Testing the effect of conspecific reproductive success on dispersal and recruitment decisions in a colonial bird: Design issues

Abstract

Factors affecting dispersal and recruitment in animal populations will play a prominent role in the dynamics of populations. This is particularly the case for subdivided populations where the dispersal of individuals among patches may lead to local extinction and 'rescue effects'. A long-term observational study carried out in Brittany, France, and involving colour-ringed Black-legged Kittiwakes (Rissa tridactyla) suggested that the reproductive success of conspecifics (or some social correlate) could be one important factor likely to affect dispersal and recruitment. By dispersing from patches where the local reproductive success was low and recruiting to patches where the local reproductive success was high, individual birds could track spatio-temporal variations in the quality of breeding patches (the quality of breeding patches can be affected by different factors, such as food availability, the presence of predators or ectoparasites, which can vary in space and time at different scales). Such an observational study may nevertheless have confounded the role of conspecific reproductive success with the effect of a correlated factor (e.g. the local activities of a predator). In other words, individuals may have been influenced directly by the factor responsible for the low local reproductive success or indirectly by the low success of their neighbours. Thus, an experimental approach was needed to address this question. Estimates of demographic parameters (other than reproductive success) and studies of the response of marked individuals to changes in their environment usually face problems associated with variability in the probability of detecting individuals and with nonindependence among events occurring on a local scale. Further, very few studies on dispersal have attempted to address the causal nature of relationships by experimentally manipulating factors. Here we present an experiment designed to test for an effect of local reproductive success of conspecifics on behavioural decisions of individuals regarding dispersal and recruitment. The experiment was carried out on Kittiwakes within a large seabird colony in northern Norway. It involved (i) the colour banding of several hundreds of birds; (ii) the manipulation (increase/decrease) of the local reproductive success of breeding groups on cliffpatches; and (iii) the detailed survey of attendance and activities of birds on these patches. It also involved the manipulation of the nest content of marked individuals breeding within these patches (individuals failing at the egg stage were expected to respond in terms of dispersal to the success of their neighbours). This allowed us to test whether a lower local reproductive success would lower (1) the attendance of breeders at the end of the breeding season; (2) the presence of prospecting birds; and (3) the proportion of failed breeders that came back to breed on the same patch the year after. In this paper, we discuss how we dealt with (I) the use of return rates to infer differences in dispersal rates; (II) the trade-off between sample sizes and local treatment levels; and (III) potential differences in detection probabilities among locations. We also present some results to illustrate the design and implementation of the experiment.

Full text

Page 1

Journal of Applied Statistics, Vol. 29, Nos. 1-4, 2002, 509- 520
Testing the eþect of conspeci®c reproductive
success on dispersal and recruitment
decisions in a colonial bird: design issues
THIERRY BOULINIER1, NIGEL G. YOCCOZ2, KAREN D. McCOY1,
KJELL EINAR ERIKSTAD2& TORKILD TVERAA2,
Universite  Pierre & Marie Curie, Paris, France, and
Research, Polar Environmental Center, N-9296, Tromsù, Norway
1Laboratoire d’Ecologie,
2Norwegian Institute for Nature
abstract
prominent role in the dynamics of populations. This is particularly the case for subdivided
populations where the dispersal of individuals among patches may lead to local extinction
and `rescue eþ ects’. A long-term observational study carried out in Brittany, France, and
involving colour-ringed Black-legged Kittiwakes (Rissa tridactyla) suggested that the
reproductive success of conspeci®cs (or some social correlate) could be one important factor
likely to aþ ect dispersal and recruitment. By dispersing from patches where the local
reproductive success was low and recruiting to patches where the local reproductive success
was high, individual birds could track spatio-temporal variations in the quality of breeding
patches (the quality of breeding patches can be aþ ected by diþ erent factors, such as food
availability, the presence of predators or ectoparasites, which can vary in space and time
at diþerent scales). Such an observational study may nevertheless have confounded the
role of conspeci®c reproductive success with the eþ ect of a correlated factor (e.g. the local
activities of a predator). In other words, individuals may have been in¯uenced directly by
the factor responsible for the low local reproductive success or indirectly by the low success
of their neighbours. Thus, an experimental approach was needed to address this question.
Estimates of demographic parameters (other than reproductive success) and studies of the
response of marked individuals to changes in their environment usually face problems
associated with variability in the probability of detecting individuals and with non-
independence among events occurring on a local scale. Further, very few studies on
dispersal have attempted to address the causal nature of relationships by experimentally
manipulating factors. Here we present an experiment designed to test for an eþ ect of local
Factors aþ ecting dispersal and recruitment in animal populations will play a
Correspondence: Thierry Boulinier, Laboratoire d’Ecologie, CNRS-UMR 7625, Universite  Pierre &
Marie Curie, 7 Quai St Bernard, F- 75005 Paris, France. E-mail: tboulini@snv.jussieu.fr
ISSN 0266-4763 print; 1360-0532 online/02/010509-12
© 2002 Taylor & Francis Ltd
DOI: 10.1080/02664760120108566

Page 2

510
T. Boulinier et al.
reproductive success of conspeci®cs on behavioural decisions of individuals regarding
dispersal and recruitment. The experiment was carried out on Kittiwakes within a large
seabird colony in northern Norway. It involved (i) the colour banding of several hundreds
of birds; (ii) the manipulation (increase/decrease) of the local reproductive success of
breeding groups on cliþpatches; and (iii) the detailed survey of attendance and activities
of birds on these patches. It also involved the manipulation of the nest content of marked
individuals breeding within these patches (individuals failing at the egg stage were expected
to respond in terms of dispersal to the success of their neighbours). This allowed us to test
whether a lower local reproductive success would lower (1) the attendance of breeders at
the end of the breeding season; (2) the presence of prospecting birds; and (3) the proportion
of failed breeders that came back to breed on the same patch the year after. In this paper,
we discuss how we dealt with (I) the use of return rates to infer diþ erences in dispersal
rates; (II) the trade-oþbetween sample sizes and local treatment levels; and (III) potential
diþ erences in detection probabilities among locations. We also present some results to
illustrate the design and implementation of the experiment.
1 Introduction
Using individually colour marked birds, most studies have either estimated survival
and related it to covariates using capture- recapture methodology, or experimentally
tested hypotheses related to life history evolution using return rates of individuals
as a surrogate for survival rate (Clobert, 1995). Relatively few experimental
studies have accounted for potential diþerences in detection probabilities among
individuals despite clear potential for bias (Martin et al., 1995; Boulinier et al.,
1997). Individuals may diþer in their probability of being detected for many
reasons, and notably because they may disperse or skip breeding. These reasons
can have important biological signi®cance, and experimental studies designed to
test what factors aþect these processes are needed.
Factors aþecting dispersal and recruitment in animal populations are likely to
play a prominent role in the dynamics of populations (Pulliam, 1996; Clobert
et al., 2001). This is particularly the case for subdivided populations where the
dispersal of individuals among patches may lead to local extinction and `rescue
eþects’ (Hanski & Gilpin, 1997). If patches of breeding habitat diþer in quality
and if the quality of these patches is temporally predictable, then it may be
particularly valuable for individuals to use information that can be sampled at the
end of a breeding season to decide where to settle for the next season (Boulinier &
Danchin, 1997). A long-term observational study carried out in Brittany, France,
and involving colour-ringed Black-legged Kittiwakes (Rissa tridactyla) suggested
that the reproductive success of conspeci®cs (or some social correlate) could be
one important factor aþecting dispersal and recruitment decisions (Danchin et al.,
1998). By dispersing from patches where the local reproductive success was low
and recruiting to patches where the local reproductive success was high, individual
birds could track spatio-temporal variation in the quality of breeding patches
(quality related to factors that can vary in space and time at diþerent scales such
as food availability, the presence of predators or ectoparasites). Such a process was
also suggested by the fact that the timing of prospecting on colonies by failed
breeders and young individuals corresponded to the period during which the best
information was available on the relative quality of breeding patches (Boulinier
et al. 1996) and that prospecting individuals were attracted to successful breeding
cliþs (Cadiou, 1999). Other recent studies have reported patterns consistent with

Page 3

Testing the eþ ect of conspeci®c reproductive success
511
the use of conspeci®c reproductive success for breeding patch selection and
dispersal (Doligez et al., 1999; Schjùrring et al., 2000; Brown et al., 2000).
Such observational studies may nevertheless confound the role of conspeci®c
reproductive success with the eþect of a correlated factor (e.g. the local activities
of a predator). In other words, individuals may have been in¯uenced directly by
the factor responsible for the low local reproductive success or indirectly by the
low success of their neighbours. Relatively few studies on dispersal have attempted
to address the causal nature of relationships by experimentally manipulating factors.
Here, we present an experiment designed to test for an eþect of local reproductive
success of conspeci®cs on behavioural decisions of individuals regarding dispersal
and local recruitment. We discuss the principle of this experiment, its implementa-
tion in the ®eld and the series of design issues it has raised.
2 Principle of the experiment and basic design
It is well known in many species that lower site ®delity is observed following
breeding failure than following breeding success (Switzer, 1997). Nonetheless, few
studies have addressed whether this was due to diþerential responses associated
with individual experiences or to diþerences in individual qualityÐlower quality
individuals being more likely to fail their reproduction and to die. A recent
experiment tested this; the reproductive success of a set of randomly chosen marked
individuals was decreased (by removing their eggs), and their return rate was
subsequently compared to the return rate of control individuals (Haas, 1998). In
the current experiment, we have applied the same type of approach. However, as
we were interested in testing for an eþect of conspeci®c reproductive success on
the decisions of individuals regarding recruitment and ®delity to local breeding
patches, we manipulated not only the reproductive success of individually colour-
marked birds, but also of groups of their neighbours on the same breeding patches.
The experiment was carried out on Kittiwakes within a large seabird colony in
northern Norway. It involved (i) the colour ringing of several hundreds of birds;
(ii) the manipulation (increase/decrease) of local reproductive success of breeding
groups on cliþ patches; and (iii) the detailed survey of attendance and activities of
birds on these patches. It also involved the manipulation of the nest content of
marked individuals breeding within these patches. This was done because we
predicted that individuals failing at the egg stage would respond in terms of
dispersal to the success of their neighbours. This experiment allowed us to test
whether a reduced local reproductive success would lower (1) the attendance of
breeders at the end of the breeding season; (2) the presence of prospecting birds;
and (3) the proportion of failed breeders that returned to breed on the same patch
the year after.
3 Design issues
3.1 Using return rates to compare dispersal rates
The basis for testing the main prediction of this experiment is that return rates to
local breeding plots can be used to infer diþerences in the proportion of individuals
that will be site-faithful versus individuals that will disperse (or not breed), and that
this response depends on the two diþerent treatments. As return rates are strongly
linked to survival rates (Martin et al. 1995), for such a comparison to make sense,

Page 4

512
T. Boulinier et al.
survival rates of birds submitted to the diþerent treatments should be assumed to
be the same, or at least the survival rate of birds from unsuccessful plots should
not be lower than that from successful plots. Survival rates of individuals from the
two treatment groups should not diþer if they are aþected in the same way by the
same series of factors. In our case, it was planned that the nest content of all
colour-marked individuals would be manipulated at the egg stage, thus this should
not have diþerentially aþected the survival of individuals between the two treatment
groups. In a recent experiment on the costs of reproduction in Kittiwakes, Golet
et al. (1998) removed egg clutches to reduce the cost of reproduction for individuals
with the prediction that these individuals would survive better than control indi-
viduals. Thus, in our case, removing eggs should not have decreased the survival
of manipulated birds due to diþerential breeding eþorts. The only diþerence
between the two treatment groups in our experiment is that their neighbours were,
or were not, put in failure at the egg stage. This could have aþected survival in
two ways: (1) the mortality risk on breeding patches could have varied due to
diþerences in the local attendance of individuals (group size eþect), and/or (2) the
diþerence in the activities of the two groups could have exposed them to diþerent
mortality risks outside breeding patches (for instance, prospecting for a new site
on diþerent patches may expose individuals to higher predation risks or parasitism).
Regarding the ®rst, it should be noted that successful and unsuccessful patches
within each plot pair were situated a maximum of 100 m away from each other,
and thus it is quite unlikely that exposure to adult mortality risk varied at that
scale. Regarding the second, diþerence in attendance and prospecting activities are
part of the predictions of the experiment. If these diþerent activities lead to
diþerential exposure to mortality risks, then this can be interpreted as part of the
diþerent costs and bene®ts of attempting to disperse. In other words, it is not a
problem for our prediction that a potential increase in mortality risk associated
with prospecting is confounded with `dispersal’ in determining the return rate of
individuals. One important advantage of such a design is that it requires the
monitoring of the activities of individuals on the study plots, but not outside, where
individuals are diýcult to locate and may vary dramatically in the probability of
being detected.
3.2 Accounting for detection probability
The fact that the probability of detecting marked individuals in the ®eld is rarely 1
and that it can diþer depending on the treatment is a critical issue to consider in
such an experiment. As explained above, in order to compare dispersal rates, we
have compared the proportion of marked birds that returned to breed (i.e. built a
nest) on their plot the year after manipulation for individuals put in failure on
successful plots and individuals put in failure on unsuccessful plots. Thus, the
experiment did not rely on the probability of ®nding individuals that dispersed or
did not breed in the year following the treatment, but conversely relied on the
ability to detect any breeding attempt by marked birds on the study plots during
the season following the treatment. Study plots were thus surveyed on regular
occasions (every three days) during the entire breeding season, and the presence
and behaviour of any marked bird was recorded. The probability of detecting
individuals involved in breeding is likely to vary among individuals (e.g. diþerence
in the proportion of time spent brooding such that colour rings are hidden in the
nest, or in the proportion of time spent at sea), but also with the timing of the

Page 5

Testing the eþ ect of conspeci®c reproductive success
513
season (e.g. nest building period versus brooding period) and the knowledge of the
observer (e.g. expectation of seeing a marked bird on a site where one has already
been recorded several times). In this context, Pollock’s robust design (Pollock, 1982)
applied between two primary sampling periods, the season of the experimental
manipulation and the following season, provided a way to estimate a rate of `locally
surviving and renesting’ while accounting for the probability of detecting individuals
using the pattern of detection versus non-detection of individuals over the series of
secondary sampling occasions within the second primary period. As we were
primarily interested in the proportion of individuals that had been recorded as
breeders and put in failure in the ®rst year and that came back and bred the second
year, we computed the ratio of the estimated number of individuals that came back
and bred the second year, estimated using a closed population model accounting
for individual detection probability using the pattern of detection of individuals
over the second breeding season, over the number of individuals that had been
recorded as breeders and put in failure in the ®rst year. This application of the
robust design is an analogue to the use of the design proposed recently for estimating
the rate of species extinction in animal communities between two points in time
(Nichols et al., 1998). We used the jackknife estimator of Burnham & Overton
(1979) for estimating the number of breeders, as this estimator permits hetero-
geneity in the probability of detecting individuals (model Mh), is relatively robust
to departures from assumptions, and the estimates and their associated standard
errors and 95% con®dence intervals can be computed directly using software
COMDYN (Hines et al., 1999; http://www.mbr-pwrc.usgs.gov/comdyn.html). This
was done independently for the group of failed individuals on successful plots
versus failed individuals on unsuccessful plots in the ®rst year. Exploratory analyses
using program CAPTURE to test for which closed population model was ®tting
best the data were carried out (Rexstad & Burnham, 1991). We did not attempt to
use capture- recapture models for open populations over several years following the
manipulation as our main interest was in what had happened in the year following
the manipulation on plots that could be heavily monitored.
3.3 Trade-oþbetween sample sizes and treatment levels
On plots where all individuals were put in failure, the sample size can be made of
a relatively large number of marked individuals. Conversely, on plots where the
treatment involved putting individuals in failure surrounded by successful indi-
viduals, the sample sizes were directly constrained by the level of success to be
maintained locally: the higher the sample size, the lower the local success, and thus
the lower the level of the treatment applied within a given patch. In the case of our
experiment, the trade-oþ between sample sizes and treatment levels had to have
been taken into account when considering what sample sizes would be required to
have the ability to detect an eþect of the treatment. We discuss below this issue by
considering the relationship between the variance of the coeýcient of association
between the treatment and the return rate, the sample size and the level of treatment.
In a simple linear regression model, the variance of the regression coeýcient is
given by (r2/nr2
variance), n is the number of observations and r2
variable X (if X is ®xed, i.e. not a random variable, as in an experiment with ®xed
treatments, it is not a true variance, but a measure of the variability of the
treatments). This expresses the fact that the variance of the regression coeýcient
x), where r2is the variance of the response variable Y (the residual
x the `variance’ of the predictor

Page 6

514
T. Boulinier et al.
will decrease as the sample size n increases and as r2
treatments), increases, and that these have identical eþects on the variance of the
regression coeýcient.
Here, however, we are dealing with a binomial response variable and must
consider a logistic regression model. In a logistic regression model, the main
diþerence comes from the fact that the variability of the response variable is now
dependent on the expected proportion piand the number of birds in each treatment
group mi: the binomial variance is equal to mipi(1 2 pi), i 51, . . ., n being the
diþerent groups of birds.
Thelogistic regression model is written
approximate variance-covariance matrix of the regression coeýcients is given by
(McCullagh & Nelder, 1989):
x (the variability of the
as Logit(p) 5 b0 + b1x, andthe
cov(b) 5(X
TW X )
21
where X is the design matrix and W a diagonal matrix containing the weight
associated with the binomial variance:
W 5diag{mipi(1 2 pi)}
In the simplest case of two proportions, p1 and p2, corresponding to two values of
the predictor variable x1and x2and the number of birds m1and m2in each group,
we have:
X 5(1x1)
(1x2)
A bit of algebra leads to a simple formula for the variance of the regression
coeýcient b1:
var(b1) 5(w1 + w2)/[w1w2(x12 x2)2]
where wi are the binomial weights equal to mipi(1 2 pi).
Whereas the term (x12 x2)2is common to the linear regression case, the weights
wire¯ect the fact that the precision of the regression coeýcient will get lower when
observed proportions get close to 0 or 1. This is illustrated in Fig. 1.
What can be seen from Fig. 1 is that the ability to detect an eþect (i.e. a
decreasing function of the variance of the regression coeýcient) is much aþected
by the expected eþect of the treatment (compare Fig. 1(a) with 1(b)), and by the
sample size and treatment level. When the expected eþect of the treatment is weak,
several combinations of sample size and treatment levels provide the same ability
to detect an eþect with the constraint that there is a trade-oþ between increasing
sample size and increasing treatment level (which can take the form, for instance,
of the dotted line in Fig. 1(a)). In that case, a low treatment level and large sample
size would do as well as a high treatment level and low sample size. When the
expected eþect of the treatment is strong, the combinations of sample size and
treatment level that provide a good ability to detect an eþect are when the level of
the treatment takes a relatively minimal value; sample sizes required an increase
on either side of this treatment level (Fig. 1(b)). Another element to be considered
when implementing such an experiment in the ®eld is that the `treatment’ level
can be modi®ed by natural causes, e.g. predation on the successful plots, which
cannot always be compensated for (for example by replacing with eggs from
manipulated nests).

Page 7

Testing the eþ ect of conspeci®c reproductive success
515
10
20
30
40
50
60
m2
0.0
0.2
0.4
0.6
0.8
1.0
x2
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
(a)
10
20
30
40
50
60
m2
0.0
0.2
0.4
0.6
0.8
1.0
x2
0.5
1
1.5
1.5
2
2.5
3
3.5
(b)
Fig. 1. Variance of the regression coeýcient b1 for diþerent combinations of the treatment value (x2)
and the number of observations (m2). Figures (a) and (b) illustrate two contrasting cases: (a) weak
treatment eþect b1 5 2 0.5; (b) strong treatment eþect b1 5 2 3. x is assumed to be equal to 0, the
intercept is taken as 2 giving a return rate for the control group equal to 1/(1 + exp( 2 2)) 5 0.88. A
strong treatment eþect would result for x 51 in a return rate equal to 1/(1 + exp( 2 2 + 3 3 1)) 50.27,
a weak treatment eþect in a return rate equal to 0.82. Dotted line: possible shape of the trade-oþ
between increasing sample size and increasing treatment level.
3.4 Experimental unit and non-independence among individuals
In this study, the experimental unit is a plot, and not the individuals within a plot.
With treatments applied following a randomized block design, one plot per pair of
plots being put in failure and the other being left/maintained in success, analyses
can be done using hierarchical generalized linear models (e.g. Lindsey, 1999).
There will be evidence for an eþect of the local reproductive success on the
attendance of individuals on failed sites, the number of prospecting birds or the
probability of returning to breed if there is a consistent eþect of the treatment
among plots.

Page 8

516
T. Boulinier et al.
4 Implementation in the ®eld
4.1 Study population
The study was conducted on the island of Hornùya (70ë22¢N, 31ë10¢E), in the
north-east of Norway, where more than 20 000 pairs of Kittiwakes breed (Erikstad
et al,. 1995). In the ®rst year of study (1998), we colour marked individually more
than 400 adult birds on a series of breeding cliþs. The cliþs were split into a series
of paired plots that were surveyed to assess the local reproductive success, the
attendance of birds on built and non-built sites, and the presence and activities of
marked individuals throughout the breeding seasons (visit to each plot every three
days). The experimental manipulation of the local reproductive success was carried
out in 1999, and in 2000 a special eþort was made early in the season (March-
April) to determine which individuals came back and bred on the plots.
4.2 Experimental treatment
Within each pair of study plots, one plot was randomly chosen to be put in failure
and the other was left/maintained in success. Manipulations of nest contents were
performed at the egg stage, where we expected the strongest response (Danchin
et al., 1998). Within plots to be put in failure, eggs were removed from all nests.
Within plots to be left in success, only a limited number of nests with colour
marked breeders were put in failure. If birds in failed nests laid a replacement
clutch, these eggs were also removed. Eggs were either used to maintain success in
other `successful’ nests or kept for analyses of yolk content as part of a parallel
study on maternal investment in eggs (Gasparini et al., 2001). Care was taken to
spend the same amount of time at each plot within a plot pair so that disturbance
would be comparable. By moving eggs into nests that had just failed, usually due
to heavy local predation by ravens Corvus corax, it was possible to cancel local
failures and maintain relatively high success on some plots. Natural failures at
diþerent stages lead nevertheless to diþerent local levels of success on `successful’
plots (the average number of chicks produced per nest on `successful’ plots varied
from 0.4 [plot 1IR] to 1.2 [plot 3E]). These levels were comparable to natural
levels observed in cliþs `in success’.
5 Preliminary results
The results presented here are preliminary and are given as an illustration of the
above discussion. They are not presented as de®nitive evidence for the patterns
that may appear, but need to be con®rmed by a more complete analyses of the
data from the described experiment. In particular, results presented concern only
®ve of the nine pairs of study plots.
5.1 Attendance of failed breeders on successful and unsuccessful plots
The attendance of individuals during the chick rearing period (after day 180) on
failed nest sites was higher on plots put in failure than on plots left/maintained in
success (ANOVA, randomized block design analysis, with plot pair as a random
factor, treatment as a ®xed eþect: F1,203 5500.23, P 50.001). We veri®ed that such
a diþerence did not occur before the egg removal treatment was applied

Page 9

Testing the eþ ect of conspeci®c reproductive success
517
10
20
30
40
50
160
60
m2
0.0
0.2
0.4
0.6
0.8
x2
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
(a)
10
Date
20
30
40
50
Date
60
0.0
140
0.2
0.4
0.6
0.8
1.0
x2
0.5
200
1
160
1.5
1.5
2
2.5
3
3.5
(b)
Date
120140 160180 200
Mean number of individuals
per nest
0
1
2
120140180200
0
1
2
Date
3EL / 3ER
3F / 3G
120180
0
1
2
Mean number of individuals
per nest
1 A / B
120140160180200
0
1
2
1 GL / GR
Fig. 2. Average number of individuals attending failed nest sites for four pairs of plots during the 1999
breeding season (pairs `3EL/3ER’, `3F/3G’, `1A/1B’, and `1GL/1GR’). Nest contents were manipulated
after day 145. j are for successful plots and s are for unsuccessful plots.
(F1,63 52.73, P 50.1033). These results are illustrated in Fig. 2, where the average
number of individuals attending a failed nest site is given for four pairs of plots
during the 1999 breeding season.
5.2 Number of prospecting birds on successful and unsuccessful plots
The number of birds attending non-built nest sites during the 1999 season gives
an indication of the temporal and spatial pattern of prospecting: for the two pairs
of plots taken as examples, prospecting on non-built sites was highest during the
chick rearing period and occurred mostly on successful plots (Fig. 3). Observations
of the behaviour of ringed birds showed that most birds on non-built sites were
prospecting birds and not oþ-duty birds resting on a site close to their nest. These
results are fully consistent with the observations reported by Boulinier et al. (1996)
regarding the timing of prospecting, and by Cadiou (1999) regarding the apparent
attraction of prospecting individuals to successful plots at the time of chick rearing.
5.3 Return rates of breeding birds on successful and unsuccessful plots
Using program CAPTURE, it appeared that model Mth, which makes the assump-
tion of a heterogeneity among individuals in the probability of being detected and

Page 10

518
T. Boulinier et al.
Date
120140 160180 200
Number of birds on non-built sites
0
2
4
6
Date
120 140160180200
0
2
4
6
3 G
3 F
Date
120140160180 200
0
2
4
6
8
10
12
14
16
Number of birds on non-built sites
Date
120 140 160 180 200
0
2
4
6
8
10
12
14
16
1 B1 A
Fig. 3. Prospecting: number of individuals attending non-built sites during the breeding season for two
pairs of plots (pair `1A/1B’ and pair `3G/3F’). Nest contents were manipulated after day 145. Most
prospecting occurred during the chick rearing period (i.e. after day 170). j are for successful plots and
s are for unsuccessful plots.
a time eþect on the probability of being detected, would be the most appropriate,
even when considering only observations carried out over the beginning of the
season. Results were nevertheless very similar when using model Mh, which is the
estimator implemented in program COMDYN. Using the corresponding jackknife
estimator, the estimated probability of detecting a breeding individual returning as
a breeder in 2000 did not diþer between the two groups, as it was 0.951
(95% c.i. 50.847 2 1.00) and 0.932 (95% c.i. 50.820 2 1.00) for the series of
unsuccessful and successful plots, respectively. Accounting for that probability of
detection, the overall proportion of experimentally failed individuals that returned
to bred on unsuccessful plots was 0.49 (95% c.i. 50.36 2 0.67) compared to 0.77
(95% c.i. 50.55 2 1.0) on successful plots. Looking within each plot pair (Table 1),
we see a trend suggesting a higher proportion of individuals returning to breed on
the successful plot (observed in four of the ®ve plot pairs). Several individuals that
did not return to breed on their plot were recorded prospecting or breeding on
other plots. This observation suggests that, indeed, a portion of birds that did not
return to breed on the same plot may have been in the process of dispersing.

Page 11

Testing the eþ ect of conspeci®c reproductive success
519
Table 1. Proportion of marked failed individuals that returned
to breed on the same plot the year after. Results are presented
for ®ve plot pairs, with the proportion on the plots that were
left/maintained in success on the left, and the proportion on
the plots that were put in failure on the right.
Proportion of failed individuals that returned
to breed on the same plot
Plot pairUnsuccessful plots Successful plots
1A/1B
2A/2B
1GL/1GR
3EL/3ER
3F/3G
Total
5/10
9/13
10/18
13/27
1/13
38/81
3/3
2/4
4/5
2/3
2/3
13/18
6 Discussion
Despite the importance of dispersal in evolutionary ecology and for understanding
the dynamics of subdivided populations, very few experimental studies have
attempted to test what factors may aþect dispersal in birds. The current work
shows that, at least at some spatial scales, it may be possible to use an experimental
approach to study factors aþecting dispersal. Such studies need speci®c designs
involving multi-sites. The advance in capture- recapture methodology clearly should
help analysing data on dispersal in such settings, but it should be stressed that
speci®cally designed observational and experimental studies are required to answer
such questions (Nichols & Kendall, 1995). The description of our experimental
approach testing the role of conspeci®c reproductive success (or some social
correlate) for breeding habitat selection shows that such experiments are not
obvious to carry out, but may provide crucial information on a wide array of
important biological processes linking the behaviour of individuals to the dynamics
of subdivided populations in ecological landscapes (Lima & Zollner, 1996). Regard-
ing the speci®c questions addressed by our experiment, a con®rmation of the
preliminary results presented here would have implications for our understanding
of factors aþecting the dynamics of individuals in space and time in relation to
environmental changes. This would have, in turn, implications for the understand-
ing of evolution ecology questions, such as the evolution of colonial breeding in
birds, but also for studies on the conservation of species living in fragmented
habitats.
Acknowledgements
We are grateful to Rob Barrett, Julien Gasparini, Ste Âphanie Jenouvrier, Armelle
Renaud, Yvan Richard and Solveig Schjùrring for their assistance in the ®eld. This
work bene®ted from support by the CNRS, the French Polar Institute (Programme
no. 333 of the `Institut Franc Ëais pour la Recherche et la Technologie Polaires’),
Franco-Norwegian Scienti®c Cooperation Programmes (`Aurora’ and PICS), and
NSERC, Canada (to KM). Our thanks are also due to the Norwegian Lighthouse
Authorities and the Finnmark County authorities for permission to work and live
on Hornùya. Permits to manipulate nest contents of Kittiwakes were granted by
the Norwegian Authority for Animal Experimentation.

Page 12

520
T. Boulinier et al.
REFERENCES
Boulinier, T. & Danchin, E. (1997) The use of conspeci®c reproductive success for breeding patch
selection in territorial migratory species, Evolutionary Ecology, 11, pp. 505-517.
Boulinier, T., Danchin, E., Monnat, J.-Y. Doutrelant, C. & Cadiou, B. (1996) Timing of prospecting
and the value of information in a colonial breeding bird, Journal of Avian Biology, 27, pp. 252-256.
Boulinier, T., Sorci, G., Clobert, J. & Danchin, E. (1997) An experimental study of the costs of
reproduction in the Kittiwake Rissa tridactyla: Comment, Ecology, 78, pp. 1284- 1287.
Brown, C. R., Bomberger Brown, M. & Danchin, E. (2000) Breeding habitat selection in cliþ
swallows: the eþect of conspeci®c reproductive success on colony choice, Journal of Animal Ecology,
69, pp. 133- 142.
Burnham, K. P. & Overton, W. S. (1979) Robust estimation of population size when capture
probabilities vary among animals, Ecology, 60, pp. 927- 936.
Cadiou, B. (1999) Attendance of breeders and prospectors on cliþs re¯ects the quality of breeding
places in the kittiwake Rissa tridactyla, Ibis, 141, pp. 321-326.
Clobert, J. (1995) Capture- recapture and evolutionary ecology: a diýcult wedding? Journal of Applied
Statistics, 22, pp. 989- 1008.
Clobert, J., Danchin, E., Dhondt, A. & Nichols, J. D. (Eds) (2001) Dispersal (Oxford, UK, Oxford
University Press).
Danchin, E., Boulinier, T. & Massot, M. (1998) Habitat selection based on conspeci®c reproductive
success: implications for the evolution of coloniality, Ecology, 79, pp. 2415- 2428.
Doligez, B., Danchin, E., Clobert, J. & Gustafsson, L. (1999) The use of conspeci®c reproductive
success for breeding habitat selection in a non-colonial, hole-nesting species, the collared ¯ycatchers,
Journal of Animal Ecology, 68, pp. 1193- 1206.
Erikstad, K. E., Tveraa, T. & Barrett, R. T. (1995) Adult survival and chick production in long-lived
seabirds: a 5-year study of the kittiwake Rissa tridactyla. In: H. R. Skjoldal, C. Hopkins, K. E.,
Eriksta, D. & H. P. Leinaas (Eds), Ecology of Fjords and Coastal Waters, pp. 471-477 (Amsterdam,
the Netherlands, Elsevier).
Gasparini, J., McCoy, K. D., Haussy, C. Tveraa, T. & Boulinier, T. (2001) Induced maternal
response to the Lyme disease spirochete Borrelia burdgorferi sensu lato in a colonial seabird, the
Kittiwake Rissa tridactyla, Proceedings of the Royal Society, London B, 1467, pp. 647-650.
Golet, G. H., Irons, D. B., Estes, J. A. (1998) Survival costs of chick rearing in black- legged
kittiwakes, Journal of Animal Ecology, 67, pp. 827-841.
Hanski, I. & Gilpin, D. (Eds) (1997) Metapopulation Biology (San Diego, California, Academic Press).
Hass, C. (1998) Eþects of prior nesting success on site ®delity and breeding dispersal: an experimental
approach, Auk, 115, pp. 929- 936.
Hines, J. E, Boulinier, T., Nichols, J. D., Sauer, J. R. & Pollock, K. H. (1999) COMDYN, software
for the dynamic of animal communities using a capture- recapture approach, Bird Study, 46,
pp. S209- 217.
Lima, S. L. & Zollner, P. A. (1996) Towards a behavioural ecology of ecological landscapes, Trends
in Ecology and Evolution, 11, pp. 131-135.
Lindsey, J. K. (1999) Models for Repeated Measurements 2nd edn (Oxford, Oxford University Press).
McCullagh, P. & Nelder, J. A. (1989) Generalized Linear Model (London, Chapman & Hall).
Martin, T. E., Clobert, J. & Anderson, D. R. (1995) Return rates in studies of life history evolution:
are biases large? Journal of Applied Statistics 22, pp. 863- 875.
Nichols, J. D. & Kendall, W. L. (1995) The use of multi-state capture-recapture models to address
questions in evolutionary ecology, Journal of Applied Statistics, 22, pp. 835-846.
Nichols, J. D., Boulinier, T., Hines, J. E. Pollock, K. H. & Sauer, J. R. (1998) Estimating rates of
species extinction, colonization and turnover in animal communities, Ecological Applications, 8,
pp. 1213- 1225.
Pollock, K. H. (1982) A capture- recapture sampling design robust to unequal catchability, Journal of
Wildlife Management, 46, pp. 752- 757.
Pulliam, H. R. (1996) Sources and sinks: empirical evidence and population consequences. In: O. E.
Rhodes Jr., R. K. Chesser & M. H. Smith (Eds), Population Dynamics in Ecological Space and Time
(Chicago, The University of Chicago Press).
Rexstad, E. & Burnham, K. P. (1991) User’s Guide for Interactive Program CAPTURE. Abundance
Estimation of Closed Animal Populations (Fort Collins, Colorado, Colorado State University).
Schjørring, S., Gregersen, J. & Bregnballe, T. (2000) Sex diþerence in criteria determining ®delity
towards breeding sites in the great cormorant, Journal of Animal Ecology, 69, pp. 214- 223.
Switzer, P. V. (1997) Past reproductive success aþects future habitat selection, Behavioral Ecology and
Sociobiology, 40, pp. 307- 317.