Carry-over effects as drivers of fitness differences in animals


Correspondence author. E-mail:


1. Carry-over effects occur when processes in one season influence the success of an individual in the following season. This phenomenon has the potential to explain a large amount of variation in individual fitness, but so far has only been described in a limited number of species. This is largely due to difficulties associated with tracking individuals between periods of the annual cycle, but also because of a lack of research specifically designed to examine hypotheses related to carry-over effects.

2. We review the known mechanisms that drive carry-over effects, most notably macronutrient supply, and highlight the types of life histories and ecological situations where we would expect them to most often occur. We also identify a number of other potential mechanisms that require investigation, including micronutrients such as antioxidants.

3. We propose a series of experiments designed to estimate the relative contributions of extrinsic and intrinsic quality effects in the pre-breeding season, which in turn will allow an accurate estimation of the magnitude of carry-over effects. To date this has proven immensely difficult, and we hope that the experimental frameworks described here will stimulate new avenues of research vital to advancing our understanding of how carry-over effects can shape animal life histories.

4. We also explore the potential of state-dependent modelling as a tool for investigating carry-over effects, most notably for its ability to calculate optimal rates of acquisition of a multitude of resources over the course of the annual cycle, and also because it allows us to vary the strength of density-dependent relationships which can alter the magnitude of carry-over effects in either a synergistic or agonistic fashion.

5. In conclusion carry-over effects are likely to be far more widespread than currently indicated, and they are likely to be driven by a multitude of factors including both macro- and micronutrients. For this reason they could feasibly be responsible for a large amount of the observed variation in performance among individuals, and consequently warrant a wealth of new research designed specifically to decompose components of variation in fitness attributes related to processes across and within seasons.


The sequential nature of biological systems means that virtually every ‘decision’ an animal makes in its life will have downstream consequences. This observation is a cornerstone of behavioural ecology and has found particular resonance in the study of resource acquisition and allocation, whereby variation in access to resources or variation in the manner in which they are allocated at one point in life has future implications for fitness (Stearns 1992; Lindstrom 1999; Metcalfe & Monaghan 2001). While a number of studies have demonstrated trade-offs between early life conditions and fitness in adulthood (e.g. Madsen & Shine 2000; Blount et al. 2003; Descamps et al. 2008), downstream effects can also operate over much shorter periods, in particular among seasons (Norris 2005; Norris & Marra 2007). Whilst ecologists have been aware that inter-seasonal effects in the form of density-dependent compensation have the potential to affect population abundance (Fretwell 1972; Sutherland 1996; Ratikainen et al. 2008), the idea that effects in one season can influence individual performance in the next season (Fig. 1) is a fundamentally different process that has remained largely untested until recently, mainly due to the difficulties associated with following individuals across the annual cycle. The advent of new technologies and approaches has enhanced the ability of ecologists to track animals throughout their annual cycles and this has led to a greater understanding of how carry-over effects can influence a range of fitness parameters. Theoretical models (e.g. Norris 2005; Runge & Marra 2005; Norris & Taylor 2006; Ratikainen et al. 2008) have predicted how such individual carry-over effects might, in turn, influence population dynamics. Therefore, it is clear that carry-over effects have the potential to influence multiple levels of biological organization from individuals to populations, and possibly community structure.

Figure 1.

 Schematic representation of two different types of seasonal interactions: carry-over effects (COEs) and seasonal compensation effects. Both phenomena can be driven by either density-dependent or -independent mechanisms during a given season (season 1 in graph). However, the difference is that for COEs to occur individuals must survive to the following season and some measure of their success is affected in the following season (season 2). In contrast, seasonal compensation effects occur when a change in population size in season 1 results in a change in the strength of density dependence in season 2. Both phenomena can lead to changes in per capita rates and subsequent population size (see text for further discussion).

Here, we review the evidence for carry-over effects in vertebrates and assess their potential impacts on individual fitness; we consider how they influence life histories and discuss the factors that might drive them. We use this information to define a framework to predict which animal groups might be most likely to exhibit carry-over effects, and suggest novel experiments designed to quantify their nature and magnitude. We also explore the role of state-dependent modelling as a valuable tool for the study of carry-over effects, and close with a prospectus for future studies.

What are carry-over effects?

Carry-over effects (COEs) have previously been defined as any event occurring in one season that influences individual performance in a non-lethal manner in a subsequent season (Norris 2005; Norris & Marra 2007). We propose a small change in this definition because ‘events’ do not necessarily account for the gradual change in certain condition indices that can accumulate over the course of a season. For example, a bird’s flight feathers may degrade slowly over winter or during migration (Barta et al. 2006), resulting in impaired foraging ability during the breeding season, which could in turn compromise the ability to provision young and reduce fitness. Similarly loss of body mass over winter due to occupying a lower-quality territory (e.g. Marra, Hobson & Holmes 1998) is arguably a gradual process. ‘Events’ certainly do have the power to propagate COEs. For example, the occurrence of bad weather conditions during migration may increase locomotion costs and reduce body reserves available for downstream breeding (e.g. Fox & Gitay 1991; Robson & Barriocanal 2008). Therefore, we define COEs as events and processes occurring in one season that result in individuals making the transition between seasons in different states (levels of condition) consequently affecting individual performance in a subsequent period.

It is also important to note that COEs are a phenomenon not a hypothesis. An a priori hypothesis may be generated from a phenomenon that is suspected to be a COE and this hypothesis can even be tested against a competing hypothesis that attempts to explain variation in fitness from within-season processes (although we are not aware of any studies that have explicitly done so). However, simply testing the ‘COE hypothesis’ in a biological system is not valid since there is no mechanistic explanation.

In many cases, COEs arise because of variation among individuals in the extent to which they have access to, or are able to utilize, resources. Consequently COEs are generally manifested when such variation in resource availability results in individuals making the transition between seasons in different levels of physical condition (state), thus affecting downstream process such as reproduction or survival. For example Bighorn Ewes (Ovis canadensis) that are heavier prior to breeding tend to have higher reproductive success than those that are lighter at the end of the non-breeding period (NBP) (Festa-Bianchet 1998) whilst Black-Tailed Godwits (Limosa limosa islandica) from higher-quality wintering habitats arrived at the Icelandic breeding grounds earlier and had higher reproductive success than those from poorer-quality wintering areas (Gunnarsson et al. 2005).

The majority of studies to date describing COEs generally use body mass or fat mass at the end of winter as an index of condition to explain subsequent asymmetries in reproductive success (see Table 1). However, fat mass is certainly not the only determinant of an animal’s condition, as feather quality in terms of damage and abrasion caused over time (Barta et al. 2006), protein reserves, lean mass, parasite load and immunocompetence at the end of one season may also be immensely important in determining performance in subsequent seasons (see McNamara & Houston 1996 and references therein).

Table 1.   A selection of studies from a variety of taxa describing known carry-over effects (COEs) (a) as well as a list of studies that could potentially represent COEs, but uncertainty about the mechanism underlying observed variation in fitness has precluded us from doing so (b). We also present a list of studies describing developmental effects that are not COEs because they operate between life stages and not between seasons (c) and examples of studies (d) that appear to link reproductive success with a previous season’s environmental variation, but are not COEs because the mechanism operates intra as opposed to inter-seasonally
  1. ‘Trait’ descriptors: BC, body condition; AT, arrival time; RR, return rate; DT, departure time; RF, reproduction frequency; RS, reproductive success; TS, testosterone; CS, clutch size; LD, laying date; GR, growth rate. ‘Mediator’ descriptors: WC, winter climate; SST, sea surface temperature; NBS FA, non-breeding season food availability; FA, food availability; WH, winter habitat; FS, fat stores. ‘TF’ (Time Frame) descriptors: IS, inter-seasonal; LS, between life stages; W, within season.

(a) True carry-over effects
MammalElk Cervus elaphus (L.)RFFAISElk show a threshold condition of 8–10% bodyfat which if not reached by the end of the non-breeding season results in significant decline in probability of pregnancy. Also, dietary quality affected timing of conception with individuals in the ‘low’ nutrition group falling pregnant later than the medium or high groups. This suggests it may have taken longer to reach reproductive threshold condition on a lower quality dietCook et al. (2004)
Gray whale Eserichtius robustus (Lilljeborg)BC/RSFAISPoorer calving success in winter when the previous summer’s ice cover restricts foraging, resulting in individuals being in poorer physical condition prior to giving birthPerryman et al. (2002)
Bighorn ewe Ovis canadensis (Shaw)RSFAIS1 kg of extra mass by mid-September increased the chance of weaning a lamb the following year, irrespective of average adult massFesta-Bianchet (1998)
BirdLight-bellied Brent goose Branta bernicla hrota (O. F. Müller)RSNBS FAISIndividuals rearing young in year t were less likely to successfully rear young in year t + 1, as reproduction in the previous year reduced their ability to reach the resource threshold for reproduction the following year. Such a COE results in reproductive skipping in alternate yearsInger et al. (in press)
Dark-bellied Brent goose Branta bernicla bernicla (O. F. Müller)CS/RFNBS FAISSize of fat reserves at end of non-breeding season determines size of clutch post-migration. Individuals with smallest fat stores are unlikely to breed at allEbbinge & Spaans (1995)
Grey-Headed Albatross Diomedea chrysostoma (Forster)RFFAISIndividuals successfully breeding in year t will not do so in t + 1, indicating a COE from not being able to reach the threshold breeding condition in two subsequent yearsPrince et al. (1994)
Cassin’s auklet Ptychoramphus aleuticus (Pallas)RSNBS FAISIndividuals with a pre-breeding diet comprising poorer-quality food sources bred later and laid smaller eggs than those eating higher quality foodsSorensen et al. (2009)
Red knot Calidris canutus rufa (L.)RSNBS FAISShowed correlation between amount of nutrient stores accumulated at staging sites prior to departure for breeding grounds and subsequent breeding successBaker et al. (2004)
Blue Tit Cyanistes caeruleus (L.)LD/RSNBS FAISBirds receiving supplementary food over winter had advanced laying dates and higher fledgling success than birds from sites receiving no supplemental foodRobb et al. (2008a,b)
Barn swallow Hirundo rustica (L.)AT/RSNBS FAISHigher winter food availability, assessed by change in Normalized Difference Vegetation Index (NDVI) between years, resulted in earlier arrival on the breeding grounds, increased frequency of second clutches and more fledged offspring per seasonSaino et al. (2004)
Black-tailed godwit Limosa limosa islandica (Brehm)AT/RSWHISBirds from high-quality coastal wintering sites arrived in Iceland earlier and had higher breeding success than those from relatively poorer-quality inland sitesGunnarsson et al. (2005)
Black-tailed Godwit Limosa limosa islandicaATWHISIndividuals from sites with stable populations and high spring prey-intake rates arrived on the Icelandic breeding grounds significantly earlier than those from sites with lower prey-intake rates. This in turn may translate into improved breeding successGill et al. (2001)
Black-tailed Godwit Limosa limosa islandicaATWHISIndividuals from traditionally occupied wintering sites arrive earlier than those from wintering sites that have recently been expanded into, suggesting that there is a cost to occupying poorer quality habitatGunnarsson et al.(2006)
Blackcap Sylvia atricapilla (L.)AT/RSWHISIndividuals from more northerly wintering grounds arrived earlier, produced larger clutches and fledged more young than those from ore distant southerly wintering groundsBearhop et al. (2004)
Black-throated Blue Warbler Dendroica caerulescens (Gmelin)BCWHISIndividuals in better condition were found to have spent the winter in forest rather than scrub habitats. This will result in these individuals having more resources to devote to reproductionBearhop et al. (2005)
American Redstart Setophaga ruticilla (L.)BC/ATWH/NBS FAISIndividuals from higher-quality mangrove winter habitat arrived earlier and in better physical condition than those from relatively poorer-quality scrub habitatMarra, Hobson & Holmes (1998)
American Redstart Setophaga ruticillaDT/RRWH/NBS FAISBirds upgraded from low quality scrub to high quality mangrove winter habitat departed earlier for spring migration and showed a higher return rate than non-upgraded control birdStudds & Marra (2005)
American Redstart Setophaga ruticillaAT/RSWH/NBS FAISIndividuals from mangrove habitat arrived earlier and fledged more young than those from scrub habitatNorris et al. (2004)
Avocet Recurvirostra avosetta (L.)ATWHISIndividuals wintering at sites closer to the breeding grounds arrived earlier and fledged more young than those wintering further awayHotker (2002)
Cormorant Phalacrocorax carbo (L.)ATWHISIndividuals wintering <300 km from the colony arrived earlier than those wintering >300 km away. Lifetime reproductive success of known early-arriving females was roughly 1·8 times higher than that of those witnering >300 km awayBregnballe, Frederiksen & Gregersen (2006)
(b) Possible carry-over effects
MammalFur seal Arctocephalus pusillus doriferus (Wood Jones)RSSSTISPup production was negatively affected by differences in winter sea surface temperature (SST) and also by the previous summer’s SSTGibbens & Arnould (2009)
Arctic fox Alopex lagopus (L.)RSNBS FAISLitter size is reduced in years when winter food availability is limited. The authors hypothesize that number of eggs at ovulation or number of embryos that implant will vary with size of fat storesAngerbjörn et al. (1995)
Arctic fox Alopex lagopusRSNBS FAISDens receiving supplemental food over winter had higher breeding success than control (unsupplemented dens)Angerbjörn et al. (1991)
Wolverine Gulo gulo (L.)RFNBS FAISAll females receiving supplementary winter food bred in 3 consecutive years, compared to only 2 of 11 non-supplemented femalesPersson (2005)
ReptileGreen turtle Chenlonia mydas (L.)RFNBS FA?ISRe-nesting interval varies from 3 to 5 years and may be determined by food availability on foraging grounds i.e. Poor food availability may increase the time between nesting attempts until the organism reaches a threshold conditionBroderick et al.(2001)
InvertebrateNepthys caeca (Fabricus) & Nepthys hombergi (Savigny)RSWCISSpawning success was higher following warmer winters. The authors suggest this may be due to oosporption prior to spawning in colder winters in response to variable resource availabilityOlive et al. (1997)
FishPlaice Pleuronectes platessa (L.)RSNBS FAISSome fish given lower-quality diets in the months prior to spawning skipped reproduction entirely, whilst no individuals receiving high-quality diets did soKennedy et al. (2008)
FishWhite Crappie Poxomis annularis (Rafinesque)RFNBS FAISFemales increase gonadal investment as prey availability increased 4 months prior to spawning. Some females in low and medium nutrition groups skipped spawning entirelyBunnell, Thomas & Stein (2007)
BirdMediterranean kestrel Falco tinunculus (L.)RSWCISSmaller clutches were laid after rainier winters. Hatching success was higher after dry wintersCostantini, Carello & Dell’omo (2010)
Garden warbler Sylvia borin (Boddaert)TSNBS FAISBirds receiving a food limited treatment to simulate food restiction during migratory stopover had slower testicular recrudescence and decreased plasma testosterone compared to males fed ad libitum foodBauchinger, Van’t Hof & Biebach (2009)
(c) Silver spoon/maternal effects
BirdZebra finch Taeniopygia guttata (Vieillot)RSNeonatal diet qualityLSIndividuals raised on a poor-quality neonatal diet initiated clutch formation later and laid eggs at a slower rate compared to controls. The authors hypothesize that such delay could result in asynchrony between laying and peak food availability, leading to reduced reproductive successBlount et al. (2006)
MammalAmerican Red Squirrel Tamiasciurus hudsonicus (Erxleben)RSNatal FALSHigher food availability between birth and weaning and warmer spring temperatures in year of birth lead to increased reproductive success in adulthoodDescamps et al. (2008)
ReptileAmphibolurus muricatus (White) & Bassiana duperreyi (Gray)GRCorticosterone levels of eggsLSCorticosterone enhances growth rates of B. duperreyi hatchlings but inhibited those of A. muricatus hatchlingsWarner, Radder & Shine (2009)
MammalWhite-tailed deer Odocoileus virginianus (Zimmerman)GRMaternal diet quality during gestationLSOffspring from food-restricted females suffered reduced growth rate compared to non-food-restricted femalesTherrien et al. (2007)
(d) Not carry-over effects
BirdSnow petrel Pagodroma nivea (Forster)RSWCWReproductive success was higher in years where winter sea-ice cover was high. The authors suggest that whilst this may reduce foraging opportunities during winter, it may promote the survival of krill and other prey so that food abundance during the breeding period is higherOlivier et al. (2005)
Thin-billed prion Pachyptila belcheri (Mathews)SWCWLower amounts of winter sea ice cover lead to increased downstream survival of adults. Similar to Olivier et al. (2005), the authors hypothesize that sea ice in winter affects food available in summer. However, note that the effect of sea ice cover in this study operates in the opposite direction to the study aboveNevoux & Barbraud (2006)
Lesser Kestrel Falco naumanni (Fleischer)RSWCWHigher winter rainfall increased nest success rate and mean number of chicks per successful nest. Rain in winter most likely affect prey availability during incubationRodríguez & Bustamante (2003)

The ‘non-lethal’ nature of carry-over effects

Norris (2005) and Norris & Taylor (2006) emphasize that the defining feature of COEs is their non-lethal nature. Specifically, COEs influence individual success in a subsequent season, where success can be defined as a component of fitness: reproduction or survival. It is relatively easy to imagine a COE occurring as a result of some event during the non-breeding period that influences subsequent reproductive output (Table 1a). However reproductive effort can conceivably also result in mortality in the following period of the annual cycle. This phenomenon would also be considered a COE because an event in one season (summer) has influenced success (survival) the following season (winter, see section on ‘Reproductive Effort’). Thus, the COE phenomenon is defined by the fact that an individual survives the transition between seasons and some component of its fitness is affected the following season. In contrast, if an event or process in one season results in mortality within the same season, there could be consequences for the population (i.e. other individuals) in the following season in the form of density-dependent compensation but this is not considered a COE. We discuss the difference between these two phenomena below.

Processes that are not carry-over effects

Much research to date has focussed on ‘silver spoon’ effects, whereby favourable conditions in the first year of life can exert positive long-term effects in adult life, and maternal effects whereby maternal phenotype (e.g. body condition) can directly influence the performance (e.g. growth rates) of offspring (e.g. Warner, Radder & Shine 2009; see Table 1c). For example Descamps et al. (2008) found that red squirrels Tamiasciurus hudsonicus) that were born under conditions of higher food availability between birth and weaning had a higher reproductive success as adults than those born under lower food availability. Intuitively one might identify such an observation as a COE, because superficially it appears that conditions experienced as a juvenile have persisted throughout ontogeny, or ‘carried-over’ to affect fitness indices in adulthood. However silver spoon and maternal effects are not COEs, primarily because they are not inter-seasonal and in the case of maternal effects can be considered inter-generational. We do not consider any effect that is measured across different life stages to be a COE, as they can be more accurately referred to as developmental effects. For illustration we have provided a selection of examples of such developmental effects from several taxa in Table 1c.

Carry-over effects and their interaction with seasonal compensation effects

COEs occur at the individual level and are a separate phenomenon from seasonal compensation effects (but both fall under the veil of ‘seasonal interactions’; Norris & Marra 2007; Fig. 1). Seasonal compensation effects occur when there is a reduction in population size in one season that results in a change in per capita rates the following season (Norris 2005). For example, if mortality increases due to habitat loss on the wintering grounds, then the per capita breeding success may increase the following season because of density-dependent compensation. Similarly, habitat loss on the breeding grounds may cause a decline in breeding success, resulting in fewer individuals using habitat on the non-breeding grounds. However, COEs can interact with these density-dependent effects in either an antagonistic or synergistic manner (Norris 2005). Although winter habitat loss and the associated reduction in resource availability can cause increased mortality, it can also cause some of the survivors to end the winter season in poorer physical condition than if the habitat loss had not occurred (Marra, Hobson & Holmes 1998; Gill et al. 2001). In this way, COEs can lead to a reduction in per capita reproductive output because there are proportionally more individuals in the population that experience a negative COE. Conversely, if the remaining habitat is of higher average quality than before, then proportionally more individuals may experience a positive COE and, in this way, have the opposite effect to the seasonal compensation phenomenon. Thus, our ability to detect and measure COEs is likely to be heavily influenced by seasonal compensation effects that could occur simultaneously. In this sense, it is critical to estimate the strength of density dependence and the mechanism(s) that drive density-dependent reproduction and survival. It is also vital to acknowledge that the apparent absence of COEs in an animal system does not preclude the possibility that they are operating. Seasonality and habitat asymmetries generate temporal and spatial variation in resource availability, whilst density in the following season can interact with these COEs to create variation in individual fitness.

While it is clear that COEs and seasonal compensation effects influence population size in the following season through different mechanisms (season 2 in Fig. 1), it should be noted that either density-dependent and density-independent processes can drive either COEs or seasonal compensation effects in the season in which they are initiated (season 1 in Fig. 1). For example, a density independent event such as poor weather conditions during migration could increase energetic costs resulting in poorer condition on arrival at the breeding grounds (i.e. a COE; Fox & Gitay 1991; Robson & Barriocanal 2008) and it could also result in significant mortality during migration resulting in fewer individuals arriving on the breeding grounds (i.e. a seasonal compensation effect). To our knowledge, no studies have linked COEs to density-dependent effects in the previous season.

How widespread are carry-over effects?

Whilst the majority of studies describing COEs are on birds, there are in fact studies from numerous taxa including invertebrates, reptiles, mammals and fish. We have provided a summary of these in Table 1a. However, identifying COEs is immensely difficult. To date few studies explicitly mention the term ‘carry-over effects’, and those that do are generally from research on avian species (e.g. Norris et al. 2004; Brown & Sherry 2006; Sorensen et al. 2009). By searching indirectly for studies mentioning ‘winter’ and ‘reproductive success’, we were able to find several studies from a range of taxa that demonstrated evidence of COEs, and we have provided these as a non-exhaustive list in Table 1. Many studies established a mechanistic link between a non-breeding season process and subsequent reproductive success, which we have labelled ‘True COEs’, (Table 1a). Several studies found that animals skip reproduction in some years and hypothesized that non-breeding season food availability was likely responsible for determining reproductive frequency (e.g. Green turtles Chelonia mydas; Broderick, Godley & Hays 2001). However because they did not directly quantify non-breeding season food availability we cannot be certain of the mechanism driving the observed differences between individuals and as such have placed them in a separate section labelled ‘Possible COEs’ (Table 1b). We acknowledge that the majority of COEs described to date, and thus the majority of the examples in this review are from ornithological studies, but argue that this is due in part to the difficulty in identifying COEs in some animal systems. For example in mammals, poor overwinter resource availability can lead to decreased reproductive success (Table 1b). This could be due to poor nutrition/condition prior to breeding affecting the probability of conception (e.g. Bighorn ewes, Festa-Bianchet 1998), which we consider to be a COE because state upon breeding is likely a product of processes in the previous season. However because mammals are able to delay implantation of embryos, or even resorb embryos during gestation (Weir & Rowlands 1973), it is plausible that non-breeding season food availability can exert direct effects on fitness by controlling how many embryos a female ‘allows’ to complete a full term of gestation. Therefore whilst some studies report a link between non-breeding season food availability and downstream reproductive success (Table 1b), we cannot irrefutably label these COEs because the mechanism by which this food limitation controls reproductive success is unknown. Studies from other taxa also report links between non-breeding season variables and later fitness that do not qualify as COE. For example Olivier et al. (2005) found that in years with greater ice cover over winter, snow petrels (Pagodroma nivea) had higher reproductive success the following summer. Superficially it appears a non-breeding season environmental variable has exerted a downstream effect on fitness, and it is tempting to label this a COE. In reality, the authors hypothesize that because sea ice promotes the overwinter survival of krill, more ice in winter ensures a larger supply of food during breeding and chick-rearing, resulting in increased reproductive success (Olivier et al. 2005). This is not a COE because the effect of food supply is operating intra- as opposed to inter-seasonally. We therefore urge caution when trying to characterize COEs without a full understanding of the mechanism behind observed differences in fitness indices. In spite of this any study linking non-breeding season variables (either exogenous variables such as habitat quality e.g. Nilsen, Linnell & Andersen (2004), or endogenous variables such as change in mass e.g. Cook et al. (2004)) with downstream performance has the potential to be a COE and with a basic understanding of the mechanism involved should be labelled as such. Due to the difficulty associated with identifying COEs, we consider the list of examples in Table 1 to be non-exhaustive as there are likely many more studies describing COEs without having identified them as such.

Processes that drive carry-over effects

While the role of habitat quality in generating COEs in migrants has recently been reviewed (Norris & Marra 2007; Ratikainen et al. 2008), many of the principles described here will equally apply to non-migratory taxa and there are likely to be a number of other potentially interacting drivers in addition to habitat. In this section, we review the major known mediators of COEs in a variety of species. We argue that several likely drivers of COEs remain little studied and poorly understood.

Social status, density and habitat quality

COEs arise when individuals make the transition between seasons in different states, either in terms of stores of a dietary component (e.g. protein) or in terms of bodily condition (e.g. immunocompetence). There are three main mediators that can drive these asymmetries in state: social status, density and habitat quality. Each of these mediators could appear in their own section and be given individual treatment. However, in many animal systems all three are intimately linked and it is the interaction between them that can propagate changes in condition parameters – such as the increase in fat stores over the non-breeding season, or gradual decline in immunocompetence. For instance, in any animal system where dominant individuals monopolize resources and exclude subordinates there is potential for the generation of asymmetries in resource access among individuals (Catry et al. 2004; Norris et al. 2004; Lu & Zheng 2007), which in turn can propagate COEs. In American Redstarts dominant individuals regularly exclude subordinates from high-quality mangrove habitats during winter where insect abundance is higher than in drier scrub habitat (Norris et al. 2004). This permits them better access to resources (food) and as a result means they are in better condition at the start of breeding than scrub-living birds (Marra, Hobson & Holmes 1998). COEs mediated by social hierarchies should be density-dependent because the monopolization of higher quality resources by dominants compromises access by other individuals, resulting in them receiving a smaller ‘share’. Subordinates (females and juveniles in the case of the American redstart; Marra, Hobson & Holmes 1998) are therefore forced into poorer quality territories with lower resource abundance or quality. The dominants experience a positive carry-over effect from superior overwinter nutrition allowing them to initiate breeding in better physical condition, whilst subordinates experience a negative carry-over effect from inadequate overwinter nutrition leading to suboptimal body condition during breeding (Marra, Hobson & Holmes 1998; Norris et al. 2004; Studds & Marra 2005; Norris & Taylor 2006). In this example, all three mediators are in operation. Dominant individuals (of higher social status) exclude subordinates from the superior quality habitat (thus securing the best resources) whilst simultaneously reducing density in their habitat and increasing it in the poorer habitat where subordinates have been forced. Consequently separating the effect of a single mediator would be difficult. For instance to link density dependence irrefutably to COEs, one would have to place the same individuals in the same patch of habitat in two consecutive years under low and high density respectively. To date this has not been done, and arguably is logistically difficult to achieve due to issues with manipulating density in the wild in a consistent fashion whilst controlling other factors such as food availability.

Although we have used the example of a bird species here, dominance-mediated habitat segregation is widespread among vertebrates and has been observed in a variety of taxa including mammals (e.g. Murray, Mane & Pusey 2007), reptiles (e.g. Calsbeek & Sinervo 2002) and fish (e.g. Stradmeyer et al. 2008). For example, chimpanzees (Pan troglodytes) of higher social rank have been shown to exclude subordinates from their territories, especially during times of food scarcity (Murray, Mane & Pusey 2007), obtain a higher quality diet (Murray, Eberly & Pusey 2006) and show higher reproductive success (Pusey, Williams & Goodall 1997). Thus, we hypothesize that COEs resulting from dominance asymmetries are likely to be a common phenomenon.

Reproductive effort

Most studies on COEs tend to focus on non-breeding season processes that exert a measurable outcome during the breeding season (such as reproductive success e.g. Ebbinge & Spaans 1995). However the converse effect, whereby breeding season processes exert effects in the non-breeding season, is less frequently investigated. One possible scenario in which COEs might arise is if reproductive effort causes a decrease in survival during the following winter. For example, Daan, Deerenberg & Dijkstra (1996) found that increasing the work rate of breeding adults by experimentally increasing brood size resulted in increased overwinter mortality of European kestrels (Falco tinnunculus). Research on starlings (Sturnus vulgaris) has shown that individuals moulting under conditions of decreasing daylength (as would be experienced if moulting had been delayed until after completion of reproduction), completed their moult more quickly and thus had a poorer quality plumage compared to those that had initiated moult earlier and took longer to complete (equivalent to non-breeders; Dawson et al. 2000). Similar work on blue tits (Parus caeruleus) has shown that birds delaying moult in this fashion consequently had higher thermoregulatory costs the following winter due to poor feather quality (Nilsson & Svensson 1996). Reproduction can effect changes in an organism’s state such as it’s energy reserves, feather quality (e.g. Dawson et al. 2000) and immunocompetence (e.g. Hanssen, Folstad & Erikstad 2003), but the important point here is that consequences of that change may not be realized until the following season. The effects of changes in reproductive output on survival the following season presents an interesting scenario because this COE could simply be manifested through the classic life-history trade-off (i.e. increased reproductive output at the expense of subsequent survival/longevity). However, the mean reproductive effort of the population could also be changed through, for example, an increase in nest predation rates. Such events may not only decrease the number of individuals moving into the non-breeding period (seasonal compensation effect) but also increase the probability of survival of those individuals who failed to breed. It is also worth noting that reproduction is not the only factor that can mediate COEs manifested through reduced downstream survival. It is also feasible that inclement weather or poor food availability during breeding can increase mortality during return migration or on the wintering grounds, irrespective of reproductive effort.

The examples we have used above refer to survival (e.g. Daan, Deerenberg & Dijkstra 1996; Nilsson & Svensson 1996), but it is equally feasible that other indirect correlates of fitness can result from COEs. For example reduced feather quality in the non-breeding season may reduce foraging efficiency, or reduce ability to defend a higher-quality territory (or both), and thus result in lower body condition by the end of the winter period. These in turn may propagate further COEs by affecting reproductive output in the next breeding season – see ‘Capital Breeding’ in the next section.

Life history and ecological characteristics and the strength of carry-over effects

Capital breeding

Many animals finance reproduction from energy stores (predominantly fat) gained in the months prior to the breeding season (‘capital’ breeding strategy), rather than using energy gained concurrently during reproduction (‘income’ breeding strategy, Drent & Daan 1980; Houston et al. 2007). As a general rule among capital breeders, individuals with larger fat stores upon initiation of reproduction tend to have the greatest reproductive success (Prop & Black 1998), with fitness penalties incurred by those individuals who initiate reproduction with lower/insufficient energy stores (e.g. Ebbinge & Spaans 1995; Drent et al. 2003). As such, COEs are expected to be common among capital breeders because any event which hinders the rate or amount of resource acquisition/storage prior to the breeding season has the potential to reduce the amount of energy subsequently available for reproduction (e.g. Inger et al. 2006a,b; Inger et al. 2008). Poor pre-breeding rates of energy intake have been shown to adversely affect reproduction in a wide range of vertebrate taxa including birds (e.g. Bolton, Houston & Monaghan 1992; Ebbinge & Spaans 1995; Gill et al. 2001; Prop, Black & Shimmings 2003), mammals (e.g. Festa-Bianchet 1998; Cook et al. 2004; Persson 2005) and fish (e.g. Rideout, Rose & Burton 2005; Bunnell, Thomas & Stein 2007; Kennedy et al. 2008). Gray whales (Eschrichtius robustus), rely almost entirely on stored fats for reproduction and tend to have poorer calving success in winter when the previous summer’s ice cover restricts foraging, resulting in poorer physical condition prior to giving birth (Perryman et al. 2002). Note that in this case it is a density independent process (environmental variation) that has caused the initial restriction in individual access to resources, but this effect is in turn modified by density-dependent processes, because the net reduction in per capita food availability is affected by the population size. Similarly COEs have also been described in species that rely on food caches. The habit of caching food was described by McNamara, Houston & Krebs (1990) as ‘functionally analogous’ to storing body fat for future reproduction. Persson (2005) found that the size of the winter food cache may be an important determinant of reproductive success in the wolverine Gulo gulo, which, in physiological terms, is an income breeder.

It is clear that separating animals into ‘capital’ and ‘income’ breeder categories is immensely difficult, mostly because many recognize that such a distinction is not a dichotomy but a continuum (e.g. Drent, Fox & Stahl 2006). We do not go into the details of this distinction here, but instead recommend Stephens et al. (2009) for an excellent in-depth treatment on the use of these ‘labels’ in ecology and subsequent consequences for quantifying reproductive costs. The problem of the capital/income distinction is epitomized by the example of Warner et al. (2008), who found that in the lizard Amphibolurus muricatus tended to use lipids from ‘capital’ and proteins from ‘income’ during egg production and as such qualified as both type of breeder. In the context of COEs, this means that simply assessing one index of state in the non-breeding season (such as fat mass) may not in fact explain any of the variation in downstream reproductive success because it is in fact being controlled by a completely different factor (perhaps protein (muscle) stores). The study of Warner et al. (2008) illustrates the complexity of the reproductive process and shows that broad classification at the organismal level as either capital or income breeder is not immensely useful, especially when attempting to identify COEs. Instead one must look at the dietary components (protein, fat, antioxidants, minerals etc.) individually to assess what is used as capital or income. Variation among species in the extent to which they use stores of these components will undoubtedly have large consequences for susceptibility to COEs. Whilst the relative contributions of protein and fat have been studies in several species (e.g. Gauthier, Bety & Hobson 2003, Warner et al. 2008), there are many other dietary components that can determine an organism’s state by the end of the non-breeding season. We consider these in the section ‘Non-Energetic Mediators of Carry-Over Effects’.

Time limitation

Whether resident or migratory, all animals are subject to time constraints within the annual cycle and must often co-ordinate timing of reproduction with environmental factors in order to maximize reproductive success and survival. For example female mink (Mustela vison) have been shown to synchronize breeding so that lactation occurs with high availability of Pacific salmon carcasses (Ben-David et al. 1997). Similarly passerine bird species often adjust lay date so that hatching of offspring coincides with peak abundance of caterpillars, thus matching peak food demand from offspring with peak food availability and increasing offspring fitness (e.g. Naef-Daenzer & Keller 1999; Both et al. 2009). Failure to synchronize demand for, and availability of, food in such a manner can result in significant declines in reproductive success or survival probability, and optimal timing can often be complicated by constraints on the parent (e.g. Drent & Daan 1980; Rowe, Ludwig & Schluter 1994). COEs can arise when non-breeding season processes affect an organism’s ability to time reproduction to coincide with these ephemeral resource pulses, such as caterpillar emergence or spawning of salmon. Positive COEs can arise when non-breeding season processes permit an animal to accurately match timing of breeding with peak resource supply, potentially resulting in increased reproductive success. Conversely, negative COEs arise when these same processes delay reproduction to cause a temporal mismatch between demand for, and supply of food. Constraints on time are even more pronounced in migratory animals due to a number of factors such as adverse weather (Bety, Gauthier & Giroux 2003; Bety, Giroux & Gauthier 2004; Gunnarsson et al. 2006), time taken to travel (Bearhop et al. 2005), and mid-journey staging for refuelling (Schaub & Jenni 2001; Bearhop et al. 2004), which can interact to delay arrival on the breeding grounds and thus time available to raise young. Moreover species breeding at higher latitudes tend to have shorter breeding windows because of the associated reduction in the period of favourable climatic conditions (Gauthier 1993). Early birth is thought to be advantageous because it allows juveniles longer growth periods, enabling them to sequester more energetic/nutritional resources and/or gain more experience by the time winter arrives, thereby increasing the chances of survival (Lepage, Gauthier & Menu 2000). Thus, in migratory animals any process which delays arrival at the breeding grounds can generate COEs because of the subsequent delay in initiation of reproduction. For example in migratory birds, low rates of food intake on wintering/staging grounds (Gill et al. 2001) and/or adverse weather conditions (Bety, Giroux & Gauthier 2004) can both act to lengthen migration and/or delay departure. Even among ‘capital-breeding’ migrants with suitable levels of macronutrient stores, time constraints may act to reduce fitness if individuals arrive in good condition, but with insufficient time to breed (Ebbinge & Spaans 1995).

In summary, time constraints within the annual cycle can generate COEs when non-breeding season processes prevent reproduction from occurring during periods of favourable environmental quality, either in terms of food availability (e.g. Ben-David et al. 1997; Both et al. 2009) or weather conditions (Lepage, Gauthier & Menu 2000; Bety, Gauthier & Giroux 2003). This effect may be particularly pronounced in migratory species, which need to co-ordinate migratory timing with periods of favourable climatic conditions and food availability at multiple stages along migratory routes.

Non-energetic mediators of carry-over effects

It is highly unlikely that energy is the only nutritional currency capable of driving COEs, and therein lies the largest gap in our understanding. All vertebrates are capable of storing non-energy dietary components, including macronutrients such as protein and micronutrients such as antioxidants, and thus may be forced to rely on stored ‘capital’ during periods of limited dietary supply. To complicate things further, some micronutrients such as antioxidants are actually stored in fat (Klasing 1998) and so in species classically considered to be ‘capital breeders’ there is likely to be a high degree of correlation between the liberation of stored energy and stored antioxidants. Here we identify specific groups of micronutrients, describe the biological processes that they control (e.g. egg formation) and thus suggest the mechanisms by which they may propagate COEs. We argue that research into these potential mechanisms and their interaction with previously established energy-mediated COEs is vital in order to fully understand how individual-level processes scale up to population-level effects.


Among micronutrients that could potentially mediate COEs, antioxidants are particularly interesting because they are thought to be an important currency that underpins life history trade offs (see Catoni, Peters & Schaefer 2008 for a review). Reactive oxygen species are produced as by-products of metabolism and have the potential to seriously damage DNA, proteins and lipids if they exceed the capacity of antioxidants to inactivate them (von Schantz et al. 1999). This process, known as oxidative stress, has been implicated in the aetiology of many diseases, impaired reproduction, and the ageing process (von Schantz et al. 1999). Certain antioxidants such as vitamin E, carotenoids and anthocyanins can only be obtained through the diet, and therefore supplies may be limiting for adequate antioxidant defence (von Schantz et al. 1999; Catoni, Peters & Schaefer 2008). Such oxidative stress (induced experimentally) has been suggested to be causally linked with over-winter mortality in adult European kestrels (Falco tinnunculus) (Daan, Deerenberg & Dijkstra 1996). In addition, the direct allocation of antioxidants to reproductive activities (e.g. provisioning of eggs in oviparous species (Blount, Houston & Moller 2000; Surai et al. 2001); maintenance of sperm quality (Blount, Moller & Houston 2001; Helfenstein et al. 2010); and allocation to sexual ornamentation (von Schantz et al. 1999; Pike et al. 2007) means such antioxidants are unavailable for somatic maintenance. Importantly, as with energy intake, the availability of diet-derived antioxidants such as carotenoids varies seasonally and spatially (Slagsvold & Lifjeld 1985; Isaksson, Von Post & Andersson 2007) and certain important antioxidants (e.g. vitamin E) can be stored in organs for subsequent use (Royle et al. 1999). We therefore might expect strong antioxidant driven COEs, particularly in individuals or species that invest highly in gamete production, or are subject to strong sexual selection.

Essential amino acids

As is the case with many antioxidants, essential amino acids cannot be synthesized de novo by animals and must be obtained from dietary sources. Among these the sulphur-containing amino acids (SAAs) methionine and cysteine can often be limited in the environment (Klasing 1998). In birds SAAs are essential for both feather synthesis and egg production and as a consequence are often stored in muscle to be mobilized when needed (Houston, Donnan & Jones 1995; Wylie, Robertson & Hocking 2003). Thus seasonal variation in the availability of SAAs could similarly act as a driver of COEs.


Several minerals are essential for reproduction and, as in the examples above, they could underpin some COEs as access to them can vary both spatially and seasonally. For example seasonal fluctuations in magnesium and phosphorous availability reduce reproductive success in domestic sheep and cattle (Yokus, Cakir & Kurt 2004; Yokus & Cakir 2006) and low calcium availability has been shown to increase lead uptake in Zebra finches Taeniopygia guttata (Dauwe et al. 2004). This causes several downstream effects including reduced male fertility, reduced hatching success (Dauwe et al. 2004) and reduced immunocompetence in females (Snoeijs et al. 2005). Many essential minerals can also be stored for later use (Klasing 1998) and female white-tailed ptarmigans Lagopus leucurus accumulate extra calcium in their leg bones several weeks prior to breeding, most likely as a response to the elevated calcium demand associated with egg formation (Larison, Crock & Snow 2001).


It is clear that variation in access to micronutrients has the potential to generate COEs. However, empirical studies of such effects are lacking. For example white-tailed ptarmigans (Larison, Crock & Snow 2001) could incur fitness penalties in the form of negative COEs if they stored insufficient amounts of calcium during the pre-breeding period. Further research into such fields is vital, because if winter micronutrient availability has the capacity to affect summer reproductive success, this could have an influence on life histories and populations equal to that of energy-driven COEs. If this is the case, then simple metrics such as body mass may not be enough to predict or measure the strength of COEs. Data from physiological samples, such as blood and tissue where practical, would allow us to derive measures of both circulating antioxidant levels and stores, for example. Conceivably two individuals could be of similar body mass, and thus appear superficially to be in similar condition, but in practice have vastly different abilities to deal with systemic challenges such as oxidative stress. Estimates of both micronutrient and macronutrient stores in the non-breeding season are vital in order to fully understand the interactions between different drivers of COEs.

Measuring the strength of carry-over effects

Separating the relative effects of intrinsic and extrinsic quality

Whilst differential resource access is undoubtedly a major factor underpinning differences in breeding performance driven by COEs, the mechanisms causing such differences are likely to be derived from a combination of intrinsic (the quality of the individual) and extrinsic (the quality of the habitat in which it lives) factors. Whilst this is recognized, to date it has proved immensely difficult to tease apart their relative contributions (Daunt et al. 2006). Norris & Marra (2007) argued that understanding how individuals of different intrinsic quality will react to changes in habitat quality is vital in order to be able to predict how population abundance will in turn be affected. In concordance with this, we argue that one cannot accurately model population dynamics without first estimating the magnitude of COEs operating in that population (both through delayed density-dependent and density-independent processes). In turn, one cannot accurately model COEs without first estimating the magnitude of intrinsic and extrinsic quality effects.

Studds & Marra (2005) experimentally upgraded subordinate American Redstarts from a poor to a high quality winter habitat and found they maintained body mass, departed earlier and had a higher return rate than those that were not upgraded. While this is compelling evidence for the importance of extrinsic factors (habitat quality in this case), there is no way of controlling which individuals actually occupy the vacant territories in this situation. This is important as it is easy to imagine that the individuals that first occupy the vacant territories are themselves of higher quality than the average for all potential new occupants (since they are best able to defend the territories). Similarly the reverse experiment cannot be performed in the wild because high quality individuals will not remain in poor quality habitats. Being able to disentangle the relative contributions of extrinsic and intrinsic quality effects is crucial if we are to understand the true role that COEs play in driving individual fitness and population dynamics. Therefore, we provide an outline of two novel approaches designed to investigate this important question.

Manipulating extrinsic quality in the field

An alternative to the Studds & Marra (2005) upgrade approach mentioned above would be to manipulate territories rather than individuals. For example, animals could be allowed to remain in their chosen territories, with territory quality being manipulated through the provision of supplemental food or removal of key resources (such as food sources, roost sites etc.). The advantage of this approach is that dominants and subordinates of known intrinsic quality can be selectively targeted. Calsbeek & Sinervo (2002) used the side-blotched lizard (Uta stansburiana) to test the ideal despotic distribution theory by manipulating territories of individual lizards to be of varying quality with respect to ‘thermal resources’. They were then able to derive measures of survival, territory size and fitness post-manipulation. Conveniently, male side-blotched lizards display one of three throat-colours based on their dominance and territory size (Sinervo & Lively 1996), and thus assessment of relative intrinsic quality is far more straightforward than in other systems. Manipulating territories is also potentially much simpler here than for a bird living in a mangrove habitat for example. Individuals can be marked (Calsbeek & Sinervo 2002), and consequently it is easy to see when they have been displaced from upgraded territories, or how territory size has changed post-manipulation. There are several experimental possibilities in this instance. One could homogenize the territories of higher and lower intrinsic-quality individuals (i.e. provide similar amounts of thermal resources), and measure relative fitness whilst taking into account how dominants may try to maintain larger territories. Similarly, it would be possible to target certain individuals specifically for an upgrade of nutritional quality of habitat, by provisioning directly within the core of its territory. Animal systems like the side-blotched lizard where there is social signalling offer considerable potential for testing the relative importance of intrinsic versus extrinsic quality in driving COEs, purely because these signals can give a preliminary proxy for dominance status. Numerous taxa display social signals in a similar fashion and thus could potentially be suitable, including other reptiles (e.g. Martin & Lopez 2009), mammals (e.g. Bergman, Ho & Beehner 2009; Marty et al. 2009) and birds (e.g. Murphy et al. 2009a,b).

Manipulating extrinsic quality in captivity

The main problem with manipulating habitat quality in the field is that while it is relatively easy to enhance the quality of a poor territory, it is much more difficult to reduce the quality of a good territory. Moreover it is questionable whether a high quality individual would remain in a reduced quality territory in a wild situation, or not simply alter its territory size to compensate (Calsbeek & Sinervo 2002). This raises the possibility of experiments conducted in captivity. Individuals could be held in enclosures within which habitat quality, size of territory and access to mating opportunities can be controlled. Dominance could be assessed in captivity (e.g. Murphy et al. 2009b) and subsequently assigned to one of four treatments: high quality individual-high quality habitat; high quality individual-low quality habitat; low quality individual-high quality habitat; low quality individual-low quality habitat. This could be carried out on a variety of species and treatments could be achieved in a number of ways, for example via provision of extra food, increased costs of thermoregulation, increased costs of foraging, increased costs of predation risk etc., whilst response variables would be linked to patterns of mass acquisition or reproductive success. This experiment would rather neatly allow an assessment of the relative importance of dominance versus extrinsic quality in explaining COEs.

Variations on a theme: testing for other mediators

Given that we have already identified several other potential mediators of COEs, it would be feasible to use similar experiments to those outlined above as a template to investigate them. For example it would be relatively straightforward to manipulate the quality of the diet/habitat with respect to antioxidants or mineral provision while maintaining a constant calorie intake and thus disseminating the relative contributions of micro- and macronutrient components of the pre-breeding diet to fitness.

Future directions for carry-over effect research: state-dependent modelling

Much of the empirical work on COEs has yet to take advantage of state-dependent life history theory (Mcnamara & Houston 1996, 2008) already embraced by those working on long-term downstream effects. This is rather surprising since integrating the two fields is likely to provide a number of avenues for progress. Indeed, some of the more recent annual routine models can predict the timing of events in an individual’s life within the context of the entire breeding cycle and how they are influenced by variation in food abundance and density (Houston & Mcnamara 1999; Barta et al. 2006, 2008; Fero et al. 2008) and thus could lend themselves very well to the study of seasonal interactions. Barta et al. (2008) suggest that for organisms living in seasonal environments it is likely that the timing of events such as initiation of moult, breeding and migration is a key determinant of individual fitness, hinting at carry over effects, since moult in most birds happens many months prior to breeding. The utility of state-dependent modelling lies in its ability to parameterize these factors and then calculate the optimal timing of each of these behaviours, thus maximizing fitness and reducing the negative aspects of COEs. Therefore they also allow us to identify which of a given set of variables exerts the most powerful influence on COEs. To date, state-dependent modelling has been used to study optimal moult strategies in both non-migratory (Barta et al. 2006) and migratory birds (Barta et al. 2008). However it could be easily applied to investigating factors such as optimal rates of overwinter resource accumulation, one of the main driving factors of COEs in the current literature. More importantly, density-dependent effects can also be incorporated (Houston & Mcnamara 1999; Barta et al. 2006, 2008), and thus by varying the strength and form of density dependence it is theoretically possible to separate changes in population size due to COEs (i.e. reduced reproductive output, not mortality) from those due to inter-seasonal density effects resulting from mortality events (Norris & Marra 2007; Barta et al. 2008; Ratikainen et al. 2008), which to date has proved difficult to achieve (Norris & Marra 2007). Theoretical work such as this would allow us to model a multitude of determinants of state (for example fat, protein and antioxidant stores) and for each one calculate when in the annual cycle their limitation is most likely to propagate COEs, as well as which groups of individuals are most likely to be affected. More importantly it may help us to identify the nature of COEs in a specific system. For example fitness costs can be incurred either by reduced breeding effort in one breeding season (i.e. reduced clutch or litter) or by making fewer total reproductive attempts over sequential years (i.e. reproductive skipping, see Mcnamara et al. 2004). Theoretical work investigating optimal annual routines may be crucial in determining which of these is most likely and thus guide future empirical work in the study and identification of COEs. The versatility of this approach, and its ability to investigate optimal strategies at the individual level for a variety of parameters in the context of the entire annual cycle, makes it a potentially very powerful tool for the study of COEs. We suggest that research effort in this field is vital for the advancement of our understanding of how COEs operate, and more importantly how they interact with other demographic processes such as density-dependent compensation.


Whilst COEs have been identified in a wide variety of taxa, their interactions with, and ability to be masked by, simultaneously occurring inter-seasonal density effects means they are probably much more common than previously thought. Until now most work on COEs has focused on energy as the key currency using relatively simplistic but easily obtained measures of state such as body mass. However it seems probable that several types of micronutrients, antioxidants in particular, could play important roles in the mediation of COEs. This warrants further investigation and will require physiological sampling, at least in terms of blood sampling for assays of circulating levels of micronutrients. Indeed further investigations are likely to highlight just how complex these phenomena are, as multiple drivers and factors will interact to modify their severity. COEs are likely to explain a significant amount of variation among individual life histories across a variety of species and we are only just beginning to understand the key processes involved. Empirical research on non-avian taxa should be a priority, and will improve our ability to predict the occurrence and strength of COEs in vertebrate taxa as a whole. The experimental procedures we have outlined here represent only a handful of possibilities for investigating COEs, and we offer them in the hope that they will stimulate a wealth of vital new research in this field. Similarly we suggest that tools such a state-dependent modelling are likely to be one of the most useful means of predicting vulnerability to COEs at different stages of the annual cycle, and thus should be a focus of future research. Understanding the manner in which COEs influence individual fitness and how they interact with density-dependent processes will significantly advance our understanding of population dynamics in animal systems, with important consequences for conservation and management across a range of vertebrate taxa.


We are immensely grateful to Alasdair Houston, Silke Bauer and one anonymous reviewer for insightful comments that substantially improved an earlier version of this manuscript. XH was funded by a NERC studentship with a CASE partnership with the Wildfowl and Wetlands Trust.