a aaééckkfifi ,
3r»? ..

5.3. ..

.:.r.¢l:.

.fi.

, w

. ..l

33...
lJt.H.¥2:.il
as.

7.}: l , 2..
34:53:53. VP $.53.
flu“.

I’i‘. 4|
11.42.. 9 .
(4.1:... r.

1 1':
x) “in.

yr.

 

.;
t “Hanan”:
I: 14.32.?
i... .
r1... K A

 

..vi.“\ itve-L‘I: t
. e .133. “no.

,2 i r. .1-..
.. in...
75533.,

.x.l‘.|.$.r2

. If... 5|..Ih.i l.

X: i. 4:3530 )

0.4.?! f.
.

3 ,i
L... .[
Ixti t?
xxx.‘ r: .
.‘v .) L. 5:0qu
81...-4\.!l

.Tx.:.
.2. . :

 

 

 

Eagééé§ué5% ‘ ...3_._:_%. . .V V . ., H. , _ : 5.}:

 

THESES

 

fwn UBRARY
‘K P Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

3261033

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRCJDateOuepss-p. 15

 

THE EVOLUTION AND MAINTENANCE OF COOPERATION IN NATURAL
AND ARTIFICIAL POPULATIONS

By

Robert Olendorf

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Zoology

2001

ABSTRACT

THE MAINTENANCE AND EVOLUTION OF COOPERATION IN NATURAL
AND ARTIFICIAL POPULATIONS

By

Robert K. Olendorf

Cooperative behaviors, where an individual benefits another at its own
expense, pose a special problem for biologists. Although populations of
cooperative individuals have higher average fitness than populations of
uncooperative individuals, cooperative populations are open to invasion by
uncooperative individuals. Proposed mechanisms for how cooperation can
resist invasion by uncooperative individuals fall into four broad categories, kin
selection, group selection, reciprocal altruism and by-product mutualism. The
primary goals of my research was to: 1) determine the mechanism responsible
for the maintenance of two cooperative behaviors in red-winged blackbirds; 2)
explore the role extra-pair paternity plays in cooperation 3) use simulations to
show how cooperative behaviors become established in a population of

uncooperative individuals.

Reduced aggression among territorial males is potentially a form of cooperation
because individuals are tempted to cheat by unilaterally expanding their territory
or by seeking extra-pair copulations on their neighbor’s territory. I used
simulated defections to determine if reduced aggression was a form of
reciprocal altruism. I used paternity analysis coupled with behavioral

observations of territorial behavior to estimate the frequency of cheating and to

determine if males respond to cuckoldry by their neighbors. The simulated
defections showed that males responded to simulated defections by increasing
territorial aggression towards the neighbor, a result indicative of reciprocal
altruism. My results also show that males are more aggressive towards males
that have successfully cuckolded them. Additional evidence suggests that
males may be able to assess a neighbor’s ability to cuckold rather than directly

detect cuckoldry.

I also used simulated defections combined with paternity analysis of nestlings to
simultaneously test between reciprocal altruism and by-product mutualism.
Previous studies have provided evidence for either hypothesis although no
study has tested for both in the same population. Males in the population I
studies appear to cooperative nest defense primarily as a form of reciprocal
altruism. There was no evidence of by-product mutualism in the form of males
defending nests on other territories in which they had obtained extra-pair

fertilizations.

Simulations using genetic algorithms show that population structure enhances
the evolution of cooperation from an uncooperative population. Populations
composed of small subpopulations achieved higher rates of cooperation than
populations composed of large subpopulations. Additionally, population
structure influenced the strategies that evolved. The common strategies in the
smallest populations cooperated almost unconditionally. The most common

strategies in lightly larger populations, however, were similar to Tit-For—Tat.

ACKNOWLEDGEMENTS

I would like to thank Amy Curry, Maxine Henry, Brian Black and Natalie Dubois
for assistance in the field. We would also like to thank Lisle Gibbs for guidance
on primer selection and blood storage. The hard working members of the
Scribner Lab, especially Scott Libants, Jennifer Warrilow, Eric Olle and
Merideth Bartow, were instrumental to the success of this study. Their guidance
and aid in the lab was indespensable. Paul Sotherland provided much needed
instruction in drawing blood and building walk-in traps. I would also like to thank
my committee, Tom Getty, Kim Scribner, Scott Robinson and Alan Tessier for
their patience and guidance throughout my graduate career. Most especially, I
would like to thank my wife Szu-Yu Chao, and my parents Karen and Frank

Schmitt for their love and support.

iv

TABLE OF CONTENTS

List of Tables ..................................................................................................... vii
List of Figures ................................................................................................... viii

Chapter 1: Male Reproductive Success in Red-Winged Blackbirds: Lack of

Trade-Off Due To Female Choice ...................................................................... 1
Abstract ....................................................................................................... 1
Introduction ................................................................................................. 3
Methods ...................................................................................................... 6
Results ...................................................................................................... 10
Discussion ................................................................................................. 18

Male Quality, Territory Settlement and Reproductive Success .......... 20
Literature Cited ......................................................................................... 23

Chapter 2: Reduced territorial aggression among “dear enemies” in red-winged

blackbirds: Tit-for-tat behavior with cryptic defectors ....................................... 27
Abstract ..................................................................................................... 27
Introduction ............................................................................................... 29
Methods .................................................................................................... 34

Study Site and Population .................................................................. 34
Capture of birds, marking and blood collection .................................. 35
Observation of territorial behavior ...................................................... 36
Paternity analyses .............................................................................. 37
Recording and editing of songs .......................................................... 39

 

Simulated Detection Trials ................................................................. 39

Statistical Analysis ............................................................................. 43
Results ...................................................................................................... 46
Discussion ................................................................................................. 50

Do red-winged blackbirds play TFT like strategies? ........................... 53

Cheating among “Dear Enemies” ....................................................... 55

Neighbor recognition and “Dear Enemies” ......................................... 58
Literature Cited ......................................................................................... 59

Chapter 3: Cooperative Nest Defense in Red-winged Blackbirds: Simultaneous

Tests for Reciprocal Altruism and Mutualism ................................................... 65
Abstract ..................................................................................................... 65
Introduction ............................................................................................... 67
Methods .................................................................................................... 72

Study Site and Population .................................................................. 72
Collection of Background Data ........................................................... 72
Paternity analyses .............................................................................. 74
Recording of songs ............................................................................ 76
Observation of cooperative nest defense ........................................... 76
Experimental tests for retaliation ........................................................ 78
Statistical Analysis ............................................................................. 80
Results ...................................................................................................... 83
Discussion ................................................................................................. 88
Nest Defense as TFT ......................................................................... 91

vi

Nest Defense as Mutualism ............................................................... 93
Simultaneous tests for reciprocal altruism and mutualism ................. 95
Literature Cited ......................................................................................... 96

Chapter 4: The Effect of Populations Structure on the Evolution of Cooperation

and Cooperative Strategies In Artificial Populations ....................................... 101
Abstract ................................................................................................... 101
Introduction ............................................................................................. 1 03
The Model ............................................................................................... 109

The Genetic Code ............................................................................ 109
The Simulation ................................................................................. 113
Experimental Design ........................................................................ 1 15
Results .................................................................................................... 116
Randomly Generated Agents in a mutualistic environment ............. 116
Pure TFT in PD environments .......................................................... 119
Randomly Generated Agents in the PD environment ...................... 119
Discussion ............................................................................................... 123
Literature Cited ....................................................................................... 129

vii

LIST OF TABLES

CHAPTER 1

Table I. The results of a PCA analysis performed on four physical traits
measured on red-winged blackbirds. PC1 is associated with size and PC is
associated with culmen length relative to weight .............................................. 16

Table II. Correlation coefficients between a male physical traits and measures
of reproductive success .................................................................................... 17

CHAPTER 2

Table I. Component loading and percent of variance explained by the
correspondence analyses performed on observational data in 1998 and 1999
(a) and experimental data (b). .......................................................................... 45

CHAPTER 3

Table l. Component loading and percent of variance explained by the
correspondence analyses performed on data taken from crow presentation at a)
individual nests and b) territorial boundaries. ................................................... 82

CHAPTER 4

Table l. Examples of chromosomes along with the associated strategies. The
names of commonly discussed strategies are also shown. ............................ 112

viii

LIST OF FIGURES

CHAPTER 1

Figure 1. The spatial arrangement of the ponds used in this study. Ponds were
30 m in diameter and approximately 3 m maximum depth. Ponds on the same
row were approximately 5 meters apart while rows were approximately 10
meters apart. Lines across ponds show territory sizes on ponds with multiple
territories. Ponds without and divisions held a single territory. ........................... 8

Figure 2. The association between the number of EPFs obtained by a male and a. the
number of young fledged from a male’s territory or b. number of nests on a male’s
territory. N = 19. .......................................................................................................... 12

Figure 3. The association between the percent of a male’s young that were cuckolded
and a. the number of young fledged from a male’s territory and b. number of nests on a
male’s territory. N = 19. ............................................................................................... 13

Figure 4. The relationship between total male reproductive success (number of
fertilizations over the entire site) and a. Young fledged on a male’s territory; b.
EPCs obtained by a male and c. proportion of young on a male's that was
cuckolded. N = 19 males. ................................................................................. 14

Figure 5. Proposed relationships between size, EPCs, on-territory reproduction,
cuckoldry and reproductive success. Words in gray represent unmeasured
variables that thought to play a role in male reproductive success. Lines with
one arrow show relationships where significant correlations were found and
causality is inferred. Double-headed arrows show relationships where causality
cannot be inferred. Solid lines indicate positive relationships; dashed lines
represent negative relationships. ..................................................................... 15

CHAPTER 2

Figure 1. Schematic and time line of the experiment. Two territories are shown.
Pre-experiment observations were performed between April 24 and April 30. All
trials were conducted between May 1 and May 8. Numbers in the pond
schematic indicate the order playbacks. Number 1 is the pre-invasion trial,
number two is the invasion trial and number 3 is the post-invasion trial. Circles
indicate that the neighbor’s song was played at that position, the star indicates
that a stranger’s song was played. The time line shows an example of the
timing of the observations and playbacks for one trial ...................................... 41

ix

Figure 2. Aggression by focal males toward playbacks in response to
experimental treatment. Experimental treatments simulated territorial invasions
by neighbors; control treatments simulated invasion by stranger males. Position
of the speakers is given above each pair of bars. The song played in a playback
is shown inside or above each bar, N = Neighbor, S = Stranger. Error bars give
standard error. N = 6 in both treatments. *** P < 0.001. ................................... 47

Figure 3. Aggression by the focal male directed towards its neighbor before and
after simulated invasions in response to treatment. Males in experimental
treatments had songs from a neighbor played within its boundaries. Males in
control treatments had songs from a strange male played within its boundaries.
Error bars give standard error. N = 6 for both treatments. ** P < 0.01 ............. 49

Figure 4. Average aggression by each male toward neighbors who sired young
on its territory and those that had not. Each line represents one male and each
circle represents the average aggression towards neighbors in that category.
Three males are not shown because one was cuckolded by all of its neighbors
and two males were not cuckolded by any of their neighbors. N = 19 males... 51

Figure 5. Aggression by territorial males in response the total number of extra-
pair fertilizations obtained by a neighbor across the entire population. The
circles show all pair wise interactions (N=99). The dark dashed line shows the
over all trend the solid gray lines give show the trend lines for each individual
male (N: 22). ................................................................................................... 52

CHAPTER 3

Figure 1. Nest defense by the focal male in response to treatment.

Experimental treatments simulated defections by the neighbor male, control
treatments allowed cooperative defense by both males. Error bars show :1: 1SE.
N: 6 for both treatments. *** P < 0.001. ........................................................... 84

Figure 2. Nest defense behavior by neighbor males in pre- and post- defection
trials in response to treatment. In experimental trials the neighbor male was
held in a covered cage for the duration of the experiment. In control trials the
neighbor was captured and released. Error bars show a: 1 SE. N = 6 for both
treatments. ....................................................................................................... 86

Figure 3. The level of nest defense by males at nests on other territories where
they had EPFs and a randomly chosen male neighboring the same territory that
was the same distance from the nest. N = 25. Error bars show :1: 1 SE. .......... 87

Figure 4. Average number of helpers at nests with cuckolded offspring
compared to nests without cuckolded offspring. Each circle represents the

average number of helpers at cuckolded and uncuckolded nests. Lines connect
data from within a territory. N = 19. .................................................................. 89

Figure 5. Nest defense by territorial males in response to cuckolded offspring in
their nests. Circles represent average defense by the territorial male at
cuckolded and uncuckolded nests. Lines connect data from within the same
territory. N = 19 ................................................................................................ 90

CHAPTER 4

Figure 1. The structure of the genetic algorithm. A chromosome is constructed
of an array of characters. Individual genes within a chromosome delineated by
slashes. Each gene consists of a receptor and effector portion. The receptor
starts at the left slash and continues to the first colon. The receptor is a pattern
to be matched by an opponent’s behavioral history. The c’s and d’s correspond
to previous plays, the most recent on the right. Asterisks (*) match any
combination of characters or no characters. The receptor is the first c or d to
the right of a colon; the remainder of the string is ignored until the next
functional gene is encountered. ..................................................................... 110

Figure 2. Level of cooperation achieved after 500 generation by randomly
generated populations in a mutualistic environment varying dispersal and
subpopulation size. The surface was fit using DWLS smoothing. .................. 117

Figure 3. The level of cooperation achieved in all populations in response to
subpopulation size in mutualistic environments. ............................................ 118

Figure 4. Level of cooperation achieved after 500 generation by initially pure
TFT populations in a PD environment varying dispersal and subpopulation size.
The surface was fit using DWLS smoothing. .................................................. 120

Figure 5. Level of cooperation achieved after 500 generation by randomly
generated populations in a PD environment varying dispersal and
subpopulation size. The surface was fit using DWLS smoothing. .................. 121

Figure 6. Level of cooperation in relation to subpopulation size. All runs are
included regardless of dispersal. .................................................................... 122

Figure 7. The effect of population structure on the types of strategies that

evolved in the populations is shown. TFT = tit-for—tat, ALLD = always defect,
ALLC = always cooperate, ONED = test for defection on the first play only... 124

xi

CHAPTER 1:

MALE REPRODUCTIVE SUCCESS IN RED-WINGED BLACKBIRDS: LACK
OF TRADE-OFF DUE TO FEMALE CHOICE

Robert Olendorf
Kellogg Biological Station
Michigan State University
3700 E Gull Lake Dr.
Hickory Corners, Ml 49060
Kim Scribner
Department of Fisheries and Wildlife

Michigan State University
East Lansing, MI 48823

ABSTRACT

Molecular analysis of paternity has demonstrated that realized male
reproductive success may be substantially different from on-territory nesting
success. The contribution of extra-pair paternity to male reproductive success,
however, is seldom obvious. Males that obtain large harems or productive
territories may suffer reduced on-territory reproductive success because they
must invest more time and energy into defense and maintenance. Males who
invest less in acquisition of territory and defense may seek more extra-pair
copulations. Alternatively, males with good territories and large harems may be
more attractive to females because they can provide extra-pair young with

either better resources or better genes. We used analysis of microsatellite loci

to genetically determine the mating success of males in relation to their
observed mating success. We looked for evidence that successful males are
physical different from those who are not. We also studied the effect that male
density had on the reproductive success of males. Males that were successful
in fledging young on their own territory also obtained the most EPFs and were
cuckolded less. We found that size (estimated using PCA analysis) and wing
chord correlated positively with number of young fledged on the territory, nests
on the territory and number of EPCs achieved by the male but not total
reproductive success. Males with neighbors on the same pond were cuckolded
more than males who held an entire pond as a territory. We found little
evidence for a trade-off in male reproductive strategies. The effects of female

choice likely over shadow any trade-off faced by male red-winged blackbirds.

INTRODUCTION

With recent advancements in molecular paternity analysis it has become
apparent that observed reproductive success and realized reproductive success
may not be the same (Gibbs et al., 1990; Petrie and Kempenaers, 1998;
Weatherhead and Boag, 1997). Off-territory reproduction, in the form of extra-
pair fertilizations (EPFs), may either decrease or augment a male’s on-territory
reproductive success. Determining effect of extra-pair paternity on reproductive
success is important because estimating total male reproductive success by
using territory production may not be accurate if extra-pair paternity is frequent
(Gibbs et al., 1990). Additionally, extra-pair paternity can alter the degree of
sexual selection by increasing or decreasing the variance in reproductive

success among males (Weatherhead and Boag, 1997).

Extra-pair paternity might decrease a male’s reproduction if there is a trade-off
between reproduction on a male’s own territory and ability to achieve extra—pair
copulations (EPCs). Males that are able to obtain productive territories and
large harems may have to spend more time protecting their territory and harem
against intruding males. Males that are successful at reproducing on their own
territory might therefore be less able to gain extra-pair copulations and might
also suffer more from cuckoldry on their own territory (Hasselquist and
Sherman, 2001; Westneat, 1993b). Additionally, males that invest more time

into seeking EPCs may suffer from higher cuckoldry, reduced on territory

mating success or reduced parental care (Sherman and Morton, 1988;

Westneat, 1988; Westneat, 1993a)

In contrast, there are also reasons to expect positive relationships between a
male on-territory reproductive success and off-territory reproductive success.
Extra-pair paternity would increase with on-territory reproductive success if
females choose males that were able to establish territories on productive
habitat (Hasselquist and Sherman, 2001 ). In this case, females control the
frequency of extra-pair fertilization in their young (Gray, 1996). Alternatively, a
positive correlation might also result even if males control extra-pair paternity. If
the variance in male quality is sufficiently high, low quality males could do
relatively worse even though they suffer proportionally less from cuckoldry. This
would result in a positive correlation when all males are considered together

(Stearns, 1989; Stearns, 1992).

Red-winged blackbirds are an ideal species to study such questions. This
species exhibits a high degree of variability in reproductive success, due
primarily to variation in habitat (Turner and McCarty, 1998; Weatherhead and
Robertson, 1977). In addition, they have been shown to engage in significant
levels of extra-pair copulation (Gray, 1997a; Gray, 1997b; Moller, 2000;

Weatherhead et al., 1994).

Female red-winged blackbirds have been shown to seek extra-pair copulations
and control the level of extra-pair paternity in their nest (Gray, 1996; Gray,
1997a). However, the relationship between a male on-territory and off-territory
reproductive success is unclear. Weatherhead (1997) found that male off-
territory reproduction was positively correlated with on-territory reproduction and
negatively correlated with the level of cuckoldry on a male’s territory. Gibbs
(1990) on the other hand, found that while off-territory reproduction correlated
negatively with cuckoldry there was no relationship between on-territory

reproduction and off-territory reproduction.

In this study, we examine the relationship between extra-pair paternity, on
territory paternity and fledging success within a territory to test for trade-offs in
population a population with almost no nest predation. This allows us to
examine the relationship between off-territory and on-territory reproductive
success without the potentially confounding affects of cooperative nest defense
(Gray, 1997b; Weatherhead et al., 1994). We also look for evidence that male
quality, measured by size, is associated with either a male’s success at raising
fledglings or a male’s ability to gain extra-pair fertilizations (EPFs) or the

number of young produced off territory.

METHODS

This study was conducted at the Kellogg Biological Station Experimental Pond
Facility, Hickory Corners, Michigan (42° 24' N, 85° 24' W) over four years, 1996-
1998. We only performed paternity analysis in 1998. Each of 18 ponds was
approximately 30 m in diameter and 3 m deep. The ponds were arranged in
three rows of six. Within rows, ponds were approximately 5 meters apart, rows
were spaced approximately 10 meters apart (Figure 1). The margin of each

pond was densely vegetated, predominately with cattails, Typha Iatifolia.

Although the ponds were constructed primarily for aquatic research, they were
consistently colonized by red-winged blackbirds. During this study, one to four
males settled each pond with each male holding a harem of one to six females.
A chain link fence surrounded the facility keeping out most predators; therefore

nesting success was very high (near 100%).

We marked every territorial male with a unique combination of three colored leg
bands and a numbered aluminum. We then measured weight, culmen length,
tarsus length and wing chord of each male and determined its age using

plumage criteria (Pyle, 1997).

We mapped each male’s territory using behavioral criteria (Beletsky and Orians,

1987). The entire pond was defined as a male’s territory when he was the only

6

resident male on that pond. On ponds with more than one male, we defined
territorial boundaries by where two neighboring males counter-sang (both males
within 10 m of each other and singing in alternate order), the limit of the
resident’s movement, or the center of overlap between two neighboring males’
movement. We estimated male harem size as the maximum number of

simultaneously active nests.

We searched for nests daily. We located most nests as they were being built
and found no nests later than a week after completion, so we are reasonably
certain we found all nests. We marked the location of each nest with flagging
tape at the edge of the pond and we monitored nests every 4 to 8 days until the
fate was determined. Although females often raise multiple broods in this
population, we incorporated data from only the first brood to limit pseudo-

replication and differences among broods.

We obtained 50-80 ml of blood from all territorial males and many females. We
also took approximately 0.5 ml of blood from nestling when they were 8 days
old. Blood was drawn from the brachial vein, and immediately placed in 800 ml

of “Queen’s” lysis buffer (Seutin et al., 1991).

DNA was extracted from the blood samples using Proteinase K digestion
followed by extraction in 7.5M NH4A0C and precipitation in isopropyl alcohol.

The DNA was washed once more in 70% ethyl alcohol.

OOOSOS
OOOOQO
OO©QO®

Figure 1. The spatial arrangement of the ponds used in this study.
Ponds were 30 m in diameter and approximately 3 m maximum
depth. Ponds on the same row were approximately 5 meters apart
while rows were approximately 10 meters apart. Lines across
ponds show territory sizes on ponds with multiple territories.
Ponds without and divisions held a single territory.

We determined paternity of nestlings using six microsatellite loci. Four of the
loci (Om 5, Om 10, Om 21, Om 31) were developed for great-tailed grackles
(Ouiscalus mexicanus) (Gibbs et al., 1997), Dpp 16 for yellow warblers
(Dendroica petechia) (Dawson et al., 1997) and Map 10 for brown-headed

cowbirds (Molothrus ater) (Gibbs et al., 1997).

We assayed genetic variation at these loci using PCR amplification in 25 pL
reaction volumes. Two different reaction conditions were required for the six
loci. Om 10, Dpu 16 and Map 10 were amplified using 250 ng of template DNA,
2 pmol of each primer (fluorescently labeled fon/vard primer), 500 uM dNTPs,
and .75 U Taq polymerase. Om 5, Om 21 and Om 31 were amplified using 125
ng template DNA, 1.25 pmol of each primer (fluorescently labeled forward
primer), 625 uM dNTPs and .75 U Taq polymerase. All reactions were
performed in buffer containing 10 mM Tris-HCI pH 8.3, 1.5 mM MgC|2, 30mM
KCI, 0.01% gelatin, 0.01% NP-40, 0.01 Triton X-100. The PCR product was run
on 6% polyacrylamide gels and visualized using an FMBIO gel scanner. We
assigned genotypes to all individuals based on internal lane size standards and
individuals of known genotype run concurrently o the gel. All gels were scored

by hand by both authors and verified using FMBIO image analysis software.

We included only territorial males as potential fathers, since previous work has
shown that floating males rarely gain fertilizations (Gibbs et al., 1990). We

determined the paternity of all offspring employing either exclusion criteria or

maximum likelihood methods using Cervus (Marshall et al., 1998). Using the six
loci described above, the probability of assigning the wrong father to an
offspring was 0.7%. We were able to establish paternity for all but two of 252
offspring for which we obtained DNA. For 217 of the young, there was only one
non-excluded candidate parent. Using maximum likelihood estimates of
paternity, we were able to assign paternity to most of the remaining offspring
with 95% confidence level (26 offspring), but we also included assignments at

the relaxed confidence of 80% (7 offspring).

All correlations shown are Pearson’s correlations. We arcsine transformed
proportional data (i.e. proportion of young cuckolded). All significance values

are two tailed.

RESULTS

The overall rate of extra-pair paternity in 1998 was 36.2% of 252 offspring and
52.4% of 93 nests for which we obtained DNA. Only 2 of 252 offspring
appeared to be sired from males not on our study site. Among the extra-pair

offspring, a neighbor was the genetic father 88% of the time.

Each male fledged an average of 13.5 (15.50 SD) offspring on its territory. The
average total reproductive success for each male, based on paternity analysis

was 11 (16.15 SD) offspring. On average each male sired 4.1 (223.57 SD) extra-

10

pair offspring. The overall effect of extra-pair paternity was to slightly reduce
each male’s reproductive success while causing a moderate increase in the

variance in reproductive success.

Ponds were the primary determinant of nest success in our population. The
return rate by males onto the same pond was 27% over the four years of the
study. However, ponds were consistent among years in number of nestlings

fledged (Kendalls tau, T17 = 0.82, P < 0.001) as well as higher probability of nest

success (Kendall’s tau, 1'17 = 0.73, P < 0.001).

There was a positive relationship the number of young fledged and number of
EPFs obtained by that male on other territories (r19 = 0.58, P < 0.01; Figure 2a).
However, there was no relationship between the number of nests on a territory
and number of EPFs obtained by that male (r19 = -0.13, NS; Figure 2b). The
highly successful male in Figure 1a did not appreciably alter the correlation
when excluded from the analysis (r19 = .48, P < 0.05). Number of young fledged
on a territory was negatively correlated with proportion of nests cuckolded on
that territory (r19 = -0.70, P < 0.001; Figure 3a). However, there was no
association between number of nests on a territory and nests that were
cuckolded on that territory (r19 = -0.36, NS; Figure 3b). Total male reproductive
success (total number of fertilizations obtained on the study site) was positively
correlated with both the number of young fledged on the territory (r19 = 0.92, P <

0.001; Figure 4a) and number of EPFs obtained (r19 = 70, P < 0.0001; Figure

11

 

 

 

 

 

 

 

 

 

16
14‘ a' o
12‘
2
310‘
8 8'
I“:
a 5‘
DJ
4.
2.
0
0 30
YoungFledged
16F
14, b. o
12‘
8
£101
5 4
0 8 ° 2 .
u“! . .
o. 6‘
I.” O
4‘ ____°//’/——o
21 O O o
O
0 ¢ . 2 . . . . . .
0 1 2 3 4 5 6 7 8 9 10

Nests On Territory

Figure 2. The association between the number of
EPFs obtained by a male and a. the number of young
fledged from a male’s territory or b. number of nests
on a male’s territory. N = 19.

12

% Young Cuckolded

% of Young Cuckolded

0.80
0.70 *
0.60 '
0.50 ‘
0.40 '
0.30 .
0.20 .

 

0.101

000 . . . : .
0 5 10 1 5
Young Fledged

 

 

 

 

és 30

 

0.80
0.701
060* . °
0.50
0.40
0.30'
0.20
0.10
0.00

 

O
O

 

 

 

moo

4 5
Nests 0n Territory

Figure 3. The association between the percent of a male’s young that
were cuckolded and a. the number of young fledged from a male’s territory
and b. number of nests on a male’s territory. N = 19.

13

 

30

20

15

10

 

Total Reproductlve Success

 

 

 

0 5 i0 i5 20 és
Young Fledged on Territory

00
O

 

30

 

Total Reproductlve Success

 

 

00 é a 6 0 4T0 12 i4 16

EPO: Obtained by Male

 

 

30
25
20 . '
15

10

 

5

Total Reproductlve Success

 

 

 

000 oh oh 0b 0) 03 oh 07 03

% of Young Cuckolded

Figure 4. The relationship between total male
reproductive success (number of fertilizations
over the entire site) and a. Young fledged on
a male’s territory; b. EPCs obtained by a
male and c. proportion of young on a male’s
that was cuckolded. N = 19 males.

14

4b). Male reproductive success was negatively correlated with proportion of

offspring cuckolded on the territory (r19 = -0.67, P < 0.005; Figure 4c).

We performed a PCA analysis on the size data from every male caught in 1998.
The first principle component accounted for 45.2% of the variance and the
component variables all loaded such that an increase in PC1 was associated
with an increase in size. We therefore used PC1 as a measure of size. PC2
accounted for 24.5% of the variation and was associated with culmen size

relative to weight (Table 1).

Male size (PC1) correlated positively and significantly with the number of EPFs
obtained, number of young fledged and total reproductive success. However,
male size did not correlate with proportion of young cuckolded. Wing cord was
the only component variable to correlate significantly with any measure of
reproductive success. PC2 did not correlate significantly with any measure of

reproductive success (Table 2).

Males with neighbors on the same pond raised fewer young to fledging (r19 = -
0.513, P < 0.025) and had fewer nests (r19 = -0.571, P < 0.01). The proportion
of young cuckolded on a territory increased with the number of territories on a
pond (r19 = 0.52, P < 0.25). However, males with more neighbors did not

achieve more EPFs (r19 = -0.32, NS).

15

Table 1. The results of a PCA analysis performed
on four physical traits measured on red-winged
blackbirds. PC1 is associated with size and PC2
is associated with culmen length relative to weight.

 

PCt PC2
Weight 0.68 0.61
Wing Cord 0.84 0.08
Ta rsus 0.62 -0.14
Culmen 0.51 -0.77

 

% Variance Explained 45.19 24.49

16

Table 2. Correlation coefficients between a male physical traits and
measures of reproductive success.

PC1 P02 Wing Weigh Tarsm Ctlmen

 

 

EPFS QDtai ned 0.54 ** -0.22 0.23 0.07 0.25 0.06
YOU'g Fledged 0.55 “ 0.1 0 0.45 * 0.22 0.19 -0.07
Masts on Territory 0.48 " 0.12 0.62 “* 0.37 0.10 0.12
Tdal Reprodtctiye Stoce$ 0.46 " -0.1 1 -0.17 0.19 0.30 -0.19
°/o Yougg Cuzkdcbd 0.16 0.07 0.12 0.1 3 0.25 0.05

 

* P <0.05; '* P <0.025; *" P < 0.001

17

DISCUSSION

We found that there was little evidence for trade-offs between on-territory
reproduction and off-territory reproduction in male red-winged blackbirds. Males
who fledged the most young on their own territory were also better at both
obtaining their own EPFs and preventing their own young from being

cuckolded.

We also found that male size correlated with number of nests on the territory,
young fledged and the number of EPCs obtained. However, male size did not

correlate with cuckoldry on the male’s own territory.

A trade-off between within territory reproductive success and extra-pair
paternity is expected because the more resources an individual invests into one
mode of reproduction, the less it should be able to invest in others (Stearns,
1989; Stearns, 1992). It seems reasonable therefore to expect that males who
invest time and energy into obtaining and defending high quality territories
might be less successful in gaining EPFS. Similarly, males who find themselves
on low quality territories might invest more energy into seeking EPFs

(Hasselquist and Sherman, 2001).

Despite this expectation, we found no direct evidence of a trade-off. Males who

fledged the most young on their own territory were also most successful in

18

gaining EPFs. Additionally, males that fledged the most young suffered
proportionately less from cuckoldry. Weatherhead & Boag (1997) found similar
results in a Washington population of red-winged blackbirds. Weatherhead &
Boag also found positive correlations between current and future reproduction

as well. There are two possible explanations for these positive correlations.

First, a trade-off can be masked by large variation in quality among individuals
(Stearns, 1989; Stearns, 1992). Trade-offs may be occurring, but they remain
hidden because high quality males from high quality territories always
outperform males from low quality territories. A correlation among all males
would therefore yield a positive relationship between territory quality and
measures, such proportion of young cuckolded. We have some evidence that
this might occur in our population. We found considerable variation in
reproductive success among males, possibly enough to account for such an
effect. We also found that the proportion of young cuckolded on territories with
neighbors on the same pond was higher than on ponds with a single territory.
Since ponds with multiple territories were also the most productive, the overall

effect was to mask the trade-off and show a positive correlation.

The other possible explanation for the lack of a trade-off is female choice. Nest
success in this population is largely dependent on the pond on which the
territory is located (Eckert and Weatherhead, 1987; Weatherhead and

Robertson, 1977). We found a high concordance among years in pond

19

productivity while at the same time only 25% of the males returned to a territory.
Females therefore appear to settle on the most productive ponds. Females that
are unable to settle on a high quality territory may “make the best of a bad

situation” by seeking EPCs with males on high quality ponds. Females are able
to gain greater foraging access on these high quality territories and may benefit

from additional nest defense from other territorial males (Gray, 1997b).

Females may also benefit genetically from EPCs if they mate mostly with males
on high quality territories. We found evidence that larger males are more likely
to obtain higher quality territories than smaller males and larger males obtain
more EPFS than smaller males. If size were heritable then the young would be
larger and more likely to obtain high quality territories in the future. These
results corroborate Gray’s (1997a) results that females benefit genetically by

seeking EPCs.

Male Quality, Territory Settlement and Reproductive Success

Male quality correlated significantly with territory productivity and several
measure of reproductive success. We found a positive relationship between
male size and pond productivity. Larger males fledged more young on their
territory and had more nests on their territory. Previous studies suggest that
male-male competition determines settlement pattern and females choose

nesting sites based on-territory quality (Picman, 1987; Searcy and Yasukawa,

20

1995). It therefore appears that larger males are better able to obtain good
territories and females choose mates based on their territory. Other studies
have also found that large males enjoy higher reproductive success on their
own territory (Rohwer et al., 1996; Searcy, 1979). Size would therefore appear
to influence female choice indirectly through territory settlement. Large males
are better able to obtain high quality territories and females then choose to

settle on high quality territories (Figure 5).

In addition to higher on-territory reproductive success, larger males on our
study site also achieved more EPCs than smaller males. This suggests that not
only are larger males able to settle on better territories but they are also able to
gain copulations with females, both on their territory and off. This result
suggests that females not only seek EPCs with males to gain material benefits
(Gray, 1997b; Weatherhead et al., 1994) but also obtain genetic benefits. If
females mate with large males then the young are likely to be larger as well.
The offspring would therefore be more likely to obtain larger territories and more
matings. Female choice therefore appears to operate directly on male
reproductive success by increasing a male’s off-territory reproduction (Figure
5). This is consistent with the “good genes” hypothesis (Kempenaers et al.,
1997; Moller, 2000) and corroborates other studies that suggest that females

seek EPCs in order to gain genetic benefits (Gray, 1997a).

21

Cuckoldry

/
' A
I
l I \
/ ' \
r V .

Frfiit‘gfifn
' 'r' -15.

 

RS 4— # EPCs 4

 

I, t. J...)
K.“ . .04"

, l

I
On-territory A’ 1,,”th
Reproduction t;- t, .31...

~
~—_-‘-’

Size

Figure 5. Proposed relationship between size, EPCs, on-
territory reproduction, cuckoldry and reproductive success.
Words in gray represent unmeasured variables that thought
to play a role in male reproductive success. Lines with one
arrow show relationships where significant correlations
were found and causality is inferred. Double headed arrows
show relationships where causality cannot be inferred. Solid
lines indicate positive relationships, dashed lines represent
negative relationships.

22

We found little evidence that there is a trade-off between on-territory and off-
territory reproduction. Large males fledged the most offspring from their
territories, a smaller proportion of their young were cuckolded and they were
able to obtain more EPFs. This is consistent with other studies that show
positive relationships between other aspects of male reproductive success
(Hasselquist and Sherman, 2001; Weatherhead and Boag, 1997). We also
show that size correlates with all aspects of male reproductive success except
proportion of young cuckolded. Taken together, this study supports the
conclusions that females choose territories based on territory quality but seek

EPCs based on male quality.

LITERATURE CITED

Beletsky LD, Orians GH, 1987. Territoriality Among Male Red-Winged
Blackbirds .1. Site Fidelity and Movement Patterns. Behavioral Ecology
and Sociobiology 20:21 -34.

Dawson RJG, Gibbs HL, Hobson KA, Yezerinac SM, 1997. Isolation of
microsatellite DNA markers from a passerine bird, Dendroica petechia
(the yellow warbler), and their use in population studies. Heredity 79:506-
514.

Eckert CG, Weatherhead PJ, 1987. Male Characteristics, Parental Quality and
the Study of Mate Choice in the Red-Winged Blackbird (Age/aius

Phoeniceus). Behavioral Ecology and Sociobiology 20:35-42.

23

Gibbs HL, Miller P, Alderson G, Sealy SG, 1997. Genetic analysis of brown-
headed cowbirds Molothrus ater raised by different hosts: data from
mtDNA and microsatellite DNA markers. Molecular Ecology 6:189-193.

Gibbs HL, Weatherhead PJ, Boag PT, White BN, Tabak LM, Hoysak DJ, 1990.
Realized reproductive success of polygynous Red-winged Blackbirds
revealed by DNA markers. Science 250:1394-1397.

Gray EM, 1996. Female control of offspring paternity in a western population of
red-winged blackbirds (Agelaius phoeniceus). Behavioral Ecology and
Sociobiology 38:267-278.

Gray EM, 1997a. Do female red-winged blackbirds benefit genetically from
seeking extra-pair copulations? Animal Behaviour 53:605-623.

Gray EM, 1997b. Female red-winged blackbirds accrue material benefits from
copulating with extra-pair males. Animal Behaviour 53:625-639.

Hasselquist D, Sherman PW, 2001. Social mating systems and extrapair
fertilizations in passerine birds. Behavioral Ecology 12:457-466.

Kempenaers B, Verheyen G, Dhondt AA, 1997. Extrapair paternity in the blue tit
(Parus caeruleus): female choice, male characteristics, and offspring
quality. Behavioral Ecology 82481-492.

Marshall TC, Slate J, Kruuk LEB, Pemberton JM, 1998. Statistical confidence
for likelihood-based paternity inference in natural populations. Molecular
Ecology 7:639-655.

Moller AP, 2000. Male parental care, female reproductive success, and

extrapair paternity. Behavioral Ecology 1 1 :161-168.

24

Petrie M, Kempenaers B, 1998. Extra-pair paternity in birds: explaining the
variability in heritable male fitness. TREE 13:52-58.

Picman J, 1987. Territory Establishment, Size, and Tenacity By Male Red-
Winged Blackbirds. Auk 104:405-412.

Pyle P, 1997. Identification guide to North American birds. Bolinas, California:
Slate Creek Press.

Rohwer S, Langston N, D. G, 1996. Body size and harem size in red-winged
blackbirds: manipulating selection with sex-specific feeders. Evolution
50:2049-2065.

Searcy WA, 1979. Male Characteristics and pairing success in red-winged
blackbirds. The Auk 96:353-363.

Searcy WA, Yasukawa K, 1995. Polygyny and sexual selection in red-winged
blackbirds. Princeton: Princeton University Press.

Seutin G, White BN, Boag PT, 1991. Preservation of avian blood and tissue
samples for DNA analyses. Canadian Journal of Zoology 69:82-90.

Sherman PW, Morton ML, 1988. Extra-Pair Fertilizations in Mountain White-
Crowned Sparrows. Behavioral Ecology and Sociobiology 22:413-420.

Stearns SC, 1989. Trade-offs in life-history evolution. Functional Ecology 3:259-
268.

Stearns SC, 1992. The Evolution of Life Histories. New York: Oxford University

Press.

25

Turner AM, McCarty JP, 1998. Resource availability, breeding site selection,
and reproductive success of red-winged blackbirds. Oecologia 1132140-
146.

Weatherhead PJ, Boag PT, 1997. Genetic estimates of annual and lifetime
reproductive success in male red-winged blackbirds. Ecology 78:884-
896.

Weatherhead PJ, Montgomerie R, Gibbs HL, Boag PT, 1994. The cost of extra-
pair fertilizations to female Red-winged Blackbirds. Proceedings of the
Royal Society of London, Series B 258:315-320.

Weatherhead PJ, Robertson RJ, 1977. Harem size, territory quality, and
reproductive success in the red-winged blackbird (Agelaius phoeniceus).
Canadian Journal of Zoology 55:1261-1267.

Westneat DF, 1988. Male Parental Care and Extrapair Copulations in the Indigo
Bunting. Auk 105:149-160.

Westneat DF, P.W. Sherman, 1993a. Parentage and the evolution of parental
behavior. Behavioral Ecology 4:66-77.

Westneat DF, 1993b. Polygyny and extrapair fertilizations in eastern Red-

winged Blackbirds (Agelaius phoeniceus). Behavioral Ecology 4:49-60.

26

CHAPTER 2:

REDUCED TERRITORIAL AGGRESSION AMONG “DEAR ENEMIES” IN RED-
WINGED BLACKBIRDS: TIT-FOR-TAT BEHAVIOR WITH CRYPTIC
DEFECTORS

Robert Olendorf
Tom Getty
Kellogg Biological Station
Michigan State University
3700 E Gull Lake Dr.
Hickory Corners, MI 49060
Kim Scribner
Department of Fisheries and Wildlife

Michigan State University
East Lansing, MI 48823

ABSTRACT

While competing for similar resources, neighboring territorial males may benefit
by cooperating to achieve common goals. Reduced aggression among
territorial neighbors is one form of cooperation where individuals respect one
another’s territorial boundaries. Each neighbor benefits by being able to reduce
the time and energy spend defending its boundaries. However, if defections
(cheating) are cryptic the frequency of defection might be higher than expected
when defection is more conspicuous. We used simulated territorial invasions to
test for retaliation by red-winged blackbirds in response to defection by

neighbors. We also used analysis of microsatellite regions of DNA to determine

27

the degree of cheating, in the form of cuckoldry, that might occur between
neighbors and to determine if males increased their aggression towards
cuckolding neighbors. We found that males increased aggression towards
neighbors when we simulated intrusions by that neighbor, but not when we
simulated intrusions by strange males. The over all rate of cuckoldry was
relatively high suggesting a high rate of defection among males. Males were
more aggressive towards neighbors that had sired offspring in their nest, but
aggression was also positively correlated with a neighbor’s overall ability to gain
extra-pair copulations. The results of our study support the hypothesis that
reduced aggression among “dear enemies” is a form of reciprocal altruism
using TFT like strategies. We argue that males do not respond directly to
cheating by neighbors but rather react more aggressively towards males that

are “attractive” to females.

28

INTRODUCTION

In territorial species where males hold closely adjacent territories, neighbors
may often be in a simultaneously competitive and cooperative relationship.
While competing for resources such as territory, mates or food, neighboring
males also share common enemies such as nest predators and other rival
males. They might therefore benefit by working together to accomplish shared
goals. Neighboring males locked in this simultaneously competitive and
cooperative relationship are known as “clear enemies” (Fisher, 1954; Getty,

1987).

Consistent with this hypothesis is the finding that territorial neighbors often
behave less aggressively with each other than with strangers at their territorial
boundary (Brindley, 1991; Stoddard et al., 1991; Ydenberg et al., 1988). By
reducing aggression with each other, “dear enemies” may benefit by reducing
the amount of time and energy each male spends guarding its territory.
However, there is a temptation for each neighbor to cheat by either unilaterally
expanding its territory or sneaking copulations with its neighbor’s mates and

realizing at least a short term gain (Getty, 1987).

Situations where cooperation is mutually beneficial but there is still a temptation
to cheat are often modeled in game theory using the Prisoner’s Dilemma (PD).

The PD was first developed in the economic and social sciences by Von

29

Neuman (1953). The PD was subsequently adapted for use in evolutionary
theory by Axelrod and Hamilton (1981) who developed the model using the
framework of evolutionarily stable strategies (Maynard Smith, 1982). In the PD,
two players may either cooperate or defect. If both players cooperate, they both
receive a high payoff (R) while mutual defection rewards both players with a
lower payoff (P). If one player defects while the other cooperates the detecting
player receives the highest possible payoff (T) while the cooperating player
receives the lowest possible payoff (S). In other words, the payoff structure
must conform to the inequality (T>R>P>S) (Von Neuman and Morgenstein,
1953). A second condition (T + S < 2R) is frequently imposed as well. This
condition ensures that individuals cannot do better than R by alternating
between the T and S payoffs. When the game is played only once or a
determinate number of times, defection (ALLD) is the only ESS. On the other
hand, if there is some probability of future interactions the game is transformed
into a new meta-game known as the Iterated Prisoner’s Dilemma (IPD)
consisting of repeated bouts of the PD. In the IPD, ALLD is no longer the only

effective strategy to play (Axelrod, 1980a; Axelrod, 1980b).

The most commonly discussed strategy for the IPD is tit-for tat (TFT) (Axelrod,
1980a; Axelrod, 1980b; Axelrod and Hamilton, 1981 ). TFT is an example of
reciprocal altruism where individuals take turns performing altruistic acts
towards each other (Trivers, 1971 ). TFT is a good strategy to play in the IPD for

several reasons. It is a cooperative strategy allowing it to obtain the

30

cooperator’s payoff in a population of cooperators. It cooperates on the first
move and cooperates if the opponent cooperates on subsequent moves. The
cooperative nature of the TFT strategy allows it to obtain the
Cooperate/Cooperate payoff in a population of TFT players or other cooperative
strategies. Just as importantly, it is retaliatory strategy. If the opponent defects
(cheats), TFT immediately retaliates by detecting in the next round. This aspect
of the TFI' strategy limits the damage a detecting strategy can inflict and allows
a population of TFT to resist invasion by uncooperative strategies. This is
because an uncooperative strategy would achieve higher fitness in any single
interaction with TFT, TFT maintains higher overall fitness if the frequency of
individuals playing TFT in a population is above some critical threshold.
Additionally, the more likely individuals are to interact in the future the less
impact the initial defection has, further enhancing the stability of TFT. Finally,
TFI' is a forgiving behavior. It the opponent cooperates after it has detected,
TFT will cooperate in the following round (Axelrod and Hamilton, 1981). This
provides some protection against mistakes if the opponent continues to

cooperate.

In order to determine if an apparently altruistic behavior is a case of reciprocal
altruism, it is necessary to show retaliation in response detection (Connor,
1986; Rothstein and Pierotti, 1988). Simply demonstrating reciprocity is
insufficient because many examples of by-product mutualism appear to be

reciprocal in nature (Connor, 1986). By-product mutualism is a special case of

31

mutualism where a behavior appears to be altruistic on the surface, but on
closer inspection the behavior actually benefits the “altruist” immediately and
only benefits the recipient as a side effect. By-product mutualism frequently
appears reciprocal as well, in which case it is referred to as pseudo-reciprocity
(Connor, 1986). Lack of retaliation in response to defections would suggest that
the behavior is in fact a form of mutualism and there is no temptation to cheat

(Connor, 1986; Connor, 1995; Rothstein and Pierotti, 1988).

Many studies have shown retaliation in response to induced or simulated
defections (reviewed in Dugatkin, 1997). For instance, predator inspection in
guppies, Poecilia reticulata, (Dugatkin, 1991; Dugatkin and Alfieri, 1991a;
Dugatkin and Alfieri, 1991 b) and sticklebacks, Gasterosteus aculeatus,
(Milinski, 1987), food sharing in vampire bats, (Wilkinson, 1984) and reduced
aggression among territorial hooded warblers, Dendroicha petechia, (Godard,
1993). Godard (1993) showed that hooded warblers, Dendroica citrina, use TFT
like strategies in territorial relationships. Male hooded warblers increased their
aggressiveness towards playbacks of a neighbor’s song at their common
boundary after simulated intrusions by that neighbor, but not after simulated
intrusions by a strange male. To date this is the only experimental
demonstration of TFT like strategies being used in reduced aggression among

“dear enemies”.

32

 

Most studies of “dear enemy" relationships emphasize that neighbors respect
each others borders by not expanding their own territory. While this is likely
true, neighbors may also respect one another’s boundaries by refraining from
seeking extra-pair copulations (EPCs) with their neighbor’s mates. Many
studies in recent years have shown that extra-pair paternity is common in many
species of birds (summarized in Moller, 2000). The cryptic nature of cuckoldry
could allow a higher level of cheating within a stable system of cooperation.
TFT and similar strategies are generally considered “honest” strategies but little
attention has been paid to the effect that an inability to detect defectors might

have on cooperation.

In this study, we determined if red-winged blackbirds (Agelaius phoeniceus)
exhibit TFT-like behavior by reducing their aggression toward their neighbors.
We used simulated invasions by neighbor and stranger males to test for levels
of retaliation in response to detections in “dear enemy” relationships. In
addition, determined whether males alter their behavior towards their neighbor
in response to simulated invasions the day following the trial by observing
natural territorial behavior before and after experimental manipulations. We also
looked for evidence of cheating by determining the overall rate of extra-pair
paternity in the population using analysis of microsatellite loci to establish
paternity of nestlings. Red-winged blackbirds are known to have high level of
extra-pair paternity (Gray, 1997; Weatherhead et al., 1994). Finally, we tested

the hypothesis that “dear enemy” relationships are influenced by extra-pair

33

paternity. If males are able to detect extra-pair paternity, then they should

increase their level of aggression toward cuckolding males.

METHODS

Study Site and Population

This study was conducted at the Kellogg Biological Station Experimental Pond
Facility, Hickory Corners, Michigan (42° 24' N, 85° 24' W) from 1998 and 1999.
Each of 18 ponds was approximately 30 m in diameter and 3 m deep. The
ponds were arranged in three rows of six. Within rows, ponds were
approximately 5 meters apart and rows were spaced approximately 10 meters
apart. The margin of each pond was densely vegetated, predominately with

cattails (Typha latifolia).

Although the ponds were constructed primarily for aquatic research, they were
consistently colonized by red-winged blackbirds. During this study, one to four
males settled each pond and one to six females settled on each territory. A
chain link fence surrounded the facility keeping out most predators, therefore

nesting success was very high (near 100%).

34

 

Capture of birds, marking and blood collection

We began capturing males as they arrived using both walk-in traps baited with
corn and with mist nets. Bait piles for the walk-in traps consisted of cracked
corn and were placed so that each pile was equidistant from four ponds, except
on the edges of the array where they were placed equidistant from two ponds.

Males and females from all adjacent ponds readily visited the bait piles.

Upon capture, we marked it each bird with a unique combination of three
colored leg bands and a numbered aluminum band. Age and sex were
determined using plumage criteria (Pyle, 1997). In 1998, we drew 50-80 ul of
blood from the brachial vein of all territorial males and immediately stored the
sample in 800 pl of “Queen’s” lysis buffer (Seutin et al., 1991). We drew
approximately 0.5 pl of blood from nestlings when they were 8 days old and

immediately placed samples in lysis buffer.

We mapped each male’s territory using behavioral criteria. The entire pond was
defined as a male’s territory when he was the only resident male on that pond.
On ponds with more than one male, we defined territorial boundaries where two
neighboring males counter-sang (both males within 10 m of each other and
singing in alternate order), at the limit of the resident’s movement or at the

center of the overlap of activity between two neighboring males.

35

We searched for nests daily after the arrival of females. We located most nests
as they during construction and all nests were located within a week of
completion. Therefore, we are reasonably certain that we found all nests. We
marked the location of each nest with flagging tape at the edge of the pond and
we monitored nests every 4 to 8 days until it’s fate was determined. The
majority of first brood nests were started on or near May 1. Although females
frequently attempt a second clutch, we incorporated data from only the first

brood to limit pseudo-replication and differences among broods.

Observation of territorial behavior

We observed male territorial behavior in 1998 and 1999. In 1998, we observed
each male twice in random order from a blind approximately 10 m from the
male’s territory. All observations were made between 6 and 10 am during the
first week of May when egg laying was at its peak. In 1999, we observed all
territorial males for 20 minutes (twice for 10 minutes) in late April, just prior to
the experiments described below. This is slightly before the period when males
were most susceptible to being cuckolded (Gray, 1997) but allowed us to
complete pre-trial observations so that we could perform experiments when egg
laying was at it’s peak. We also performed and additional 20 minute observation

the day following the experiment.

36

An observational period consisted of a 3-minute acclimation period followed by
the observational period. We recorded the position of the male, time spent at
that position, the number of songs, song displays, counter-songs, fights and
chases at that position and whether the male was foraging or vigilant. We also
recorded the identity of the other male for counter-songs, fights and chases. In
the few instances where the male was absent for an observation, we returned

the next day to observe that male.

Paternity analyses

We determined paternity of nestlings using six microsatellite loci. Four of the
loci (Om 5, Om 10, Om 21, Om 31) were developed for great-tailed grackles
(Ouiscalus mexicanus) (Gibbs et al., 1997), Dpu 16 for yellow warblers
(Dendroica petechia) (Dawson et al., 1997) and Map 10 for brown-headed

cowbirds (Molothrus ater) (Gibbs et al., 1997).

DNA was extracted from the blood samples using Proteinase K digestion
followed by extraction in 7.5M NH4AOC and precipitation in isopropyl alcohol.

The DNA was washed once more in 70% ethyl alcohol.

We assayed genetic variation at these loci using PCR amplification in 25 uL
reaction volumes. Two different reaction conditions were required for the six
loci. Om 10, Dpp. 16 and Map 10 were amplified using 250 ng of template DNA,

37

2 pmol of each primer (fluorescently labeled fonNard primer), 500 uM dNTPs,
and .75 U Taq polymerase. Om 5, Om 21 and Om 31 were amplified using 125
ng template DNA, 1.25 pmol of each primer (fluorescently labeled fonlvard
primer), 625 uM dNTPs and .75 U Taq polymerase. All reactions were
performed in buffer containing 10 mM Tris-HCI pH 8.3, 1.5 mM MgClz, 30mM
KCI, 0.01% gelatin, 0.01% NP-40, 0.01 Triton X-100. The PCR product was run
on 6% polyacrylamide gels and visualized using an FMBIO gel scanner. We
assigned genotypes to all individuals based on size comparisons with internal
standards and individuals of known genotype run on the same gel. All gels were
scored by hand by two authors (RO & KS) and verified using FMBIO image

analysis software.

We included only territorial males as potential fathers, since previous work has
shown that floating males rarely gain fertilizations (Gibbs et al., 1990). We were
able to establish paternity for all but two of 252 offspring for which we obtained
DNA. For 217 of the young, there was only one non-excluded candidate parent.
We determined the paternity of the remainder of the offspring employing
maximum likelihood methods using Cervus (Marshall et al., 1998). We
performed maximum likelihood estimates of the remaining offspring using error
rates of 0% and 0.5% resulting in probabilities of false inclusion (assigning the
wrong father to the nestling) of 0.7% and 10.5% respectively. We were able to

assign paternity to most of the remaining offspring with 95% confidence level

38

(26 offspring). but we also included assignments at the relaxed confidence of

80% (7 offspring).

Recording and editing of songs

We recorded the territorial songs of all territorial males in early April using a
Sony Audio Acoustica 815a shotgun microphone and a Marantz PMD222 tape
recorder. All recordings were made from a blind between 6 am and 10 am on

calm days.

We uploaded several examples of each male’s territorial song into .wav files at
16 bit resolution. We then filtered the samples to eliminate background noise
with minimal distortion of the signal and standardized all songs to equal
maximum amplitude using Cool Edit© sound editing software. We constructed
playback files for each male by splicing 10 seconds of silence between each
example of a male’s song (minimum of 4 examples/male) resulting in a song

rate of 6 songs/minute. This is within our observed range of male singing rate.

Simulated Defection Tn'a/s

We performed simulated territorial intrusions using a modified version of the

playback experiment performed by Godard (1993; Figure 1) to test whether

39

males retaliated in response to detection by a neighbor. We conducted all trials
during the first week of May after we had observed every male for 20 minutes.
We conducted all trials between 6 and 10 am and only used males that had
neighbors on the same pond. We performed all trials from a blind approximately
10 m from the focal male’s territory using a laptop computer running Cool Edit©
software and Panasonic battery powered computer speakers to generate the
stimulus. We set the volume by ear so that it was similar in volume to natural

male songs and volume was kept constant throughout the experiment.

Prior to running trials, we began catching males using walk-in traps. Upon
capture of a suitable male designated as the neighbor male and chose another
male on the same pond as focal male. In order to avoid pseudo-replication we
used each male only once as a focal male and only once as a neighbor male.
We then assigned the pair to experimental and control groups (explained
below) in alternating order. The neighbor male held in a covered cage in a
building nearby so that it would not interact with the focal male during the

experiment.

Each trial consisted of three playbacks referred to as pre-invasion, invasion and
post-invasion playbacks (Figure 1). We waited 15 minutes after setting up the
apparatus to perform the first playback and we waited 45 minutes between

each successive playback. Each playback lasted 3 minutes.

40

  
   

Focal Mole
ajow JoqufigaN

   

O
0 45minsfi 45mins (E)

Pre-Experiment
Observations

 

 

A A A A A A
April 24th 6; May 1st -¢\ .& May 8m
«‘9’ o (6‘ s“ s

196‘ 6'60 4° (‘9‘ 8&0
@fiée’d do o‘vkfied ‘9
Q o oqoé 090

Figure 1. Schematic and time line of the experiment. Two territories
are shown. Pre-experiment observations were performed between
April 24 and April 30. All trials were conducted between May 1 and
May 8. Numbers in the pond schematic indicate the order playbacks.
Number1 is the pre-invasion trial, number two is the invasion trial
and number 3 is the post-invasion trial. Circles indicate that the
neighbors song was played at that position, the star indicates that a
stranger’s song was played. The time line shows an example of the
timing of the observations and playbacks for one trial.

41

We placed the speaker on the neighbor's side of the territorial boundary for pre-
and post-invasion playbacks while in invasion trials we placed the speaker
approximately 10 m within the focal male’s territory. In experimental treatments,
we played the neighbor’s song in all three playbacks (NNN). For control
treatments, we used the neighbor’s song for pre and post invasion trials, but
used a stranger’s song for invasion trials (NSN). The stranger’s song was a
male’s song from the same population but not adjacent to the focal male’s
territory. To avoid pseudo-replication a male was used only once as a
stranger’s song (Kroodsma, 1989; Kroodsma et al., 2001). By using a song
from the same population for a control, we ensured that the song could

potentially be perceived as a viable threat to the focal male’s territory.

For each trial we recorded latency of response to the playback (defined as an
approach of 10 m or half the distance to the speaker) and closest approach to
the speaker. We also recorded the number of songs, displays, hovers over the
speaker, and any calls given by the male. Upon completion of the trial, we
released the captured male and verified that it had returned to its territory by the

afternoon following the experiment.

The day following an experiment, we observed the focal male’s territorial
behavior for 20 minutes as outlined above. We therefore have five points in time
recorded for each replicate: a pre-trial observation, response to pre-invasion,

invasion and post invasion playbacks, and a post-trial observation. In once

42

case, the neighbor male was not present during the post-trial observation so we

eliminated this trial from the analyses.

Statistical Analysis

We used correspondence analysis (CA) to reduce the number of variables
analyzed. CA is similar to principle component analysis (PCA) in that it finds
orthogonal axes (latent variables) that best explain the variation in multi-
dimensional data. Correspondence analysis, however, may be more
appropriate than PCA in situations where the component variables are not
linearly related (Terbraak, 1985). For instance, aggressive behavior towards
neighbors can be measured using several component variables. Some
behaviors may only be exhibited at low levels of aggression (such as songs or
displays) while others may only be exhibited at higher levels of aggression
(such as physical attacks). This situation would violate the assumptions of PCA
but would be well suited to analysis using CA. Additionally CA will often do a

better job of explaining the variance in these situations.

To analyze the observational datawe first determined each male’s territorial
behavior towards each of its neighbors from the behavioral observations. We
determined the amount of time a male spent within 5 m of a neighbor’s territory,
how many songs and displays it performed in that area and how many counter-

songs, chases and fights it had with that neighbor. We then performed separate

43

correspondence analyses for each year of observational data, 1998 and 1999.
Behavioral observations within a year were analyzed together to facilitate

comparisons.

We analyzed both the first correspondent dimension (CD1) and the second
correspondent dimension (CD2) (Table I). CD1 accounted for 45-50% of the
variation in the data and the component variables always loaded such that CD1
was associated with increased aggression. We therefore refer to CD1 as
aggression. CD2 did not yield consistent results between years for the
observational data. Additionally, CD2 never yielded significant results in any of
the statistical tests for observational or experimental data. We therefore do not

refer to it further.

To test for treatment effects in the experiment we used repeated measures
analysis of variance (repeated measures ANOVA). Each playback or
observation was considered a repeated measure of experimental or control
treatments. We used pooled variance t-tests to test for significant differences
within playbacks or observations when the repeated measures ANOVA yielded

significant differences.

Since each male had several neighbors, there were multiple pair-wise
observations for each male. We used nested ANOVA to test for differences in

aggression between males towards neighbors who had cuckolded them and

44

 

Table l. Component loadings and percent of variance explained
by the correspondence analyses performed on observational data
in 1998 and 1999 (a) and experimental data (b).

 

 

CD1 CD2 CD1 CD2
Time Spent at Border 0.18 0.03 0.29 0.77
Time Spent Foraging -0.21 0.20 -0.56 -0.65
Time Spent Off Territory -0.17 0.04 -0.24 005
Songs at Border 0.28 0.02 0.81 -0.07
Counter Songs 0.16 0.06 0.48 -0.05
Dis plays 0.28 0.04 0.57 -0.08
% Variation Explained 44.49 22.76 48.13 25.35

 

 

 

b.
Experimental Data
1999.00

CD1 CD2
Latency of Response 056 -0.43
Closest Approach -0.55 034
Songs 0.05 0.21
Hovers 0.72 -0.38
Dis plays 0.51 0.68
Calls 0.68 -0.48
% Variation Explained 50.85 28.40

 

45

males that had not. Levels of aggression between cuckolding and non-
cuckolding neighbors were nested within each male. We used analysis of
covariance to test for a relationship between aggression by a male towards a
neighbor and the neighbor’s ability to gain extra-pair fertilizations (EPFS) across
the entire population and aggression. In this case, each male was used as a
categorical variable and number of EPFs obtained by the neighbor was the

covariate. All ANOVAS were run as random effects models in SAS.

RESULTS

Males responded more aggressively to playbacks at the territorial boundary 45
minutes after simulated intrusion by a neighbor than after simulated intrusion by
a stranger (repeated measures ANOVA: Treatment Fm = 12.914, P < 0.005;
Trial F222 = 0.066, NS; Treatment * Trial F222 = 15.103, P < 0.001; Figure 2).
Males from both treatment groups behaved identically in the pre-invasion trial (t-
test: t 11:1.036, N.S.). Males behaved more aggressively in both the invasion
trial (t-test: t 11 = 4.276, P < 0.001) and in the post-invasion trial (t-test: tn =
6.084, P < 0.001). Although overall aggression differed significantly among
treatments, the component behaviors did not differ significantly (t-test: t1o, MS
in all cases). For each variable, however, the difference between treatments
was in the predicted direction, the cumulative effect of which leads to the
significant differences observed when analyzing aggression. Males therefore

increased aggressive behavior against playbacks of their neighbor’s song after

46

 

2 . . Invasion
A III Experiment
.— I 1
8 1.5 - 1
c
.9 1 ~
in
3 Boundary
6: 0.5 - .-
2 N N S N N N
l

 

Treatment ***
Bo d
a Control *** “" “I”

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Pre-Invasion Invasion Post-Invasion
Playback

Figure 2. Aggression by focal males toward playbacks in
response to experimental treatment. Experimental treatments
simulated territorial invasions by neighbors, control
treatments simulated invasion by stranger males. Position of
the speakers is given above each pair of bars. The song
played in a playback is shown inside or above each bar, N =
Neighbor, S = Stranger. Error bars give standard error. N = 6
in both treatments. *** P < 0.001.

47

simulated invasion by the neighbor, but did not exhibit increased aggression to
the playbacks of their neighbor’s song after simulated invasions by strangers.
Additionally, focal males did not react aggressively against stranger playbacks
during the invasion trials but reacted very aggressively against neighbor

playbacks in invasion trials.

The results above show that males react more aggressively to playbacks of
their neighbor after detection, but these results do not show that males alter
their behavior towards their actual neighbor. We therefore used observations of
territorial behavior between the focal and neighbor male before and after
simulated defections. The level of aggression toward neighbor males increased
significantly for both treatments, however the increase was greater for males in
the experimental group (repeated measures ANOVA: Treatment FUD = 2.147,
NS; Trial F130: 15.551, P < 0.001; Treatment * Trial Fug: 17.425, P < 0.005;
Figure 3). Except for display rate, which differs slightly in the other direction, the
component behaviors differ in the expected direction but not significantly (t-
test", MS. in all cases). Males therefore not only respond more aggressively
towards playbacks of their neighbor’s song in response to defection by their
neighbor, but also behave more aggressively towards their neighbor the day

following the experiment.

We collected blood from 252 offspring, 36.2% of which were products of extra-

pair fertilizations. Of 93 nests, 52.4% contained cuckolded young. To determine

48

 

Aggression (CD1)

 

**
0.5 ~

 

  
  

 

 

 

 

 

 

-0.5 -
'1 ' Treatment
0 Control
-1-5 - I:I Experimental
-2

 

Pre-Trial Post-Trial
Observation

Figure 3. Aggression by the focal male directed towards its
neighbor before and after simulated invasions in response to
treatment. Males in experimental treatments had songs from a
neighbor played within its boundaries. Males in control
treatments had songs from a strange male played within its

boundaries. Error bars give standard error. N = 6 for both
treatments. ** P < 0.01

49

it the level of aggression between males was correlated with cuckoldry, we
analyzed interactions between neighboring males on the same pond or closely
adjacent ponds. To control for distance effects of both aggression and
cuckoldry, we did not include neighbors from different rows. The territorial
behavior between males and the paternity analysis were performed separately
and effectively blind of each other. The results show that males are more
aggressive towards neighbors that have successfully cuckolded them (nested
ANOVA: male F1849: 1.24, NS; cuckoldry(male) F1942: 2.32, P < 0.05; Figure

4).

We reasoned that total EPFS obtained by a male across the entire population
would correlate with a male’s attractiveness to females. We therefore tested to
see if there was a correlation between aggression towards a neighbor and the
total number of EPFs that neighbor obtained. There was a positive correlation
between aggression towards a neighbor and that neighbor’s ability to gain EPFs
(ANCOVA; male F19, 59 = 0.826, NS; neighbor’s EPFs F159 = 4.465, P < 0.05;

male * neighbor’s EPFs F1g_ 59 = 0.828, NS; Figure 5).

DISCUSSION

The results of our study parallel those of Godard’s (1993) study. Red-winged
blackbirds, like hooded warblers, use TFT-like strategies in reducing their

aggression toward their neighbors. In addition, our study extends Godard’s

50

Aggression (CD1)

 

 

 

 

 

 

 

 

No Yes
Cuc koldy

Figure 4. Average aggression by each male toward
neighbors who sired young on its territory and those
that had not. Each line represents one male and each
circle represents the average aggression towards
neighbors in that category. Three males are not shown
because one was cuckolded by all of its neighbors and
two males were not cuckolded by any of their
neighbors. N = 19 males.

51

 

 

Aggresion (001)
O

 

 

 

4 6 8 10 12 14 16

Number of EPFs Obtained by Neighbor

Figure 5. Aggression by territorial males in response the
total number of extra-pair fertilizations obtained by a
neighbor across the entire population. The circles show
all pair wise interactions (N=99). The dark dashed line
shows the over all trend the solid gray lines give show the
trend lines for each individual male (N: 22).

52

work in several ways. Although Godard mentions that hooded warblers seemed
to act more aggressively towards their neighbors after simulated intrusions, this
was not tested explicitly. We show that the response to detection is not just
directed towards the experimental playbacks, but also towards the neighboring
male the day following the experiment. We also evaluated effect of cheating on
“dear enemy” relationships. Our study suggests that males are sensitive to
cheating by their neighbors and increase their aggression towards cuckolding
neighbors. This response may, however, reflect the capability of males to
accurately determine the ability of its neighbors to obtain EPFs rather than a
male’s ability to detect cuckoldry directly. Males appeared to increase their
defense towards “sexy” neighbors that were able to gain large numbers of EPFs
overall and would presumably be better able than “less sexy” neighbors to

cuckold his offspring.

Do red-winged blackbirds play TFT like strategies?

The three crucial traits that make TFT such an effective strategy to play in the
IPD are; 1) it cooperates with cooperative strategies 2) it retaliates in response
to detection and 3) forgives when the neighbor again cooperates (Axelrod and
Hamilton, 1981). This study suggests that red-winged blackbirds have at least 2
of these three attributes, they are initially cooperative and retaliate in response

to detection. They do not, however, appear especially quick to forgive.

53

 

 

 

Males were generally cooperative in that their levels of aggression towards their
neighbors were low compared to their aggression towards males that have
detected. In both pre-trial observations and in pre-defection trials, focal red-
winged blackbirds showed little aggression towards one another or to the
playback. Most of their time was spent singing from a perch well within their
territory, foraging or chasing females. In contrast, males increased their
aggression markedly in response to the invading song from a neighboring male,

while not reacting at all to the song of an invading stranger.

Males retaliate against cheaters following territorial invasions. In response to
simulated invasions by their neighbors, male red-winged blackbirds significantly
increased their level of aggression towards that neighbor in post-invasion trials
and continued to behave more aggressively the following day. This increase
was pronounced, focal males approached the speaker very closely, often
perching on top of it and appeared to search the area closely for the offending

male. This result is consistent with Godard’s (1993) results.

Although we did not specifically study forgiveness, red-winged blackbird males
do not appear to quickly forgive an invasion by their neighbor. In this study,
where we simulated invasion by neighboring males, focal males continued to
behave more aggressively towards their neighbors on the day following the
experiment compared to males where we simulated invasions by stranger

males. Although not explicitly testing this either, Godard (1993) noted that

54

males also appeared to show increased aggression towards neighbors
following simulated invasions. The concept of forgiveness, however, is not well
developed. Clearly returning to background levels of aggression would be
interpreted as forgiveness, but there is no a priori way of knowing how long

increased levels of aggression should last.

Perhaps the best interpretation of our result is that after the apparent defection
by a neighbor, the pair must start over and negotiate reduced aggression anew.
This process may take a few days through incremental investment in
cooperation (Roberts and Sherratt, 1998). A better framework to understand
forgiveness in continuous games (?) may be to model forgiveness using
negotiation rules as opposed to action rules. Negotiation rules can model the
rate at which aggression decreases as well as the final level of reduced
aggression. Action rules, on the other hand, model the response to defections

(McNamara et al., 1999).

Cheating among “Dear Enemies”

Implicit in TFT is a low rate of cheating (detecting), because cheating should be
selected against in a TFT (or TFI' like) population. Within any single interaction,
a cheater will always achieve higher fitness than an individual playing TFT.
However, TFT individuals enjoy a higher long-term payoff in the IPD, because

every time two TFT strategists meet they receive the cooperator’s payoff. It the

55

frequency of TFT individuals is high enough, TFT will be favored over
uncooperative strategies (Axelrod and Hamilton, 1981). This of course assumes
that cheating is detected. If cheating is difficult to detect, as is the case with
cuckoldry, cheaters may go undetected. However, if males invade their
neighbor’s territory to expand their own territory, cheating should easily be

detected.

The resource being guarded in our study or the motivation behind any territorial
invasions that might occur is unknown. Godard (1993) and Getty(1987) assume
that males are guarding their physical territory. However, their reasoning does

not exclude the possibility that they may be guarding access to their females as

well.

Our study shows that males behave more aggressively towards neighbors that
have cuckolded them compared to males that had not. This is a surprising

result given that extra-pair copulations (EPCs) are cryptic events. Our study and
Gibbs’ (1990) show that most EPFs come from neighbors. In addition, Gray
(1996) observed that 78% of EPCs occur off the territory and that females
appear to actively solicit them and control the level of extra-pair paternity in their
young. These results suggest that males may be unable to directly determine
which males have cuckolded them. Males may, however, be able to assess
which neighbors are most attractive to females and behave more aggressively

towards those “sexy” males.

56

How males are able to assess on another is uncertain. Larger and older males
tend to sire more total young (Weatherhead and Boag, 1995), however it is
unknown if males are responsive to size. The evidence for physical characters
contributing to territory acquisition is weak (Eckert and Weatherhead, 1987a;
Eckert and Weatherhead, 1987b) and may therefore not factor into a male’s
assessment of a neighbors ability to gain EPCs. Alternatively, males may
assess other male’s behavior to gauge their resource holding potential
(Freeman, 1987) through behavioral interactions with other males and may use
similar strategies and monitor their neighbors’ behavior to assess a neighbors’

ability to achieve EPCs.

Despite the fact that red-winged blackbirds do not appear to be particularly
honest in one aspect of their “dear enemy” relationships, they continue to
maintain relatively low levels of aggression with each other. Perhaps the
benefits of reduced aggression are great enough that males will tolerate a
certain level of cuckoldry. In environments were there was a strong advantage
to cooperating, guppies were more tolerant of detection (more forgiving)
compared to environments were cooperation was less beneficial (Dugatkin and
Alfieri, 1992). Additionally, since females appear to seek EPCs away from the
territory, there may be little males can do about it except to seek their own

EPCs

57

Neighbor recognition and “Dear Enemies”

Several studies have shown that territory owners react more strongly to
strangers on their border than to neighbors (Brindley, 1991; Fox and Baird,
1992; Godard, 1991; Stoddard et al., 1991). By reducing aggression towards
their neighbors, males can reduce the cost of territorial defense. On the other
hand, strange males on the boundary may be new males attempting to
establish a territory that could intrude onto the resident male’s territory and are

therefore present more of a risk.

One striking result of our study is the focal male’s lack of interest in strangers
present their territory. Although our study never determined the behavior of
males towards strangers on their territorial boundaries, hooded warblers react
strongly to strangers on the border (Godard, 1991) but not to strange males are
invading (Godard, 1993).lt is possible that strange males within the territory are
not perceived as a threat. In fact, in our study non-territorial males were
frequently observed on territories and were seldom harassed by the resident
male. We also never observed new territories being added after initial
establishment in our population. Perhaps stranger males in the territory are not
perceived as a threat because they are unable to usurp an entire territory and

are unlikely to gain EPCs from the resident females.

58

This is the first study to show that the increased aggression displayed to
playbacks in simulated invasions is also transferred to the neighboring male
that appeared to invade. While males respond strongly to the songs of
neighbors on their territory, they do not react strongly to the songs of strange
males on their territory. This may be a cost saving trait as stranger males may
present little risk to territorial males. We found a relatively high level of
cuckoldry in this population that we interpret to be cheating and there is a
positive relationship between cuckoldry and levels of aggression. Males behave
more aggressively to males that have succeeded in gaining EPCs on its
territory than those that did not. However, males also behaved more
aggressively towards males that were most successful in gaining EPCs overall
suggesting that males do not detect cuckoldry directly but rather are able to
asses a neighbors attractiveness to females and react more aggressively

towards attractive males.

LITERATURE CITED

Axelrod R, 1980a. Effective choice in the prisoner's dilemma. Journal of Conflict
Resolution 2423-25.

Axelrod R, 1980b. More effective choice in the prisoner's dilemma. Journal of
Conflict Resolution 24:379-403.

Axelrod R, Hamilton WD, 1981. The evolution of cooperation. Science

211:1390-1396.

59

 

Brindley EL, 1991. Response of European Robins to Playback of Song -
Neighbor Recognition and Overlapping. Animal Behaviour 41 :503-512.

Connor RC, 1986. Pseudo-reciprocity: Investing in Mutualism. Animal
Behaviour 34:1562-1584.

Connor RC, 1995. The benefits of mutualism: a conceptual framework.
Biological Review 70:427-457.

Dawson RJG, Gibbs HL, Hobson KA, Yezerinac SM, 1997. Isolation of
microsatellite DNA markers from a passerine bird, Dendroica petechia
(the yellow warbler), and their use in population studies. Heredity 79:506-
514.

Dugatkin LA, 1991. Dynamics of the TIT FOR TAT strategy during predator
inspection in the guppy (Poecilia reticulata). Behavioral Ecology and
Sociobiology 29:127-132.

Dugatkin LA, 1997. Cooperation Among Animals. New York: Oxford University
Press.

Dugatkin LA, Alfieri M, 1991a. Guppies and the TIT FOR TAT strategy:
preference based on interaction. Behavioral Ecology and Sociobiology
28:243-246.

Dugatkin LA, Alfieri M, 1991b. Tit-For—Tat in guppies (Poecilia reticulata): the
relative nature of cooperation and detection during predator inspection.

Evolutionary Ecology 5:300-309.

60

Dugatkin LA, Alfieri M, 1992. lnterpopulational differences in the use of the Tit-
For-Tat strategy during predator inspection in the guppy, Poecilia
reticulata. Evolutionary Ecology 6:519-526.

Eckert CG, Weatherhead PJ, 1987a. Competition For Territories in Red-Winged
Blackbirds - ls Resource-Holding Potential Realized. Behavioral Ecology
and Sociobiology 20:369-375.

Eckert CG, Weatherhead PJ, 1987b. Male Characteristics, Parental Quality and
the Study of Mate Choice in the Red-Winged Blackbird (Agelaius-
Phoeniceus). Behavioral Ecology and Sociobiology 20:35-42.

Fisher J, 1954. Evolution and bird socialilty. In: Evolution as a process (Huxley
J, Hardy AC, Ford EB, eds). London: Allen & Unwin.

Fox SF, Baird TA, 1992. The Dear Enemy Phenomenon in the Collared Lizard,
Crotaphytus- Collaris, With a Cautionary Note On Experimental
Methodology. Animal Behaviour 44:780-782.

Freeman S, 1987. Male Red-Winged Blackbirds (Agelaius-Phoeniceus) Assess
the RHP of Neighbors By Watching Contests. Behavioral Ecology and
Sociobiology 21:307-31 1.

Getty T, 1987. Dear enemies and the Prisoner's Dilemma: Why should
territorial neighbors form defensive coalitions? American Zoologist
27:327-336.

Gibbs HL, Miller P, Alderson G, Sealy SG, 1997. Genetic analysis of brown-
headed cowbirds Molothrus ater raised by different hosts: data from

mtDNA and microsatellite DNA markers. Molecular Ecology 6:189-193.

61

Gibbs HL, Weatherhead PJ, Boag PT, White BN, Tabak LM, Hoysak DJ, 1990.
Realized reproductive success of polygynous Red-winged Blackbirds
revealed by DNA markers. Science 250:1394-1397.

Godard R, 1991. Long-Term-Memory of Individual Neighbors in a Migratory
Songbird. Nature 350:228-229.

Godard R, 1993. Tit for tat among neighboring hooded warblers. Behavioral
Ecology and Sociobiology 33:45-50.

Gray EM, 1996. Female control of offspring paternity in a western population of
red-winged blackbirds (Agelaius phoeniceus). Behavioral Ecology and
Sociobiology 382267-278.

Gray EM, 1997. Do female red-winged blackbirds benefit genetically from
seeking extra-pair copulations? Animal Behaviour 53:605-623.

Kroodsma DE, 1989. Suggested experimental designs for song playbacks.
Animal Behaviour 37:600-609.

Kroodsma DE, Byers BE, Goodale E, Johnson S, Liu WC, 2001.
Pseudoreplication in playback experiments, revisited a decade later.
Animal Behaviour 61 :1029-1033.

Marshall TC, Slate J, Kruuk LEB, Pemberton JM, 1998. Statistical confidence
for likelihood-based paternity inference in natural populations. Molecular
Ecology 7:639-655.

Maynard Smith J, 1982. Evolution and the Theory of Games. Cambridge:

Cambridge University Press.

62

McNamara JM, Gasson CE, Houston Al, 1999. Incorporating rules for
responding into evolutionary games. Nature 401 :368-371.

Milinski M, 1987. TIT FOR TAT in sticklebacks and the evolution of cooperation.
Nature 325:433-435.

Moller AP, 2000. Male parental care, female reproductive success, and
extrapair paternity. Behavioral Ecology 1 1 :161-168.

Pyle P, 1997. Identification guide to North American birds. Bolinas, California:
Slate Creek Press.

Roberts G, Sherratt TN, 1998. Development of cooperative relationships
through increasing investment. Nature 394:175-179.

Rothstein SI, Pierotti R, 1988. Distinctions among reciprocal altruism, kin
selection, and cooperation and a model for the initial evolution of
beneficent behavior. Ethology and Sociobiology 9:189-209.

Seutin G, White BN, Boag PT, 1991. Preservation of avian blood and tissue
samples for DNA analyses. Canadian Journal of Zoology 69:82-90.

Stoddard PK, Beecher MD, Horning CL, Campbell SE, 1991. Recognition of
Individual Neighbors By Song in the Song Sparrow, a Species With Song
Repertoires. Behavioral Ecology and Sociobiology 29:211-215.

Terbraak CJF, 1985. Correspondence-Analysis of Incidence and Abundance
Data - Properties in Terms of a Unimodal Response Model. Biometrics
41 1859-873.

Trivers RL, 1971. The evolution of reciprocal altruism. The Quarterly Review of

Biology 46:35-57.

63

Von Neuman J, Morgenstein O, 1953. Theory of games and economic
behavior. Princeton, NJ: Princeton University Press.

Weatherhead PJ, Boag PT, 1995. Pair and Extra-Pair Mating Success Relative
to Male Quality in Red-Winged Blackbirds. Behavioral Ecology and
Sociobiology 37:81 -91.

Weatherhead PJ, Montgomerie R, Gibbs HL, Boag PT, 1994. The cost of extra-
pair fertilizations to female Red-winged Blackbirds. Proceedings of the
Royal Society of London, Series B 258:315-320.

Wilkinson GS, 1984. Reciprocal food sharing in the vampire bat. Nature
308:181-184.

Ydenberg RC, Giraldeau LA, Falls JB, 1988. Neighbors, Strangers, and the

Asymmetric War of Attrition. Animal Behaviour 36:343-347.

64

CHAPTER 3:

COOPERATIVE NEST DEFENSE IN RED-WINGED BLACKBIRDS:
SIMULTANEOUS TESTS FOR RECIPROCAL ALTRUISM AND MUTUALISM

Robert Olendorf
Kellogg Biological Station
Michigan State University
3700 E Gull Lake Dr.
Hickory Corners, Ml 49060
Kim Scribner
Department of Fisheries and Wildlife

Michigan State University
East Lansing, MI 48823

ABSTRACT

Male red-winged blackbirds frequently cooperate in defense against a nest
predator. Previous studies suggest that such cooperation can result from either
reciprocal altruism or mutualism. While theoretically distinct, the two
mechanisms are not mutually exclusive and may act in concert to favor high
levels of cooperation. We tested for both tit-for-tat like strategies (a form of
reciprocal altruism) and by-product mutualism. We used simulated defections to
test for reciprocal altruism and determined paternity based on microsatellite
analysis to examine potential causal mechanisms underlying mutualistic
interactions among males. Results show that male red-winged blackbirds

cooperate in nest defense primarily as a form of reciprocal altruism. In trials

65

where we simulated defections by neighboring territorial males, males reduced
their level of defense against a nest predator at a common territorial boundary
the day following the simulated detection. In contrast, males where defense was
not manipulated increased their level of defense over the same period. We
found no evidence that male red-winged blackbirds defend nests where they
have sired young or that they reduce defense at cuckolded nests on their own
territory. Considered with other studies, the results of this study suggest that
different forms of cooperation may evolve in different populations. This
suggests that other studies that test for both reciprocal altruism and mutualism

may be of considerable value.

66

INTRODUCTION

Frequently when a potential nest predator enters a breeding colony of birds, it is
“mobbed” by more than one adult. By cooperating in nest defense, otherwise
competitive neighbors may benefit from reduced nest predation. However, there
may also be temptation to cheat, relying on others to drive off the nest predator
and avoiding the costs associated with nest defense. Such simultaneously
competitive and cooperative relationships between neighboring males are often

referred to as “dear enemy” relationships (Fisher, 1954; Getty, 1987).

Red-winged blackbirds, Agelaius phoeniceus, are colonial and males often
develop relationships both within and across breeding seasons (Beletsky,
1989). Long-term associations can favor the evolution of “dear enemy”
relationships through reciprocal altruism (Getty, 1987; Trivers, 1971).
Reciprocal altruism is a form of cooperation where an individual helps another
when there is a reasonable expectation that the other individuals will
reciprocate helpful act in the future. With repeated interactions, both individuals

will achieve higher fitness than individuals that don not cooperate.

If cooperative nest defense is primarily a form of reciprocal altruism, then it can
by modeled using the iterated prisoner’s dilemma (IPD). The IPD is an
extension of the prisoner’s dilemma (PD). The PD was first used to model

economic and social behavior (Von Neuman and Morgenstein, 1953) and later

67

applied to evolutionary theory (Axelrod and Hamilton, 1981; Maynard Smith,
1982). Trivers (1971) makes use of the concept in developing the theory of
reciprocal altruism. In the PD and IPD, if both individuals cooperate, they both
receive higher fitness (R) than when both detect (P). If one individual defects
while the other cooperates, the detector receives the highest possible payoff (T)
while the cooperative individual receives the lowest payoff (S). The resulting
scenario conforms to the inequality (T>R>P>S). An additional constraint (2R >
T+S) is often included to insure that two individuals cannot do better than
always cooperating by alternating detecting and cooperating behaviors. Pure
cooperation is not evolutionary stable in the PD, but in the IPD, several stable
cooperative strategies have been found (Axelrod and Dion, 1988; Axelrod and

Hamilton, 1981).

The most commonly discussed of which is the tit-for—tat strategy (TFT),
popularized by Axelrod (1980a; 1980b) and further developed in an evolutionary
context by Axelrod and Hamilton (1981) and as a potential evolutionary stable
strategy (Maynard Smith, 1982). TFT is a strategy where an individual
cooperates on the first move, and copies it’s opponent’s last move thereafter.
TFT is a good strategy in the IPD. In a population of individuals engaging in
TFT-like behavior, or other cooperative strategies an individual always
cooperates, yielding a high payoff. In the presence of a detecting strategy,
however, an individual limits the damage by only cooperating on the first move

and defecting thereafter.

68

Several studies show that birds are capable of recognizing their neighbors
(Brindley, 1991; Godard, 1991; Stoddard et al., 1991). Beletsky and Orians
(1989) show that males with familiar neighbors enjoy higher nesting success
than males with unfamiliar neighbors suggesting that familiarity might favor
cooperation among neighbors. These studies suggest that red-winged
blackbirds might use TFT-like strategies in cooperative nest defense.
Weatherhead (1995), however, found the reverse pattern, whereby males with
at least one unfamiliar neighbor had higher nest success due to reduced nest

predation.

An alternative to reciprocal altruism described above that may be consistent
with the Weatherhead (1995) study is pseudo-reciprocity. Pseudo-reciprocity is
a form of mutualism where individuals appear to take turns helping one another,
while in fact, each individual benefits directly from helping (Connor, 1986;
Connor, 1995). Weatherhead (1994)showed that nests that containing young
from extra-pair matings benefited from higher nest success than nests that
contained only young from the resident male. Weatherhead suggested that
males defend nests on other territories when they have sired young in them.
Gray (1997) showed that males are more likely to help defend nests where they
have sired extra-pair offspring. These helpers would appear to be incurring a

cost by helping at another male’s nests while in fact they are defending their

69

own young from nest predators. Any benefits that accrue to the recipient would

be a by-product of the helper’s behavior.

Pseudo-reciprocity and TFT both appear reciprocal in that individuals help their
neighbors and are in turn helped by their neighbors. Simply demonstrating
reciprocity, therefore, does not distinguish between reciprocal altruism and
mutualism (Connor, 1995; Rothstein and Pierotti, 1988). To demonstrate that a
behavior is an example of TFT it is therefore necessary to show retaliation in
response to detection. Lack of retaliation would suggest that there is no
temptation to defect, that the behavior is directly beneficial to the helper and
would be consistent with mutualism. Demonstrating the benefits, however,
would constitute stronger evidence for mutualism. There are now many studies
that have shown retaliation in response to simulated or induced defections and
there are also several studies that have demonstrated mutualism (see

Dugatkin, 1997, for a complete review)

Although theoretically distinct, mutualism and TFT are not mutually exclusive. It
is feasible that mutualism may support a certain level of cooperation, but higher
levels of cooperation might be achieved through TFT-like relationships acting in
concert with mutualistic interactions resulting in increased frequency of
cooperative interactions. For example, a helper may defend a neighbor’s nest if
the helper had sired young in that nest. This of course would benefit both

individuals directly. That same helper might also help at the nest regardless of

70

the paternity of the young if the neighbor could be counted on to help in the

future.

Although there are many studies that test for either TFT-like behavior or
mutualism, none simultaneously test for both. Experiments that simulate or
induce defections to test for retaliation look for reduced cooperation by the
“duped” individual. If there is a mutualistic component to the interaction,
however, some level of cooperation may remain undetected by simulated
detections. In order to detect mutualistic interactions, it is therefore necessary to
show that there is a benefit to cooperating regardless of the behavior of the

other individual.

In this study, we looked for evidence of both TFT and mutualism in cooperative
nest defense in red-winged blackbirds. To test for TFT-like strategies, we
looked for evidence of retaliation in response to simulated defections. To test
for mutualism, we look for evidence that males are more likely to aid in defense

of nests on other territories in which they have sired young.

71

METHODS

Study Site and Population

This study was conducted at the Kellogg Biological Station Experimental Pond
Facility, Hickory Corners, Michigan (42° 24' N, 85° 24' W).The site consisted of
18 ponds that were approximately 30 m in diameter and 3 m deep. The ponds
were arranged in three rows of six. Within rows, ponds were approximately 5
meters apart, rows were spaced approximately 10 meters apart. The margin of
each pond was densely vegetated predominately with cattail (Typha Iatifolia).
Although constructed primarily for aquatic research, red-winged blackbirds
readily colonized the ponds. During this study, one to four males settled each
pond and one to six females settled on each territory. A chain link fence
surrounded the facility keeping out most predators, nesting success was

therefore nearly 100%.

Collection of Background Data

We captured birds using both walk-in traps baited with corn and with mist nets.
Bait piles were placed so that each pile was equidistant from four ponds, except
on the edges of the array where they were placed equidistant from two ponds.

Males and females from all adjacent ponds readily visited the bait piles.

72

Upon capture, we marked each individual with a unique combination of three
colored leg bands for visual identification in the field and with a numbered
aluminum band. Age and sex was determined using plumage criteria (Pyle,
1997). In 1999 we obtained 50-80 ul of blood from the brachial vein of all
territorial males and samples were immediately stored in 800 pl of “Queen’s”
lysis buffer (Seutin et al., 1991). We obtained approximately 50 ml of blood from
nestlings when they were 8 days old and immediately stored the samples in

lysis buffer.

We used behavior criteria to map each male’s territory. The entire pond was
defined as a male’s territory when he was the only male there. On ponds with
more than one male, we defined territorial boundaries using the following
criteria. We estimated boundaries to be at locations where two neighboring
males counter-sang (both males were within 10 m of each other and sang
repeatedly in response to each other’s songs), at locations marking the limit of
the resident’s movements or at locations approximating the center of overlap

between two neighboring males’ area of activity.

We searched for nests when females arrived. The majority of first brood nests
were started on or near May 1. We located most nests as they were built, and
all nests were located within a week of completion. We found no nests later in

the nesting cycle so we are reasonably certain we found all nests. We marked

73

the location of each nest with flagging tape on the edge of the pond. Nests were
monitored every 4 to 8 days until the fate was determined. Egg and nestling
stages lasted approximately 11 days each and the nesting season ended in

early July when the second broods fledged.

Paternity analyses

We determined paternity of nestlings using six microsatellite loci. Four of the
loci (Om 5, Om 10, Om 21, Om 31) were developed for great-tailed grackles
(Ouisca/us mexicanus) (Gibbs et al., 1997), Dpp 16 for yellow warblers
(Dendroicha petechia) (Dawson et al., 1997) and Map 10 for brown-headed

cowbirds (Molothrus ater) (Gibbs et al., 1997).

DNA was extracted from the blood samples using Proteinase K digestion
followed by NH4AOC extraction and precipitation in isopropyl alcohol. The DNA
was washed once more in 70% ethyl alcohol. We assayed genetic variation at
each of six loci using PCR amplification in 25 uL reaction volumes. Two
different reaction conditions were required for the six loci. Om 10, Dpp. 16 and
Map. 10 were amplified using 250 ng of template DNA, 2 pmol of each primer
(fluorescently labeled forward primer), 500 uM dNTPs, and .75 U Taq
polymerase. Om 5, Om 21 and Om 31 were amplified using 125 ng template

DNA, 1.25 pmol of each primer (fluorescently labeled forward primer), 625 0M

74

dNTPs and .75 U Taq polymerase. All reactions were preformed in buffer
containing 10 mM Tris-HCI pH 8.3, 1.5 mM MgCl2, 30mM KCI, 0.01% gelatin,
0.01% NP-40, 0.01 Triton X-100. The PCR product was run on 6%
polyacrylamide gels and visualized using an FMBIO gel scanner. Genotypes
were assigned to all individuals based on size comparisons with internal lane
standards and individuals of known genotype run concurrently on each gel. All
gels were scored by hand by both authors and verified using FMBIO image

analysis software.

We included only territorial males as potential fathers; previous work has shown
that floating males rarely if ever gain fertilizations (Gibbs et al., 1990). We were
able to establish paternity for all but two of 252 offspring for which we obtained
DNA. For 217 young (87%) there was only one non-excluded candidate parent.
The remainder of the offspring were assigned paternity based on maximum
likelihood estimates of paternity using Cervus (Marshall et al., 1998). We were
able to assign paternity to 26 of remaining the offspring with 95% confidence
level, but we also included assignments at the relaxed confidence of 80% for
seven offspring. The two offspring for which no male could be assigned
paternity were mismatched at two loci or more for all candidate parents.
Although we could still assign paternity based on maximum likelihood

estimates, we felt it was more likely that an unknown male sired these offspring.

Recording of songs

We recorded the territorial songs of all territorial males in early April 1999 using
an Audio Technica AT815a shotgun microphone and a Marantz PMD222 tape

recorder. All recordings were made from a blind from 6 am to 10 am.

After recordings were made, we uploaded several examples of territorial songs
from each male into .wav files at 16 bit resolution. We then filtered as much
noise as was possible from the recordings without reducing the quality of the
song and standardized all songs to equal amplitude using Cool Edit © sound
editing software. We then spliced a minimum of four song examples from each
male together, separating each example with 10 seconds of silence resulting in

a call rate of approximately six songs/minute.

Observation of cooperative nest defense

In order to determine each male’s willingness to defend nests on other
territories, in 1998 we performed presentations of stuffed crows at nests 1 to 4
days after hatching. Red-winged blackbirds substantially increased their
willingness to defend after hatching and we were therefore more likely to
observe helping in nest defense at this time. To minimize affects of repeated

exposure to the stuffed crow (Knight and Temple, 1986), presentations

76

performed on the same day were separated in distance by at least one pond.
All presentations were performed between 6 and 10 am, when males were most
likely to be on their territories. If any of the neighbors were not present for a

presentation, we returned the next day to perform a presentation.

For each trial, we positioned a stuffed crow, mounted on a pole perched
posture, within 1 meter of the nest. The crow was initially covered with a cloth
with string attached to it. We then observed any reaction to the covered crow for
3 minutes from a blind 10 m from the crow. In no case was there any reaction to
the covered crow. After 3 minutes, the cloth was pulled off and we observed

any reaction to the crow by all neighbors for 3 minutes.

For each male responding to the crow, we recorded that male’s identity, latency
to approach, closeness of approach, number of vocalizations of reach type,
hovers, dives and strikes. Red-winged blackbirds are known to use as many as
seven defense calls (Knight and Temple, 1988; Orians, 1961) but for analyses
we kept a tally for only the most common calls, the “Teer”, “Seet” and “T iti”
(Knight and Temple, 1988, and pers. obs. RO). We combined all other calls
together excluding the ubiquitous “Chit” call which red-winged blackbirds use
constantly. The “Chit” call is given almost constantly and is likely a contact call
between male and female (Yasukawa, 1989). We therefore reasoned it would

contribute little information on nest defense.

77

Experimental tests for retaliation

In 1999, we used simulated defections by neighboring males to test for
retaliation. In order to reduce the effects of paternity and placement of nests we
performed all trials at the territorial boundaries between males on the same
pond. Placing the crow at territorial boundaries had the additional benefit that
both males should benefit equally from defense so that any changes in behavior
were attributable to the experiment. All trials were performed in late May when a

substantial number of the nests contained nestlings.

To establish the baseline level of cooperative defense for each pair of males we
first performed a crow presentation at every territorial boundary in the
population (pre-defection presentation). The crow was covered and males were
allowed to acclimate for 3 minutes. We then uncovered the crow and observed
the behavior of both neighbors for 3 minutes. For each male we recorded the
latency of response, closest approach, hovers, dives, strikes, “Teer”, “Seet” and

“T iti” calls as well as summing up all other calls given excluding the “Chit” call.

After establishing initial levels of defense among neighbors, we performed
simulated defections. Before performing experiments, we began capturing
males. When we captured a suitable, it was designated as the neighbor male,
we determined it’s neighbor (the focal male) and assigned them to experimental

or control groups in alternating order. Males were used only once as focal

78

males and neighbor males although some males were used as both focal and
neighbor males. Captured males assigned to control groups were taken to a
nearby building and released. Captured males assigned to experimental groups
were taken to the same building, kept in a covered cage and given generous

supplies of cracked corn and water.

We waited 30 minutes to perform the defection presentation and in the case of
pairs assigned to control groups verified that the neighboring male had returned
to the territory. We placed the covered crow on the pair’s territorial boundary
and placed a set of speakers well within the neighbor male’s territory. We
waited an additional 15 minutes to allow the males to recover from the

disturbance and performed the defection presentation.

In control trials, we played ambient sounds recorded near the pond facility but
well away from any red-winged blackbirds. To avoid pseudo-replication, a
different example of ambient sounds was used for each playback (Kroodsma et
al., 2001). In experimental trials, we played the neighboring male’s territorial
song. The goal in experimental trials was to make it appear as though the
neighboring male was present but refusing to participate in defense against the
crow. We observed the response of the focal male for 3 minutes recording the
same data we recorded during pre-defection presentations. After completion of
this playback, the captured male was released and we verified that it had

returned by the afternoon.

79

The day following a simulated defection presentation, we performed the post-
defection presentation. This presentation was conducted the same as the pre-
defection presentation. The covered, stuffed crow was placed at the territorial
boundary. We retreated to a blind and waited 3 minutes. We then uncovered

the crow and recorded both males’ reaction to the crow for 3 minutes.

Statistical Analysis

Since our behavioral observations consisted of many measures of nest defense
behavior for each individual, we used correspondence analysis (CA) to reduce
the number of dimensions in our data. CA is similar to principal component
analysis (PCA) in that it reduces the number of dimensions in a multivariate
data set. CA however is more appropriate than PCA when the component
variables are not linearly related (Terbraak, 1985). In the case of nest defense,
CA is more appropriate because certain component behaviors may only be
expressed at low levels of aggression, while others may only be expressed at
higher levels of aggression. In addition, CA explains a greater proportion of the

variance under these circumstances.

We performed separate correspondence analyses on data from presentations
made at individual nests(where we tested for a male’s willingness to defend

nests in which they had extra-pair copulations) and data from presentations at

80

territorial boundaries (where we tested for TFT-like relationships among
neighboring males). The correspondence analysis on presentations at territorial
boundaries included data from both the focal and the neighbor male for each
pair of neighbors. In all cases, the component variables correlated with the first
correspondent dimension (CD1) such that an increase in CD1 was associated
with an increase in nest defense (Table I). For clarity, we therefore refer to CD1
as defense. The second correspondent (CD2) dimension never correlated
consistently with the component behaviors between data sets nor was it easily
interpretable (Table 1). Nonetheless, we subjected CD2 to identical statistical
tests as CD1, but CD2 never yielded significant results. We therefore do not

refer to CD2 further.

Since each focal male in the detection experiment was used in pre-defection,
defection and post defection presentations, we used repeated measures
analysis of variance (ANOVA) to analyze the results. Presentations were
blocked within male and we tested for effects of treatment, presentation and the
interaction between the two. Likewise, we analyzed the affect of the
experimental treatment on the neighbor male’s behavior using repeated

measures ANOVA.

To determine if neighboring males were more likely to defend an off territory
nest in which they had sired offspring we used a paired t-test. For each nest or

territory with extra-pair offspring, we paired the genetic father of cuckolded

81

Table l. Component loadings and percent of variance explained
by the correspondence analyses performed on data taken from
crow presentation at a) individual nests and b) territorial boundaries

a. Presentations at Nests

 

 

CA1 CA2
Approach -0.900 0.028
Latency -0.873 0.061
Teer 0.629 -0.419
Titi 0.351 0.233
Seet 0.302 0.787
Hover 0.639 0.1 10
Dives 0.478 0.1 00
Strikes 0.519 -0.900
% Variance 43.05 17.55

 

b. Presentations at Boundaries

 

 

CA1 CA2
Latency -0.744 0.034
Approach -0.531 0.296
Teer 0.518 -0.458
TiTi 0.516 0.803
Seet 0.352 -0.200
Hover 0.324 0.277
Dive 0.438 -0.365
Strike 0.240 -0.388
% Variance 37.80 28.65

 

82

offspring with a randomly chosen neighbor using a pseudo-random number
generator. The neighbor’s territory was the same distance from the nest and

had not cuckolded that nest.

Because each male’s territory consisted of several nests, the resident male
might influence nest defense at nests within a territory. Therefore, for tests
involving multiple nests within a territory we used nested ANOVAs to partition

variance between nests within a territory and variation at the level of territory.

RESULTS

The results of the detection experiment revealed that males reduce their level of
defense at their territorial boundary in response to their neighbor’s lack of
defense (repeated measures ANOVA, treatment Fm = 2.571, NS, trial F222 =
0.429, NS, treatment * trial F222 = 4.132, P < 0.05; Figure 1). Defense by the
focal male was almost identical in the pre-defection presentation of the crow.
Males in the experimental group significantly decreased their level of nest
defense in the detection presentation (t-test, T11 = 3.4276, P < 0.001; Figure 1)

and also the day following defection (t-test, T11 = 6.084, P < 0.001; Figure 1).

Decreased expression of nest behavior observed for males in the experimental
treatment might result from differences in behavior by neighbor males between

control and experimental treatments. We therefore compared defense behavior

83

Nest Defense (CD1)

 

1.2

 

  
 

 

 

 

 

 

 

 

Treatment ... m
1 - 0 Control
0.8 _ E] Experimental
0.6 ~
0.4 d
0.2 -
0 .
.. . I
-0.4 4
-0.6 -
-0.8 . a
Pre-defectlon Defectlon Post-defection
Trial

Figure 1. Nest defense by the focal male in response to

treatment. Experimental treatments simulated defections by the
neighbor male, control treatments allowed cooperative defense
by both males. Error bars show :I: 1SE. N: 6 for both treatments.
*** P < 0.001.

84

of neighbor males between treatments in both pre- and post- defection
presentations. We found no difference in defense behavior by neighbor males
in response to the treatment or between trials (repeated measures ANOVA,
treatment Fm = 0.047, NS, trial F1,“ = 0.882, NS, treatment * trial Fm =
0.044, NS; Figure 2). The results of the detection experiment therefore appear

to result only from the response of the focal males to defection of their neighbor.

We found no difference in nest defense between a cuckolding male’s defense
at a nest on another territory and a randomly chosen neighbor who had not
cuckolded that nest (paired t-test, T24 = .654, NS; Figure 3). It is possible that
males are able to gain copulations with more than one female on a territory and
only a few of those copulations result in fertilizations. We therefore compared
average level of defense for all nests on a territory by a male that gained
fertilizations on that territory and a randomly chosen neighbor that had not
gained any fertilizations. There was again no difference in nest defense
between males that had gained fertilizations on a territory and those that had

not (paired t-test, T20 = 0.323, NS).

If females solicit copulations from several males in order to gain additional
parental care in the form of increased nest defense, then nests with cuckolded
young might receive more helpers than those without cuckolded young. We
found a significant association between the number of males defending a crow

and cuckoldry (nested ANOVA, territory F1319 = 1.708, NS, cuckoldry

85

Nest Defense(CDt)

 

 

    

 

 

 

 

 

 

 

1.2 —
Treatment 1'
1 ‘ " 0 Control
El Experimental
0.8 -
J
0 6 ; 33420
C
0.4 0
0.2 ~
0 - l—
-0.2 I
Pre-Defection Post-Detection
Trial

Figure 2. Nest defense behavior by neighbor males in pre- and
post- defection trials in response to treatment. In experimental
trials the neighbor male was held in a covered cage for the
duration of the experiment. In control trials the neighbor was
captured and released. Error bars show :I: 1 SE. N = 6 for both
treatments.

86

 

 

 

 

 

 

 

 

 

02-

E

g -0.4-

8

it -0.6l

G)

D

I”: -0.8‘ I

z I I
-1 I
-1.2

 

Genetic Father Paired Male

Figure 3. The level of nest defense by males at nests on other territories
where they had EPF’s and a randomly chosen male neighboring the
same territory that was the same distance from the nest. N = 25. Error
bars show :I: 1 SE.

87

(territory) F19, 32 = 2.498, P < 0.05; Figure 4) the difference however is the
opposite of the predicted direction. Nests without EPFs had more helpers than

those with EPFs.

Territorial males may reduce their defense of nests on their territory if they are
able to determine that the nest contains cuckolded offspring. We again used a
nested analysis to account for the fact that males have multiple nests on each
territory. Territorial males did not alter their level of nest defense in response to
cuckoldry (nested ANOVA, territory F1349 = 1.659, NS, cuckoldry (territory)

F1932: 1.751, NUS, Figure 5).

DISCUSSION

We found that males cooperate in nest defense primarily as a form of reciprocal
altruism. In simulated defections, males decreased defense against a nest
predator when their neighbor appeared to defect by not defending, but
increased defense in control trials. There was no association between nest
defense and cuckoldry suggesting that mutualism plays little role in cooperative

nest defense.

88

Number Of Helpers

 

 

 

 

 

 

 

 

No Yes
Cuckoldry

Figure 4. Average number of helpers at nests with cuckolded
offspring compared to nests without cuckolded offspring. Each
circle represents the average number of helpers at cuckolded and
uncuckolded nests. Lines connect data from within a territory. N =
19.

89

Nest Defense (CD1)

 

 

 

 

 

(15‘

 

 

 

 

-1 I
No Yes
Cuckoldry

Figure 5. Nest defense by territorial males in response to
cuckolded offspring in their nests. Circles represent average
defense by the territorial male at cuckolded and uncuckolded
nests. Lines connect data from within the same territory. N = 19

90

Nest Defense as TFT

If individuals engage in play TFT-like behavior as a form of cooperation then
they should generally cooperate, but in response to detection, should retaliate
by not cooperating. Additionally, long relationships enhance the advantage to
individuals that use TFT-like strategies (Axelrod and Hamilton, 1981). Our study
shows that male red-winged blackbirds reduce their level of defense when it
appears their neighbor is not cooperating. Retaliation indicates TFT-like
strategies and suggests that individuals that do not cooperate receive a short-
term benefit (Rothstein and Pierotti, 1988). This is consistent with Beletsky and
Orians’ (1989) findings that males that had been neighbors in previous years
suffered less from nest predation. Although they did not specifically test for
TFT-like relationships they surmised that the reduced predation they observed
resulted from cooperative nest defense. Unfortunately, in the year for which we
have defense data, not enough males returned from the previous year to

determine if familiar males were more likely to cooperate.

It could be argued that the reduction in defense we observed was in response
to increased risk to the focal male when the detecting male did not participate in
defense. By defending alone, the focal male was likely at greater risk of injury
than when the neighbor aided in defense (Hamilton, 1971). Similar arguments
have been proposed to explain apparent retaliatory behavior by guppies and

sticklebacks in predator inspection. In these studies, guppies or sticklebacks

91

approach a predatory fish and the fishes reflection in a mirror is used to
simulate another fish that either approaches simultaneously with or lags behind
the focal fish (Dugatkin, 1988; Milinski, 1987; Milinski et al., 1990). An
individual’s unwillingness to approach the predator, however, can be attributed
to selfish herd like mutualism rather than tit-for tat-like behavior (Lazarus and
Metcalfe, 1990; Masters and Waite, 1990). Although, subsequent experiments
have further strengthened the case for TFT in predator inspection (Milinski,
1990; Milinski, 1992; Milinski, 1996) the initial criticisms are nonetheless

potentially valid in this study.

It is unlikely increased risk to the focal male caused the results found in our
study. Increased risk may have caused the focal male to decreases nest
defense in the detection presentation, but it does not explain why males
continued to show decreased nest defense in the post defection presentation.
Especially since the neighbor male had returned and was defending normally
by then. The only explanation therefore is that males reduced nest defense in
the post presentation trial because they were retaliating in response to the

perceived detection by their neighbor.

Males in control treatments increased their defense considerably between pre-
defection presentations and detection and post-defection presentations. This is
likely caused by repeated presentation of the crow over the course of a trial.

Males may defend more aggressively with repeated exposure to models of nest

92

predators because they learn that they have “successfully" driven it off with no
negative consequences (Knight and Temple, 1986). This was an unavoidable
artifact of our design, but does not influence the interpretation of our results
since we presented the crow equally to both treatments. In fact, it makes the

decrease in defense by males in the experimental treatment more convincing.

Nest Defense as Mutualism

Other studies have suggested that there is a mutualistic component to
cooperative nest defense in red-winged blackbirds (Gray, 1997; Weatherhead
et al., 1994). Both of these studies suggest that males help at other male’s
nests because they have sired offspring in those nests. Gray (1997) showed
that males are more likely to defend at nests where they have cuckolded
offspring and Weatherhead et al. (1994) show that nests that contain extra-pair
offspring enjoy higher nesting success. This conforms to by-product mutualism
because the helping males are most likely reducing nest predation for offspring

of the territorial male as well as it’s own.

In contrast to the studies outlined above, our study suggests that mutualism
played little role in nest defense in our population. Males did not defend at nests
where they had cuckolded offspring (Figure 3). Nor did cuckolded nests appear
to attract more helping males than nests that had not been cuckolded (Figure

4). In fact, there was some tendency for nest that uncuckolded nests to attract

93

more helping males. This result is surprising given that there would seem to be
an obvious advantage to defending your own young from a nest predator.
Territorial males were also unresponsive to levels of cuckoldry on their own
territory and did not reduce their defense at cuckolded nests as expected
(Figure 5). This again is contrary to results in other populations (Weatherhead

etaL,1994)

Differing rates of cuckoldry cannot explain these differences. We found a similar
rate of cuckoldry in our population as Gray, approximately 33% of offspring and
50% of nests were cuckolded, although this rate is somewhat higher than
Weatherhead’s population. Nor is it likely that males are not responsive to rates
of cuckoldry. Other work in this population has shown that males react more
aggressively to neighbors that have gained EPFs on their territory (Olendorf

and Scribner, unpublished data).

Weatherhead (1995) found that in a population of red-winged blackbirds in
Ontario, female nesting success did not increase with male familiarity. Beletsky
and Orians (1989) found that familiarity enhanced breeding success in
population of red-winged blackbirds in Washington. Weatherhead suggested
that one possible reason for differences between his and Beletsky and Orians’

was different density of red-winged blackbirds and nest predators.

94

In our study, site mammalian predators are rare because they are fenced out.
Crows are common in the area and frequently forage at the pond facility when
red-winged blackbirds are not nesting there. During the nesting season,
however, crows are seldom seen in the pond facility. The crows remain in the
area, because the are frequently heard in the trees approximately 200 m away
and when the red-winged blackbirds leave after breeding, the crows return
immediately. Possibly, early defense by the red-winged blackbirds had
succeeded in reducing risks to predation and all the males (and females and

offspring as well) were benefiting from low predation rates.

Simultaneous tests for reciprocal altruism and mutualism

As outlined above, there is evidence for considerable heterogeneity in the
existence of cooperative nest defense. Other studies (Beletsky, 1989;
Weatherhead, 1995) specifically addressed the relationship between familiarity
among males and nest success or the relationship between extra-pair paternity
and nest success (Gray, 1997; Weatherhead et al., 1994). Although the
presumed mechanism among all these studies, was cooperative nest defense
among males, the cooperation can result either from mutualism (Gray, 1997;
Weatherhead et al., 1994) or reciprocal altruism (Beletsky, 1989; Weatherhead,

1995).

95

In this study, we tested for both mutualism and reciprocal altruism. We found
evidence for reciprocal altruism (retaliation and TFT-like strategies) but not
mutualism (no relationship between EPFs and nest defense). Our results
contribute to growing evidence for heterogeneity in cooperative nest defense
among red-winged blackbirds and suggest that in the future it would be valuable
to test for both reciprocal altruism and mutualism in similar studies. It would be
especially interesting to see if in populations where male nest defense varies in

response to cuckoldry, there is also evidence for reciprocal altruism.

LITERATURE CITED

Axelrod R, 1980a. Effective choice in the prisoner's dilemma. Journal of Conflict
Resolution 2423-25.

Axelrod R, 1980b. More effective choice in the prisoner's dilemma. Journal of
Conflict Resolution 24:379-403.

Axelrod R, Dion D, 1988. The further evolution of cooperation. Science
242:1385-1390.

Axelrod R, Hamilton WD, 1981. The evolution of cooperation. Science
211:1390-1396.

Beletsky LD, 1989. Familiar neighbors enhance breeding success in birds.
Proceedings of the National Academy of Sciences, USA 86:7933-7936.

Brindley EL, 1991. Response of European Robins to Playback of Song -

Neighbor Recognition and Overlapping. Animal Behaviour 41:503-512.

96

 

Connor RC, 1986. Pseudo-reciprocity: Investing in Mutualism. Animal
Behaviour 34:1562-1584.

Connor RC, 1995. The benefits of mutualism: a conceptual framework.
Biological Review 70:427-457.

Dawson RJG, Gibbs HL, Hobson KA, Yezerinac SM, 1997. Isolation of
microsatellite DNA markers from a passerine bird, Dendroica petechia
(the yellow warbler), and their use in population studies. Heredity 79:506-
514.

Dugatkin LA, 1988. Do guppies play TIT FOR TAT during predator inspection
visits? Behavioral Ecology and Sociobiology 23:395-399.

Dugatkin LA, 1997. Cooperation Among Animals. New York: Oxford University
Press.

Fisher J, 1954. Evolution and bird socialilty. In: Evolution as a process (Huxley
J, Hardy AC, Ford EB, eds). London: Allen & Unwin.

Getty T, 1987. Dear enemies and the Prisoner's Dilemma: Why should
territorial neighbors form defensive coalitions? American Zoologist
272327-336.

Gibbs HL, Miller P, Alderson G, Sealy SG, 1997. Genetic analysis of brown-
headed cowbirds Molothrus ater raised by different hosts: data from
mtDNA and microsatellite DNA markers. Molecular Ecology 6:189-193.

Gibbs HL, Weatherhead PJ, Boag PT, White BN, Tabak LM, Hoysak DJ, 1990.
Realized reproductive success of polygynous Red-winged Blackbirds

revealed by DNA markers. Science 250:1394-1397.

97

Godard R, 1991. Long-Term-Memory of Individual Neighbors in a Migratory
Songbird. Nature 350:228-229.

Gray EM, 1997. Female red-winged blackbirds accrue material benefits from
copulating with extra-pair males. Animal Behaviour 53:625-639.

Hamilton WD, 1971. Geometry for the Selfish Herd. Journal of Theoretical
Biology 31:295-311.

Knight RL, Temple SA, 1986. Why does intensity of avian nest defense
increase during the nesting cycle. Auk 103:318-327.

Knight RL, Temple SA, 1988. Nest-defense behavior in the Red-winged
Blackbird. Condor 90:193-200.

Kroodsma DE, Byers BE, Goodale E, Johnson S, Liu WC, 2001.
Pseudoreplication in playback experiments, revisited a decade later.
Animal Behaviour 61 :1029-1033.

Lazarus J, Metcalfe NB, 1990. Tit-for-tat cooperation in sticklebacks: a critique
of Milinski. Animal Behaviour 39:987-1002.

Marshall TC, Slate J, Kruuk LEB, Pemberton JM, 1998. Statistical confidence
for likelihood-based paternity inference in natural populations. Molecular
Ecology 7:639-655.

Masters WM, Waite TA, 1990. Tit-for-tat during predator inspection, or shoaling.
Animal Behaviour 39:603-604.

Maynard Smith J, 1982. Evolution and the Theory of Games. Cambridge:

Cambridge University Press.

98

Milinski M, 1987. TIT FOR TAT in sticklebacks and the evolution of cooperation.
Nature 325:433-435.

Milinski M, 1990. No alternative to tit-for-tat cooperation in sticklebacks. Animal
Behaviour 39:989-991.

Milinski M, 1992. Predator inspection: cooperation or "safety in numbers"?
Animal Behaviour 43:679-680.

Milinski M, 1996. By-product mutualism, Tit-for—Tat reciprocity and cooperative
predator inspection: a reply to Connor. Animal Behaviour 51 :458-461.

Milinski M, Kulling D, Kettler R, 1990. Tit for Tat: sticklebacks (Gasterosteus
aculeatus) "trusting" a cooperating partner. Behavioral Ecology 1:7-11.

Orians GH, 1961. The ecology of blackbird (Agelaius) social systems.
Ecological Monographs 31:285-312.

Pyle P, 1997. Identification guide to North American birds. Bolinas, California:
Slate Creek Press.

Rothstein SI, Pierotti R, 1988. Distinctions among reciprocal altruism, kin
selection, and cooperation and a model for the initial evolution of
beneficent behavior. Ethology and Sociobiology 9:189-209.

Seutin G, White BN, Boag PT, 1991. Preservation of avian blood and tissue
samples for DNA analyses. Canadian Journal of Zoology 69:82-90.

Stoddard PK, Beecher MD, Homing CL, Campbell SE, 1991. Recognition of
Individual Neighbors By Song in the Song Sparrow, a Species With Song

Repertoires. Behavioral Ecology and Sociobiology 29:211-215.

99

 

Terbraak CJF, 1985. Correspondence-Analysis of Incidence and Abundance
Data - Properties in Terms of a Unimodal Response Model. Biometrics
41:859-873.

Trivers RL, 1971. The evolution of reciprocal altruism. The Quarterly Review of
Biology 46:35-57.

Von Neuman J, Morgenstein O, 1953. Theory of games and economic
behavior. Princeton, NJ: Princeton University Press.

Weatherhead PJ, 1995. Effects of female reproductive success of familiarity
and experience among male red-winged blackbirds. Animal Behaviour
49:967-976.

Weatherhead PJ, Montgomerie R, Gibbs HL, Boag PT, 1994. The cost of extra-
pair fertilizations to female Red-winged Blackbirds. Proceedings of the
Royal Society of London, Series B 258:315-320.

Yasukawa K, 1989. The costs and benefits of a vocal signal: the nest-
associated 'Chit' of the female red-winged blackbird, Agelius

phoeniceusi. Animal Behaviour 38:866-874.

100

CHAPTER 4:

THE EFFECT OF POPULATIONS STRUCTURE ON THE EVOLUTION OF
COOPERATION AND COOPERATIVE STRATEGIES IN ARTIFICIAL
POPULATIONS

Robert Olendorf
Kellogg Biological Station
Michigan State University

3700 E Gull Lake Dr.
Hickory Corners, Ml 49060

ABSTRACT

Cooperative populations usually achieve higher average fitness than
uncooperative populations. Cooperative populations are susceptible to invasion
by uncooperative individuals, however, because uncooperative individuals may
potentially achieve higher fitness if the selective environment conforms to the
prisoners’ dilemma. Four general solutions to this problem have been proposed,
group selection, kin selection, reciprocal altruism and by-product mutualism.
These explanations only deal with the maintenance of cooperation but do not
explain how cooperation can arise in a population of uncooperative individuals.
Previous work has suggested that population structure can be instrumental in
facilitating the evolution of cooperation. These studies, however, did not vary
the degree of population structure, nor did they allow for a large array of
possible strategies to evolve. In this study, I use genetic algorithms in a

simulation of structured populations to determine the degree to which

101

population structure can affect the frequency of cooperation and the types of
strategies that evolve. The simulations show that populations composed of
small subpopulations achieved higher levels of cooperation. Populations with
the smallest subpopulations evolved strategies that were almost unconditional
cooperators, while slightly larger subpopulations evolved strategies similar to tit-
for-tat. This study shows that differences in subpopulation structure are
influence both the quantity and quality of cooperation that evolves in a

population.

102

INTRODUCTION

The existence of cooperative behaviors in nature presents a problem for
biologists. If a cooperative behavior imposes a cost on the individual exhibiting
the behavior while at the same time benefiting another individual, natural
selection should work to eliminate that behavior from the population. To persist
in the population the cooperative individual must therefore achieve some benefit
that compensates for the apparent loss of fitness. Explanations of how the
benefit is realized fall into four general categories, group-selection (Wilson,
1975; Wynne-Edwards, 1965), kin-selection (Hamilton, 1963), reciprocal
altruism (Trivers, 1971) and by-product mutualism (Connor, 1986). Each of
these explanations makes certain assumptions about population structure and
the pay-off structure so that cooperation is favored. In this study, I use
simulations using populations of genetic algorithms to study the role of
population structure in determining how cooperation evolves and what

strategies result.

Two selective environments can lead to the evolution of cooperation, the
Prisoner’s Dilemma (PD) and mutualism. The PD was developed for use in the
economic and social sciences by Flood and Dresher for the Rand Corporation
and popularized by Von Neuman (1953). In the PD two players can make one
of two choices, cooperate or defect. If both players cooperate then they both

receive a high payoff (R). If both players choose to defect, then both players

103

receive a low payoff (P). If one player defects while the other cooperates, than
the detecting players gets the highest possible payoff (T), while the other payoff
receives the lowest possible payoff (S). The pay-off structure therefore results
in the inequality (T > R > P > S). This creates an interesting tension, if both
players cooperate than they both do well. If, however, the one player is
reasonably certain the other will cooperate, then he should defect. The other
player should therefore also choose defect so as not to receive the lowest
payoff. Therefore, the only stable equilibrium is mutual defection. Group
selection, kin selection and reciprocal altruism alter the rules of this game so

that cooperation can evolve in certain circumstances.

Mutualism results when the inequality described above is altered so that (R >
T,S > P). Since cooperation always pays better than not cooperation,

cooperation should easily evolve in.

Cooperation can evolve through group selection provided the rate of mutation is
low, and the population is highly structured with limited migration among small
subpopulations. Populations are assumed ancestrally selfish. Mutation gives
rise to rare cooperative individuals and genetic drift allows some populations to
become fixed for the cooperative trait. Cooperative populations enjoy higher
fitness than selfish populations and produce more migrants. The greater
number of migrants coming for cooperative subpopulations results in enhanced

colonization of empty habitats (Wynne-Edwards, 1965) or increased migration

104

to occupied habitats in some situations (Wilson, 1975). Over time cooperative

subpopulations come to dominate selfish subpopulations.

Kin-selected cooperation results when individuals interact with related
individuals. Any fitness costs associated with cooperative behaviors are
compensated for by the increased fitness enjoyed by relatives (Hamilton, 1963).
Lack of dispersal and small population size can cause related individuals to
interact more often than would be expected otherwise. Kin-selection and group

selection can therefor work together to favor cooperation.

If individuals have a great enough probability of playing the game again, the
game is transformed into a new game known as the Iterated Prisoner’s
Dilemma (IPD). Because the players may play the game more than once,
cooperative strategies can often outperform always defect (ALLD) by
reciprocating altruistic acts (Trivers, 1971). By conducting a series of
tournaments where people submitted strategies to play in the IPD, Axelrod
(1980a; 1980b) ignited interest into effective strategies to play in the IPD. In the
first tournament submitted strategies played against one another in a round
robin tournament (Axelrod, 1980a). Much to his surprise tit-for-tat (TFT), one of
the simplest strategies submitted won. TFT cooperates on the first move, and
echoes the opponent’s previous move thereafter. In a subsequent tournament
where strategies competed in a more evolutionary context, TFT again did very

well. However, several other strategies, most similar to TFT, also performed

105

well (Axelrod, 1980b). Axelrod and Hamilton (1981) then went on to apply
evolutionary game theory (Maynard Smith, 1982) to argue that TFT was an
ESS strategy. Other studies have since shown that while TFI' is very robust it is
not strictly an ESS because other strategies can do as well as or possibly even
slightly better than TFT in the IPD (Boyd and Lorberbaum, 1987; Dugatkin and

Wilson, 1991; Nowak and Sigmund, 1993; Posch, 1999).

Despite the robustness of these strategies in pure populations, none of these
strategies can easily invade a population of selfish individuals. When
cooperators are rare in a population, they most often interact with selfish
individuals. TFI' for instance would cooperate on the first iteration of the game
and receive the lowest possible payoff while the selfish individual would receive
the highest payoff. Although it would detect thereafter, the initial loss of fitness
relative to the selfish individual would prevent it from increasing in frequency
through natural selection. If the population is structured and the individual
subpopulations are small, genetic drift can allow the frequency of TFT can pass

a threshold where it becomes favored by frequency dependent selection.

Following the example of Axelrod’s (1980b, 1980 #35) tournaments many
recent studies of the evolution of cooperation rely on simulations (e.g. Brauchli
et al., 1999b; Brembs, 1996; Nowak and Sigmund, 1993; Posch, 1999;
Sigmund and Nowak, 1998). Simulations have several advantages over

analytical solutions when studying the evolution of cooperation. Analytical

106

models can easily describe two competing strategies in a population, When
additional strategies are added, however, analytical models become difficult or
even impossible to solve. Simulation models overcome this difficulty by directly
modeling the strategies as individual agents in a population. Individual agents
interact based on their assigned strategies and reproduce according the fitness
achieved through those interactions. More recently, genetic algorithms (GAs)
are being used to model the evolution of cooperative behaviors (Sigmund,

1998).

GAs are essentially computer programs that give instructions on how an
artificial agent should behave (Holland, 1975). Although initially developed by to
by computer scientists to solve complex problems that cannot be solved
analytically, they are increasingly used by biologists to study evolutionary
processes (Mitchell, 1998). A large number of agents (individual variants of a
GA) are created, each variant is a potential solution to a problem. The structure
of the algorithm varies widely and is dependent on the nature of the problem.
They can be as simple as a vector of number to be used as parameters to
actual computer programs. In all cases, each GA is potential solution to the
problem and the fitness of each GA is determined by its ability to solve the
problem. The most successful GAs either go through differential reproduction or
differential death so that the GAs that best solve the problem increase in
frequency. A fraction of the agents are mutated altering their algorithm and the

process is repeated. Mutations allow the testing of new potential solutions that

107

are similar to already successful solutions and help to fully explore the adaptive
landscape. The similarity of these techniques to biological evolution is not
accidental. Computer scientists purposely copied nature reasoning that
evolution is a problem solving system using individual life forms as potential
solutions. Formal studies into effective mechanisms for selecting good
algorithms in fact suggest that mechanisms that most closely model biology are

among the most effective at finding good solutions to complex problems.

Genetic algorithms are a natural tool to use to study the evolution of behavior
provided an adequate system can be devised to generate the behaviors of
interest. Several recent studies have used genetic algorithms to show that a
certain strategies is an effective solution to the PD (e.g. Nowak and Sigmund,
1993; Posch, 1999). These algorithms, however, are designed to easily code
for the strategy under study. All of these studies assume that there is no need

to have a memory more than three interactions in the past.

Several simulation studies have also shown that population structure is an
important factor in the evolution of cooperation (Brauchli et al., 1999a;
Killingback and Doebeli, 1998; Mitteldorf and Wilson, 2000). These studies,
however, only use at most a handful of known strategies to study this
relationship. In this study, I develop a scheme for generating algorithms that
can code for a large variety of strategies, including most commonly discussed

strategies. I addition, there is virtually no limit on length of memory. Using

108

structured, populations I will determine under which conditions of population
size and dispersal the evolution of cooperation is favored. Additionally, I will
determine if population structure can influence the types of strategies that
evolve. For instance, always cooperate might evolve under circumstances
favorable to group selection, but TFT might evolve in somewhat less structured

populations.

THE MODEL

The Genetic Code

The model presented in this study uses genetic algorithms to determine the
strategies of the individuals. The principal behind this system of generating
strategies was to create simple mechanism that would easily and succinctly
code for many commonly discussed strategies in the literature as well as a
great number of strategies that may not have been previously studied. In
addition, this system allows for easy interpretation of the “DNA” strings to

determine the strategies for which they code.

The basic units of the system are a set of five characters (c, d, :, I, *) arrayed
along a one-dimensional array, the “chromosome” (Figure 1). The structure of

the chromosome consists of a series of “genes” along the chromosome. Each

109

 

 

chddc*ccd:*cc:d{c:ccdlcdclcc Chromosome
ddc*ccdr’Ebzd [EKGenes
*c c

ddc*ccd c

Receptor Effector Receptor Effector

 

 

 

 

Figure 1. The structure of the genetic algorithm. A chromosome is
constructed of an array of characters. Individual genes within a
chromosome delineated by slashes. Each gene consists of a receptor
and effector portion. The receptor starts at the left slash and continues
to the first colon. The receptor is a pattern to be matched by an
opponent’s behavioral history. The c’s and d’s correspond to previous
plays, the most recent on the right. Asterisks (*) match any
combination of characters or no characters. The receptor is the first c
or d to the right of a colon, the remainder of the string is ignored until
the next functional gene is encountered.

110

gene starts at a slash and ends at the next slash. The structure of each gene
consists of a receptor and an effector. The receptor consisted of a series of c’s,
d’s and *’s starting from the left most slash to the first colon. The c’s and d’s
reflect the behavioral history of an opponent, the most recent interactions being
on the right. The asterisks can stand for any number of characters. For
example, a receptor of d*c would match any history where the opponent
defected on the first move and cooperated on the most recent move, any
intervening interactions are ignored. The effector portion of a gene consists
simply of the first c or d to the right of the first colon. A single gene, therefor,
consists of a pattern on the left of a colon that, if matched, elicits a response
specified on the right of the colon (Figure 1). The set of genes on a
chromosome, reading from left to right, specifies an individual’s repertoire of
stimulus/response sets or its strategy. In many instances, genes are not
interpretable in this system and uninterpretable genes are ignored. If more than
one pattern matches the opponent’s behavioral history, the first matching
pattern from the left is used. If no patterns match, the individual plays defect.
Setting the default response to defect is analogous to the hypothesized

primitive condition of detection in animals.

Strategies such as tit-for-tat, always cooperate and always defect are easily
written and easily recognized using this system (Table I). Additionally, a wide
variety of other strategies can also be generated at random. This system does

ignore the individuals own behavior and the payoffs received. It therefore

111

Table I. Examples of chromosomes along with the associated
strategies. The names of commonly discussed strategies are
also show.

 

 

 

 

 

 

COW

Chromosome Receptor : Effector Pairs Strategy Abbreviation
/':d/ (' : 0) Always Detect ALLD
/':c/ (' : 0) Always Cooperate ALLC
/‘d:d/':c/ (”d : d).(' : c) Tit-for-tat TFT
/‘dd:d/':c/ (”dd : d),(' : c) Generous TFT GTFT

R n oml t Chr m
Chromosome Receptor : Effector Pairs
'd:d (‘ : d)
'///:/dd/cd:'d'd/l/cdc:d/cdd (cd: d),(cdc : d),(‘ : d)
d‘Id:::d'/:'d‘:: (d : d),(' : d)
c:'c'::cd:::c'd"d//’ d/://c‘///d/‘dcc’d::cd:/: ('dcc'd : c),(' : d)

':/: d'/ d:'/ c’:cc/ ccdc:dc'cdcc'/ dcdd'l cdc'd c),(ccdc : d),(' :d)

 

112

cannot code for strategies such as Pavlov (Posch, 1999) where an individual
must be aware of its own history and the payoff structure as well as its’
opponent’s play history. Such strategies, however, require additional cognitive
skills on the part of biological animals such as memory of both its own and it’s
opponents behavioral history or knowledge of the payoff structure. This results
in an analogous increase in memory usage by the agents. One of the goals of
this study is to determine the simplest conditions under which cooperation might
evolve. It would be interesting to study such things in the future but it is beyond

the scope of this study.

The Simulation

All populations in this study were composed subpopulations arranged in a 10 X
10 array. l varied population size within subpopulations (carrying capacity) and
dispersal between subpopulations. Populations were first initialized by either
randomly generating 100 character chromosomes for each individual or

supplying all individuals with a specified strategy.

Two individuals were selected randomly with replacement from within a sub-
population to interact. Each interaction consisted of a number of bouts. After
each bout, a pseudo-random number was generated. If the number was less
than 1/15, the interaction ended and new individuals were drawn. The process

was repeated N times (N = subpopulation size).

113

Its chromosome and the opponent’s behavior determined each individual’s
behavior. The fitness for each interaction was determined using the defined
pay-off structure (i.e. Prisoner’s Dilemma). The mean fitness per interaction was

used to select among strategies to standardize for interaction length.

After interactions were complete, individuals were selected to were subjected
mutation and crossing-over. I set the mutation rate to be 1 X 10'6
mutations/character. This rate is within the observed mutation rate in natural
populations (Nei, 1987). This rate of mutation also resulted in good rates of
evolution when simulating simple problems such as evolving a population of
pure defectors to pure cooperators in a mutualistic environment with no
crossing-over. I allowed substitution, insertion and deletion mutations to occur. I
set the rate of crossing-over to be 10% of the population in a generation. Higher
rates of crossing-over appeared to hinder the ability of a population to find the

fitness peak in the simple problem outlined above.

After mutation and crossing over, individuals reproduced based on their relative
fitness within each sub-population (soft selection). Individuals then dispersed
from subpopulations. The number of dispersing individuals was equal to the
surplus number of agents in the subpopulation. l varied dispersal rate varied by
varying mortality among dispersers from 0 to 100%. In other words, if a

subpopulation had a carrying capacity of 10, and after reproduction contained

114

15 agents, 5 agents would attempt to disperse. If the dispersal rate were 0.1,
then on average only 0.5 agents would successfully disperse. It a sub-
population was still over carrying capacity following dispersal, individuals were
removed at random until carrying capacity was reached (density independent
mortality). Each individual’s fitness was reset to the initial value and the process

was repeated for the specified number of generations.

Experimental Design

To determine the affect of population structure on the evolution of cooperation, I
varied subpopulation size (10, 25, 50, 100) and dispersal rate (0, 0.0001, 0.001,
0.01, 0.1, 1) in a fully crossed design. Ten runs of 500 generations were

completed for each combination of parameters for a total of 240 runs.

I used the frequency of cooperative interactions as a measure of cooperation. I
also determined the most commonly used strategies in each setting. With
randomly generated chromosomes, there was typically a wide diversity of
chromosomes at the end of each simulation. There was considerably less
variation in strategies although the number of different strategies was still
formidable. Much of that variation however, resulted from similar strategies that
differed in a single gene that had little impact on the function of the agent. The
simplest form of TFT is coded by; I*d:dl*:c/. Addition of a new gene such as

/cddcd:cl*d:d/*:cl would not alter the behavior of the agent because the

115

 

receptor of the additional gene is so specific. To facilitate analysis, I therefore
grouped similar strategies together. Granted, this is a subjective process, so I

was conservative in grouping strategies together.

Three different combinations of initial strategy compositions and selective
environments were used. I first ran randomly generated individuals in a
mutualistic environment to ensure that cooperation could evolve in each setting.
I also ran pure TFT populations in a PD environment to establish that TFT was
resistant to invasion. I then ran randomly generated individuals in a PD
environment to determine how population structure affected the evolution of

cooperation from selfish population.

RESULTS

Randomly Generated Agents in a mutualistic environment

All populations achieved high levels of cooperation in a mutualistic environment,
reaching 99% cooperation. Despite the generally higher levels of cooperation
achieved in the mutualistic environment, there was a significant affect of
subpopulation size but there was no effect of dispersal (multiple regression,
dispersal T = 0.0002, N.S., subpopulation size T = -13.736, P < 0.001, R2 =

0.465, F2222 = 96.621, P < 0.001; Figure 2 & 3). The highest levels of

116

Cooperation

 

Figure 2. Level of cooperation achieved after 500 generation

by randomly generated populations in a mutualistic environment
varying dispersal and subpopulation size. The surface was fit
using DWLS smoothing.

117

 

 

 

‘l

O

c

2 .: ’

«..:

s I

or

Q -I

00'99 '

O

0 . o

o 0

O
8

0.98 ‘

 

 

 

I I T T

0 20 40 60 80 100 120
Subpopulation Size

Figure 6. The level of coopration achieved in all populations in response to
subpopulations size in mutualistic environments.

118

cooperation were found in populations composed of subpopulations of 10
agents and decreased slightly as subpopulation size increased. The most

common strategies in all runs were effectively equivalent to ALLC.

Pure TFT in PD environments

TFT proved to be resistant to invasion by other strategies in all environments
(Figure 4). There was a very slight dip in levels of cooperation in populations
with low dispersal and high population size but the change was very small.

There appeared to be little increase in the frequency of other strategies.

Randomly Generated Agents in the PD environment

When starting with randomly generated chromosomes, the initial level of
cooperation within a population was approximately 2% of all interactions. After
500 generations, the average level of cooperation among all runs was 13%.
Subpopulation size was the most important factor influencing the evolution of
cooperation while dispersal had no affect (regression, dispersal T = -0.498,
NS, population size T = --8.008, P < 0.001, R2 = 0.214, F2237 = 96.621, P <
0.0001; Figure 5 & 6). Populations consisting of subpopulations of 10 agents
had the highest level of cooperation at 40% of interactions. Cooperation

decreased with increasing subpopulation size until at 100 individuals per

119

 

Cooperation

 

 

Figure 4. Level of cooperation achieved after 500 generation

by initially pure TFT populations in a PD environment varying
dispersal and subpopulation size. The surface was fit using DWLS
smoothing.

120

Cooperation

_L
O.

9
00

0.6

 

Figure 5. Level of cooperation achieved after 500 generation
by randomly generated populations in a PD environment varying

dispersal and subpopulation size. The surface was fit using DWLS
smoothing.

121

1.2'

0.8“
0.61
0.4'

Cooperation

0.2‘

 

“MW. 0 “.0 Q
0

O

r

 

l

0 20 40 60 80 100
Subpopulation Size

Figure 6. Level of cooperation in relation to subpopulation size. All runs are
included regardless of dispersal.

122

subpopulation, the levels of cooperation was identical to randomly generated

agents.

The combination of dispersal and subpopulation size influenced the type of
strategies that evolved (Figure 4). When dispersal and populations size were
low, the most common strategies were similar to always cooperate strategy
(ALLC) but often differed in one aspect. These strategies always cooperated
unless the opponent detected on the first play, in which case it always defected
(ONED). TFT evolved most commonly in intermediate levels of dispersal and
subpopulation size (Figure 4). Always defect (ALLD) was the most common
strategy to evolve in populations defined by large subpopulations and high

dispersal.

DISCUSSION

The simulations presented in this study verify the broad prediction that in PD
environments, population structure can facilitate the evolution of cooperation
both through reciprocal altruism and group selection. We also show that at least
in the short term, novel cooperative strategies can evolve. The simulations also
corroborate the expectation that the TFT strategy is a robust strategy, not easily

invaded by other populations.

123

Dispersal

 

 

1 TFT ALLD/TFT ALLD ALLD
0.1 TFT TFT ALLD ALLD
0.01 ONED TFT ALLD ALLD

0.001 ONED TFT/ONED ALLD ALLD

0.0001 ALLC ONED ALLD ALLD

0 ONED ALLC ALLD ALLD
1O 25 50 100

Population Size

Figure 7. The effect of population structure on the types of strategies
that evolved in the populations is shown. TFT = tit-for-tat, ALLD =
always defect, ALLC = always cooperate, ONED = test for defection on
the first play only.

124

Several recent studies have demonstrated that population structure can
facilitate the evolution of cooperation (Brauchli et al., 1999a; Killingback and
Doebeli, 1998; Mitteldorf and Wilson, 2000). These studies corroborate earlier
analytical studies that predict such a relationship (Axelrod and Hamilton, 1981;
Wilson, 1975). Small subpopulations allow genetic drift to occur so that
strategies such as TFI' can reach a critical threshold where frequency
dependent selection begins to favor them (Boorman and Levitt, 1973). It has
also been argued that population structure can facilitate the evolution of ALLC-

like strategies via group selection (Wilson, 1977; Wynne-Edwards, 1965).

In this study, subpopulation size appears to be the dominant force influencing
the rate at which populations evolve cooperative behaviors. This is a somewhat
surprising result since dispersal determines how distinct subpopulations are. It
is likely that this result is an artifact of the simulation. Even in runs where
dispersal was set at one, individuals were only able to move to neighboring
demes. This limited amount of dispersal may not have been able to disrupt local
population structure. If individuals had dispersed more widely, dispersal would
probably have had a greater effect. Nonetheless, this study supports the
hypothesis that population supports the evolution of both group selected

cooperation and reciprocal altruism.

Kin selection may also have played a role in the evolution of cooperation in

these simulations. Several generations of breeding within small populations

125

leads to inbreeding. The greater than average relatedness among the
individuals in a subpopulation suggests that altruism is probably always directed
towards kin. This makes teasing apart the forces of kin selection and group
selection difficult since inbreeding will always occur in small populations. There
is one important difference between kin selection, as described by Hamilton,
and kin selection that might occur in small populations. In classical kin selection
individuals take advantage of heterogeneity of relatedness within a population
(Hamilton, 1963). Kin selection in a small population there may be little
heterogeneity of relatedness among individuals. It may therefore be more
appropriate to refer to the evolution of cooperation in small populations to be

evolving as a result of group selection.

This study also suggests that group selection can favor cooperation in
conditions where subpopulations are very small and distinct. However, one
novel strategy evolved, ONED. This strategy was the same as ALLC except
that reverted to always detect if it’s opponent detected on the first move. This
strategy has the ability to detect ALLD early on much like TFT. However, it has
two serious weaknesses. First, it only checks for detection on the first move.
Strategies that cooperated on the first move and defected aftenlvards would
have a strong advantage against ONED. Second, it does not forgive so that it
may lose the opportunity to cooperate with more cooperative strategies that
happen to defect on the first move. It is likely that this is a transitory strategy,

and possibly is a step on the way to evolving a true TFT strategy. All of our runs

126

 

lasted for only 500 generations. Longer runs would determine the ability of

ONED to persist.

Brauchli et al. (1999a) found similar results. They ran simulations of all
stochastic strategies with one round memories. Their simulations show, as in
my simulations, that spatial structure can not only influence the level of
cooperation but also the strategies used the evolve. In addition, their spatially
structured populations were invaded by transient strategies that facilitated the
establishment of more robust strategies such as PAVLOV (win stay, lose shift)

or TFI’.

Nowak & Sigmund (1993) show that the Pavlov strategy cannot invade a
population of detectors without first without another cooperative strategy such
as TFT establishing itself first. These results suggest that ONED in our
populations is quite possibly a transient strategy, that could possibly facilitate

the evolution of other more robust cooperative strategies.

One problem with structured populations where a variety of different
populations can coexist is noise. Although two different strategies might be
cooperative, differences between the strategies might lead to mutual defection.
Brauchli et al. (1999b) found that generous strategies were often favored under
these circumstances. Their model, however, used stochastic strategies.

Individuals could be forgiving by reducing the probability of detecting in

127

 

response to detecting. Grim (1995) and Sigmund & Nowak (1992), however,
also found that spatial structure favored generous strategies in reactive

strategies.

Generosity did not develop in my simulations. The most likely explanation for
this is the short length of the simulations. Although high levels of cooperation
were achieved under a variety of conditions, there may still be more robust
strategies that had yet to arise. It is also possible, however, that the potentially
long memory used in my simulations allowed strategies to forgo the potential

cost of generosity.

The results of this study conform to the results from other studies. Spatial
structure enhanced the evolution of cooperation, although I found that
subpopulation size, but not subpopulation structure contributed to the evolution
of cooperation. The simulations presented here extend previous studies by
showing that not only does spatial structure facilitate the cooperation of
cooperation, but that varying levels of population structure altars the trajectory

and possibly the final composition of strategies in the population.

128

LITERATURE CITED

Axelrod R, 1980a. Effective choice in the prisoner's dilemma. Journal of Conflict
Resolution 2423-25.

Axelrod R, 1980b. More effective choice in the prisoner's dilemma. Journal of
Conflict Resolution 242379-403.

Axelrod R, Hamilton WD, 1981. The evolution of cooperation. Science
211:1390-1396.

Boorman SA, Levitt PR, 1973. A frequency-dependent natural selection model
for the evolution of social cooperation networks. Proceedings of the
National Academy of Sciences 70:187-189.

Boyd R, Lorberbaum JP, 1987. No pure strategy is evolutionarily stable in the
repeated Prisoner's Dilemma game. Nature 327:58-59.

Brauchli K, Killingback T, Deobeli M, 1999a. Evolution of Cooperation in
Spatially Structured Populations. Journal of Theoretical Biology 200:405-
417.

Brauchli K, Killingback T, Doebeli M, 1999b. Evolution of cooperation in
spatially structured populations. Journal of Theoretical Biology 200:405-
417.

Brembs B, 1996. Chaos, cheating and cooperation: potential solutions to the
Prisoner's Dilemma. Oikos 76:14-24.

Connor RC, 1986. Pseudo-reciprocity: Investing in Mutualism. Animal

Behaviour 34: 1 562-1 584.

129

Dugatkin LA, Wilson DS, 1991. Rover - a Strategy For Exploiting Cooperators in
a Patchy Environment. American Naturalist 138:687-701.

Grim P, 1995. The greater generosity of the spatialized prisoner's dilemma.
Journal of Theoretical Biology 173:353-359.

Hamilton WD, 1963. The evolution of altruistic behavior. American Naturalist
97:354-356.

Holland JH, 1975. Adaptation in natural and artificial systems, 1992 Edition ed.
Cambridge, Massachusetts: The MIT Press.

Killingback T, Doebeli M, 1998. Self-organized criticality in spatial evolutionary
game theory. Journal of Evolutionary Biology 191 :335-340.

Maynard Smith J, 1982. Evolution and the Theory of Games. Cambridge:
Cambridge University Press.

Mitchell M, 1998. An introduction to genetic algorithms. Cambridge,
Massachusetts: The MIT Press.

Mitteldorf J, Wilson DS, 2000. Population Viscosity and the Evolution of
Cooperation. Journal of Theoretical Biology 204:481-496.

Nei M, 1987. Molecular Evolutionary Genetics. New York: Columbia University
Press.

Nowak M, Sigmund K, 1993. A Strategy of Win Stay, Lose Shift That
Outperforms Tit-For—Tat in the Prisoners-Dilemma Game. Nature 364256-
58.

Nowak MA, Sigmund K, 1992. Tit-For-Tat in Heterogeneous Populations.

Nature 355:250-253.

130

 

Posch M, 1999. Win-stay, lose-shift strategies for repeated games - Memory
length, aspiration levels and noise. Journal of Theoretical Biology
198:183-195.

Sigmund K, 1998. Complex adaptive systems and the evolution of
reciprocation. Ecosystems 1:444-448.

Sigmund K, Nowak MA, 1998. The complexity of cooperation: Agent-based
models of competition and collaboration. Nature 392:457-457.

Trivers RL, 1971. The evolution of reciprocal altruism. The Quarterly Review of
Biology 46:35-57. l

Von Neuman J, Morgenstein O, 1953. Theory of games and economic
behavior. Princeton, NJ: Princeton University Press.

Wilson DS, 1975. A theory of group selection. Proceedings of the National
Academy of Sciences 72: 1 43-1 46.

Wilson DS, 1977. Structured demes and the evolution of group-advantageous
triats. American Naturalist 11 1:157-185.

Wynne-Edwards VC, 1965. Self-regulating systems in populations of animals.

Science 147:1543-1548.

131