QUANTITATIVE GENETICS, SELECTION, MATE CHOICE AND RED SQUIRREL
BEHAVIOR IN A FLUCTUATING ENVIRONMENT
by
Ryan W. Taylor

A DISSERTATION
Submitted
to Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Zoology
Ecology, Evolutionary Biology and Behavior
2012

ABSTRACT
QUANTITATIVE GENETICS, SELECTION, MATE CHOICE AND RED SQUIRREL
BEHAVIOR IN A FLUCTUATING ENVIRONMENT
by
Ryan W. Taylor
Consistent individual differences in behaviour, and behavioural correlations within and across contexts, are referred to as animal personalities. These patterns of variation have been identified
in many animal taxa and are likely to have important ecological and evolutionary consequences.
My dissertation focuses on understanding how individual behavioral variation in free-living North
american red squirrels (Tamiasciurus hudsonicus) might be maintained in the face of natural selective pressures. I investigated the proximate sources of variation in red squirrel behavior by estimating sources of genetic, maternal and environmental variation, and the ultimate effect of behavior
on reproductive success. I used a Bayesian animal model approach to estimate genetic parameters
for aggression, activity and docility and found support for low heritabilities (0.08-0.12), and cohort
effects (0.07-0.09), as well as low to moderate maternal effects (0.07-0.15) and permanent environmental effects (0.08-0.16). I also found evidence of a substantial positive genetic correlation
(0.68) and maternal effects correlation (0.58) between activity and aggression, providing evidence
of genetically based behavioral correlations in red squirrels. These results provide evidence for the
presence of heritable variation in red squirrel behavior, but also emphasize the role of other sources
of variation, including maternal effects, in shaping patterns of variation and covariation in behavioral traits. I then tested the hypothesis that ecological changes through time lead to fluctuating
selection, which maintains variation in behavioral traits. Linear selection gradients on red squirrel
dam aggression and activity significantly fluctuated in sign across years depending on the level of
competition among juveniles. Selection on aggression and activity also differed among components of fitness, between the sexes and included nonlinear components between and within traits
that also changed through time. These results suggest that repeatable and heritable individual dif-

ferences in red squirrel behavior could be maintained by complex fluctuations in natural and sexual
selection. Finally, I tested the hypotheses that mating chases provide the opportunity for both female and male red squirrels to select for context-dependent good genes and complementary genes
for their offspring’s recruitment. Specifically, I predicted that aggressive mates would be preferred
in high-juvenile competition years, but disfavored in low competition years, and that mate choice
for docility would be disassortative with low-docility squirrels preferring high-docility mates and
vice-versa. I did not find support for adaptive context-dependent mate choice by females, but I
did find support for male mate choice for complementary genes that was mediated through which
female mating chase males attended. Male red squirrels attended mating chases disassortatively
by docility, which would enhance fitness due to stabilizing selection on docility during juvenile
recruitment. I also found evidence for post-copulatory selection for less aggressive males. These
results highlight that even in systems with very high operational sex ratios, male choice is a factor
that needs to be considered. In this dissertation I have shown that behavioral traits in red squirrels
are important with strong evolutionary implications. Red squirrel behavior is heritable, subject to
maternal effects and under strong selection that varies with environmental fluctuation and across
components of reproductive success. This research has provided strong evidence that fluctuating
selection could maintain variation in behavior. These findings provide a foundation for future work
to elucidate mechanisms and continue exploring how and why behavioral variation is maintained.

ACKNOWLEDGEMENTS

I would like to thank Andrew McAdam for being a great scientific mentor and advisor throughout my graduate studies. I’m especially appreciative that Andrew is persistently supportive and
positive. I would also like to thank Kay Holekamp for welcoming me into her lab after Andrew
moved to the University of Guelph. Kay provided excellent academic and scientific mentoring and
I learned a lot from her and from the great students she attracts. I thank Stand Boutin and Murray
Humphries, along with Andrew McAdam for their leadership of the Kluane Red Squirrel Project
and their valuable input input into planning and carrying out my dissertation research. Thanks
to all of the Kluane Red Squirrel Project graduate students, post-docs and technicians, especially
Ainsley Sykes, Ben Dantzer, Bastien Ferland-Raymond, Quinn Fletcher, Christina Guillemont,
Sara Hanner, Abe Katzen, Jeff Lane, Veronica Mari, Jo McEvoy, Eryn McFarlane, Skyler Shatkin
and Carolyn Troudou. I thank Adi Boon for starting the study of personality in Kluane red squirrels
and allowing my use of the data she collected.
I thank my parents, John and Delia, for introducing me to and teaching me about the diversity
and wonder of the natural world. Thanks to my brother, Gavin, for your encouragement and for
visiting me in Kluane.
Thank you to my wife, Doro, for your patience, support and good humor.

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
CHAPTER 1 GENERAL INTRODUCTION . . . . . . . . .
1.1 What is animal personality? . . . . . . . . . . . . . . . .
1.2 North American red squirrels (Tamiasciurus hudsonicus)
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4 A note about writing style . . . . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

1
2
5
6
7

LOW HERITABILITIES, BUT GENETIC AND MATERNAL
CORRELATIONS BETWEEN RED SQUIRREL BEHAVIOURS
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

8
8
13
16
18
23

FLUCTUATING AND NONLINEAR SELECTION ON BEHAVIOR
IN A WILD POPULATION OF RED SQUIRRELS . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Behavioral Traits . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1 Temporal variation and annual selection . . . . . . . . . . . . . . . . .
3.3.2 Interactions with Competition and Path Analysis . . . . . . . . . . . .
3.3.3 Linear and nonlinear selection on ARS . . . . . . . . . . . . . . . . . .
3.3.4 Components of fitness . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.5 Sexual Conflict . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.6 Fitness surface estimation . . . . . . . . . . . . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1 Fluctuating selection . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Competition and the Pace-of-Life Hypothesis . . . . . . . . . . . . . .
3.4.3 Fitness Components . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

28
28
31
32
33
37
37
44
44
47
48
48
50
50
53
53
56

CHAPTER 2
2.1
2.2
2.3
2.4
2.5

CHAPTER 3
3.1
3.2

3.3

3.4

CHAPTER 4
4.1
4.2

DISASSORTATIVE MATE CHOICE ON BEHAVIOR IN RED
SQUIRRELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

v

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

63
64
66
67
67
69
69
72
72
73

CHAPTER 5 CONCLUSIONS AND FUTURE DIRECTIONS . . . . . . . . . . .
5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Are there consistent individual differences in behavioral plasticity? . . .
5.2.2 Investigation of links between behavior and life-history, energetics and
hormone levels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.

.
.
.
.

77
77
78
78

4.3

4.4

4.2.1 Siring success and parturition date
4.2.2 Behavioral Traits . . . . . . . . .
4.2.3 Mating Chase Observations . . . .
4.2.4 Components of siring success . .
4.2.5 Mate choice . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . .
4.3.1 Distance . . . . . . . . . . . . . .
4.3.2 Male Mate Choice . . . . . . . .
4.3.3 Female Mate Choice . . . . . . .
Discussion . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

. . 79

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
APPENDIX A

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

APPENDIX B

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

APPENDIX C

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

LITERATURE CITED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

vi

LIST OF TABLES

Table 2.1

Heritability of behavior measured in wild populations, including those from
entirely wild populations (w) and those estimated from offspring raised in captivity (c). (−) Parameter was not estimated; NS: parameter was estimated but
found not significant; S: a significant genetic correlation was found between
this trait and another behavioral trait. . . . . . . . . . . . . . . . . . . . . . . . 10

Table 2.2

First principle component loadings for behaviors from an open field arena test
(OF PC1) and a mirror-image stimulation test (MIS PC1) in North American
red squirrels. Behaviors were measured as percentage of time unless otherwise
noted. Latencies were log transformed prior to principle component analysis.
Additional principle component axis are provided in Appendix Table A.3 . . . . 19

Table 2.3

Effects of habituation, day of year and study area on red squirrel docility,
aggression and activity. Year trial number is the trial number counted from
the start of each year, while life trial number is the trial number counted over
each individual’s entire life. Trial numbers for docility include all handling
events, even those in which docility was not measured. 95% credible intervals
are given in parentheses and those that exclude zero are indicated in bold. The
effects of the Kloo and Sulphur study areas are assessed relative to the Agnes
study area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Table 2.4

Heritability (h2 = VA /VP ), maternal effects (m2 = VM /VP ), permanent environmental effects (PE = VPE /VP ), cohort effects (C = VC /VP ), repeatability
([VA + VM + VPE + VC ]/VP ), and the mean trait value. Variances were estimated using a trivariate model. Variances are calculated as the mode of the
posterior distribution with 95% credible intervals in parentheses and are bound
above zero. Coefficients of variation (CV = 100 × standard deviation / mean)
are given for docility. Because aggression and activity are scores from a principal component analysis using a correlation matrix the trait means are 0 and
coefficients of variation cannot be calculated. . . . . . . . . . . . . . . . . . . . 22

Table 2.5

Additive genetic and phenotypic variances, covariances and correlations (Gmatrix and P-matrix) of red squirrel personality traits. Variances are indicated
along the diagonal, the upper triangle contains correlations and the lower triangle covariances. Variances are calculated as the mode of the posterior distribution with 95% credible intervals in parentheses and are bounded above
zero. Correlations and covariances that were different from zero (based on
95% credible intervals) are indicated in bold. . . . . . . . . . . . . . . . . . . . 24

vii

Table 2.6

Maternal, permanent environmental and cohort (birth year) variances, covariances and correlations of red squirrel personality traits. Variances are indicated along the diagonal, the upper triangle contains correlations and the lower
triangle covariances. Variances are calculated as the mode of the posterior distribution with 95% credible intervals in parentheses and are bound above zero.
Correlations that were different from zero (based on 95% credible intervals)
are indicated in bold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Table 3.1

The effect of year on selection for dam and sire behavioral traits through annual reproductive success. Significance was calculated with Wald χ 2 tests
from an analysis of deviance. GLMMs were fitted with squirrel identity as a
random effect and assumed a Poisson error distribution. . . . . . . . . . . . . . . 37

Table 3.2

Standardized linear selection gradients ± standard errors for dam and sire
behavioral traits on annual reproductive success. Significance was based on
standard errors generated by jackknifing. The sample size (n) indicates the
number of individuals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Table 3.3

Reduced models of selection on behavioral traits for dam annual reproductive
success, fecundity and offspring overwinter survival. The full GLMM models
included linear, pairwise interaction and quadratic terms and were reduced by
removing the least significant term until all remaining terms were significant.
Interaction and quadratic terms were removed prior to linear terms starting
with the least significant. Study-area-year combination was a random effect
and a Poisson error distribution was assumed. Random effect grid-year variances for high-competition years: ARS = 0.34, Fecundity = 0, OWS = 0.84,
and for low-competition years: ARS = 0.11, Fecundity = 0.11, OWS = 0. . . . . 41

Table 3.4

Reduced models of selection on behavioral traits for sire annual reproductive
success, siring success and offspring overwinter survival. The full GLMM
models included linear, pairwise interaction and quadratic terms and were reduced by removing the least significant term until all remaining terms were
significant. Interaction and quadratic terms were removed prior to linear terms
starting with the least significant. Study-area-year combination was a random
effect and a Poisson error distribution was assumed. . . . . . . . . . . . . . . . . 42

Table 3.5

Vector of standardized directional selection gradients (β ) for annual reproductive success, and the matrix of standardized quadratic and correlational
selection gradients (γ). Linear and quadratic selection gradients were estimated in separate regressions. Quadratic selection coefficients were doubled
to give quadratic selection gradients (Stinchcombe et al. 2008). Significance
was based on standard errors generated by jackknifing. . . . . . . . . . . . . . . 43

Table 3.6

The M matrix of eigenvectors from the canonical analysis of γ for annual
reproductive success. The eigenvalue (λ ) of each eigenvector (m) is given in
the first column. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
viii

Table 3.7

Summary statistics for standardized linear selection gradients (β ). A high
ratio of mean of absolute values of selection gradients to the absolute value of
the mean of selection gradients indicates that fluctuations in selection reduced
overall directional selection (sensu Kingsolver and Diamond 2011). Large
standard deviations among selection gradients relative to the mean of their
standard errors indicates, along with significant changes in sign (Table 2 and
Figure 1), that variation in selection was not due to sampling error (sensu
Morrissey and Hadfield 2012). The frequency of sign changes was calculated
as the number of changes in direction between successive years relative to
n − 1, where n is the total number of years (sensu Siepielski et al. 2011). . . . . . 52

Table 4.1

Sample sizes for each study area-year. Total number of individuals living on
the study area, and the number of individuals for which we had behavioral
data are given. Behavior was measured in 2005 and 2008. . . . . . . . . . . . . 67

Table 4.2

Reduced models of male attendance, siring success and copulation success of
males that attended a mating chase, and siring success of males that copulated.
The full GLMM models are described in the methods. The least significant
terms were removed, starting with the highest order. Male identity and mating
chase identity were included as random effects. A binary error distribution
was assumed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Table 4.3

Reduced models of male copulation order , siring success after controlling
for copulation order, copulation number and siring success after controlling
for copulation number. The full GLMM models are described in the methods.
The least significant terms were removed, starting with the highest order. Male
identity and mating chase identity were included as random effects. A binary
error distribution was assumed for the siring success models, and a Poisson
error distribution was assumed for the copulation order and copulation number
models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Table A.1 Between-observer reliability, means and variances of specific behaviors recorded
during open field and mirror-image stimulation tests. Correlations were calculated for each behavior between two observers’ scores for the same trial.
Behaviors with a reliability greater than 0.7 were used in further analysis.
Mirror image stimulation behaviors are indicated above, open field behaviors
are below. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Table A.2 Summary statistics for the entire multi-generational red squirrel pedigree, pedigree of individuals informative for docility and pedigree of individuals informative for activity and aggression. . . . . . . . . . . . . . . . . . . . . . . . . 82
Table A.3 Principle component loadings for behaviors from an open field arena test in
North American red squirrels. Behaviors were measured as percentage of time
unless otherwise noted. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

ix

Table A.4 Principle component loadings for behaviors from an mirror-image stimulation
test in North American red squirrels. Behaviors were measured as percentage of time unless otherwise noted. Latencies were log transformed prior to
principle component analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Table A.5 Candidate univariate models of sources of variation in behavior in North American red squirrels. Fixed effects (FE) are trial number, quadratics for trial
number, study area, day of year and observer. Random effects are individual identity (VPE ), additive genetic (VA ), maternal effects (VM ) and cohort
effects (VC ). Deviance Information Criterion (DIC) estimates are given for
each model. Models within 2 DIC of the best model were considered to have
equal support (bold). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Table A.6 Candidate univariate models of sources of variation in behavior in North American red squirrels. Fixed effects (FE) are trial number, quadratics for trial
number, study area, day of year and observer. Random effects are individual identity (VPE ), additive genetic (VA ), maternal effects (VM ) and cohort
effects (VC ). Deviance Information Criterion (DIC) estimates are given for
each model. Models within 2 DIC of the best model were considered to have
equal support (bold). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Table A.7 Candidate trivariate animal models of sources of (co)variation of the behavioral traits aggression, activity and docility. Fixed effects (FE) are trial number, quadratics for trial number, study area, day of year and observer. Random
effects are individual identity (VPE ), additive genetic (VA ), maternal effects
(VM ) and cohort effects (VC ). Deviance Information Criterion (DIC) estimates
are given for each model. Models within 2 DIC of the best model were considered to have equal support. . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Table B.1 The effect of competition on linear selection for dam and sire behavioral traits
through annual reproductive success. Coefficients are from a GLMM with
ARS as the response and each study-area-year combination as a random effect
and a Poisson error distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Table B.2 Vector of standardized directional selection gradients(β ) for fecundity or siring success, and the matrix of standardized quadratic and correlational selection gradients (γ). Linear and quadratic selection gradients were estimated
in separate regressions. Quadratic selection coefficients were doubled to give
quadratic selection gradients (Stinchcombe et al. 2008) . . . . . . . . . . . . . . 88
Table B.3 Vector of standardized directional selection gradients(β ) for offspring overwinter survival, and the matrix of standardized quadratic and correlational selection gradients (γ). Linear and quadratic selection gradients were estimated
in separate regressions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

x

Table C.1 Reduced models of male attendance, siring success and copulation success of
males that attended a mating chase, and siring success of males that copulated.
The full GLMM models are described in the methods. The least significant
terms were removed, starting with the highest order. Male identity and mating
chase identity were included as random effects. A binary error distribution
was assumed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

xi

LIST OF FIGURES

Figure 1.1

Terminology used in publications on animal personality from 2000 to 2010.
Plotted are the annual number of publications containing each keyword in the
title or abstract. Plural and American spelling were included in the search.
I performed a search of the title, abstract and keywords of publications in
behavioral journals in the Web of Science, or Science Citation Index Expanded, using the terms ‘personality trait’, ‘animal personality’, ‘behavioral
syndrome’, ‘consistent individual differences’, ‘behavioral type’, ‘coping
style’, and ‘temperament’. The journals searched were: Animal Behavior,
Behavioral Ecology, Behavioral Ecology and Sociobiology, Behavior, Behavioral Processes, American Naturalist, Biology Letters, Ecology, Ecology
Letters, Evolution, Functional Ecology, Journal of Animal Ecology, Evolution, Functional Ecology, Journal of Animal Ecology, Journal of Evolutionary Biology, Proceedings of the Academy of Sciences of the United States of
America, Proceedings of the Royal Society B, Trends in Ecology and Evolution, Quarterly Review of Biology, Biological Reviews and Philosophical
Transactions of the Royal Society B. . . . . . . . . . . . . . . . . . . . . . . .

3

Figure 3.1

Directional selection gradients ± standard errors of annual reproductive success for dam aggression (MIS), activity (OF) and struggle rate (ST). 75th
percentiles are given for directional selection gradient magnitudes compiled
by Kingsolver et al. (2001). * P < 0.05, . P < 0.1. . . . . . . . . . . . . . . . . 39

Figure 3.2

Path models depicting direct and indirect effects of parental behavior on
annual reproductive success (ARS) in high and low-competition environments. Indirect effects of behavior are mediated through fecundity (or siringsuccess) and offspring overwinter survival (OWS). Variation due to error for
fecundity, siring-success and OWS was high (U = 0.97 to 0.99) and omitted
for simplicity. Dashed lines depict negative coefficients, and line-width is
proportional to the standardized coefficients (see scale). . . . . . . . . . . . . . 45

Figure 3.3

Fitness surfaces of ARS for dam aggression and activity in high and lowcompetition years. Selection favored aggression (β = 0.55±0.14, P < 0.05)
and disfavored activity (β = −0.36 ± 0.20, P < 0.1) in high-competition
years. In low-competition years selection favored activity (β = 0.30±0.12, P <
0.05), but disfavored aggression (β = −0.28 ± 0.12, P < 0.05). Surfaces
were visualized using thin plate splines. Surfaces and data points are conditioned on docility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

xii

Figure 3.4

Fitness surfaces of ARS for sire aggression and activity in high and lowcompetition years. Selection on aggression was disruptive in high-competition
years (γii = 1.67 ± 1.20, P > 0.1), but stabilizing in low-competition years
(γii = −0.72 ± 0.28, P < 0.05). Selection on activity fluctuated in the opposite direction and was stabilizing in high-competition years (γii = −0.48 ±
0.47, P > 0.1), but disruptive in low-competition years (γii = 0.72±0.28, P <
0.05). Surfaces were visualized using thin plate splines. Surfaces and data
points are conditioned on docility. . . . . . . . . . . . . . . . . . . . . . . . . . 47

Figure 3.5

Fitness surfaces for sire behavioral traits in high competition years. Fitness
surfaces are plotted for 1st , 50th and 99th percentiles of the third behavioral
trait. Surfaces were visualized using thin plate splines . . . . . . . . . . . . . . 49

Figure 3.6

The relationships between selection gradients for dam and sire aggression.
On the y-axes are selection gradients for dam aggression through annual reproductive success. On the x-axes are selection gradients for sire aggression
for a) siring success(r = −0.77, CI = −0.96 to −0.04) and b) offspring overwinter survival (r = 0.62, CI = −0.25 to 0.94). . . . . . . . . . . . . . . . . . 54

Figure 4.1

Three-way interaction plot of the interaction between male docility, female
docility and competition on male attendance. . . . . . . . . . . . . . . . . . . . 73

Figure 4.2

The effect distance on male attendance to a mating chase. The dashed line
indicates predicted probability of attendance at −1, and the solid line at
+1, standard deviation of juvenile competition. In high juvenile-competition
years the effect of distance is stronger. . . . . . . . . . . . . . . . . . . . . . . 74

xiii

Chapter 1
GENERAL INTRODUCTION

Darwinian natural selection presents a paradox that if only the fittest survive, then why do less fit
genotypes persist, or why does selection not erode all genetic variation? The evolutionary maintenance of variation in fitness-related traits has been a major topic in evolutionary biology since it
was first popularized by Fisher (Fisher, 1930; Barton and Turelli, 1989; Houle, 1992; Merilä and
Sheldon, 2000; Kruuk et al., 2008). A number of recent meta-analysis have provided some insights
(e.g. linear selection on survival is generally weaker than on fecundity; Kingsolver et al. 2012) but
some expected generalized patterns have not appeared or are subject to interpretation. For example,
Siepielski et al. (2011) found that temporal fluctuations in selection was prevalent, but Morrissey
and Hadfield (2012) disagree and argue that most of the temporal variation can be explained by
measurement error. Additionally, relatively few studies have examined environmental correlates
of selection, which often requires selection measured over multiple years, or selection on behavior
(Kingsolver et al., 2012). More detailed analysis of the interaction between ecology and evolution
will be necessary to understand how evolution maintains genetic variation.
Behavioral ecologists are increasingly recognizing that individual animals display consistently
different behavioral tendencies across time and contexts. The study of individual behavioral consistency is not new per se, and is a major area of research in a number of behavioral fields (e.g.,
personality in humans and domesticated animals or behavioral genetics; Fraser et al. 2001; Gosling
and John 1999; Henderson 1986; Boissy 1995). However, until recently most studies in behavioral
ecology did not address variation at an individual level or across multiple contexts, instead comparing population means within a single context (Sih and Bell, 2008). Behavior has been shown
to be under selection in a variety of taxa (reviewed in Smith and Blumstein (2008)) and can be
considered a trait closely related to fitness.

1

1.1

What is animal personality?

The study of individual animal behavior from an evolutionary ecology perspective started to gain
momentum around the turn of the millennium while being promoted by a number influential reviews (Wilson et al., 1994; Gosling, 2001; Sih et al., 2004b; Réale et al., 2007, 2010a). As a result,
there has been a rapid increase in research quantifying and comparing behavioral variation both
within and between individuals, but this progress has also produced a daunting array of terminology and concepts (Figure 1.1). A number of recent articles have provided useful synopses of
terminology (e.g. Réale et al. 2007, 2010a; Sih et al. 2004b; Stamps and Groothuis 2010; Sih and
Bell 2008), but inconsistent use and overlap remains.
The terms ‘personality trait’, ‘behavioral syndrome’, and ‘animal personality’ are the most
common terms but are used interchangeably, and not consistently in the literature. For example,
‘behavioral syndrome’ was first used (Stamps, 1991) and is most often used to describe a correlation between two distinct behavioral traits (e.g. aggression and boldness), but in a major review
Sih et al. (2004b) defined behavioral syndrome as a “a suite of correlated behaviors across multiple (two or more) observations”; and clarified in Sih and Bell (2008) that this includes the same
behavior across multiple contexts (e.g. activity in low-resource environments and activity in highresource environments). Thus ‘behavioral syndrome’ can be used to describe a very broad array of
behavioral phenomena. Similarly, ‘behavior’ describes a very broad array of quantifiable phenomena, and the distinction between one behavior and another is not always clear (i.e. is aggression
towards a predator “aggression” or “boldness”?).
In any field of study consistent understanding and application of terms and the concepts that
they refer to is necessary for clear scientific communication and collaboration. In new fields of
biology, there may be multiple terms used for the same concept because researchers do not have
a reliable framework of terminology to consult. For example, when defining “adaptation” there
was great confusion regarding different conceptual meanings of the term: meanings that applied
to either historical evolution or to current processes (Reeve and Sherman, 1993). The consistent
application of a concept is important because it clarifies the explicit assumptions that underlie the
2

Annual number of publications

30
Personality Trait
Animal Personality
Behavioral Syndrome
Consist. Indiv. Diff.
Behavioral Type
Coping Style
Temperament

25
20
15
10
5

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

0

Year

Figure 1.1: Terminology used in publications on animal personality from 2000 to 2010. Plotted
are the annual number of publications containing each keyword in the title or abstract. Plural and
American spelling were included in the search. I performed a search of the title, abstract and
keywords of publications in behavioral journals in the Web of Science, or Science Citation Index
Expanded, using the terms ‘personality trait’, ‘animal personality’, ‘behavioral syndrome’,
‘consistent individual differences’, ‘behavioral type’, ‘coping style’, and ‘temperament’. The
journals searched were: Animal Behavior, Behavioral Ecology, Behavioral Ecology and
Sociobiology, Behavior, Behavioral Processes, American Naturalist, Biology Letters, Ecology,
Ecology Letters, Evolution, Functional Ecology, Journal of Animal Ecology, Evolution,
Functional Ecology, Journal of Animal Ecology, Journal of Evolutionary Biology, Proceedings of
the Academy of Sciences of the United States of America, Proceedings of the Royal Society B,
Trends in Ecology and Evolution, Quarterly Review of Biology, Biological Reviews and
Philosophical Transactions of the Royal Society B.

3

ideas (Fauth et al., 1996). The current state of terminology use in the animal personality literature
is in danger of confusing those outside of the field and newcomers who rely on the terminology to
organize the behavioral phenomenon being studied. For these reasons I use these terms sparingly
and instead focus on describing the behavioral phenomenon that was quantified.
The fundamental principal of all personality related concepts is that individual differences in
behavior exist and are quantifiable. Behavioral traits can then be treated as any other trait and all
the tools evolutionary biologists have developed can be applied. Furthermore behavioral traits can
be used to study many of the broader questions in ecological and evolutionary biology.
SectionFluctuating environments
Environmental heterogeneity on temporal and spatial scales is a widespread if not universal
factor the evolutionary history of populations (Bell, 2010). Due to climactic, ecological and social variation individuals encounter many different environments and situations throughout their
life. Evolutionary theory predicts that individuals should exhibit optimal behaviors in each situation and therefore should be behaviorally flexible when the optimal behavior varies across the
situations they encounter (Stearns, 1989). However, evidence shows that individuals do not adjust
behaviorally as much as we would expect indicating that behavioral plasticity can be limited (Sih
et al., 2004b; Réale et al., 2007; Bell et al., 2009). Perhaps one of the more puzzling facets of behavioral consistency is that consistency may be maladaptive. For example, aggressiveness towards
competitors may benefit an individual via access to resources while a similar level of aggressiveness towards predators might increase chance of death. In a high resource year active individuals
may be able to consume more resources than those that are less active, but a consistently active and
energetic individual is more likely to starve when resources are scarce than an individual that can
tailor its behavior to the amount of resources available. The fact that variation in behavioral consistency is maintained in populations makes it a particularly interesting trait with which to study
the maintenance of variation.

4

1.2

North American red squirrels (Tamiasciurus hudsonicus)

The recent finding of individual variation in activity, aggressiveness and docility in red squirrels
(Boon et al., 2007, 2008) provides an excellent opportunity to study the proximate (i.e. genetic) and
ultimate factors of behavioral variation in a species for which there exists strong ecological variation. Boon et al. showed that activity (measured via open-field trials) was correlated with nestling
growth rates (grams per day) and that aggressiveness (measured via mirror-image stimulation) was
correlated with pup survival in nest and overwinter (Boon et al., 2007). These results show that
personality is associated with other phenotypes and affects fitness. Importantly, the magnitude and
direction of these correlations varied across years (Boon et al., 2007).
Northern populations of red squirrels experience a high degree of annual fluctuation in available
resources due to the synchronization of white spruce (Picea glauca) cone production. Spruce trees
typically mast once every 4-6 years and drive population density fluctuations in squirrels with low
density preceding the mast and peak density following the mast. The combination of low density
and high resources during a mast event results in weak selection for individual quality traits in
juvenile squirrels (e.g. parturition date, growth rate), but in the year following a mast selection
for quality traits is high due to high densities and scarce resources (Andrew McAdam, in prep).
Additionally, this variation in environmental conditions experienced by squirrels in early life has
been shown to have long lasting fitness consequences (Descamps et al., 2008). These cohort effects
may result in differences in residual reproductive value across years that may select for different
personalities (sensu Wolf et al. 2007). For instance the likelihood that an individual born during
a mast year will survive to reproduce in a future mast year is lower than individuals born in the
year after a mast. The combination of detailed life-history data (collected since 1988; McAdam
et al. 2007), personalities associated with fitness and a variable and quantifiable environment make
northern red squirrels an excellent model system in which to study the generation and maintenance
of behavioral variation.

5

1.3

Outline

In this dissertation I investigate the proximate sources of variation in red squirrel behavior and
the ultimate effect of behavior on reproductive success. I use the same behavioral methods as
Boon et al. (described in detail in Chapter 2). In Chapter 2 I leverage the extensive pedigree
that has been assembled for Kluane red squirrels to characterize the quantitative genetic source
of behavioral variation in a wild population of red squirrels. I used a Bayesian animal model
approach to estimate additive genetic, maternal, cohort and permanent environmental sources of
(co)variation in individual behavior and showed that red squirrel behaviors are heritable and have
maternal, cohort and environmental effects. Additionally, I showed that aggression and activity are
genetically and maternally correlated.
I then investigated whether fluctuating selection could maintain variation in behavior (Chapter
3). I first showed that there was a significant effect of year on selection for female aggression and
activity. I then estimated linear and non-linear selection gradients for annual reproductive success
separately for each of 8 years to show that selection for female aggression and activity fluctuated
significantly in sign. I performed a canonical rotation of the non-linear selection gradients to reveal
nonlinear multivariate relationships on fitness, which exposed nonlinear stabilizing selection contrasting sire activity with docility in high juvenile-competition years. Finally, I used generalized
linear mixed models to partition selection on behavior through annual reproductive success into
selection through fecundity (or for males, siring success) and offspring overwinter survival. Selection on dam behavioral traits acted entirely through offspring overwinter survival, but selection on
sire traits acted through siring success and offspring overwinter survival.
In my final chapter (4), I addressed a specific hypothesis that red squirrel mating chases are
opportunities for female squirrels to select good genes for their offspring, and that attendance to
mating chases is an opportunity for male squirrels to select good genes (Chapter 4). I did not find
evidence of female mate choice, but I did find that males attended mating chases disassortatively
by docility, which would enhance fitness due to stabilizing selection on docility during juvenile
recruitment.
6

1.4

A note about writing style

Chapter 2 has been published and is reprinted in this dissertation with only changes to American
English from the original British English. The title of chapter 2 remains in British English. I use
“we” throughout this dissertation to recognize that this work results from collaboration with my
co-authors.

7

Chapter 2
LOW HERITABILITIES, BUT GENETIC AND MATERNAL
CORRELATIONS BETWEEN RED SQUIRREL BEHAVIOURS

Taylor, R. W., Boon, A. K., Dantzer, B., Réale, D., Humphries, M. M., Boutin, S., Gorrell, J.
C., Coltman, D. W., and McAdam, A. G. 2012. Low heritabilities, but genetic and maternal
correlations between red squirrel behaviours. Journal of Evolutionary Biology 25 : 614 − 624.

Author Contributions
R. W. Taylor and A. G. McAdam developed the concept of the paper. R. W. Taylor, A. K. Boon,
and B. Dantzer collected behavioral data. All authors contributed to the collection of pedigree data.
J. C. Gorrell and D. W. Coltman performed the paternity assignment. R. W. Taylor performed all
other analyses and wrote the paper. All other authors provided intellectual insight and comments.

2.1

Introduction

Consistent individual differences in behavior and behavioral correlations within and across contexts have been identified in many animal taxa (Sih et al., 2004a; Réale et al., 2010a). The persistence of individual differences in behavior has stimulated much research (Gosling, 2001; Sih et al.,
2004b,a; Réale et al., 2007, 2010a), because behavior was previously thought to be very labile,
and have been increasingly accepted as important traits with ecological and evolutionary consequences (Réale et al., 2010a). Animal personalities have been shown to have important fitness
effects in a number of systems (Dingemanse and Réale, 2005; Smith and Blumstein, 2008) and
balancing selection has been proposed as a mechanism that could maintain these individual differences (Dingemanse et al., 2004; Boon et al., 2007; Wolf et al., 2007). For example, Dingemanse
et al. (2004) found that adult great tit (Parsus major) overwinter survival was related to exploratory

8

behavior and that this relationship fluctuated in direction across years. Understanding how natural
selection shapes patterns of variation in animal personalities, however, requires an understanding
of the underlying genetic structure of these important traits.
The genetic and environmental sources of (co)variation in specific repeatable behaviors, or
personality traits, have rarely been characterized for wild populations (Dochtermann and Roff,
2010), despite the widely recognized importance of genetic variation for evolutionary processes
(Roff, 1997; Lynch and Walsh, 1998). A long history of study in captive animals has established
that heritable variation in behavior is common in laboratory populations (Stirling et al., 2002).
However, because these populations have been bred in captivity for many generations and are
not exposed to the degree of environmental variation found in the wild, these heritability estimates
may not be representative of wild populations (Weigensberg and Roff, 1996; Hoffmann and Merilä,
1999; Hoffmann, 2000; Conner et al., 2003). The number of studies that have measured heritability
of personality traits in wild populations is small, but growing (Table 2.1), yet more estimates will
be necessary before broad patterns emerge. For example, it is unclear whether certain personality
traits or suits of correlated behaviors are more heritable than others, and which sources of variation
in personality traits are more sensitive to changes in environmental circumstances (i.e., genotype
× environment, maternal × environment interactions).

9

Table 2.1: Heritability of behavior measured in wild populations, including those from entirely wild populations (w) and those
estimated from offspring raised in captivity (c). (−) Parameter was not estimated; NS: parameter was estimated but found not
significant; S: a significant genetic correlation was found between this trait and another behavioral trait.
Species

h2

Behavior

Ovis canadensis (Sheep) (w)
Parus major (Tits) (w)
Gasterosteus aculeatus (Sticklebacks) (c)
Population 1

Boldness
Exploration
Activity
Aggression
Boldness
Population 2
Activity
Aggression
Boldness
Euprymna tasmanica (Squids) (c)
Boldness
Antipredator context
Activity
Reactivity
Feeding context
Boldness
Activity
Reactivity
Gasterosteus aculeatus (Sticklebacks) (c)
Multiple traits
Antipredator context
Multiple traits
Sialia mexicana (Bluebirds) (w)
Aggression
Dispersal
Ovis canadensis (Sheep) (w)
Docility
Boldness
Parus major (Tits) (w)
Exploration
Marmota flaviventris (Marmots) (w)
Vigilance
Sprint speed
Thamnophis ordinoides (Garter Snakes) (c) Anti-predator

10

m2

rG

Reference

0.21
0.22 to 0.41
0.05
0.01
0.04
0.16
0.14
0.002
0.21
0.67
0.89
NS
NS
NS
0.06 to 0.27
0.15 to 0.32
0.34
0.52
0.65
0.39
0.23
0.08
0.21
0.54 to 0.65

−
−
−
−
−
−
−
−
NS
NS
NS
NS
NS
NS
−
−
NS
NS
NS
NS
−
NS
NS
−

−
−
S
S
S
S
S
S
−
−
−
−
−
−
S
S
S
S
S
S
−
NS
NS
S

Réale et al. 2000
Dingemanse et al. 2002
Bell 2005

Sinn et al. 2006

Dingemanse et al. 2009
Duckworth and Kruuk 2009
Réale et al. 2009
Quinn et al. 2009
Blumstein et al. 2010
Brodie 1993

Correlated behaviors (sensu Sih et al. 2004a,b) are an important component of animal personalities and quantifying genetic correlations between behaviors will provide important insights
into the functional integration of behaviors (Cheverud, 1996) and potential constraints on behavioral evolution. A genetic behavioral correlation would suggest that independent adaptation of
the behavioral traits involved could be constrained (Roff and Fairbairn 2007; but see Blows and
Hoffmann 2005). Most previous studies have quantified only phenotypic correlations between behavioral traits (Dochtermann and Roff, 2010), but the degree or sign of a phenotypic correlation
may not match the underlying genetic correlation (Roff, 1997; Kruuk et al., 2008). So far, there
have been few estimates of genetic correlations among personality traits in wild populations but
those that have estimated genetic correlations have generally found results consistent with patterns
of variation observed at the phenotypic level (Bell, 2005; Réale et al., 2009; Blumstein et al., 2010).
For example, a number of studies have found positive phenotypic correlations between aggression
and boldness (Koolhaas et al., 1999; Sih et al., 2004a; Carere et al., 2005; Boon et al., 2007; Sih
and Bell, 2008; Réale et al., 2009) and, where examined, positive genetic correlations between
boldness and aggression (Bell, 2005; Dingemanse et al., 2007; Réale et al., 2009). Additionally, in
a meta-analysis Dochtermann (2011) found a strong correlation between phenotypic and genetic
correlations for behavioral traits. However, the aggression-boldness syndrome is not ubiquitous
and has been shown to depend on environmental context (Bell, 2005; Dingemanse et al., 2007).
Genetic correlations may be the result of pleiotropic relationships (Conner and Via, 1993; Blows
and Hoffmann, 2005). Alternatively, natural selection for optimal trait combinations (correlational
selection) may produce genetic correlations through linkage disequilibrium (Sinervo and Svensson, 2002), as hypothesized for correlated behaviors by Wolf et al. (2007); Stamps (2007); Biro
and Stamps (2008); Duckworth and Kruuk (2009); Houston (2010).
Environmental conditions during an individual’s development can also have consistent long
lasting effects (Lindstrom, 1999). These cohort effects can be important sources of trait variation
especially when the environment substantially fluctuates on an annual basis. Environmental effects
shared by relatives may also play a role in shaping variation in personalities and, if not accounted

11

for, may bias estimates of heritabilities (Kruuk and Hadfield, 2007). Parental effects are a special
case in which the environment provided by one or both parents results in similar offspring phenotypes (Mousseau and Fox, 1998; Mousseau et al., 2009), which can cause an overestimation of
additive genetic effects when the effects of a common parental environment are not considered
(Falconer and Mackay, 1996; Kruuk and Hadfield, 2007). These effects can be especially strong
in species where parents provide extended care to their offspring, as in the case of mammals where
mothers typically nurse their young for extended periods (Reinhold, 2002). Although documented
in captivity (Forstmeier et al., 2004), maternal (or parental) effects have rarely been estimated
for personality traits in wild populations and have never been shown to account for a substantial
proportion of the variation in those traits (Table 2.1). The potential for maternal effects to shape
evolutionary processes has received a lot of recent attention (Mousseau et al., 2009) and they may
play an important role in the evolution of personality traits, but this has yet to be thoroughly examined.
Personality traits and a behavioral syndrome were recently identified in a population of red
squirrels (Tamiasciurus hudsonicus Erxleben; Boon et al. 2007), for which an extensive pedigree
has been established (e.g., Réale et al. 2003b; McFarlane et al. 2011). Significant phenotypic
correlations between aggressive behavior towards a mirror-image, activity in an open-field arena
and activity in response to handling (i.e. docility) indicated the presence of a behavioral syndrome
where aggressive squirrels tended to be more active and less docile (Boon et al., 2008), which
is similar to the aggressive-boldness syndrome that has been identified in many taxa (Tulley and
Huntingford, 1988; Koolhaas et al., 1999; Careau et al., 2010).
Boon et al. (2007) also found that natural selection on activity and aggression fluctuated in
direction and magnitude across years, a mechanism by which genetic and phenotypic variation
in red squirrel personality traits could be maintained. Understanding how patterns of behavioral
variation respond to natural selection, however, depends also on the levels of genetic variation and
covariation in these traits. We, therefore, estimated heritabilities, maternal effects and sources of
environment variation (e.g., cohort, permanent environmental and residual effects) in aggression,

12

activity and docility, as well as genetic correlations among these traits for this population of red
squirrels using a Bayesian animal model approach (Hadfield, 2010).

2.2

Methods

A population of wild red squirrels has been monitored since 1989 in the southwest Yukon (61◦ N,
138◦ W), and a detailed description of the population and general methods can be found in McAdam
et al. (2007). T. hudsonicus is a small (150 − 250 g), diurnal, semi-arboreal rodent in the family
Sciuridae that is present in much of forested North America (Steele, 1998). Individuals of both
sexes defend exclusive year-round territories (Smith, 1968). This territoriality allows for complete enumeration of the study population through targeted trapping and behavioral observations.
Though trappability is related to red squirrel personality traits (Boon et al., 2008), the ability to
target individuals minimizes sampling bias by ensuring that all individuals in the population are
sampled (Biro and Dingemanse, 2009). Each squirrel in the study area was uniquely marked with
numbered ear-tags as nestlings or at first capture after emergence from the natal nest, and followed
from birth until death. Nests of lactating females were entered once immediately after parturition,
then again when pups were approximately 25 days age, to tag pups and collect tissue samples for
paternity analysis (Lane et al., 2008b).
We measured personality traits of individuals from three study areas. The Kloo and Sulphur
study areas have been monitored continuously since 1989. In contrast, the Agnes study area has
been monitored only recently. Since the autumn of 2004, ad libitum peanut butter has been experimentally provided to every squirrel in the Agnes study area between October and May of each
year.
Following Boon et al. (2007), we used three behavioral tests to measure red squirrel personality
traits. The first test, an open field (OF) test, was used to measure an individual’s activity, exploration and behavioral stress response in a novel environment (Walsh and Cummins, 1976; Martin
and Réale, 2008). The second test was a mirror-image stimulation (MIS) that measured aggression
in response to the individual’s reflection (Svendsen and Armitage, 1973). The third test, conducted
13

during routine handling events, measured docility as the struggle rate of an individual confined
in a mesh handling bag. We performed 556 OF and MIS trials on 183 female and 183 male red
squirrels and 3122 struggle rate tests on 291 female and 301 male squirrels over a four-year period
(2005 and 2008 − 2010). The mean number of OF and MIS trials per individual was 1.4 (range 1
to 5) with a mean interval of 261 days (range 12 to 1435) and the mean date of trials was July 2
(range April 29 to September 24). The mean number of struggle rate trials per individual was 6.2
(range 1 to 44) with a mean interval of 56 days (range 0 to 1436 days) and the mean date of trials
was June 22 (range April 29 to September 24). Individual squirrels were tested within one hour
of being trapped on their territory. To measure docility, the captured squirrel was immediately
transferred from the trap into a handling bag and placed onto dry ground and the proportion of
time the squirrel spent struggling was measured over 30 seconds. If the ground was wet or snowy
a foam or cloth barrier was positioned between the squirrel and the ground. After the docility test
was performed, ear-tag numbers, mass, and reproductive status were recorded. If the squirrel was
to be tested in the OF or MIS test, the squirrel was then transferred into the arena through a sliding
door to begin the OF trial. The testing arena for the OF and MIS tests was a 60 × 80 × 50cm
white corrugated-plastic box with a clear acrylic lid through which the behavior of the squirrel was
recorded with a digital video camera. Four blind holes were placed in the floor to provide the focal
squirrel with the opportunity to explore. A 45 × 30 cm mirror at one end of the arena was exposed
during the MIS portion of the session. OF and MIS trials were both performed in the same testing
session. OF performance was tested first so that it would also serve as a habituation period for the
MIS trial. After 7.5 minutes the mirror was exposed to start the 5-minute MIS trial. At the conclusion of the session the squirrel was released where it was trapped, the number of feces deposited in
the arena counted and the arena was cleaned with 70% ethanol. These procedures were the same
as those performed by Boon et al. (2007).
We quantified each squirrel’s behavior during the trials by scoring the videotaped trials using
JWatcher Video 1.0 (Blumstein and Daniel, 2007) and the same ethogram as described in Boon
et al. (2007). During the OF trial, we recorded latencies, rates and proportions of time the squirrel

14

spent engaged in a variety of activities that we did not consider to be unique behaviors or traits,
but which we hoped would collectively represent behaviors such as activity or aggression. These
behavioral measurements included the proportion of time spent walking, sniffing, chewing (gnawing at arena sides), rearing, grooming, scanning (a clear movement of the head in a side to side
manner), and still. These measurements were mutually exclusive. We also recorded the proportion of time spent hanging from the arena lid, which is mutually exclusive with walking, rearing,
grooming, and still but not mutually exclusive to sniffing, chewing, or scanning. In addition to
these proportions, we recorded rates of jumping and rates of interactions with the false holes. The
measurements recorded during MIS trials were proportions of time spent in the third of the arena
closest to the mirror (front), and farthest from the mirror (back), and the proportion of time spent
stretching towards the mirror. We also recorded the rate of aggressive contact with the mirror (attacks), the rate of grunting vocalizations, the rate of crouches (tail positioned over head with hairs
erect) and the latency in seconds until the first attack and first approach towards the mirror. A. K.
Boon performed and analyzed all trials in 2005 (study 1). Trials from 2008, 2009, and 2010 (study
2) were performed by R. W. Taylor and 4 assistants and analyzed by R. W. Taylor and 5 assistants.
We tested the inter-observer reliability of our measurements by calculating the correlation between
two observers’ scores of the same trial (56 trials were scored by multiple observers). We removed
measurements with a reliability of less than 0.7 (see Appendix Table A.1; Martin and Bateson
1993).
Maternity has been determined with certainty by ear-tagging juveniles prior to first emergence
from their natal nest since 1989 on the Kloo and Sulphur study areas and since 2002 on the Agnes
study area; adoption in red squirrels is extremely rare (Gorrell et al., 2010). Tissue samples for paternity analysis have been collected since 2003. Paternity was assigned based on 16 microsatellite
loci using CERVUS 3.0 (Kalinowski et al., 2007) with matches accepted at 95% or greater probability and no more than 1 mismatch (detailed in Gunn et al. 2005; Lane et al. 2008b). The complete
pedigree included 7086 individuals, 819 of which were informative for the docility phenotype, and
451 were informative for OF and MIS phenotypes (Appendix Table A.2).

15

2.3

Statistical Analysis

We did not consider each of our measurements from the behavioral trials to be unique behaviors
and given the intercorrelation that necessarily results from mutually exclusive scores, we did not
attempt to interpret them as unique behaviors. Instead our goal was to collect many measurements
that we hoped would provide a reliable overall assessment of the behavior of squirrels under these
conditions that have previously been found to have important ecological and evolutionary consequences (Boon et al., 2007). We, therefore, used principal component analysis to reduce the
redundancy among our measurements and to identify the dominant axes of behavioral variation
in the OF and MIS trials. Principal components were calculated separately for the OF and MIS
measurements using a correlation matrix (Table 2.2; results). We obtained very similar loadings to
Boon et al. (2007), confirming that the correlation matrices are consistent across years so principal
components were calculated for all years combined. All further analyses used the scores calculated
from the first principal component loadings for each trial and will be referred to as OF PC1 and
MIS PC1.
We estimated the variance components for the squirrel struggle rate, and their first principal
components scores for the OF and MIS trials (interpreted as docility, activity and aggressiveness,
respectively; see results and Boon et al. 2007) using a mixed-effect ‘animal model’, which allows for variance structures associated with pedigrees (Henderson, 1984; Lynch and Walsh, 1998;
Kruuk, 2004; Wilson et al., 2009). We fitted the animal models using a Markov Chain Monte
Carlo for Generalized Linear Mixed Models (MCMCglmm) analysis in the R statistical package
(Hadfield, 2010; R Development Core Team, 2011). We assumed that all study areas functioned as
a single population and, therefore, estimated common variance components across all populations
in all analyses. Phenotypic (VP ), additive genetic (VA ), maternal (VM ), permanent environmental
(VI ) and cohort (VC ) variances and their covariances were estimated using a trivariate MCMCglmm
model with individual, additive genetic, dam and birth year as random effects. Variance components were estimated as the mode of the posterior distribution and 95% credible intervals are given.
√
Covariances were rescaled as correlations (r = Cova,b / VaVb ). The variances for aggression and
16

activity were not transformed into coefficients of variation because they were estimated from principal component scores and so had mean values of 0. Heritabilities, maternal effects, permanent
environmental effects were calculated for each MCMC sample by dividing the relevant variance
component by the total phenotypic variance (VA +VM +VI +VC ) and the mode and 95% credible
intervals of these posterior distributions are reported. Repeatabilities were estimated as the mode
of a posterior distribution generated by dividing the among-individual variance (genetic and nongenetic) by the sum of among- and within-individual variances (Lessells and Boag, 1987) for each
MCMC sample.
Covariances were supported when 95% credible intervals excluded zero. This is not applicable
to variances because they are bounded above zero, so we determined support of variances by
comparing deviance information criterion (DIC) values of the fitted models (DIC values for each
model are provided in Appendix Tables A.5, A.6 & A.7). DIC can be viewed as the Bayesian
equivalent to the Akaike information criterion (AIC) and the rules of thumb developed for AIC
(Burnham and Anderson, 1998) appear to also work well for DIC (Spiegelhalter et al., 2002). In
all cases the most complete models were within 2 DIC of the best model (Appednix Tables A.5,
A.6 & A.7) so we report values from the models that include all random effects to avoid biased
parameter estimates.
Because missing values are not allowed in random effects, unique dams were generated for
individuals whose dam was unknown and treated as founders in the pedigree. This generation of
dams allowed us to use all the information available in our estimates of the variance components,
but assumes that all individuals with unknown dams were unrelated.
To account for the effects of habituation we included as covariates for the OF and MIS scores
a fixed effect term for lifetime and yearly trial number. For struggle rate we included lifetime and
yearly handling event, which included handling events where no struggle test was performed, because handling for routine data collection is similar to the struggle rate trial. We also included a
quadratic term for trial numbers to account for a non-linear response to repeated trials. To control
for effects of seasonality and study area we included day of year and study area as continuous and

17

categorical fixed effects, respectively. Finally, we included observer as a fixed effect in the docility models. Following Wilson (2008) we only attempted to control for methodological variation
(measurement error) through the inclusion of fixed effects in our models, and did not attempt to
account for other biological sources of variation (e.g. age, sex, birth year, reproductive status or
mass) to avoid removing phenotypic variation that might be relevant to natural selection.
Priors for the reported models were slightly informative and generated by partitioning the variance in phenotype evenly among each random term and were given a low degree of belief (V =
diag(n)∗VP /r, nu= 0.2; Hadfield and Hadfield 2011). Altering the priors so that VP was not evenly
distributed had no effect on the results. We evaluated convergence by visually inspecting time series plots of the model parameters and assessing autocorrelation values (all were < 0.1 for reported
results). The posterior distribution of the animal models was sampled every 500 iterations after a
burn-in period of 50,000 iterations for a total of 2,000 samples.

2.4

Results

Red squirrel responses to the open field arena varied from active to sedentary behavior and the major axis of behavioral variation (Table 2.2) was best described as activity in a novel environment.
After release into the arena active individuals immediately began walking, sniffing, jumping, hanging and chewing while sedentary individuals remained still with longer latency until first movement
(Table 2.2). This variation in activity was captured by the first principal component (OF PC1) and
explained 32% of the behavioral variation in the open field test (Table 2.2). Many individuals were
able to hang from the top corners of the arena by clinging to seams in the walls of the arena. This
hanging was often accompanied by chewing directed at the walls and corner of the arena. The second principal component for the open field test separated individuals that spent much of the trial
hanging and chewing from those who did not and explained 16% of the behavioral variation (Appendix Tables A.3, A.4 for all PC loadings). Because the second principal component explained
relatively little variation and its biological relevance is not readily apparent we did not perform any
further analyses on this axis of variation.
18

Table 2.2: First principle component loadings for behaviors from an open field arena test (OF
PC1) and a mirror-image stimulation test (MIS PC1) in North American red squirrels. Behaviors
were measured as percentage of time unless otherwise noted. Latencies were log transformed
prior to principle component analysis. Additional principle component axis are provided in
Appendix Table A.3
Behavior
Walk
Jump rate
Hole rate
No. Pellets
Hang
Chew
Groom
Still
SD
% Total variance

OF PC1
0.49
0.44
0.31
0.29
0.25
0.24
−0.06
−0.52

Behavior
Front
Attack rate
Back
Attack latency
Approach latency

1.71
0.36

MIS PC1
0.49
0.37
−0.41
−0.47
−0.48

1.67
0.56

Red squirrel behavior in the arena immediately and noticeably changed upon exposure to their
mirror image. Individuals varied in their response with some immediately approaching and attacking the mirror and others retreating to the opposite end of the arena and adopting a passive
posture. This variation in behavior was reflected in the first principal component for the mirror image stimulation trial (MIS PC1) which differentiated aggressive interactions directed at the mirror
(e.g. approaching and attacking the mirror) from avoidance (e.g. retreating from the mirror and
staying in the back of the arena) and explained 56% of the total variation in behavioral response to
the mirror image (Table 2.2).
Red squirrels varied in their response to restraint in the mesh handling bag with some individuals struggling for the entire 30 seconds of the test and others remaining entirely still. We have
interpreted this behavior as a measure of docility.
The complete univariate models for struggle rate, which included permanent environment, additive genetic, dam and cohort as random effects, had the lowest DIC score and the complete
models for OF PC1 and MIS PC1 were within 2 DIC of the best models (Appendix Tables A.5 &
A.6). The most complete trivariate model also had the lowest DIC score (Appendix Table A.7).

19

Because our focus was on achieving the best estimates for each (co)variance we evaluated the most
complete trivariate model to avoid estimates that may be confounded in the reduced models.
For all three behavioral measures (OF PC1; MIS PC1; struggle rate) the addition of identity
as a random effect greatly improved the fit of the model (Appendix Tables A.5 & A.6). Along
with substantial repeatabilities (Table 2.4), this demonstrates that consistent individual differences
in behavior exist for all three of these variables that we have interpreted as representing activity,
aggression and docility (see above). Red squirrels were less active with repeated lifetime trials
and less docile with repeated yearly and lifetime handling events, however there was no effect of
trial number on aggression (Table 2.3). There was a quadratic component to the effect of handling
events on docility such that the effect of each successive handling event diminished. There was a
small positive effect of day of year on activity and docility. Red squirrels on the Kloo and Sulphur
study areas were more active, and squirrels on the Sulphur study area were more aggressive than
squirrels on the Agnes study area.

20

Table 2.3: Effects of habituation, day of year and study area on red squirrel docility, aggression and activity. Year trial number is the
trial number counted from the start of each year, while life trial number is the trial number counted over each individual’s entire life.
Trial numbers for docility include all handling events, even those in which docility was not measured. 95% credible intervals are given
in parentheses and those that exclude zero are indicated in bold. The effects of the Kloo and Sulphur study areas are assessed relative to
the Agnes study area.
Docility
Intercept
Life Trial No.
Life Trial No.2
Year Trial No.
Year Trial No.2
Day of Year
Kloo
Sulphur

Aggression

15.494 (13.614 to 17.944)
–0.057 (–0.088 to –0.011)
0.00043 (0.00005 to 0.00078)
–0.204 (–0.312 to –0.138)
0.004 (0.002 to 0.005)
0.010 (0.004 to 0.018)
1.484 (0.201 to 2.503)
0.784 (–0.506 to 2.227)

–0.187 (–1.666 to 0.943)
0.784 (–0.588 to 1.834)
0.650 (–0.285 to 1.705) –0.841 (–1.929 to –0.100)
–0.101 (–0.357 to 0.108)
0.148 (–0.050 to 0.377)
–0.118 (–1.894 to 1.852) –0.962 (–3.045 to 0.405)
–0.012 (–0.651 to 0.383)
0.234 (–0.237 to 0.749)
–0.002 (–0.007 to 0.003)
0.005 (0.001 to 0.010)
0.295 (–0.124 to 0.840)
0.565 (0.102 to 0.972)
0.617 (0.046 to 1.019)
1.020 (0.563 to 1.467)

21

Activity

Table 2.4: Heritability (h2 = VA /VP ), maternal effects (m2 = VM /VP ), permanent environmental effects (PE = VPE /VP ), cohort effects
(C = VC /VP ), repeatability ([VA +VM +VPE +VC ]/VP ), and the mean trait value. Variances were estimated using a trivariate model.
Variances are calculated as the mode of the posterior distribution with 95% credible intervals in parentheses and are bound above zero.
Coefficients of variation (CV = 100 × standard deviation / mean) are given for docility. Because aggression and activity are scores
from a principal component analysis using a correlation matrix the trait means are 0 and coefficients of variation cannot be calculated.
h2

m2

PE

C

Repeatability

0.07 (0.03 to 0.10)
0.09 (0.03 to 0.18)
0.15 (0.05 to 0.26)

0.16 (0.08 to 0.21)
0.07 (0.03 to 0.20)
0.08 (0.04 to 0.21)

0.07 (0.03 to 0.17)
0.07 (0.03 to 0.23)
0.09 (0.04 to 0.27)

0.41 (0.36 to 0.49)
0.44 (0.33 to 0.56)
0.51 (0.40 to 0.63)

CVA

CVM

CVPE

CVC

Mean

15.54 (9.37 to 18.54)

10.67 (7.21 to 13.81)

16.99 (12.78 to 19.97)

10.24 (6.75 to 18.89)

18.65

Docility
0.09 (0.05 to 0.19)
Aggression 0.12 (0.03 to 0.22)
Activity
0.08 (0.03 to 0.19)

Docility

22

Heritabilities were low for all traits (0.09 − 0.12) as were cohort effects (0.07 − 0.09), maternal
effects on aggression and docility (0.07 − 0.09) and permanent environmental effects on activity
and aggression (0.07 − 0.08). However, maternal effects on activity (0.15) and permanent environmental effects on docility (0.16) were nearly twice as strong as the other effects (Table 2.4).
Activity and aggression were positively phenotypically correlated (rP = 0.40; 0.26 to 0.50),
while docility was negatively correlated with both activity (rP = −0.20; −0.30 to −0.08) and
aggression (rP = −0.12; −0.25 to −0.01), confirming the presence of a behavioral correlation
(Table 2.5). The genetic correlations were in the same direction as the phenotypic correlations
but varied in strength. There was a strong positive genetic correlation between aggression and
activity (rG = 0.68; 0.12 to 0.87), and a moderate negative correlation between aggression and
docility (rG = −0.49; −0.81 to 0.07) and between activity and docility (rG = −0.45; −0.71 to
0.20), though the credible intervals between docility and the other traits overlapped with zero (Table 2.5). There were also a maternal effect correlations (rM = 0.58; 0.01 to 0.81) and a permanent
environmental effect correlation (rPE = 0.61; 0.03 to 0.83) between activity and aggression and
a permanent environmental effect correlation between activity and docility (rPE = −0.45; −0.74
to 0.01) that overlapped with zero (Table 2.6). We did not find any support for cohort effects
correlations (Table 2.6).

2.5

Discussion

Theoretical studies of the evolution or maintenance of animal personalities often make the assumption that personalities are heritable (reviewed in Dingemanse and Wolf 2010). For instance
Wolf et al. (2007) proposed that correlational selection between life-history and personality traits
can give rise to suites of personality traits maintained by frequency-dependent selection. Similarly, Stamps and Biro (Stamps, 2007; Biro and Stamps, 2008) hypothesized that selection will
favor certain combinations of productivity (e.g. growth rates, or fecundity) and behavior leading
to stable personalities. If variable selection is to contribute to the generation and maintenance of

23

Table 2.5: Additive genetic and phenotypic variances, covariances and correlations (G-matrix and
P-matrix) of red squirrel personality traits. Variances are indicated along the diagonal, the upper
triangle contains correlations and the lower triangle covariances. Variances are calculated as the
mode of the posterior distribution with 95% credible intervals in parentheses and are bounded
above zero. Correlations and covariances that were different from zero (based on 95% credible
intervals) are indicated in bold.
Genetic
Docility
Docility
Aggression
Activity

Aggression

Activity

8.40 (3.05 to 11.95)
–0.50 (–1.71 to 0.21)
–0.45 (–1.39 to 0.27)

–0.49 (–0.81 to 0.07)
0.35 (0.11 to 0.74)
0.11 (0.01 to 0.47)

–0.45 (–0.71 to 0.20)
0.68 (0.12 to 0.87)
0.23 (0.09 to 0.57)

Aggression

Activity

Phenotypic
Docility

Docility
63.35 (58.02 to 72.33) –0.12 (–0.25 to –0.01)
Aggression –1.61 (–3.68 to –0.13)
3.25 (2.72 to 3.94)
Activity
–2.48 (–4.34 to –0.96) 1.31 (0.77 to 1.63)

–0.20 (–0.30 to –0.08)
0.40 (0.26 to 0.50)
2.89 (2.43 to 3.63)

personalities, then the personality traits on which selection acts must be heritable.
We found low heritabilities that were smaller than typically found for behavioral traits (mean
h2 = 0.30 ± 0.02 reviewed by Mousseau and Roff 1987; Stirling et al. 2002) and substantially
lower than our estimates of repeatability, which sets the upper bound for heritability (Hoffmann,
2000; Dohm, 2002; Bell et al., 2009). Small heritabilities may be the consequence of stabilizing
or directional selection eroding additive genetic variation or due to large environmental variances
(Barton and Turelli, 1989; Roff, 2007).
We found support for habituation in activity and docility, but not for aggression. The novelty
of the open-field arena is an important component of the testing environment, which is altered with
repeated exposure and may explain why individuals grew less active with subsequent trials (Martin
and Réale, 2008; Archer, 1973). Red squirrels also grew less docile with increased handling events
both within year and over their lifetime showing that the intensity of their reaction to handling
decreased with experience.
Red squirrels experience large yearly fluctuations in their main food source, seeds from white
spruce cones (Picea glauca; McAdam and Boutin 2003a; Boutin et al. 2006; LaMontagne et al.
24

Table 2.6: Maternal, permanent environmental and cohort (birth year) variances, covariances and
correlations of red squirrel personality traits. Variances are indicated along the diagonal, the
upper triangle contains correlations and the lower triangle covariances. Variances are calculated
as the mode of the posterior distribution with 95% credible intervals in parentheses and are bound
above zero. Correlations that were different from zero (based on 95% credible intervals) are
indicated in bold.
Maternal
Docility

Aggression

Docility
3.96 (1.81 to 6.63) –0.22 (–0.65 to 0.25)
Aggression –0.24 (–0.91 to 0.33) 0.24 (0.10 to 0.59)
Activity
–0.41 (–1.13 to 0.21) 0.16 (–0.03 to 0.43)

Activity
–0.31 (–0.67 to 0.16)
0.58 (0.01 to 0.81)
0.47 (0.16 to 0.78)

Permanent Environmental
Docility
Docility
10.3 (5.68 to 13.86)
Aggression –0.3 (–1.53 to 0.44)
Activity
–0.4 (–1.66 to 0.06)

Aggression

Activity

–0.26 (–0.71 to 0.19)
0.29 (0.10 to 0.70)
0.05 (–0.05 to 0.44)

–0.45 (–0.74 to 0.01)
0.61 (0.03 to 0.83)
0.34 (0.12 to 0.64)

Aggression

Activity

Cohort Effect
Docility
Docility
Aggression
Activity

3.64 (1.59 to 2.41) –0.02 (–0.65 to 0.58) –0.16 (–0.70 to 0.49)
–0.02 (–1.40 to 1.09) 0.20 (0.09 to 0.89)
0.16 (–0.55 to 0.73)
–0.07 (–1.59 to 0.97) 0.01 (–0.32 to 0.38) 0.30 (0.09 to 0.95)

2005), and fluctuations in these resources have long lasting cohort effects on red squirrel fitness
and life-history traits (Descamps et al., 2008). We found support for small cohort effects (approximately equal to VA ) on all three personality traits. Permanent environmental effects represented a
larger proportion of individual variation in docility (approximately twice VA ) but low amounts of
variation in activity and aggression (approximately equal to VA ). Permanent environmental effects
represent the effects of an individual’s environment that are consistent over the individual’s lifetime
(Roff, 1997). Individual red squirrels typically maintain a consistent territory over their lifetime
(Smith 1968; but see Boutin et al. 1993; Price and Boutin 1993) and the environmental effect of the
quality of their territory could contribute to permanent environmental effects on behavior. However, because red squirrels rarely change territories (Berteaux and Boutin, 2000) ‘individual’ and
‘territory’ are too confounded to reliably distinguish in the analysis of these personality data.

25

Determining whether phenotypic correlations adequately reflect underlying genetic correlations requires large sample size (Kruuk, 2004) and consequently the credible intervals around our
estimates were wide. The phenotypic correlations between docility and aggression (rP = −0.12)
and docility and activity (rP = −0.20) were low and the 95% credible intervals for genetic, maternal, permanent environmental and cohort correlations overlapped with zero. However, the overlap
was small for the genetic correlation between docility and aggression (rG = −0.49) and the permanent environmental effect correlation between docility and activity (rG = −0.45). We were able
to detect a more strongly supported genetic correlation (rG = 0.68), a maternal effects correlation
(rM = 0.58) and a permanent environmental correlation (rPE = 0.61) between aggression and activity. Together these results show that the degree to which genetic and other correlations can be
inferred from phenotypes alone can depend on the particular traits being considered even within
a single class of traits within a single species (Kruuk and Hadfield, 2007). Furthermore, other
sources of covariation such as maternal and permanent environmental covariation may be important components of observed phenotypic correlations between behaviors. Here we were unable
to assess whether these genetic correlations were due to pleiotropy or linkage disequilibrium, but
are currently performing selection analyses to determine whether contemporary patterns of correlational selection are consistent with the strong positive genetic correlations that we found among
activity and aggression.
We found stronger maternal effects for activity than docility and aggression and maternal effects on activity and aggression were correlated. The maternal effects correlation shows that the
effect of maternal environment on activity is strongly correlated with the effect on aggression, but
as maternal effects were weak the magnitude of these correlations may be misleading as other
sources of variation play a large role in determining the phenotypes of offspring. Maternal effects
have been found for personality traits in captive populations (Forstmeier et al., 2004; van Oers
et al., 2004) and have been examined in a few wild populations (Table 2.1), however, to our knowledge, this is the first documentation of maternal effects on a personality trait in a wild population.
We did not have the power to further separate the maternal effects into maternal environmental and

26

maternal genetic effects so can not reject the possibility that part of the maternal variance estimated
here is of genetic origin (Wilson and Réale, 2006).
Maternal hormonal responses to environmental variation and consequent differential early hormone exposure of offspring (i.e. hormone-mediated maternal effects) could have persistent consequences on personality traits and generate the correlations between personality traits that we
observed. For example, increased early exposure to androgens can increase aggression (Dloniak
et al., 2006; Eising et al., 2006; Mann and Svare, 1983) or simultaneously increase both aggression
and activity (Pasterski et al., 2007). In contrast, heightened early exposure to glucocorticoids may
decrease activity (Koolhaas et al., 1999; Wilcoxon and Redei, 2007). The effects of early exposure
to heightened androgens or glucocorticoids on personality traits can persist into adulthood (Eising
et al., 2006) and perhaps across generations through epigenetic programming of neuroendocrine
traits (Champagne, 2008; Meaney, 2001; Weinstock, 2008). We are currently investigating the
hormonal responses of female red squirrels to environmental variation and associations between
maternal hormone concentrations and the behavioral attributes of their offspring.
The persistence of consistent individual variation and covariation in behavior across a wide
range of taxa has led to many adaptive hypotheses that explicitly or implicitly assume sufficient
underlying genetic variation for these personalities to evolve or be maintained (e.g. (Stamps, 2007;
Wolf et al., 2007; Biro and Stamps, 2008; Houston, 2010)). Here we have not only provided
evidence of the genetic basis to personality in red squirrels, but have also identified maternal effects
as a potentially important source of variation in the personality of a wild vertebrate. Further studies
of the inheritance of personality traits in a variety of wild organisms are needed before general
patterns will emerge regarding differences in the magnitude of genetic, maternal and environmental
sources of variation among personality traits. These might also shed further light on the differences
that appear to exist between sources of variation in behaviors in the wild (Stirling et al., 2002) and
other traits typically studied from a quantitative genetic perspective (Houle, 1992). Such studies
are needed if we are to evaluate the potential for current and future models to describe the evolution
of persistent individual differences in behavior of wild animals.

27

Chapter 3
FLUCTUATING AND NONLINEAR SELECTION ON BEHAVIOR IN A WILD
POPULATION OF RED SQUIRRELS

Taylor, R. W., Humphries, M. M., Boutin, S., Gorrell, J. C., Coltman, D. W., and McAdam, A.
G., Fluctuating and nonlinear selection on behavior in a wild population of red squirrels. To be
submitted.

Author Contributions
R. W. Taylor and A. G. McAdam developed the concept of the paper. R. W. Taylor collected
behavioral data. All authors contriubted to the collection of fitness data. J. C. Gorrell and D. W.
Coltman performed the paternity assignment. R. W. Taylor performed all other analyses and wrote
the paper. All other authors provided intellectual insight and comments.

3.1

Introduction

Fluctuating selection has long been proposed as a mechanism by which genetic variation might be
maintained (Levins, 1968; Ellner and Hariston Jr, 1994; Ellner and Sasaki, 1996) despite strong
directional selection being common (Kingsolver et al., 2001). If selection fluctuates strongly in
direction over time then overall directional selection may be reduced, limiting any erosion of genetic variation (Bell, 2010). Fluctuating selection has been documented in a few systems (Bell,
2010), and enough longitudinal studies of selection have been undertaken that Siepielski et al.
(2009) were able to conduct a meta-analysis documenting the prevalence of variation in selection
(but see Morrissey and Hadfield 2012). However, new longitudinal studies of the ecological basis
of variation in selection will be necessary for a better understanding of fluctuating selection.

28

Although behavior was previously thought to be very labile, the persistence of individual differences in behavior has stimulated much recent research (Gosling, 2001; Sih et al., 2004a,b; Réale
et al., 2007, 2010a), with persistent behavioral differences increasingly accepted as important traits
with ecological and evolutionary consequences (Sih et al., 2004a,b; Réale et al., 2010a). Individual
differences in behavior have been shown to have important fitness effects in a number of systems
(Dingemanse and Réale, 2005; Smith and Blumstein, 2008) and fluctuating selection has been proposed as a mechanism that might maintain these individual differences (Dingemanse et al., 2004;
Boon et al., 2007; Dingemanse and Wolf, 2010).
Behavioral aggression and activity are are often positively correlated, both phenotypically and
genetically, as part of a ‘proactive-reactive’ behavioral syndrome (Koolhaas et al., 1999; Réale
et al., 2010b). At the ‘proactive’ end of the continuum individuals are more aggressive, active, have
a lower stress response and explore superficially. Recently, Réale et al. (2010b) advocated for the
integration of behavior into the pace-of-life syndrome concept, which has its roots in the concepts
of r- and K-selection (MacArthur and Wilson, 1967). Réale et al. posit that ‘proactive’ (Koolhaas
et al., 1999) individuals will outperform ‘reactive’ individuals in r-selected low-competition environments with the opposite in K-selected high-competition environments. As a result, the direction
of selection on pace-of-life strategies, and associated behavioral traits, may fluctuate in responses
to changes in important environmental conditions.
Consistent individual differences in behavior have been previously documented in a population
of North American red squirrels (Tamiasciurus hudsonicus) near Kluane national park, Canada
(Boon et al., 2007; Taylor et al., 2012). Furthermore, northern populations of red squirrels experience large fluctuations in available resources due to the synchronization of white spruce (Picea
glauca) cone production (Boutin et al., 2006). Additionally Boon et al. (2007) found evidence that
the relationship between dam behavior and juvenile growth rates and survival was not consistent
across years. This presents an ideal system for a comprehensive analysis of fluctuating selection
on behavior. We hypothesized that ecological changes in resource production through time led
to fluctuating selection, which could potentially maintain the variation in behavioral traits that we

29

have previously documented.
Individual differences in red squirrel behavior have been quantified using open-field trials, a
mirror-image stimulation test and handling tests (Boon et al., 2007; Taylor et al., 2012) and have
been interpreted as measurements of activity, aggression and docility. Activity characterizes motor
movements in response to a novel environment and has been considered a measure of exploration
in other systems (e.g. Dingemanse et al. 2002; Careau et al. 2008; Walsh and Cummins 1976). In
Kluane red squirrels aggression and activity are positively genetically correlated, which is consistent with a ‘proactive-reactive’ behavioral syndrome (Taylor et al., 2012). Docility is negatively
phenotypically correlated with aggression and activity (Boon et al., 2007; Taylor et al., 2012), but
is not simply the opposite of either and is best thought of as the response to stressful confinement.
Docile individuals would fall on the reactive end of the spectrum (Réale et al., 2010b).
Spruce masting results in large fluctuations in available cones, the main food source for northern red squirrels (Fletcher et al. in prep.), that can vary annually across 3 orders of magnitude
(Boutin et al., 2006). McAdam et al. (in prep) have shown that these resource pulses have strong
effects on juvenile competition for recruitment, which results in episodic alternating r- and Kselected environments. Thus, the general pace-of-life syndrome model proposed by Réale et al.
(2010b) would predict that high activity, high aggression and low docility are favored in r-selected,
high resource, ‘fast’ pace-of-life environments, with the opposite phenotypes favored in ‘slow’ Kselected environments.
However, the biology of the red squirrel system necessitates an alternate prediction for aggression. Because red squirrels are highly territorial (Smith, 1968), and a territory is required for
overwinter survival (Larsen and Boutin, 1994), juvenile recruitment depends on the availability of
vacant territories and not food resource availability per se. After emergence from their nest, juveniles search for territory vacancies, a process that often involves interactions with other juvenile
squirrels. It is not clear how red squirrels resolve conflicts, as physical contact and fighting is very
rarely observed (Dantzer et al., 2012), but it is reasonable to suspect that more aggressive individuals will be better able to win these contests over vacant territories. When competition for vacant

30

territories is high in low resource environments there are many juveniles for every vacancy, which
increases the probability of intense and direct competition among juveniles for limited vacancies.
Therefore, we predict that high-competition, low-resource, environments will favor aggressive and
docile individuals but select against active individuals.

3.2

Methods

A population of wild red squirrels was monitored since 1989 in the southwest Yukon (61◦ N,
138◦ W), and a detailed description of the population and general methods can be found in McAdam
et al. (2007). Red squirrels are small (150 − 250 g), diurnal, semi-arboreal rodents that inhabit
much of forested North America (Steele, 1998). White spruce cone production has been quantified, using visual cone counts since 1989 (LaMontagne et al., 2005). Individuals of both sexes
defend exclusive year-round territories (Smith, 1968). All squirrels in each of two study areas
were uniquely marked with numbered ear-tags as nestlings or at first capture after emergence from
the natal nest, and followed from birth until death. Nests of lactating females were entered immediately after parturition when young were counted and weighed, then again when young were
approximately 25 days age to tag and re-weigh the young for the estimation of growth rates (grams
per day; (McAdam and Boutin, 2003a)). Parturition dates were calculated based on weight loss of
the mother recorded during trapping or from the weight of young during the first nest entry.
We used annual reproductive success (ARS), calculated as the number of offspring that survived overwinter to the following spring (March 1st), as our measure of fitness. The study areas
are large relative to the dispersal distance of juvenile squirrels (Berteaux and Boutin, 2000) and
previous comparisons of the survival of juveniles on the edge of our study areas compared to survival of juveniles in the core supports our assumption that disappearance represents death (Kerr
et al., 2007; McAdam et al., 2007). We also investigated selection for two components of annual
reproductive success: fecundity (or siring success for males) and offspring overwinter survival
(OWS). Fecundity was calculated as the total number of offspring produced by a female each
season; similarly, siring success was the total number of offspring sired by a male each season.
31

OWS was calculated as the proportion of offspring produced by an individual that survived to the
following spring, and therefore could not be determined for individuals that did not produce any
offspring. Maternity was assigned during nest entries (see above) and paternity was assigned based
on 16 microsatellite loci using CERVUS 3.0 (Kalinowski et al., 2007) with matches accepted at
95% or greater probability and no more than one mismatch (detailed in Gunn et al. 2005; Lane
et al. 2008b).

3.2.1

Behavioral Traits

We measured behavioral traits of individuals from two study areas that have been monitored continuously since 1989. Following Boon et al. (2007); Taylor et al. (2012), we used three behavioral
trials to measure red squirrel behavioral traits. Here we give a brief description of the behavioral
methods, but full details can be found in Boon et al. (2007); Taylor et al. (2012). The first test, an
open field (OF) trial in a portable arena, was used to measure an individual’s activity, exploration
and stress response in a novel environment (Walsh and Cummins, 1976; Martin and Réale, 2008).
The second test was a mirror-image stimulation (MIS) that measured sociability and aggression
(Svendsen and Armitage, 1973). The testing arena for the OF and MIS tests was a 60 × 80 × 50cm
white corrugated-plastic box with a clear acrylic lid through which the behavior of the squirrel was
recorded with a digital video camera. Four blind holes were placed in the floor to provide the focal
squirrel with the opportunity to explore. A 45 × 30 cm mirror at one end of the arena was exposed
during the MIS portion of the session. OF and MIS trials were both performed in the same testing
session (trial). OF performance was tested first so that it would also serve as a habituation period
for the MIS trial. Individual squirrels were tested within one hour of being trapped on their territory. The third test, conducted during routine handling events, measured docility as the struggle
rate of individuals confined in a mesh handling bag. To measure docility, squirrels were immediately transferred from the trap into a handling bag and placed onto dry ground and the proportion of
time the squirrel spent struggling was measured over 30 seconds. The exclusive territoriality of red
squirrel allowed for complete enumeration of the study population through targeted trapping and

32

behavioral observations. Importantly the ability to target individuals allows us to avoid sampling
bias (Biro and Stamps, 2008; Boon et al., 2008). In 2005, 2008, 2009 and 2010 we performed 359
OF and MIS trials on 105 female and 130 male red squirrels and 2266 docility tests on 141 female
and 209 male squirrels for which we had fitness data.
We quantified each squirrel’s behavior by scoring the rate or proportion of time each squirrel
spent performing specific behaviors from the videotaped trials using JWatcher Video 1.0 (Blumstein and Daniel, 2007) and the same ethogram as Boon et al. (2007); Taylor et al. (2012). We did
not consider each of our measurements from the behavioral trials to be unique behaviors, and given
the intercorrelation that necessarily results from mutually exclusive scores, we did not attempt to
interpret them as unique behaviors. Instead our goal was to collect many measurements that we
hoped would provide a reliable overall assessment of the behavior of squirrels under these conditions that have previously been found to have important ecological and evolutionary consequences
(Boon et al., 2007; Taylor et al., 2012). We used the principal component loadings from Taylor
et al. (2012) to calculate behavioral scores for each trial. All further analyses used the scores calculated from the first principal component loadings for each trial and will be referred to as activity
(OF principal component 1) and aggression (MIS principal component 1.)

3.2.2

Selection

We controlled for a number of methodological sources of variation in behavior prior to evaluating selection on the behavioral traits by extracting best linear unbiased predictors (BLUPs) from
linear mixed-effects models with individual as the random effect (Pinheiro and Bates, 2000). To
account for effects of habituation we included as covariates, for the activity and aggression scores,
fixed effect terms for lifetime and yearly trial number. For struggle rate we included lifetime and
yearly handling event, which included handling events where no struggle test was performed, because handling for routine data collection is similar to the struggle rate trial. We also included a
quadratic term for trial numbers to account for a nonlinear response to repeated trials. To control
for effects of seasonality and study area we included day of year as a continuous fixed effect and

33

study area as a categorical fixed effect. Finally, we included observer as a fixed effect in the docility models. Following (Wilson, 2008), we only attempted to control for methodological variation
(measurement error) through the inclusion of fixed effects in our models, and did not attempt to
account for other biological sources of variation (e.g. age, sex, birth year, reproductive status or
mass) to avoid removing phenotypic variation that might be relevant to selection. Results from
these models are presented and discussed further by Taylor et al. (2012). The behavioral traits
analyzed here are, therefore, equivalent to those studied by Taylor et al. (2012).
To test whether selection on behavior fluctuated across years, we fitted generalized linear
mixed-models of male or female ARS that included the three behavioral traits (activity, aggression and docility) and year as a fixed factor. To test for temporal variation in selection we included
interactions between year and the behavioral traits. We modeled selection for each sex separately.
To account for pseudoreplication (Hurlbert, 1984) we fitted the models with study area as a fixed
effect and individual as a random effect. We tested the significance of the interactions between year
and the behavioral traits using a type II analysis of deviance. All generalized linear mixed-models
for annual reproductive success were fitted assuming a Poisson error distribution using a log link
function (Bolker et al., 2008).
We estimated linear selection differentials and gradients (Lande and Arnold, 1983) using linear
regression models of relative fitness separately for each year. Relative fitness was calculated and
traits were standardized for each study area-year combination, but standardized data from multiple
study areas were combined for phenotypic selection analyses within each year. Included as covariates in the selection gradient models were the three behavioral traits. We also estimated selection
gradients using a model that additionally included three life-history traits (offspring growth rate,
litter size and parturition date) that have been shown to affect fitness in red squirrels (McAdam and
Boutin, 2003a; Réale et al., 2003a). The inclusion of the life-history traits did not substantially
affect the selection gradients estimated from just the behavioral traits, so to reduce the number of
terms in the model we only included the behavioral traits. Standard errors for selection gradients
were generated using a delete-one jackknife procedure (Mitchell-Olds and Shaw, 1987).

34

We calculated summary statistics describing yearly selection gradients to facilitate comparison
of our results to recent meta-analysis of variation in selection (Kingsolver and Diamond, 2011;
Morrissey and Hadfield, 2012; Siepielski et al., 2011). If the ratio ‘mean of absolute values of
selection gradients’ : ‘absolute value of the mean of selection gradients’ is high, then fluctuating
selection reduced overall directional selection (sensu Kingsolver and Diamond 2011). A large
standard deviation among selection gradients relative to the mean of their standard errors indicates
that variation in selection was not due to sampling error (sensu Morrissey and Hadfield 2012). The
frequency of sign changes was calculated as the number of changes in direction between successive
years relative to n − 1, where n is the total number of years (sensu Siepielski et al. 2011).
We assessed how selection on the three behavioral traits was affected by the amount of competition among juveniles for vacant territories. Competition was measured for each year as the total
number of offspring produced in the population divided by the total number that recruited into the
population (the sum of each dam’s ARS). We first quantified the effects of this putative agent of
selection for dams and sires separately by examining the relationship between linear selection on
the behavioral traits and competition using generalized linear mixed models (GLMMs). ARS was
used as the dependent variable, and the three behavioral traits, competition, and all trait interactions with competition were included as covariates. Individual identity was included as a random
effect. We found strong and significant interactions between competition and linear selection for
dam behavioral traits (Appendix B.1). Our sample sizes were not sufficient to include additional
interactions between nonlinear selection and competition in the GLMMs without overfitting the
models. So, we continued our analysis of linear and nonlinear selection separately for high and
low competition environments, using data pooled for each sex across high and low-competition
years. We defined high-competition years to be those with juvenile competition greater than the
median, whereas we considered years with juvenile competition lower than the median to be lowcompetition years.
To better understand how behavior could affect ARS we performed a path analysis for each sex
in each environment. We included direct effects of parental behavior on ARS and indirect effects

35

through fecundity (or siring-success) and OWS. Coefficients for each path model were estimated
using linear regression of standardized trait values on relative fitness (Kingsolver and Schemske,
1991). We did not evaluate the significance of the path models, but instead used them as a guide for
further analysis using GLMMs. We evaluated selection separately for high- and low-competition
years using GLMMs that included all three behavioral traits, quadratic terms for each behavioral
trait and pairwise interactions between the behavioral traits. These models were reduced in a stepwise manner by removing the least significant term until further removal altered the significance
of the remaining terms. Non-significant second-order interaction terms were removed prior to
any first-order terms. We estimated linear and nonlinear selection gradients separately (Lande and
Arnold, 1983) for each sex from data pooled across all high-competition years and data pooled
across all low-competition years. Quadratic selection coefficients were doubled to give quadratic
selection gradients (Lande and Arnold, 1983; Stinchcombe et al., 2008). To further investigate the
mechanism of selection we analyzed the effect of behavior on two components of ARS: fecundity
(or siring success) and offspring overwinter survival (OWS). We used GLMMs with a Poisson
error distribution to estimate the effect of behavior on dam fecundity and male siring success. To
estimate the effect of behavior on OWS we used GLMMs with a binary error distribution weighted
by fecundity.
To further examine multivariate selection we performed a canonical rotation of the γ-matrix
to reveal nonlinear multivariate relationships on fitness (Phillips and Arnold, 1989; Blows and
Brooks, 2003). The resulting eigenvalues represent the nonlinearity of new combinations of the
original traits and the eigenvectors represent the contribution of the original traits to the eigenvalues. Higher magnitude eigenvalues indicate greater curvature along the axis of selection and
negative eigenvalues indicate a concave surface while positive values indicate a convex surface
(Phillips and Arnold, 1989). The significance of the canonical axes was estimated by entering the
eigenvectors into a quadratic regression on the original fitness values (Blows and Brooks, 2003).
Surfaces were visualized using thin-plate splines and the smoothing parameter was chosen to minimize general cross-validation scores (Phillips and Arnold, 1989; Brodie III et al., 1995). Unless

36

Table 3.1: The effect of year on selection for dam and sire behavioral traits through annual
reproductive success. Significance was calculated with Wald χ 2 tests from an analysis of
deviance. GLMMs were fitted with squirrel identity as a random effect and assumed a Poisson
error distribution.
Dams
χ2
Year
52.89
Study Area
2.04
Activity
0.80
Aggression
0.05
Docility
0.11
Year×Activty
24.31
Year×Aggression 25.79
Year×Docility
8.75

Sires

Df

P

χ2

Df

P

7
1
1
1
1
7
7
7

< 0.0001
0.15
0.37
0.82
0.74
< 0.005
< 0.001
0.27

41.85
0.00
1.71
0.32
0.03
11.60
14.22
5.76

6
1
1
1
1
6
6
6

< 0.0001
0.98
0.19
0.57
0.87
0.07
0.03
0.45

otherwise noted all analysis were performed using R 2.15.0 (R Development Core Team, 2012).
Generalized linear mixed models were fitted using the lme4 package (Bates et al., 2012), type II
anova was performed using the car package (Fox and Weisberg, 2011), fitness surfaces were visualized using the mgcv package (Wood, 2003, 2006) and the lattice package (Sarkar, 2008); other
figures were plotted using the ggplot2 package (Wickham, 2009).

3.3
3.3.1

Results
Temporal variation and annual selection

The addition of pairwise interactions between year and the three behavioral traits significantly
improved the models of dam and sire ARS showing that selection was not consistent across the
eight years that we studied (Table 3.3.1).
Therefore, we estimated linear selection gradients separately for each year. Standardized selection gradients for dam behavioral traits (Table 3.2 & Figure 3.1) fluctuated in sign across years
and were significantly positive for activity in 2004 (β = 0.65, CI = 0.05 to 1.25) but negative in
2009 (β = −0.97, CI = −1.78 to −0.16).

37

Table 3.2: Standardized linear selection gradients ± standard errors for dam and sire behavioral traits on annual reproductive success.
Significance was based on standard errors generated by jackknifing. The sample size (n) indicates the number of individuals.
Dams

Sires

Year

Aggression

Activity

Docility

n

Aggression

Activity

Docility

2003
2004
2005
2006
2007
2008
2009
2010

−0.13 ± 0.59
−0.70 ± 0.31∗
−0.42 ± 0.19∗
0.62 ± 0.26∗
−0.10 ± 0.25
0.78 ± 0.35∗
0.52 ± 0.29†
0.19 ± 0.24

0.82 ± 0.53
0.65 ± 0.29∗
0.30 ± 0.23
−0.14 ± 0.39
0.23 ± 0.33
−0.48 ± 0.35
−0.97 ± 0.40∗
0.10 ± 0.21

−0.11 ± 0.33
0.35 ± 0.25
0.15 ± 0.17
−0.37 ± 0.27
0.09 ± 0.27
−0.22 ± 0.40
−0.51 ± 0.20∗
0.03 ± 0.13

18
26
50
48
40
45
35
34

0.01 ± 0.50
0.17 ± 0.38
−0.30 ± 0.47
−0.12 ± 0.24
1.06 ± 0.65
1.06 ± 1.66
0.00 ± 0.21

−1.18 ± 0.64†
−0.58 ± 0.39
0.22 ± 0.33
0.12 ± 0.26
−1.09 ± 0.51∗
0.20 ± 0.96
−0.20 ± 0.25

0.09 ± 0.41
0.12 ± 0.26
0.11 ± 0.31
0.27 ± 0.20
0.19 ± 0.41
0.30 ± 0.30
−0.35 ± 0.21†

∗

P < 0.05, † P < 0.01

38

n
53
68
97
93
75
109
84

Standardized Selection Gradient ± SE

1.5

q

*
*

1.0

Activity
Aggression
Docility

.

*
q
q

q

0.5

q

0.0

q

q
q

−0.5
q

*

*

−1.0
*

−1.5

*

2003

2004

2005

2006

2007

2008

2009

2010

Year
Figure 3.1: Directional selection gradients ± standard errors of annual reproductive success for
dam aggression (MIS), activity (OF) and struggle rate (ST). 75th percentiles are given for
directional selection gradient magnitudes compiled by Kingsolver et al. (2001). * P < 0.05, . P <
0.1.

Selection for aggression was positive in 2006 (β = 0.62, CI = 0.11 to 1.14) and 2008 (β =
0.78, CI = 0.07 to 1.50), but negative in 2004 (β = −0.70, CI = −1.34 to −0.06) and 2005
(β = −0.42, CI = −0.81 to −0.03). Selection for docility was never significantly positive, but
was significantly negative in 2009 (β = −0.51, CI = −0.92 to −0.10). Selection on activity and
selection on docility occurred in the same direction in each of the eight years (correlation for all
years, γ = 0.76, CI = 0.12 to 0.95), but selection on activity was in the opposite direction of
aggression for six of those years (2004 to 2009; correlation between activity and aggression for
all years, γ = −0.79, CI = −0.96 to −0.20). Linear selection gradients for sire behavioral traits
also fluctuated in sign across years but were only significant for activity in 2008 (β = −1.09, CI
= −2.11 to −0.08; Table 3.2). Sire linear selection gradients for the three behavioral traits based
on annual reproductive success as our measure of fitness were not correlated with each other or

39

with dam selection gradients (but see Sexual conflict below).

40

Table 3.3: Reduced models of selection on behavioral traits for dam annual reproductive success, fecundity and offspring overwinter
survival. The full GLMM models included linear, pairwise interaction and quadratic terms and were reduced by removing the least
significant term until all remaining terms were significant. Interaction and quadratic terms were removed prior to linear terms starting
with the least significant. Study-area-year combination was a random effect and a Poisson error distribution was assumed. Random
effect grid-year variances for high-competition years: ARS = 0.34, Fecundity = 0, OWS = 0.84, and for low-competition years: ARS =
0.11, Fecundity = 0.11, OWS = 0.
High Competition
Fitness Component

Term

Annual
Reproductive
Success

Intercept
Aggression
Activity
Docility

Fecundity*

Intercept

Offspring
Overwinter
Survival

∗

Intercept
Aggression
Activity
Docility
Docility2
Aggression×Activity

Estimate

Z

Low Competition
P

Estimate

−1.13 ± 0.26 −4.28 < 0.001
0.19 ± 0.14
0.48 ± 0.18
2.68
0.007 −0.22 ± 0.11
−0.32 ± 0.18 −1.77
0.076
0.23 ± 0.11
−0.26 ± 0.15 −1.77
0.077
1.22 ± 0.04
−2.42 ± 0.42
0.61 ± 0.23
−0.45 ± 0.24
−0.41 ± 0.21
−0.20 ± 0.15
0.41 ± 0.19

No terms significantly explained fecundity

41

Z
1.34
−2.08
2.14

P
0.182
0.038
0.032

1.44 ± 0.12

11.62 < 0.001

−5.79 < 0.001 −0.95 ± 0.12
2.68
0.007 −0.44 ± 0.18
−1.88
0.061
0.39 ± 0.18
−1.93
0.054
−1.35
0.178
2.16
0.031

−7.93 < 0.001
−2.42
0.016
2.19
0.028

27.24 < 0.001

Table 3.4: Reduced models of selection on behavioral traits for sire annual reproductive success, siring success and offspring
overwinter survival. The full GLMM models included linear, pairwise interaction and quadratic terms and were reduced by removing
the least significant term until all remaining terms were significant. Interaction and quadratic terms were removed prior to linear terms
starting with the least significant. Study-area-year combination was a random effect and a Poisson error distribution was assumed.
High Competition
Fitness Component

Term

Low Competition

Estimate

Z

Annual
Reproductive
Success

Intercept
Aggression
Activity
Docility
Aggression2
Activity2
Docility2
Aggression×Activity

−2.65 ± 0.34
−0.08 ± 0.32
0.10 ± 0.32
0.37 ± 0.31
0.36 ± 0.19
−0.35 ± 0.34
−0.19 ± 0.23
−0.48 ± 0.27

−7.71
−0.25
0.32
1.19
1.87
−1.02
−0.83
−1.77

< 0.001 −1.14 ± 0.22
0.803
0.45 ± 0.32
0.752 −0.56 ± 0.26
0.235
0.062 −0.46 ± 0.24
0.306
0.406
0.076

Siring Success

Intercept
−0.93 ± 0.16
Aggression
−0.35 ± 0.19
Activity
0.01 ± 0.17
Docility
0.26 ± 0.13
2
Aggression
0.32 ± 0.11
Aggression×Activity −0.40 ± 0.16

−5.73
−1.85
0.06
1.90
2.79
−2.51

< 0.001
0.064
0.953
0.058
0.005
0.012

Offspring
Overwinter
Survival

Intercept
Aggression
Docility
Docility2

−1.67 ± 0.21 −7.79
0.45 ± 0.18
2.53
0.30 ± 0.22
1.33
−0.30 ± 0.15 −1.99

42

P

Estimate

−0.42 ± 0.30
0.40 ± 0.26
−0.61 ± 0.22
−0.08 ± 0.15
−0.21 ± 0.16

< 0.001 −0.70 ± 0.13
0.011
0.182
0.046

Z

P

−5.25 < 0.001
1.39
0.164
−2.16
0.031
−1.93

0.054

−1.41
1.56
−2.78
−0.56
−1.33

0.158
0.118
0.006
0.578
0.184

−5.35 < 0.001

Table 3.5: Vector of standardized directional selection gradients (β ) for annual reproductive success, and the matrix of standardized
quadratic and correlational selection gradients (γ). Linear and quadratic selection gradients were estimated in separate regressions.
Quadratic selection coefficients were doubled to give quadratic selection gradients (Stinchcombe et al. 2008). Significance was based
on standard errors generated by jackknifing.
High Competition
Dams
β
Aggression
Activity
Docility

Aggression

0.55 ± 0.14∗
−0.36 ± 0.20†
−0.31 ± 0.14∗

0.13 ± 0.41
0.14 ± 0.35
−0.10 ± 0.23

Sires
Activity

0.20 ± 0.50
−0.02 ± 0.27

Docility

β

Aggression

Activity

Docility

−0.22 ± 0.27

0.43 ± 0.48
−0.08 ± 0.29
0.27 ± 0.16†

1.67 ± 1.20
−0.66 ± 0.51
0.10 ± 0.40

−0.48 ± 0.47
0.23 ± 0.25

−0.40 ± 0.29

Low Competition
Dams
β
Aggression
Activity
Docility

Aggression

−0.28 ± 0.12∗
0.30 ± 0.12∗
0.12 ± 0.10

0.13 ± 0.26
−0.14 ± 0.20
−0.14 ± 0.15

Sires
Activity
0.31 ± 0.29
0.20 ± 0.18

Docility

0.16 ± 0.18

∗ P < 0.05, † P < 0.01

43

β

Aggression

0.02 ± 0.18
−0.60 ± 0.24∗
−0.11 ± 0.15

−0.72 ± 0.28∗
−0.14 ± 0.24
0.08 ± 0.24

Activity
0.74 ± 0.43†
0.03 ± 0.32

Docility

0.10 ± 0.22

3.3.2

Interactions with Competition and Path Analysis

We found strong and significant interactions between juvenile competition and linear selection
for dam behavioral traits (Appendix B.1). This finding prompted us to split the dataset into high
and low-competition years to investigate selection on different components of fitness using path
analysis (see above). The path diagrams presented in figure 3.2
illustrate the direct and indirect linear effects of aggression, activity and docility on ARS. Dam
behavior affected ARS indirectly through OWS, whereas sire behavior affected ARS both through
siring-success and OWS. The only direct effect of behavior on ARS was from sire aggression in
high-competition years. The proportion of variation in ARS explained by the models for each sex
in each competitive environment was high (R2 = 0.7 to 0.9) because fecundity/siring-success and
OWS are, not surprisingly, large linear components of ARS. However, behavior explained little of
the variation in fecundity/siring-success or OWS (R2 = 0.02 to 0.06). The path analysis does not
model nonlinear or multivariate selection, but we used it as a blueprint for further analyses that
included nonlinear selection.

3.3.3

Linear and nonlinear selection on ARS

We evaluated selection on dam and sire behavior separately for high and low-competition environments using GLMMs (Tables 3.4 and 3.5). In high-competition years there was selection
for increased dam aggression (b = 0.48 ± 0.18, Z = 2.68, P = 0.007; Table 3.4; Figure 3.3)
but selection for reduced activity (b = −0.32 ± 0.18, Z = −2.04, P = 0.042) and docility (b =
−0.26 ± 0.13, Z = −1.99, P = 0.047). In low-competition years selection on dam aggression
(b = −0.24 ± 0.11, Z = −2.10, P = 0.035) and activity (b = 0.23 ± 0.11, Z = 2.17, P = 0.03)
was reversed.
We investigated the effect of competition on nonlinear selection for dam and sire behavior
through ARS separately for high and low-competition environments using GLMMs (Tables 3.3
and 3.4). The addition of quadratic and correlational interaction terms uncovered selection through
sire ARS (Table 3.4; γ-matrix in table 3.5) that was not evident from linear interactions with com44

Aggression

Aggression
Dams
Fecundity

Fecundity

Activity

ARS
OWS

U

U
Scale
< 0.1
< 0.2
< 0.4
< 0.6
< 0.8
< 1

Docility
High Competition
Aggression
Siring
Success

Activity

U

OWS

Docility
Low Competition
Aggression
U

Siring
Success

ARS

Activity

Activity
OWS

OWS
Sires
Docility

Docility

Figure 3.2: Path models depicting direct and indirect effects of parental behavior on annual
reproductive success (ARS) in high and low-competition environments. Indirect effects of
behavior are mediated through fecundity (or siring-success) and offspring overwinter survival
(OWS). Variation due to error for fecundity, siring-success and OWS was high (U = 0.97 to 0.99)
and omitted for simplicity. Dashed lines depict negative coefficients, and line-width is
proportional to the standardized coefficients (see scale).

45

High Competition

Low Competition

●
●

●
●

●
●

●
●
●

ARS

●

2

●

●

●
●

●
●

1

●
●

0

●
●
●
●
●

●
● ●
●
●● ● ● ● ●
● ●
●
● ●
●
●
●
●

●

●

ARS

3

●
●
●
●
●
●
●

●
●●
●
●
●
● ● ●● ●
●● ● ● ●● ● ● ●
●●
●
●
●
●
●
● ● ●●
●
●
●
●
●
●
●
● ●●
●
●●
● ● ●
●
●
● ● ●● ● ● ● ●
●
●
●
●
●
●
●● ●
●
●
●
●
●●
●●
●
●●
●
●
●

1

0
−1
Activity

−2

6
5
4
3
2
1
0

●

1
Activity 0

1

0
−1 Aggression

●
●

●

●

●
●
●
●
●
●
● ●
●
●
● ●
●
●
●●
●
● ●●
●
●
●
●
●●
●
● ●
● ●
●●
●
● ●
●
●
● ●●
●●
●
● ● ●●
●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●●
●
●
● ●
●
●● ●
●
● ●
● ● ● ●
●
●●
●
●
●● ●
●
●
●
●
●
●
●
●●
●● ●
●
●
●
●●
●
●
● ● ● ●●
● ●
●

●

−1
−2

●
●

●

1

−1

2

0
Aggression

Figure 3.3: Fitness surfaces of ARS for dam aggression and activity in high and low-competition
years. Selection favored aggression (β = 0.55 ± 0.14, P < 0.05) and disfavored activity
(β = −0.36 ± 0.20, P < 0.1) in high-competition years. In low-competition years selection
favored activity (β = 0.30 ± 0.12, P < 0.05), but disfavored aggression
(β = −0.28 ± 0.12, P < 0.05). Surfaces were visualized using thin plate splines. Surfaces and
data points are conditioned on docility.

petition (Table 3.3) or the path analysis (Figure 3.2). In high-competition years we found negative
correlational selection between activity and aggression (b = −0.35±0.16, Z = −2.15, P = 0.032).
That is, sires with low activity and high aggression had higher fitness while sires with high activity
and low aggression had lower fitness (Figure 3.4).
In addition, we found evidence for disruptive quadratic selection for aggression (b = 0.31 ±
0.12, Z = 2.56, P = 0.01) and stabilizing selection for docility (b = −0.34±0.17, Z = −2.02, P =
0.043) in high-competition years. A quadratic term for stabilizing selection on activity (b =
−0.42 ± 0.24, Z = −1.77, P = 0.076) in high-competition years was also marginally significant.
In low-competition years low activity was favored (b = −0.60 ± 0.16, Z = −3.71, P < 0.001) and
we found stabilizing selection for aggression (b = −0.43 ± 0.15, Z = −2.83, P = 0.005).
We estimated linear and nonlinear selection gradients for dam and sire behavior separately in

46

High Competition

Low Competition

q
q

3

q

ARS

2

q
q
q
q
q
q
q
q
q q q
q qq
q q
q
qq q
q
qq
q q q
q
q q q q qq
qq qq q q q
qq
q q
q
q
q qq
q q
q
q
q
q
q qq qq qq
q
q q q q
q
q
q
q q
q qqqq q
q q
q
q qq q
qq q
q q
qqq
q q qq q qqqq q q q
q
q qqq q
q
qq
q
q qq q q
qq
qq
q q
q
q
qq qqq qqq q q q
qq
q q
q
q
q
qq qq q
q
qqqq q q q q q q q
qq q q
q qq
q
q
q
q
q
qq q q q qq q q q
q
q q
q q
q
q
q qq q q q qqqq q
q
q
q q qq
q q
q qqq q q
q qq
q q
qq
q
q
qq
q
q qq q
q

1
0
−1
2

ARS

q

q

q
q

1

0
Activity−1

2

−2

−1

5
4
3
2
1
0
−1
2

q
q
q
q
q
q
qq
q
q
q
q
q q
q
q
q
q
q
q
q
q
q
q qq
q
q
q
q
q q q
q
q
q
q
q q q
q
q q
q
q
q qq
q
q
q
q
q
q
q q
q
q q qq q q
q
q
q
qq
qq
q
q
q
q q qq q
q q
q
qq
q qq q q
q q
q
q
q
q q q qq
q
q
qq q q q
q
qq
q
q
qq
qq
q q
q
q
q
q q
q
q
q
q

q

q

1

0
Activity −1

1
0
Aggression

q

q

q

q

q
q

q

q

1

−1

2

0
Aggression

Figure 3.4: Fitness surfaces of ARS for sire aggression and activity in high and low-competition
years. Selection on aggression was disruptive in high-competition years
(γii = 1.67 ± 1.20, P > 0.1), but stabilizing in low-competition years
(γii = −0.72 ± 0.28, P < 0.05). Selection on activity fluctuated in the opposite direction and was
stabilizing in high-competition years (γii = −0.48 ± 0.47, P > 0.1), but disruptive in
low-competition years (γii = 0.72 ± 0.28, P < 0.05). Surfaces were visualized using thin plate
splines. Surfaces and data points are conditioned on docility.

high and low-competition environments (Table 3.5). The direction and shape of selection was
consistent with GLMMs (above). Visual inspection of the fitness surfaces (Figure 3.3) showed that
the positive quadratic selection on dam activity in low-competition environments was not strictly
disruptive (sensu Mitchell-Olds and Shaw 1987) because only one stationary maximum appeared
to exist.

3.3.4

Components of fitness

We continued investigating selection on the components of fitness separately for high and lowcompetition environments using GLMMs (Tables 3.3 and 3.4). We did not find any effect of
behavior on dam fecundity in either high or low-competition environments. Selection on behavior
for dam offspring OWS was very similar to selection for dam ARS (Table 3.3) with the addition of
47

significant correlational selection between activity and aggression (b = 0.32 ± 0.16, Z = 2.00, P =
0.046) in high-competition years. These results suggest that selection on dam behavior acts through
offspring OWS and not fecundity (Figure 3.2).
In low-competition environments there was no selection on sire behavior through offspring
OWS, so selection through ARS acted entirely through selection for siring-success (Table 3.3).
This contrasted with selection for sire behavior in high-competition environments. We found negative correlational selection between aggression and activity (b = −0.24 ± 0.06, Z = −4.11, P <
0.001), meaning that sires with dissimilar aggression and activity had higher siring-success. We
also found disruptive selection on aggression (b = 0.23 ± 0.04, Z = 5.38, P < 0.001) and positive
linear selection on docility for siring-success (b = 0.24 ± 0.05, Z = 4.39, P < 0.001). Selection on
aggression for offspring OWS was positive (b = 0.45 ± 0.18, Z = 2.53, P = 0.011), and we found
stabilizing selection for docility (b = −0.30 ± 0.15, Z = −1.99, P = 0.046). Selection gradients
for dam and sire OWS and fecundity or siring-success are presented in Appendix B.

3.3.5

Sexual Conflict

Selection on dam aggression through ARS was negatively correlated with selection on sire aggression through siring success (r = 0.77, CI = −0.96 to −0.04; Figure 3.5a). In contrast, we found
a positive correlation between selection for dam aggression through dam ARS and sire aggression
through OWS, though it was not significant (r = 0.62, CI = −0.25 to 0.94; Figure 3.5b).

3.3.6

Fitness surface estimation

We performed a canonical rotation of the γ matrix to investigate multivariate fitness surfaces
(Phillips and Arnold, 1989; Blows and Brooks, 2003).The M matrix of eigenvectors and their
associated eigenvalues are presented in Table 3.6. Eigenvectors for dam and sire behavioral traits
in the high-competition years and for sire traits in low-competition years had both positive (disruptive) and negative (stabilizing) eigenvalues suggesting that the fitness surfaces contained at least
one saddle and that no stationary point exists (Blows et al., 2003; Phillips and Arnold, 1989). In
48

Aggression

Docility

Aggression

Activity
Docility

Docility

Activity

Activity

Aggression

Figure 3.5: Fitness surfaces for sire behavioral traits in high competition years. Fitness surfaces
are plotted for 1st , 50th and 99th percentiles of the third behavioral trait. Surfaces were visualized
using thin plate splines

low-competition years the eigenvalues for dam traits were all positive suggesting that alternative
maxima may exist on the fitness surface. We found significant nonlinear selection on sire traits in
high and low-competition years so we will interpret these selection surfaces in more detail. The m1
eigenvector for sires in high-competition years represented strong convex selection on aggression,
which can be readily visualized in Figures 3.4 and 3.5. The remaining eigenvectors represent convex positive correlational selection (m2 ) and negative correlational selection (m3 ) between activity
and docility. Plotted apart these surfaces would be saddles with peaks at the +, + corners in the
case of m2 , and peaks at the +, − corners for m2 . However, they interact in complex ways and are
best interpreted through visualization (Figure 3.5). The m3 eigenvector for sire low-competition
years almost entirely reflects aggression and contrasts with a non-significant, positive eigenvector
(m1 ) in high-competition years. This pattern of selection reflects the fluctuating positive and negative γii gradients for aggression from the quadratic regression and does not reflect any correlational
selection.

49

Table 3.6: The M matrix of eigenvectors from the canonical analysis of γ for annual reproductive
success. The eigenvalue (λ ) of each eigenvector (m) is given in the first column.
High Competition
Dams
λ
m1
m2
m3

0.247
0.095
−0.228

Sires

Aggression

Activity

Docility

λ

−0.528
0.837
0.144

−0.846
−0.534
−0.001

0.076
−0.122
0.990

1.717
−0.325
−0.605∗

Aggression

Activity

Docility

0.989
0.061
−0.135

−0.147
0.504
−0.851

0.016
0.862
0.507

Activity

Docility

−0.999
−0.016
0.049

−0.018
0.999
−0.049

Low Competition
Dams
λ
m1
m2
m3
∗

Aggression

0.395
0.342
0.134 −0.603
0.077
0.720

Sires

Activity

Docility

λ

−0.822
−0.564
−0.082

−0.456
0.564
0.689

0.743†
0.097
−0.728∗

Aggression
0.048
0.050
0.998

P < 0.05, † P < 0.01

3.4
3.4.1

Discussion
Fluctuating selection

We found strong temporal fluctuations in selection on dam aggression and activity that involved
significant changes in the direction of selection across years. These results confirm and add support
to Boon et al. (2007) conclusion that selection for activity and aggression fluctuates across years.
Moreover, the magnitude of linear selection gradients for dam aggression and activity repeatedly
exceeded 95th percentile of selection gradients assembled by Kingsolver et al. (2001). In a metaanalysis of a large dataset, Siepielski et al. (2009) found that selection commonly varies from year
to year, but they were quick to acknowledge that measurement error may contribute substantially
to this variation. Morrissey and Hadfield (2012) re-analyzed a portion of Siepielski et al’s dataset
to show that the majority of variation in selection could be explained by measurement error, which
led to their conclusion that selection was actually “remarkably consistent in time” (Morrissey and
Hadfield, 2012). Nevertheless, fluctuating selection clearly plays a strong role in a number of
natural systems (e.g. Darwin’s finches; Grant and Grant 2002; see also Bell 2010) and remains

50

an intuitive and plausible mechanism for the maintenance of genetic variation. Our results, along
with a previous study showing that red squirrel behavioral traits are heritable (Taylor et al., 2012)
suggest that temporally fluctuating selection could play a large role in maintaining variation in
red squirrel behavior despite strong associations with fitness, a result that is consistent with the
oligogenic rather than infinitesimal model of selection (Bell, 2010).
Linear selection on sire aggression and activity did not fluctuate significantly through ARS as
it did for dam traits, but the mean frequency of changes in sign was the same for dam and sire
aggression and activity (0.5; Table 3.7) and the standard deviation of the selection coefficients
were very similar for dams (0.56) and sires (0.58). However, the mean of standard errors for sire
selection coefficients (0.53) were higher than those for dams (0.32), which may be due to selection
acting on multiple components of ARS in sires and nonlinear selection (see below).

51

Table 3.7: Summary statistics for standardized linear selection gradients (β ). A high ratio of mean of absolute values of selection
gradients to the absolute value of the mean of selection gradients indicates that fluctuations in selection reduced overall directional
selection (sensu Kingsolver and Diamond 2011). Large standard deviations among selection gradients relative to the mean of their
standard errors indicates, along with significant changes in sign (Table 2 and Figure 1), that variation in selection was not due to
sampling error (sensu Morrissey and Hadfield 2012). The frequency of sign changes was calculated as the number of changes in
direction between successive years relative to n − 1, where n is the total number of years (sensu Siepielski et al. 2011).
Dams

Sires

Aggression
Mean |β |
|Mean (β )|
SD of β
Mean SE of β
Frequency of β sign changes

Activity

Docility

Aggression

Activity

Docility

0.43
0.10
0.53
0.31
0.43

0.46
0.07
0.59
0.34
0.57

0.23
0.07
0.29
0.25
0.71

0.51
0.36
0.60
0.48
0.67

0.51
0.36
0.60
0.48
0.67

0.20
0.10
0.22
0.30
0.17

52

3.4.2

Competition and the Pace-of-Life Hypothesis

The opportunity for selection through offspring viability (survival over first winter, and subsequent
recruitment into the population) is strong in this system (McAdam and Boutin, 2003b), so we
hypothesized that variation in competition among juvenile squirrels would explain variation in the
direction of selection for behavioral traits. We found that selection favored aggressive dams when
competition was high and active dams when competition was low. These results are consistent with
our predictions that low-competition, high-resource environments would favor active individuals
and high-competition environments would favor aggressive individuals. A more integrated study
of the effect of competition on relationships between behavior, physiology and life-history will be
necessary to understand how these results fit into the pace-of-life framework (Réale et al., 2010b).
This opposing selection is surprising given that aggression and activity are positively genetically correlated, but opposing linear selection does not imply negative correlational selection. In
fact, GLMMs and surface-visualization indicated positive correlational selection for dam aggression and activity in high-competition years (Table 3.5 and Figure 3.3). Thus, correlational selection
on dam aggression and activity within high-competition years could help to maintain the genetic
correlation despite opposing patterns of linear selection. This positive correlational selection on
female behavior, however, was opposed by negative correlational selection on sire aggression and
activity in high-competition years (Table 3.5) indicating sexual conflict within the population (Figure 3.6; Trivers 1972; Chapman et al. 2003). These results highlight the complexity of interactions
between the environment, ecology, sex and the effects of behavioral phenotypes on fitness.

3.4.3

Fitness Components

Selection on dam behavioral traits for annual reproductive success acted almost entirely through
offspring overwinter survival (Table 3.3); we did not find any significant effect of dam behavior
on fecundity (correlations between dam selection gradients for ARS and OWS ranged from 0.94
to 0.99 all P < 0.001). Selection on sire behavior acted both through both siring success and
OWS. Sexual selection on sire behavior through siring success might act through a combination of
53

a

q

q

q

q
q

q

0.0

q

0.5
Dam ARS

Dam ARS

0.5

q

q

−0.5
−0.2 0.0
0.2
Siring Success

q

0.0

q

q

−0.5

q

b

q

0.4

−0.5
0.0
0.5
Sire Offspring OWS

Figure 3.6: The relationships between selection gradients for dam and sire aggression. On the
y-axes are selection gradients for dam aggression through annual reproductive success. On the
x-axes are selection gradients for sire aggression for a) siring success(r = −0.77, CI = −0.96 to
−0.04) and b) offspring overwinter survival (r = 0.62, CI = −0.25 to 0.94).

male-male competition (Lane et al., 2008a) and, potentially, also female mate-choice (see below).
Parental behavioral could directly affect OWS if parents intervened in offspring competition for
vacant territories or preparation for winter. However, these behaviors have not been observed,
so if they occur they would be exceedingly rare. Dams occasionally bequeath their territory to
a juvenile (Price and Boutin, 1993; Berteaux and Boutin, 2000), and this could be considered a
direct reflection of the mother’s behavior, but Boon et al. (2007) found no relationship between
bequeathal and dam behavior. In this population it is more likely that parental behavior influences
OWS indirectly either through correlated parental care, or through parent-offspring phenotypic
resemblance.
Because sires do not provide parental care, selection on sire behavior through OWS almost
surely acts through heritable behavioral components. We can test this assumption and hypothesize
how selection acts on dam behavior by comparing sire and dam selection gradients across years.
For selection that acts through heritable components we would expect selection through OWS
to be similar for sires and dams and we find that linear selection gradients for aggression were
marginally positively correlated (r = 0.67, CI = −0.16 to 0.95, P = 0.1). This suggests that
some component of selection for aggression could act through effects of heritable behaviors on
54

offspring survival. We also found stabilizing selection for sire docility, and marginally significant
stabilizing selection for dam docility through OWS in high-competition environments. We were
unable to estimate nonlinear selection gradients separately for each year, so we cannot compare
those, but the similarity in selection on sire and dame docility through offspring survival also hint
that selection on parental docility acts through heritable components. Finally, GLMMs did not
detect any selection for sire activity through OWS and we found no correlation between sire and
dam linear selection gradients for activity (r = 0.23, CI = −0.57 to 0.80, P = 0.6), which suggests
that selection for dam activity does not act through a heritable component and instead acts through
correlated traits or maternal effects. A further piece of evidence that supports this hypothesis
is that, in this population, there are stronger maternal effects on activity than aggression and that
aggression is more strongly heritable than activity (Taylor et al., 2012). Ideally we would also have
been able to directly measure selection on juvenile behavioral traits, but trappability of juveniles is
low immediately after emergence and juvenile death prior to measurement could cause substantial
bias (e.g. “the invisible fraction”; Grafen 1988; Hadfield 2008). The relative importance of parental
behavior and offspring behavior to offspring survival could be assessed directly in future studies by
measuring juvenile behavior before emergence or by looking for evidence of a genetic correlation
between personality and survival using the analysis of a multigenerational pedigree (Hadfield,
2008; Sinervo and McAdam, 2008) even if phenotypic behavior are not available for offspring.
Red squirrels perform elaborate and demanding mating chases in which males compete for opportunities to mate with each female on her one day of estrus (Smith, 1968), providing opportunity
for male-male competition (Lane et al., 2008a) and female mate choice. Lane et al. (2008a) characterized the mating system as scramble competition and identified home range size and the number
of oestrous females found by searching males as positive predictors of siring success. Additionally, Lane et al. (2008a) hypothesized that behavior might influence search ability. Here we do not
test this hypothesis directly but we do show that behavioral traits are also important predictors of
siring success and that their effects on siring success vary with environmental conditions. Future
analysis using detailed behavioral data collected during mating chases should directly address how

55

selection acts on behavior through mating chases.
Nonlinear selection on sire aggression through ARS fluctuated from convex in high-competition
years to concave in low-competition years. Visual inspection of the selection surfaces (Figure 3.4)
show two stationary points in high-competition years, meaning that selection was disruptive, and
a ridge in low-competition years exhibiting stabilizing selection (Mitchell-Olds and Shaw, 1987).
The opposite was true for sire activity with a concave surface in low-competition years and a
convex surface in high-competition years. The concave surface appears to be stabilizing, but the
convex surface has one stationary point so exhibits nonlinear directional selection.
Canonical rotation of the y-matrix exposed nonlinear stabilizing selection contrasting sire activity with docility in high-competition years (Table 3.6). Interpreting these values can be difficult,
so we visualized the interactions between the three traits by plotting the activity by docility surface
for three levels of aggression (Figure 3.5). It is clear from these graphs that the fitness surface for
sire traits in high-competition years is complex. There has been some discussion of the utility of
matrix diagonalization methods advocated by Blows (Blows, 2007a,b; Conner, 2007; Kruuk and
Garant, 2007). We found that canonical rotation highlighted nonlinear selection that was not evident in the y-matrix for selection on sire behavior in high-competition environments. So, in this
case it was clearly a useful tool in alerting us to multivariate-nonlinear selection (Blows, 2007a,b).
However, interpreting the resulting nonlinear selection in a biological meaningful manner remains
challenging (Conner, 2007; Kruuk and Garant, 2007).

3.4.4

Conclusion

We tested the hypothesis that ecological changes through time lead to fluctuating selection, which
might maintain variation in behavioral traits. Our results show that linear selection on red squirrel
dam aggression and activity is sometimes very strong, but fluctuates in sign across years depending
on the level of competition among juveniles for vacant territories. We also found that selection on
red squirrel behavior varies among components of fitness and between the sexes, and includes
nonlinear components between and within traits. These results, combined with previous work

56

showing that these behavioral traits are heritable (Taylor et al., 2012), suggest that behavioral
variation can be maintained by complex patterns of variation in selection across years, among
fitness components and between the sexes. Furthermore, congruent directional selection through
offspring overwinter survival on male and female aggression and stabilizing selection on docility
suggest that selection is acting on heritable components of aggression and docility in contrast to
selection on activity, which seems to be affected by either correlated maternal care or maternal
effects on behavior.

57

Chapter 4
DISASSORTATIVE MATE CHOICE ON BEHAVIOR IN RED
SQUIRRELS

Taylor, R. W., Lane, J. E., Humphries, M. M., Boutin, S., Gorrell, J. C., Coltman, D. W., and
McAdam, A. G., Disassortative mate choice on behavior in red squirrels. To be submitted.

Author Contributions
R. W. Taylor and A. G. McAdam developed the concept of the chapter. R. W. Taylor collected
behavioral trial data. R. W. Taylor and J. E. Lane collected mating chase behavioral data. All
authors contriubted to the collection of fitness data. J. C. Gorrell and D. W. Coltman performed
the paternity assignment. R. W. Taylor performed all other analyses and wrote the paper. A. G.
McAdam provided comments.

4.1

Introduction

Sexual selection occurs when individuals vary in reproductive success (Darwin, 1871). Sexual
selection can arise through intra-sexual competition or through inter-sexual mate choice (Bateson,
1983; Andersson and Iwasa, 1996; Kokko et al., 2003). With high investment in gametes and
high post-fertilization care, individuals should be more selective in their mates to maximize fitness
through benefits provided by mates (Darwin, 1871; Bateson, 1983; Andersson and Iwasa, 1996).
Benefits can be directly beneficial, as in food, shelter or protection, or indirectly beneficial through
an increase in the genetic quality of offspring. For example, female side-blotched lizards (Uta
stansburiana) prefer mates with experimentally improved high-quality territories, presumably for
the direct benefits offered by the territory (Calsbeek and Sinervo, 2002). But females also prefer-

58

entially fertilize male offspring with sperm from large males, presumably for the indirect benefit
of the large-male genes that confer high fitness (Calsbeek and Sinervo, 2002).
The direct and indirect benefits of potential mates may be context-dependent and may also
vary with environmental conditions. For example, in flagfish (Jordanella floridae) female preference for direct benefits of paternal care depends on salinity (Hale, 2008). Alternatively the mate
preference of female side-blotched lizards (Uta stansburiana) for direct benefits depends on the
social environment (Alonzo and Sinervo, 2001). Good genes might not only be context-dependent,
but their indirect benefits may also depend on the particular genotype of the choosing individual.
Mate preference for indirect, genetic benefits can be the same for all individuals in which case the
benefit of good genes is independent of the receiver’s genes. In such cases mate choice results in
the direct inheritance of good genes, which give an additive genetic benefit to offspring (and indirectly, parental) fitness. Alternatively, the benefit of genes may depend on interactions between the
male and female genomes (i.e. genetic compatibility), in which case different individuals might
exhibit different adaptive mate choice decisions. In this scenario mate choice is for compatible
genes, which could potentially operate on either a genome wide scale or at individual loci (Tregenza and Wedell, 2000). In these cases the genetic effect on fitness would be non-additive and
involve dominance and/or epistasis (Neff and Pitcher, 2005). For example, side-blotched lizards
exist in distinct genetic morphs that are differentiated by life-history and reproductive strategies
(Sinervo and Lively, 1996; Sinervo et al., 2000). Female side-blotched lizards exhibit assortative
mate choice by preferring males morphs with the same life-history strategy as their own (Bleay and
Sinervo, 2007). However, after their first clutch, females of the fast life-history morph switched
their preference to exhibit disassortative mate choice, suggesting that the compatibility of genes
can be context dependent (Bleay and Sinervo, 2007).
Typically gametic and post-fertilization investment is greater for females than males, leading to
female mate choice and strong competition among males for reproductive success (Darwin, 1871;
Trivers, 1972). However, whenever there are costs or constraints that limit male reproductive output and females vary in quality, then we can expect male mate choice to evolve as well. Male costs

59

could accrue from parental care (Sargent et al., 1986), sperm production (Nakatsuru and Kramer,
1982) or mate searching including predation risks (Magnhagen, 1991). Male choice has been documented numerous times (e.g. Herdman et al. 2004; Bateman and Fleming 2006); however, as
emphasized by Barry and Kokko (2010), male choice may not evolve if mating opportunities are
scarce even when there is variation in female quality and high costs to reproduction.
Consistent individual differences in behavioral traits are widespread (Réale et al., 2007) and
have been shown to be heritable (Taylor et al., 2012; Drent et al., 2002) and under selection in
a number of species (reviewed in Smith and Blumstein 2008). Mate choice based on behavioral
traits has been documented in humans and in animals (reviewed in Schuett et al. (2010), though
few studies have shown inter-individual variation in mate choice based on behavior Schuett et al.
(2010). In one notable exception, high-exploratory female zebra finches (Taeniopygia guttata)
were found to prefer high-exploratory males, but low-exploratory females had no preference for
male exploration (Schuett et al., 2011).
Red squirrels (Tamiasciurus hudsonicus) have been shown to exhibit consistent individual differences in aggression (response to a mirror image), activity in a novel environment (open field
arena) and docility (time spent still while being handled; Boon et al. 2007). Additionally, these
traits are heritable, experience maternal effects, and aggression and activity are positively genetically correlated in a behavioral syndrome (Taylor et al., 2012). Selection on these red squirrel
behavioral traits was also found to differ among components of fitness, between the sexes and includes linear and nonlinear components between and within traits that fluctuate over time (Taylor
et al. in prep; Chapter 3). Taylor et al. found that both aggressive males and intermediately docile
males produced offspring with enhanced overwinter survival. Males do not provide parental care,
but aggression and docility are heritable so presumably these good behavior genes were inherited by offspring, which enhanced their survival. Even if this is not the case, and other genes are
the targets of selection for overwinter survival, aggression and docility could provide indirect, but
reliable cues for the selection of good genes.
Fluctuations in selection on aggression have been linked to annual variation in the strength of

60

juvenile competition for vacant territories (Taylor et al. in prep). Red squirrels defend exclusive
territories year-round (Smith, 1968) and a territory is necessary for overwinter survival (Larsen
and Boutin, 1994), so juvenile squirrels born in the spring and summer must compete for territory
ownership in order to survive the winter and recruit into the population. Survival to one year of age
accounts for 60% of a female’s contribution to population growth (McAdam et al., 2007) and as a
result the opportunity for selection on this stage of life is very high (McAdam and Boutin, 2003b).
Variation in juvenile competition for territory vacancies is influenced by adult overwinter survival
and food availability (McAdam et al. in prep). Using the ratio of juveniles produced in a year to
the number that survived overwinter as an index of competition, Taylor et al. (in prep; Chapter
3) found that selection on dam activity and aggression fluctuated significantly in sign across years
depending on intensity of competition among juvenile squirrels for vacant territories. Female red
squirrels can anticipate autumn abundance of their main food source, seeds from white spruce
cones (Picea glauca), and produce more offspring in anticipation of higher offspring recruitment
due to higher resource availability (Boutin et al., 2006). Male red squirrels increase energy expenditure during the mating season in advance of autumns with high cone production (Lane et al.,
2009). Additionally, red squirrels adjust their behavior in response to changes in population density (Dantzer et al., 2012). The ability to anticipate autumn resource levels and assess current
population density indicates that red squirrels have the capacity to anticipate upcoming juvenile
competition for territory vacancies. Because male red squirrels do not provide parental care, any
female mate choice should be based on indirect (genetic) benefits provided by males. Good genes
should enhance offspring recruitment for the upcoming juvenile competitive environment.
Red squirrels perform elaborate and demanding mating chases where up to a dozen males vie
for opportunities to mate with each female on her single day of estrus (Smith, 1968; Lane et al.,
2008b). In this scramble mating system, the ability of males to locate females is a strong predictor
of male siring success (Lane et al., 2008a) and energy expenditure of males during the mating
season approaches that of females during lactation (Lane et al., 2009). Female red squirrels mate
with many males (mean 6.9; McFarlane et al. 2011; Lane et al. 2008b) during their day of estrous.

61

During mating chases a dominant male follows the estrous female, only leaving her to drive away
any other males that approach (Smith, 1968). How males establish dominance is unknown because
contact among males is rare (Smith, 1968). During mating chases there is considerable variation
in how fast and far the female travels (area covered ranges from 45 m2 to 56,020 m2 ), indicating
that females may be able to manipulate the characteristics of the mating chase to attract males or
to favor certain males over others. Lane et al. (2008b) hypothesized that, if multiple mating is
adaptive, then the benefits of multiple mating should outweigh its costs, but found no benefits of
multiple mating on fertilization assurance, infanticide avoidance, litter allelic diversity or parental
relatedness (Lane et al., 2008b). However (Lane et al., 2008b) also found no detectable costs of
multiple mating for females in the form of reduced female overwinter survival. Here we consider
the hypothesis that mating chases provide a mechanism that allows female mate choice for highquality mates, and test this hypothesis with predictions about context-specific indirect benefits for
offspring overwinter survival.
Here we make explicit predictions based on our prior selection analyses (Taylor et al. in prep.;
Chapter 3) about patterns of male red squirrel attendance at mating chases (male mate choice) and
the siring success of males that attend a female’s mating chase (female choice). Parental aggression was favored in high-competition environments, but was a disadvantage in low-competition
environments (Taylor et al. in prep; Chapter 3). As a result, we first predicted that, in highcompetition years, both males and females should prefer more aggressive mates because of the
benefits of aggressive alleles for offspring survival under those conditions. In low-competition
years less aggressive mates should be favored. Second, in high-competition years intermediate
values of paternal docility increased offspring survival whereas sires with very high or low docility
had low offspring survival (i.e. stabilizing selection), but no significant selection on docility was
found in low-competition environments (Taylor et al. in prep.; Chapter 3). Therefore, we predicted
that preference for docility would depend on the chooser’s docility. Specifically, a highly docile
individual should prefer a less docile mate so that the additive effect of the mate’s genes would
moderate offspring docility. Conversely, a less docile individual should prefer a more docile mate,

62

resulting is disassortative mate choice for docility.

4.2

Methods

A population of North American red squirrels has been monitored since 1989 in the southwest
Yukon (61◦ N, 138◦ W), and a detailed description of the population and general methods can be
found in (McAdam et al., 2007). Red squirrels are small (150 − 250 g), diurnal, semi-arboreal
rodents that inhabit much of forested North America (Steele, 1998). Individuals of both sexes
defend exclusive year-round territories (Smith, 1968). Each squirrel in each of two study areas
was uniquely marked with numbered ear-tags as nestlings or at first capture after emergence from
the natal nest, and followed from birth until death. To allow for observational identification, unique
combinations of colored wires and/or pipe-cleaners were threaded through each squirrel’s ear-tags.
Stakes placed every 30m in a grid allowed for spatial locations of squirrels to be recorded.

4.2.1

Siring success and parturition date

Nests of lactating females were entered and tissue samples were taken from pups immediately after
parturition and again at 25 days of age for paternity analysis. Paternity was assigned based on 16
microsatellite loci using CERVUS 3.0 (Kalinowski et al., 2007) with matches accepted at 95%
or greater probability and no more than one mismatch (detailed in Gunn et al. 2005; Lane et al.
2008a). We evaluated siring success both as a continuous and as a binary variable. Since we found
no substantial differences in our results, we present only the results from siring success as a binary
response (i.e. sired or did not sire offspring). Parturition was identified based on weight loss of
the mother recorded during trapping and parturition dates were calculated from trapping records or
from the weight of young during the first nest entry. For mating chases that were not observed we
calculated the mating chase date as 35 days prior to parturition (Steele 1998; S. Boutin, personal
communication).

63

4.2.2

Behavioral Traits

We measured behavioral traits of individuals from two study areas that have been monitored continuously since 1989. Following Boon et al. (2007); Taylor et al. (2012), we used two behavioral
trials to measure aggression and docility. Here we give a brief description of the behavioral methods, but full details can be found in Boon et al. (2007); Taylor et al. (2012). The aggression test was
a mirror-image stimulation (MIS) test in a portable arena that measured the aggressive response
to a mirror image (Svendsen and Armitage, 1973). In the same testing session (trial), prior to the
MIS test an open field (OF) test was used to measure an individual’s activity in a novel environment
(Walsh and Cummins, 1976; Martin and Réale, 2008), but also served as a habituation period for
the MIS test. The testing arena for the OF and MIS tests was a 60 × 80 × 50cm white corrugatedplastic box with a clear acrylic lid through which the behavior of the squirrel was recorded with a
digital video camera. Four blind holes were placed in the floor to provide the focal squirrel with
the opportunity to explore. A 45 × 30 cm mirror at one end of the arena was exposed during the
MIS portion of the session. Individual squirrels were tested within one hour of being trapped on
their territory. The docility test, conducted during routine handling events, measured docility as
the struggle rate of individuals confined in a mesh handling bag. To measure docility squirrels
were immediately transferred from the trap into a handling bag and placed onto dry ground and
the proportion of time the squirrel spent struggling was measured over 30 seconds. The exclusive
territoriality of red squirrel allowed for complete enumeration of the study population through targeted trapping and behavioral observations. Importantly the ability to target individuals allows us
to avoid sampling bias (Biro and Dingemanse, 2009; Boon et al., 2008). We performed 556 OF
and MIS trials on 183 female and 183 male red squirrels and 3122 struggle rate tests on 291 female
and 301 male squirrels over a four-year period (2005 and 2008 − 2010).
We quantified each squirrel’s MIS behavior by scoring the rate or proportion of time each
squirrel spent performing specific behaviors (e.g. attacks, time spent in the closest 1/3 of the arena
to the mirror) from the videotaped trials using JWatcher Video 1.0 (Blumstein and Daniel, 2007)
and the same ethogram as Boon et al. (2007); Taylor et al. (2012). We did not consider each of our

64

measurements from the behavioral trials to be unique behaviors and given the intercorrelation that
necessarily results from mutually exclusive scores, we did not attempt to interpret them as unique
behaviors. Instead our goal was to collect many measurements that we hoped would provide a
reliable overall assessment of the behavior of squirrels under these conditions that have previously
been found to have important ecological and evolutionary consequences (Boon et al., 2007; Taylor
et al., 2012). We used the principal component loadings from (Taylor et al., 2012) to calculate
behavioral scores for each trial. All further analyses used the scores calculated from the first
principal component loadings for each trial, which were interpreted as the aggressive response to
a mirror image and referred to as aggression.
We controlled for a number of methodological sources of variation in behavior prior to evaluating mate choice through the behavioral traits by extracting best linear unbiased predictors (BLUPs)
from linear mixed-effects models with individual as the random effect (Pinheiro and Bates, 2000).
To account for effects of habituation we included as covariates for the docility and aggression
scores fixed effects terms for lifetime and yearly trial number. For struggle rate we included lifetime and yearly handling event, which included handling events where no struggle test was performed, because handling for routine data collection is similar to the struggle rate trial. We also
included a quadratic term for trial numbers to account for a nonlinear response to repeated trials.
To control for effects of seasonality and study area we included day of year as a continuous fixed
effect and study area as a categorical fixed effect. Finally, we included observer as a fixed effect in
the docility models. Following (Wilson et al., 2009), we only attempted to control for methodological variation (measurement error) through the inclusion of fixed effects in our models, and did not
attempt to account for other biological sources of variation (e.g. age, sex, birth year, reproductive
status or mass) to avoid removing phenotypic variation that might be relevant to selection. Results
from these models are presented and discussed further by Taylor et al. (2012). The behavioral traits
analyzed here are, therefore, equivalent to those studied by Taylor et al. (2012).

65

4.2.3

Mating Chase Observations

In order to assess courtship behavior and copulation success we observed red squirrel mating in
2003, 2004, 2005 and 2008. See (Lane et al., 2008a) for additional methodology. A red squirrel
mating chase begins when the female emerges in estrus from her nest early in the day, and continues
until she returns to her nest at the end of the day. Male red squirrels actively search out estrous
females during the breeding season by visiting female territories and assessing olfactory, visual and
vocal cues for signs of estrus (Smith, 1968; Lane et al., 2008a). Upon locating a receptive female,
males compete for access to the female, which often leads to a congregation of 2 to 7 males vying
for access to the female at once. At any time one male is ‘dominant’ to all others and attempts
to monopolize access to the female by chasing off other males, yet male squirrels rarely contact
each other (R. Taylor, personal observation) and conspicuous wounding is not prevalent during the
mating season (Smith, 1968).
To identify mating chases we located each female twice daily before noon using radio-telemetry.
Mating chases were identified by the existence of male squirrels on a female’s territory and her tolerance of their presence. To quantify mating chase behavior we used radio-telemetry to follow the
female squirrel throughout the chase and employed a combination of scan sampling for attending males and all-occurrence sampling of mating behavior (Altmann, 1974; Martin and Bateson,
1993). We recorded all dominant and subordinate males, mounting, attempted mounting, chasing,
periods out of site or underground, vocalizations, and other noteworthy events as well as spatial
locations. Copulations were judged to have occurred when male squirrels successfully mounted a
female or when he accompanied the female underground for a minimum of 60s. Copulations of
tree squirrels generally last less than 60 seconds (Koprowski, 2007), which is adequate for fertilization in red squirrels (J. E. Lane & S. Boutin, unpublished data). This criterion has been used
previously in this (e.g. Lane et al. 2008b,a) and other (Waterman, 1998) systems.

66

Table 4.1: Sample sizes for each study area-year. Total number of individuals
living on the study area, and the number of individuals for which we had
behavioral data are given. Behavior was measured in 2005 and 2008.
Total
number of
individuals
Study Area
Kloo
Kloo
Kloo
Kloo
Sulphur
Sulphur
Sulphur
4.2.4

Year
2003
2004
2005*
2008*
2003
2004
2005*

Number of
individuals
with behavior

Proportion
with
behavior

Females

Males

Females

Males

Females

Males

23
10
8
21
18
12
12

34
27
22
47
28
19
16

4
3
6
20
10
9
12

8
10
16
46
5
7
12

0.17
0.30
0.75
0.95
0.56
0.75
1.00

0.24
0.37
0.73
0.98
0.18
0.37
0.75

Components of siring success

We evaluated four potential components of siring success. Each of these components is also a
mechanism by which mate choice might occur. Attendance was assessed as whether a male squirrel
was present during a female’s mating chase. We assessed both copulation success (binary: yes or
no) and copulation number (the number of discrete copulations). Copulation order was the order
in which successive males copulated with a female. Multiple successive copulations by the same
male were counted as a single copulation event when calculating the copulation order. When a male
copulated with a female in more than one session (i.e. copulation events separated by the female’s
copulation with another male) his copulation order was determined by his earliest copulation.

4.2.5

Mate choice

In order to assess mate choice we generated a dataset consisting of all parental combinations for
each grid-year consisting of individuals for which we had behavioral data (See Table 4.1 for sample
sizes and Appendix Table C.1 for data structure).
For example the data subset for the study area Kloo in 2010 consisted of 840 rows, one row

67

for each possible combination of male (40 individuals) and female (21 individuals). This data
structure allowed us to test, using generalized linear mixed models (GLMMs), for effects of male
and female behavior as well as interactions between male and female behavior on siring success
and components of siring success. In all models we included the distance between the male and
female territories as a fixed effect. To account for pseudoreplication (Hurlbert, 1984), we fitted
the models with mating chase (analogous to female identity) and male identity as random effects.
Models for male attendance, copulation, and binary siring success were fitted assuming a binomial
error distribution (logit link function). Models for copulation order, number of copulations and
siring success, measured as the number of offspring sired, were fitted assuming a Poisson error
distribution (log link function). All traits were standardized to a mean of zero and unit variance
prior to analysis. GLMMs were reduced using backward reduction. All analysis were performed
using R 2.15.0 (R Development Core Team, 2012). Generalized linear mixed models were fitted
using the lme4 package (Bates et al., 2012),
We tested for mate choice on siring success and the four components of siring success: mating
chase attendance, copulation success, copulation order and siring success (Lane et al., 2008a).
To test our prediction that aggression would be favored in high-competition years we included
aggression, competition and an interaction between competition and aggression in the models. To
test our prediction that individuals would prefer mates with opposite docility in high-competition
years we included an interaction between male and female docility and a three-way interaction
with male and female docility and the degree of juvenile competition for vacant territories in that
year. A positive interaction between male and female docility would indicate assortative mating,
whereas a negative interaction would indicate disassortative mating. We included the three-way
interaction between male docility, female docility and competition, because we expected to see
disassortative mating (male docility × female docility interaction) only in high competition years
(i.e. three-way interaction: male docility × female docility × competition).
For the analysis of attendance, copulation success, copulation order and copulation number we
subsetted the dataset to only include females whose mating chases we followed. For the models

68

of copulation success we subsetted the data to only include males that attended a chase, which
modeled copulation success given attendance at the chase. For the models of copulation order and
copulation number we subsetted the data to only include males who successfully copulated, which
modeled the effect of copulation order and number after controlling for copulation success. For
the model of attendance we included the Julian date of the mating chase to control for seasonal
variation in the probability of a male attending a mating chase (Lane et al., 2008a). In this model
only, we included female aggression because attendance represents male mate-choice. In all other
models we included male aggression because we were modeling female choice. Models for male
and female choice included both male and female docility because we expected an interaction
between the chooser’s docility and their mate preference.

4.3

Results

4.3.1

Distance

Males that lived farther from females were less likely to attend her mating chase (b = −2.03 ±
0.14, Z = −14.51, P < 0.001; Table 4.2).

69

Table 4.2: Reduced models of male attendance, siring success and copulation success of males that attended a mating chase,
and siring success of males that copulated. The full GLMM models are described in the methods. The least significant terms
were removed, starting with the highest order. Male identity and mating chase identity were included as random effects. A
binary error distribution was assumed.
Random Variances
Response

Term

Estimate

Male attendance

Intercept
Julian day of chase
Distance
Female docility
Male docility
Competition
Competition×Female docility
Competition×Male docility
Male Docility×Female docility
Competition×Male docility×Female docility

Siring success
of attending
males

Male id.

Chase id.

−1.96 ± 0.44 −4.48 < 0.001
0.19 ± 0.08
2.23
0.026
−2.03 ± 0.14 −14.51 < 0.001
−0.12 ± 0.34 −0.34
0.732
−0.40 ± 0.44 −0.90
0.37
−0.10 ± 0.05 −1.83
0.067
0.03 ± 0.04
0.65
0.514
0.07 ± 0.06
1.18
0.239
−0.90 ± 0.37 −2.45
0.014
0.12 ± 0.04
2.70
0.007

0.952

0

Intercept
Number of sires
Distance

−2.51 ± 0.17 −15.04 < 0.001
0.72 ± 0.18
4.07 < 0.001
−0.45 ± 0.14 −3.09
0.002

0.885

0

Copulation
success of
attending males

Intercept
Number of copulatory partners
Distance

1.51 ± 0.32
−0.29 ± 0.09
−0.10 ± 0.05

4.67 < 0.001
−3.35 < 0.001
−2.08
0.038

0

0.764

Siring success
of copulating
males

Intercept
Number of sires
Male aggression

−1.85 ± 0.22
−0.33 ± 0.21
−0.88 ± 0.31

−8.52 < 0.001
−1.60
0.111
−2.81
0.005

0.467

0

70

Z

P

Table 4.3: Reduced models of male copulation order , siring success after controlling for copulation
order, copulation number and siring success after controlling for copulation number. The full GLMM
models are described in the methods. The least significant terms were removed, starting with the highest
order. Male identity and mating chase identity were included as random effects. A binary error
distribution was assumed for the siring success models, and a Poisson error distribution was assumed for
the copulation order and copulation number models.
Random Variances
Response

Term

Copulation order

Intercept
Docility

Siring success
controlled for
copulation order

Intercept
Copulation order
Number of sires
Male aggression

Copulation number

Intercept
Number of partners

Siring success
controlled for
copulation number

Intercept
Number of sires
Male aggression

Estimate

Male id.

Chase id.

1.31 ± 0.06 22.34 < 0.001
−0.10 ± 0.05 −1.98
0.048

0.051

0.073

−2.02 ± 0.24 −8.39 < 0.001
−0.84 ± 0.24 −3.48 < 0.001
0.59 ± 0.23
2.59
0.009
−1.05 ± 0.34 −3.11
0.002

0.165

0

5.23 < 0.001
2.34
0.019

0.051

0.147

−1.96 ± 0.24 −8.24 < 0.001
0.61 ± 0.23
2.70
0.007
−1.03 ± 0.33 −3.10
0.002

0.388

0

0.83 ± 0.16
0.05 ± 0.02

71

Z

P

The effect of distance persisted among males that attended a mating chase; males that lived
farther from females had lower copulation success (b = −0.1 ± 0.05, Z = −2.08, P < 0.038)
and lower siring success given attendance (b = −0.40 ± 0.15, Z = −2.79, P = 0.005). After
controlling for copulation success there was no longer any effect of distance on siring success
(b = −0.25 ± 0.25, Z = −0.98, P = 0.3). An interaction effect between distance and juvenile
competition on attendance was marginally significant, so was removed during the backwards model
selection procedure. To explore this effect more directly, we fitted a separate GLMM of male
attendance with only competition, distance and the interaction between juvenile competition and
distance as predictors, and found a significant negative interaction between distance and juvenile
competition such that the effect of distance was stronger in years with high juvenile competition
(Figure 4.2; Intercept = −1.97 ± 0.27, Z = −7.39, P < 0.001; distance = −1.18 ± 0.20, P <
0.001,t = −6.01; competition = −0.14 ± 0.04, P < 0.001,t = −3.53; distance × competition
= −0.13 ± 0.03,t = −3.91, P < 0.001; Random effect variance: Male identity = 1.69 ± 1.30,
mating chase identity = 0.10 ± 0.32).

4.3.2

Male Mate Choice

We found a significant positive three-way interaction between juvenile competition, male docility
and female docility (b = 0.12 ± 0.04, Z = 2.71, P = 0.007; Table 4.2; Figure 4.1) indicating that
males attended mating chases disassortatively based on docility and that the disassortative attendance was strongest in low juvenile competition years. After removing the three-way interaction
term, we no longer found a significant negative interaction between male and female docility on
attendance. We also found that males were more likely to attend mating chases that occurred later
in the season (b = 0.19 ± 0.08, Z = 2.26, P = 0.03).

4.3.3

Female Mate Choice

We found no effect of male aggression or any interaction between male docility and female docility
on the siring success (continuous or binary) of males that attended a mating chase (stats). After
72

1

Probability of Attendance

0.8

Male
Low
High
Low
High

0.6

Female
High
Low
Low
High

0.4

0.2

0
Low Competition

High Competition

Figure 4.1: Three-way interaction plot of the interaction between male docility, female docility
and competition on male attendance.

sub-setting the data to only include males that copulated we found that more aggressive males had
lower siring success (b = −0.87±0.31, Z = −2.83, P = 0.005), but this effect did not interact with
the degree of juvenile competition. This negative effect of aggression persisted after controlling for
copulation order and after controlling for number of copulations. Docile males copulated earlier
in order than less docile males (b = −0.13 ± 0.05, Z = −2.46, P = 0.01; Table 4.2) and we found
a negative effect of copulation order on siring success (b = −0.84 ± 0.24, Z = −3.48, P < 0.001)
meaning that males that mated with a female earlier in her day of estrus were more likely to sire
some of her offspring.

4.4

Discussion

Despite strong associations between sire personality and offspring survival (Taylor et al. in prep;
Chapter 3), we did not find support for our hypothesis that mating chases are mechanisms by which
73

0.8
0.4

0.6

High Juvenile Competition
Low Juvenile Competition

0.2
0.0

Probability of attendance

1.0

qq q
q qqqq qqqq q q
q qq
qq q q qq
qqqqqq qqqq qq qq qq q qqq
q q q qq q q
q qqqq q qq q q
q
q q q q q qq q
q
qqqqqqqqqqqqqqqqq q qqq q
q
q qqqqq qqq q qq q
qqqqqqqqqqqqqqqqqqqqq q qqq q q
q qqqqqq qq qqq q q
q qqqqqqqqqqqqq q qq q q
qq q
qq q qq
q
qq
q qq q q q
qqqqqqqq q q q
qq qqqqqqqqqqqq q
qqqqqqqqqqqqq qqqqq qq qq q
qqqqqqqqqqq
q
q qq q qq
qq q q qq
q qqq q qqq
qqqq q qq qqq q q q q q q q
q q
q q q q
qqqqqq qq qqq qqqq qqq q q q q
q q qqq qq qq q
qq qqq
q
qq q
q qq qqqqqqq qq qq q
qq qqq q q q q qq q
q q
q
qqqqqqqqqqqqq q q qq q q
qq q
q
q
q q qq qq q q q q q
q q q qq q q q
qqq qqq q qqq q qqq q q
q qqqq qqqq qqqq qqq q q q q
qqq qqq qq qq qq qq q
q
q
q q
q
q
qq q q qq qqq q
q
q
qq q q
qqqqqqqq qqq qqqq q q qqq q q
qqqqqqqqq qq qq q q
qq qq qq qq q
q
qq q qq qqq qq q q

q
q
q q qqq qq qqqqqqqqqqqqq qqqq qq qq
q
q
q q
qq q
qqqq qqq qqqqqqqqqqqqqqqqqq qqqq qqq q q
q q q qq
q
q q q qqqq qq qqqqq
qq qqqqq qqqqqqqqqqqqqqqqqqqqqqqqqq q qq
q q qqqqqqqqqqqq qq q
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq qqqq q q q q
q
q qq qqqqqqqqqqqqqqqq q qqqq q q q
qq qqqqq qqqqqqqqqqqqqqqq qq q q q
qq
qq q qq qqqqqqq qqqqqq q q
q qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q q qq q
q q q qq qq q q q
q q q
qqq qqqqqqqq qq q q q q q q q
qq qqqqqqqqqqqqqqqqqqqqqq qqqq q q
qqqqqqqqqqqqqqqqqqqqqqqqqqq q q qq qq
q q
q qq q q q q qqq q q q q q
q
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q q q q
q q q q qq q q
qq q qqqqqqqqqqqqqq qqqq qqqqq q q q
q
qqq q q q qqq qqq qq q qqqq q
qqqqqqqqqqqqqqqqqqqqqqqqqqq qqq qq
qqqqqqqqqq qqqqqq qq qqqqq q q
qq qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq qqqq q
q qq
q qqq q q q
qqqqqqqqqqqqqqqqqqqqq qqqqq q q q q
qqq qqqqqqqqqqqqqq q qq q q q
q
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q qq
qq qqqqq q qqqq q q qq
qqqqqqqq qqqqqqqqqqq qq q q qq q q
q qq q q q
q
q
qq qq qqqqqqqqqqqq q qq qqq q q q
q
qq
q qqqqqqq qqqqqqqq qqqq q qqq qq q
q q
qqqqq q qq qq
qqqqqqq qqqqqqqqqqqqqqqqqqqq q q q q q q qqq
qqqqq qqqqqqqqqqqqqq qqqqqqqqqqqq qqq q qqq
qqqqqqqq qq q q q qq q q q q q
q qqqqqq qq qq qq q q
qq q q
q
q qqq q qqqqqqqqqqqq q q q qq qq q
q q qq q q
q q
qqqqq q q q qqqqqqqqqqq q qqqqqqqq q
q qq qqq qqqqqqqqq qq q qq qqq q q
q
q qq q qq qq qq
q
qqqqq qqqqqqqqqqqqqqqq qqqqqqqqqqq qq qqq
q q qq qq qqqqqqq q qq q qq q q q
q
qqq qq qqqq qqqqqq qq q qqqqq q qqq q qqq
q qqqq qqqqqqqqqqqqqqqqqqqqqqqqqqqqq qq qqq qq q q
q qqqq qq q qq
q qqqqq qqqq qqqqqqqqqqq q qqq q q q
q
q qqqqqqqqqq qqqqqqqqqqqqqqqqqqqqqq q q qqq qqq
q
q qqqq q qqqqqqq q qqqqqqq qqq q
q qq q qqqqqq q qqq q q
q q q qq qqq q qq q q
qqqqqqqqq qqqqqqqqqqqqqqqqqqqqqq q qq
qqqq qqqqq qqq q qqq q q q q
qq q q
q qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q qq q
q q
qq qq q
q qqqq qqqqqqqqqqqqqqqqq q q
qqqqqqqqqqqqqqqqqqqqq qqq q q q
q q qq q qq q q q q qq qq qqqqq qqq qqq qq q
q q qq q q
qq qqq qqq qqqqq q qqqqqqq qqqqqqqq qqqq qqqq q q
q q qq qq q q q q qq q qq q qqq qqq
q
q q qq
q q q
q
q
q

0

200

400

600

800

Distance (meters)
Figure 4.2: The effect distance on male attendance to a mating chase. The dashed line indicates
predicted probability of attendance at −1, and the solid line at +1, standard deviation of juvenile
competition. In high juvenile-competition years the effect of distance is stronger.

female squirrels can acquire good genes for offspring overwinter survival. In fact we found no evidence for female mate choice, but our analysis was narrowly targeted to specific predictions based
on selection results from Taylor et al. (in prep; Chapter 3) rather than a survey for any evidence
of female mate choice. There may be traits other than aggression or docility, behavioral or otherwise (e.g. energetic or life-history), that females are choosing. It is also possible that male-male
competition may swamp our ability to detect female mate choice. We did, however find evidence
for male mate choice based on docility via mating chase attendance. Taylor et al. (in prep; Chapter 3) found that males with intermediate levels of docility sired offspring with higher overwinter
survival. Thus we inferred, because males do not provide paternal care that these offspring are
74

favored because they inherited good genes from their sires. Following this line of reasoning we
predicted disassortative mating for docility whereby individuals would prefer a mate with opposite
docility in order to produce offspring with moderate docility. We found a negative interaction between male docility and female docility on attendance, showing that males attended mating chases
disassortatively based on docility. Because stabilizing selection for docility was only present in
high-competition environments we additionally predicted that disassortative mate choice for docility should be stronger in high-competition years. However, we found a positive three-way interaction between male docility, female docility and competition. This interaction can be interpreted
as disassortative mating based on docility that is mediated by the third variable, competition, and
indicates that the two-way interaction decreased in strength with competition, the opposite of our
prediction. Why should male red squirrels be picky and not take advantage of every mating opportunity? There are costs in energy, time, territorial defense and predation to attending mating
chases and to identifying when a female is in estrus. Male red squirrels must travel to each female’s
territory in order to ascertain her reproductive status in advance of her day of estrus. These costs
are clearly represented by the effect of distance on mating chase attendance, but it is surprising
that males that live close to a female do no always attend her chase. Approximately one third of
mating chases were concurrent with at least one other mating chase (i.e. on the same day) and,
because sperm precedence is important, the benefits of attending a second mating chase in one day
may be very small. Male squirrels were also more likely to attend a particular mating chase later
in the season when the number of females that had not had their mating chase was lower. There
may be a point at which the cost of keeping track of additional females outweighs any benefits of
attending an additional mating chase at which point males could prioritize which female’s chases
have the highest benefit. Males attend more mating chases in low-competition years, which may
be why we find stronger mate choice in those years. Males also increase their energetic expenditure during the mating season (Lane et al., 2009) in anticipation of elevated autumn resources
and associated high juvenile recruitment. High juvenile competition years are the result of high
spring (adult) density and low autumn resources (McAdam et al. in prep). There may be higher

75

costs to assessing the reproductive status of distant females when population densities are high
due to potentially increased intruder pressure and fewer gaps between territories in which squirrels could travel unmolested. In fact we do find that in high competition years there is a stronger
negative effect of distance on male attendance (Figure 4.2). Our finding that disassortative mating
is stronger in low juvenile-competition years may be due not to a larger benefit of mate choice,
but to a greater opportunity to be selective. We did not find any evidence of mate-choice based on
aggression. We had predicted that aggressive mates would be preferred in high-competition environments because Taylor et al. (in prep; Chapter 3) found positive selection for parental aggression
in high-competition years. We did find significant effects of aggression on siring success after controlling separately for copulation success, copulation order and copulation number, but there was
no significant interaction between aggression and competition, and the effect of aggression on siring success was negative. A negative relationship between aggression and sperm competition has
been reported in a number of fish species (Alonzo and Warner, 2000; Le Comber et al., 2003;
Spence and Smith, 2005; Rezucha and Smith, 2012), and Koyama and Kamimura (2000) found
a negative relationship between aggression and sperm motility in mice. Therefor our results may
represent a trade-off between aggression and sperm quality. Conclusion We tested the hypotheses
that mating chases and male attendance at mating chases are mechanisms by which red squirrels
can acquire context dependent good and complementary genes for their offspring’s recruitment.
We did not find support for the mating chase hypothesis, but we did find support for mate choice
through attendance by male red squirrels for complementary genes. Male red squirrels attended
mating chases disassortatively by docility, presumably because stabilizing selection for recruitment
favored juveniles with moderate docility. However, disassortative attendance was stronger in low
juvenile-competition years, which did not match our prediction. We also surprisingly identified
post-copulatory selection for less aggressive males, which may be due to a trade-off between aggression and sperm quality. The role female red squirrels play in mating chases remains a mystery
that we hope future studies will uncover.

76

Chapter 5
CONCLUSIONS AND FUTURE DIRECTIONS

5.1

Conclusions

In this dissertation I investigated the proximate sources of variation in red squirrel behavior by
estimating sources of genetic, maternal and environmental variation, and the ultimate effect of behavior on reproductive success. In Chapter 2, I used a Bayesian animal model approach to estimate
genetic parameters for aggression, activity and docility and found support for low heritabilities
(VA /VP = 0.08 − 0.12), and cohort effects (Vcohort /VP = 0.07 − 0.09), as well as low to moderate maternal effects (VM /VP = 0.07 − 0.15) and permanent environmental effects (VPE /VP =
0.08 − 0.16). I also found evidence of a substantial positive genetic correlation (0.68) and maternal effects correlation (0.58) between activity and aggression, providing evidence of genetically
based behavioral correlations in red squirrels. These results provide evidence for the presence of
heritable variation in red squirrel behavior, but also emphasize the role of other sources of variation,
including maternal effects, in shaping patterns of variation and covariation in behavioral traits. I
then tested the hypothesis that ecological changes through time lead to fluctuating selection, which
maintains variation in behavioral traits (Chapter 3). Linear selection gradients on red squirrel dam
aggression and activity significantly fluctuated in sign across years depending on the level of competition among juveniles. Selection on aggression and activity also differed among components of
fitness, between the sexes and included nonlinear components between and within traits that also
changed through time. These results suggest that repeatable and heritable individual differences in
red squirrel behavior could be maintained by complex fluctuations in natural and sexual selection.
Finally, I tested the hypotheses that mating chases provide the opportunity for both female and male
red squirrels to select for context-dependent good genes and complementary genes for their offspring’s recruitment (Chapter 4). Specifically, I predicted that aggressive mates would be preferred

77

in high-juvenile competition years, but disfavored in low competition years, and that mate choice
for docility would be disassortative with low-docility squirrels preferring high-docility mates and
vice-versa. I did not find support for adaptive context-dependent mate choice by females, but I
did find support for male mate choice for complementary genes that was mediated through which
female mating chase males attended. Male red squirrels attended mating chases disassortatively
by docility, which would enhance fitness due to stabilizing selection on docility during juvenile
recruitment. I also found evidence for post-copulatory selection for less aggressive males. These
results highlight that even in systems with very high operational sex ratios, male choice is a factor
that needs to be considered. In this dissertation I have shown that behavioral traits in red squirrels
are important with strong evolutionary implications. Red squirrel behavior is heritable, subject to
maternal effects and under strong selection that varies with environmental fluctuation and across
components of reproductive success. This research has provided strong evidence that fluctuating
selection could maintain variation in behavior (Chapter 2). Incidentally, while also showing that
mate choice may reduce variation (Chapter 3). These findings provide a foundation for future work
to elucidate mechanisms and continue exploring how and why behavioral variation is maintained.
There are major areas of research regarding red squirrel personality that have yet to be investigated
(some of which are outlined below), but these results also expose targeted questions that can be
asked (as in Chapter 3).

5.2
5.2.1

Future Directions
Are there consistent individual differences in behavioral plasticity?

There is increasing evidence that individuals differ in their behavioral responsiveness to environmental variation (reviewed in Dingemanse et al. 2010) and that behavioral plasticity is linked
with fitness (e.g. Betini and Norris 2011). Red squirrels have evolved in a variable environment
and plasticity in response to environmental variation has been documented for life-history traits
(McAdam and Boutin 2003b; Boutin et al. 2006; Dantzer et al. in prep). In Chapter 2 I found

78

evidence of behavioral plasticity in aggression and docility in the form of habituation as well as
in response to seasonal change (i.e. day of year). Future studies should test whether behavioral
plasticity occurs in response to variation in density and resources. Additionally, using individual behavioral reaction norms (Dingemanse et al., 2010) future studies should investigate whether
individuals vary in their plasticity and if that variation is under selection.

5.2.2

Investigation of links between behavior and life-history, energetics and hormone levels.

There has been growing interest in how behavior co-evolves with life-history, energetics and hormones. A number of hypotheses positing explanations for consistent individual differences rely
on life-history variation as the source of behavioral variation (Wolf et al., 2007; Biro and Stamps,
2008; Dingemanse and Wolf, 2010). Similarly, links between energetics and personality have also
been proposed (Careau et al., 2008). Recently Réale et al. (2010b) integrated these hypotheses into
the pace-of-life syndrome (POLS) framework (Ricklefs and Wikelski, 2002). Whether mediated
by proximate or ultimate mechanisms these hypotheses posit that behavioral variation may arise
through selection for optimal combinations of behavior and life-history, energetic or hormonal
traits. Studies of this fundamental component of behavioral theory are sorely needed as it has
rarely been tested in the wild (but see Réale et al. 2009). The Kluane red squirrel system provides
an excellent opportunity to study these questions due to the multi-disciplinary approach already
underway.

79

APPENDICES

80

Appendix A

Table A.1: Between-observer reliability, means and variances of specific behaviors recorded
during open field and mirror-image stimulation tests. Correlations were calculated for each
behavior between two observers’ scores for the same trial. Behaviors with a reliability greater
than 0.7 were used in further analysis. Mirror image stimulation behaviors are indicated above,
open field behaviors are below.
Behavior

Reliability (r2 )

Crouch Rate
0.50
Front
0.86
Attack Rate
0.97
Grunt Rate
0.00 *
Stretch
0.32
Back
0.99
Attack Latency
1.00
Approach Latency 0.94
Hole Rate
Jump Rate
Chew
Still
Hang
Groom
Rear
Scan
Walk
Sniff
∗

Kept
Y
Y

Y
Y
Y

0.90
1.00
0.80
0.85
0.99
0.94
0.13
0.67
0.71
0.39

Y
Y
Y
Y
Y
Y

Y

Grunt rate was not consistently recorded

81

Mean

Variance

0.45
0.77
26.13 1023.99
0.66
2.15
0.93
15.99
0.81
3.12
42.59 1605.63
217.76 13472.92
138.71 16610.15
1.90
4.26
5.68
31.75
14.10
3.70
8.94
12.70
16.79
4.93

2.41
19.82
109.33
705.5
251.44
17.17
99.28
217.98
133.26
18.00

Table A.2: Summary statistics for the entire multi-generational red squirrel pedigree, pedigree of
individuals informative for docility and pedigree of individuals informative for activity and
aggression.
Entire Pedigree

OF & MIS Pedigree

819
444
311
61
486
425
255
194
258
154
119
88
8
361
2.166
1.968
0.003

Records
7086
Maternities
5620
Paternities
1549
Full sibs
1056
Maternal sibs
27501
Maternal half sibs
26445
Paternal sibs
5051
Paternal half sibs
3995
Maternal grandmothers
3543
Maternal grandfathers
1080
Paternal grandmothers
458
Paternal grandfathers
342
Maximum pedigree depth
8
Founders
1442
Mean maternal sibsip size
6.862
Mean paternal sibsip size
3.688
Mean pairwise relatedness
0.001

Docility Pedigree

451
247
146
21
194
173
90
69
135
63
53
28
7
193
1.871
1.780
0.004

Table A.3: Principle component loadings for behaviors from an open field arena test in North
American red squirrels. Behaviors were measured as percentage of time unless otherwise noted.
Open Field Behavior

PC1

Walk
Jump Rate
Hole Rate
No. Pellets
Hang
Chew
Groom
Still

0.490 –0.259 0.119 0.063
0.436 –0.057 0.106 0.298
0.307 –0.495 0.131 0.080
0.288 –0.089 0.023 –0.678
0.253
0.634 –0.227 0.333
0.240
0.381 –0.107 –0.563
–0.061 –0.344 –0.914 –0.011
–0.517 –0.109 0.241 –0.118

Std. Deviation
Prop. Variance

1.706
0.364

PC2

1.130
0.160

PC3

0.993
0.123

82

PC4

0.964
0.116

PC5

PC6

–0.045
–0.138
0.452
–0.601
–0.144
0.626
–0.011
–0.004

–0.199 –0.576
–0.587 0.580
0.561 0.339
0.219 0.199
0.400 0.175
–0.249 0.077
–0.114 0.089
–0.134 0.365

0.875
0.096

0.774
0.075

PC7

0.619
0.048

PC8
–0.549
0.089
–0.051
–0.012
–0.401
–0.094
–0.144
–0.705
0.389
0.019

Table A.4: Principle component loadings for behaviors from an mirror-image stimulation test in
North American red squirrels. Behaviors were measured as percentage of time unless otherwise
noted. Latencies were log transformed prior to principle component analysis.
MIS Behavior

PC1

Front
Attack Rate
Back
Attack Latency
Approach Latency

0.491 –0.268
0.373 0.622
–0.407 0.612
–0.469 –0.400
–0.484 0.081

Std. Deviation
Prop. Variance

1.668
0.557

PC2

0.948
0.180

83

PC3

PC4

0.161
0.606
–0.369
0.387
0.567

0.813 –0.006
–0.143 –0.292
0.519 –0.235
0.070 –0.683
0.211 0.627

0.789
0.124

0.611
0.075

PC5

0.567
0.064

Table A.5: Candidate univariate models of sources of variation in behavior in North American red
squirrels. Fixed effects (FE) are trial number, quadratics for trial number, study area, day of year
and observer. Random effects are individual identity (VPE ), additive genetic (VA ), maternal effects
(VM ) and cohort effects (VC ). Deviance Information Criterion (DIC) estimates are given for each
model. Models within 2 DIC of the best model were considered to have equal support (bold).
Open field (OF)

DIC

OF∼FE+VPE +VM +VC
OF∼FE+VPE +VA +VM +VC
OF∼FE+VPE +VM
OF∼FE+VPE +VA +VM
OF∼FE+VPE +VC
OF∼FE+VPE +VA +VC
OF∼FE+VPE
OF∼FE+VPE +VA
OF∼FE+VC
OF∼FE
OF∼ 1

1916.2
0
1916.4
0.2
1919.7
3.5
1920.6
4.4
1931.9 15.7
1932.9 16.7
1935.7 19.5
1936.3 20.1
2028.9 112.7
2037.9 121.7
2110.7 194.5

Mirror image (MIS)

DIC

MIS∼FE+VPE +VA +VC
MIS∼FE+VPE +VA
MIS∼FE+VPE +VC
MIS∼FE+VPE
MIS∼FE+VPE +VA +VM +VC
MIS∼FE+VPE +VM +VA
MIS∼FE+VPE +VM
MIS∼FE+VPE +VM +VC
MIS∼FE+VC
MIS∼FE
MIS∼FE

2018.6
2019.7
2019.7
2019.9
2020.4
2021.3
2022.8
2023.0
2079.3
2082.1
2084.5

84

∆DIC

∆DIC
0
1.1
1.1
1.3
1.8
2.8
4.2
4.5
60.7
63.5
65.9

Table A.6: Candidate univariate models of sources of variation in behavior in North American red
squirrels. Fixed effects (FE) are trial number, quadratics for trial number, study area, day of year
and observer. Random effects are individual identity (VPE ), additive genetic (VA ), maternal effects
(VM ) and cohort effects (VC ). Deviance Information Criterion (DIC) estimates are given for each
model. Models within 2 DIC of the best model were considered to have equal support (bold).
Struggle rate (SR)

DIC

SR∼FE+VPE +VA +VM +VC
SR∼FE+VPE +VA +VM
SR∼FE+VPE +VA +VC
SR∼FE+VPE +VA
SR∼FE+VPE +VM +VC
SR∼FE+VPE +VM
SR∼FE+VPE +VC
SR∼FE+VPE
SR∼FE+VC
SR∼FE
SR∼FE

25140.6
0
25141.4
0.8
25142.2
1.6
25142.8
2.2
25143.2
2.7
25145.2
4.6
25149.2
8.7
25151.1
10.5
26452.0 1311.4
26515.0 1374.4
26856.6 1716.0

∆DIC

Table A.7: Candidate trivariate animal models of sources of (co)variation of the behavioral traits
aggression, activity and docility. Fixed effects (FE) are trial number, quadratics for trial number,
study area, day of year and observer. Random effects are individual identity (VPE ), additive
genetic (VA ), maternal effects (VM ) and cohort effects (VC ). Deviance Information Criterion (DIC)
estimates are given for each model. Models within 2 DIC of the best model were considered to
have equal support.
Model

DIC

∆DIC

(OF, MIS, SR)∼FE+VPE +VA +VM +VC
(OF, MIS, SR)∼FE+VPE +VM +VC
(OF, MIS, SR)∼FE+VPE +VA +VM
(OF, MIS, SR)∼FE+VPE +VA +VC
(OF, MIS, SR)∼FE+VPE +VM
(OF, MIS, SR)∼FE+VPE +VC
(OF, MIS, SR)∼FE+VPE +VA
(OF, MIS, SR)∼FE+VPE

29220.86
29228.14
29234.71
29235.83
29240.49
29250.52
29254.60
29267.78

0
7.28
13.85
14.97
19.63
29.66
33.74
46.92

85

Appendix B

86

Table B.1: The effect of competition on linear selection for dam and sire behavioral traits through annual reproductive success.
Coefficients are from a GLMM with ARS as the response and each study-area-year combination as a random effect and a Poisson error
distribution.
Dams
Term
Intercept
Competition
Aggression
Activity
Docility
Competition×Aggression
Competition×Activity
Competition×Docility

Estimate

Z

−1.33 ± 0.23
−4.06 ± 0.70
0.75 ± 0.23
−0.30 ± 0.20
−0.30 ± 0.16
2.55 ± 0.69
−1.30 ± 0.60
−1.02 ± 0.44

Sires
P

−5.66
−5.83
3.28
−1.51
−1.88
3.69
−2.17
−2.30

< 0.001
< 0.001
0.001
0.13
0.06
< 0.001
0.03
0.02

Random effect grid-year variances: Dam = 0.035, Sire = 0.129

87

Estimate

Z

−2.05 ± 0.17 −11.75
−0.63 ± 0.16 −3.88
0.02 ± 0.20
0.08
−0.22 ± 0.18 −1.22
−0.02 ± 0.15 −0.15
−0.10 ± 0.17 −0.59
0.07 ± 0.15
0.45
0.01 ± 0.14
0.05

P
< 0.001
< 0.001
0.94
0.22
0.88
0.56
0.65
0.96

Table B.2: Vector of standardized directional selection gradients(β ) for fecundity or siring success, and the matrix of standardized
quadratic and correlational selection gradients (γ). Linear and quadratic selection gradients were estimated in separate regressions.
Quadratic selection coefficients were doubled to give quadratic selection gradients (Stinchcombe et al. 2008)
High Competition
Dams
β
Aggression
Activity
Docility

Aggression

−0.08 ± 0.04†
−0.01 ± 0.04
−0.02 ± 0.04

0.20 ± 0.08∗
−0.03 ± 0.06
−0.09 ± 0.07

Sires
Activity
0.03 ± 0.09
0.14 ± 0.07†

Docility

0.04 ± 0.07

β

Aggression

Activity

Docility

−0.07 ± 0.18
0.05 ± 0.14
0.26 ± 0.09∗

0.67 ± 0.41
−0.33 ± 0.20
−0.20 ± 0.17

−0.10 ± 0.21
0.11 ± 0.13

0.07 ± 0.15

Activity

Docility

Low Competition
Dams
β
Aggression
Activity
Docility

Aggression

−0.02 ± 0.05
0.09 ± 0.05†
0.10 ± 0.04∗

0.04 ± 0.11
−0.01 ± 0.08
−0.09 ± 0.07

Sires
Activity

−0.05 ± 0.11
0.06 ± 0.07

Docility

β

Aggression

−0.01 ± 0.06

0.15 ± 0.17
−0.48 ± 0.16∗
0.01 ± 0.13

−0.63 ± 0.21∗
−0.21 ± 0.20
−0.04 ± 0.18

* P < 0.05, † P < 0.1

88

0.33 ± 0.30
0.00 ± 0.20

0.15 ± 0.20

Table B.3: Vector of standardized directional selection gradients(β ) for offspring overwinter survival, and the matrix of standardized
quadratic and correlational selection gradients (γ). Linear and quadratic selection gradients were estimated in separate regressions.
High Competition
Dams
β
Aggression
Activity
Docility

0.53 ± 0.14∗
−0.28 ± 0.19
−0.29 ± 0.14∗

Aggression
0.02 ± 0.37
0.28 ± 0.28
0.00 ± 0.20

Sires
Activity

0.06 ± 0.46
−0.09 ± 0.26

Docility

β

Aggression

Activity

Docility

−0.22 ± 0.28

0.20 ± 0.22
−0.14 ± 0.19
−0.07 ± 0.16

0.07 ± 0.68
−0.18 ± 0.26
−0.06 ± 0.23

−0.77 ± 0.34∗
0.52 ± 0.26∗

−0.39 ± 0.25

Low Competition
Dams
β
Aggression
Activity
Docility

Aggression

−0.31 ± 0.12∗
0.24 ± 0.12†
0.00 ± 0.09

0.12 ± 0.29
−0.13 ± 0.22
−0.07 ± 0.15

Sires
Activity
0.27 ± 0.28
0.07 ± 0.16

Docility

0.09 ± 0.16

* P < 0.05, † P < 0.1

89

β
−0.20 ± 0.21
0.12 ± 0.19
−0.07 ± 0.12

Aggression
0.38 ± 0.46
0.06 ± 0.33
0.42 ± 0.35

Activity
0.27 ± 0.53
−0.03 ± 0.26

Docility

0.07 ± 0.27

Appendix C

90

Table C.1: Reduced models of male attendance, siring success and copulation success of males that attended a mating chase, and siring
success of males that copulated. The full GLMM models are described in the methods. The least significant terms were removed,
starting with the highest order. Male identity and mating chase identity were included as random effects. A binary error distribution
was assumed.
Record

Grid Year

Competition

F ID

Julian

Attendees Partners Sires

F Agg.

F Act.

F Doc.

1
2
3

KL2005
KL2005
KL2005

3.3
3.3
3.3

4329
4329
4706

93
93
103

9
9
12

Record

M ID

M Agg.

1
2
3

4559
4029
4559

−0.6
−0.2
−0.6

Record

M-F Distance (m)

1
2
3

289.3
114.0
162.6

7
7
2

3
3
2

−0.5
−0.5
−0.2

−0.6
−0.6
−0.2

3.0
3.0
7.7

M Act. M Doc. SS

BS

Attend

Copulate

Cop. Order

No. Cops.

−0.5
−1.1
−0.5

0
1
0

Y
N
NA

Y
N
NA

3
NA
NA

2
NA
NA

3.2
4.2
3.2

0
1
0

91

LITERATURE CITED

92

LITERATURE CITED

Alonzo, S. and Sinervo, B. 2001. Mate choice games, context-dependent good genes, and genetic
cycles in the side-blotched lizard, Uta stansburiana. Behavioral Ecology And Sociobiology
49:176–186.
Alonzo, S. and Warner, R. 2000. Allocation to mate guarding or increased sperm production in a
Mediterranean wrasse. American Naturalist 156:266–275.
Altmann, J. 1974. Observational study of behavior: sampling methods. Behaviour 49:227–267.
Andersson, M. and Iwasa, Y. 1996. Sexual selection. Trends in Ecology and Evolution 11:53–58.
Archer, J. 1973. Tests for emotionality in rats and mice: a review. Animal Behaviour 21:205–235.
Barry, K. L. and Kokko, H. 2010. Male mate choice: why sequential choice can make its evolution
difficult. Animal Behaviour 80:163–169.
Barton, N. H. and Turelli, M. 1989. Evolutionary quantitative genetics: how little do we know?
Annual Review of Genetics 23:337–370.
Bateman, P. and Fleming, P. 2006. Males are selective too: mating, but not courtship, with sequential females influences choosiness in male field crickets (Gryllus bimaculatus). Behavioral
Ecology And Sociobiology 59:577–581.
Bates, D., Maechler, M., and Bolker, B., 2012. lme4: Linear mixed-effects models using S4
classes.
Bateson, P., 1983. Mate Choice. Cambridge University Press, Cambridge, UK.
Bell, A. M. 2005. Behavioural differences between individuals and two populations of stickleback
(Gasterosteus aculeatus). Journal of Evolutionary Biology 18:464–473.
Bell, A. M., Hankison, S. J., and Laskowski, K. L. 2009. The repeatability of behaviour: a
meta-analysis. Animal Behaviour 77:771–783.
Bell, G. 2010. Fluctuating selection: the perpetual renewal of adaptation in variable environments.
Philosophical Transactions of the Royal Society B 365:87–97.
Berteaux, D. and Boutin, S. 2000. Breeding dispersal in female North American red squirrels.
Ecology 81:1311–1326.
Betini, G. S. and Norris, D. R. 2011. The relationship between personality and plasticity in tree
swallow aggression and the consequences for reproductive success. Animal Behaviour .
Biro, P. A. and Dingemanse, N. J. 2009. Sampling bias resulting from animal personality. Trends
in Ecology and Evolution 24:66–67.

93

Biro, P. A. and Stamps, J. A. 2008. Are animal personality traits linked to life-history productivity?
Trends in Ecology and Evolution 23:361–368.
Bleay, C. and Sinervo, B. 2007. Discrete genetic variation in mate choice and a conditiondependent preference function in the side-blotched lizard: implications for the formation and
maintenance of coadapted gene complexes. Behavioral Ecology 18:304–310.
Blows, M. W. 2007a. A tale of two matrices: multivariate approaches in evolutionary biology.
Journal of Evolutionary Biology 20:1–8.
Blows, M. W. 2007b. Complexity for complexity’s sake? Journal of Evolutionary Biology 20:39–
44.
Blows, M. W. and Brooks, R. 2003. Measuring nonlinear selection. American Naturalist 162:815–
820.
Blows, M. W., Brooks, R., and Kraft, P. 2003. Exploring complex fitness surfaces: Multiple
ornamentation and polymorphism in male guppies. Evolution 57:1622–1630.
Blows, M. W. and Hoffmann, A. A. 2005. A reassessment of genetic limits to evolutionary change.
Ecology 86:1371–1384.
Blumstein, D. T. and Daniel, J. C., 2007. Quantifying behavior the JWatcher way. Sinauer Associates, Sunderland (MA).
Blumstein, D. T., Lea, A. J., Olson, L. E., and Martin, G. A. 2010. Heritability of anti-predatory
traits: vigilance and locomotor performance in marmots. Journal of Evolutionary Biology
23:879–887.
Boissy, A. 1995. Fear and fearfulness in animals. Quarterly Review of Biology 70:165–191.
Bolker, B., Brooks, M., Clark, C. W., and Geange, S. 2008. Generalized linear mixed models: a
practical guide for ecology and evolution. Trends in Ecology and Evolution .
Boon, A. K., Réale, D., and Boutin, S. 2007. The interaction between personality, offspring fitness
and food abundance in North American red squirrels. Ecology Letters 10:1094–1104.
Boon, A. K., Réale, D., and Boutin, S. 2008. Personality, habitat use, and their consequences for
survival in North American red squirrels Tamiasciurus hudsonicus. Oikos 117:1321–1328.
Boutin, S., Tooze, Z., and Price, K. 1993. Post-breeding dispersal by female red squirrels (Tamiasciurus hudsonicus): the effect of local vacancies. Behavioral Ecology 4:151–155.
Boutin, S., Wauters, L. A., McAdam, A. G., Humphries, M. M., Tosi, G., and Dhondt, A. A. 2006.
Anticipatory reproduction and population growth in seed predators. Science 314:1928–1930.
Brodie, E. 1993. Homogeneity of the genetic variance-covariance matrix for antipredator triats in
two natural populations of the garter snake Thamnophis ordinoides. Evolution 47:844–854.
Brodie III, E. D., Moore, A., and Janzen, F. 1995. Visualizing and quantifying natural selection.
Trends in Ecology and Evolution 10:313–318.
94

Burnham, K. and Anderson, D., 1998. Model selection and inference. a practical informationtheoretic approach. Springer.
Calsbeek, R. and Sinervo, B. 2002. Uncoupling direct and indirect components of female choice in
the wild. Proceedings Of The National Academy Of Sciences Of The United States Of America
99:14897–14902.
Careau, V., Réale, D., Humphries, M. M., and Thomas, D. W. 2010. The pace of life under
artificial selection: personality, energy expenditure, and longevity are correlated in domestic
dogs. American Naturalist 175:753–758.
Careau, V., Thomas, D., Humphries, M. M., and Réale, D. 2008. Energy metabolism and animal
personality. Oikos 117:641–653.
Carere, C., Drent, P., Privitera, L., Koolhaas, J., and Groothuis, T. 2005. Personalities in great tits,
Parus major: stability and consistency. Animal Behaviour 70:795–805.
Champagne, F. A. 2008. Epigenetic mechanisms and the transgenerational effects of maternal
care. Frontiers In Neuroendocrinology 29:386–397.
Chapman, T., Arnqvist, G., Bangham, J., and Rowe, L. 2003. Sexual conflict. Trends in Ecology
and Evolution 18:41–47.
Cheverud, J. 1996. Developmental integration and the evolution of pleiotropy. American Zoologist
36:44–50.
Conner, J. K. 2007. A tale of two methods: Putting biology before statistics in the study of
phenotypic evolution. Journal of Evolutionary Biology 20:17–19.
Conner, J. K., Franks, R., and Stewart, C. 2003. Expression of additive genetic variances and covariances for wild radish floral traits: Comparison between field and greenhouse environments.
Evolution 57:487–495.
Conner, J. K. and Via, S. 1993. Patterns of phenotypic and genetic correlations among morphological and life-history traits in wild radish, Raphanus raphanistrum. Evolution 47:704–711.
Dantzer, B., Boutin, S., Humphries, M. M., and McAdam, A. G. 2012. Behavioral responses of
territorial red squirrels to natural and experimental variation in population density. Behavioral
Ecology And Sociobiology .
Darwin, C., 1871. The Descent of Man, and Selection in Relation to Sex. Murray, London.
Descamps, S., Boutin, S., Berteaux, D., McAdam, A. G., and Gaillard, J.-M. 2008. Cohort effects
in red squirrels: The influence of density, food abundance and temperature on future survival
and reproductive success. Journal Of Animal Ecology 77:305–314.
Dingemanse, N. J., Both, C., Drent, P., Oers, K. v., and van Noordwijk, A. 2002. Repeatability and
heritability of exploratory behaviour in great tits from the wild. Animal Behaviour 64:929–938.

95

Dingemanse, N. J., Both, C., Drent, P., and Tinbergen, J. 2004. Fitness consequences of avian
personalities in a fluctuating environment. Proceedings Of The Royal Society B-Biological
Sciences 271:847–852.
Dingemanse, N. J., Kazem, A. J. N., Réale, D., and Wright, J. 2010. Behavioural reaction norms:
animal personality meets individual plasticity. Trends in Ecology and Evolution 25:81–89.
Dingemanse, N. J. and Réale, D. 2005. Natural selection and animal personality. Behaviour
142:1159–1184.
Dingemanse, N. J., van der Plas, F., Wright, J., Réale, D., Schrama, M., Roff, D. A., van der
Zee, E., and Barber, I. 2009. Individual experience and evolutionary history of predation affect
expression of heritable variation in fish personality and morphology. Proceedings Of The Royal
Society B-Biological Sciences 276:1285–1293.
Dingemanse, N. J. and Wolf, M. 2010. Recent models for adaptive personality differences: a
review. Philosophical Transactions of the Royal Society B 365:3947–3958.
Dingemanse, N. J., Wright, J., Kazem, A. J. N., Thomas, D. K., Hickling, R., and Dawnay, N.
2007. Behavioural syndromes differ predictably between 12 populations of three-spined stickleback. Journal Of Animal Ecology 76:1128–1138.
Dloniak, S., French, J., and Holekamp, K. 2006. Rank-related maternal effects of androgens on
behaviour in wild spotted hyaenas. Nature 440:1190–1193.
Dochtermann, N. A. 2011. Testing Cheverud’s conjecture for behavioral correlations and behavioral syndromes. Evolution 65:1814–1820.
Dochtermann, N. A. and Roff, D. A. 2010. Applying a quantitative genetics framework to behavioural syndrome research. Philosophical Transactions of the Royal Society B 365:4013–
4020.
Dohm, M. 2002. Repeatability estimates do not always set an upper limit to heritability. Functional
Ecology 16:273–280.
Drent, P., van Oers, K., and van Noordwijk, A. 2002. Realized heritability of personalities in the
great tit (Parus major). Proceedings Of The Royal Society B-Biological Sciences 270:45–51.
Duckworth, R. A. and Kruuk, L. E. B. 2009. Evolution of genetic integration between dispersal
and colonization ability in a bird. Evolution 63:968–977.
Eising, C. M., Muller, W., and Groothuis, T. 2006. Avian mothers create different phenotypes by
hormone deposition in their eggs. Biology Letters 2:20–22.
Ellner, S. and Hariston Jr, N. G. 1994. Role of overlapping generations in maintaining genetic
variation in a fluctuating environment. American Naturalist 143:403–417.
Ellner, S. and Sasaki, A. 1996. Patterns of genetic polymorphism maintained by fluctuating
selection with overlapping generations. Theoretical Population Biology 50:31–65.

96

Falconer, D. and Mackay, T. F. C., 1996. Introduction to quantitative genetics. Longman, London.
Fauth, J., Bernardo, J., Camara, M., Resetarits, W., VanBuskirk, J., and McCollum, S. 1996.
Simplifying the jargon of community ecology: A conceptual approach. American Naturalist
147:282–286.
Fisher, R. A., 1930. The Genetical Theory Of Natural Selection. Oxford University Press, Oxford.
Forstmeier, W., Coltman, D. W., and Birkhead, T. 2004. Maternal effects influence the sexual
behavior of sons and daughters in the zebra finch. Evolution 58:2574–2583.
Fox, J. and Weisberg, S., 2011. An R Companion to Applied Regression. Sage, Thousand Oaks
CA, second edition.
Fraser, D., Gilliam, J., Daley, M., Le, A., and Skalski, G. 2001. Explaining leptokurtic movement distributions: Intrapopulation variation in boldness and exploration. American Naturalist
158:124–135.
Gorrell, J. C., McAdam, A. G., Coltman, D. W., Humphries, M. M., and Boutin, S. 2010. Adopting
kin enhances inclusive fitness in asocial red squirrels. Nature Communications 1:1–4.
Gosling, S. 2001. From mice to men: What can we learn about personality from animal research?
Psychological Bulletin 127:45–86.
Gosling, S. and John, O. 1999. Personality dimensions in nonhuman animals: a cross-species
review. Current Directions in Psychological Science 8:69–75.
Grafen, A., 1988. On the use of data on lifetime reproductive success in T. Clutton-Brock, ed.
Reproductive success. The University of Chicago Press, Chicago, IL.
Grant, P. R. and Grant, B. R. 2002. Unpredictable evolution in a 30-year study of Darwin’s finches.
Science 296:707–711.
Gunn, M. R., Dawson, D., Leviston, A., Hartnup, K., Davis, C., Strobeck, C., Slate, J., and Coltman, D. W. 2005. Isolation of 18 polymorphic microsatellite loci from the North American red
squirrel, Tamiasciurus hudsonicus (Sciuridae, Rodentia), and their cross-utility in other species.
Molecular Ecology Notes 5:650–653.
Hadfield, J. D. 2008. Estimating evolutionary parameters when viability selection is operating.
Proceedings Of The Royal Society B-Biological Sciences 275:723–734.
Hadfield, J. D. 2010. MCMC methods for multi-response generalized linear mixed models: the
MCMCglmm R package. Journal Of Statistical Software 33:1–22.
Hadfield, J. D. and Hadfield, M. 2011. Package ’MCMCglmm’ .
Hale, R. 2008. Evidence that context-dependent mate choice for parental care mirrors benefits to
offspring. Animal Behaviour 75:1283–1290.
Henderson, C., 1984. Applications of linear models in animal breeding. University of Guelph.

97

Henderson, N. 1986. Predicting relationships between psychological constructs and genetic characters - an analysis of changing genetic influences on activity in mice. Behavior Genetics
16:201–220.
Herdman, E. J. E., Kelly, C. D., and Godin, J. 2004. Male mate choice in the guppy (Poecilia
reticulata): Do males prefer larger females as mates? Ethology 110:97–111.
Hoffmann, A. A., 2000. Laboratory and field heritabilities: some lessons from Drosophila. Pages
200–218 in T. A. Mousseau, B. Sinervo, and J. Endler, eds. Adaptive Genetic Variation in the
wild. Oxford University Press.
Hoffmann, A. A. and Merilä, J. 1999. Heritable variation and evolution under favourable and
unfavourable conditions. Trends in Ecology and Evolution 14:96–101.
Houle, D. 1992. Comparing evolvability and variability of quantitative traits. Genetics 130:195–
204.
Houston, A. 2010. Evolutionary models of metabolism, behaviour and personality. Philosophical
Transactions of the Royal Society B 365:3969–3975.
Hurlbert, S. 1984. Pseudoreplication and the design of ecological field experiments. Ecological
Monographs 54:187–211.
Kalinowski, S. T., Taper, M. L., and Marshall, T. C. 2007. Revising how the computer program
CERVUS accommodates genotyping error increases success in paternity assignment. Molecular
Ecology 16:1099–1106.
Kerr, T. D., Boutin, S., LaMontagne, J. M., McAdam, A. G., and Humphries, M. M. 2007. Persistent maternal effects on juvenile survival in North American red squirrels. Biology Letters
3:289–291.
Kingsolver, J., Diamond, S., Siepielski, A., and Carlson, S. 2012. Synthetic analyses of phenotypic
selection in natural populations: lessons, limitations and future directions. Evolutionary Ecology
26:1101–1118.
Kingsolver, J. G. and Diamond, S. E. 2011. Phenotypic selection in natural populations: What
limits directional selection? American Naturalist 177:346–357.
Kingsolver, J. G., Hoekstra, H., Hoekstra, J., Berrigan, D., Vignieri, S., Hill, C., Hoang, A., Gibert,
P., and Beerli, P. 2001. The strength of phenotypic selection in natural populations. American
Naturalist 157:245–261.
Kingsolver, J. G. and Schemske, D. W. 1991. Path analyses of selection. Trends in Ecology and
Evolution 6:276–280.
Kokko, H., Brooks, R., Jennions, M., and Morley, J. 2003. The evolution of mate choice and
mating biases. Proceedings of the Royal Society of London. Series B, Biological Sciences
270:653–664.

98

Koolhaas, J., Korte, S., de Boer, S., van der Vegt, B., van Reenen, C., Hopster, H., de Jong, I.,
Ruis, M., and Blokhuis, H. 1999. Coping styles in animals: current status in behavior and
stress-physiology. Neuroscience And Biobehavioral Reviews 23:925–935.
Koprowski, J. L., 2007. Alternative Reproductive Tactics and Strategies of Tree Squirrels. Pages
86–95 in J. Wolff and P. W. Sherman, eds. Rodent Societies: An Ecological & Evolutionary
Perspective. University of Chicago Press, Chicago, IL.
Koyama, S. and Kamimura, S. 2000. Influence of social dominance and female odor on the sperm
activity of male mice. Physiology & behavior 71:415–422.
Kruuk, L. E. B. 2004. Estimating genetic parameters in natural populations using the ’animal
model’. Philosophical Transactions of the Royal Society B 359:873–890.
Kruuk, L. E. B. and Garant, D. 2007. A wake-up call for studies of natural selection? Journal of
Evolutionary Biology 20:30–33.
Kruuk, L. E. B. and Hadfield, J. D. 2007. How to separate genetic and environmental causes of
similarity between relatives. Journal of Evolutionary Biology 20:1890–1903.
Kruuk, L. E. B., Slate, J., and Wilson, A. J. 2008. New answers for old questions: The evolutionary
quantitative geneticsof wild animal populations. Annual Review of Ecology, Evolution, and
Systematics 39:525–548.
LaMontagne, J., Peters, S., and Boutin, S. 2005. A visual index for estimating cone production for
individual white spruce trees. Canadian Journal of Forest Research 35:3020–3026.
Lande, R. and Arnold, S. J. 1983. The measurement of selection on correlated characters. Evolution 37:1210–1226.
Lane, J. E., Boutin, S., Gunn, M. R., and Coltman, D. W. 2008a. Sexually selected behaviour: red
squirrel males search for reproductive success. Journal Of Animal Ecology 78:296–304.
Lane, J. E., Boutin, S., Gunn, M. R., Slate, J., and Coltman, D. W. 2008b. Female multiple mating
and paternity in free-ranging North American red squirrels. Animal Behaviour 75:1927–1937.
Lane, J. E., Boutin, S., Speakman, J. R., and Humphries, M. M. 2009. Energetic costs of male
reproduction in a scramble competition mating system. Journal Of Animal Ecology 79:27–34.
Larsen, K. W. and Boutin, S. 1994. Movements, survival, and settlement of red squirrel (Tamiasciurus hudsonicus) offspring. Ecology 75:214–223.
Le Comber, S., Faulkes, C., Formosinho, J., and Smith, C. 2003. Response of territorial males
to the threat of sneaking in the three-spined stickleback (Gasterosteus aculeatus): a field study.
Journal Of Zoology 261:15–20.
Lessells, C. and Boag, P. 1987. Unrepeatable repeatabilities - a common mistake. Auk 104:116–
121.

99

Levins, R. 1968. Evolution in changing environments; some theoretical explorations Monographs
in population biology, no. 2.
Lindstrom, J. 1999. Early development and fitness in birds and mammals. Trends in Ecology and
Evolution 14:343–348.
Lynch, M. and Walsh, B., 1998. Genetics and analysis of quantitative traits. Sinauer, Sunderland
(MA).
MacArthur, R. H. and Wilson, E. O., 1967. The theory of island biogeography. Princeton Univ Pr.
Magnhagen, C. 1991. Predation risk as a cost of reproduction. Trends in Ecology and Evolution
6:183–185.
Mann, M. A. and Svare, B. 1983. Prenatal testosterone exposure elevates maternal aggression in
mice. Physiology & behavior 30:503–507.
Martin, J. G. A. and Réale, D. 2008. Temperament, risk assessment and habituation to novelty in
eastern chipmunks, Tamias striatus. Animal Behaviour 75:309–318.
Martin, P. and Bateson, P., 1993. Measuring behaviour: an introductory guide. Cambridge University Press, Cambridge, 2nd edn. edition.
McAdam, A. and Boutin, S. 2003a. Effects of food abundance on genetic and maternal variation
in the growth rate of juvenile red squirrels. Journal of Evolutionary Biology 16:1249–1256.
McAdam, A. G. and Boutin, S. 2003b. Variation in viability selection among cohorts of juvenile
red squirrels (Tamiasciurus hudsonicus). Evolution 57:1689–1697.
McAdam, A. G., Boutin, S., Sykes, A. K., and Humphries, M. M. 2007. Life histories of female red
squirrels and their contributions to population growth and lifetime fitness. Ecoscience 14:362–
369.
McFarlane, S. E., Lane, J. E., Taylor, R. W., Gorrell, J. C., Coltman, D. W., Humphries, M. M.,
Boutin, S., and McAdam, A. G. 2011. The heritability of multiple male mating in a promiscuous
mammal. Biology Letters 7:368–371.
Meaney, M. 2001. Maternal care, gene expression, and the transmission of individual differences
in stress reactivity across generations. Annual Review of Neuroscience 24:1161–1192.
Merilä, J. and Sheldon, B. C. 2000. Lifetime reproductive success and heritability in nature.
American Naturalist 155:301–310.
Mitchell-Olds, T. and Shaw, R. 1987. Regression analysis of natural selection: Statistical inference
and biological interpretation. Evolution 41:1149–1161.
Morrissey, M. B. and Hadfield, J. D. 2012. Directional selection in temporally replicated studies
is remarkably consistent. Evolution 66:435–442.
Mousseau, T. and Fox, C. W., 1998. Maternal effects as adaptations. Oxford University Press,
Oxford.
100

Mousseau, T., Uller, T., Wapstra, E., and Badyaev, A. V. 2009. Evolution of maternal effects: past
and present. Philosophical Transactions of the Royal Society B 364:1035.
Mousseau, T. A. and Roff, D. A. 1987. Natural selection and the heritability of fitness components.
Heredity 59:181–197.
Nakatsuru, K. and Kramer, D. L. 1982. Is sperm cheap? Limited male fertility and female choice
in the lemon tetra (Pisces, Characidae). Science 216:753–755.
Neff, B. and Pitcher, T. 2005. Genetic quality and sexual selection: an integrated framework for
good genes and compatible genes. Molecular Ecology 14:19–38.
Pasterski, V., Hindmarsh, P., Geffner, M., Brook, C., Brain, C., and Hines, M. 2007. Increased aggression and activity level in 3-to 11-year-old girls with congenital adrenal hyperplasia (CAH).
Hormones And Behavior 52:368–374.
Phillips, P. C. and Arnold, S. J. 1989. Visualizing multivariate selection. Evolution 43:1209–1222.
Pinheiro, J. and Bates, D., 2000. Mixed-Effects Models in S and S-PLUS. Springer, New York,
NY.
Price, K. and Boutin, S. 1993. Territorial bequeathal by red squirrel mothers. Behavioral Ecology
4:144–150.
Quinn, J., Patrick, S., Bouwhuis, S., Wilkin, T., and B C Sheldon, B. C. 2009. Heterogeneous
selection on a heritable temperament trait in a variable environment. Journal Of Animal Ecology
78:1203–1215.
R Development Core Team, 2011. R: A Language and Environment for Statistical Computing.
Vienna, Austria.
R Development Core Team, 2012. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing.
Réale, D., Berteaux, D., and McAdam, A. G. 2003a. Lifetime selection on heritable life-history
traits in a natural population of red squirrels. Evolution 57:2416–2423.
Réale, D., Dingemanse, N. J., Kazem, A. J. N., and Wright, J. 2010a. Evolutionary and ecological approaches to the study of personality. Philosophical Transactions of the Royal Society B
365:3937–3946.
Réale, D., Gallant, B., Leblanc, M., and Festa-Bianchet, M. 2000. Consistency of temperament in
bighorn ewes and correlates with behaviour and life history. Animal Behaviour 60:589–597.
Réale, D., Garant, D., Humphries, M. M., Bergeron, P., Careau, V., and Montiglio, P.-O. 2010b.
Personality and the emergence of the pace-of-life syndrome concept at the population level.
Philosophical Transactions of the Royal Society B 365:4051–4063.
Réale, D., Martin, J. G. A., Coltman, D. W., Poissant, J., and Festa-Bianchet, M. 2009. Male
personality, life-history strategies and reproductive success in a promiscuous mammal. Journal
of Evolutionary Biology 22:1599–1607.
101

Réale, D., McAdam, A. G., Boutin, S., and Berteaux, D. 2003b. Genetic and plastic responses of a
northern mammal to climate change. Proceedings Of The Royal Society B-Biological Sciences
270:591–596.
Réale, D., Reader, S., Sol, D., McDougall, P., and Dingemanse, N. J. 2007. Integrating animal
temperament within ecology and evolution. Biological Reviews 82:291–318.
Reeve, H. K. and Sherman, P. W. 1993. Adaptation and the goals of evolutionary research.
Quarterly Review of Biology 68:1–32.
Reinhold, K. 2002. Maternal effects and the evolution of behavioral and morphological characters:
A literature review Indicates the importance of extended maternal care. Journal of Heredity
93:400–405.
Rezucha, R. and Smith, C. 2012. Personality traits, reproductive behaviour and alternative mating
tactics in male European bitterling, Rhodeus amarus. Behaviour .
Ricklefs, R. and Wikelski, M. 2002. The physiology/life-history nexus. Trends in Ecology and
Evolution 17:462–468.
Roff, D. and Fairbairn, D. J. 2007. The evolution of trade-offs: where are we?
Evolutionary Biology 20:433–447.

Journal of

Roff, D. A., 1997. Evolutionary quantitative genetics. Chapman & Hall, New York.
Roff, D. A. 2007. A centennial celebration for quantitative genetics. Evolution 61:1017–1032.
Sargent, R., Gross, M., and Van Den Berghe, E. 1986. Male mate choice in fishes. Animal
Behaviour 34:545–550.
Sarkar, D., 2008. Lattice: Multivariate Data Visualization with R. Springer, New York.
Schuett, W., Godin, J., and Dall, S. R. X. 2011. Do female zebra finches, Taeniopygia guttata,
choose their mates based on their ’personality’? Ethology 117:908–917.
Schuett, W., Tregenza, T., and Dall, S. R. X. 2010. Sexual selection and animal personality.
Biological Reviews 85:217–246.
Siepielski, A. M., DiBattista, J., and Carlson, S. 2009. It’s about time: the temporal dynamics of
phenotypic selection in the wild. Ecology Letters 12:1261–1276.
Siepielski, A. M., DiBattista, J. D., Evans, J. A., and Carlson, S. M. 2011. Differences in the
temporal dynamics of phenotypic selection among fitness components in the wild. Proceedings
Of The Royal Society B-Biological Sciences 278:1572–1580.
Sih, A. and Bell, A. M. 2008. Insights for behavioral ecology from behavioral syndromes. Advances in the Study of Behavior 38:227–281.
Sih, A., Bell, A. M., and Johnson, J. 2004a. Behavioral syndromes: an ecological and evolutionary
overview. Trends in Ecology and Evolution 19:372–378.

102

Sih, A., Bell, A. M., Johnson, J., and Ziemba, R. 2004b. Behavioral syndromes: an integrative
overview. Quarterly Review of Biology 79:241–277.
Sinervo, B. and Lively, C. 1996. The rock-paper-scissors game and the evolution of alternative
male strategies. Nature 380:240–243.
Sinervo, B. and McAdam, A. G. 2008. Maturational costs of reproduction due to clutch size and
ontogenetic conflict as revealed in the invisible fraction. Proceedings Of The Royal Society
B-Biological Sciences 275:629–638.
Sinervo, B., Svensson, E., and Comendant, T. 2000. Density cycles and an offspring quantity and
quality game driven by natural selection. Nature 406:985–988.
Sinervo, B. and Svensson, E. I. 2002. Correlational selection and the evolution of genomic architecture. Heredity 89:329–338.
Sinn, D. L., Apiolaza, L. A., and Moltschaniwskyj, N. 2006. Heritability and fitnessrelated consequences of squid personality traits. Journal of Evolutionary Biology 19:1437–1447.
Smith, B. R. and Blumstein, D. T. 2008. Fitness consequences of personality: a meta-analysis.
Behavioral Ecology 19:448.
Smith, C. 1968. The adaptive nature of social organization in the genus of three squirrels Tamiasciurus. Ecological Monographs 38:31–63.
Spence, R. and Smith, C. 2005. Male territoriality mediates density and sex ratio effects on
oviposition in the zebrafish, Danio rerio. Animal Behaviour 69:1317–1323.
Spiegelhalter, D., Best, N., Carlin, B., and van der Linde, A. 2002. Bayesian measures of model
complexity and fit. Journal of the Royal Statistical Society: Series B (Statistical Methodology)
64:583–639.
Stamps, J. 1991. Why evolutionary issues are reviving interest in proximate behavioral mechanisms. American Zoologist 31:338–348.
Stamps, J. A. 2007. Growth-mortality tradeoffs and ’personality traits’ in animals. Ecology Letters
10:355–363.
Stamps, J. A. and Groothuis, T. 2010. Developmental perspectives on personality: implications
for ecological and evolutionary studies of individual differences. Philosophical Transactions of
the Royal Society B 365:4029–4041.
Stearns, S. 1989. The evolutionary significance of phenotypic plasticity. BioScience 39:436–445.
Steele, M. A. 1998. Tamiasciurus hudsonicus. Mammalian Species Pages 1–9.
Stinchcombe, J. R., Agrawal, A. F., Hohenlohe, P. A., Arnold, S. J., and Blows, M. W. 2008. Estimating nonlinear selection gradients using quadratic regression coefficients: Double or nothing
? Evolution 62:2435–2440.

103

Stirling, D., Réale, D., and Roff, D. A. 2002. Selection, structure and the heritability of behaviour.
Journal of Evolutionary Biology 15:277–289.
Svendsen, G. and Armitage, K. 1973. Mirror-image stimulation applied to field behavioral studies.
Ecology 54:623–627.
Taylor, R. W., Boon, A. K., Dantzer, B., Réale, D., Humphries, M. M., Boutin, S., Gorrell, J. C.,
Coltman, D. W., and McAdam, A. G. 2012. Low heritabilities, but genetic and maternal correlations between red squirrel behaviours. Journal of Evolutionary Biology 25:614–624.
Tregenza, T. and Wedell, N. 2000. Genetic compatibility, mate choice and patterns of parentage:
Invited review. Molecular Ecology 9:1013–1027.
Trivers, R. 1972. Parental investment and sexual selection 136.
Tulley, J. and Huntingford, F. 1988. Additional information on the relationship between intraspecific aggression and anti-predator behaviour in the three-spined stickleback, gasterosteus aculeatus. Ethology 78:219–222.
van Oers, K., Drent, P., de Jong, G., and van Noordwijk, A. 2004. Additive and nonadditive
genetic variation in avian personality traits. Heredity 93:496–503.
Walsh, R. and Cummins, R. 1976. The open-field test: a critical review. Psychological Bulletin
83:482–504.
Waterman, J. 1998. Mating tactics of male Cape ground squirrels, Xerus inauris: consequences of
year-round breeding. Animal Behaviour 56:459–466.
Weigensberg, I. and Roff, D. A. 1996. Natural heritabilities: can they be reliably estimated in the
laboratory? Evolution 50:2149–2157.
Weinstock, M. 2008. The long-term behavioural consequences of prenatal stress. Neuroscience
And Biobehavioral Reviews 32:1073–1086.
Wickham, H., 2009. ggplot2: elegant graphics for data analysis. Springer New York.
Wilcoxon, J. S. and Redei, E. E. 2007. Maternal glucocorticoid deficit affects hypothalamicpituitary-adrenal function and behavior of rat offspring. Hormones And Behavior 51:321–327.
Wilson, A. J. 2008. Why h2 does not always equal VA/VP? Journal of Evolutionary Biology
21:647–650.
Wilson, A. J. and Réale, D. 2006. Ontogeny of additive and maternal Genetic effects: Lessons
from domestic mammals. American Naturalist 167:E23–E38.
Wilson, A. J., Réale, D., Clements, M., Morrissey, M., Postma, E., Walling, C., Kruuk, L. E. B.,
and Nussey, D. 2009. An ecologist’s guide to the animal model. Journal Of Animal Ecology
79:13–26.
Wilson, D., Clark, AB, Coleman, K., and Dearstyne, T. 1994. Shyness and boldness in humans
and other animals. Trends in Ecology and Evolution 9:442–446.
104

Wolf, M., van Doorn, G. S., Leimar, O., and Weissing, F. J. 2007. Life-history trade-offs favour
the evolution of animal personalities. Nature 447:581–584.
Wood, S. N. 2003. Thin plate regression splines. Journal of the Royal Statistical Society: Series
B (Statistical Methodology) 65:95–114.
Wood, S. N., 2006. Generalized additive models. an introduction with R. CRC Press.

105