacks by dominant females, b) displays by dominant females, and c) displays by dominant males in control groups (n=22) and removal treatments (n=21). Boxplots show the median and quartiles, as well as the minimum and maximum values, as well as minimum and maximum values excluding >1.5 times the interquartile range (marked with circles; diamonds indicate values >3 times the interquartile range).

73

Figure 13: Boxplot of a) displays and b) attacks towards the stimulus fish by non-focal subordinate group members in the removal treatment (n=21) and control groups (n=22). Boxplots show the median and quartiles, as well as the minimum and maximum values, as well as minimum and maximum values excluding >1.5 times the interquartile range (marked with circles; diamonds indicate values >3 times the interquartile range).

74

Results:

Does group member status influence their responses to the stimulus fish?

Dominant females and the sex- and size-matched subordinates performed more aggressive attacks towards the stimulus fish than did dominant males or smaller subordinates in control groups (F3,78=6.62, P<0.001; Figure 11). Dominant males performed significantly fewer aggressive displays than any of the other group members (F3,78=6.56, P=0.001; Figure 11). Male stimulus fish in control groups received more aggressive displays than female stimulus fish overall (F1,78=4.298, P=0.041), but the interaction of individual status and stimulus fish sex was not significant (F3,78=0.11, P=0.954). There was no relationship between the number of overt attacks a stimulus fish received and its sex in control groups (F1,78=0.08, P=0.776).

How does subordinate removal influence aggression towards the stimulus fish?

Dominant females performed fewer attacks towards the stimulus fish in the removal treatment groups relative to dominant females in unmanipulated groups (Treatment: X2=15.18, p<0.001; Figure 12a). In both treatments, dominant females performed fewer overt attacks towards male stimulus fish than towards female stimulus fish (Stimulus fish sex: X2=4.66,

P=0.031; Figure 12a). There was also a non-significant trend towards dominant females performing fewer aggressive displays towards stimulus fish in the removal treatment (Treatment:

X2=3.18, P=0.074; Figure 12b). Dominant males very rarely performed overt attacks and they were performed by only a few individuals; thus overt attacks by dominant males were not considered in subsequent analyses. Dominant males performed fewer threat displays against male stimulus fish in removal treatment groups than they did in the control groups, but dominant males did not differ in their behaviour towards female stimulus fish depending on treatment

(Treatment*Stimulus fish sex: X2=3.98, P=0.046, Figure 12c). In contrast to dominant fish, the

75 smaller non-focal subordinates performed more aggressive displays (Treatment: X2=4.06,

P=0.044; Figure 12a) and more overt aggressive attacks (Treatment: X2=8.28, P=0.004; Figure

13b) towards the stimulus fish in removal groups relative to control groups.

Group size did not affect the number of aggressive displays performed towards the stimulus fish by the dominant male (X2=0.21, P=0.65) or the number of overt attacks and aggressive displays by the dominant female (Attacks: X2=0.19, P=0.66; Displays: X2=0.35,

P=0.55). Neither the sex of the stimulus fish, nor the number of subordinates in the group had a significant effect on the number of overt attacks (Stimulus fish sex: X2=0.03, P=0.87; number of subordinates: X2=0.03, P=0.85) or aggressive displays (Stimulus fish sex: X2=0.79, P=0.37;

Number of subordinates: X2=0.06, P=0.81) performed by smaller non-focal subordinate fish.

Discussion:

Our results were consistent with the hypothesis that stimulus fish were perceived as potential joiners to the group. Size- and sex-matched subordinates and dominant females from intact groups exhibited more frequent aggressive behaviour than did other group members.

Subordinate group members matched in terms of size and sex would face the greatest conflict over status with the stimulus fish, and several other studies have shown that agonistic interactions in N. pulcher are greatest between individuals close in size (15, 66, 146, 183).

Dominant females face potential reproductive competition from female joiners, and indeed were more aggressive toward female stimulus fish than towards male stimulus fish. Sex- specific aggression towards unfamiliar individuals of the same sex and status has also been described in the group-living false clown anemonefish, Amphiprion ocellaris (92). Lewis (170) observed a similar pattern in female sifaka (Propithecus verreauxi verreauxi). Females in Lewis’ study behaved aggressively towards female non-group members, and encouraged males to reside in the group through affiliative behaviours. Lewis (170) suggested that female group members 76 facilitate membership depending on the sex of potential joiners. While dominant males could eventually face reproductive competition with successfully joining male intruders, they did not differ in their defense against male or female stimulus fish in control groups, and after a removal were less aggressive towards male intruders. This finding suggests dominant males do not consider visiting stimulus fish to be reproductive competitors.

Both male and female dominants were less aggressive toward the stimulus fish in groups from which a subordinate had been removed. This result is inconsistent with the hypotheses that stimulus fish were perceived as cuckholders or predators of eggs or fry. We suggest that dominants may benefit from restoring the number of large subordinates in a group after an experimental removal. Allowing new group members to join and restoring a group back to its previous size may increase alloparental care (64, 88) or dilute predation risk (1, 101, 192).

Indeed, Schaffner and French (88) found that dominant female marmosets, Callithrix kuhli, from small groups behaved less aggressively towards strangers, and suggested that dominant individuals may facilitate group joining events by being tolerant of unfamiliar individuals. These benefits of increased group size could have also applied in the control treatment. However, dominants might benefit less from visitors in control groups for two reasons: 1) they may face diminishing returns of additional subordinates as group size increases beyond the resources/shelters available in that territory and 2) there may be increased conflict amongst subordinates as group size increases (33, 35, 183, 193, 194), which could detract from alloparental care and the time dedicated to maintaining and defending the territory (106), or attract attention from predators (174-177).

The behaviour of smaller subordinates was also consistent with the hypothesis that simulated visitors were perceived as potential joiners to the group. Smaller subordinates behaved more aggressively towards the stimulus fish in the removal treatment than in control groups. We

77 suggest that the subordinates in our experiment perceived that they had increased in rank due to the removal, and thus increased their aggression towards unfamiliar individuals threatening their newly acquired rank (183). Interestingly, if small subordinates in the reduced groups did indeed perceive a change in their rank, this occurred quite rapidly (<15 minutes after the removal).

Behavioural and physiological indicators of perceived changes in rank have been shown to arise quite rapidly in another cichlid fish, Astatotilapia burtoni (195). It is also possible that smaller subordinates were compensating for the lost defensive efforts of the missing focal subordinate group member, although Brouwer et al. (106) found no evidence that remaining subordinate group members increased their amount of territory defense after a similar subordinate removal experiment. We also found no evidence that the overall size of the group had an effect on subordinate aggression.

Our results are most congruent with the hypothesis that visitors were perceived as prospective joiners to the group, however we do not suggest visitors would have been accepted into the group on their initial visit. While N. pulcher often disperse and subsequently join groups

(82, 179), the process by which joining events occur is not well understood. Joining events have been observed only a handful of times in nature (158) and the response to these immigrants was initially aggressive in nature (S. Balshine, pers. obs.). Zöttl et al. (180) found that immigrating fish that faced high rates of aggression from a breeding pair were generally rejected from the group. However, many of these groups were visited multiple times each hour, and it appears that successful joiners are familiar with their new group by the time they join.

It is unclear why male stimulus fish faced less aggression from dominant females in general, or why dominants of both sexes appear to be more tolerant of them than female stimulus fish after a removal. Stiver et al. (2006) found that removed dominant males were much more likely to be replaced by joining individuals, while dominant female status was more often claimed

78 by a female already in the group. Also, male N. pulcher disperse further (82) and more frequently

(26, 179) than do females. Sex differences in dominant response may simply result from visiting subordinate males being a more common occurrence for N. pulcher groups.

Our results were not consistent with visitors being perceived as egg or fry predators.

Although subordinates increased their rates of aggression after a removal, which is consistent with compensation for a lost group member, dominant fish, which are the most invested in the current offspring, performed fewer aggressive acts after removal. Desjardins et al. (29) also found that group members responded differently to unfamiliar N. pulcher than to heterospecific predators of adult or juvenile N. pulcher, further suggesting that the group did not perceive our stimulus fish as potential predators.

Our study complements previous work on between-group movement in cooperative species (88, 158, 166, 190, 196-198) by examining the factors underlying current group response to a visitor. Here we present evidence that the behaviour of individual group members differed depending on status and sex, and were consistent with stimulus fish being perceived as potential joiners. Smaller subordinates’ responses depended on their relative rank, while dominants’ response depends on whether a subordinate has recently been removed from the group and, for dominant females, on the sex of the stimulus fish. Previous studies have suggested that subordinate N. pulcher are familiar with neighboring groups (70, 158), that they strategically reduce investment in their current group prior to dispersal (180), and that movement between groups can occur rapidly after a position in the hierarchy is vacated (33, 179, 199). Our results highlight that group joining decisions, and thus emerging group structure, are the product of both joiner preferences and the responses of current group members.

79

Chapter 6: Strategic decision making in a three-player sequential game: order effects

influence the outcome in cooperatively breeding groups.

Abstract: In cooperatively breeding groups, members may be in conflict regarding reproductive skew within the group. While subordinates may accrue direct fitness gains through reproduction, subordinate reproduction may be associated with net losses in fitness for same-sex dominants through reproductive competition and lost alloparental care. In contrast, opposite-sex dominants may experience net fitness gains if they parent subordinate offspring. Using a sequential game, we modeled this conflict in a group consisting of a dominant male, a dominant female, and a subordinate female. Subordinate females had a choice to reproduce, dominant females had a decision to “police” subordinate reproduction in groups and destroy subordinate clutches they found, and dominant males had a decision to “patrol” groups and prevent policing dominant females from destroying clutches. We compared the outcomes of the six possible sequences in which these decisions could be made and determined that in cases in which one individuals’ optimal strategy led to a negative shift in fitness for another group member, individuals who moved first did better in the game. Additionally, these results may explain the evolution of signals to inform group mates of strategies prior to the onset of situations which might lead to conflict in groups.

80

Introduction:

Cooperatively breeding animal groups include subordinate individuals who limit or forego their own reproduction but provide alloparental care for the offspring of dominant group members (22-25). While alloparental care provides clear benefits to dominant individuals (101,

107), dominants may face costs from reproductive same sex subordinates. Reproductive subordinates may compete for fertilization (e.g. sneaker males - 200) and the offspring of subordinates may compete with those of dominants for limited resources in the group (12). On the other hand, reproductive opposite-sex subordinates may provide additional fitness gains for dominants as additional mates (e.g., increased number of offspring, increased offspring diversity, fertilization assurance – 116, 201-203). Thus, dominant males and dominant females, if present, may face different costs and benefits from the presence and behavior of potentially reproductive subordinates.

Dominant individuals can limit direct reproduction by subordinates through preemptive aggression (109), infanticide (90), or eviction (80, 118). If dominants attempt to limit reproduction by same sex subordinates, this may result in conflict with opposite sex dominants that are potential parents of subordinate offspring. In the present study we examine the outcome of the conflict between the dominants arising from subordinate reproduction, and the effects of that outcome on subordinate reproduction using a three player sequential game theoretical model.

Additionally, we investigate the relevance of order effects in the outcome of the game. In game theoretical models, the order in which players make decisions can have drastic impacts on the outcome of sequential games (204-206). If the moves of earlier players are known to subsequent players, and influence the potential payoffs of their moves, subsequent players should apply this information to their decisions (e.g., 207). Game theoretical models focused on animal behavior rarely discuss the implications of order effects in the outcome of games (208). In many

81 cases individual players make single “sealed bids” which they cannot change in response to their partners’ decisions. Simultaneous sealed bids may not reflect interactions in natural systems, and the order in which individuals actually make decisions may have drastic impacts on the outcome of the game (207, 209). For instance, if a same-sex dominant preemptively prevents subordinate reproduction (e.g. temporary eviction - 109) they are able limit future decisions by other players to those which maximize their own fitness. Similarly, subordinates may be at an advantage depending on the order of individual decisions. For instance, particularly helpful subordinates may be more likely to successfully reproduce in groups after previous interactions with dominants (e.g. the pay-to-reproduce hypothesis - 90). When competitors have knowledge of how other players will respond to their moves, but do not have common interest in the outcome of the game (e.g. Land or sea - 209), individuals may do best by making the final decision in games

(e.g. punishment - 53). However, if there is common interest between players (e.g. related individuals - 210, 211) and individuals can anticipate other players subsequent moves, then it may be better to make decisions earlier in the game (“reactive commitment” - 209, 212).

Using a three-player sequential game theory approach, we investigated how the costs and benefits associated with subordinate reproduction in a cooperatively breeding system influence the behavior of individual group members. We first investigate under what conditions subordinate females attempt to reproduce, dominant females will destroy clutches produced through subordinate reproduction (policing), and when dominant males will defend clutches produced through subordinate reproduction (patrolling). We then describe how the order in which these decisions are made influences the outcome of the game.

82

Parameter Definition Range Examined θ Success rate of "policing" by DF 0 to 1.0 ɸ Success rate of "patrolling" by DM 0 to 1.0

CPOL Cost of policing to DF 0 to 2.0

CPAT Cost of patrolling to DM 0 to 2.0 Probability SF will perform z alloparental care if successfully 0 to 1.0 Benefit of alloparental care to DF A offspring 1.0 to 1.5 M Size of SF clutch relative to that of 0 to 1.0 Competition penalty to DF offspring β if SF reproduces 0 to 1.0 r Relatedness 0 to 1.0

Table 8: Parameters and boundaries of parameter space explored in models.

The Model:

Our game theoretical model consisted of three players, a dominant male, a dominant female, and a subordinate female. Both the dominant and subordinate females were capable of reproduction, and it was assumed the dominant male would father any offspring produced by either female. Subordinate females have a choice whether to attempt to reproduce (lay a clutch of eggs), dominant females have a choice whether to “police” subordinate reproduction (a policing dominant female limits subordinate reproduction by searching for and destroying the subordinate female’s clutch with a probability θ; the effort of policing imposes a cost on dominant females equal to CPOL). Lastly, dominant males have a choice whether to “patrol” the group and defend subordinate female clutches from dominant female policing at a cost equal to CPAT. Patrolling dominant males successfully defend subordinate female clutches against dominant females policing subordinate female reproduction with a probability equal to ɸ.

83

Dominant female reproductive success, if she reproduces without competition from the subordinate female and without alloparental care from the subordinate female, is normalized to 1.

Figure 14: Extensive form depiction of a sequential game in which subordinate females make the first decision, followed by the dominant female, and finally the dominant male.

84

M represents the normalized direct reproductive success gained by the subordinate female if she successfully reproduces in the group. If the subordinate female reproduces, the offspring of the dominant female may face competition for food or shelter. Dominant female reproductive success, under competition with the subordinate female is represented by β, and varies between 0

(no dominant offspring survive) and 1 (all dominant offspring survive) in our model. If β exceeds

1, dominant female offspring do better in the presence of subordinate female offspring, which could result from mechanisms such as dilution effects; however we do not explore this possibility. If a subordinate female does not reproduce, she is assumed to provide alloparental care to the dominant female’s offspring which provides increased dominant direct reproductive success to some value, A. Subordinate females whose clutches are destroyed may also engage in alloparental care; they do so with probability z. Dominant and subordinate females may be related to one another (r), and this can lead to indirect fitness gains for both females. See Table 8 for a summary of all parameters and the range for each which we deemed biologically relevant.

Fitness functions for all three group members are based on gains derived from dominant and subordinate female reproduction minus the costs of patrolling or policing, respectively.

Fitness benefits derived from dominant female reproduction is reduced by 1-β if subordinate females successfully reproduce, which occurs with a probability of (1-θ)+(θ*ɸ), the sum of the probabilities that policing by the dominant female fails, or patrolling by the dominant male successfully counters policing by the dominant female. Subordinate females who attempt to reproduce are successfully policed with a probability of θ(1-ɸ). Dominant males gain fitness benefits from both dominant and subordinate females regardless of relatedness between each female whereas both dominant and subordinate female can only gain fitness benefits from the other females’ reproduction if r>0. See Table 9 for the fitness functions for each player in each possible outcome of the game.

85

In order to explore the possible fitness outcomes we developed a three player game in extended form (Figure 14). In Figure 14, subordinate females make their decision first, followed by the dominant female, who is then followed by the dominant male. Using backwards induction, we first solve for the optimal decision for the final player (the player that decides last), for any combination of earlier-deciding players’ decisions in which it may find itself. Then we move up the next-to-last player, and solve for its optimal decisions for any earlier-deciding players’ decisions, given the expected responses of later players.

Outcome SF strategy DF strategy DM strategy SF fitness payoff vii Don’t Repro. Don't Police Patrol rA ii Don’t Repro. Don't Police Don't Patrol rA viii Don’t Repro. Police Patrol rA-rC POL iv Don’t Repro. Police Don't Patrol rA-rCPOL v Reproduce Don't Police Patrol rβ+M i Reproduce Don't Police Don't Patrol rβ+M vi Reproduce Police Patrol [rzA( θ(1- ɸ) )]+[r(1-z)( θ(1- ɸ) )]+[rβ((1- θ)+( θɸ) )]+[M((1- θ)+(θɸ)) ]-rC POL iii Reproduce Police Don't Patrol [rzA( θ) ]+[r(1-z) θ]+[rβ(1-θ) ]+[M(1- θ) ]-rC POL

Outcome SF strategy DF strategy DM strategy DF fitness payoff vii Don’t Repro. Don't Police Patrol A ii Don’t Repro. Don't Police Don't Patrol A viii Don’t Repro. Police Patrol A -C POL iv Don’t Repro. Police Don't Patrol A-C POL v Reproduce Don't Police Patrol β+rM i Reproduce Don't Police Don't Patrol β+rM vi Reproduce Police Patrol [zA( θ(1- ɸ)) ]+[(1-z)( θ(1- ɸ)) ]+[β((1- θ)+( θɸ)) ]+[rM((1- θ)+(θɸ)) ]-C POL iii Reproduce Police Don't Patrol [zA θ]+[(1-z) θ]+[β(1-θ)]+[rM(1- θ)]-C POL

Outcome SF strategy DF strategy DM strategy DM fitness payoff vii Don’t Repro. Don't Police Patrol A -C PAT ii Don’t Repro. Don't Police Don't Patrol A viii Don’t Repro. Police Patrol A -C PAT iv Don’t Repro. Police Don't Patrol A v Reproduce Don't Police Patrol β+M -C PAT i Reproduce Don't Police Don't Patrol β+M vi Reproduce Police Patrol [zA( θ(1- ɸ)) ]+[(1-z)( θ(1- ɸ)) ]+[β((1- θ)+( θɸ)) ]+[M((1- θ)+(θɸ))]-C PAT iii Reproduce Police Don't Patrol [zA θ]+[(1-z) θ]+[β(1- θ) ]+[M(1- θ)]

Table 9: Fitness outcomes for each of the eight potential outcomes of the game. Lowercase roman numerals correspond to the legend in Figures 15 and 16. 86

In this example, we begin with the four possible decisions made by the dominant male.

So long as CPAT>0, for both decisions (a) and (b) in Figure 14, the dominant male should choose to not patrol, as:

퐴 > 퐴 − 퐶푃퐴푇 (1)

Similarly, for decision (c) in Figure 1, the dominant male should not patrol as:

훽 + 푀 > 훽 + 푀 − 퐶푃퐴푇 (2)

For decision (d) in Figure 1, the dominant male should patrol when:

푧퐴휃(1− 휑) + (1 − 푧)휃(1 − 휑) + 훽((1 − 휃)(휃휑))+ 푀((1 − 휃)(휃휑))− 퐶푃퐴푇 (3) > 푧퐴휃 + (1 − 푧)휃 + 훽(1 − 휃) + 푀(1 − 휃)

In this scenario, patrolling is the best strategy when the average costs of failing to successfully patrol are greater than the costs of patrolling. Patrolling could be favored if the costs of doing so are low (CPAT→0), if dominant males are particularly effective at doing so (ɸ→1), or if subordinate reproduction M is valuable relative to dominant female reproductive losses due to subordinate reproduction (A-β).

퐶푃푂퐿 < 휃휑(훽 + 푀 + 푧 − 훽휃 − 푀휃 − 퐴푧 − 1) + 휃(훽 + 푀) − 훽 − 푀 (4)

Moving up the game tree, dominant females now face the decision whether to police or not police given the expected responses of dominant males. In decision (e), dominant females should choose to not police, as:

퐴 > 퐴 − 퐶푃푂퐿 (5)

87

Whether or not dominant males are patrolling, in decision (f) dominant females’ choice between policing and not policing will depend on whether the costs of allowing the subordinate female to reproduce (β) are less than the benefits of alloparental care (A) after accounting for the possibility of failing to successfully police. Policing should be the optimal strategy when:

퐶푃푂퐿 < 휃(퐴푧 − 푟푀 − 훽 − 푧 + 1) (6)

퐶푃푂퐿 < 휃휑(푟푀+ 훽 + 푧 − 푟푀휃 − 훽휃 − 퐴푧 − 1) + 휃(퐴푧 − 푧 + 1) − 푟푀 − 훽 (7)

depending on whether dominant males are predicted to patrol. Note that this could occur in several ways. First, alloparental care could be extremely valuable to dominant females (A>>1).

Secondly, policing could be highly effective (θ→1) or is inexpensive (CPOL→0).

Lastly, subordinate females are left with the decision whether to attempt to reproduce or not. If dominant males are not patrolling and dominant females are not policing, then this decision depends on whether indirect fitness gains acquired through alloparental care outweigh possible direct fitness gains. Subordinate females should choose to forego reproduction if:

푀 푟 > 퐴 − 훽 (8)

If subordinate females are unrelated to dominant females (r = 0), there are no circumstances in which not reproducing would be favored over reproducing.

We utilized this same approach to investigate all six models representing each possible decision sequence which could be made by the three players in order to identify the impact of the order of these decisions on the outcome of the game. The sequence decisions were made in each model is summarized in Table 10. Equations associated with this model were solved using Maple

88

17.00 (Maplesoft, Waterloo Maple Inc.), and optimal outcomes for each of the six possible trees were determined using MatLab R2014b (The Mathworks Inc.) for various parameters of interest.

Player 1 Player 2 Player 3 Model 1 DM DF SF Model 2 DM SF DF Model 3 DF SF DM Model 4 DF DM SF Model 5 SF DM DF Model 6 SF DF DM

Table 10: Sequence in which individuals made decisions in each of the six games. Player 1 decided first, followed by player 2 and finally player 3.

89

Figure 15: Outcome of the six sequential games plotted against relatedness (r) and the costs of policing

(CPOL). Parameters were set to: A=1.5, β=0.8, θ=0.8, ɸ=0.8, z=0.8, M=0.5, CPAT=0.5. See Table 9 for the decisions made and fitness functions associated with each of these outcomes.

90

Figure 16: Outcome of the six sequential games plotted against relatedness (r) and the costs of patrolling

(CPAT). Parameters were set to: A=1.5, β=0.8, θ=0.8, ɸ=0.8, z=0.8, M=0.9, CPOL=0.05. See Table 9 for the decisions made and fitness functions associated with each of these outcomes.

91

Results:

Outcomes of models regardless of the order of play:

In all six models, under certain parameters the outcome of the game was for subordinate females to reproduce, and for dominant males and females to not patrol or police (respectively) (zone i –

Figures 15, 16). This zone occupied the space within which:

푀 푟 < (9) 퐴 − 훽

퐶푃푂퐿 > 휃(퐴푧 − 푟푀− 훽 − 푧 + 1) (10)

Generally, this zone depended on the subordinate females benefitting more from direct fitness gains than they did from indirect fitness gains through the dominant female, and on the costs of policing subordinate reproduction (CPOL) for the dominant female exceeding the net losses of subordinate females breeding rather than providing alloparental care.

A second zone which emerged regardless of the order in which decisions were made was characterized by subordinate females not reproducing, and dominant males and females not patrolling or policing subordinate behavior (respectively) (zone ii – Figure 15). This zone occupied the parameter space within which:

푀 푟 > 퐴 − 훽 (11)

퐶푃푂퐿 > (훽 + 푟푀)(1− 휃) + 휃(1 − 푧) − 퐴(1 − 휃푧) (12)

Generally, this zone depended on the indirect fitness gains subordinate females could acquire through alloparental care exceeded the direct fitness gains they could achieve through reproducing. 92

A third zone which emerged regardless of the order in which decisions were made was characterized by subordinate female reproduction, dominant female policing subordinate reproduction, and dominant males not patrolling groups (zone iii – Figures 15, 16). This zone occupied the parameter space in which:

퐶푃푂퐿 > (훽 + 푟푀)(1− 휃) + 휃(1 − 푧) − 퐴(1 − 휃푧) (13)

Generally, this zone depended on the dominant females’ net losses associated with policing subordinate females attempting to reproduce being less than the net losses of allowing subordinate females to reproduce.

When subordinate females made decisions before dominant females:

In the three models in which subordinate female decisions preceded those of dominant females, zone iii also was limited to the parameter space in which subordinate females would choose to reproduce in spite of dominant females policing subordinate behavior. At this threshold, subordinate females would attempt to reproduce if:

푀(−1 + 휃) 푟 < 휃(퐴푧 − 훽 − 푧 + 1) − 퐴 + 훽 (14)

In these three models, subordinate females would attempt to reproduce so long as direct fitness gains of reproducing in spite of policing outweighed indirect fitness gains acquired through providing alloparental care to dominant females. In these three models, if subordinate females had already decided not to reproduce, there was no reason for dominant females to expend the effort to police. When subordinate females did reproduce, dominant females would police when the costs of policing exceeded those which characterized zone i outlined above.

93

When dominant females made decisions before subordinate females:

When dominant female decisions preceded those of subordinate females, a fourth zone emerged in which subordinate females policed subordinate reproduction, subordinate females did not attempt to reproduce, and dominant males did not patrol groups (zone iv – Figure 15). This zone occurred in the parameter space in which:

퐶푃푂퐿 < 퐴 − 훽 + 푟푀 (15) In these three models, when dominant females had already committed to policing, subordinate females attempted to reproduce at the same threshold as subordinate females did in the models in which they acted first (Equation 14).

When dominant males made decisions before dominant females:

Throughout the vast majority of the parameter space which we examined in this model, dominant males’ optimal strategy was to not patrol the group regardless of the order in which players made their decisions. However, in games in which dominant males made their decision before dominant females, subordinate clutches were close in size to those of dominant females

(M→1), and both policing and patrolling were inexpensive (CPOL→0, CPAT→0) dominant males would patrol groups. In this zone subordinate females would reproduce, and dominant females would not police (Figure 16 – zone v). Dominant male patrolling was their optimal strategy when:

퐶푃퐴푇 < 휃(훽 + 푀 + 푧 − 퐴푧 − 1) (16)

Within this range it was not favorable for dominant females to police when:

퐶푃푂퐿 > 휃휑(푟푀+ 훽 − 푟푀휃 − 퐴푧 − 푧 − 1) + 휃(퐴푧 − 훽 − 푧 + 1) − 푀 (17)

94

Below this threshold however, when policing subordinate reproduction was so inexpensive that the net gains of policing despite patrolling dominant males exceeded those associated with allowing subordinate females to reproduce uninhibited in the group. In this area of parameter space (Figure 16, zone vi) subordinate females reproduced, and dominant females policed despite dominant males having already decided to patrol.

Discussion:

In general, each threshold at which subordinate females’ and dominant females’ optimal strategy was to reproduce rather than not reproduce or police subordinate reproduction rather than not do so (respectively) occurred when:

푁푒푡 푑푖푟푒푐푡 푓푖푡푛푒푠푠 푔푎푖푛푠 푟 < 푁푒푡 푖푛푑푖푟푒푐푡 푓푖푡푛푒푠푠 푔푎푖푛푠 (18)

When one individual’s decision followed that of another, the latter individual had more information on which to base its decision, which in some cases changed where this threshold occurred. When dominant females knew whether subordinate females had reproduced, their threshold at which to begin policing subordinate reproduction matched the threshold at which subordinate females actually reproduced. Similarly, when a dominant female’s decision to police subordinate reproduction preceded the subordinate female’s actual decision to reproduce, subordinate females could then reproduce without the threat of policing in parameter space they otherwise would not attempt to do so.

Hirshleifer (209) describes “reactive commitments” in which players commit to specific strategies in response to previous players decisions. In this framework, individuals playing earlier in the game are at an advantage because subsequent players are forced to play in response to what

95 other players have already committed to. In our game individuals making initial decisions could determine their optimal strategy based on perfect knowledge of how subsequent players would respond to their decision. Whether dominant or subordinate females made their decision first or second left the individual that plays earlier in a better position to maximize their own fitness. For example, when subordinate individuals decided first, they knew whether dominant females would police or not if they did, or did not reproduce. With this information regarding the dominant females “reactive commitment”, subordinate females could select the option that resulted in the greatest fitness gains.

The advantage of making earlier decisions depends on some common interest between players (e.g. Battle of the sexes - 209). In our game, dominant and subordinate females have at least some common interest as long as r>0. When r=0 in our game, dominant and subordinate females’ interests are entirely antagonistic and subordinate females optimal decision was always to reproduce regardless of the behavior of the dominant female, and dominant females would do best to police subordinate reproduction so long as the costs of policing are less than the expected losses of not doing so.

Dominant males’ decisions to patrol groups and protect threatened subordinate clutches only occurred in our model when dominant males’ decisions preceded dominant female’s decisions, and only then when:

퐶푃퐴푇 < 푁푒푡 푓푖푡푛푒푠푠 푔푎푖푛푠 푣푖푎 푆퐹 − 푁푒푡 푓푖푡푛푒푠푠 푙표푠푠푒푠 푣푖푎 퐷퐹 (19)

Subordinate female clutches had to be large in order for dominant males to invest in patrolling behavior. Under no circumstances would dominant males choose to patrol if dominant females had already decided to police. As was the case in interactions between the dominant and subordinate female when r>0, making the first decision was advantageous for whichever player

96 moved first. One fascinating result of our model was that while a dominant male’s decision to patrol never influenced a subordinate female’s decision to reproduce, by deterring the dominant female from policing he did influence the amount of fitness a subordinate female could expect from reproducing. Thus dominant male behavior influenced the degree of reproductive skew amongst females in the group.

Male participation in within-group interactions varies widely across taxa, but there is some evidence of males intervening in the conflict of female group members, presumably for their own benefit. In the shell brooding cichlid fishes Neolamprologus multifasciatus and

Lamprologus ocellatus males intervene in conflict between resident and potential joiner females

(213, 214). In the cooperatively breeding cichlid fish Neolamprologus pulcher, dominants do intervene in conflicts between other group members (66, personal observation). Our model suggests that dominant males should only patrol groups under specific conditions in which policing and patrolling are minimally costly, and subordinate clutches are especially valuable.

Cooperatively breeding species may rarely experience these conditions in natural settings, as groups face persistent threats from conspecific intruders, heterospecific competitors, and predators which may have greater fitness impacts than extra pair reproduction within the group

(22-24). Additionally, in many species subordinates are physiologically less fecund (85, 215).

This finding of our model may explain why males do frequently intervene in the conflict between females in polygynous species such as N. multifasciatus and L. ocellatus. In these species males maintain small territories containing closely grouped shells which females compete over. While females face clear costs from other females breeding (216), male’s fitness may improve greatly with the addition of more mates, particularly if they can economically intervene in female-female conflict.

97

Subordinates in many cooperatively breeding systems are physiologically or behaviorally inhibited from reproducing (89, 121, 127, 128, 215, 217). Physiological inhibition in particular may serve as an honest signal of a decision to not reproduce (217). In most cases dominants can presumably detect subordinate reproduction relatively easily. Dominants’ decisions to police or patrol, particularly if they precede subordinate’s decision to reproduce may be more difficult to detect. Subordinates may be reproductively inhibited through pre-emptive aggression (109) or subsequent punishment (53) or eviction (80). In meerkats, Suricata suricatta, dominant females will temporarily evict large subordinate females from the group in order to prevent them from reproducing in the group (109). Mechanisms such as this may serve as signals of a dominant’s decision to police subordinate reproduction or patrol groups. In most cooperatively breeding systems, groups consist of more than one subordinate group member. If so, subordinates may also gain information about whether dominants should be expected to police or patrol through observing the experiences of their fellow group members.

Empirical and theoretical work suggests subordinates may forego reproduction in order to reproduce in the group later in life (90, 218). Our model did not consider the impact of previous or future interactions, which likely influence both dominant and subordinate decisions. If a tradeoff exists between reproducing and growth (and presumably being more fecund later in life) it may benefit subordinates to forego reproduction even if dominants would not resist it. Our model also did not consider the possibility of dispersal or eviction, both of which are central to many models of reproductive skew (105, 118) and may lower the threshold at which subordinates are permitted to reproduce in groups.

Our models reflect both the concession of reproductive opportunities by dominants and the possibility of dominants having limited control over subordinate reproduction (103) in establishing the degree of reproductive skew in groups. Further, our model indicates that the order

98 in which group members make decisions influences the groups eventual “consensus decision”

(219) regarding reproductive skew. In groups in which individual member’s optimal outcomes are inversely related, committing to and making other players aware of ones strategy before competing players can do so is advantageous. This advantage may explain the evolution of preemptive signals of subsequent behavior such as aggressive displays or threats to maintain stability in animal groups (63, 79). Our results are consistent with the concept that reproductive skew is the product of a tug-of-war or transaction between group members (85, 207, 220, 221) which third parties can influence in some cases. Further, our model shows that the resultant distribution of reproductive opportunities within groups reflects not only the costs and benefits associated with individual behavior, but whether group members have the advantage of anticipating the subsequent decisions of other group members.

99

Chapter 7: Conclusion

The social environments is an emergent characteristic of groups which is influenced by the direct and third-party (or higher order) interactions of individual group members.

Through these pathways, group members may shift the social environment for themselves and other group members. What constitutes an “optimal” social environment may differ among group members and certain group members may be better-able to shift the social environment to reflect their own interests. The potential for such differences is most clearly represented in the results I describe in Chapter 5, in which individuals of different social status varied widely in their rate of aggression towards potential group joiners, and also shifted their rates of aggression in conflicting ways when the group’s makeup had been experimentally manipulated. These results suggest that dominant group members benefit from restoring group size after the removal of a subordinate, while subordinate group members benefit from their promotion in social rank after a removal.

Direct interactions within dyads can vary between groups. These direct interactions may influence individual participation in other behaviors. In Chapter 2, I found that female residents adjust their participation in defense against intruders depending on the asymmetry in power between the resident male and female. Less- powerful females (those much smaller than their large mates) performed significantly more defensive behaviors than more-powerful females (those similar in size to their large 100 mates). Individuals may also be able to influence the nature of their dyadic interactions by adjusting their own behavior. The results of Chapter 2 show that residents receive more aggression from their mates when they participate less in territory defense. In

Chapter 3, subordinate females who were more affiliative and submissive towards dominant females also received less aggression from them, and the results of Chapter 4 indicate that subordinate females who frequently performed affiliative behaviors towards dominant females had lower cortisol levels.

The impact of direct interactions was also reflected in the outcome of game theoretical models (Chapter 6). In these models, the expected response of other group members influenced the individual behavior of group members. When dominant females were expected to destroy the clutches produced by subordinates (“policing”), subordinate females at times adjusted under what circumstances they would attempt to reproduce.

Similarly, if subordinate females were expected to reproduce, dominant females could adjust under what circumstances they would put in the effort to police. Where individual’s optimal social environment differ (but individuals do have some common interest regardless of its outcome), this model shows that making decisions earlier in the game allows individuals to maximize their fitness.

Individual behavior and dyadic interactions can also have indirect effects on other group members. As with direct dyadic interactions, these third party effects may best reflect the interests of more-powerful individuals. These third-party effects were the focus of Chapter 3, which determined that subordinates have different influences on dominant interactions depending on subordinate sex. Subordinates of both sexes faced

101 more aggression from same-sex dominants when the size asymmetry between dominant males and females was large. Subordinate males who faced more aggression from dominant males also behaved more submissively towards those males. In contrast, more- submissive and affiliative subordinate females faced less aggression from dominant females when size differences between dominants were relatively large, but received more aggression from dominant females and were less submissive and affiliative when the size difference between dominants was smaller.

Third party effects were also evident in the outcome of the game theoretical model in Chapter 6. While dominant males in the model rarely defended subordinate clutches from policing by dominant females (“patrolling”), when they did so, it did not influence a subordinate female’s decision to attempt to reproduce. A dominant male’s decision to police did influence the amount of direct fitness subordinates could expect to acquire however, as dominant females would often not police if dominant males had already decided to patrol.

Members of groups are presumed to benefit from living socially, however these benefits may not be equal for all group members. Each group member’s net gains from being a part of a group is influenced by the social environment in the group, which is an emergent property of the group resulting from the behavior, interactions, and indirect influences of the other group members. Asymmetries in the power of dominants have implications on territory defense as well as the behavior and dyadic interactions between subordinates and same-sex dominants. Each of these may influence group productivity and subordinate decisions regarding reproduction and group membership. The conflicting

102 interests of group members may prohibit or facilitate the recruitment of additional group members, and the net costs or benefits of recruits for current group members. Thus, the social environment in groups not only impacts individuals currently living in groups, but also the offspring of those individuals and the persistence of the group as a whole.

103

References

1. Foster WA, Treherne JE. 1981. Evidence of the dilution effect in the selfish herd from fish predation on a marine insect. Nature. 293:466-467.

2. Roberts G. 1996. Why individual vigilance declines as group size increases. Animal Behavior, 51:1077-1086.

3. Krause J, Ruxton GD. 2002. Living in groups. Oxford, UK: Oxford University Press.

4. Cresswell W, Quinn JL. 2011. Predicting the optimal prey group size from predator hunting behaviour. Journal of Animal Ecology, 80:310-319.

5. Ioannou CC, Bartumeus F, Krause J, Ruxton GD. 2011. Unified effects of aggregation reveal larger prey groups take longer to find. Proceedings of the Royal Society B, 278:2985-2990.

6. Berger J. 1978. Group size, foraging, and antipredator ploys: an analysis of bighorn sheep decisions. Behavioral Ecology and Sociobiology, 4:91-99.

7. Pitcher TJ, Magurran AE, Winfield IJ. 1982. Fish in larger shoals find food faster. Behavioral Ecology and Sociobiology, 10:149-151.

8. Gartska WR, Camazine B, Crews D. 1982. Interactions of behavior and physiology during the annual reproductive cycle of the red-sided garter snake (Thamnophis sirtalis parietalis). Herpetologica, 38:104-123.

104

9. Krause J, Godin JGJ. 1995. Predator preferences for attacking particular prey group sizes: consequences for predator hunting success and prey predation risk. Animal Behavior, 50:465-473.

10. Skogland T. 1983. The effects of density dependent resource limitation on size of wild reindeer. Oecologia, 60:156-168.

11. Thirgood SJ. 1990. Alternative mating strategies and reproductive success in fallow deer. Behaviour, 116:1-10.

12. Hodge SJ, Bell MBV, Mwanguhya F, Kyabulima S, Waldick RC, Russell AF. 2009. Maternal weight, offspring competitive ability, and the evolution of communal breeding. Behavioral Ecology. 20:729-735.

13. Beaugrand J, Boulet C, Payette D. 1991. Outcome of dyadic conflict in male green swordtail fish, Xiphophorus helleri: effects of body size and prior dominance. Animal Behavior. 41:417-424.

14. Buston PM. 2003. Social hierarchies: size and growth modification in clownfish. Nature, 424:145-146.

15. Hamilton IM, Heg D, Bender N. 2005. Size differences within a dominance hierarchy influence conflict and help in a cooperatively breeding cichlid. Behaviour, 142:1591- 1613.

16. Kutsukake N, Castles DL. 2004. Reconciliation and post-conflict third-party affiliation among wild chimpanzees in the Mahale Mountains, Tanzania. Primates, 45:157-165.

105

17. Kutsukake N. 2007. Conspecific influences on vigilance behavior in wild chimpanzees. Int. J. Primatol. 28:907-918.

18. Flack JC, Krakauer DC, de Waal FBM. 2005. Robustness mechanisms in primate societies: a perturbation study. Proceedings of the Royal Society B, 272:1091-1099.

19. Flack JC, Girvan M, de Waal FBM, Krakauer DC. 2006. Policing stabilizes construction of social niches in primates. Nature, 439:426-429.

20. Beekman M, Komdeur J, Ratnieks FLW. 2003. Reproductive conflicts in social animals: who has power? Trends in Ecology and Evolution, 18:277-282.

21. Flack JC, Krakauer DC. 2006. Encoding power in communication networks. The American Naturalist, 168:E87-E102.

22. Choe JC, Crespi BJ, eds. 1997. The evolution of social behaviour in insects and arachnids. Cambridge University Press.

23. Koenig WD, Dickenson JL, eds. 2004. Ecology and evolution of cooperative breeding in birds. New York: Cambridge University Press.

24. Solomon NG, French JA. 1997. Cooperative breeding in mammals. New York: Cambridge University Press.

25. Taborsky M. 1984. Broodcare helpers in the cichlid fish brichardi – their costs and benefits. Animal Behavior, 32:1236-1252.

26. Wong MYL, Balshine S. 2011. The evolution of cooperative breeding in the African cichlid fish, Neolamprologus pulcher. Biology Reviews, 86:511-530.

106

27. Mitchell JS, Jutzeler E, Heg D, Taborsky M. 2009. Dominant members of cooperatively-breeding groups adjust their behaviour in response to the sexes of their subordinate. Behaviour, 146:1665-1686.

28. Mitchell JS, Jutzeler E, Heg D, Taborsky M. 2009. Gender differences in the costs that subordinate group members impose on dominant males in a cooperative breeder. Ethology, 99:139-149.

29. Desjardins JK, Stiver KS, Fitzpatrick JL. Balshine S. 2008. Differential responses to territory intrusions in cooperatively breeding fish. Animal Behaviour, 75:595-604.

30. Hadley ME. 1996. Endocrinology. Prentice Hall, Upper Saddle River, NJ.

31. Williams GC. 1966. Natural selection, the costs of reproduction, and a refinement of Lack’s Principle. The American Naturalist, 100:687-690.

32. Sargent, RC, Gross MR. 1985. Parental investment decision rules and the Concorde fallacy. Behavioral Ecology and Sociobiology, 17:43-45.

33. Balshine-Earn S. 1995. The costs of parental care in Galilee St Peter’s fish, Sarotherodon galilaeus. Animal Behaviour, 50, 1-7.

34. Wedell N, Kvarnemo C, Lessells CM, Tregenza T. 2006. Sexual conflict and life histories. Animal Behaviour, 71:999-1011.

35. Grantner A, Taborsky M. 1998. The metabolic rates associated with resting, and with the performace of agonistic, submissive and digging behaviours in the cichlid fish

107

Neolamprologus pulcher (Pisces: Cichlidae). Journal of Comparative Physiology B, 168:427-433.

36. Dale S, Gustaven R, Slagvold T. 1996. Risk taking during parental care: a test of three hypotheses applied to the pied flycatcher. Behavioral Ecology and Sociobiology, 39:31-42.

37. Ghalambor CK, Martin TE. 2001. Fecundity-survival trade-offs and parental risk- taking in birds. Science, 292:494-497.

38. Trivers RL. 1972. Parental investment and sexual selection. In Sexual Selection and the Descent of Man, B. Campbell, ed. Heinemann, London.

39. Clutton-Brock T. 1991. The evolution of parental care. Princeton University Press. Princeton, NJ.

40. Maynard Smith J. 1977. Parental Investment: A prospective analysis. Animal Behaviour, 25:1-9.

41. Bruintjes R, Taborsky M. 2011. Size-dependent task specialization in a cooperative cichlid in response to experimental variation of demand. Animal Behaviour, 81:387-394.

42. Komdeur J. 2006. Variation in individual investment strategies among social animals. Ethology, 112:729-747.

43. Kendeigh CS. 1952. Parental care and its evolution in birds. Illinois biological monographs; v. 22, nos. 1-3. University of Illinois Press, Urbana.

108

45. Itzkowitz M, Santangelo N, Richter M. 2001. Parental division of labour and the shift from minimal to maximal role specializations: an examination using a biparental fish. Animal Behaviour, 61:1237-1245.

46. McKaye KR, Murry BA. 2008. Sex role differentiation in brood defense by a Nicaraguan cichlid fish, Amphilophus xiloanensis. Caribbean Journal of Science, 44:13- 20.

47. Keenleyside MHA. 1989. Mate desertion in relation to adult sex ratio in the biparental cichlid fish Herotilapia multispinosa. Animal Behaviour, 31:683-688.

48. van Dijk RE, Székely T, Komdeur J, Pogány Á, Fawcett TW, Weissing FJ. 2011. Individual variation and the resolution of conflict over parental care in penduline tits. Proceedings of the Royal Society B, rspb20112297.

49. Itzkowitz M. 2005. Is the selection of sex-typical parental roles based on an assessment process? A test in the monogamous convict cichlid fish. Animal Behaviour, 69:95-105.

50. Hunt J, Simmons LW. 2002. Behavioural dynamics of biparental care in the dung beetle Onthophagus taurus. Animal Behaviour, 64:65-75.

51. Hinde CA. 2006. Negotiation over offspring care? – a positive response to partner- provisioning rate in great tits. Behavioral Ecology, 17:6-12.

52. Johnstone RA, Hinde CA. 2006. Negotiation over offspring care – how should parents respond to each other’s efforts? Behavioral Ecology, 17:818-827.

109

53. Clutton-Brock TH, Parker GA. 1995. Punishment in animal societies. Nature. 373:209-216.

54. Bergmüller R, Johnstone RA, Russell AF, Bshary R. 2007. Integrating cooperative breeding into theoretical concepts of cooperation. Behavioural Processes, 76:61-72.

55. Houston, AI, Davies NB. 1985. The evolution of cooperation and life history in the dunnock, Prunella modularis. RM Sibly, RH Smith Eds. Behavioural Ecology, Blackwell Scientific, Oxford.

56. Harrison F, Barta Z, Cuthill I, Székely T. 2009. How is sexual conflict over parental care resolved? A meta-analysis. Journal of Evolutionary Biology, 22:1800-1812.

57. Barta Z, Houston AI, McNamara JM, Szekely T. 2002. Sexual conflict about parental care: the role of reserves. The American Naturalist, 159:687-705.

58. Houston AI, Székely T, McNamara JM. 2005. Conflict between parents over care. Trends in Ecology and Evolution, 20:33-38.

59. Jones KM, Ruxton GD, Monaghan P. 2002. Model parents: is full compensation for reduced partner nest attendance compatible with stable biparental care. Behavioral Ecology, 13:838-843.

60. Yuma M, Narita T, Hori M, Kondo T. Food resources of shrimp-eating cichlid fishes in Lake Tanganyika. Fish biology in Japan: an anthology in honour of Hiroya Kawanabe, 371-378.

110

61. Balshine-Earn S, Neat F, Reid H, Taborsky M. 1998. Paying to stay or paying to breed? Field evidence for direct benefits of helping behavior in a cooperatively breeding fish. Behavioral Ecology, 9:432-438.

62. Reddon AR, Voisin MR, Menon N, Marsh-Rollo SE, Wong MYL, Balshine S. 2011. Rules of engagement for resource contests in a social fish. Animal Behaviour, 82:93-99.

63. Wong MYL, Buston PM, Munday PL, Jones GP. 2007. The threat of punishment enforces peaceful cooperation and stabilizes queues in a coral-reef fish. Proceedings of the Royal Society of London B, 274:1093-1099.

64. Balshine S, Leach B, Neat FC, Reid H, Taborsky M, Werner N. 2001. Correlates of group size in a cooperatively breeding cichlid fish (Neolamprologus pulcher). Behavioral Ecology and Sociobiology, 50:134-140.

65. Heg D, Brouwer L, Bachar Z, Taborsky M. 2005. Large group size yields group stability in the cooperatively breeding cichlid Neolamprologus pulcher. Behaviour, 142:1615-1641.

66. Dey CJ, Reddon AR, O’Connor CM, Balshine S. 2013. Network structure is related to social conflict in a cooperatively breeding fish. Animal Behaviour, 85:395-402.

67. Itzkowitz M. 1990. Heterospecific intruders, territorial defense and reproductive success in the beaugregory damselfish. Journal of Experimental Marine Biology and Ecology, 140:49-59.

68. Haley MP, Müller CR. 2002. Territorial behaviour of beaugregory damselfish (Stegastes leucostictus) in response to egg predators. Journal of Experimental Marine Biology and Ecology, 273:151-159.

111

69. Garvy KA, Hellmann JK, Ligocki IY, Reddon AR, Marsh-Rollo SE, Hamilton IM, Balshine S, O’Connor CM. 2014. Sex and social status affect territorial defence in a cooperatively breeding cichlid fish, Neolamprologus savoryi. Hydrobiologia, doi:10.1007/s10750-014-1899-0

70. Hellmann JK, Hamilton IM. 2014. The presence of neighbors influences defense against predators in a cooperatively breeding cichlid. Behavioral Ecology, doi:10.1093/beheco/aru001

71. Coeckelberghs V. 1975. Territorial, spawning and parental behaviour of Lamprologus brichardi Poll 1974 (Pisces, Cichlidae). Annales de la Société Royale Zoologique de Belgique, 105, 73-86.

72. Kalas K. 1975. Zur Ethologie von Lamprologus brichardi (Pisces, Cichlidae) Trewavas und Poll 1952 unter besonderer Berücksichtigung des Sozialverhaltens. Diplomarbeit, Zoological Institute of the University of Giessen, West Germany.

73. Taborsky M. 1982. Brutpflegehelfer beim Cichliden Lamprologus brichardi, Poll (1974): einen Kosten/Nutzen-Analyse. – PhD Dissertation, University of Vienna.

74. Blumstein,DT, Evans CS, Daniel JC. 2000. JWatcher 0.9. An Introductory User’s Guide. Animal Behaviour Laboratory, Macquarie University. http://www.jwatcher.ucla.edu.

75. Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society B, 57:289- 300.

112

76. Narum SR. 2006. Beyond Bonferroni: less conservative analyses for conservation genetics. Conservation genetics, 7:783-787.

77. Hamilton IM, Ligocki IY. 2012. The extended personality: indirect effects of behavioral syndromes on the behavior of others in a group-living cichlid. Animal Behaviour, 84, 659-664.

78. Eggert AK, Reinking M, Müller JK. 1998. Parental care improves offspring survival and growth in burying beetles. Animal Behavior, 55:97-107.

79. Cant MA. 2011. The role of threats in animal communication. Proceedings of the Royal Society B, 278:170-178.

80. Cant MA, Hodge SJ, Bell MBV, Gilchrist JS, Nichols HJ. 2010. Reproductive control via eviction (but not the threat of eviction) in banded mongooses. Proceedings of the Royal Society B, doi: rspb20092097.

81. Stiver KA, Dierkes P, Taborsky M, Balshine S. 2004. Dispersal patterns and status change in a co-operatively breeding cichlid Neolamprologus pulcher: evidence from microsatellite analyses and behavioral observations. Journal of Fish Biology, 65:91-105.

82. Stiver KA, Desjardins JK, Fitzpatrick JL, Neff B, Quinn JS, Balshine S. 2007. Evidence for size and sex-specific dispersal in a cooperatively breeding cichlid fish. Molecular Ecology, 16:2974-2984.

83. Limberger D. 1983. Pairs and harems in a cichlid fish Lamprologus brichardi. Z. Tierphychol. 62:115-144.

113

84. Desjardins JK, Fitzpatrick JL, Stiver KA, Van Der Kraak GJ, Balshine S. 2008. Costs and benefits of polygyny in the cichlid Neolamprologus pulcher. Animal Behaviour, 75:1771-1779.

85. Heg D, Hamilton IM. 2008. Tug-of-war over reproduction in a cooperatively breeding cichlid. Behavioral Ecology and Sociobiology, 62:1249-1257.

86. Siemens MV. 1990. Broodcare or egg cannibalism by parents and helpers in Neolamprologus brichardi (Poll 1986) (Pisces: Cichlidae): a study on behavioural mechanisms. Ethology, 84:60-80.

87. Kokko H, Johnstone RA, Clutton-Brock TH. 2001. The evolution of cooperative breeding through group augmentation. Proceedings of the Royal Society in London B, 268:187-196.

88. Schaffner CM, French JA. 1997. Group size and aggression: ‘recruitment incentives’ in a cooperatively breeding primate. Animal Behaviour, 54:171-180.

89. Heg D, Jutzeler E, Bonfils D, Mitchell JS. 2008. Group composition affects male reproductive partitioning in a cooperatively breeding cichlid. Molecular Ecology, 17:4359-4370.

90. Heg D, Jutzeler E, Mitchell JS, Hamilton IM. 2009. Helpful female subordinate cichlids are more likely to reproduce. Plos One, 4:e5458.

91. Boydston EE, Morelli TL, Holekamp KE. 2001. Sex differences in territorial behavior exhibited by the spotted hyena (Hyaenidae, Crocuta crocuta) Ethology. 107:369-385.

114

92. Iwata, E, Manbo J. 2013. Territorial behaviour reflects sexual status in groups of false clown anemonefish (Amphiprion ocellaris) under laboratory conditions. Acta Ethologica, 16:97-103.

93. Stacey PB. 1986. Group size and foraging efficiency in yellow baboons. Behavioral Ecology and Sociobiology, 18:175-187.

94. Pitcher TJ, Parrish JK. 1993. Functions of shoaling behaviour in teleosts, in Pitcher TJ, ed. Behaviour of Teleost Fishes, 363-439.

95. Heg D, Bender N, Hamilton IM. 2004. Strategic growth decisions in helper cichlids. Proceedings of the Royal Society B, 27: S505-S508

96. Horn HS. 1968. The adaptive significance of colonial nesting in the Brewer’s blackbird (Euphagus cyanocephalus). Ecology, 49:682-694.

97. Creel S, Creel NM. 1995. Communal hunting and pack size in African wild dogs, Lycaon pictus. Animal Behavior, 50:1325-1339.

98. Creel S, Creel NM, Mills MGL, Monfort SL. 1997. Rank and reproduction in cooperatively breeding African wild dogs: behavioral and endocrine correlates. Behavioral Ecology, 8:298-306.

99. Johnstone RA. 2000. Models of reproductive skew: A review and synthesis. Ethology, 106:5-26.

100. Reyer HU. 1984. Investment and relatedness: a cost/benefit analysis of breeding and helping in the pied kingfisher (Ceryle rudis). Animal Behavior, 32:1163-1178.

115

101. Zöttl M, Frommen JG, Taborsky M. 2013. Group size adjustment to ecological demand in a cooperative breeder. Proceedings of the Royal Society B, 280:1-9.

102. Stiver., KA, Fitzpatrick JL, Desjardins JK, Balshine S. 2009. Mixed parentage in Neolamprologus pulcher groups. Journal of Fish Biology, 74:1129-1135.

103. Clutton-Brock TH. 1998. Reproductive skew, concessions and limited control. TREE, 13:288-292.

104. Reeve HK, Keller L. 2001. Tests of reproductive-skew models in social insects. Annu. Rev. Entomol. 46:347-385.

105. Hamilton IM, Heg D. 2007. Clutch-size adjustments and skew models: effects on reproductive partitioning and group stability. Behavioral Ecology, 18:467-476.

106. Brouwer L, Heg D, Taborsky M. 2005. Experimental evidence for helper effects in a cooperatively breeding cichlid. Behavioral Ecology, 16:667-673.

107. Russell AF, Langmore NE, Cockburn A, Astheimer LB, Kilner RM. 2007. Reduced egg investment can conceal helper effects in cooperatively breeding birds. Science, 317:941-944.

108. Hellmann JK, Ligocki IY, O’Connor CM, Reddon AR, Garvy KA, Marsh-Rollo SE, Gibbs HL, Balshine S, Hamilton IM. In revision. The effects of within group and within colony traits on parentage loss in a cooperatively breeding cichlid fish.

109. Young AJ, Carlson AA, Monfort SL, Russell AF, Bennett NC, Clutton-Brock T. 2006. Stress and the suppression of subordinate reproduction in cooperatively breeding meerkats. Proceedings of the National Academy of Sciences, 103:12005-12010.

116

110. Fitzpatrick JL, Desjardins JK, Stiver KA, Montgomerie R, Balshine S. 2008. Male reproductive suppression in the cooperatively breeding fish Neolamprologus pulcher. Behavioral Ecology, 17:25-33.

111. Hamilton,, IM, Heg D. 2008. Sex differences in the effect of social status on the growth of subordinates in a co-operatively breeding cichlid. Journal of Fish Biology, 72:1079-1088.

112. Awata S, Kohda M. 2004. Parental roles and the amount of care in a bi-parental substrate brooding cichlid: the effect of size differences within pairs. Behaviour, 141:1135-1149.

113. Ligocki IY, Earley RL, Hellmann JK, Hamilton IM. In revision. Variation in glucocorticoid levels in relation to direct and third-party interactions in a social cichlid fish.

114. Pankhurst NW, Conroy AM. 1987. Size-fecundity relationships in the orange roughy, Hoplostethus atlanticus. 21:295-299.

115. Beacham TD, Murray CB. 1993. Fecundity and egg size variation in North American Pacific salmon (Oncorhynchus) Journal of Fish Biology, 42:485-508.

116. Yasui Y. 1998. The ‘genetic benefits’ of female multiple mating reconsidered. TREE, 13:246-250.

117. Rubenstein DR. 2007. Female extrapair mate choice in a cooperative breeder: trading sex for help and increasing offspring heterozygosity. Proceedings of the Royal Society B, 274:1895-1903.

117

118. Johnstone RA, Cant MA. 1999. Reproductive skew and the threat of eviction: a new perspective. Proceedings of the Royal Society B, 266: 275-279.

119. Fischer S, Zöttl M, Groenewoud F, Taborsky B. 2014. Group-size-dependent punishment of idle subordinates in a cooperative breeder where helpers pay to stay. Proceedings of the Royal Society B, 281:1-9.

120. Hacklӓnder K, Möstl E, Arnold W. 2003. Reproductive suppression in female Alpine marmots, Marmota marmota. Animal Behaviour, 65:1133-1140.

121. Bender N, Heg D, Hamilton IM, Bachar Z, Taborsky M, Oliveira RF. 2006. The relationship between social status, behaviour, growth and steroids in male helpers and breeders of a cooperatively breeding cichlid. Hormones and Behavior, 50:173-182.

122. Kutsukake N. 2009. Complexity, dynamics and diversity of sociality in group-living mammals. Ecol. Res, 24:521-531.

123. Sapolsky RM, 1992. Neuroendocrinology of the stress resonse, in: Becker JB, Marc Breedlove S, Crews D, eds. Behavioral endocrinology. MIT press, 287-324.

124. Klausner H, Heimberg M. 1967. Effect of adrenalcortical hormones on release of triglycerides and glucose by liver. Am. J. Physiology 212:1236-1246.

125. Rotllant J, Tort L, 1997. Cortisol and glucose responses after acute stress by net handling in the sparid red porgy previously subjected to crowding stress. J. Fish Biol. 51:21-28.

118

126. Wingfield JC, Hegner RE, Lewis DM. 1991. Circulating levels of luteinizing hormone and steroid hormones in relation to social status in the cooperatively breeding white-browed sparrow weaver, Plocepasser mahali. J. Zool. 225:43-58.

127. Schoech SJ, Mumme RL, Moore MC. 1991. Reproductive endocrinology and mechanisms of breeding inhibition in cooperatively breeding Florida scrub jays (Aphelocoma c. coerulescens). Condor 354-364.

128. Creel S, Creel N, Wildt DE, Monfort SL, 1992. Behavioural and endocrine mechanisms of reproductive suppression in Serenge dwarf mongooses. Anim. Behav. 43:231-245.

129. Abbott DH, Keverne EB, Bercovitch FB, Shively CA, Mendoza SP, Saltzman W, Snowdon CT. Ziegler TE, Banjevic M, Garland Jr. T, Sapolsky RM. 2003. Are subordinates always stressed? A comparative analysis of rank differences in cortisol levels among primates. Horm. Behav. 43:67-82.

130. Creel S, 2001. Social dominance and stress hormones. TREE. 16:491-497.

131. Goymann W, Wingfield JC. 2004. Allostatic load, social status and stress hormones: the costs of social status matter. Anim. Behav. 67:591-602.

132. Bronson FH, Eleftherious BE. 1964. Chronic physiological effects of fighting in mice. Gen. Comp. Endocrinol. 4:9-14.

133. McLeod PJ, Moger WH, Ryon J, Gadbois S, Fentress JC. 1996. The relation between urinary cortisol levels and social behaviour in captive timber wolves. Can. J. Zool. 74:209-216.

119

134. Seeman TE, McEwen BS, Rowe JW, Singer BH, 2001. Allostatic load as a marker of cumulative biological risk: MacArthur studies of successful aging. PNAS. 98:4770- 4775.

135. McGregor PK, Peake TM. 2000. Communication networks: social environments for receiving and signalling behaviour. Acta Ethol. 2:71-81.

136. Oliveira RF, McGregor PK., Latruffe C, 1998. Know thine enemy: fighting fish gather information from observing conspecific interactions. Proc. R. Soc. B. 265:1045- 1049.

137. Earley RL, Dugatkin LA 2002. Eavesdropping on visual cues in green swordtail (Xiphophorus helleri) fights: a case for networking. Proc. R. Soc. Lond. B. 269:943-952.

138. Doutrelant C, McGregor PK, Oliveira RF. 2001. The effect of an audience on intrasexual communication in male Siamese fighting fish, Betta splendens. 12:283-286.

139. Heg D, Bergmüller R, Bonfils D, Otti O, Bachar Z, Burri R, Heckel G, Taborsky M. 2006. Cichlids do not adjust reproductive skew to the availability of independent breeding options. Behavioral Ecology, 17:419-429.

140. Buchner AS, Sloman KA, Balshine S. 2004. The physiological effects of social status in the cooperative breeding cichlid Neolamprologus pulcher. J. Fish Biol. 65:1080- 1095.

141. Mileva VR, Fitzpatrick JL, Marsh-Rollo S, Gilmour KM, Wood CM, Balshine S. 2009. The stress response of the highly social African cichlid Neolamprologus pulcher. Physiol. Biochem. Zool. 82:720-729.

120

142. Bender N, Heg-Bachar Z, Oliveira RF, Canario AVM, Taborsky M. 2008. Hormonal control of brood care and social status in a cichlid fish with brood care helpers. Physiol. Behav. 94:349-358.

143. Ellis T, James JD, Stewart C, Scott AP. 2004. A non-invasive stress assay based upon measurement of free cortisol released into the water by rainbow trout. J. Fish Biol. 65:1233-1252.

144. Oliveira RF, Canario AV, Bshary R, 1999. Hormones, behaviour and conservation of littoral fishes: current status and prospects of future research. Behav. Cons. Litt. Fishes. 149-178.

145. Scott AP, Hirschenhauser K, Bender N, Oliveira R, Earley RL, Sebire M, Ellis T, Pavlidis M, Hubbard PC, Huertas M, Canario A. 2008. Non-invasive measurement of steroids in fish-holding water: important considerations when applying the procedure to behaviour studies. Behaviour, 145:1307-1328.

146. Wong SC, Dykstra M, Campbell JM, Earley RL. 2008. Measuring water-borne cortisol in convict cichlids (Amatitlania nigrofasciata): is the procedure a stressor? Behav. 145:1283-1305.

147. Ellis T, Sanders MB, Scott AP. 2013. Noninvasive monitoring of steroids in fishes. Wien Tierӓrzti Monat-Vet Med Austria. 100:255-269.

148. Scott AP, Ellis T. 2007. Measurement of fish steroids in water – a review. Gen. Comp. Endocrinol. 153:392-400.

149. Zar JH. 1996. Biostatistical analysis, third ed. Prentice-Hall, Upper Saddle River, NJ.

121

150. Sopinka NM, Fitzpatrick JL, Desjardins JK, Stiver KA, Marsh-Rollo SE, Balshine S. 2009. Liver size reveals social status in the African cichlid Neolamprologus pulcher. J. Fish Biol. 75:1-16.

151. Alberts SC, Sapolsky RM, Altmann J. 1992. Behavioral, endocrine, and immunological correlates of immigration by an aggressive male into a natural primate group. Horm. Behav. 26:167-178.

152. Wootten JT, 1992. Indirect effects, prey susceptibility, and habitat selection: impacts of birds on limpets and algae. Ecol. 73: 981-991.

153. Menge BA. 1995. Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecol. Mono. 65:21-74.

154. Fehr E, Fischbacher U. 2004. Third-party punishment and social norms. Evol Hum. Behav. 25: 63-87.

155. Vucetich JA, Peterson RO, Waite TA. 2004. Raven scavenging favours group foraging in wolves. Anim. Behav. 67: 1117-1126.

156. Packer C, Scheel D, Pusey AE. 1990. Why lions form groups: food is not enough. Am. Nat. 136:1-19.

157. Cockburn A. 1998. Evolution of helping behavior in cooperatively breeding birds. Ann. Rev. Ecol. Syst. 29:141-177.

122

158. Bergmüller R, Heg D, Peer K, Taborsky M. 2005. Extended safe havens and between-group dispersal of helpers in a cooperatively breeding cichlid. Behav. 142:1643- 1667.

159. Kitchen DM, Beehner JC. 2007.. Factors affecting individual participation in group- level aggression among non-human primates. Beh. 144:1551-1581.

160. Rohwer S. 1978. Parent cannibalism of offspring and egg raiding as a courtship strategy. Am. Nat. 112:429-440.

161. Goodall J. 1986. The chimpanzees of Gombe: patterns of behavior. Harvard University Press, Cambridge.

162. Taborsky M. 1994. Sneakers, satellites, and helpers: parasitic and cooperative behavior in fish reproduction. Advances in the study of behavior, Vol. 23.

163. Hamilton IM, Dill LM. 2003. The use of territorial gardening versus kleptoparasitism by a subtropical reef fish (Kyphosus cornelii) is influenced by territory defendability. Beh. Ecol. 14:561-568.

164. Greenwood PJ. 1980. Mating systems, philopatry and dispersal in birds and mammals. Animal Behavior, 28:1140-1162.

165. Taborsky M. 1985. Breeder-helper conflict in a cichlid fish with broodcare helpers: an experimental analysis. Behav. 95:45-75.

166. Rood JP. 1987. Dispersal and intergroup transfer in the dwarf mongoose. In: Mammalian dispersal patterns. The effects of social structure on population genetics (Chepko-Sade BD, Halpin ZT, eds). University of Chicago Press, Chicago, 85-103.

123

167. Rood JP. 1990. Group size, survival, reproduction, and routes to breeding in dwarf mongooses. Anim Behav. 39:566-572.

168. Komdeur J. 1992. Importance of habitat saturation and territory quality for evolution of cooperative breeding in the Seychelles warbler. Nature. 358:493-495.

169. Girman DJ, Mills MGL, Geffen E, Wayne RK. 1997. A molecular genetic analysis of social structure, dispersal, and interpack relationships of the African wild dog (Lycaon pictus). Behav. Ecol. Sociobiol. 40:187-198.

170. Lewis RJ. 2008. Social influences on group membership in Propithecus verreauxi verreauxi. Int. J. Primatol. 29:1249-1270.

171. Alexander RD. 1974. The evolution of social behavior. Ann. Rev. Ecol. Systematics. 5:325-383.

172. Sherman PW. 1985. Alarm calls of Belding’s ground squirrels to aerial predators: nepotism or self-preservation? Behav Ecol Sociobiol. 17:313-323.

173. Côté IM, Poulin R. 1995. Parasitism and group size in social animals: a meta- analysis. Behavioral Ecology, 6:159-165.

174. Rimmer DM, Power G. 1978. Feeding response of Atlantic salmon (Salmo salar) alevins in flowing and still water. J. Fisheries Board Can. 35:329-332.

175. Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Can J Zool. 68:619-640.

124

176. Magurran AE, Seghers BH. 1991. Variation in schooling and aggression amongst guppy (Poecilia reticulata) populations in Trinidad. Beh. 118:214-234.

177. Martel G, Dill LM. 1995. Influence of movement by coho salmon (Oncorhynchus kisutch) parr on their detection by common mergansers (Mergus merganser). Ethol. 99:139-149.

178. von Siemens M. 1990. Broodcare or egg cannibalism by parents and helpers in Neolamprologus brichardi (Poll 1986) (Pisces: Cichlidae): a study on behavioural mechanisms. Ethol. 84:60-80.

179. Stiver KA, Fitzpatrick J, Desjardins JK, Balshine S. 2006. Sex differences in rates of territory joining and inheritance in a cooperatively breeding cichlid fish. Anim Behav. 71:449-456.

180. Zöttl M, Chapuis L, Freiburghaus M, Taborsky M. 2013. Strategic reduction of help before dispersal in a cooperative breeder. Biology Letters, doi: 10.1098/rsbl.2012.0878.

181. Dey CJ, Tan QYJ, O’Connor CM, Reddon AR, Caldwell JR, Balshine S. 2015. Dominance network structure across reproductive contexts in the cooperatively breeding cichlid fish Neolamprologus pulcher. Current Biology, 61:45-54.

182. Griesser M, Nystrand M, Eggers S, Ekman J. 2008. Social constraints limit dispersal and settlement decisions in a group-living bird species. Behavioral Ecology, 19:317-324.

183. Wong MYL, Balshine S. 2011. Fight for your breeding right: hierarchy re- establishment predicts aggression in a social queue. Biol Lett. 7:190-193.

125

184. Konings A. 2005. Back to nature guide to Tanganyika cichlids. Cichlid Press, El Paso.

185. Trewavas E. 1983. Tilapine Fishes of the Genera Sarotherodon, Oreochromis and Kanakilia. British Museum of Natural History, London, UK.

186. Werner NY, Balshine S, Leach B, Lotem A. 2003. Helping opportunities and space segregation in cooperatively breeding cichlids. Behavioral Ecology, 14:749-756.

187. Hick K., Reddon AR, O’Connor CM, Balshine S. 2014. Strategic and tactical fighting decisions in cichlid fishes with divergent social systems. Behav. 151:47-71.

188. Reddon AR, O’Connor CM, Marsh-Rollo SE, Balshine S, Gozdowska M, Kulczykowska E. 2015. Brain nonapeptide levels are related to social status and affiliative behaviour in a cooperatively breeding cichlid fish. R. Soc. Open Sci. 2:140072.

186. Hert E. 1985. Individual recognition of helpers by the breeders in the cichlid fish Lamprologus brichardi (Poll, 1974). Z. Tierpsychol. 68:313-325.

187. Taborsky M, Hert E, von Siemens M, Stoerig P. 1986. Social behaviour of Lamprologus species: functions and mechanisms. Ann. Kon. Mus. Mid. Afr. Zool. Wetensch. 251:7-11.

188. Bergmüller R, Taborsky M. 2005. Experimental manipulation of helping in a cooperative breeder: helpers ‘pay to stay’ by pre-emptive appeasement. Anim. Behav. 69:19-28.

189. Bergmüller R, Taborsky M. 2007. Adaptive behavioural syndromes due to strategic niche specialization. BMC Ecol. 7:12 doi:10.1186/1472-6785-7-12.

126

190. Jordan LA, Avolio C, Herbert-Read JE, Krause J, Rubenstein DI, Ward AJW. 2010. Group structure in a restricted entry system is mediated by both resident and joiner preferences. Behav. Ecol. Sociobiol. 64:1099-1106.

191. Desjardins JK, Fitzpatrick JL, Stiver KA, Van Der Kraak GJ, Balshine S. 2011. Lunar and diurnal cycles in reproductive physiology and behavior in a natural population of cooperatively breeding fish. J. Zool. 285:66-73.

192. Hebblewhite M, Pletscher DH. 2002. Effects of elk group size on predation by wolves. Can. J. Zool. 80:800-809.

193. Rannala BH, Brown CR. 1994. Relatedness and conflict over optimal group size. TREE. 9:117-119.

194. Taborsky M, Grantner A. 1998. Behavioural time-energy budgets of cooperatively breeding Neolamprologus pulcher (Pisces: Cichlidae). Anim Behav. 56:1375-1382.

195. Chen CC, Fernald RD. 2011. Visual information alone changes behavior and physiology during social interactions in a cichlid fish (Astatotilapia burtoni). PLOSOne. 6:e20313.

196. Zack S, Rabenold KN. 1989. Assessment, age and proximity in dispersal contests among cooperative wrens: field experiments. Anim. Behav. 38:235-247.

197. Le Vin A.L, Mable BK, Arnold KE. 2010. Kin recognition via phenotype matching in a cooperatively breeding cichlid, Neolamprologus pulcher. Anim Behav. 79:1109- 1114.

127

198. Reddon AR, Balk D, Balshine S. 2011. Sex differences in group-joining decisions in social fish. Anim Behav. 82:229-234.

199. Fitzpatrick JL, Desjardins JK, Stiver KA, Montgomerie R, Balshine S. 2006. Male reproductive suppression in the cooperatively breeding fish Neolamprologus pulcher. Behav. Ecol. 17:25-33.

200. Oliveira RF, Taborsky M, Brockmann HJ, eds. 2008. Alternative reproductive tactics: an integrative approach. Cambridge University Press.

201. Stockley P, Searle JB, MacDonald DW, Jones CS. 1993. Female multiple mating behaviour in the common shrew as a strategy to reduce inbreeding. Proceedings of the Royal Society, B. 254:173-179.

202. Nakamura M. 1998. Multiple mating and cooperative breeding in polygynadrous alpine accentors. I. Competition among females. Animal Behaviour. 55:259-275.

203. Nakamura M. 1998. Multiple mating and cooperative breeding in polygynandrous alpine accentors. II. Male mating tactics. Animal Behaviour. 55:277-289.

204. Rapoport A. 1997. Order of play in strategically equivalent games in extensive form. International Journal of Game Theory. 26:113-136.

205. Muller RA, Sadanand A. 2003. Order of play, forward induction, and presentation effects in two-person games. Experimental Economics. 6:5-25.

206. Weber RA, Camerer CF, Knez M. 2004. Timing and virtual observability in ultimatum bargaining and “weak link” coordination games. Experimental Economics. 7:25-48.

128

207. Hamilton IM. 2013. The effects of behavioral plasticity and leadership on the predictions of optimal skew models. Behavioral Ecology. 24:444-456.

208. Dugatkin LA, Reeve HK eds. 1998. Game theory and animal behavior. Oxford University Press.

209. Hirschleifer J. 2001. Game-theoretic interpretations of commitment. Evolution and the capacity for commitment. 77-93.

210. Hamilton WD. 1964. The genetical evolution of social behaviour. I. Journal of Theoretical Biology, 7:1-16.

211. Hamilton WD. 1964. The genetical evolution of social behaviour II. Journal of Theoretical Biology, 7:17-52.

212. Klein DB, O’Flaherty B. 1993. A game-theoretic rendering of promises and threats. Journal of Economic Behavior and Organization. 21:295-314.

213. Walter B, Trillmich F. 1994. Female aggression and male peace-keeping in a cichlid fish harem: conflict between and within the sexes in Lamprologus ocellatus. Behavioral Ecology and Sociobiology. 34:105-112.

214. Schradin C, Lamprecht J. 2000. Female-biased immigration and male peace-keeping in groups of the shell dwelling cichlid fish Neolamprologus multifasciatus. Behavioral Ecology and Sociobiology. 48:236-242.

215. Sherman PW, Jarvis JUM, Alexander RD, eds. 1991. The Biology of the Naked Mole Rat. Princeton University Press.

129

216. Brandtmann G, Scandura M, Trillmich F. 1999. Female-female conflict in the harem of a snail cichlid (Lamprologus ocellatus): behavioral interactions and fitness consequences. Behaviour. 136:1123-1144.

217. Hamilton IM. 2004. A commitment model of reproductive inhibition in cooperatively breeding groups. Behavioral Ecology. 4:585-591.

218. Emlen ST. 1982. The evolution of helping. I. An ecological constraints model. The American Naturalist. 119:29-39.

219. Conradt L, Roper TJ. 2005. Consensus decision making in animals. TREE. 20:449- 456.

220. Langer P, Hogendoorn K, Keller L. 2004. Tug-of-war over reproduction in a social bee. Nature. 428:844-847.

221. Buston PM, Zink AG. 2009. Reproductive skew and the evolution of conflict resolution: a synthesis of transactional and tug-of-war models. Behavioral Ecology. 24:444-456.

130

Appendix A: N. pulcher Ethogram (Chapter 4)

131

Category Behavior Description Attacks Ram Charges opponent and makes physical contact. Chase Chases opponent.

Display Fin Raise Dorsal fin of erect while facing opponent. Slow Approach Slowly approaching opponent, often associated with other display behaviors. Puffed Throat Opercular plates covering gills flared out.

Other Agonistic Avoid Moving away from approaching conspecific before or in response to aggressive act. Block An individual (typically more-dominant individual) moving between two other conflicting individuals and "corralling" one away from conflict.

Submissive Tail Quiver Rapidly moving tail in proximity to a conspecific.

Affiliative Join Swims into proximity of another fish who remains in proximity of each other. Distinguishable from a "slow approach" in the lack of aggression or avoidance by either party. Bump Approaches conspecific and gently "bumps" head into the underside of conspecific.

Reproductive Soft touch Individual is positioned parallel to conspecific, gently bites at flank of conspecific. In the present study only observed by dominant pair, typically occurs in or around breeding shelter in the days preceding spawning.

132

Appendix B: Principal Component Loadings (Chapter 4)

133 a. Dominant Male - Dominant Female

DMDF DMDF DMDF DMDF DMDF DMDF DM →DF Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Block 0.116 0.304 -0.139 0.710 0.117 0.293 Slow Approach 0.545 0.579 -0.180 0.481 0.120 -0.102 Chase 0.346 0.134 -0.777 0.313 -0.287 -0.015 Ram 0.863 0.236 -0.221 -0.094 -0.134 0.001 Soft Touch 0.113 -0.857 0.276 0.109 0.254 -0.138 Fin Raise 0.071 0.888 0.160 0.183 0.089 -0.016 Puffed Throat 0.806 -0.176 0.158 0.121 0.344 0.245 Join -0.022 -0.363 -0.067 0.794 0.246 0.078 DF→DM Slow Approach -0.045 0.193 0.885 -0.307 0.066 0.062 Chase -0.473 -0.212 -0.316 -0.216 0.279 -0.491 Ram 0.160 -0.041 0.025 0.146 0.078 0.913 Fin Raise 0.273 0.808 0.321 -0.212 0.003 -0.031 Puffed Throat 0.739 0.301 -0.182 -0.021 0.001 0.401 Avoid -0.064 -0.005 0.030 0.237 0.917 0.030 Bump -0.146 0.019 0.812 0.404 -0.258 0.005 Tail Quiver 0.785 0.166 0.043 0.359 -0.290 -0.106 Join -0.759 0.108 0.234 0.067 0.000 -0.129 b. Dominant Male - Subordinate Female

DMSF DMSF DMSF DM→SF Factor 1 Factor 2 Factor 3 Slow Approach 0.791 0.074 0.180 Chase 0.733 -0.420 -0.308 Fin Raise 0.062 0.033 0.951 Join 0.135 0.919 0.036 SF→DM

Bump 0.633 0.357 -0.276 Tail Quiver 0.709 0.018 0.241 Join 0.743 0.262 -0.091

c. Dominant Female - Subordinate Female DFSF DFSF DFSF DF→SF Factor 1 Factor 2 Factor 3 Slow Approach 0.799 -0.011 0.002 Chase 0.348 -0.441 0.645 Join -0.253 0.158 0.793 Fin Raise 0.817 0.059 0.037 SF→DF

Fin Raise 0.881 -0.008 -0.226 Bump 0.049 0.765 0.195 Tail Quiver 0.777 0.393 0.015 Join 0.159 0.770 -0.203

134