Patterns of incubation and nest-site attentiveness in relation to organochlorine (PCB) contamination in glaucous gulls


  • J.O. Bustnes,

    Corresponding author
    1. Norwegian Institute for Nature Research, Division of Arctic Ecology, The Polar Environmental Centre, N-9296 Tromsø, Norway;
      Jan Ove Bustnes, Norwegian Institute for Nature Research, Division of Arctic Ecology, The Polar Environmental Centre, N-9296 Tromsø, Norway (fax 47 77 75 04 01;
    Search for more papers by this author
  • V. Bakken,

    1. Norwegian Polar Institute, c/o University of Oslo, Zoological Museum, Sarsgt. 1, N-0562 Oslo, Norway; and
    Search for more papers by this author
  • K.E. Erikstad,

    1. Norwegian Institute for Nature Research, Division of Arctic Ecology, The Polar Environmental Centre, N-9296 Tromsø, Norway;
    Search for more papers by this author
  • F. Mehlum,

    1. Norwegian Polar Institute, c/o University of Oslo, Zoological Museum, Sarsgt. 1, N-0562 Oslo, Norway; and
    Search for more papers by this author
  • J.U. Skaare

    1. National Veterinary Institute, PO Box 8156 Dep., N-0033 Oslo, Norway
    Search for more papers by this author

Jan Ove Bustnes, Norwegian Institute for Nature Research, Division of Arctic Ecology, The Polar Environmental Centre, N-9296 Tromsø, Norway (fax 47 77 75 04 01;


  • 1Although experimental studies show that organochlorines (OC) can affect bird behaviour, field assessments are invariably confounded by ecological differences between contaminated and uncontaminated sites. The behaviour of individual birds in the field has rarely been related to the contaminant burden.
  • 2We examined individual patterns of incubation and nest-site attentiveness in relation to OC burden, measured as polychlorinated biphenyl (PCB) concentration in the blood, of 27 glaucous gulls Larus hyperboreus in two breeding areas at Bear Island, in the north-eastern Atlantic.
  • 3Blood PCB concentrations ranged from 52 ng g-1 to 1079 ng g-1 (wet weight). There were significant differences between the two breeding areas, and females had significantly lower concentrations than males.
  • 4Gull behaviour differed significantly between breeding areas and sexes independently of PCB. Females incubated more than males (54% vs. 46%) but spent more time away from the nest site than males, both overall (23% vs. 12%) and when not incubating (50% vs. 21%). They were also absent for longer periods (4·5 vs. 2·8 h). Moreover, length of incubation bouts (6·4 vs. 4·4 h), the amount of time absent from the nest site when not incubating (51% vs. 25%) and length of absences (5·6 vs. 1·8 h) differed between breeding areas, probably due to different feeding specializations.
  • 5After controlling for these area and sex effects, the proportion of time absent from the nest site when not incubating, and the number of absences, were both significantly related to blood concentration of PCB.
  • 6Increased absence from the nest site in individual glaucous gulls with high blood concentrations of OC suggests effects on reproductive behaviour. We speculate that endocrine disruption or neurological effects might be involved, leading to increased energetic costs during incubation and reduced reproductive output.


In ecotoxicology a central question is whether pollution impairs normal behaviour in wildlife populations. A general conclusion from a wide variety of experimental and observational studies is that heavy metals and various organochlorines (OC) induce changes in behaviour, but there is little evidence of serious threats to populations (reviewed by Peakall 1985, 1992, 1996). However, the laboratory tests commonly used to quantify the behavioural effects of contaminants lack ecological realism (Peakall 1996). There is thus a great need to study behaviours that can be linked to population parameters, such as reproduction and survival. This includes the ability of birds to conduct complex behaviours such as food gathering, pairing or chick rearing. For instance, the nesting period is critical in avian reproduction because both the offspring and the parents are vulnerable to predation and starvation. Any changes in behaviours related to incubation and nest defence might have an effect on reproductive output (Monaghan & Nager 1997; Thomson, Monaghan & Furness 1998).

OC such as dichlorodiphenyltrichloroethane (DDT) and polychlorinated biphenyls (PCB) are global contaminants and have a variety of negative effects on wildlife. At high concentrations such OC are lethal, but at sublethal levels they may have neurological effects (Peakall 1985; Hoffman, Rice & Kubiak 1996) or various effects on the endocrine system (Colborn, vom Saal & Sato 1993). Endocrine disruption may affect reproductive behaviour because the breeding cycle is hormonally controlled (Farner & Wingfield 1980). The behavioural effects of OC in birds include decreased parental attentiveness, impaired courtship behaviour, and subtle neurological effects such as impaired avoidance behaviour. Some studies have shown correlations between OC such as dichlorodiphenyldichloroethylene (DDE) and PCB and the parents’ ability to incubate and defend the nest, but the results have not been consistent (Fyfe, Risebrough & Walker 1976; Fox & Donald 1980; Peakall et al. 1980; Kubiak et al. 1989; Barron, Galbraith & Beltman 1995; Peakall 1996). Furthermore, a problem with many field studies is that the pollutant burdens of the individuals involved have not been known. Instead behaviours are often compared between heavily polluted sites and less polluted reference areas (Fox et al. 1978; McCarty & Secord 1999). This approach does not control for differences in environmental factors between areas, such as feeding conditions and predation, which often influence behaviour. Behavioural differences due putatively to environmental contaminants might in reality be caused by other factors.

To our knowledge no field study has measured OC contaminant levels and recorded the behaviour of the same individuals. In this study, we examined how the reproductive behaviour of individual glaucous gulls Larus hyperboreus Gunnerus 1767 varied in relation to their blood PCB concentration. We used PCB because they are the most common OC in the glaucous gull (Bourne & Bogan 1972; Gabrielsen et al. 1995; Sagerup et al. 2000). Even if this approach is also correlational, it allows control for covariates such as sex and breeding area. The experimental poisoning of wild birds would allow similar control, but it would carry ethical and logistical problems.

The study was carried out during 1998 at Bear Island (Fig. 1) in the north-eastern Atlantic, where glaucous gulls are heavily contaminated by PCB. Aberrant behaviours and unexplained deaths have been reported among glaucous gulls on the island, and it is assumed that the high levels of PCB are contributing factors (Bourne & Bogan 1972; Gabrielsen et al. 1995).

Figure 1.

The two study areas at Bear Island, north-eastern Atlantic.

In glaucous gulls, both males and females participate in the care of offspring, including incubation, nest defence and feeding (Cramp & Simmons 1983). They are predatory, breeding in colonies where unattended eggs and small young are often eaten by conspecifics (Ferguson-Lees 1963; Barry & Barry 1990). This necessitates close co-ordination of incubation shifts and feeding trips between mates to ensure offspring survival. If one parent fails to synchronize its behaviour, the mate may be forced to leave the nest unattended to feed. In addition, glaucous gulls are very aggressive at the nest and parents that are present but not incubating (nest-site attentiveness) show strong protective behaviour (Løvenskiold 1954; Ferguson-Lees 1963). We assume that the time spent incubating and the time spent attending the nest site when not incubating are good measurements of reproductive investments, and our interpretation is that low incubation effort and/or frequent absence from the nest, associated with high PCB levels, are adverse effects.

To unravel the effects of PCB on nesting behaviour we recorded the proportions of the time that individuals with various burdens incubated, and measured the length and number of their incubation intervals. We also recorded nest-site attentiveness as the proportion of time spent away from the nest when not incubating, and measured the length and number of absences. In all cases we controlled for confounding variables.

Study area and methods

Bear Island (74°30′ N, 19°01′ E) (Fig. 1) is 178 km2 and the topography is described by Bakken & Mehlum (1988). The southern part holds one of the largest concentrations of breeding seabirds in the north-eastern Atlantic, and especially important are the populations of common guillemots Uria aalge Pontoppidan 1763, Brünnichs guillemots Uria lomvia Linnaeus 1758, kittiwake Rissa tridactyla Linnaeus 1758, fulmar Fulmarus glacialis Linnaeus 1761 and little auk Alle alle Linnaeus 1758, in addition to about 2000 pairs of glaucous gulls (Bakken & Mehlum 1988). Two distinct breeding areas of glaucous gull were selected for behavioural observation (Fig. 1). One area (Revdalen) was situated on the edge of the main seabird cliffs (hereafter referred to as the seabird cliff), about 100–150 m above sea level. The second area (Kvalrossbukta) was located about 2–2·5 km from the main seabird cliffs and < 50 m above sea level (hereafter referred to as sea level) (Fig. 1). In both areas the birds are colonial and not territorial, except for the nest site.

All gulls included in this study were caught in both 1997 and 1998, using a nest trap. The trap consisted of a snare placed on the edge of the nest bowl and attached to an elastic rope. The trap was triggered by a radio-transmitter. In 1997, the birds were equipped with alpha-coded PVC rings and numbered steel leg rings for individual identification. In 1998 they were caught between 3 and 19 (mean = 12·2 ± 4·3 SD) days after behavioural observations. Blood was sampled from the wing vein (c. 10 ml) and the birds were weighed to the nearest 10 g. In gulls the best morphometric measurement reflecting body size is skull length (head + bill). As a body condition index we used body weight/(head + bill)3 (Coulson et al. 1983).

In 1998, nests were marked with numbered flags and observed from the start of egg laying until hatching. Incubation stage was defined as the number of days between the laying of the first egg and the day of behavioural observations. Clutch size was recorded as the number of eggs in the nest at the time of behavioural observation.

Sex determination

Sex was determined by size, males being larger than females (Cramp & Simmons 1983). As there was no overlap between males and females in the skull length and bill length (head + bill) on Bear Island, we assumed that birds with a combination of bill longer than 61·5 mm and head + bill longer than 142 mm were males (E. O. Henriksen, unpublished data).

Behavioural observations

The behavioural observations were carried out from 31 May to 5 June, when the eggs had been incubated for between 1 and 20 days (mean = 10·7 ± 0·66 SE). Glaucous gulls usually incubate for longer continuous periods of several hours. The non-incubating partner, when not on feeding trips, sits close to the nest responding aggressively to external threats. Records of whether the marked birds were incubating, attending the nest site or absent from the nest site (nest-site attentiveness) were made every hour, for 48 h. The area surrounding the nest was observed if the marked birds were not observed by the nest. This produced a data set where birds were incubating (1) or not (0), which was used to analyse the proportion of time spent incubating. Our observations also produced a data set where non-incubating individuals were present at the nest (1) or absent (0), which was used to analyse the proportion of time spent absent.

We analysed the mean length of incubation bouts and the mean absence length for each individual. Ideally, only incubation intervals and absences that were started and terminated within the observation period should be used. However, the probability of an incubation bout or absence being excluded from the analyses increases with increasing length of the bout, leading to an underestimation of the actual mean lengths. To reduce this problem we calculated, for each of the two breeding areas, the mean lengths of incubation bouts and absences starting and terminating within the observation period. We then included in the analyses all the incubation bouts and absences begun before the start of the observation, or terminated after the end of the observation period, that were longer than the mean length of the bouts for each breeding area. Thus, if a bird was incubating when we started observations, and continued for 10 h, while the mean in the breeding area was 5 h, the observation would be included in the data set. If it was less than 5 h, it was excluded.

Contaminant analyses

The PCB analyses were carried out at the Environmental Toxicology Laboratory at the Norwegian School of Veterinary Science, Oslo. The methods for determination of OC in plasma and blood cells (Bernhoft, Wiig & Skaare 1997) were used with some modifications: samples of whole blood (c. 8 g) were weighed, spiked with internal standards (PCB-29, PCB-112, PCB-207), and extracted twice with cyclohexane and acetone. The percentage extractable fat was determined gravimetrically. Thereafter the extracted fat was redissolved in 1 ml cyclohexane and a clean-up was done with ultra-pure sulphuric acid.

The extracts were automatically injected (Fisons Autosampler AS 800) on a Carlo Erba HRGC 53 Mega Series (Carlo Erba Instrumentation, Milan, Italy) gas chromatograph, equipped with a split/splitless injector and an electron capture Ni63 detector (ECD). Two columns of differing polarity and selectivity were used to obtain the desired chromatographic separation (SPB-5 and SPB-1701, 60 m, 0·25 mm internal diameter and 0·25 µm film layer; Supelco Incorporation, Bellefonte, PA), both connected to a 1-m deactivated pre-column. Details on the gas chromatographic analyses are given in Bernhoft, Wiig & Skaare (1997). The following eight PCB congeners were determined (International Union of Pure and Applied Chemistry numbers; Ballschmiter & Zell 1980): 28, 101, 99, 118, 138, 153, 170 and 180. Other compounds analysed included hexachlorbenzene (HCB), pp′-dichlorodiphenyldichloroethylene (pp′-DDE) and oxychlordane.

The laboratory has, since 1996, been accredited by the Norwegian Accreditation as the testing laboratory for OC according to the requirements of NS-EN45001 and ISO/IEC Guide 25.

The detection limits for individual compounds were determined as three times the noise level. Detection limits for individual components were between 0·004 and 0·026 ng g−1 wet weight. Quantification was performed using PCB-29, -112 and -207 as internal standards in each sample. Percentage recoveries and CV of individual PCB in spiked sheep blood varied from 75 to 118 and 3·4 to 18·1, respectively (n = 8), which are in the acceptable range set by the laboratory’s quality control system. Reproducibility was continuously tested by analysing the PCB levels in the laboratory’s own reference sample (seal blubber). The results were plotted on a control chart and were within ± 2σ (103·9 ± 11·3% of true value, CV = 10·9%, n = 7). Blank samples were included in each series to test for interference. Certified international reference materials (Community Bureau of Reference, European Union, Brussels) (CRM 349 and 350, International Council for Exploration of the Sea cod liver oil and mackerel oil) were analysed regularly with results within the given ranges.

The correlations between the congeners and the sum of PCB were extremely high (28: R2 = 0·818; 99: R2 = 0·967; 118: R2 = 0·975; 138: R2 = 0·995; 153: R2 = 0·998; 170: R2 = 0·980; 180: R2 = 0·975), except for PCB-101 (R2 = 0·154) which only made up on average 0·37% of the sum of the PCB. To extract the effects of single PCB congeners from the effects of the sum of the PCB would thus be impossible, and we therefore used the sum of the PCB as the measure of contamination burden. Blood concentrations of other compounds (Table 1) were several times lower than PCB (4–5 times for DDE, 25–30 times for HCB and 25–30 times for oxychlordane) but were also very highly correlated with PCB (R2 = 0·87 for DDE, R2 = 0·82 for HCB and R2 = 0·91 for oxychlordane).

Table 1.  Concentrations of organochlorines in blood (ng g−1, wet weight), incubation and nest-site attentiveness (mean ± SE) of glaucous gulls from two different breeding areas, on the seabird cliff (Revdalen) and close to the sea level (Kvalrossbukta), at Bear Island
 Seabird cliffSea level
 Female (n = 7)Male (n = 6)Female (n = 9)Male (n = 5)
 PCB317·4 ± 53·0630·8 ± 101·7185·1 ± 55·8240·6 ± 53·6
 DDE 76·8 ± 13·1148·4 ± 25·1 40·8 ± 9·7 55·8 ± 11·4
 HCB 20·2 ± 2·9 28·9 ± 3·9  7·2 ± 1·2  8·2 ± 2·2
 Incubation 12·6 ± 1·6 21·4 ± 1·9  8·6 ± 2·4  9·2 ± 1·9
 Percentage of time incubating 55·7 ± 2·3 49·3 ± 1·8 53·5 ± 3·9 40·8 ± 5·2
 Bout length (h)  3·5 ± 0·2  5·3 ± 1·2  7·0 ± 0·9  5·3 ± 0·3
 Number of bouts during 48 h  7·9 ± 0·6  5·5 ± 1·0  4·4 ± 0·4  4·4 ± 0·6
Nest-site attentiveness
 Percentage of time absent when not incubating 29·7 ± 9·4 19·6 ± 8·7 66·4 ± 7·2 22·2 ± 11·6
 Absence length (h)  1·6 ± 0·2  2·0 ± 0·6  6·4 ± 1·4  3·8 ± 1·8
 Number of absences during 48 h  3·7 ± 1·1  3·2 ± 1·5  3·0 ± 0·3  1·5 ± 0·6

We used the concentration of PCB in blood wet weight as a measure of PCB burden, because wet weight is usually considered most relevant for potential toxic effects (for discussions see Klaasen & Eaton 1991; Bignert et al. 1993; Henriksen, Gabrielsen & Skaare 1996). Other studies suggest a relatively high short- and long-term stability of the blood levels of OC in incubating glaucous gulls, indicating that blood concentration is a reliable, relative (compared with other individuals), measurement for body burden of an individual (Henriksen, Gabrielsen & Skaare 1998; Bustnes et al. 2001).


Statistical analyses were carried out using SAS (SAS 1990) and S-plus (Venables & Ripley 1999). PCB values were log10 transformed to approximate normal distribution.

To get robust estimates of the effect of size of the different variables, including PCB, and to compare these estimates between groups, we used parametric covariance analyses (Piegrosh & Bailer 1997). The dependent variables were (i) the proportion of the time each individual incubated and (ii) the proportion of the time non-incubating individuals were away from the colony. They were analysed using logistic regression. In these analyses, the criteria for assessing goodness-of-fit indicated over-dispersion in the data (Pearson chi-square/d.f. > 1), probably as a consequence of non-independence between successive observations and observations made on birds in the same pair (there were four pairs among the 27 studied individuals). We therefore used quasi-likelihood estimation (McCullagh & Nelder 1989). We analysed factors influencing (iv) the length of incubation intervals; (v) the number of incubation intervals; (vi) the length of absences from the colony; and (vii) the number of absences, using general linear models. The variables were transformed when necessary to meet criteria of normality, either log or square-root transformed. The covariates included in the models were breeding area, sex, incubation stage, clutch size, body condition and the following interactions: sex × breeding area, PCB × breeding area, PCB × sex. Because our sample size was small, we used Akaike’s information criterion (AIC) to select the best statistical models. AIC is defined as −2 × log likelihood +2 × number of parameters of the model, and is recommended as the most efficient tool for choosing the statistical models with the best predictive abilities (Burnham & Anderson 1992, 1998; Lindsey 1999). The best models have the lowest AIC value, and models with AIC values varying by less than 1 are considered equivalent (Burnham & Anderson 1992, 1998). We used a combination of backward and forward search together with different initial models to find the model with the lowest AIC value. When using quasi-likelihood estimation, we used QAIC (quasi-likelihood AIC) where the scale was taken as the Pearson chi-square of the most complex model divided by the associated number of degrees of freedom (Burnham & Anderson 1998). Model parameter estimates and associated confidence intervals were calculated using non-parametric bootstrapping (Efron & Tibshirani 1993), as implemented in S-Plus. Estimates of effect size were therefore not dependent on assumptions about the distribution of the response variables.


The PCB concentrations in the blood (wet weight) of glaucous gulls in this study ranged from 52·4 to 1079·0 ng g−1. PCB levels in males (n = 11) were higher than in females (n = 16) (453·5 ± 84·4 vs. 243·0 ± 41·4 ng g−1). They were also higher at the seabird cliff (n = 13) than at sea level (n = 14) (462·1 ± 69·1 vs. 205·0 ± 40·0 ng g−1) (Table 1). Both breeding area (F1,24 = 15·65, P = 0·0006) and sex (F1,24 = 6·75, P = 0·0158) had significant effects on PCB concentrations.


Females incubated more than males (mean of 54·4% vs. 45·5%) in both areas (Table 1), and the best model (lowest QAIC value) for the proportion of time spent incubating included only sex. PCB levels did not contribute to the model fit (Table 2).

Table 2.  Models explaining variation in parental behaviour in the glaucous gull by using Akaike’s information criterion (AIC). The best model is the model with the lowest AIC. The scale used to calculate the quasi-likelihood AIC (QAIC) is given for each variable. QAIC values are given for the best model (MB1), the second (MB2) and third (MB3) best models (MB1 + x or MB1x) and for the best model when removing one parameter at the time (MB1x). Changes in QAIC values as a result of adding PCB to the best model is always given
Dependent variableModelQAICNumber of parameters
Proportion of time incubating (scale = 2·30)
 MB1 = sex22·482
 MB2 = (MB1 + incubation stage)23·023
 MB3 = (MB1 + breeding area)23·313
 MB1 + PCB24·283
Proportion of time absent when not incubating (scale = 4·75)
 MB1 = sex + breeding area + PCB + body condition + clutch size + incubation stage29·847
 MB2 = (MB1 + breeding area × PCB)30·698
 MB3 = (MB1 − incubation stage)30·706
 MB1− body condition32·706
 MB1− clutch size32·716
 MB1− sex35·666
 MB1− PCB38·276
 MB1− breeding area49·206

Although the incubation bouts of females and males were of the same length (5·5 vs. 5·3 h), there was a striking difference between the two areas. At the seabird cliff, females tended to have shorter incubation bouts (3·5 h) than males (5·3 h), while at the sea level breeding area the trend was opposite (7·0 vs. 5·3 h) (Table 1). The best model for the variation in incubation bout length contained only breeding area, and PCB was not retained in the final model (Tables 3).

Table 3.  Models explaining variation in parental behaviour in the glaucous gull, by using Akaike’s information criterion (AIC). AIC values are given for the best model (MB1), the second (MB2) and third (MB3) best models (MB1 + x or MB1x) and for the best model when removing one parameter at the time (MB1x). Changes in AIC values as a result of adding PCB to the best model is always given. The best model is the model with the lowest AIC
Dependent variableModelAICNumber of parameters
Length of incubation bouts
 MB1= breeding area−47·692
 MB2 = (MB1 + sex + sex × breeding area)−47·104
 MB 3 = (MB1 + sex)−46·013
 MB1 + PCB−45·873
Number of incubation bouts
 MB1 = sex + breeding area + PCB + incubation stage + sex × PCB + sex × breeding area 29·577
 MB2 = (MB1 + body condition) 31·018
 MB3 = (MB1 + breeding area × PCB) 31·138
 MB1− sex × PCB 31·656
 MB1− incubation stage 31·906
 MB1− sex × breeding area 35·446
Absence length
 MB1= breeding area + incubation stage−22·913
 MB2 = (MB1 + sex)−21·914
 MB3 = (MB1 + clutch size)−21·104
 MB1 + PCB−20·994
Number of absences
 MB1 = sex + PCB + incubation stage + body condition + clutch size + sex × PCB−41·017
 MB2 = (MB1 + breeding area)−39·718
 MB3 = (MB1 − PCB × sex)−38·956
 MB1− incubation stage−31·206
 MB1− clutch size−28·536
 MB1− body condition−27·896

Females had a mean of six incubation bouts, while males incubated on average for five bouts during the 48-h observation period, and the number of bouts was lower for both sexes at sea level compared with the seabird cliff (Table 1). The best model, with the lowest AIC value, explaining the number of incubation bouts included four main effects and two interactions (Table 3). The main effect of PCB was not significant, but PCB had a weak effect in the interaction with sex, indicating that females were more affected by PCB than males (Tables 3, 4).

Table 4.  Parameter estimates for the best models explaining parental care behaviour in the glaucous gull. Model selection based on AIC values in Table 3. Standard errors (SE) estimated using bootstrapping (Efron & Tibshirani 1993)
Dependent variableParameter estimateSEt– valueP-value
Length of incubation bouts
 Intercept   1·85   0·1017·78< 0·01
 Breeding area  −0·44   0. 14−3·080·01
Number of incubation bouts
 Intercept   7·62   1·52 4·720·01
 Sex  −2·43   3·24−1·350·19
 Breeding area   2·85   1·03 2·770·011
 PCB  −0·002   0·003−0·600·55
 Incubation stage  −0·22   0·12−1·830·08
 Sex × breeding area  −5·77   2·24−2·56 0·03
 Sex × PCB  −0·01   0·01 1·770·09
Absence length
 Intercept   3·91   0·65 7·68 < 0·01
 Breeding area  −1·65   0·39−5·49 < 0·01
 Incubation stage   −0·12   0·04−3·31< 0·01
Number of absences
 Intercept  −1·84   1·61−1·650·11
 Sex  −0·93   0·57−2·270·04
 PCB   0·001    0·0009 2·360·03
 Incubation stage   0·06   0·03 3·27 < 0·01
 Body condition 6265·92316 3·85< 0·01
 Clutch size  −0·69   0·25−3·74 < 0·01
 PCB × sex   0·002    0·001 1·760·10

Nest-site attentiveness

Females spent more than twice as much time away from the nest site when not incubating compared with males (50·4% vs. 20·8%), and birds at sea level spent more time absent from the nest site than birds at the seabird cliff (50·6% vs. 25·0%) (Table 1). The best models for the proportion of time absent from the nest site included PCB and all five covariates, but with no interactions (Table 2). The effects of PCB were strongly positive (Table 5), i.e. birds with high PCB levels were absent more (Fig. 2).

Table 5.  Parameter estimates for the best models explaining parental care behaviour in the glaucous gull. Model selection based on QAIC values in Table 2. Standard errors (SE) estimated using bootstrapping (Efron & Tibshirani 1993)
Dependent variableParameter estimateSEt-value
Proportion of time incubating
 Intercept 0·170·10 1·72
Proportion of time absent when not incubating
 Breeding area−3·121·08−4·42
 PCB 0·0050·003 3·30
 Incubation stage−0·120·09−1·80
 Clutch size−1·170·66−2·36
 Body condition 0·0010·001 2·26
Figure 2.

The relationship between the PCB level in the blood and percentage of time absent from the nest site in glaucous gulls, when not incubating. Regression lines are based on a model with sex and breeding area as independent variables, as there were no interactions (Table 2).

Females were also absent from the nest site for significantly longer periods than males (4·5 vs. 2·8 h), with birds breeding at the sea level being absent for longer periods than those at the seabird cliff (5·6 vs. 1·8 h) (Table 1). The best model included breeding area and incubation stage (Table 3), and birds late in the incubation period were absent for shorter periods (Table 4). PCB showed no effect on this behavioural trait (Table 3).

Females were absent more often than males (3·3 vs. 2·5 times), with lower numbers of absences at the sea level compared with the seabird cliff (Table 1). The best fitting model included PCB and four covariates, in addition to the interaction PCB × sex (Table 3). The main effect of PCB was positively significant (Table 4) and birds with high burdens had a higher number of absences from the nest site than birds with lower burdens, except from females at the sea level breeding area (Fig. 3).

Figure 3.

The relationship between the PCB level in the blood and the number of absences from the nest site in glaucous gulls during a 48-h period. Lines are based on separate regressions for each sex and breeding area because of a significant interaction between sex and PCB (Table 4).


The novelty of our approach was that the level of contaminants was measured in the blood of individuals whose behaviour could be observed directly. A central finding was a positive relationship between absence from the nest site and PCB concentration and, even if our sample size was relatively small, PCB greatly improved the model fit (Tables 2 and 4). This indicates that the parental behaviours of birds with high levels of PCB are affected, possibly by neurological effects or through endocrine disruption. However, other factors also have a large influence on nesting behaviour, and to disentangle the effects of PCB it is necessary to control for these factors.

Males spent less time away from the colony than females, suggesting that they may have a more important role protecting the nest than females (Løvenskiold 1954; Ferguson-Lees 1963). Alternatively, females may be away from the nest site more often because they have to replenish their reserves after egg production, thus spending more time feeding.

The incubation bouts and absences were shorter on the seabird cliff (Revdalen), the breeding area with the highest PCB levels, than in the breeding area close to the sea level (Kvalrossbukta). In the latter, birds spent more time away from the nest site when not incubating. The main reason for this difference was probably related to diet composition, even if other factors such as diet quality may have influenced the pattern. On the seabird cliff the predominant food during incubation was seabird eggs, while in the other colonies the birds seemed to eat predominantly fish, such as capelin Mallotus villosus (Bustnes et al. 2000). Hence the birds feeding on fish flew out to sea, probably needing several hours to find food, compared with those that mostly foraged only a few hundred metres from the nest. Thus the fact that glaucous gulls (Barry & Barry 1990; Schmutz & Hobson 1998), as other gull species (Davis 1975; Pierotti & Annett 1991), show individual and colonial specialization in feeding habits, makes it difficult to study effects of contaminants between different study areas without measuring individual OC levels. The best way to control for the effects of diet differences among breeding areas, and to get a measure of the effect of PCB, was to include breeding area as a variable in the covariance analysis.

The nesting period is the most critical phase in avian reproduction, rendering both parents and offspring vulnerable to predation and starvation. Thus factors impairing nesting behaviour may cause reproductive failure. Birds with high levels of PCB were more often absent and spent a higher proportion of the time away from the nest site. A number of experimental studies in birds have shown that exposure to various OC, such as PCB, DDE and dieldrin, affect parental behaviour. This has often been attributed to neurological effects of intoxication (McArthur et al. 1983; Tori & Peterle 1983; Peakall 1985; Keith & Mitchell 1993; Hoffman, Rice & Kubiak 1996). Moreover, the reproductive cycle of birds is controlled by hormones (Farner & Wingfield 1980). OC may therefore affect reproductive behaviour through endocrine disruption (Barron, Galbraith & Beltman 1995), as several compounds are known to alter hormone levels and may act as oestrogen or thyroxine agonists or antagonists (McKinney et al. 1985; Petersen, Theobald & Kimmel 1993; Crews, Bergeron & McLachlan 1995).

Under field conditions there is little direct evidence for toxic effects on parental behaviour (Peakall 1996) or breeding performance in birds (Harris & Osborn 1981). By measuring the levels of OC in eggs, and then subsequently recording the behavioural responses to disturbance at the nest, Fyfe, Risebrough & Walker (1976) found that DDE and PCB were negatively correlated with nest defence in merlins Falco columbarius and prairie falcons Falco mexicanus. This could not, however, be confirmed in a study of merlins by Fox & Donald (1980). By comparing nesting colonies, Fox et al. (1978) found that herring gulls Larus argentatus heavily contaminated by PCB had lower incubation temperatures and lower nest attendance, in addition to decreased nest defence, than birds nesting at a lightly contaminated control site. Moreover, egg exchange experiments between low and highly PCB-contaminated herring gulls and Foster’s tern Sterna fosteri colonies have indicated that the parent’s ability to incubate eggs successfully was affected by PCB (Peakall et al. 1980; Kubiak et al. 1989). The demonstrated effects of OC from previous field studies may thus be of a nature similar to those we found for nest-site attentiveness. Experimental studies of birds exposed to various OC have shown changes in operant behaviour and ability to conduct complex behavioural patterns, supporting this conclusion (reviewed by Peakall 1985; Hoffman, Rice & Kubiak 1996).

It is well established that reductions in body fat reserves lead to increased OC concentrations in the tissue and blood (Bogan & Newton 1977; Henriksen, Gabrielsen & Skaare 1996, 1998; van den Brink, van Franeker & de Ruiter-Dijkman 1998), and birds in poor body condition generally have higher OC levels than those in good condition. An alternative explanation for the increased absence may be a confounding effect of poor body condition and not the OC concentration. However, for no behaviour did body condition weaken the effect of PCB in the statistical models.

Even if the levels of PCB found in the brains of glaucous gulls at Bear Island (Gabrielsen et al. 1995) are considerably lower than those usually considered lethal in birds (Sileo et al. 1977; Stickel et al. 1984), the individuals with high levels are well within the range expected to produce behavioural aberrations (Peakall 1985; Gabrielsen et al. 1995; Sagerup et al. 2000). In birds, the levels of PCB in the brain needed to cause behavioural change are generally about one order of magnitude below the levels that cause mortality (Peakall 1985). However, a pertinent question is whether the differences we observed were caused by specific PCB congeners or other contaminants, for example DDE. This problem cannot be resolved by this study because levels of various long-distance transported contaminants are strongly correlated in individual glaucous gulls (see the Methods). However, the brain residues of PCB and DDE necessary to cause death (310 vs. 300–400 p.p.m., respectively) and behavioural effects (5–6 vs. 3–8 p.p.m.) are of similar size (Peakall 1985). In the glaucous gull, PCB is five times higher than DDE (see the Methods).

Recent studies have found that incubation has considerable energetic costs, and an increase in these costs can lead to reduced reproductive output (Monaghan & Nager 1997; Thomson, Monaghan & Furness 1998). Although we did not observe nests being destroyed because they were left unattended, this study does indicate that birds with higher levels of PCB suffer some type of behavioural impairment. Possibly reproductive propensity is affected through endocrine disruption (Barron, Galbraith & Beltman 1995). Alternatively, neurological disorders may lead to difficulties in catching food and thus higher energy costs during incubation. If such effects are large, they may lead to reduced ability to raise young in the more energy-demanding chick period. Deaths of gulls observed early in the chick period (V. Bakken & H. Strøm, personal observation) may be related to increased costs during incubation, either because of reduced ability to catch prey or because the partner is less able to contribute to the incubation.


We are grateful to Magnus Fjeld, Øystein Miland and Kjetil Sagrup for valuable help during field work, and to Rob Barrett who commented on the paper and corrected the English. We are especially thankful to Dr Nigel Yoccoz for help with the statistical analyses. We also wish to thank Dr Dan Osborn and two anonymous referees for comments that greatly improved an earlier draft of the manuscript. The study was funded by the Norwegian Research Council (Project No. 114198/720) and the Norwegian Ministry for Environment.

Received 27 March 2000; revision received 13 February 2001