Animal Behaviour 133 (2017) 57e72

Contents lists available at ScienceDirect

Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav

Difference in arrival date at the breeding site between former pair

members predicts divorce in blue tits
Carol Gilsenan, Mihai Valcu, Bart Kempenaers*
Department of Behavioural Ecology and Evolutionary Genetics, Max Planck Institute for Ornithology, Seewiesen, Germany

a r t i c l e i n f o
Divorce occurs when both members of a breeding pair survive to the following year but then pair with
Article history: other individuals instead of reuniting. Divorce is common in birds, but its frequency can vary widely both
Received 8 April 2017 between and within species, or even between populations across years. Several explanations for divorce
Initial acceptance 10 May 2017 have been described, both adaptive and nonadaptive. Many studies have compared the breeding success
Final acceptance 4 August 2017 of faithful and divorced individuals, but fewer have considered the process of divorce, i.e. the events that
lead up to divorce. In this study, we used data from eight breeding seasons to investigate divorce in a
MS. number: 17-00309R population of blue tits, Cyanistes caeruleus, in southern Germany. To compare our results to previous
work, we first describe the frequency of divorce and compare the breeding success of divorced and
Keywords: faithful pairs. We then use data from an RFID transponder-based system, where all visits of individuals to
blue tit
any nestbox in the study site are automatically recorded throughout the year, to compare the behaviour
competition
of pairs in the interbreeding period. We found that the probability of divorce was not affected by
coordination
Cyanistes caeruleus
breeding success in Year X. However, divorce was predicted by the difference in arrival time to the study
dispersal site between the members of Year X pairs. Furthermore, during the interbreeding period, compared to
divorce their divorced counterparts, members of faithful pairs had more interactions with one another than with
extrapair paternity other individuals of the opposite sex. In Year X þ 1, faithful females started egg laying earlier, had
mate choice somewhat larger clutches and produced slightly more fledglings, than females that had divorced. We
mating behaviour propose that divorce in blue tits is a by-product of separation of the two pair members after the Year X
settlement breeding season, leading to asynchrony in the timing of settlement and pair formation in Year X þ 1.
© 2017 The Author(s). Published by Elsevier Ltd on behalf of The Association for the Study of Animal
Behaviour. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

Divorce, where one or both individuals from a previous breeding
pair form a new pair with another individual while their former
partner is still alive (Coulson, 1972), occurs in many bird species
(see review in Black, 1996). The probability of divorce depends not
only on individual life histories, but also on extrinsic factors such as
breeding site quality (e.g. Blondel, Perret, & Galan, 2000). Conse-
quently, the rate of divorce can vary widely between species (see
reviews in Rowley, 1983; Ens, Choudhury, & Black, 1996), and even
between populations of the same species (e.g. Dhondt &
Adriaensen, 1994).
In general, remaining faithful to a mate from one breeding
season to the next can provide several advantages to an individual
(see reviews in Rowley, 1983; Culina, Radersma, & Sheldon, 2014).
* Correspondence: B. Kempenaers, Department of Behavioural Ecology and
Evolutionary Genetics, Max Planck Institute for Ornithology, E. Gwinnerstr, 82319
Seewiesen, Germany.
E-mail address: b.kempenaers@orn.mpg.de (B. Kempenaers).
Reuniting or staying with a mate across multiple years means less
time and energy will be spent in finding a new mate and estab-
lishing a new pair bond (Bried & Jouventin, 2001), or in exploring
an unfamiliar territory (Black, 1996) if divorce coincides with
moving to a new breeding site. Newly formed pairs may also be less
efficient than faithful pairs in coordinating breeding and parental
behaviours (e.g. feeding offspring). Ultimately, mate fidelity may
lead to higher reproductive success. Indeed, it has been shown that
females of faithful pairs start laying earlier in the season (see re-
view in Rowley, 1983), and that faithful pairs fledge more young
than pairs containing divorced individuals (Dhondt, 2002;
Diamond, 1987).
If mate fidelity leads to higher fitness why, then, are cases of
divorce frequently observed in populations? One set of hypotheses
suggests that divorce is adaptive for at least one member of a pair;
individuals should divorce when the benefits outweigh the costs of
mate change (Choudhury, 1995). Evidence that divorce might be
adaptive comes from studies showing that at least one member of a
divorced pair improves its reproductive success (Dhondt &

http://dx.doi.org/10.1016/j.anbehav.2017.09.004
0003-3472/© 2017 The Author(s). Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
58 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
Adriaensen, 1994; Green, Krebs, & Cockburn, 2004; Linden, 1991;
Marzluff, Woolfenden, Fitzpatrick, & Balda, 1996; Orell, Rytko €nen,
& Koivula, 1994; Ramsay, Otter, Mennill, Ratcliffe, & Boag, 2000;
Streif & Rasa, 2001). The underlying reasons for improved
breeding success may be diverse, but can be summarized as
obtaining a higher quality territory (Desrochers & Magrath, 1996;
Blondel et al., 2000; the ‘habitat-mediated’ hypothesis, Newton &
Wyllie, 1996), a higher quality mate (where females may use
male quality indicator traits to make decisions about divorce, e.g.
body size or ornaments: Ryan, 1998; individual rank in a flock:
Otter & Ratcliffe, 1996) or a more compatible mate (the ‘in-
compatibility’ hypothesis, Coulson, 1966; Coulson & Thomas, 1980),
where ‘incompatibility’ can be genetic or behavioural (see Ihle,
Kempenaers, & Forstmeier, 2015). In general, this is referred to as
the ‘better option’ hypothesis (Ens, Safriel, & Harris, 1993). Thus,
divorce might be a mechanism to remedy the costs of having mated
with a genetically related mate (‘inbreeding avoidance’;
Kempenaers, Adriaensen, & Dhondt, 1998; Hatchwell, Russell, Ross,
& Fowlie, 2000), a less fertile or infertile mate (see Hasson & Stone,
2009 for a mathematical model that predicts an increase in female
infidelity as a response to certain aspects of male infertility) or an
unfaithful mate (higher risk of paternity loss for a male; Ce zilly &
Nager, 1995).
The observation that one or both former pair members increase
their fitness after a divorce can erroneously be taken as evidence
that divorce is an active behavioural ‘decision’ of at least one pair
member. Divorce might also be the consequence of other processes
that do not involve a decision to leave the previous partner. For
example, strong intrasexual conflict for access to a potential mate
can lead to a new male or female taking over a territory/mate, with
passive acceptance of the usurper as the new breeding partner
(Daniels & Walters, 2000; Heg, Bruinzeel, & Ens, 2003; Taborsky &
Taborsky, 1999; Valcu & Kempenaers, 2008). Divorce may also
occur because of timing differences in arrival and settlement of
previous pair members (‘musical chairs’ hypothesis, Dhondt &
Adriaensen, 1994) simply because of separation during the
nonbreeding season (‘accidental loss’ hypothesis, Owen, Black, &
Liber, 1988). An individual may then decide to pair with a new
partner rather than wait, possibly in vain, for their previous partner
to return (Gonzalez-Solis, Becker, & Wendeln, 1999). In these con-
texts, divorce could even be maladaptive for both members of the
previous pair, that is, both could have a lower breeding success after
the separation.
Most studies on divorce in birds have focused on the differences
in reproductive success or on between-year changes in fitness
measures between faithful and divorced pairs (see meta-analyses
in Dubois & Ce zilly, 2002; Culina et al., 2014). However, as
explained above, the results of such studies are not sufficient to
conclude that divorce is an adaptive mating tactic. The specific
behaviours of individuals in between breeding attempts that may
ultimately lead to divorce remain relatively poorly studied, even
though this is necessary to understand how and why divorce occurs
in a population.
We studied divorce in a nestbox-breeding population of blue
tits, Cyanistes caeruleus, in southern Germany. The blue tit is a small
secondary cavity-nesting passerine species that is common
throughout most of Europe. Most blue tits are socially monoga-
mous with biparental care and produce one clutch per pair per
breeding season (Perrins, 1979). The proportion of pairs that
divorce ranges from 8% to 85% between populations (Dhondt &
Adriaensen, 1994). The general aim of our study was to investi-
gate whether divorce is an active choice (a mating tactic) employed
by at least one pair member or whether it is a passive outcome of
intrasexual competition or behaviour during the nonbreeding
season. First, we investigated whether aspects of a pair's
reproductive success (lay date, clutch size and fledging success) in
Year X can predict divorce in Year X þ 1. Second, we tested whether
reproductive success differs between males and females from
faithful pairs and those that divorced (Year X þ 1), and we compare
our results with those from the literature. Third, we analyse
spatiotemporal patterns of the presence of individuals at nestboxes
between winter and the start of breeding to gain insight into the
behavioural mechanisms behind divorce. A previous study (Valcu &
Kempenaers, 2008) suggested that divorce may be the consequence
of female - female competition. If intrasexual conflict over a mate or
breeding territory is the main cause of divorce, we expected to see
visits to the future breeding nestbox by at least two competing
same-sex individuals. We also report when both members of a new
or faithful pair first visited the future breeding nestbox (or any
nestbox), providing insight into the timing of territory establish-
ment or pair formation.
METHODS
Study Site
This research was carried out as part of a long-term study on the
reproductive biology of a blue tit population in a protected forest
site in southern Germany. The study area (‘Westerholz’,
48 080 2600 N, 10 530 2900 E) is a 40 ha mixed deciduous/coniferous
forest equipped with 277 nestboxes since 2007 to monitor breeding
(converted to ‘smart’ nestboxes, hereafter ‘SNBs’, in 2009; see
Schlicht, Girg, Loe €s, Valcu, & Kempenaers, 2012 for more details).
Each year, the breeding population consists of between 60 and 100
blue tit pairs.
Field Procedures
On four occasions during each winter (in the first week of
November, December, January and February), roosting individuals
were caught inside the nestboxes at night. During the breeding
season, adults were caught in the nestbox, using an automated
trapping system or by hand, when they fed 8e10-day-old nestlings.
Unbanded birds were ringed with a unique combination of one
metal band and three plastic colour bands, and a blood sample (ca.
50 ml) from the brachial vein was taken for molecular sexing and
parentage analysis. Each individual was measured (tarsus and third
primary wing feather ± 0.5 mm, weight ± 0.1 g), and age (yearling
or adult) and sex were determined based on plumage characteris-
tics. New birds were also implanted with a passive integrated
transponder, which allowed them to be uniquely identified by the
SNBs (see Schlicht et al., 2012 and below for more details).
During the breeding season (from early March until early June),
each nestbox was visited at least weekly to check for presence of
nesting material. When a nest had reached a ‘lining’ stage (i.e. a
completed nest lined with feathers or other soft material), it was
checked daily to determine the date the first egg was laid (hereafter
‘lay date’). Weekly checks thereafter allowed us to determine clutch
size. Towards the end of incubation, around 2 weeks after clutch
completion, nestboxes were checked daily again to determine
hatch date. The number of young in each nest were counted at least
every week until fledging.
Smart Nestbox Data Collection
Each nestbox was fitted with a transponder reader and with two
light barriers, one at the inside, and one at the outside of the
entrance hole (see Schlicht et al., 2012 for technical details), such
that all entries and exits of birds (with transponder number, if
available) were recorded with date and time (based on real-time
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
clocks, recalibrated if necessary to counter time drift) throughout
the year. The system also detects individuals (by transponder ID)
when they are perched at the nest entrance without entering the
box. Visit data from each SNB were stored on a 2GB SD card, which
was exchanged manually at least once every month.
Data Analysis
For data analysis, we used the free software R (versions 3.1.2 and
later; R Development Core Team, 2014). First, we determined the
breeding status of each pair (that bred in Year X) where both
members (hereafter ‘focal individual(s)’) had survived to breed in
the subsequent year (Year X þ 1), using data collected from 2007 to
2015. Specifically, we determined how many pairs reunited (here-
after ‘faithful pair’), and how many pairs consisted of individuals
that bred with a new partner (hereafter ‘divorced pair or individ-
ual’). When comparing reproductive success between groups, var-
iables were scaled by subtracting the mean value for each year.
Male and female age were included in models as a factor with two
levels (yearling or adult). Plots of the residuals were used to check
the fits of the models, and data were transformed or models sup-
plemented with optimizers where appropriate to allow conver-
gence to be reached. P values were obtained using the package
‘lmerTest’ (Kuznetsova, Brockhoff, & Christensen, 2015).
Individuals involved in socially polygynous matings (N ¼ 22)
and data from replacement clutches (N ¼ 10) were excluded. We
also excluded two divorced pairs for which no information about
the females breeding in Year X þ 1 was available. Overall, we
include breeding data (from 2007 to 2015 inclusive) from 121 pairs:
77 divorced and 44 faithful.
The SNBs did not record nestbox visits of individuals until mid-
2009; thus, for the analysis of behaviour, we only included pairs
that first bred together from 2009 onwards (faithful: N ¼ 29;
divorced: N ¼ 60). From the pairs that remained, 10 focal in-
dividuals lost their transponder between Year X and Year X þ 1, and
were thus not detectable by the SNBs for either part or all of the
study period. We excluded pairs containing these individuals from
the analysis of behaviour (divorced N ¼ 6, faithful N ¼ 3). For four
Year X pairs (divorced N ¼ 3, faithful N ¼ 1) the Year X þ 1 breeding
box of at least one pair member did not record data, and thus these
pairs were excluded from the analyses. For one faithful female no
data were recorded, even though the Year X þ 1 breeding box was
working (male data were available), and thus we excluded this pair
from the analyses as well. This left data on 51 divorced and 24
faithful pairs. For the analyses that explored interactions of focal
individuals with other individuals during the nonbreeding period,
we also excluded individuals without a transponder (those that lost
their transponder and those that were caught for the first time
during this period).
Analysis of potential causes of divorce (Year X)
We investigated whether divorce rates were related to between-
year fluctuations in breeding pair number in the study site, which
may indicate variation in the level of competition and/or number of
potential mates available. To do this, we correlated the proportion
of divorced pairs (relative to the total number of focal pairs) in Year
X þ 1 with the change in the number of breeding pairs from Year X
to Year X þ 1 (expressed as a ratio: number of breeding pairs in Year
X þ 1 divided by the number of breeding pairs in Year X; thus,
values higher/lower than 1 represent an increase/decrease in the
breeding population). We predicted that when the number of
breeding pairs had increased between years, pairs would be more
likely to divorce, either (1) because individuals would have a larger
pool of potential partners to sample from (assuming active choice
by one of the pair members to ‘trade up’), or (2) because of
59
increased competition from potential usurpers. In contrast, when
the number of pairs had decreased, the probability of divorce would
be lower.
To test whether breeding success or individual age in one year
could predict divorce in the following breeding season, we used a
logistic regression (package ‘lme4’, Bates, Maechler, Bolker, &
Walker, 2014) with future status (coded as ‘0’ for faithful pairs
and ‘1’ for divorced pairs) as the response variable, and with lay
date, clutch size, fledging success, extrapair paternity (yes/no), and
male and female age as explanatory variables. Male and female
identity and year were included as random factors.
Analysis of consequences of divorce (Year X þ 1)
To evaluate the consequences of divorce, we first tested whether
breeding success in Year X þ 1 differed depending on pair status.
We compared faithful pairs (N ¼ 44) with either ‘divorced female’
pairs (N ¼ 77) or with ‘divorced male’ pairs (N ¼ 77), using linear
mixed models with lay date, clutch size, fledging success or the
proportion of extrapair young as the response variable, and pair
category (faithful versus divorced) as the explanatory variable. In-
dividual identity and year were included as random factors.
Because divorced individuals may have either a yearling or an adult
individual as a new breeding partner, and because partner age may
influence reproductive success, we also ran the analyses for each
sex using status as a factor with three levels (faithful pairs, divorced
individuals with a yearling new mate and divorced individuals with
an adult new mate).
We further investigated whether the presence of extrapair
young in Year X þ 1 (yes/no) differed between faithful and divorced
females, using a chi-square test. We also report instances where the
previous mate of a divorced female sired extrapair young in her
Year X þ 1 clutch, and we noted how many young the male had
sired.
Finally, we compared the change in breeding success from Year
X to Year X þ 1 for faithful and divorced individuals. We subtracted
the standardized values for each breeding parameter for Year X
from those of Year X þ 1 and used paired t tests to assess how
breeding success changed for individuals of different status.
Analysis of processes possibly leading to divorce
To study the processes that may have led to divorce, we used
visits to any nestbox as a proxy for spatiotemporal presence at the
breeding site, and compared individuals that divorced the
following breeding season with those that stayed faithful to their
partner. We defined the ‘interbreeding period’ for each individual
as the time from the end of breeding in Year X (generally defined as
1 August of Year X; in July only one visit to a nestbox had been
registered) to the beginning of breeding in Year X þ 1 (defined as 7
days before the first egg of a pair in Year X þ 1).
First, we tested for each sex separately whether divorced and
faithful individuals differed in their presence at the breeding site (at
any nestbox) during the interbreeding period, using a Fisher's exact
test for small sample sizes.
Second, for those Year X pairs where both individuals were
present at the breeding site during the interbreeding period, we
tested whether the differences in arrival date at the breeding site
between the two individuals (estimated based on the first visit of
each individual to any nestbox in the study area) predicted divorce.
We used a logistic regression model with pair status as a binary
response variable (divorced ¼ 1, faithful ¼ 0), and difference in
arrival date (in days) between pair members as the explanatory
variable.
Third, we investigated whether future divorce was related to
interactions with individuals other than the Year X partner during
the interbreeding period. We defined an ‘interaction’ as two
60 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72

individuals visiting the same nestbox within 15 min. Any individual with pair (Year X) and individual identity as random factors). For
interacting with a focal individual was considered either a ‘poten- both parts, we used all interacting future mates regardless of
tial competitor’ (if the same sex as the focal individual) or a ‘po- whether they carried a transponder throughout the interbreeding
tential mate’ (if the opposite sex to the focal individual). We period, because all individuals that interacted at least once with the
categorized the individuals that had mated (Year X) or ended up focal individual could be included, irrespective of the number of
mating (Year X þ 1) with focal individuals into ‘past mate’ or ‘future interactions.
mate’, respectively (for faithful pairs, past and future mate is, by Finally, we investigated postbreeding dispersal of focal in-
definition, the same individual). We then fitted three models to dividuals as an indication of who decided to move or was forced to
explore whether and how interactions of focal individuals are move to a new nest site or territory after the divorce. To this end, we
related to future divorce. used a linear mixed model (REML) with the distance between the
Model 1 is a logistic regression with a binary response variable Year X and Year X þ 1 breeding box (ln-transformed) as the
indicating whether a focal individual had interacted with its past response variable, and pair status (divorced female, divorced male,
mate (a score of 1) or with any other individual of the opposite sex faithful pair) as the explanatory variable. Individual identity and
(a score of 0), and pairing status and month, and their interaction, study year were included as random factors.
as explanatory variables. We ran this model for each sex separately
to test whether faithful and divorced individuals differed in how Ethical Note
frequently they interacted with their past mate, relative to the in-
teractions they had with members of the opposite sex. This could Adults were captured in a nestbox, either at night while roosting
indicate whether and when a male or female chooses to spend time or during the day (before 1500 hours) while feeding nestlings.
with its former mate or with other potential mates, or it could Caught individuals were placed in bird bags and immediately
indicate the temporal availability of other mating options. For transferred to a mobile field laboratory nearby for processing. In
divorced individuals, we also investigated how the interactions general, banding, measuring and equipping a bird with a tran-
with their future mate developed over the season, and tested sponder took less than 10 min. Blood sampling and/or transponder
whether this differed between males and females. Model 2 is a implantation did not seem to have any adverse effects on the birds.
logistic regression with a binary response variable indicating The birds were either placed back into the nestbox from which they
whether a focal individual had interacted with its future mate (a had been taken (if trapped at night), or released in their territory
score of 1) or with any other individual of the opposite sex (a score close to their breeding box (if trapped while feeding nestlings). For
of 0), and with status and month, and their interaction, as an in-depth analysis and discussion of the consequences of
explanatory variables. Model 3 is a Poisson regression with the catching during feeding, see Schlicht and Kempenaers (2015).
number of interactions a focal individual had with potential com- Permits were obtained from the Bavarian government and the
petitors (any member of the same sex) as the response variable, and Bavarian regional office for forestry (LWF).
with period (month) and pairing status as explanatory variables. If
divorce is the outcome of intrasexual competition, we expected RESULTS
that the number of interactions with competitors during the
interbreeding period would differ between faithful and divorced Frequency of Divorce
pairs. In all models, focal individual identity and breeding year were
included as random factors, and the month of interaction was re- Across all years, 64% of blue tit pairs divorced (Table 1). The
centred to have December as 0. percentage of pairs divorcing varied between 50% and 80%, but did
For each focal individual present at the breeding site during the not differ significantly between years (chi-square test with simu-
interbreeding period, we also graphically visualized all interactions lated P values for small sample sizes: c2 ¼ 5.91, P ¼ 0.56; Table 1).
(with past, future and potential mates, and with potential com- The change in the total number of breeding pairs from Year X to
petitors) over time (faithful pairs: N ¼ 25, divorced pairs: N ¼ 52, Year X þ 1 (expressed as a ratio, see Methods) did not correlate
see graphs in Fig. S1 in the Supplementary material). Based on with the number of focal pairs that divorced in Year X þ 1 (Pearson
these graphs, we defined ‘early’ divorce as those cases where the correlation: r6 ¼ 0.12, P ¼ 0.79), suggesting that the frequency of
members of a Year X pair no longer interacted during the inter- divorce was not related to the number of potential new partners.
breeding period. We defined ‘late’ divorce as those cases where
Year X pair members interacted at least once during the inter-
breeding period. For ‘late’ divorce, we noted when each member of Table 1
a Year X pair arrived at the study area, when they interacted with Frequency of divorce in a blue tit population in southern Germany
other potential mates, including their future mate, and when they Year Xþ1 Faithful Divorced Proportion No. of breeding
interacted with potential competitors. We concluded that there pairs pairs divorceda pairsb
was competition for a breeding site (nestbox) or for access to a 2007 88
potential mate whenever a member of a pair from Year X interacted 2008 9 11 0.55 96
with a same-sex individual that had also interacted with that 2009 6 6 0.50 61
individual. 2010 6 10 0.63 102
2011 4 7 0.64 79
Fourth, we investigated whether divorce could be the outcome 2012 10 12 0.55 149
of direct competition between same-sex individuals, assuming that 2013 4 11 0.73 99
body size (measured as tarsus length) reflects the likelihood to win 2014 3 12 0.80 79
in competitive interactions. Thus, for each sex separately, we 2015 2 8 0.80 112
Total 44 77 0.64
compared (1) tarsus length of the Year X and Year X þ 1 mate of
divorced individuals (paired t test), and (2) tarsus length of inter- Data were included only if both members of a pair that bred in Year X were known
acting individuals of the same sex that did or did not end up as the to breed again in Year X þ 1.
a
Proportion of pairs that divorced from the total number of focal pairs (faithful
new mate (generalized linear mixed model with mating success and divorced) that year.
(0 ¼ no mating, 1 ¼ new mate) as the response variable, with tarsus b
Total number of breeding pairs in the field site each year, excluding second
length relative to the Year X mate as the explanatory variable, and (replacement) breeding attempts.
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72 61

In total, 17 females and 21 males bred in at least 3 years. Of these, Table 3

six females (35%) and 10 males (48%) always divorced their Year X Predictors of divorce in a blue tit population in southern Germany from 2007 to 2015

partner (Table 2). We only found one instance where a faithful pair Parameter estimate SE z P
(Year X þ 1) divorced by Year X þ 2 (\10 and _4 in Table 2). All other Intercept 1.340 0.520
faithful pairs bred together until one of the individuals disappeared Lay datea 0.046 0.059 0.780 0.44
from the population (presumably died). Only one pair in our pop- Clutch sizea 0.039 0.170 0.230 0.82
ulation bred together across three breeding seasons (\6 and _1). Fledging successa,b 0.005 0.130 0.039 0.97
Female agec 0.330 0.420 0.780 0.43
Male agec 1.039 0.450 2.330 0.02
Extrapair paternityd 0.120 0.440 0.270 0.79
Potential Causes of Divorce
Logistic regression with divorce (yes/no) as dependent variable, measures of
The occurrence of complete brood failure in Year X was some- breeding success (standardized per year), and male and female age in Year X as
predictors.
what more common in future divorced pairs (10%, 8/77 pairs) than a
Estimates show the change in the odds of divorce for every unit change in the
in future faithful pairs (5%, 2/44 pairs), but this difference was not breeding parameter.
significant (Fisher's exact test: odds ratio ¼ 0.41, P ¼ 0.32). The b
The number of chicks successfully fledged (excluding cases of complete brood
probability that a pair would divorce in Year X þ 1 did not depend failure).
c
Estimate is for adults compared to yearlings.
on measures of breeding success in Year X (Table 3), nor did it d
Estimate is for clutches containing at least one extrapair young compared to
depend on female age in Year X (69%, 42/61 of yearling females those containing no extrapair young.
divorced; while 57%, 35/60 of adult females divorced). However,
yearling males were significantly more likely to divorce (78%, 39/ divorced pairs; 58%, 18/31 future faithful pairs), nor on the pro-
50) than adult males (54%, 38/71). This might suggest that females portion of extrapair young in the clutch (Appendix Table A2).
preferred older males. However, by default faithful females would
have ended up with an adult male in year X þ 1, whereas 26% of the Consequences of Divorce and Changes in Breeding Success
females that divorced a yearling ended up with another yearling
partner in Year X þ 1 (Appendix Table A1). Finally, the probability In Year X þ 1, females paired with divorced males started to lay
that a pair would divorce in Year X þ 1 did not depend on the about 2 days later than faithful females, and laid on average half an
occurrence of extrapair young in their clutch (53%, 32/60 future egg less (Table 4). This was true, independent of whether the new

Table 2
Number of individuals that bred in at least 3 years, and their pairing status or fate

Year Xþ1: 2008 2009 2010 2011 2012 2013 2014 2015

\1 Divorce Divorce y e e e e e
\2 Divorce Faithful y e e e e e
\3 Divorce Widow Widow Faithful y e e e
\4 Divorce Faithful Widow y e e e e
\5 Faithful Widow Divorce Widow y e e e
\6 Faithful Faithful Widow Divorce y e e e
\7 Divorce Divorce Divorce y e e e e
\8 e e e Divorce Faithful Widow y e
\9 e e e e e Divorce Divorce Widow
\10 Faithful Divorce y e e e e e
\11 e e Divorce Divorce Faithful Widow Divorce y
\12 e e Divorce Widow Divorce Widow y e
\13 e e Divorce Faithful y e e e
\14 e e e Divorce Divorce Divorce X y
\15 e e e Divorce Faithful Widow y e
\16 e e e e e Divorce Faithful Widow
\17 e e e e e e Divorce Divorce
_1 Faithful Faithful y e e e e e
_2 Divorce Divorce y e e e e e
_3 Widow Widow Widow Divorce Faithful Widow y e
_4 Faithful Divorce Divorce Faithful y e e e
_5 Faithful Widow Divorce Divorce y e e e
_6 Divorce Faithful Widow X Widow Widow y e
_7 e Divorce Widow Divorce y e e e
_8 e e e Widow Divorce Faithful y e
_9 e e e e e Faithful Widow Divorce
_10 e e e e e Divorce Faithful y
_11 e e e e Divorce Widow Widow Divorce
_12 e e e Divorce Faithful y e e
_13 e e e Widow Divorce Divorce Faithful y
_14 e e e Divorce Divorce Widow y e
_15 e e e e Divorce Divorce y e
_16 e e e e e e Divorce Divorce
_17 e e e e Divorce Divorce y e
_18 e e e e Widow Divorce Divorce y
_19 e e e e e Divorce Divorce Divorce
_20 e e e e e Divorce Widow Faithful
_21 e e e e e Divorce Divorce y

‘y’ indicates that an individual was no longer found breeding in our population (presumably died). ‘Widow’ indicates that an individual bred in Year X þ 1, while its Year X mate
was no longer found breeding in our population (presumably died). ‘X’ indicates that an individual's breeding status in that year was unknown.
62 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72

Table 4 Table 5
Differences in measures of breeding success (standardized within year) between Differences in measures of breeding success (standardized within year) between
faithful and divorced males faithful and divorced females

Breeding Explanatory variable Parameter SE t P Breeding Explanatory variable Parameter SE t P

parameter estimate parameter estimate

Lay date Intercepta 3.04 0.58 Lay date Intercepta 3.12 0.58
Pair statusb 1.87 0.71 2.65 0.008 Pair statusb 1.09 0.68 1.61 0.11
Intercepta 3.059 0.58 Intercepta 3.140 0.58
New mate: yearlingc 2.300 0.94 2.46 0.014 Mate age: yearlingc 1.320 0.95 1.40 0.16
New mate: adultc 1.69 0.76 2.24 0.025 Mate age: adultc 1.012 0.71 1.42 0.16
Clutch size Intercepta 0.86 0.20 Clutch size Intercepta 0.88 0.21
Pair statusb 0.55 0.24 2.25 0.024 Pair statusb 0.53 0.25 2.12 0.034
Intercepta 0.86 0.20 Intercepta 0.87 0.21
New mate: yearlingc 0.49 0.32 1.53 0.130 Mate age: yearlingc 0.35 0.35 1.001 0.320
New mate: adultc 0.57 0.26 2.19 0.028 Mate age: adultc 0.60 0.27 2.260 0.024
Fledging Intercepta 2.33 0.32 Fledging Intercepta 2.34 0.35
successd Pair statusb 0.74 0.38 1.93 0.053 successd Pair statusb 0.68 0.37 1.84 0.066
Intercepta 2.33 0.32 Intercepta 2.34 0.35
New mate: yearlingc 0.92 0.50 1.86 0.063 Mate age: yearlingc 0.59 0.51 1.14 0.250
New mate: adultc 0.65 0.41 1.59 0.110 Mate age: adultc 0.71 0.40 1.82 0.068
Extrapair Intercepta 0.073 0.019 Extrapair Intercepta 0.078 0.018
paternitye Pair statusb 0.015 0.023 0.69 0.49 paternitye Pair statusb 0.004 0.020 0.19 0.85
Intercepta 0.072 0.018 Intercepta 0.082 0.020
New mate: yearlingc 0.059 0.029 2.044 0.041 Mate age: yearlingc 0.020 0.032 0.65 0.52
New mate: adultc 0.007 0.023 0.290 0.770 Mate age: adultc 0.003 0.022 0.12 0.90

Shown are the results of linear mixed-model analysis (ML) with pair status as the Shown are the results of linear mixed-model analysis (ML) with pair status as the
explanatory variable. In the first model, pair status has two categories (faithful/ explanatory variable. In the first model, pair status has two categories (faithful/
divorced). In the second model, age of the new mate is included: faithful males are divorced). In the second model, age of the new mate is included: faithful females are
compared to divorced males paired with either a yearling or an adult female (i.e. pair compared to divorced females paired with either a yearling or an adult male (i.e. pair
status has three categories). status has three categories).
a a
Estimate for faithful individuals. Estimate for faithful individuals.
b b
Estimate for divorced males (N ¼ 77) relative to faithful males (N ¼ 44). Estimate for divorced females (N ¼ 77) relative to faithful females (N ¼ 44).
c c
Estimate for divorced males and their Year X þ 1 mates (N ¼ 25 yearling fe- Estimate for divorced females and their Year X þ 1 mates (N ¼ 24 yearling
males, N ¼ 52 adult females), relative to faithful males (N ¼ 44 per definition all males, N ¼ 53 adult males), relative to faithful females (N ¼ 44 per definition all
adult females). adult males).
d d
The number of chicks successfully fledged (excluding complete brood failures; The number of chicks successfully fledged (excluding complete brood failures;
faithful pairs 16% (7/44), divorced males 8% (6/77), c21 ¼ 1.92, P ¼ 0.17). faithful pairs 16% (7/44), divorced females 13% (10/77), c21 ¼ 0.2, P ¼ 0.66).
e e
The proportion of extrapair young in the clutch. The proportion of extrapair young in the clutch.

partner of the divorced male was a yearling or an adult female,
although the clutch size effect was no longer significant for yearling
females (Table 4). Pairs containing divorced males also produced
0.7 fewer fledglings than faithful pairs, but this was marginally
nonsignificant (Table 4). Clutches of divorced males contained a
higher proportion of extrapair young than those of faithful pairs,
but only if the new female was a yearling (Table 4).
Divorced females started to lay around 1 day later than
faithful females, but this was not significant, independent of the
age of the new partner (Table 5). Divorced females also laid
around half an egg less than faithful females, which was statis-
tically significant overall and for females that remated with an
adult male, but not for females that remated with a yearling
(Table 5). Pairs containing divorced females produced 0.7 fewer
fledglings, an effect that was marginally nonsignificant overall
and for females that remated with an adult male (Table 5).
Clutches of divorced and faithful females did not differ in the
proportion of extrapair young.
We also compared the Year X þ 1 breeding success of pairs
containing divorced individuals with that of pairs containing wid-
owed individuals (from the same years 2007e2015; widowed
males N ¼ 159; widowed females N ¼ 155). Divorced and widowed
individuals did not differ in any of the measures of breeding success
(Appendix Table A3).
Both divorced females and divorced males were significantly
more likely to mate with adults than with yearlings in Year X þ 1
(69% of adults for divorced females, chi-square test: c21 ¼ 10.92,
P < 0.001; 68% of adults for divorced males, chi-square test:
c21 ¼ 9.47, P ¼ 0.002; Appendix Table A1). This was not because
there were more adults in the population; the proportion of adult
individuals that mated with divorced individuals was significantly
larger than the total proportion of adult individuals in the
breeding population (population-wide 51% adult males, divorced
females: c21 ¼ 9.94, P ¼ 0.002; population-wide 48% adult fe-
males, divorced males: c21 ¼ 12.31, P < 0.001). Most of the adult
individuals that mated with divorced males and females in Year
X þ 1 had previous breeding experience in the field site (females
of divorced males: 76%, 40/52; males of divorced females: 71%, 38/
53). Widowed individuals were also more likely to mate with
adults in Year X þ 1 (Appendix Table A1), and did not differ
significantly from divorced individuals in this respect (66% of
adults for widowed females, chi-square test comparing divorced
and widowed females: c21 ¼ 0.21, P ¼ 0.65; 57% of adults for
widowed males, chi-square test comparing divorced and widowed
males: c21 ¼ 2.04, P ¼ 0.15).
The likelihood of extrapair paternity (at least one egg sired by an
extrapair male) in Year X þ 1 clutches did not differ between
faithful females and divorced females (c21 ¼ 0.77, P ¼ 0.38; faithful
pairs 18/42, 43%; divorced female pairs 26/75, 35%), or between
faithful females and females mated with divorced males
(c21 ¼ 0.091, P ¼ 0.76; divorced male pairs 30/75, 40%). Interest-
ingly, there were four instances (5% of divorced females) where a
divorced female's past mate sired extrapair young in her Year X þ 1
clutch (3/10 eggs, 2/8, 2/11 and 3/11; all other eggs were sired by
the within-pair male, except for the last clutch where another
extrapair male sired one egg). In all but the last case, the individuals
had divorced early, and in two of the four cases, the former partners
nested in neighbouring nestboxes.
There were no significant changes in breeding success from Year
X to Year X þ 1 for individuals that divorced, but faithful pairs
significantly advanced their lay date, and laid somewhat larger
clutches (Fig. 1).
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
Lay date
Clutch size
Faithful
Divorced females
Divorced males
Fledging success
–4 –3
Year X+1 - Year X (Z scores)
Figure 1. Changes in standardized breeding parameters from Year X to Year X þ 1 for faithful pairs, divorced females and divorced males. The dotted vertical line indicates no
change. Shown are means with 95% confidence intervals and results of paired t tests. For fledging success, complete brood failures either in Year X or in Year X þ 1 are excluded.
Processes Leading to Divorce
Of those individuals that ended up divorced (51 pairs), one male
(2%) and one female (2%) were absent from the breeding site during
the entire interbreeding period. In comparison, one male (4%) that
was in a faithful pair (N ¼ 24) was absent from the breeding site
during the entire interbreeding period. These individuals were first
detected at their breeding box less than 7 days before the appear-
ance of the first egg of their clutch. The likelihood of presence at the
breeding site during the interbreeding period did not differ be-
tween the sexes (Fisher's exact test P ¼ 0.62), and did not depend
on pairing status (P ¼ 1.0).
For those pairs where both individuals were present at the study
area during the interbreeding period, the probability of divorce
increased as the difference in arrival time between the two in-
dividuals increased (Fig. 2a).
In nine of 23 (39%) faithful pairs, both members arrived at the
breeding site within 3 days of one another (Fig. 2c), whereas this
was only true for three of 44 (7%) divorced pairs (Fig. 2b; chi-square
test: c21 ¼ 10.73, P ¼ 0.001). In eight cases (N ¼ 7 faithful pairs;
N ¼ 1 divorced pair), both pair members arrived on the same day.
Males usually arrived at the breeding site before females, both for
faithful pairs (81% of the 16 cases where the individuals did not
arrive on the same day) and for divorced pairs (72% of the 43 cases
where the individuals did not arrive on the same day). In 13 of 16
(81%) cases where the male arrived after their mate, the pair ended
up divorced, even if the female visited the former breeding box
before the male had arrived (3/13 cases, 23%).
The process of divorce and remating is reflected in the
observed interactions: faithful individuals were more likely to
have a higher proportion of interactions with their Year X mate
(relative to interactions with other individuals of the opposite sex)
as the interbreeding period progressed than divorced individuals
(Fig. 3a, b; also see Appendix Tables A4 and A5). Faithful males
were more likely to interact with their Year X þ 1 mate (relative to
interactions with other individuals of the opposite sex) during the
interbreeding period than divorced males (Fig. 3d), and the
63
t43 = 2.48, P = 0.017
t75 = 0.73, P = 0.47
t75 = –0.99, P = 0.33
t43 = –1.78, P = 0.082
t76 = –0.22, P = 0.83
t76 = –0.16, P = 0.87
t35 = –1.46, P = 0.15
t60 = 0.79, P = 0.43
t62 = 0.20, P = 0.85
–2 –1 0 1 2
likelihood of such interactions increased significantly among all
males as the season progressed. In divorced males, the proportion
of interactions with their Year X þ 1 mate increased more over the
interbreeding period than in faithful males (Appendix Table A4),
suggesting that the re-pairing process took place during this time.
During the interbreeding period, faithful females interacted more
with their Year X þ 1 mate (relative to interactions with other
individuals of the opposite sex) than divorced females, although
this was not significant (Table A5). The proportion of interactions
with the Year X þ 1 mate increased significantly over the season,
similarly for faithful and divorced females (Table A5). Overall,
compared to divorced females, divorced males interacted rela-
tively more with other potential mates than with their new (Year
X þ 1) mate (Fig. 3c, d, Appendix Table A6). The increase in the
proportion of interactions with the Year X þ 1 mate over the
interbreeding period was somewhat steeper for divorced males
than for divorced females (Appendix Table A6). Faithful and
divorced individuals did not differ in the number of interactions
they had with potential competitors during the interbreeding
period (Fig. 4; see also Appendix Tables A4 and A5).
To describe events related to divorce or reunification of pairs in
Year X þ 1, we plotted all interactions of both pair members with
opposite-sex (previous, future or potential mates) and same-sex
individuals (potential competitors) during the interbreeding
period (see Fig. 5 for examples; graphs for all other pairs are given
in the Supplementary Material).
In 63% of the cases of divorce, the former pair no longer inter-
acted during the interbreeding period, suggesting that the divorce
had occurred early (Fig. 6). In the remaining cases, the pair mem-
bers were still observed together during the interbreeding period,
suggesting that the divorce occurred later. In eight (40%) of those
pairs (nine individuals), at least one pair member interacted with
one to three same-sex individuals that had also interacted with the
focal individual's past mate, suggesting that intrasexual competi-
tion could have been the cause of the divorce.
For seven of the faithful pairs (29%, eight individuals), we
observed interactions with same-sex individuals. Hence, the
64 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72

1 (a)

Predicted probability of divorce

0.75

0.5

0.25

0 50 100 150 200

Absolute difference in arrival (days)

Male arrived first Female arrived first

12 (b)
10

2
Number of pairs

Male arrived first Female arrived first

12 (c)
10

0
–250 –225 –200 –175 –150 –125 –100 –75 –50 –25 0 25 50 75 100 125 150 175 200 225
Difference in arrival (number of days)

Figure 2. Differences in arrival date at the study site (first observation at a nestbox during the interbreeding period) of Year X pair members that ended up divorced or remained
faithful. (a) The predicted probability of divorce increased by 0.5 (estimate 0.022, 0.009 SE) per day increase in difference of arrival time, logistic regression: Wald's z ¼ 2.43,
P ¼ 0.015, N ¼ 67. Shown are the model estimates with 95% confidence intervals. (b, c) Histograms show the distribution of differences in arrival date between members of (b)
divorced pairs and (c) faithful pairs.

potential occurrence of intrasexual competition was not signifi- transformed mean: 50 m, N ¼ 77) did not disperse further than
cantly different between faithful pairs and pairs that divorced late faithful males (b ± SE ¼ 0.04 ± 0.14, t ¼ 0.28, P ¼ 0.78).
(chi-square test: c21 ¼ 0.72, P ¼ 0.4).
There were no differences in tarsus length between the Year X DISCUSSION
and X þ 1 mates of divorced individuals (paired t test: divorced
males: t8 ¼ 0.54, P ¼ 0.61; divorced females: t10 ¼ 0.15, We found that the probability of divorce in blue tit pairs in
P ¼ 0.88). The probability that a female that interacted with a focal Westerholz was not affected by breeding success in Year X, sug-
male ended up mating with this male did not depend on her tarsus gesting that divorce did not result from mating decisions in
length relative to that of the Year X female (Table 6). Males that response to lower mate or territory quality. In Year X þ 1, faithful
interacted with a focal female were more likely to end up as the pairs had a somewhat higher breeding success than individuals
new mate when they had a smaller tarsus length compared to the that had divorced: faithful females started egg laying earlier, had
Year X male (Table 6). somewhat larger clutches and produced slightly more fledglings.
Females that ended up divorced in Year X þ 1 dispersed signif- The only significant predictor of divorce was the difference in
icantly further (back-transformed mean: 93 m, N ¼ 77; linear arrival time to the study site between the members of Year X pairs
mixed model with ln-transformed distance: b ± SE ¼ 0.57 ± 0.14, (as estimated by SNB visits). Faithful pairs were more likely to
t ¼ 4.05, P < 0.001) from their Year X nestbox than females that arrive simultaneously at the breeding site. Moreover, we found that
remained faithful (back-transformed mean: 52 m, faithful pairs members of faithful pairs had a higher proportion of interactions
N ¼ 44; b0 ± SE ¼ 3.96 ± 0.12). In contrast, divorced males (back- with one another (compared to interactions with other individuals
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72 65

(a) Divorced N : 5 4 0 0 12 34 38 21 (c) Divorced N : 5 4 0 0 12 34 38 22

Proportion of interactions with Year X+1 mate

Proportion of interactions with Year X mate

Faithful N : 4 0 2 2 8 14 22 18 Faithful N : 4 0 2 2 8 14 22 28
1

0.75

0.5

0.25

0
(b) Divorced N : 3 6 0 2 14 32 38 22 (d) Divorced N : 3 6 0 2 14 32 38 22
Faithful N : 5 0 4 3 6 13 22 17 Faithful N : 5 0 4 3 6 13 23 17
1

0.75

0.5
Divorced
0.25 Faithful
0
Sept Oct Nov Dec Jan Feb Mar Apr Sept Oct Nov Dec Jan Feb Mar Apr
Month Month

Figure 3. Differences in proportion of interactions of (a) females and (b) males of faithful and divorced pairs with their Year X mate, compared to interactions with other individuals
of the opposite sex, and (c) females and (d) males of faithful and divorced pairs with their Year X þ 1 mate, compared to interactions with other individuals of the opposite sex, over
the interbreeding period. Shown are means and standard errors. Sample sizes are numbers of focal individuals per month that interacted with members of the opposite sex. Faithful
individuals' Year X mates and Year X þ 1 mates are, by definition, the same individual.

(a) + Divorced N: 0/0 1/6 3/6 0/0 0/0 0/12 6/34 16/40 5/24
Faithful N: 0/0 0/4 1/1 2/3 2/2 0/8 0/14 1/22 1/18
Number of interactions with potential competitors

(b) Divorced N: 2/2 1/3 4/8 4/4 7/8 7/16 7/32 5/39 4/23
Faithful N: 0/0 0/5 0/0 0/4 1/3 4/7 3/13 3/22 1/17

Faithful
Divorced

1
Aug Sept Oct Nov Dec Jan Feb Mar Apr
Month

Figure 4. Differences in number of interactions with potential competitors over the interbreeding period. Shown are means and standard errors for (a) females and (b) males of
faithful and divorced pairs. Sample sizes are numbers of focal individuals per month that interacted with members of the same sex, out of the total number of focal individuals
interacting with individuals in that month.

of the opposite sex) throughout the interbreeding period than their that in general the process of pair bonding with the new mate
divorced counterparts. In fact, 63% of divorced pairs no longer occurred during this period.
interacted during the interbreeding period, suggesting that the
divorce might have taken place earlier. In divorced pairs, the pro- Breeding Success and Parental Age as Potential Causes of Divorce
portion of interactions with their Year X þ 1 mates was overall
lower than in faithful pairs. However, in divorced individuals, the Our results suggest that the probability of divorce by Year X þ 1
proportion of interactions with their Year X þ 1 mates steeply is unrelated to the pair's breeding success in Year X, which is in line
increased between January and the start of breeding, suggesting with results from previous studies on tits and other bird species
66 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72

15
(a) +
12
9

Early divorce
6
3
0
15
12 (b)
9
6
3
0
15
(c) +
Number of interactions

12
9
6

Late divorce
3
0
15
12 (d)
9
6
3
0
15
12 (e) +
9
6

Faithful pair
3
0
15
12 (f)
9
6
3
0
1 Sept 1 Oct 1 Nov 1 Dec 1 Jan 1 Feb 1 Mar 1 Apr
Date

Relation Potential competitor Potential mate Year X+1 mate Year X mate

Figure 5. Graphs illustrating interactions of the (a, c, e) female and (b, d, f) male of a blue tit pair during the interbreeding period. Each symbol in each graph corresponds to a
different individual that interacted with the focal individual. The solid black vertical line indicates the day of arrival of that individual at the study site; the solid red vertical line
indicates the date of the first egg in Year X þ 1. Symbols are jittered to prevent visual clumping of data. (a, b) Pair that divorced early, i.e. no more interactions between the Year X
partners were observed. (a) The female had arrived at the study site in September and soon after started interacting with its new mate, and briefly also with one other potential
mate. Later in the season (February), the female interacted with two other potential competitors, one of which became the new mate of her Year X partner. (b) The male of the Year
X pair only arrived at the study site in March, and 18 days later started interacting with his new mate. (c, d) Pair that divorced late, i.e. was still observed together during the
interbreeding period. (c) In this example, the female also arrived at the study site considerably earlier than the male, and interacted with two other potential mates, one of which
was her future mate, before the first interaction with her former mate. The former pair was observed together for 14 days, after which the female only interacted with her new mate
(and one potential competitor). (d) The male started interacting with his new mate 7 days after the last interaction with his former mate. (e, f) Faithful pair. (e) The female arrived 84
days later than (f) the male, and they both first interacted with one another. Both pair members interacted with two other males, and the male also interacted with three other
females.

(Dhondt, Adriaensen, & Plompen, 1996; García-Navas & Sanz, 2011;
Harvey, Greenwood, & Perrins, 1979; Saitou, 2002; Valcu &
Kempenaers, 2008). Because our study site is an island of high-
quality habitat (the only forest patch in the area dominated by
oak), variation in breeding success related to territory or mate
quality may be relatively low compared to the larger-scale blue tit
population, and may therefore not be related to the probability of
divorce (see also Ramsay et al., 2000; Valcu & Kempenaers, 2008).
Moreover, several studies have suggested that remaining faithful or
divorcing can both be strategies to improve breeding success in
different contexts unrelated to Year X breeding success (Dhondt &
Adriaensen, 1994; Pampus, Schmidt, & Wiltschko, 2005). Here, we
only considered the number of offspring produced, and not their
zilly, Pre
quality (Ce ault, Dubois, Faivre, & Patris, 2000), for example
as reflected in postfledging survival and subsequent recruitment.
Apart from these considerations, our results do not support the
hypothesis that divorce is the result of a decision to switch partners
based on past breeding success.
Adult males were significantly less likely to divorce than year-
ling males (see also Pyle, Sydeman, & Hester, 2001), while divorce
was independent of female age. This may reflect female preference
for more experienced or older males, for example, because they are
better at providing parental care or because they have shown their
ability to survive (Searcy, 1979). However, after divorce, 26% of fe-
males ended up with another yearling mate (Table A1). Alterna-
tively, older males may be more successful than yearling males in
defending their mate or territory against intruders, or they may be
better at keeping track of their Year X mate during the inter-
breeding period (see below).
Competition or Partner Availability as Potential Causes of Divorce
More than half of all blue tit pairs divorced each year. The pro-
portion of pairs that divorced was not related to between-year
changes in the total number of breeding pairs. This suggests that
neither variation in the level of competition nor variation in the
availability of new partners (both should be higher in high-density
years) explained variation in divorce rate between years.
We found evidence for intrasexual competition in at least 40% of
pairs that divorced late, and in at least 29% of pairs that had
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72 67

At least one individual of Year X

pair present in interbreeding period
N = 77

Faithful Divorced
N = 25 N = 52

♂ interacting with ♀ interacting with

Early divorce Late divorce
other ♂♂ other ♀♀
N = 32 (62%) N = 20 (38%)
N = 4* (16%) N = 4 (16%)

♂ arrived late ♀ arrived late

N = 7 (4§) N = 13 (4§)

♂ interacting with ♀ interacting with ♂ interacting with ♀ interacting with

other ♂♂ other ♀♀ other ♂♂ other ♀♀
N = 1 (14%) N = 3 (43%) N = 4* (31%) N = 1* (8%)

Figure 6. Schematic overview of the frequency of different behavioural scenarios related to divorce, based on detection of individuals, marked with a transponder, at a nestbox in
the study site. Early divorce implies that the former partners never interacted during the interbreeding period. Interactions with same-sex individuals during the time when the
partners were still observed together suggests intrasexual competition for the territory or mate. Note that these cases are underestimated, because individuals without a tran-
sponder were not recorded. * In some cases, an individual's future mate was not equipped with a transponder during (part of) the interbreeding period. x Number of known cases
where the Year X mate had already interacted with potential mates before the arrival of its former partner.

Table 6 contra Valcu and Kempenaers (2008), who suggested that divorce
Mating success in relation to tarsus length of potential mates (individuals that in a Viennese population of blue tits was the result of female - fe-
interacted with the focal male or female) relative to tarsus length of the focal in-
dividual's Year X mate
male competition. Valcu and Kempenaers (2008) reported that the
new mates of divorced males were larger than the males' former
Estimate SE z P mates (see also Hogstad, 1989), suggesting that the new females
Female mating Intercept 2.95 5.40 might have been dominant and had seized the territory/male by
successa Tarsus length of 3.94 5.40 0.73 0.47 expelling the former mate.
potential mate
Male mating Intercept 8.87 5.07
successb Tarsus length of 9.88 5.07 1.95 0.05 Timing of Separation of Former Breeding Partners
potential mate

Both faithful and divorced pairs are included. Results from a generalized linear
mixed model with mating success (0 ¼ no mating, 1 ¼ Year X þ 1 mate) as the
response variable. Pair ID (Year X) and individual ID were included as random
effects.
a another) during the interbreeding period (Fig. 6). The asynchro-
Interacting females: N ¼ 124, divorced males: N ¼ 38, faithful males: N ¼ 24.
b
Interacting males: N ¼ 130, divorced females: N ¼ 41, faithful females: N ¼ 24.
remained faithful. However, the number of interactions with po-
tential competitors at the nestbox did not differ between faithful or
divorced males or females (Fig. 4), and there was no difference in
body size (measured as tarsus length) between the old and new
mate of divorced individuals. Moreover, the probability that a
known competitor ended up mating with an individual did not
increase when it was larger relative to the former (Year X) mate.
Taken together, we found no evidence for the hypothesis that direct
male - male or female - female competition played a role in divorce,
We found that in at least 43% of Year X pairs (at least 63% of Year
X pairs that divorced), the pair members no longer interacted with
one another (i.e. did not visit a nestbox within 15 min of one
nous arrival times to the field site of the individuals of these pairs,
coupled with this lack of interactions, indicates that in these cases
the ‘decision’ to divorce must have been made long before the start
of the next breeding season. Early divorce might then simply be a
consequence of the former pair members' decisions about where
and with whom (in which flock) to spend the interbreeding period.
This may or may not be related to individual qualities or alternative
mating opportunities.
In at least 37% of Year X pairs that divorced, the former pair
members still interacted at least once during the interbreeding
period (‘late divorce’; Fig. 6). This implies that the process of divorce
68 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
was still ongoing after both former pair members returned to the
study site. Although the exact cause of divorce remains unclear, it
might still be linked to the late arrival of one of the pair members
(see below), and it might involve direct competition between the
former mate and the new mate (e.g. Fig. 5c).
Timing of Return to the Breeding Site Explains Divorce
The probability that a pair divorced was strongly related to the
difference in arrival time at the study site (estimated as the first
visit to a nestbox) of the former pair members (Fig. 2a). Pairs that
arrived more simultaneously were much more likely to breed
together again, whereas early arrival of the male (more common),
and even more so early arrival of the female, often resulted in
divorce (Fig. 2b, c). Indeed, in 13 of 16 cases where the male of the
Year X pair arrived at the breeding site after the female, the pair
ended up divorcing. These observations are in line with previous
studies that have monitored individual movements in the
nonbreeding season by direct observations (Dhondt et al., 1996;
Gunnarsson, Gill, Sigurbjo € rnsson, & Sutherland, 2004; Handel &
Gill, 2000). Dhondt et al. (1996) found that great tit, Parus major,
pairs divorced when one member arrived at the field site after its
former mate had already formed a new pair. Handel and Gill (2000,
p. 480) described that in black turnstones, Arenaria melanocephala,
‘both sexes return to former breeding territories, a male courts all
females until paired, and a female pairs with the unattached male
closest to her previous territory if her former mate fails to return’,
and they refer to this as the ‘bet-hedging’ hypothesis. In our study,
individuals of at least eight late-divorced pairs had already inter-
acted with potential mates or with their future mate before their
Year X mate had arrived at the field site. In those cases where the
male of the Year X pair arrived after the female, three females had
already visited their Year X breeding box before their former mate
arrived at the breeding site. This suggests that, at least in some
instances, divorce would not have occurred if the former mate had
arrived more synchronously.
In winter, species such as the blue tit often join mixed-species
flocks (Cramp, Perrins, & Brooks, 1993) to improve foraging effi-
ciency and/or predator defence (Berner & Grubb, 1985; Hogstad,
1989). Previous studies showed that divorce occurred more often
in blue or great tit pairs whose members did not belong to the same
winter flocks (Dhondt et al., 1996; Saitou, 2002). Thus, individuals
may actively decide to remain faithful through the decision to stay
with their former partner in the same winter flock, possibly with
active maintenance of the pair bond throughout the winter (e.g.
Colquhoun, 1942; Drent, 1983; Pampus et al., 2005). Remaining
together and interacting throughout the interbreeding period may
improve breeding success by increasing survival chances of both
individuals, by strengthening the pair bond, and by both in-
dividuals deciding to return to the nest site and start breeding
earlier. Indeed, several studies have suggested that individuals that
remain together during the interbreeding period might increase
their probability of survival by boosting their dominance rank in
mixed flocks (Culina, Lachish, Pradel, Choquet, & Sheldon, 2013;
Nicolai, Sedinger, Ward, & Boyd, 2012; Pampus et al., 2005).
Alternatively, divorce may just be a side-effect of decisions related
to habitat use and flock membership, as well as decisions about
when to return to the breeding site, without any pair bond
behaviour during the interbreeding period.
If former pair members have been separated after the breeding
season (for example, ended up in different winter flocks) by chance
or by decision, it is not unlikely that they will return to their former
territory at different times in the spring (Dhondt et al., 1996). For a
short-lived species such as the blue tit, it is probably adaptive to
pair with a new mate rather than wait for the previous mate to
return, because the probability that the former mate has died is
high, and delaying or forgoing breeding has strong negative fitness
consequences (Perrins, 1965). Thus, after having arrived at the
breeding area, a blue tit might benefit from forming a new pair,
even if there is a cost to divorce (see below).
Previous studies have highlighted the importance of the social
environment for both the formation and maintenance of pair bonds
in birds (Colquhoun, 1942; Culina, Hinde, & Sheldon, 2015 and
references therein). Culina et al. (2015) reported that male great tits
that associated with a lower proportion of females in the winter
before Year X, and whose Year X mate had a low rank among those
females, were more likely to divorce by Year X þ 1. Our results
suggest that divorce in blue tits might occur as a by-product of the
pair members' presence in winter flocks, and their arrival at the
breeding site (Fig. 2). If staying together during winter or arrival
time at the breeding site is an individual or pair-specific ‘trait’, then
we should rarely see instances where a pair that bred together in
the first 2 or more years, divorced the year after. We found that 65%
of females and 48% of males had mixed breeding strategies over the
years, that is, in some years they remained faithful, while in other
years they divorced (Table 2). Several studies have shown that in-
dividual consistency (repeatability) in mate preferences is low (see
meta-analysis by Bell, Hankison, & Laskowski, 2009), but, as we
discuss above, divorce might not directly reflect mate preferences.
Of the two focal individuals in our study that were faithful in year
X þ 1 and could have bred together in at least one more year, there
was one instance where the pair divorced in Year X þ 2 and one
instance where they remained together (Table 2).
Consequences of Divorce
Our results suggest that faithful pairs did somewhat better in
Year X þ 1 than pairs that divorced. Faithful females started egg
laying earlier, laid a larger clutch and produced more fledglings
than the new females of divorced males (Table 4), or than divorced
females (Table 5). When a divorced male remated with a yearling
female, he also suffered higher paternity loss (Table 4). Interest-
ingly, some of the effects we found were significant even when the
divorced individuals' new partner was an adult (Tables 4, 5). Most
of the new adult partners of the divorced individuals had previous
breeding experience in the study site (females of divorced males:
76%, 40/52; males of divorced females: 71%, 38/53). This implies
that the lower breeding success in divorced pairs is not solely due
to age effects or lack of breeding experience of the new mates.
Divorced individuals were more likely to mate with adults in Year
X þ 1, but this was also true for widowed individuals (Table A1).
This implies that being locally experienced per se makes it more
likely that an individual will end up with an older partner. Our
results are in accordance with earlier studies showing that
reproductive success increases with pair bond duration (see re-
views in Rowley, 1983; Fowler, 1995; Ens et al., 1996), which seems
true in both short-lived species similar to the blue tit (Griggio &
Hoi, 2011; Perrins & McCleery, 1985), and in longer-living spe-
cies (Coulson, 1966; van de Pol, Heg, Bruinzeel, Kuijper, & Verhulst,
2006). A recent meta-analysis (Culina et al., 2014) comprising 81
studies of divorce in 64 bird species showed that, overall, in-
dividuals that divorced had a lower Year X þ 1 breeding success
than individuals in faithful pairs. In addition, the meta-analysis
showed a similar reduction in breeding success for divorced and
widowed individuals, suggesting that there is an intrinsic cost to
mating with a new partner (see also Coulson, 1972; Ens et al.,
1996), regardless of the reason for the mate change. Indeed, we
also found no significant differences in measures of breeding
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
success in Year X þ 1 between divorced and widowed individuals
(Table A3).
Moreover, divorced individuals did not improve their breeding
success from Year X to Year X þ 1, while faithful pairs advanced
their laying date (Fig. 1). Previous studies on divorce in birds have
suggested that in many instances at least one pair member initiates
the separation to improve its breeding success with another mate
(Ens et al., 1993; Orell et al., 1994; Streif & Rasa, 2001; Dhondt,
2002; see also reviews in ; Rowley, 1983; Johnston & Ryder, 1987;
Choudhury, 1995; McNamara & Forslund, 1996). Previous studies
on divorce in blue tits reported conflicting results as to which sex
benefits from the mate change, and thus has seemingly initiated the
divorce. For example, Dhondt and Adriaensen (1994) suggested
that because females improved their breeding success after divorce,
females chose to change males. Valcu and Kempenaers (2008),
however, reported that male blue tits breeding in Vienna seemed to
benefit from divorce, while there was a cost to females. The meta-
analysis by Culina et al. (2014) showed that divorced individuals
generally increased their breeding success from Year X to Year
X þ 1, but the conclusion depended on the method of analysis.
Overall, our results suggest that divorce is not a strategy to improve
breeding success, but rather an outcome with negative fitness
consequences.
Faithful individuals may simply avoid costs related to time and
energy spent prospecting for a new territory or mate (Ramsay
et al., 2000), courting the new mate (Coulson, 1966), and in
competitive encounters that come with mate change. This in-
vestment may delay the start of laying, and in turn lead to a
smaller clutch in divorced individuals. The interactions observed at
the nestboxes confirm the hypothesis that divorced individuals
spent more time or energy remating. During the period of intense
territorial defence, nest site selection and nest building (between
February and April), faithful individuals interacted almost exclu-
sively with their mates, whereas divorced individuals still inter-
acted with other potential mates (the proportion of interactions
with their Year X þ 1 mates increased to reach a maximum of
around 0.8 in April; Fig. 3). From January to April, divorced females
interacted with an average of 2.4 potential mates (2.0 when
excluding the Year X mate), while faithful females interacted on
average with only 0.6 potential mates. Similarly, divorced males
interacted with an average of 2.1 potential mates other than their
Year X þ 1 mate (1.6 when excluding the Year X mate), while
faithful males interacted with an average of 0.6 potential mates.
These observations suggest that in early spring divorced in-
dividuals still needed time for pair bond development, had to deal
with competitive interactions between potential mates, or showed
‘bet-hedging’ behaviour (Handel & Gill, 2000), that is, they
continued to interact with multiple individuals of the opposite sex
until later in the season.
Males and females differed in dispersal distance from Year X to
Year X þ 1 (see also Valcu & Kempenaers, 2008). Both faithful pairs
and divorced males were site-faithful between years: they
dispersed on average 52 m and 50 m, respectively, which corre-
sponds to moving to a neighbouring nestbox in our study site. This
again suggests that it is unlikely that male - male competition plays
a role in divorce; that is, divorce is not the result of a male losing his
territory after interactions with a same-sex intruder. In contrast,
divorced females dispersed significantly further: on average around
90 m (see also Blondel et al., 2000; Pampus et al., 2005). Similarly,
Valcu and Kempenaers (2008) reported that divorced blue tit fe-
males bred in Year X þ 1 territories that overlapped significantly
less with their Year X territory than those of faithful pairs. They also
reported that females that had dispersed on average produced
more fledglings than females that had remained resident, pre-
sumably because they moved to a territory with a higher oak
69
density (the study site showed a north - south gradient in habitat
quality). However, in Westerholz oak trees are spread more ho-
mogeneously throughout the study site. Indeed, within the study
area, there was no evidence for spatial autocorrelation for mean
clutch size (a good measure of territory quality, see Valcu &
Kempenaers, 2008 and references therein) or for mean number of
young fledged (see Fig. S2 in the Supplementary Material), sug-
gesting homogeneity in habitat quality. Thus, it seems unlikely that
dispersal per se causes changes in breeding success between years.
Moreover, the observed change did not differ between divorced
males (which do not move) and divorced females (Fig. 1).
Furthermore, for divorced females, the change in breeding success
from Year X to Year X þ 1 did not correlate with the distance they
dispersed (lay date: Pearson correlation r74 ¼ 0.08, P ¼ 0.49;
clutch size: r75 ¼ 0.12, P ¼ 0.31; fledging success: r75 ¼ 0.10,
P ¼ 0.40). Finally, we compared mean clutch size and mean number
of young fledged across all study years (2007e2016; data stan-
dardized within year) between the Year X and the Year X þ 1
nestbox of the divorced females, with data from the focal female
excluded. If females that divorced had ‘upgraded’ to a higher quality
territory, we would expect that the mean clutch size and fledging
success of their Year X þ 1 boxes would be higher than those of Year
X boxes. However, this was not the case (paired t tests for N ¼ 65
divorced females, each female only considered once; difference in
mean clutch size: Year X þ 1 box e Year X box ¼ 0.17 ± 1.37 SD,
t64 ¼ 1.00, P ¼ 0.32; difference in mean number of fledglings
(including complete brood failures): 0.01 ± 3.27, t64 ¼ 0.02,
P ¼ 0.99; difference in number of fledglings (excluding complete
brood failures): 0.04 ± 2.08, t57 ¼ 0.16, P ¼ 0.88). Although fe-
male - female aggression during the early breeding season is
common in blue tits (Kempenaers, 1995), we found little evidence
that divorced females dispersed as a result of retreat from
competitive encounters with other females (as suggested by Valcu
& Kempenaers, 2008, see above).
Extrapair Mating and Divorce
Pairs that had at least one extrapair young in their Year X brood
were not more likely to divorce than those without extrapair young
(Table 3), suggesting that there is no direct relationship between
female decisions to engage in extrapair copulations and divorce
(contra Cezilly & Nager, 1995).
In Year X þ 1, there was no difference in the probability that a
brood contained extrapair young between divorced females, fe-
males mated with divorced males or faithful females. However, the
Year X þ 1 broods of four divorced females (three early divorced,
one late divorced) contained extrapair young sired by her former
(Year X) mate. In the late-divorce case, the focal female arrived 100
days after her former male, but visited her Year X breeding box on
multiple occasions. This suggests that the male was already paired
with a new female by the time the focal female had arrived, but that
she still copulated with him. If the asynchrony in arrival times
between the former pair members led in turn to asynchrony in the
onset of laying of the former and new mate, the male may have had
a window of opportunity to copulate with his former mate. Indeed,
males may seek extrapair matings when their own female is no
longer fertile (Canal, Jovani, & Potti, 2012; García-Navas et al., 2015;
but see Araya-Ajoy, Dingemanse, & Kempenaers, 2016). However,
we think this explanation is unlikely, because extrapair paternity
patterns were unrelated to breeding synchrony in this and other
blue tit populations (Schlicht, Valcu, & Kempenaers, 2015 and ref-
erences therein). In at least two of the three early divorced pairs,
the early arriving females had already paired with their new mate
by the time their Year X male arrived (in the third case, no box visit
data were available until 6 days before the appearance of the first
70 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72
egg). In all cases, the extrapair males were either first- or second-
order neighbours in Year X þ 1. Although these anecdotes are
intriguing, it still needs to be shown whether the observed number
of cases (N ¼ 4) differs from the number of cases expected, given
that most extrapair sires in blue tits are first- or second-order
neighbours (Kempenaers et al., 1992; Schlicht et al., 2015), but
this requires a larger sample size.
Conclusions and Outlook
Our results and those of previous studies suggest that to un-
derstand divorce it is not sufficient to consider the start and end-
points, namely breeding success in Year X and Year X þ 1, but that it
is necessary to examine the social environment and processes of
settlement and pair formation before the first breeding attempt
(Culina et al., 2015), and in between breeding attempts. Our study
shows that divorce in blue tits may arise as a by-product of sepa-
ration of the two pair members after the Year X breeding season,
perhaps because they joined different winter flocks, thus leading to
asynchrony between the two individuals in the timing of arrival,
settlement and pair formation in Year X þ 1. Future studies on
nonmigratory populations that follow the behaviour of pair mem-
bers after breeding (e.g. using radiotracking) might help us gain an
even better understanding of the processes that ultimately lead to
divorce.
Acknowledgments
We are grateful to all members of the field team, and especially
to Agnes Türk, Andrea Wittenzellner and numerous field assistants,
for catching, processing and collecting data on the birds over the
years. We thank Peter Loe €s and Peter Skripsky for maintaining the
‘smart nestboxes’ and other equipment, Alexander Girg for geno-
typing, and two anonymous referees for constructive comments on
the manuscript. B.K. and M.V. designed the study, C.G. analysed the
data with the help of M.V., and B.K. and C.G. wrote the paper with
input from M.V. C.G. is a member of the International Max Planck
Research School for Organismal Biology. This work was funded by
the Max Planck Society (to B.K.).
References
Araya-Ajoy, Y. G., Dingemanse, N. J., & Kempenaers, B. (2016). Timing of extra-pair
fertilizations: Within-pair fertilization trade-offs or pair synchrony spillovers?
Behavioral Ecology, 27, 377e384.
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2014). lme4: Linear mixed-effects
models using ‘Eigen’ and S4. R package version 1.1-7. http://cran.r-project.org/
web/packages/lme4/index.html.
Bell, A., Hankison, S. J., & Laskowski, K. L. (2009). The repeatability of behaviour: A
meta-analysis. Animal Behaviour, 77, 771e783.
Berner, T. O., & Grubb, T. C. (1985). An experimental analysis of mixed-species
flocking in birds of deciduous woodland. Ecology, 66, 1229e1236.
Black, J. M. (1996). Introduction: Pair bonds and partnerships. In J. M. Black (Ed.),
Partnerships in Birds: The Study of Monogamy (pp. 3e20). Oxford, U.K.: Oxford
University Press.
Blondel, J., Perret, P., & Galan, M. J. (2000). High divorce rates in corsican blue tits:
How to choose a better option in a harsh environment. Oikos, 89, 451e460.
Bried, J., & Jouventin, P. (2001). Site and mate choice in seabirds: An evolutionary
approach. In E. A. Schreiber, & J. Burger (Eds.), Biology of Marine Birds (pp.
263e305). New York, NY: CRC Press.
Canal, D., Jovani, R., & Potti, J. (2012). Male decisions or female accessibility?
Spatiotemporal patterns of extra pair paternity in a songbird. Behavioral Ecol-
ogy, 23, 1146e1153.
zilly, F., & Nager, R. G. (1995). Comparative evidence for a positive association
Ce
between divorce and extra-pair paternity in birds. Proceedings of the Royal So-
ciety of London Series B: Biological Sciences (London), 262, 7e12.
zilly, F., Pre
Ce ault, M., Dubois, F., Faivre, B., & Patris, B. (2000). Pair-bonding in birds
and the active role of females: A critical view of the empirical evidence.
Behavioural Processes, 51, 83e92.
Choudhury, S. (1995). Divorce in birds: A review of the hypotheses. Animal
Behaviour, 50, 413e429.
Colquhoun, M. K. (1942). Notes on the social behaviour of blue tits. British Birds, 35,
234e240.
Coulson, J. C. (1966). The influence of the pair-bond and age on the breeding
biology of the kittiwake gull Rissa tridactyla. Journal of Animal Ecology, 35,
269e279.
Coulson, J. C. (1972). The significance of the pair-bond in the kittiwake Rissa tri-
dactyla. International Ornithological Congress, 15, 423e433.
Coulson, J. C., & Thomas, C. S. (1980). A study of the factors influencing the duration
of the pair-bond in the Kittiwake Gull Rissa tridactyla. International Ornitho-
logical Congress, 27, 823e833.
Cramp, S., Perrins, C. M., & Brooks, D. J. (Eds.). (1993). Handbook of the Birds of
Europe, the Middle East, and North Africa: The Birds of the Western Palearctic.
Oxford, U.K.: Oxford University Press.
Culina, A., Hinde, C. A., & Sheldon, B. C. (2015). Carry-over effects of the social
environment on future divorce probability in a wild bird population. Pro-
ceedings of the Royal Society of London Series B: Biological Sciences (London), 282.
http://dx.doi.org/10.1098/rspb.2015.0920, 20150920.
Culina, A., Lachish, S., Pradel, R., Choquet, R., & Sheldon, B. C. (2013). A multievent
approach to estimating pair fidelity and heterogeneity in state transitions.
Ecology and Evolution, 3, 4326e4338.
Culina, A., Radersma, R., & Sheldon, B. C. (2014). Trading up: The fitness con-
sequences of divorce in monogamous birds. Biological Reviews, 90,
1015e1034.
Daniels, S. J., & Walters, J. R. (2000). Between-year breeding dispersal in red-
cockaded woodpeckers: Multiple causes and estimated cost. Ecology, 81,
2473e2484.
Desrochers, A., & Magrath, R. D. (1996). Divorce in the European blackbird: Seeking
greener pastures? In J. M. Black (Ed.), Partnerships in Birds: The Study of
Monogamy (pp. 177e191). Oxford, U.K.: Oxford University Press.
Dhondt, A. A. (2002). Changing mates. Trends in Ecology and Evolution, 17, 55e56.
Dhondt, A. A., & Adriaensen, F. (1994). Causes and effects of divorce in the blue tit
Parus caeruleus. Journal of Animal Ecology, 63, 979e987.
Dhondt, A. A., Adriaensen, F., & Plompen, W. (1996). Between- and within-
population variation in mate fidelity in the great tit. In J. M. Black (Ed.), Part-
nerships in Birds: The Study of Monogamy (pp. 235e248). Oxford, U.K.: Oxford
University Press.
Diamond, J. M. (1987). A Darwinian theory of divorce. Nature, 329, 765e766.
Drent, P. J. (1983). The functional ethology of territoriality in the great tit (Parus major
L.) (Doctoral Thesis). Groningen, The Netherlands: University of Groningen.
Dubois, F., & Ce zilly, F. (2002). Breeding success and mate retention in birds: A
meta-analysis. Behavioral Ecology and Sociobiology, 52, 357e364.
Ens, B. J., Choudhury, S., & Black, J. M. (1996). Mate fidelity and divorce in
monogamous birds. In J. M. Black (Ed.), Partnerships in Birds: The Study of
Monogamy (pp. 344e401). Oxford, U.K.: Oxford University Press.
Ens, B. J., Safriel, U. N., & Harris, M. P. (1993). Divorce in the long-lived and
monogamous oystercatcher Haematopus ostralegus: Incompatibility or choosing
the better option? Animal Behaviour, 45, 1199e1217.
Fowler, G. S. (1995). Stages of age-related reproductive success in birds: Simulta-
neous effects of age, pair-bond duration and reproductive experience. American
Zoologist, 35, 318e328.
García-Navas, V., Ferrer, E. S., C aliz-Campal, C., Bueno-Enciso, J., Barrientos, R.,
Sanz, J. J., et al. (2015). Spatiotemporal and genetic contingency of extrapair
behaviour in a songbird. Animal Behaviour, 106, 157e169.
García-Navas, V., & Sanz, J. J. (2011). Females call the shots: Breeding dispersal and
divorce in blue tits. Behavioral Ecology, 22, 932e939.
Gonzalez-Solis, J., Becker, P. H., & Wendeln, H. (1999). Divorce and asynchronous
arrival in common terns, Sterna hirundo. Animal Behaviour, 58, 1123e1129.
Green, D. J., Krebs, E. A., & Cockburn, A. (2004). Mate choice in the brown thornbill
(Acanthiza pusilla): Are settlement decisions, divorce, and extrapair mating
complementary strategies? Behavioral Ecology and Sociobiology, 55, 278e285.
Griggio, M., & Hoi, H. (2011). An experiment on the function of the long-term pair
bond period in the socially monogamous bearded reedling. Animal Behaviour,
82, 1329e1335.
€rnsson, T., & Sutherland, W. J. (2004). Pair bonds:
Gunnarsson, T. G., Gill, J. J., Sigurbjo
Arrival synchrony in migratory birds. Nature, 431, 646.
Handel, C. M., & Gill, R. E., Jr. (2000). Mate fidelity and breeding site tenacity in a
monogamous sandpiper, the black turnstone. Animal Behaviour, 60, 471e481.
Harvey, P. H., Greenwood, P. J., & Perrins, C. M. (1979). Breeding area fidelity of great
tits Parus major. Journal of Animal Ecology, 48, 305e313.
Hasson, O., & Stone, L. (2009). Male infertility, female fertility and extrapair
copulations. Biological Reviews of the Cambridge Philosophical Society, 84,
225e244.
Hatchwell, B. J., Russell, A. F., Ross, D. J., & Fowlie, M. K. (2000). Divorce in coop-
eratively breeding long-tailed tits: A consequence of inbreeding avoidance?
Proceedings of the Royal Society of London Series B: Biological Sciences, 267,
813e819.
Heg, D., Bruinzeel, L. W., & Ens, B. J. (2003). Fitness consequences of divorce in the
oystercatcher, Haematopus ostralegus. Animal Behaviour, 66, 175e184.
Hogstad, O. (1989). Social organization and dominance behaviour in some Parus
species. Wilson Bulletin, 101, 254e262.
Ihle, M., Kempenaers, B., & Forstmeier, W. (2015). Fitness benefits of mate choice for
compatibility in a socially monogamous species. PLoS Biology, 13, e1002248.
http://dx.doi.org/10.1371/journal.pbio.1002248.
Johnston, V. H., & Ryder, J. P. (1987). Divorce in larids: A review. Colonial Waterbirds,
10, 16e26.
 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72 71

Kempenaers, B., Verheyen, G. R., van den Broeck, M., Burke, T., van Broeckhoven, C., Appendix
& Dhondt, A. (1992). Extra-pair paternity results from female preference for
high-quality males in the blue tit. Nature, 357, 494e496.
Kempenaers, B. (1995). Polygyny in the blue tit: Intra- and inter-sexual conflicts.
Table A1
Animal Behaviour, 49, 1047e1064.
Kempenaers, B., Adriaensen, F., & Dhondt, A. A. (1998). Inbreeding and divorce in Partner age of divorced and widowed males and females in Year X and Year X þ 1
blue and great tits. Animal Behaviour, 56, 737e740. Partner age Year X Partner age Year Xþ1
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2015). LmerTest: Tests in
linear mixed effects models. http://cran.r-project.org/web/packages/lmerTest/ Nyearling Nadult
index.html.
Linden, M. (1991). Divorce in great tits - chance or choice? An experimental Divorced female mated to
approach. American Naturalist, 138, 1039e1048. Yearling male 10 (26%) 29
Marzluff, J. M., Woolfenden, G. E., Fitzpatrick, J. W., & Balda, P. B. (1996). Breeding Adult male 14 (37%) 24
partnerships of two New World jays. In J. M. Black (Ed.), Partnerships in Birds: Divorced male mated to
The Study of Monogamy (pp. 138e161). Oxford, U.K.: Oxford University Press. Yearling female 15 (36%) 27
McNamara, J. M., & Forslund, P. (1996). Divorce rates in birds: Predictions from an Adult female 10 (29%) 25
optimization model. American Naturalist, 147, 609e640. Widowed female lost
Newton, I., & Wyllie, I. (1996). Monogamy in the sparrowhawk. In J. M. Black (Ed.), Yearling male 19 (37%) 33
Partnerships in Birds: The Study of Monogamy (pp. 249e267). Oxford, U.K.: Adult male 34 (33%) 69
Oxford University Press. Widowed male lost
Nicolai, C. A., Sedinger, J. S., Ward, D. H., & Boyd, W. S. (2012). Mate loss affects
Yearling female 28 (41%) 41
survival but not breeding in black brant geese. Behavioral Ecology, 23,
Adult female 39 (43%) 51
643e648.
Orell, M., Rytko€nen, S., & Koivula, K. (1994). Causes of divorce in the monogamous
willow tit, Parus montanus, and its consequences for reproductive success.
Animal Behaviour, 48, 1143e1154. Table A2
Otter, K., & Ratcliffe, L. (1996). Female-initiated divorce in a monogamous songbird: Predictors of divorce in a blue tit population in southern Germany from 2007 to 2015
Abandoning mates for males of higher quality. Proceedings of the Royal Society of
London Series B: Biological Sciences, 263, 351e354. Parameter estimate SE z P
Owen, M., Black, J. M., & Liber, H. (1988). Pair bond duration and timing of its for-
mation in barnacle geese Branta leucopsis. In M. W. Weller (Ed.), Waterfowl in Intercept 1.21 0.5
Winter (pp. 23e38). Minneapolis, MN: University of Minnesota Press. Lay datea 0.047 0.059 0.793 0.43
Pampus, M., Schmidt, K., & Wiltschko, W. (2005). Pair bond and breeding success in Clutch sizea 0.053 0.168 0.315 0.75
blue tits Parus caeruleus and great tits Parus major. Ibis, 147, 92e108. Fledging successa,b 0.007 0.128 0.055 0.96
Perrins, C. M. (1965). Population fluctuations and clutch-size in the great tit, Parus Female agec 0.362 0.427 0.846 0.40
major L. Journal of Animal Ecology, 34, 601e647. Male agec 0.979 0.449 2.182 0.03
Perrins, C. M. (1979). British Tits. London, U.K.: Collins. Extrapair paternityd 0.184 1.688 0.109 0.91
Perrins, C. M., & McCleery, R. H. (1985). The effect of age and pair bond on the
breeding success of great tits Parus major. Ibis, 127, 306e315. Logistic regression with divorce (yes/no) as dependent variable, and measures of
van de Pol, M., Heg, D., Bruinzeel, L. W., Kuijper, B., & Verhulst, S. (2006). Experi- breeding success (standardized per year), and male and female age in Year X as
mental evidence for a causal effect of pair-bond duration on reproductive predictors.
a
performance in oystercatchers (Haematopus ostralegus). Behavioral Ecology, 17, Estimates show the change in the odds of divorce for every unit change in the
982e991. breeding parameter.
Pyle, P., Sydeman, W. J., & Hester, M. (2001). Effects of age, breeding experience, b
The number of chicks successfully fledged (excluding cases of complete brood
mate fidelity and site fidelity on breeding performance in a declining popula- failure).
tion of Cassin's auklets. Journal of Animal Ecology, 70, 1088e1097. c
Estimate is for adults compared to yearlings.
R Development Core Team. (2014). R: A language and environment for statistical d
Estimate is for the proportion of extrapair young in a clutch.
computing. Vienna, Austria: R Foundation for Statistical Computing. http://
www.r-project.org.
Ramsay, S. M., Otter, K. A., Mennill, D. J., Ratcliffe, L. M., & Boag, P. T. (2000). Divorce
and extrapair mating in female black-capped chickadees (Parus atricapillus): Table A3
Separate strategies with a common target. Behavioral Ecology and Sociobiology, Differences in measures of breeding success in Year X þ 1 (standardized within year)
49, 18e23. between widowed and divorced individuals
Rowley, I. (1983). Re-mating in birds. In P. Bateson (Ed.), Mate Choice (pp. 331e360).
Breeding parameter Explanatory Parameter SE t P
Cambridge, U.K.: Cambridge University Press.
Ryan, M. J. (1998). Sexual selection, receiver biases, and the evolution of sex dif- variable estimate
ferences. Science, 281, 1999e2003. _ Lay date Intercepta 1.67 0.29
Saitou, T. (2002). Factors affecting divorce in the great tit Parus major. Ibis, 144, Pair statusb 0.57 0.45 1.26 0.21
311e316.
Clutch size Intercepta 0.29 0.12
Schlicht, L., Girg, A., Loe €s, P., Valcu, M., & Kempenaers, B. (2012). Male extra-pair
Pair statusb 0.07 0.19 0.37 0.71
nestlings fledge first. Animal Behaviour, 83, 1335e1343.
Fledging successd Intercepta 2.05 0.18
Schlicht, E., & Kempenaers, B. (2015). Immediate effects of capture on nest visits of
breeding blue tits, Cyanistes caeruleus, are substantial. Animal Behaviour, 105, Pair statusb 0.27 0.28 0.97 0.33
63e78. Extrapair paternitye Intercepta 0.11 0.01
Schlicht, L., Valcu, M., & Kempenaers, B. (2015). Male extraterritorial behaviour Pair statusb 0.02 0.02 1.10 0.27
predicts extrapair paternity pattern in blue tits, Cyanistes caeruleus. Behavioral \ Lay date Intercepta 1.78 0.46
Ecology, 26, 1404e1413. Pair statusc 0.21 0.44 0.47 0.64
Searcy, W. A. (1979). Female choice of mates: A general model for birds and its Clutch size Intercepta 0.33 0.14
application to red-winged blackbirds (Agelaius phoeniceus). American Naturalist, Pair statusc 0.06 0.22 0.30 0.77
114, 77e100. Fledging successd Intercepta 1.60 0.21
Streif, M., & Rasa, A. (2001). Divorce and its consequences in the common blackbird Pair statusc 0.09 0.31 0.29 0.77
Turdus merula. Ibis, 143, 554e560. Extrapair paternitye Intercepta 0.10 0.01
Taborsky, B., & Taborsky, M. (1999). The mating system and stability of pairs in kiwi
Pair statusc 0.001 0.02 0.07 0.94
Apteryx spp. Journal of Avian Biology, 30, 143e151.
Valcu, M., & Kempenaers, B. (2008). Causes and consequences of breeding dispersal Shown are the results of linear mixed model analysis (ML) with pair status (wid-
and divorce in a blue tit, Cyanistes caeruleus, population. Animal Behaviour, 75, owed/divorced) as the explanatory variable.
1949e1963. a
Estimate for widowed individuals.
b
Estimate for divorced males (N ¼ 77) relative to widowed males (N ¼ 159).
c
Appendix A. Supplementary data Estimate for divorced females (N ¼ 77) relative to widowed females (N ¼ 155).
d
The number of chicks successfully fledged (excluding complete brood failures:
widowed males 11% (18/159), divorced males 8% (6/77), c21 ¼ 0.71, P ¼ 0.40; wid-
Supplementary data related to this article can be found at http:// owed females 14% (21/155), divorced females 13% (10/77), c21 ¼ 0.01, P ¼ 0.91).
dx.doi.org/10.1016/j.anbehav.2017.09.004. e
The proportion of extrapair young in the clutch.
72 C. Gilsenan et al. / Animal Behaviour 133 (2017) 57e72

Table A4 Table A6
Interactions with past and future mate and with potential competitors for divorced Interactions with future mate for divorced males and females in relation to date
and faithful males and in relation to date
Parameter estimate SE z P
Parameter estimate SE z P
Interactions with future matea
Interactions with past matea Interceptb 3.35 1.97
Interceptb 0.58 0.91 Divorced males 7.87 2.3 3.42 <0.001
Faithful males 1.41 0.93 1.52 0.13 Monthc 1.08 0.14 7.76 <0.001
c
Month 1.23 0.2 6.17 <0.001 Divorced males: monthc 2.40 0.40 5.99 <0.001
Faithful males: monthc 2.48 0.26 9.58 <0.001
Logistic regression with divorced individuals' interactions with future mate (yes/no)
Interactions with future matea
as dependent variable, and pairing status and month of interaction as explanatory
Interceptb 10.56 2.34
variables. For all models, both focal individual ID and year were included as random
Faithful males 11.89 1.98 6.02 <0.001
effects, and month of interaction was re-centred to have December as 0.
Monthc 3.41 0.38 9.05 <0.001 a
Based on interactions involving divorced males (N ¼ 36) and divorced females
Faithful males: monthc 2.05 0.41 4.98 <0.001
(N ¼ 38).
Interactions with potential competitorsd b
b Estimate for divorced females.
Intercept 0.23 0.15 c
The estimate shows the change in odds of interacting with future mate, versus
Faithful males 0.02 0.52 0.05 0.96
any other individual of the opposite sex, with respect to every unit change in month
Monthe 0.03 0.07 0.49 0.63
(from September to April inclusive, December ¼ 0).
Faithful males: monthe 0.02 0.23 0.07 0.95

Logistic regression with focal males' interactions with (1) past or (2) future mate
(yes/no) as dependent variable, and pairing status and month of interaction as
explanatory variables. For faithful individuals, past and future mate would be the
same individual. ‘Potential competitor’: Poisson regression model with number of
focal males' interactions with potential competitors as dependent variable, and
pairing status and month of interaction as explanatory variables. For all models,
both focal individual ID and year were included as random effects, and month of
interaction was re-centred to have December as 0.
a
Based on interactions involving divorced (N ¼ 36) and faithful (N ¼ 19) males.
b
Estimate for divorced males.
c
The estimate shows the change in odds of interacting with previous/future
mate, versus any other individual of the opposite sex, with respect to every unit
change in month (from September to April inclusive).
d
Based on interactions involving divorced (N ¼ 23) and faithful (N ¼ 8) males.
e
Month range is from August to April inclusive (December ¼ 0).

Table A5
Interactions with past and future mate and with potential competitors for divorced
and faithful females and in relation to date

Parameter estimate SE z P

Interactions with past matea

Interceptb 1.45 0.63
Faithful females 2.72 0.87 3.13 0.002
Monthc 1.06 0.18 6.04 <0.001
Faithful females: monthc 2.31 0.25 9.10 <0.001
Interactions with future matea
b
Intercept 2.01 0.89
Faithful females 2.08 1.13 1.84 0.066
Monthc 1.08 0.14 7.88 <0.001
Faithful females: monthc 0.21 0.24 0.87 0.38
Interactions with potential competitorsd
b
Intercept 0.26 0.26
Faithful females 0.26 0.46 0.57 0.57
e
Month 0.04 0.09 0.43 0.67
Faithful females: monthe 0.04 0.20 0.18 0.86

Logistic regression with focal females' interactions with (1) past or (2) future mate
(yes/no) as dependent variable, and pairing status and month of interaction as
explanatory variables. For faithful individuals, past and future mate would be the
same individual. ‘Potential competitor’: Poisson regression model with number of
focal females' interactions with potential competitors as dependent variable, and
pairing status and month of interaction as explanatory variables. For all models,
both focal individual ID and year were included as random effects, and month of
interaction was re-centred to have December as 0.
a
Based on interactions involving divorced (N ¼ 38) and faithful (N ¼ 24) females.
b
Estimate for divorced females.
c
The estimate shows the change in odds of interacting with previous/future
mate, versus any other individual of the opposite sex, with respect to every unit
change in month (from September to April inclusive).
d
Based on interactions involving divorced (N ¼ 23) and faithful (N ¼ 5) females.
e
Month range is from September to April inclusive (December ¼ 0).