Carry-over effects reveal reproductive costs in a long-distance migrant


  • Richard Inger,

    Corresponding author
    1. Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
      Correspondence authors. E-mail:,
    Search for more papers by this author
  • Xavier A. Harrison,

    1. Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
    Search for more papers by this author
  • Graeme D. Ruxton,

    1. Division of Environmental and Evolutionary Biology, Institute of Biomedical & Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK
    Search for more papers by this author
  • Jason Newton,

    1. NERC Life Sciences Mass Spectrometry Facility, Scottish Universities Environmental Research Centre, Rankine Avenue, East Kilbride G75 0QF, UK
    Search for more papers by this author
  • Kendrew Colhoun,

    1. Wildfowl and Wetlands Trust, Castle Espie Wetlands Centre, Ballydrain Road, Comber, County Down BT23 6EA, UK
    2. R.S.P.B. Northern Ireland Headquarters, Belvoir Park Forest, Belfast BT8 4QT, UK
    Search for more papers by this author
  • Gudmundur A. Gudmundsson,

    1. Icelandic Institute of Natural History, PO Box 5320, IS-125 Reykjavik, Iceland;
    Search for more papers by this author
  • Graham McElwaine,

    1. 100 Strangford Road, Downpatrick, County Down BT30 7JD, UK
    Search for more papers by this author
  • Matthew Pickford,

    1. Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
    Search for more papers by this author
  • David Hodgson,

    1. Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
    Search for more papers by this author
  • Stuart Bearhop

    Corresponding author
    1. Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
      Correspondence authors. E-mail:,
    Search for more papers by this author

Correspondence authors. E-mail:,


1. It has been known for some time that the consequences of ‘decisions’ made at one point in an animal’s life may not always be borne immediately. For example, numerous studies have demonstrated the trade-off between current and future breeding success across multiple taxa.

2. It is becoming increasingly clear that such processes may also operate among seasons, such that the conditions experienced at one point in the annual cycle may have significant downstream impacts, or ‘carry-over effects’, and this is particularly evident among migratory species. We might therefore predict that certain combinations of reproductive and migratory strategy could lead to profound carry-over effects. However, the extent to which these phenomena might generate variation in fitness within a population is unclear.

3. Here, we investigate how winter habitat selection in a long-distance migrant, with extended parental care (the Light-bellied Brent goose) is influenced by parental status and how this has a counterintuitive effect on subsequent breeding success.

4. Dominant individuals and groups generally monopolize the best quality resources. In the case of geese, families are dominant; however, our findings highlight a hidden cost to raising a family. Stable isotope analysis demonstrates that later in the non-breeding season, adults with families utilize lower quality resources than non-breeders. This is probably caused by parents being constrained in habitat choice by the lower foraging efficiency of their juveniles. Consequently, parental adults end the winter in poorer condition than non-breeders.

5. We further demonstrate that parents in one year are less likely than expected to breed again in the next year and suggest that this is caused by conditions during the non-breeding period being carried over into the breeding season. In conclusion, we demonstrate previously hidden costs to raising a family, which are likely to be important in terms of life-history evolution.


Reproduction is costly and there is a trade-off between current breeding effort and subsequent attempts (Charnov & Krebs 1974; Lessells 1991), which may preclude breeding in consecutive seasons (Clutton-Brock 1988). Intermittent breeding, or reproductive skipping has been demonstrated across multiple taxa (Cam et al. 1998; Lourdais et al. 2002; Rivalan et al. 2005; Johnston & Post 2009), and has been suggested as an adaptive strategy to maximize lifetime reproductive success (Schaffer 1974).

Parental care is one of the major costs of reproduction (e.g. Pettifor, Perrins & McCleery 1988; Jacobsen, Erikstad & Sæther 1995; Milonoff et al. 2004), and for species with extended parental care this will be most acute during the non-breeding period. Additionally, it is becoming increasing apparent that conditions experienced during one period can be ‘carried over’ such that an individual’s reproductive success can be heavily dependent on conditions during the previous non-breeding period. (Marra, Hobson & Holmes 1998; Bearhop et al. 2004; Norris 2005; Norris & Marra 2007; Robb et al. 2008). Given this, it is likely that the conditions experienced during the non-breeding period and parental status will strongly influence the relationship between current and future breeding success.

Carry-over effects are usually driven by access to resources. In social species from insects (Hölldobler 1981) to birds (Black & Owen 1989) and mammals (Mech & Boitani 2006) access to resources is heavily influenced by competitive ability, which can often be linked to social status or the sizes of kin groups (Huntingford & Turner 1987; Crofoot et al. 2007). However in groups of closely related individuals such as families, members are likely to vary in competitive ability and this could potentially lead to conflicts of interest. For example, as preferred resources become limiting, the foraging efficiencies of juveniles and adults can differ (Daunt et al. 2007), and thus may favour foraging in different habitat types. We have previously demonstrated this in Brent geese with juveniles utilizing lower quality terrestrial resources both earlier, and for longer periods, than adults (Inger et al. 2006b) (although this analysis did not consider the social class of the adults). This may mean that parental adults are forced to abandon the best quality habitats earlier than other adults in other social groupings. Hence, there is likely to be a trade-off between the costs of caring for a large family over extended periods and the benefits of increased access to resources. A trade-off of this nature is likely to be most obvious among migratory species in late winter, since it is likely to reduce pre-migratory condition, which is often closely tied to breeding success (Lundberg & Alatalo 1992).

The interaction between social class and habitat selection has been largely unexplored in migratory species because until recently it was virtually impossible to track the habitat usage of large numbers of individuals, of known status, over time. However, the advent of several stable isotope techniques has led to significant advances in tracking habitat selection of individual animals across their annual cycles (e.g. Gunnarsson et al. 2005; Studds & Marra 2005). Isotopic gradients exist between many habitats and these are reflected in consumer’s tissues, which can provide a robust assessment of dietary/habitat selection (Hobson & Clark 1992; Bearhop et al. 2001, 2002). For example, the isotopic gradients between marine and terrestrial habitats are particularly pronounced and have been used to infer habitat selection in a number of species (Bearhop et al. 1999; Inger et al. 2006b, c; Bodey et al. 2009).

In this study, we examine the costs of extended parental care in a long-distance migrant and ask whether differential habitat selection among parental and non-parental adults can precipitate carry-over effects in the subsequent breeding season. We consider two sets of competing hypotheses:

  • 1Conventionally, larger social groups are expected to be dominant, hence parental adults in family groups should (i) have access to the best quality marine resources, (ii) will be in better pre-migratory condition, and (iii) will be more likely to reproduce again in the next year than non-parental adults.
  • 2Parental adults are constrained by the foraging inefficiency of their offspring, and are thus forced to move to more abundant, lower quality terrestrial resources earlier than other non-parental adults, hence adults in family groups should (i) have restricted access to the best quality marine resources, (ii) be in poorer pre-migratory condition, and (iii) be less likely to reproduce in the subsequent breeding season.

To discriminate between these hypotheses, first we assess habitat utilization as inferred by stable isotope analysis. Secondly, we quantify body condition and how it changes throughout the winter and how it differs amongst social classes. Thirdly, we measure differences in foraging efficiencies between adults and juveniles within family groups. Finally, we use an 8-year resighting data base to quantify the effects of rearing a family on breeding success in the following year.

Materials and methods

Study population

The East Canadian High Arctic population of light-bellied Brent Geese, Branta bernicla hrota (O.F. Müller) consists of c. 40 000 individuals (Irish Brent Goose Research Group, unpublished data), which stage in Iceland and the vast majority winter around the coast of Ireland. Brent geese feed preferentially in the intertidal zone on the marine angiosperm Zostera spp., and green algae (Enteromorpha spp. and Ulva lactuca L). However, by mid-winter these resources are depleted at most sites and increasing numbers of birds switch to feeding on terrestrial grasses including agricultural and recreational habitats (Inger et al. 2006a, b, c).

Sample collection

Birds were captured at multiple sites using cannon nets during winter from February 2001 to April 2009 (Table 1) and marked with unique coloured leg rings. Birds were aged as adults or juveniles via plumage characteristics and sexed by cloacal examination, and measurements of skull length, maximum wing chord and body mass taken. A body condition index (BCI) was calculated by regressing the first principle component extracted from a principal component analysis of maximum wing chord and skull length (explaining 75% of the variation) upon body mass (r2 = 0·38, < 0·001) and taking the residuals. Between October 2003 and October 2006 blood samples were taken and separated into plasma and cells for stable isotope analysis (see Inger et al. 2006b for full details).

Table 1.   Cannon net catches by date, site and country; United Kingdom (UK) or Republic of Ireland (ROI)
DateCatch siteCountryn
February 2001WexfordROI32
February 2002Strangford Lough SouthUK26
October 2003Strangford Lough WestUK9
December 2004Strangford Lough WestUK30
February 2004Strangford Lough WestUK132
November 2004Strangford Lough WestUK13
December 2004Strangford Lough WestUK43
January 2005DundrumUK93
February 2005Strangford Lough WestUK32
April 2005WexfordROI34
April 2005Strangford Lough WestUK37
October 2005Strangford Lough EastUK30
November 2005Strangford Lough WestUK11
January 2006Strangford Lough WestUK63
January 2006DundrumUK67
January 2006Strangford Lough SouthUK82
January 2006Strangford Lough WestUK4
February 2006DungarvanROI44
February 2006Carlingford LoughROI15
April 2006Tralee BayROI52
April 2006Strangford Lough WestUK16
November 2006Strangford Lough WestUK36
January 2007Strangford Lough SouthUK42
January 2007DundrumUK75
February 2007Tralee BayROI28
October 2007Strangford Lough EastUK34
January 2008DundrumUK107
February 2008Killala BayROI149
October 2008Strangford Lough WestUK1
November 2008Strangford Lough WestUK9
December 2008Strangford Lough SouthUK15
December 2008DungarvanROI147
February 2009WexfordROI66
March 2009CastlemaineROI55
April 2009Tralee BayROI34
 Total 1663

Re-sighting of birds

Marked birds were re-sighted using high-quality telescopes between February 2001 and March 2009, by a network of experienced observers in Ireland, Iceland, Great Britain and France. The location, flock size, habitat and associated birds were also recorded. Associates (see below) were defined as conspecifics near to the focal birds, which moved in coordination with the focal birds consistently during the observation period (usually >10 min). Where associated birds were also ringed, the ring combination for the associate was also recorded. For adults in family groups, the number of associated juveniles was recorded. Re-sighting data were rigorously checked for quality/accuracy and any inconsistent or ambiguous data were discarded. Quality checked data were then entered into a data base for further analysis.

Re-sighting data base analysis

Adult birds were assigned to one of three social classes for each season: (i) parental adult, having juvenile(s) in that season, (ii) one of a non-breeding pair, being found in association with another adult bird, or (iii) singleton, not associated with other individuals. We assumed that individuals did not alter their social status during any particular over-winter period. Loss of entire broods is rare and as each assignment is based on multiple observations this is a reasonable assumption. Assignments were made as follows; for each ringed adult, we interrogated the resightings data base, and identified all records for that individual. If the bird was not seen, or no associations were recorded in a particular season then no social class was defined for that season. Individuals explicitly recorded as unassociated were defined as singletons. Birds recorded in association with another adult bird, but with no juveniles were classified as non-breeding pairs. Birds recorded with another associated adult and juveniles were defined as parental adults. Multiple (>2) independent re-sightings of a bird with the same associated individual (or individuals) were taken as confirmation of association, and these data were utilized in subsequent analyses. In instances where conflicts existed between resightings, data were excluded from the analysis hence we can be confident all of the associations identified for this study represent actual social groups.

We independently verified a subset of these associations in two ways. First, a number of known family groups have been captured and ringed on the breeding grounds in Arctic Canada, and we were able to identify correctly them as family groups after they had migrated to Ireland. Secondly, and most persuasively we used microsatellite data (Harrison et al. in press). A total of 1128 individuals were genotyped across 15 microsatellite loci. We then used Colony v2·0 (Wang, 2004) to build family pedigrees of related geese. We considered only assignments with a probability of 1. Colony assignments resulted in 60 family groups of which 70% could be matched to resightings of ringed family groups in the field. These data suggest that field observations of associated adults and juveniles are highly likely to represent a group of first order relatives.

Individuals for which we had data on social class for two consecutive seasons were further categorized as consecutive or non-consecutive breeders, with consecutive breeders being defined as parental adults in years n and n + 1.

Stable isotopic analysis

Blood plasma were freeze-dried, ground into a homogenous powder, and weighed into tin cups. Analysis was carried out at the East Kilbride node of the Natural Environment Research Council’s Life Sciences Mass Spectrometry Facility. Stable carbon and nitrogen isotope measurements were carried out using continuous flow isotope ratio mass spectrometry (CF-IRMS), using a Costech (Milan, Italy) ECS 4010 elemental analyser interfaced with a Thermo Electron (Bremen, Germany) Delta Plus XP mass spectrometer. Stable isotope ratios are reported as δ values and expressed in ‰, according to the following: δX = [(Rsample/Rstandard) − 1] × 1000, where X is 13C or 15N and R is the corresponding ratio 13C/12C or 15N/14N, and Rstandard is the ratio of the international references PDB for carbon and AIR for nitrogen. Replicate analyses of internal laboratory standard gelatine and alanine during measurements reported in this study yielded standard deviations better than 0·3 for δ15N and 0·24 for δ13C.

Isotopic mixing models

We used the siar package in r to calculate the proportion of marine derived food in the diet of individuals (Jackson et al. 2008; Inger et al. 2010; Parnell et al. 2010). siar uses a Bayesian approach to determine the proportion of different sources in the diet of consumers. It differs from earlier multiple source mixing models used for underdetermined systems (Isosource; Phillips & Gregg 2003) in that the output are true probability density functions and hence measurements of central tendency represent the most probable solution and can hence be used in down-stream statistical analyses. siar also allows the user to incorporate natural variability within the models, which is propagated through the model resulting more honest output as they reflect the variability within the system. Here, we used the mean and standard deviation for each food source (Zostera spp., Ulva lactuca, Enteromorpha spp., and terrestrial grasses; see Inger et al. 2006b, c for values) as the source values. Trophic enrichment factors (TEFs) (Δ15N = 3·2‰, Δ13C = 1·45‰) were derived from literature sources (see Inger et al. 2006b, c), which we know to be sound as the results of stable isotope mixing models accurately reflect the foraging behaviour seen in the field (Inger et al. 2006b). It is important to recognize that the use of TEFs which are not species and tissue-specific have been criticized (Caut et al. 2009). However, siar can incorporate variation in TEFs into the model and we used a large standard deviation of 1‰ for the TEFs used. siar can also incorporate variability in elemental concentration dependence within the models, although this was not necessary in this case as the C : N ratios where very similar across sources.

We used the SIARSOLO command to carry out the analysis as we had a single data point for each individual for each isotope. The total marine contribution (TMC) to the diet was calculated by summing the median value for each marine source (Zostera spp., Ulva lactuca, Enteromorpha spp.) from the siar output.

Foraging efficiencies

To identify any differences in foraging efficiency between adult and juvenile geese, and how these varied within a winter, we quantified both peck and pace rates whilst birds were actively feeding on intertidal resources at multiple sites (n = 13) in both October and December 2008. To control for differences in resource density we only used adults and juveniles in the same family group, which associate closely and thus are likely to be foraging on resources of similar density. For each individual, both peck and pace rates were measured over a 2-min period (see Inger et al. 2006a for further details).

Statistical analysis

Factors affecting body condition were explored using a general linear model (GLM) with BCI as the dependant variable and season of capture, country of capture (Northern Ireland or the Republic of Ireland), social class (parent, non-breeding pair, or singleton) and sex as factors, and time (both linear and quadratic terms for time, day number, day 1 = 1st October) as covariates. Model simplification was achieved by backwards elimination of non-significant terms, which removed sex and country from the model. We examined the role of brood size (ranging from one to six juveniles) on BCI of adult parents using a GLM with brood size and, to control for seasonal changes, capture month and season as factors in the model. To examine factors affecting TMC to the diet, a GLM was constructed with TMC as the dependant variable and social class [two levels; parent or non-breeding pair, singletons’ were dropped due to small sample size (n = 2)], season and sex as factors, and time (linear and quadratic terms for time) as covariates. Sex and the linear term for day number were removed due to non-significance. The relationship between TMC and time was fitted with both linear and quadratic models and the best fits were chosen using the methods outlined by Crawley (2007). Difference between adult and juvenile foraging efficiencies in the early and mid-winter was examined using Generalized Liner Mixed Models, using lmer in the lme4 r package. Peck and pace rates were used as the dependant variable with age and month as fixed factors and family as a random factor, with a Poisson error distribution. Differences in breeding frequency between the whole data set and a subset of individuals breeding in consecutive seasons were identified using Pearson’s chi-squared test with Yate’s continuity correction. Expected frequency of breeders and non-breeders was calculated from the whole data set. GLMs were carried out in spss (V.15; Chicago, IL, USA), all other analyses were carried out in the r statistical environment, R Development Core Team (2008).



A total of 1663 geese (1089 adults) were caught using cannon nets in Northern Ireland and the Republic of Ireland between 13 February 2001 and 8 April 2009. The total number of resightings during this period was 49 755, at 377 sites throughout the Republic of Ireland and Northern Ireland. Between 26 October 2003 and 17 October 2005, a subset of 281 geese was sampled for stable isotope analysis. Of these, 102 were assigned to a social group and included in subsequent analyses.

Body condition/mass trajectory

Body condition index (GLM r2 = 0·50, n = 424) varied significantly over time (day number). The linear term (F1,393 = 100·0, P < 0·001) highlighted a general decrease in BCI throughout the winter, although the quadratic term (F1,393 = 34·4, P < 0·001) showed a recovery in BCI in late winter associated with pre-migratory fattening. BCI also differed between social classes (F2,393 = 16·1, P < 0·001), and we found a strong interaction between social class and time (quadratic term; F2,393 = 9·8, P < 0·001, linear term; F2,393 = 15·1, P < 0·001) indicating that the patterns of change in body condition differed amongst social classes. Parental adults also end the winter (day > 150) in significantly lower body condition than singletons and non-breeding pairs (F1,22 = 5·5, P = 0·029). Singletons and non-breeding pairs start the winter in better body condition than parental adults, which have a higher BCI during mid-winter. To illustrate this interaction, quadratic and linear models were fitted to the relationship between BCI and time for the different social classes. For both the non-breeding pairs (quadratic r2 = 0·47, linear r2 = 0·37, F1,188 = 37·1, P < 0·001) and singletons (quadratic r2 = 0·57, linear r2 = 0·46, F1,105 = 27·9, P < 0·001) the quadratic model provided the best fit, whereas for parents the quadratic model did not provide a better fit than the linear model for BCI over time (quadratic r2 = 0·24, linear r2 = 0·25, F1,114 = 0·1, P = 0·7) (Fig. 1). The GLM also found significant differences in BCI between seasons (F7,393 = 2·3, P = 0·029) and a significant interaction between season and social class (F7,393 = 2·2, P = 0·036). Considering parental adults, only a second GLM (r2 = 0·36) indicated that BCI was significantly influenced by capture month (F6,100 = 6·78, P > 0·001) and negatively influenced by brood size (F5,100 = 2·49, P = 0·036), but no seasonal effects (F4,100 = 0·841, P = 0·503) were found.

Figure 1.

 Winter body condition index trajectories with time (where 1 = 1 October) for different social class; parental adults (red; y = −1·41x + 169·17), non-breeding pairs (blue; y = 0·019x2 + −5·59 + 362) and singletons (green, y = 0·02x2 + −6·32 + 421).

Stable isotopic diet analysis

The TMC (GLM r2 = 94·9) to the diet of individual geese changed significantly during seasons (F2,90 = 66·1, P < 0·001). Importantly, social class (parent/non-breeding pair, insufficient data for singletons) (F1,90 = 50·6, P < 0·001), time (day number quadratic term) (F1,90 = 26·4, P < 0·001) and the interaction between social class and time (quadratic term; F1,90 = 66·7, P < 0·001, linear term; F1,90 = 32·5, P < 0·001) were also significant terms in the model. Thus, it is clear that different social groups have differential access to marine resources. The relationship between TMC and time was, as with the mass trajectories, modelled using quadratic polynomials (Fig. 2). For parental adults the quadratic model was a significantly better fit (r2 = 0·77; F1,51 = 16, P < 0·001) than the linear model (r2 = 0·70). Similarly, for non-breeding pairs the quadratic model (r2 = 0·96) produced a better fit to the data (F1,41 = 122, P < 0·001) than the linear model. The two relationships intersected at day 35 (4 November) and day 121 (29 January) with the parental adults having the higher level of marine resource in their diet during this period, whilst the non-breeding pairs had a high level of marine resources outside this period. Over the whole season, the proportion of marine resources in the diet of parental adults was 0·51 (SD = 0·23), where as this was much higher in non-breeding pairs at 0·70 (SD = 0·21).

Figure 2.

 Total marine contribution to the diet as derived from stable isotope ratios using siar over time (where 1 = 1 October); parental adults (red, y = −5 × 10−5x2 + 5 × 10−3x + 0·759), and non-breeding pairs (blue y = −2·3 × 10−5x2 + 0·895, linear coefficient = 0).

Foraging efficiencies

We found significant increase in peck rate between October (n = 96) and December (n = 78) (z = −26·50, P < 0·001), and significant differences between age classes (z = −9·15, P < 0·001), although the significant interaction between month and age (z = 6·39, P < 0·001) indicates that this is mostly because of increased adult peck rates (compared to juveniles) during December (Fig. 3b).

Figure 3.

 Foraging rates of adult and juvenile geese in early (October) and late (December) season intertidal foraging. (a) Pace rates and (b) peck rates.

Pace rates also increased significantly between October and December (z = −17·71, P < 0·001) and again there was a difference between age classes (z = 8·69, P < 0·001). Juveniles had a higher pace rate (indicating more searching behaviour) than adults, but relative pace rates did not vary between months as reflective by the non-significant interaction between month and age (z = −2·0, P = 0·842; Fig. 3a).

Frequency of consecutive breeding

For all years, 2001–2009, we considered the occurrence of individuals bringing juveniles to wintering grounds, indicating successful breeding, in consecutive seasons. The proportion of individuals with a family in season n + 1 when they had bred in season n was lower than expected when compared to the proportion of breeders found in the whole data set. We found that only 27% of these individuals bred in consecutive years compared to 73% that did not breed in the year following rearing a family, which is significantly different (χ21 = 7·16, P = 0·0075) from the expected frequency of 40% breeders and 60% non-breeders as calculated from the whole data set (Fig. 4).

Figure 4.

 Proportion of parents to non-breeding pairs (NBP) in year n + 1. Dark columns were successful breeders in the previous season (n). Light column indicate the proportion of parents and NBPs calculated for all years. Successful breeders in year n are less likely to be parents in year n + 1, but more likely to be non-breeders.


Our findings illustrate a previous unknown cost of extended parental care and provide a mechanism by which carry-over effects can influence both individual fitness and population dynamics. These findings are likely to have implications for numerous species in multiple taxa where long-term family bonds are maintained or where there are long periods of dependency, including other bird species (Radford & Ridley 2006) and mammals (Dahle & Swenson 2003; Noren 2007). As prolonged care of young may create conflicts of interest, which ultimately generate increased and potentially cryptic reproductive costs for parents.

Studies from a range of species, from social insects to birds and mammals, have shown that dominance can be a function of group size, and in turn a key determinant of access to the best quality resources (Jarman 1974; Black & Owen 1989; Heinsohn 1991; Tanner 2006). Here, we challenge that assumption and find that for much of the winter adults with families (which are dominant at other times of year (Tinkler, Montgomery & Elwood 2007a, b)) are constrained by the foraging inefficiencies of their juveniles and thus, we suggest, are forced to utilize suboptimal but super abundant resources. Hence, parental adults end the winter in lower pre-migratory body condition than non-breeding adults. Critically, we demonstrate that adults with families in one year are less likely to breed successfully in the following year, and suggest that this is a carry-over effect associated with poor pre-migratory body condition (Fig. 5). Such carry-over effects should however be considered within the context of a population that is likely to experience substantial stochasticity in reproductive success, including years of very low reproduction output. The manner in which individual level carry-over effects might interact with variable annual reproductive success to influence population dynamics remains a key focus of future research.

Figure 5.

 Flow diagram illustrating the carry-over effects that cause successful breeders in year n to have a lower probability of breeding in year n + 1.

At the population level, body condition changes throughout the winter followed a quadratic trajectory associated with amount of daylight available for foraging. When however social class is included in the analysis it is clear that parental adults differ considerably (and significantly) in body condition trajectory to non-breeders. Non-breeders begin the winter in better condition likely due to a combination of no or limited breeding effort, earlier departure from the breeding grounds and longer staging periods in Iceland. They also end the winter in better condition than parental adults, which is critical as pre-migratory condition is a critical factor in success of future breeding attempts.

Brent geese will always preferentially forage on, and achieve better body condition by utilizing marine resources (Ganter 2000; Inger et al. 2006b, 2008). This is reflected in our results, with parental adults achieving a better body condition in mid-winter (Fig. 1) which is coupled with a period of higher marine intake (Fig. 2). This is almost certainly due to their dominance and ability to monopolize marine resources, consistent with the results of previous studies (Tinkler, Montgomery & Elwood 2007a, b). When the whole of the winter is considered however, we find that non-breeders have 20% higher levels of marine resources in their diet. Similar findings have been demonstrated in several other vertebrates where non-breeders, free from the constraints of parenthood can maintain better body condition than breeders (Woodroffe & Macdonald 1995; Ruusila, Ermala & Hyvarinen 2000; Koivula et al. 2003; Moyes et al. 2006).

The most likely mechanism limiting parental adults’ access to marine resources is the effects of differences in intrafamilial foraging efficiencies, with juveniles having inferior foraging skills (Daunt et al. 2007). Charman (1979) demonstrated that, in Brent geese, when marine Zostera spp. cover falls to below c. 15% adults increase feeding rates (peck rate), presumably to maintain intake rates, compensating for increased search time. Juveniles, however, are unable to increase their intake rates (Charman 1979) due to a lack of proficiency (Marchetti & Price 1989; Heinsohn 1991; Daunt et al. 2007), and are therefore more likely to be driven to seek out new foraging opportunities (Sutherland, Jones & Hadfield 1986). Our results support these data. Juveniles had a greater search times (as indicated by pace rate Fig. 3b) throughout the study period and where unable to increase their intake rates (as indicated by peck rates Fig. 3a) by the same amount as adults to compensate for resource depletion later in the year.

In this study, we have demonstrated how winter habitat utilization can have significant consequences for individual body condition, which almost certainly contributes, via carry-over effects, to the occurrence of intermittent breeding found in this population. It is becoming increasingly apparent that carry-over effects may explain a significant amount of variation in a number of life-history traits. For example, reproductive performance and body condition have consistently been showed to be influenced by conditions experienced in the previous season in many avian systems and are becoming apparent in a number of other taxa (X.A. Harrison, J.D. Blount, R. Inger & S. Bearhop, unpublished data).

Here, we find that some of the drivers of breeding success start to be expressed almost immediately after the previous breeding season. The costs of brood rearing in the Arctic mean that parents arrive on the wintering grounds in poorer condition than other adults. During the winter, parents are then further constrained by the lower foraging efficiencies of their offspring. Indeed, our results demonstrate that over the course of the winter parents have 20% less of the most profitable marine resources in their diets. We have previously shown that access to marine resources is directly related to body condition in this species (Inger et al. 2008), and our current results support this, with parents ending the winter in poorer body condition than non-breeders.

Thus, our study shows that the influence of carry-over effects on short-term fitness measures appears to depend on both the migratory and reproductive strategies. Although such patterns are likely to be most obvious in capital breeding migrants, they might be expected in a range of species, including non-migrants. Extended parental care always has the potential to detrimentally impact resource acquisition, which may consequently affect body condition and thus future breeding attempts as has previously been demonstrated in non-avian taxa (Festa-Bianchet 1998; Festa-Bianchet, Gaillard & Jorgenson 1998; Persson 2005). Our results support this hypothesis with parental adults having a lower than expected probability of being successful in the next breeding attempt. Whether this is a strategic decision, with individuals in suboptimal condition forgoing breeding, or if this represents failed breeding remains unclear, suggesting a clear direction for future research.


We thank Kerry Mackie, Alex Portig, Hugh Thurgate, Gerry Murphy and all the ring readers who made this work possible. R.I. was funded by a NERC studentship with a WWT CASE award and NERC standard grant NE/F021690/1. X. A. H was funded by a NERC studentship with a WWT CASE award.