ANIMAL BEHAVIOUR, 2002, 64, 619–628
doi:10.1006/anbe.2002.3079, available online at http://www.idealibrary.com on
Female choice and the benefits of mate guarding by male
mallards
ELLEN S. DAVIS
Department of Zoology, University of Wisconsin, Madison
(Received 12 December 2000; initial acceptance 17 May 2001;
final acceptance 18 February 2002; MS. number: A8944R )
Pair formation and breeding in many species of waterfowl are separated both temporally and spatially.
Most studies of female choice in this group have focused on male characteristics at the time of pairing,
with less attention given to how mate choice affects breeding season outcomes. In this study I compared
pairing success, male plasma testosterone level and mate-guarding ability of male mallards, Anas
platyrhynchos, in two experiments. In the first experiment females and males were group housed with
equal sex ratios, thus allowing all of these males to pair. At the same time, an equal number of males was
housed in groups without access to females and remained unpaired. In this experiment testosterone levels
of paired and unpaired males during autumn (baseline) and spring (breeding) did not differ, indicating
that the process of pair formation and breeding does not cause elevated spring testosterone levels in
males. However, testosterone did temporarily decrease in paired males during the winter (pair formation)
season. In the second experiment groups were male biased, allowing only half of the males to pair. Here
paired males had significantly higher testosterone levels than unpaired males during the breeding season,
but not during the preceding autumn. Together the results of these experiments indicate that successful
pair formation predicts but does not alter male testosterone level during the breeding season. I also found
that females paired to males with high levels of testosterone were missing fewer feathers due to forced
copulation attempts by nonmates, suggesting that females may choose males based on their mateguarding abilities.

2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.
Numerous studies of mate choice in mallards, Anas
platyrhynchos, have indicated that the courtship activity
level of the male is a consistent predictor of female
mate choice (Bossema & Kruijt 1982; Bossema & Roemers
1985; Holmberg et al. 1989; but see Brodsky et al. 1988).
Other characteristics such as plumage quality (Klint 1980,
1985; Bossema & Kruijt 1982; Holmberg et al. 1989;
Weidmann 1990), corticosterone (B) level (Sorenson et al.
1997) and body condition (Wishart 1983; Holmberg et al.
1989) may also be important in this and other duck
species, but the evidence for body size (Bossema &
Roemers 1985; Holmberg et al. 1989) and dominance
(Wishart 1983; Brodsky et al. 1988; Sorenson &
Derrickson 1994) is more conflicting. However, all of
these studies have considered only those characteristics of
males displayed at the time of pair formation, with little
attention to male phenotype during the breeding
attempt. Pairs may remain together for as much as
Correspondence: E. S. Davis, Psychology Department, Brogden Hall,
1202 W. Johnson Street, Madison, WI 53706 U. S. A. (email:
[email protected])
0003–3472/02/$35.00/0

6 months or longer each year (Hepp & Hair 1983), during
which time males and females synchronize activities,
and males begin to display mate-guarding behaviour by
remaining close to the female and alert for conspecific
intrusions and predators. Mate guarding is important
to female breeding condition and success (Ashcroft
1976; Ankney & MacInnes 1978; Seymour & Titman
1978; Titman 1981; Gauthier 1987; Lamprecht 1989;
Pattenden & Boag 1989) and is energetically costly for
males (Ashcroft 1976; Titman 1981; Lamprecht 1989).
Furthermore, in many waterfowl species, mate guarding
is the only male contribution to the reproductive
effort, aside from donating sperm (Goodburn 1984).
Therefore, it would seem important to females to
choose mates who guard them well. Moreover, as
mate guarding is an aggressive behaviour, and aggression is generally under the control of testosterone (T)
in birds (reviewed by Harding 1983), mate choice,
male T levels and mate-guarding behaviour may all be
associated. In this study I explore the possible link
between autumn female mate choice and spring
mate-guarding ability, as well as possible hormonal
619
2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved.
620
ANIMAL BEHAVIOUR, 64, 4
correlates throughout the pair formation and breeding
seasons.
Mate Guarding and Testosterone
Mate guarding in waterfowl is a behaviour that is likely
to increase both male and female fitness by serving
primarily to increase female foraging efficiency (Ashcroft
1976; Seymour & Titman 1978; Titman 1981; Gauthier
1987; Lamprecht 1989), which may occur via increased
male vigilance (Lamprecht 1989) or the dominance of
pairs over unpaired birds (Raveling 1970; Paulus 1983;
Hepp & Hair 1984). Females with increased nutrient
reserves, in turn, lay eggs earlier in the breeding season
(Pattenden & Boag 1989) and are more likely to survive
the breeding effort (Ankney & MacInnes 1978). In some
species mate guarding also protects females from forced
extrapair copulation (FEPC) attempts by nonmates
(Barash 1977; Mineau & Cooke 1979; Goodburn 1984;
Seymour 1990; Sorenson 1994a, b), thereby also protecting paternity. In addition, it has been proposed that mate
guarding may help to maintain the pair bond and protect
the female from depredation (McKinney 1988). In this
study, I used protection against FEPC attempts as a proxy
for overall mate-guarding ability because birds in a
provisioned, captive setting spend less time actively
foraging, and therefore foraging differences are not easily
measured. Forced extrapair copulation, however, is very
common in such densely populated groups (Titman &
Lowther 1975; Seymour 1990; personal observation), and
protection from FEPC attempts rather than increased
foraging efficiency is likely to be the major benefit
of mate guarding that females receive in a captive
population.
Mate guarding is an aggressive activity, and aggression
in mallards is generally well correlated with testosterone
(Etienne 1964; Balthazart & Stevens 1975; Schmedemann
& Haase 1984); it is therefore likely that mate guarding
may be influenced by T as well. Conversely, there are
several ways in which experience can influence T levels.
For example, during the breeding season, the T levels of
territorial males rise in response to a male intruder, a
phenomenon known as the ‘challenge effect’ (Wingfield
& Moore 1987; Wingfield et al. 1990; Wingfield 1994).
Testosterone level may also be influenced by stress. The
hormone corticosterone (B) increases quickly in response
to stresses such as handling, and can suppress T levels
when elevated (Siegel 1971, 1980; Harvey et al. 1980), or
B may directly affect behaviour. Care must be taken
therefore when attempting to evaluate the influence of
endogenous T levels on behaviour. It is possible that
mate-guarding behaviour may be under the control of T,
but it is important to demonstrate that neither stress nor
activities such as pair formation, copulation and mate
guarding are affecting T levels.
This study examines two main questions that serve to
link mate choice to mate-guarding behaviour. First, do
males chosen by females have higher T levels than males
not chosen? Second, are males with higher T levels better
at mate guarding?
METHODS
Study Site and Population
All experiments were conducted at the Hillebrand Rare
Bird Ranch in Cross Plains, Wisconsin, U.S.A. In spring
1997, I raised ducklings from eggs collected from nests
on the property for experiment 1. Owing to high nest
depredation that year, it was also necessary to purchase
40 day-old ducklings from Whistling Wings Mallard Farm
in Hanover, Illinois, U.S.A. I used only first-year birds
in the study to control for age and experience effects.
Offspring of these birds were used beginning in the spring
of 1998 for experiment 2.
In the autumn of each year, all males and females were
individually and redundantly marked with coloured nasal
markers and matching leg bands. All birds were pinioned
and housed in covered, wire mesh outdoor pens with
unlimited access to food and to water in small wading
pools for drinking and bathing. To control for bird
density, pens containing both males and females as
described below were approximately twice as large
(100 m2 and 110 m2) as pens containing males only
(50 m2 and 56 m2). The birds were exposed to ambient
light and temperatures throughout the study. Water in
the wading pools was heated as necessary to prevent
freezing.
Pairing Effects: Experiment 1
Experiment 1 was conducted to determine whether pair
formation or breeding activities can affect male T levels
during the breeding season. Here I randomly assigned
males to be either paired or unpaired. In November 1997,
during the autumn courtship season, each of 34 males
was assigned to one of four adjacent home pens without
regard to relatedness, as it was largely unknown. There
were nine males in each of two pens and eight in each of
the other two pens. Each pen was visually isolated from
all other pens by means of opaque barriers. After giving
the males 2 weeks to become accustomed to the pen and
group, I took baseline blood samples from each male to
ensure that there were no hormone differences between
groups at this time. As mentioned above, stress can affect
steroid hormone levels. To avoid possible stress effects
due to prolonged disturbance within a pen, I sampled
blood as described below from only one randomly
selected male per pen per day beginning 21 November
until they all were sampled (2 December). All males were
bled between 1 and 3 h postsunrise throughout the entire
study to minimize the effects of diurnal fluctuations
in T seen in this species (Balthazart 1976; Balthazart
& Hendrick 1979). Mean sampling duration for each
sampling period is reported in Table 1.
Females were introduced into two of the pens on
6 December (eight and nine females, respectively) such
that the sex ratio within those two pens was 1:1. These
pens are referred to as ‘paired-male’ pens. To ensure that
birds in the paired-male pens actually paired and to
identify mates, males in these pens were observed for
evidence of pair formation using 3-min focal sampling
DAVIS: FEMALE CHOICE AND MATE GUARDING
Table 1. Summary of regression of hormone levels on sampling duration
Sampling duration
Testosterone
Corticosterone
Mean±SE
F
r2
P
F
r2
P
Experiment 1
Nov (unpaired)
Jan (early pairing)
Feb (late pairing)
April (breeding)
192.9±11.4
173.2±9.6
155.6±7.8
137.0±6.3
0.06
0.45
0.46
0.15
0.002
0.014
0.014
0.005
0.81
0.51
0.50
0.70
28.1
18.6
12.6
1.7
0.47
0.37
0.28
0.06
<0.0001
0.0002
0.001
0.2
Experiment 2
Nov (unpaired)
April (breeding)
128.4±9.9
162.1±12.7
2.81
0.54
0.076
0.024
0.10
0.47
1.8
0.3
0.05
0.01
0.19
0.76
Sampling period
techniques. Potential mallard pairs are identified easily
by: (1) the female inciting display, which indicates
preference for a particular male (Lorenz 1953; Weidmann
1956; Johnsgard 1961; McKinney 1992; Sorenson &
Derrickson 1994); (2) precopulatory displays (Weidmann
1956; McKinney et al. 1983); and (3) copulation. Pairing
was also assessed by the (4) proximity and (5) synchronization of activities of a male and female (McKinney
1992). A pair bond was considered to have formed if the
same two birds met any of these criteria during three
successive focal sampling periods. These observations
continued at least twice per week through the end of
March, and at least once per week thereafter to confirm
the identity of pairs using the same criteria as above.
Much of the behaviour recorded involved the two latter
criteria, but the first three criteria, which are more overt
indications of mate choice and pair formation, helped to
confirm the identity of pairs. Fifteen of the 17 females
incited to their presumed mates, 15 pairs performed
precopulatory displays and nine attempted or successfully performed at least one copulation during focal
sampling observations. Several instances of mate switching did occur during the observation period, using the
criteria described above, but all individuals in the pairedmale pens had permanently paired by the beginning of
the spring sampling period.
As there is some evidence that male mallards can assess
the breeding condition of females (Sorenson 1994a;
McKinney & Evarts 1998), it was necessary to allow the
unpaired males to observe female activities. Therefore, I
removed all visual barriers after the autumn sampling so
that males in the two unpaired-male pens were still
isolated physically but not visually from females, and
were therefore able to observe female behaviour and
breeding condition. Also, to assay hormone levels across
the pair-formation and breeding seasons, I drew
additional blood samples from the males in January,
February and April 1998 (see below). I was unable to
successfully obtain a January blood sample from one
unpaired male.
Data were also recorded on male–male interactions to
monitor levels of aggression in the pens and ensure the
well being of individuals. Preliminary observations indicated very low levels of male–male interactions during
the autumn and winter in the unpaired-male pens, presumably owing to the absence of females. Because overall
interactions were infrequent in these pens and due to
time limitations, unpaired males were observed systematically in 30-min sessions beginning February using
ad libitum sampling techniques. By the beginning of
the breeding season, however, male–male aggression
increased in one of the unpaired-male pens, and three
males were removed from that pen before the study
ended because they were severely harassed by other
males; the final sample size was therefore 31.
Pairing Effects: Experiment 2
Whereas the goal of experiment 1 was to determine
whether pair formation and breeding activities could
influence T levels, the goal of experiment 2 was to
determine whether female choice was associated with
either autumn or spring T level. The methodology in the
second year was similar to that of the first year overall
except that all males were initially housed together and
competed for a limited number of females. In November
1998, I assigned 36 males to one large home pen and gave
them 2 weeks to become accustomed to the pen and
group. I again drew baseline blood samples, as described
below, on four different days between 26 November and
6 December 1998. Because the previous results of experiment 1 indicated that neither handling nor sampling
duration affected T levels (see Results), I sampled multiple
males from the same pen on the same day. After all males
were sampled, I placed 18 females into the pen so that the
sex ratio was male biased at 2:1, which was similar to the
overall sex ratio in experiment 1. To identify pairs as they
formed, birds were observed at least two times per week as
described above, but this time females were the focal
individuals as they were fewer in number. As in experiment 1, most of the data recorded involved more subtle
indications of pair formation, but of the 12 pairs that met
the pairing criteria, eight of the females had incited
to their mates, 11 pairs were observed performing
precopulatory displays and four were observed to have
attempted or successfully performed at least one
copulation. Two females died over the course of the
winter from unknown causes and all males survived.
By the beginning of spring, only 12 of the 16 remaining
females appeared to be reliably paired. On 23 March, I
removed all 24 unpaired males and the four unpaired
females from this home pen (now the paired-male pen) to
621
622
ANIMAL BEHAVIOUR, 64, 4
create a paired-male and unpaired-male group similar to
experiment 1. At this time, I also randomly selected 12 of
the unpaired males to be placed in an adjacent, unpairedmale home pen. The remaining unpaired birds were
transferred to a distant pen that was visually isolated from
the two home pens and were not used further in this
study. I did not include four of the 24 males in this
selection process either because they demonstrated
homosexual behaviour such as precopulatory displays
towards one another (Schutz 1965), or because they had
been paired with one of the females who had died,
making their pair status difficult to determine. I took
blood samples from the males again in April 1999 to assay
breeding season hormone levels.
Blood Sampling and Hormone Assays
To draw a blood sample a field assistant and I stood
outside of the home pen and first visually located the
randomly determined target male. We then entered
the pen and the assistant captured the male with a net.
The male was rolled up in a towel with only his head and
neck exposed, and I then drew a 2.5-cc blood sample from
the jugular vein using a heparinized syringe and 22–gauge
needle. I timed this process from the moment we entered
the pen until the sample was drawn. During the autumn
sampling of experiment 2, I sampled subsequent males
while remaining in the home pen and timed the process
from the point at which the field assistant began to move
towards the next target male. Thus, I recorded both the
sampling duration (time involved in chasing and
sampling a particular male) as well as the time of day that
blood was drawn from each male. After all the samples
were drawn for the given day, I centrifuged them for
20 min, and transferred the plasma portions to 2-ml
storage vials. I placed these vials in a small cooler with ice
packs until they could be stored in a 20 C freezer later
that day.
All samples were analysed for T and B in the summer
following their collection at the University of Winsconsin
Regional Primate Research Center Assay Laboratory. I
extracted the samples and separated the progesterone,
dihydrotestosterone, T and B fractions by celite chromatography (Abraham et al. 1972) before assaying using
radioimmunoassay techniques (Robinson et al. 1975;
Moore 1986) with internal recoveries. Testosterone antibody was purchased from Holly Hill Farms in Hillsboro,
Oregon, U.S.A. and the B antibody was purchased from
Endocrine Sciences Products of Calbasas Hills, California,
U.S.A. Laboratory staff conducted the assay validations.
Serial dilutions of the plasma pool (N=7) gave parallelism
to the standards for T and B with no differences in slopes
(P>0.05). Accuracy for T and B was 104.0 and 99.2%,
respectively. Sensitivities of the assays were 5 pg/tube and
36 pg/tube for T and B, respectively.
I analysed two levels of pooled samples for each
hormone for the samples from experiment 1. The intraand interassay coefficients of variation (CVs) for B were
5.97 and 14.84%, respectively, for the low pool, and
3.19% and 10.77% for the high pool. The intra- and
interassay CVs for T were 3.50 and 7.91%, respectively,
for the low pool, and 2.02 and 7.42% for the high pool. I
analysed one pooled sample for each hormone for experiment 2 samples. The intra- and interassay CVs for B were
4.01% and 18.48%, respectively. The intra- and interassay
CVs for T were 4.94 and 9.87%, respectively.
Mate-guarding Success
Protection from FEPC attempts may be assessed by
measuring the plumage quality of the female. A male
typically grabs the head or neck feathers of a female as he
mounts to copulate. If the attempt is forced, the female
often loses some of these feathers as she struggles to
escape. Females that frequently are subjected to FEPC
attempts often have bald and sometimes bleeding areas
on the head, neck or back (Titman & Lowther 1975;
McKinney & Evarts 1998; personal observation). Males
often attempt to fend off FEPC attacks on their mates
with varying success (Mineau & Cooke 1979; McKinney
et al. 1983; personal observation). Therefore, the amount
of feather loss shown by a female can be used as an
indicator of the ability of her mate to successfully guard
her from FEPC attempts. Approximately 1 month after
females began laying eggs, I captured each female with
the help of a field assistant and measured the length and
width of the areas on the head, neck and back of the
female from which feathers had been pulled. Feathers
pulled from these areas are not replaced until the postnuptial moult, which follows the breeding season. I then
calculated the area of feather loss for each area and
summed them to yield one measure of total feather loss
for each female. All the females were captured and
measured on the same day. Mate-guarding success was
measured only during experiment 2.
RESULTS
Handling Effects
Sampling duration
Correlations between sampling duration and hormone
levels for both years are reported in Table 1. Of the six
sampling periods, B levels were correlated with sampling
duration only for the first three sampling periods, presumably because sampling duration decreased with
experience, such that sampling was completed in the
later sampling periods before B increased significantly.
There was no correlation between sampling duration and
T for any sampling period. For the sake of further analysis, I used B residuals for all experiment 1 sampling
periods presented above.
Time-of-day Effects
Mallard T level follows a diurnal cycle, with an early
morning peak in winter (Balthazart 1976), and both an
early morning peak and higher peak in early afternoon in
spring (Balthazart & Hendrick 1979). In the autumn
sampling period of experiment 2, I sampled as many as
10 males per day, a process that took nearly an hour. I
DAVIS: FEMALE CHOICE AND MATE GUARDING
Pairing Effects: Experiment 1
Testosterone
Figure 1 shows the mean plasma T levels for paired and
unpaired males across all sampling periods. A nested
ANOVA indicated there was no effect due to the individual pens, allowing the four pens to be combined into
two treatments: paired and unpaired. Repeated measures
ANOVA revealed an overall difference between T levels
of paired and unpaired males (ANOVA: F1,28= 6.9581,
P=0.014). Pairwise two-tailed randomization tests
between paired and unpaired males showed that there
was a highly significant difference in January T levels
during early pairing (P=0.0003), but no difference for the
other three sampling periods (P=0.50, 0.12 and 0.63 for
November, February and April, respectively).
The T levels of the males in the unpaired-male pens
increased significantly between November and each
of the other sampling periods (one-tailed paired t test:
t15 = 2.49, P=0.012; t16 = 3.62, P=0.001; and
t13 = 3.54, P=0.002 for comparisons with January,
February and April, respectively). Likewise, unpaired
males had significantly different T levels between January
and April (t12 = 2.21, P=0.023) and between February
and April (t13 = 2.07, P=0.029), but not between January
and February (t15 = 0.27, P=0.396).
Only 14 of the 17 males in the paired-male pens were
paired by the January sampling and so only those 14 T
levels were used in analyses involving the January
sampling. All pairwise comparisons of T levels between
sampling periods were significantly different, using onetailed paired t tests: November to January, t13 =1.94,
P=0.037; November to February, t16 = 1.78, P=0.045;
November to April, t16 = 3.13, P=0.003; January to
February, t13 = 3.12, P=0.004; January to April,
t13 = 4.79, P=0.0002; and February to April, t16 = 1.89,
P=0.038. All of these differences represent increases in T
over time except for the difference between November
and January, which represents a decrease. Moreover, of
4
Paired
Unpaired
3
Backtransformed T (ng/ml)
therefore examined the data for time-of-day effects.
Regression analysis of autumn T level versus time of
day as measured by minutes postsunrise revealed a
highly significant negative effect (F1,34 =19.9, r2 =0.37,
P=0.0001). However, there was no such effect on B during
the autumn (F1,34 =0.862, r2 =0.025, P=0.36). There was
also no effect in the spring on either T (F1,22 =0.0819,
r2 =0.004, P=0.78) or B (F1,22 =0.0247, r2 =0.001, P=0.88). I
used the residuals obtained from this analysis for all
subsequent analyses involving autumn T levels reported
above.
Overall T levels were significantly higher in the
spring of experiment 2 than experiment 1 (F1,53 =4.05,
P=0.049). However, birds were sampled on average
slightly later in the day in experiment 1 than experiment
2 (mean time postsunriseSD: 138.112.41 min and
100.123.8 min, respectively). When sampling time
postsunrise was included in the ANOVA model as a
covariate, the difference in T level between experiments
disappeared (F1,52 =1.64, P=0.21).
2
1
0
November
(autumn)
January
February
(winter)
April
(spring)
Figure 1. Testosterone (T) level and pair status (experiment 1).
Means are backtransformed from ln-transformed data. Bars indicate
standard errors, and means lacking bars indicate errors too small to
show clearly. Major results of post hoc analysis include significant
differences between paired and unpaired males only in January
(P=0.0003). Also, the T levels of unpaired males increased significantly between November and January (P=0.012), whereas the T
levels of paired males decreased between those sampling periods
(P=0.037). Sample size was 17 for each group with two exceptions,
the sample size for unpaired males was 16 in January and 14 in April.
See text for complete information and statistical results.
the 14 males that were paired by January, 10 showed
a decrease in T level from November, whereas only
two unpaired males showed a decrease in T between
November and January.
Corticosterone
Figure 2 shows the mean B residuals for paired and
unpaired males across all sampling periods. Repeated
measures ANOVA yielded no significant difference
between paired and unpaired males in their overall B
residuals (ANOVA: F1,28 =0.372, P=0.55). Corticosterone
residuals were somewhat larger in April, a fact that is
consistent with the lack of significant correlation with
sampling duration for that sampling. Similarly, T was not
correlated with B for any sampling period (regression:
F1,32 =0.866, r2 =0.026, P=0.36; F1,31 =0.004, r2 < 0.001,
P=0.95; F1,32 =1.01, r2 =0.031, P=0.32; F1,29 =0.323,
r2 =0.011, P=0.57 for November, January, February and
April, respectively).
Pairing Effects: Experiment 2
Figure 3 shows no association between T and pair status
in the autumn (ANOVA: F1,20 =0.917, P=0.35) for the 24
623
ANIMAL BEHAVIOUR, 64, 4
8
0.4
P = 0.034
30
T of mate (ng/ml)
B residuals (ng/ml)
0
:T
:B
20
4
0
–0.4
Paired
Unpaired
–0.6
November
(autumn)
January
February
(winter)
April
(spring)
Figure 2. Mean residuals of ln-transformed corticosterone (B) data.
There was no significant difference between B levels of paired and
unpaired males for any sampling period. Bars indicate standard
errors. Sample sizes are as in Fig 1.
5.0
10
2
–0.2
Paired
Unpaired
P = 0.022
B of mate (ng/ml)
6
0.2
< Median
> Median
(well guarded)
(poorly guarded)
Female feather loss
0
Figure 4. Male hormone level (T: testosterone; B: corticosterone)
versus female feather loss. Means and standard errors of hormone
levels of mates of females with feather loss less than or greater than
the median. Females with feather loss less than the median had
mates with higher T levels than females with feather loss greater
than the median, but there was no difference in their mates’ B levels.
The sample size was six in each group.
autumn were included in the analysis (F1,34 =0.0229,
P=0.88). However, by spring, paired males had significantly higher T levels than unpaired males (F1,20 =6.1,
P=0.022), in contrast to experiment 1.
Mate-guarding Success
4.0
Backtransformed T (ng/ml)
624
Regression of feather loss on male spring T level did not
yield a significant linear relationship (F1,10 =2.61,
r2 =0.207, P=0.14). However, I also divided the females
into two groups based on the median amount of feather
loss (Fig. 4), and then calculated the mean T level of their
respective mates. Females that were relatively well
guarded (feather loss below the median) had mates with
significantly higher T levels than females that were less
well guarded (feather loss above the median) (two-tailed
t test: t10 =2.477, N1 =N2 =6, P=0.034,). No relationship
was observed, however, between B and feather loss (twotailed t test: t10 = 1.309, N1 =N2 =6, P=0.24). Thus,
although the relationship was not linear, males with
higher T levels appeared to be more successful at mate
guarding than males with lower T levels.
3.0
2.0
1.0
P = 0.35
0.0
Autumn
Spring
Figure 3. Testosterone (T) and pair status (experiment 2). Means
and standard errors of autumn and spring T levels are shown for
paired and unpaired males. The sample size was 12 in each group.
males used in both autumn and spring of experiment 2.
Nor was there an association between autumn T and pair
status when the 12 unpaired males used only in the
DISCUSSION
The results of this study indicate that females prefer males
during the autumn pair formation season that will have
high breeding season T levels in the spring, and that these
males are more successful at mate guarding than males
with low T. This result is important because it not only
identifies one possible function of female choice in
waterfowl, but it also describes a choice that is predictive.
DAVIS: FEMALE CHOICE AND MATE GUARDING
That is, females are making decisions in the autumn that
bear on their own future reproductive success in the
spring.
Mate Choice and Mate-guarding Success
Males that paired successfully when competing for a
limited number of females during the autumn and winter
courtship season (experiment 2) had higher spring T
levels than unsuccessful males. This result contrasts
with the results of experiment 1: when pair status was
randomly predetermined, there was no difference in
spring T levels between paired and unpaired males, indicating that pair activities during the breeding season,
such as copulation and mate guarding, did not affect T
levels, as occurs in numerous other species (for a review
see Wingfield 1994). In contrast, Hirschenhauser et al.
(2000) found that pair status did appear to modulate T
levels in male greylag geese. However, paired and
unpaired males in that study were self-selected (as in
experiment 2 here), and it is instead possible that the
female geese preferred high T males.
It is also possible that the difference in results between
experiments 1 and 2 was caused by differences in male–
male aggression owing to differences in sex ratios
between experiments (1:1 versus 2:1), but this seems
unlikely for several reasons. First, when the time of day
that sampling occurred was included in the analyses, no
difference in spring T levels existed between years, indicating that possible differences in competition between
years did not result in overall differences in T levels.
Rather, the apparent difference in overall T levels
between experiments (cf. Figs 1 and 3) is likely due to the
fact that T levels fall throughout the morning during the
spring in mallards (Balthazart & Hendrick 1979), and
males in experiment 1 were sampled slightly later in the
morning on average than males in experiment 2. Second,
the winter hormonal profile of paired males in experiment 1, in which the sex ratio was 1:1, closely mirrored
the results of Sorenson et al. (1997). In that study the
T levels of chosen male northern pintails, A. acuta,
decreased following mate choice trials in which the sex
ratio was 3:1 (discussed more fully below). Thus, the
effects of competition on hormone levels found in northern pintails were similar to those observed in experiment
1 of the current study despite the extreme difference in
sex ratios between studies. Finally, all unpaired males in
experiment 2 were removed and housed separately from
the paired birds prior to the spring breeding season,
making the sex ratios and presumably the aggression
levels during the spring similar in both experiments when
the spring blood samples were taken.
Males with higher spring T levels were more successful
at mate guarding than males with lower T levels. Several
studies of passerine birds have found that male T level is
correlated with the amount of time males spend mate
guarding (white-crowned sparrows, Zonotrichia leucophrys:
Moore 1984; barn swallows, Hirundo rustica: Saino &
Møller 1995). However, the present study is the first
in which T levels were compared to a measure
of mate-guarding success. That T was associated with
mate-guarding success is not a surprising result, as mateguarding is an aggressive activity, and, as mentioned
above, aggression in reproductive birds is typically under
the influence of T (Etienne 1964; Balthazart & Stevens
1975; Schmedemann & Haase 1984; Wingfield & Moore
1987; Wingfield et al. 1990).
The data in the current study together suggest that
sexual selection may operate on male mate-guarding
behaviour in waterfowl. Indeed, it has been proposed that
the fitness benefits of mate guarding in migratory waterfowl may drive both the monogamous mating system and
early timing of pair formation in this group (Rohwer
& Anderson 1988; Oring & Sayler 1992). In addition,
Sorenson (1992) found that polygynous white-cheeked
pintails, A. bahamensis bahamensis, are more effective at
mate guarding than males that have only one mate.
Assuming that males that attract and secure multiple
mates indicate preferred males, her results suggest that
selection for mate-guarding behaviour may also operate
on sedentary waterfowl species.
It is also possible that some females in the present study
were better than others at avoiding FEPC attempts, and
that such females paired assortatively with males that had
high spring T levels. This interpretation still begs the
question of why females would choose high T males if
not for their aggressive defence. One possibility is that
females are choosing the indirect benefit of good genes.
However, this hypothesis appears unlikely in this species,
as Cunningham & Russell (2000) showed that differences
in offspring condition in this species could not be attributed to paternal effects, and Cunningham & Cheng
(1999) demonstrated no postinsemination female preference based on sperm genotype. Moreover, the females of
waterfowl species do not solicit extrapair copulations,
as might be expected if good genes were important
(Birkhead & Møller 1996).
Autumn T Levels and Pair Formation
No difference in autumn T levels was found between
paired and unpaired males in either year, although
autumn T levels were negatively correlated with the time
of day that blood was sampled in experiment 2. This
correlation was unlikely to be due to a stress response,
however, because there was no such association between
B and either time of day or sampling duration. Rather,
the observed correlation was more likely the result
of the natural diurnal fluctuation in T level in mallards
(Balthazart 1976; Balthazart & Hendrick 1979), in which
winter androgen levels peak in the morning and fall
steadily throughout the day.
The lack of association in experiment 2 between
autumn T level and pair status was unexpected, given
that several studies directly associated autumn and winter
T levels with female preference (Klint 1985; Klint et al.
1989), or indicated that the displays themselves may
be under the control of testosterone or its metabolite,
dihyrotestosterone (Phillips & McKinney 1962; Etienne
1964; Balthazart & Hendrick 1979; Schmedemann &
Haase 1985; Klint et al. 1989). There is also evidence that
at least one of the major courtship displays in mallards is
625
626
ANIMAL BEHAVIOUR, 64, 4
used agonistically among males (Davis 1997), indicating
that aspects of mallard courtship are aggressive. It may be
that female preference is a separate issue from actual pair
formation, but despite the lack of evidence in the current
study, T may still be a physiological link between courtship, pair formation and mate-guarding behaviour.
Regardless, mate guarding is an energetically costly
activity (Ashcroft 1976; Titman 1981; Lamprecht 1989),
and courtship activity is affected by male body condition
(Klint et al. 1989; Holmberg et al. 1989). Therefore,
courtship activity may be the best predictor females have
of both male condition and mate-guarding ability in the
spring.
Although the data in experiment 1 indicate that
behaviour such as copulation and mate guarding do not
cause elevated T levels in paired males during the breeding season, the process of pair formation evidently caused
a decrease in T during the autumn and winter courtship
season. This result supports the findings of Sorenson
et al. (1997), in which the T level of the closely related
northern pintail, (A. acuta), showed a significant decrease
within 2 h after being introduced to and chosen by a
female, whereas the T levels of unchosen males did
not show this decrease. Thus, upon being chosen as
mates, the T levels of males decreases very quickly and
remains lower than unpaired males for at least several
weeks.
Together, the results of this study and that of Sorenson
et al. (1997) suggest a previously unknown feedback
mechanism involving courtship and pair formation in
ducks. Although much work on this system is still necessary, the simplest scenario that best reconciles the literature is that: (1) androgens (T or DHT, or both) are
positively associated with male courtship display activity
in ducks; (2) females prefer males that court actively, and
thus, prefer males that are in good condition and have
relatively high androgen levels; (3) once a pair bond
begins to form, a feedback mechanism triggers a decrease
in androgen level in the male; and (4) the male decreases
his courtship activity in favour of pair formation activities. Sorenson et al. (1997) posited that the decrease in T
was owing to an increase in B levels caused by male–male
competition for the female. If so, it is not likely that
suppression is maintained by B as paired males in the
current study did not have significantly higher B levels
than unpaired males in January, although T levels were
lower in paired males. Such a feedback mechanism is a
classic example of behaviour affecting hormone physiology (see Lehrman 1965 for an early synthesis), but its
involvement in waterfowl courtship and pair formation is
new.
In summary, males that paired successfully during the
autumn had higher spring T levels than unpaired males.
Although breeding season activities probably did not
elevate T levels in paired males, initial pair formation
during the autumn and winter did lead to a temporary
decrease in T levels. Also, high T males were better at mate
guarding than low T males. These results together suggest
that females choose males that will have high T in the
spring because they will mate-guard more effectively than
low T males.
Acknowledgments
I thank my graduate committee for their patience, advice,
guidance and feedback over the years: Jeffrey Baylis,
Jack Hailman, Robert Jeanne, Robert Bleiweiss, Charles
Snowdon and Thomas Kurtz. I am also thoroughly
indebted to Catherine Marler and Lisa Sorenson their
help with methodology, and to David Abbott, Toni
Ziegler and Daniel Wittwer of the Wisconsin Regional
Primate Research Center Assay Services for recognizing
that mallards might be distantly related to primates.
Helpful comments on earlier versions of this manuscript
were made by J. Baylis, C. Marler, Janet Bester-Meredith,
Brian Trainor, Laurene Ratcliffe and two anonymous
referees. I was also fortunate to have a wonderful group of
undergraduate volunteers, who suffered through early
mornings in the cold Wisconsin winter. This work was
supported by grants from Sigma Xi, the Animal Behavior
Society and monies from the UW-Madison Zoology
Department.
References
Abraham, G. E., Buster, J. E., Lucas, L. A., Corrales, P. C. & Teller,
R. C. 1972. Chromatographic separation of steroid hormones for
use in radioimmunoassy. Analytical Letters, 5, 509–517.
Ankney, C. D. & MacInnes, C. D. 1978. Nutrient reserves and
reproductive performance of female lesser snow geese. Auk, 95,
459–471.
Ashcroft, R. E. 1976. A function of the pairbond in the common
eider. Wildfowl, 27, 101–105.
Balthazart, J. 1976. Daily variations of behavioural activities and of
plasma testosterone levels in the domestic duck. Journal of
Zoology, 180, 155–173.
Balthazart, J. & Hendrick, J. C. 1979. Relationships between the
daily variations of social behavior and plasma FSH, LH, and
testosterone levels in the domestic duck Anas platyrhynchos L.
Behavioural Processes, 4, 107–128.
Balthazart, J. & Stevens, M. 1975. Effects of testosterone propionate on the social behaviour of groups of male domestic ducklings
Anas platyrhynchos L. Animal Behaviour, 23, 926–931.
Barash, D. P. 1977. Sociobiology and Behavior. New York:
Elsevier.
Birkhead, T. R. & Møller, A. P. 1996. Monogamy and sperm
competition in birds. In: Partnerships in Birds: the Study of
Monogamy (Ed. by J. M. Black), pp. 323–343. Oxford: Oxford
University Press.
Bossema, I. & Kruijt, J. P. 1982. Male activity and female mate
acceptance in the mallard (Anas platyrhynchos). Behaviour, 79,
313–324.
Bossema, I. & Roemers, E. 1985. Mating strategy, including mate
choice, in mallards. Ardea, 73, 147–157.
Brodsky, L. M., Ankney, C. D. & Dennis, D. G. 1988. The influence
of male social dominance in black ducks and mallards. Animal
Behaviour, 36, 1371–1378.
Cunningham, E. J. A. & Cheng, K. M. 1999. Biases in sperm use in
the mallard: no evidence for selection by females based on sperm
genotype. Proceedings of the Royal Society of London, Series B, 266,
905–910.
Cunningham, E. J. A. & Russell, A. F. 2000. Egg investment is
influenced by male attractiveness in the mallard. Nature, 404,
74–77.
Davis, E. S. 1997. The Down-Up display of the mallard: one display,
two orientations. Animal Behaviour, 53, 1025–1034.
DAVIS: FEMALE CHOICE AND MATE GUARDING
Etienne, A. 1964. Der Einfluß von Testosteron auf das Verhalten
junger Stockerpel (Anas platyrhynchos L.) Zeitschrift für
Tierpsychologie, 21, 822–836.
Gauthier, G. 1987. The adaptive significance of territorial behavior
in breeding buffleheads: a test of three hypotheses. Animal
Behaviour, 35, 348–360.
Goodburn, S. F. 1984. Mate guarding in the mallard Anas platyrhynchos. Ornis Scandinavica, 15, 261–265.
Harding, C. 1983. Hormonal influences on avian aggressive
behavior. In: Hormones and Aggressive Behavior (Ed. by B. B.
Svare), pp. 435–476. New York: Plenum.
Harvey, S., Merry, B. J. & Phillips, J. G. 1980. Influence of stress on
the secretion of corticosterone in the duck (Anas platyrhynchos).
Journal of Endocrinology, 87, 161–171.
Hepp, G. R. & Hair, J. D. 1983. Reproductive behavior and
pairing chronology in winter dabbling ducks. Wilson Bulletin, 95,
675–682.
Hepp, G. R. & Hair, J. D. 1984. Dominance in wintering waterfowl (Anatini): effects on distribution of sexes. Condor, 86, 251–
257.
Hirschenhauser, K., Möstl, E., Wallner, B., Dittami, J. & Kotrschal,
K. 2000. Endocrine and behavioural responses of greylag geese
(Anser anser) to pairbond challenges during the reproductive
season. Ethology, 106, 63–77.
Holmberg, K., Edsman, L. & Klint, T. 1989. Female mate preferences and male attributes in mallard ducks, Anas platyrhynchos.
Animal Behaviour, 38, 1–7.
Johnsgard, P. A. 1961. Pair-formation mechanisms in Anas
(Anatidae) and related genera. Ibis, 102, 616–618.
Klint, T. 1980. Influence of male nuptial plumage on mate selection
in the female mallard (Anas platyrhynchos). Animal Behaviour, 28,
1230–1238.
Klint, T. 1985. Mate choice and male behaviour following castration
and replacement of testosterone in mallards (Anas platyrhynchos).
Behavioural Processes, 11, 419–424.
Klint, T., Edsman, K., Holmberg, K. & Silverin, B. 1989. Hormonal
correlates of male attractiveness during mate selection in the
mallard duck (Anas platyrhynchos). Hormones and Behavior, 23,
83–91.
Lamprecht, J. 1989. Mate guarding in geese: awaiting female
receptivity, protection of paternity or support of female feeding?
In: The Sociobiology of Sexual and Reproductive Strategies (Ed. by
A. E. Rasa, C. Vogel & E. Voland), pp. 48–60. New York: Chapman
& Hall.
Lehrman, D. S. 1965. Interaction between internal and external
environments in the regulation of the reproductive cycle of the
ring dove. In: Sex and Behavior (Ed. by F. A. Beach), pp. 355–380.
New York: J. Wiley.
Lorenz, K. 1953. Comparative studies on the behaviour of Anatinae.
(Reprinted from Avicultural Magazine, 57–59, 1–85). London:
Avicultural Society.
McKinney, F. 1988. Mate guarding and the confidence of paternity
in waterfowl. In: Acta XIX Congressus Internationalis Ornithologici.
Vol. 1 (Ed. by H. Oulette), pp. 411–419. Ottawa: University of
Ottawa Press.
McKinney, F. 1992. Courtship, pair-formation, and signal systems of
waterfowl. In: Ecology and Management of Breeding Waterfowl (Ed.
by B. D. J. Batt, A. D. Afton, C. D. Ankney, D. H. Johnson, J. D.
Kadlec & G. L. Krapu), pp. 214–250. Minneapolis: University of
Minnesota Press.
McKinney, F. & Evarts, S. 1998. Sexual coercion in waterfowl
and other birds. In: Avian Reproductive Tactics: Female and Male
Perspectives (Ed. by P. G. Parker & N. T. Burley), pp. 163–195.
Washington, D.C.: American Ornithologists’ Union.
McKinney, F., Derrickson, S. R. & Mineau, P. 1983. Forced
copulation in waterfowl. Behaviour, 86, 250–294.
Mineau, P. & Cooke, F. 1979. Rape in the lesser snow goose.
Behaviour, 70, 280–291.
Moore, M. C. 1984. Changes in territorial defense produced by
changes in circulating levels of testosterone: a possible hormonal
basis for mate-guarding in white-crowned sparrows. Behaviour,
88, 215–226.
Moore, M. C. 1986. Elevated testosterone levels during
nonbreeding-season territoriality in a fall-breeding lizard,
Sceloporus jarrovi. Journal of Comparative Physiology A, 158,
159–163.
Oring, L. & Sayler, R. D. 1992. The mating systems of waterfowl. In:
Ecology and Management of Breeding Waterfowl (Ed. by B. D. J.
Batt, A. D. Afton, C. D. Ankney, D. H. Johnson, J. D. Kadlec &
G. L. Krapu), pp. 190–213. Minneapolis: University of Minnesota
Press.
Pattenden, R. K. & Boag, D. A. 1989. Effects of body mass on
courtship, pairing, and reproduction in captive mallards. Canadian
Journal of Zoology, 67, 495–501.
Paulus, S. L. 1983. Dominance relations, resource use, and pairing
chronology of gadwalls in winter. Auk, 100, 947–952.
Phillips, R. E. & McKinney, F. 1962. The role of testosterone in the
displays of some ducks. Animal Behaviour, 10, 244–246.
Raveling, D. G. 1970. Dominance relationships and agonistic
behavior of Canada geese in winter. Behaviour, 37, 291–319.
Robinson, J. A., Scheffler, G., Eisele, S. & Goy, R. W. 1975. Effects
of age and season on sexual behavior and plasma testosterone
and dihydrotestosterone concentrations of laboratory-housed
male rhesus monkeys (Macaca mulatta). Biology of Reproduction,
13, 203–210.
Rohwer, F. C. & Anderson, M. G. 1988. Female-based philopatry,
monogamy, and the timing of pair formation in migratory waterfowl. In: Current Ornithology. Vol. 5 (Ed. by R. F. Johnston),
pp. 187–221. New York: Plenum.
Saino, N. & Møller, A. P. 1995. Testosterone correlations of mate
guarding, singing and aggressive behavior in male barn swallows,
Hirundo rustica. Animal Behaviour, 49, 465–472.
Schmedemann, R. & Haase, E. 1984. Zur Wirkung von Sexualhormonen auf das Verhalten kastrieter Stockerpel (Anas platyrhynchos
L.) (Anseriformes: Anatidae) II. Kampf und Fluchtverhalten.
Zoologischer Anzeiger, 213, 355–373.
Schmedemann, R. & Haase, E. 1985. Zur Wirkung von Sexualhormonen auf das Verhalten kastrieter Stockerpel (Anas platyrhynchos
L.) (Anseriformes: Anatidae) III. Sozial- und Sexualbalz.
Zoologischer Anzeiger, 214, 164–176.
Schutz, F. 1965. Sexuelle Prägung bei Anatiden. Zeitschrift für
Tierpsychologie, 22, 50–103.
Seymour, N. R. 1990. Forced copulation in sympatric American
black ducks and mallards in Nova Scotia. Canadian Journal of
Zoology, 68, 1691–1696.
Seymour, N. R. & Titman, R. D. 1978. Changes in activity patterns,
agonistic behavior, and territoriality of black ducks (Anas rubripes)
during the breeding season in a Nova Scotia tidal marsh. Canadian
Journal of Zoology, 56, 1773–1785.
Siegel, H. S. 1971. Adrenals, stress and the environment. Poultry
Science, 27, 327–349.
Siegel, H. S. 1980. Physiological stress in birds. BioScience, 30,
529–534.
Sorenson, L. G. 1992. Variable mating system of a sedentary
tropical duck: the white-cheeked pintail (Anas bahamensis
bahamensis). Auk, 109, 277–292.
Sorenson, L. G. 1994a. Forced extra-pair copulation in the whitecheeked pintail: male tactics and female responses. Condor, 96,
400–410.
Sorenson, L. G. 1994b. Forced extra-pair copulation and mate
guarding in the white-cheeked pintail: timing trade-offs in
an asynchronously breeding duck. Animal Behaviour, 48,
519–533.
627
628
ANIMAL BEHAVIOUR, 64, 4
Sorenson, L. G. & Derrickson, S. R. 1994. Sexual selection in the
northern pintail (Anas acuta): the importance of female choice
versus male–male competition in the evolution of sexuallyselected traits. Behavioral Ecology and Sociobiology, 35, 389–400.
Sorenson, L. G., Nolan, P. M., Brown, A. M., Derrickson, S. R. &
Monfort, S. L. 1997. Hormonal dynamics of aggression, dominance and courtship during mate choice in the northern pintail:
a test of the ‘challenge‘ hypothesis. Animal Behaviour, 54,
1117–1133.
Titman, R. D. 1981. A time-activity budget for breeding mallards
(Anas platyrhynchos) in Manitoba. Canadian Field Naturalist, 95,
266–271.
Titman, R. D. & Lowther, J. K. 1975. The breeding behavior of a
crowded population of mallards. Canadian Journal of Zoology, 53,
1270–1283.
Weidmann, U. 1956. Verhaltensstudien an der Stockente (Anas
platyrhynchos L.). I. Das Aktionssystem. Zeitschrift für
Tierpsychologie, 13, 208–271.
Weidmann, U. 1990. Plumage quality and mate choice in mallards
(Anas platyrhynchos). Behaviour, 115, 127–141.
Wingfield, J. C. 1994. Hormone-behavior interactions and mating
systems in male and female birds. In: The Differences between the
Sexes (Ed. by R. V. Short & E. Balaban), pp. 303–330. Cambridge:
Cambridge University Press.
Wingfield, J. C. & Moore, M. C. 1987. Hormonal, social and
environmental factors in the reproductive biology of free-living
male birds. In: Psychobiology of Reproductive Behavior: an Evolutionary Perspective (Ed. by D. Crews), pp. 148–175. Englewood Cliffs,
New Jersey: Prentice Hall.
Wingfield, J. C., Hegner, R. E, Duffy, A. M., Jr & Ball, G. F. 1990.
The ‘challenge hypothesis’: theoretical implications for patterns of
testosterone secretion, mating systems, and breeding strategies.
American Naturalist, 136, 829–846.
Wishart, R. A. 1983. Pairing chronology and mate selection in the
American wigeon (Anas americana). Canadian Journal of Zoology,
61, 1733–1743.