The impact of reproductive investment and early-life environmental conditions on senescence: support for the disposable soma hypothesis
Behavioural Ecology and Self-Organization, Centre for Ecological and Evolutionary Studies, University of Groningen, Groningen, The Netherlands
Centre for Ecology, Evolution and Conservation, School of Biological Sciences, University of East Anglia, Norwich, UK
Correspondence: Martijn Hammers, Behavioural Ecology and Self-Organization, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 11103, 9700 CC, Groningen, The Netherlands. Tel.: +31 50 363 5928;
Several hypotheses have been put forward to explain the evolution of senescence. One of the leading hypotheses, the disposable soma hypothesis, predicts a trade-off, whereby early-life investment in reproduction leads to late-life declines in survival (survival senescence). Testing this hypothesis in natural populations is challenging, but important for understanding the evolution of senescence. We used the long-term data set from a contained, predator-free population of individually marked Seychelles warblers (Acrocephalus sechellensis) to investigate how age-related declines in survival are affected by early-life investment in reproduction and early-life environmental conditions. The disposable soma hypothesis predicts that higher investment in reproduction, or experiencing harsh conditions during early life, will lead to an earlier onset, and an increased rate, of senescence. We found that both sexes showed similar age-related declines in late-life survival consistent with senescence. Individuals that started breeding at a later age showed a delay in survival senescence, but this later onset of breeding did not result in a less rapid decline in late-life survival. Although survival senescence was not directly related to early-life environmental conditions, age of first breeding increased with natal food availability. Therefore, early-life food availability may affect senescence by influencing age of first breeding. The disposable soma hypothesis of senescence is supported by delayed senescence in individuals that started breeding at a later age and therefore invested less in reproduction.
Lifespan is one of the most important determinants of fitness in many vertebrates in which lifetime reproductive success increases with lifespan (Clutton-Brock, 1988; Newton, 1989). Therefore, to understand the evolution of life histories, it is important to identify the factors and trade-offs that influence survival (Stearns, 1992). Senescence is defined as the decline in performance with advancing age and is caused by a progressive decline in an individual's physiological and cellular functions (Kirkwood & Austad, 2000). Individuals may show senescent declines in performance across a wide range of traits, such as foraging (Lecomte et al., 2010), immune function (Hayward et al., 2009) and body mass (Nussey et al., 2011). This reduced performance may ultimately lead to a progressive decrease in survival with age.
Senescence occurs in many taxa in the wild (Jones et al., 2008; Ricklefs, 2008; Nussey et al., 2013), but how and why individuals age remains poorly understood. Several hypotheses have been proposed to explain the evolution of senescence (Kirkwood & Austad, 2000), all of which assume that the force of natural selection becomes weaker on older age classes. The effect of selection diminishes because it is proportional to the number of individuals alive in each age class, which diminishes with age because of extrinsic mortality. The ‘antagonistic pleiotropy’ hypothesis proposes that senescence may be caused by weakened selection against late-acting deleterious mutations, leading to greater selection on genes that are beneficial during early life, but that have deleterious effects later on (Williams, 1957). A related hypothesis, the ‘disposable soma’ hypothesis, suggests that an individual should only invest in maintenance of the soma for as long as it has a realistic chance of survival (Kirkwood, 1977; Kirkwood & Holliday, 1979; Kirkwood & Austad, 2000). Higher investment in reproduction is expected at younger ages at the cost of investment into maintenance and repair, even though this may lead to higher levels of damage to cells and tissues and, therefore, senescence (Kirkwood & Rose, 1991). The accumulation of this damage during early life may then be manifested in a late-life decline in survival. Thus, the antagonistic pleiotropy and disposable soma hypotheses predict a trade-off between investment during early life and late-life survival (Kirkwood & Rose, 1991). The main difference between these two hypotheses is that the antagonistic pleiotropy hypothesis focuses on the effects of genes with pleiotropic effects, whereas the disposable soma hypothesis focuses on the optimal allocation of resources to reproduction, self-maintenance and repair (Kirkwood & Austad, 2000).
Clearly, to understand the evolution of senescence, it is important to investigate how early-life environmental conditions and reproductive investment influence survival senescence (the progressive age-dependent decline of survival probability). Several studies have shown that survival senescence occurs in a wide variety of taxa, including birds and mammals (Jones et al., 2008; Péron et al., 2010), insects (Bonduriansky & Brassil, 2002) and fish (Reznick et al., 2004). However, few studies have investigated which factors contribute to differences in senescent declines in survival probability between individuals within a population. Factors that are predicted to affect the onset and the rate of survival senescence include early-life reproductive investment and early-life environmental conditions. Several studies have shown that individuals that allocate more resources to reproduction during early life (e.g. produce more offspring) suffer from increased survival senescence later in life (Luckinbill et al., 1984; McCleery et al., 1996; Orell & Belda, 2002; Reid et al., 2003; Pettay et al., 2005; Reed et al., 2008), but see Gaillard et al. (2003). One way to invest less in reproduction during early life is to delay breeding until well into adulthood. Hence, individuals that start breeding at a later age are expected to show delayed senescence and/or a less rapid decline in late-life survival compared with individuals that start breeding at a younger age. Furthermore, the environmental conditions experienced during early life are predicted to influence senescence. For example, animals experiencing poor conditions during early life (e.g. low food availability or high local breeding density) may show steeper declines in late-life survival (Nussey et al., 2007), but see Millon et al. (2011).
The effective study of survival senescence in the wild requires long-term, longitudinal data sets of individually marked animals living in closed populations with low levels of extrinsic mortality (Monaghan et al., 2008). Here, we use the long-term data set (1981–2011) on the Cousin Island (29 ha, 4°20′S, 55°40′E) population of the Seychelles warbler (Acrocephalus sechellensis). In this population, the majority (> 96%; Richardson et al., 2001) of individuals have been individually colour-marked, and there is virtually no interisland dispersal (Komdeur et al., 2004). This lack of dispersal provides a rare opportunity to follow nearly all individuals within a single population from birth to death. Importantly, predation on adults is absent, thus individuals have relatively long lives (life expectancy at fledging = 5.5 years (Komdeur, 1991), maximum lifespan=17 years (Barrett et al., 2013)). Food availability is an important determinant of environmental conditions, with survival declining with lower food abundance in territories (Komdeur, 1991) and with higher local breeding density (Brouwer et al., 2006). Although survival senescence has not yet been investigated, previous studies have indicated that reproductive senescence occurs in this species, with reproductive output in females declining beyond 6 years of age (Komdeur, 1996c; Hammers et al., 2012). Although Seychelles warblers can breed successfully in socially monogamous pairs, cooperative breeding occurs frequently (Komdeur, 1992; Richardson et al., 2001) and is primarily driven by the shortage of high-quality breeding vacancies (Komdeur, 1992; Komdeur et al., 1995). Not all mature offspring start independent breeding at the same age. Mature offspring often become subordinates within a territory until a breeding position becomes vacant. These individuals often gain both direct and indirect benefits from being a subordinate within a territory (Richardson et al., 2002). Komdeur (1992) showed that 1-year-old Seychelles warblers may produce more offspring over their lifetimes by remaining in a high-quality territory for several years as a subordinate helper than when they would by starting breeding immediately in a lower quality territory. Indeed, individuals that were born in high-quality territories were more likely to stay as subordinate helpers rather than filling breeding vacancies on lower quality territories than individuals born in lower quality territories (Komdeur, 1992). An individual's eventual acquisition of a dominant breeding position within a territory may occur between 1 and 8 years of age (this study), although some adult individuals will die without ever acquiring a territory (Komdeur, 1992). Individuals that have acquired a breeding position typically remain on and defend the same stable territory until their death (99.1%, Komdeur & Edelaar, 2001), but see (Richardson et al., 2007).
In this study, we present a test of the disposable soma hypothesis of senescence. This hypothesis predicts a trade-off between investment during early-life and late-life survival, consequently, we predict that individuals that started breeding at a later age will show gentler declines in age-dependent survival probabilities and/or a delay in survival senescence because of reduced investment in early-life reproduction. In addition, we test the prediction that individuals that spent the first years of their lives in poor-quality territories with low food availability, or lived with many other individuals in a territory (so experiencing high local competition) will show lower late-life survival.
Materials and methods
Study population and data collection
The Cousin Island population of Seychelles warblers is at carrying capacity and comprises ca. 320 colour-banded adults of known age in ca. 115 territories (Brouwer et al., 2009). The population has been monitored since 1981 as part of a long-term study (Komdeur, 1992; Brouwer et al., 2012). The warbler's life history is characterized by high annual adult survival (84%), predominantly single-egg clutches (87% of clutches) and extended periods of parental care (3–6 months) (Komdeur, 1991, 1994b; Brouwer et al., 2006). The presence and dominance status of all individuals in each territory was determined during each main breeding season (June–September) from 1982 to 2011 (except in 1992 when no fieldwork was conducted, and 1982–1984 and 1991–1994, when only 68% of the territories were monitored; Brouwer et al., 2009). The ‘dominant’ male and female were defined (based on behavioural observations) as the primary, pair-bonded male and female in a territory, whereas the term ‘subordinate’ includes all other mature birds (> 8 months old) resident in the territory (Richardson et al., 2007). Typically, dominant individuals remain together in the same territory until one of the partners dies. The dead partner is then rapidly replaced by a subordinate, providing a rare opportunity for a subordinate to acquire a territory and a dominant breeding position. We recorded the birth year and the death year for each individual. As the resighting probability of breeding individuals is virtually one (0.98 ± 0.01 SE; Brouwer et al., 2010), and the probability of dispersal from the island is extremely low (0.10%; Komdeur et al., 2004), we assumed that all individuals that were not seen during the subsequent year, or thereafter, died in the first year they were missing. The sex of each individual that was sampled since 1994 was established using molecular sexing methods (Griffiths et al., 1998). Before 1994, the sex of individuals was established using a combination of body size and behaviour, which corresponds with molecular sexing (Komdeur, 1996a).
Territory quality, group size and age of first breeding
Seychelles warblers are purely insectivorous, taking insect food from the undersides of leaves and territory quality – measured in terms of insect prey availability in a territory – is important for reproductive success (Komdeur, 1991). To estimate local food availability, a territory quality index was calculated following Komdeur (1992) and van de Crommenacker et al. (2011), using the formula a·∑(cx·ix), where a is the territory size in hectares, cx is the total foliage cover for broad-leafed tree species x (sum of foliage score, see below) and ix is the mean monthly insect count for tree species x per unit leaf area (dm2). Territory size was determined by mapping observations of foraging and territorial defence behaviour at the territory borders. Foliage cover was scored by determining the presence of each tree species at 20 random points in every territory, in the following height bands: 0–0.75 m, 0.75–2 m and at 2 m intervals thereafter. Insect counts were estimated by monthly counting of the total number of insects on the undersides of 50 leaves for each tree species present in 14 different regions across the island, reflecting the regional variation in insect abundance. Because territory quality was not measured every year (measured in 1987, 1990, 1996–1999, 2003, 2004, 2006–2011), and because of fluctuations in overall territory quality between years, we calculated one index of territory quality for each territory, rather than extrapolating territory quality for the missing years. To this end, we calculated the standardized territory quality for each territory in each year (mean-centred by year and divided by the standard deviation) and averaged these values to obtain one value of quality for each territory for all years combined (Hammers et al., 2012). Because reproductive output increases with food availability, and food availability differs markedly between years (Komdeur, 1991, 1996b), population-level food availability was estimated based on the total number of birds that fledged in a certain year. Group size was the number of adult individuals (dominants and subordinates) resident in a territory and reflects the local competition for food (Brouwer et al., 2006). The age of first breeding (defined as when an individual became a dominant territory owner) provides a measure of early-life investment, as it reflects abrupt increased investment in reproduction and territorial defence. Although subordinates sometimes participate in aspects of reproductive behaviour (e.g. incubation, feeding dependent offspring, co-breeding) and territorial defence (Komdeur, 1994b; Richardson et al., 2001), the costs of territory ownership and reproduction can be assumed to be much greater for dominants compared to subordinates. For example, it was shown experimentally that dominant males, but not subordinate males, invest in energetically costly mate-guarding (Komdeur, 2001). Because vacant territories are extremely limited in this saturated population (only 16% of adults die each year; Brouwer et al., 2006), and because the loss of a territory typically means the end of the individual's reproductive career (c.f. Richardson et al., 2007), dominant individuals are expected to invest much more in territory defence compared to subordinates. In addition, almost all dominant individuals attempt to breed each year (average 98.2%; Komdeur & Daan, 2005), whereas only a proportion of subordinates contribute to nest care and territorial defence in each breeding attempt and, if so, to a lesser extent than dominants (Komdeur, 1991).
We used generalized linear mixed models with a binomial error distribution and logit link function to investigate age-dependent survival using r (2.13.0, R Development Core Team, 2011) in the package lme4 (0.999375, Bates et al., 2011). We recorded whether each individual was alive or not in each year. To control for differences in survival probabilities between years (Brouwer et al., 2006), we included year as a random effect in the analyses. The 116 birds that had been translocated from Cousin to neighbouring islands as part of the conservation programme (29 individuals in 1988, 29 individuals in 1990 and 58 individuals in 2004 (Komdeur, 1994a; Richardson et al., 2006)) were excluded from all analyses. Population-level food availability could not be established for individuals born before 1992 and so was included only for birds born after 1991. Furthermore, because the year of death could not be established precisely for birds that died in 1992, birds that were not observed in 1993 were assigned 1991 as their death year. Repeating all analyses using only individuals that were born after 1991 did not change the results of these analyses (analyses not shown).
To investigate late-life declines in survival, we restricted the data set to include only dominant individuals older than 6 years of age. We did so for two reasons: (i) 6 years demarks the mean onset of reproductive senescence in Seychelles warblers (Komdeur, 1996c; Hammers et al., 2012); (ii) all individuals that eventually became dominant breeders had acquired a territory when 7 years old, except for one individual that gained dominance at 8 years of age. We used territory quality (‘early-life territory quality’), population-level food availability (‘early-life population-level food availability’) and group size (‘early-life group size’), averaged over the first 3 years of an individual's life, as measures of early-life conditions. Along with age, the starting model (143 females and 127 males older than 6 years) contained sex, the age at which an individual became a dominant breeder, early-life territory quality, early-life population-level food availability and early-life group size. A significant main effect, along with a significant decline in survival with age, would indicate that senescence is advanced (left-shifted) or delayed (right-shifted). All two-way interactions with age were included to investigate whether the rate of decline in survival with age was associated with gender, age of first breeding or early-life conditions. Nonsignificant terms (P >0.05) were then – in order of least significance and starting with the interaction terms – removed until a model was retained with only significant terms. Repeating the model including all (significant and nonsignificant) fixed effects, but excluding nonsignificant interactions, gave similar results.
Because early-life environmental conditions may affect senescence through their effects on age of first breeding (Nussey et al., 2007), we used generalized linear models with a quasi-Poisson error distribution and log link function to investigate how age of first breeding is related to natal (first year of life) conditions. In these models, we included cohort (birth year) and sex as factors, and natal territory quality, natal population-level food availability and natal group size as covariates. These data were available for 124 individuals.
We related reproductive lifespan (number of years from the start until the end of the reproductive career) to age of first breeding (sensu Hawn et al. 2007) using parametric survival models (time to event models) with a Weibull distribution using the package Survival (2.36-5, Therneau 2011). In this analysis, we included all dominant individuals (irrespective of age). Sex was included as a covariate and individuals that were still alive in 2011 (n =228 of 942 individuals, 24%) were censored in the analyses.
Annual survival decreased sharply and progressively with age in individuals older than 6 years (Table 1, Fig. 1). Males and females showed similar declines in survival with age (Table 1). The annual survival probabilities predicted by the statistical model declined by 46% from age seven (0.85) to age 17 (0.46) (Fig. 1). Age of first breeding ranged from 1 to 8 years, with 48% (129/270) of individuals starting breeding as 1-year-olds, 35% (95/270) starting at 2 years and 17% (46/270) after 2 years of age. The age-dependent decline in survival was delayed (right-shifted) in individuals that started breeding at a later age (Table 1, Fig. 1). The coefficients of the model were almost identical, but opposite, for age and age of first breeding, suggesting that starting independent breeding 1 year later also delays the age-dependent decline in late-life survival probability for 1 year. At 7 years of age, the average survival probabilities for individuals that started breeding at 1 year of age, 2 years of age or when they were older than 2 years of age, were 0.83, 0.86 and 0.92, respectively. The trajectory of the age-dependent decline in survival (rate of survival senescence) did not differ for individuals that started breeding at different ages (no interaction between age and age of first breeding: Table 1, Fig. 1).
Table 1. The relationship between age-dependent annual survival in Seychelles warblers and age at first breeding and early-life conditions. The analyses included 143 females and 127 males older than 6 years. One hundred and 29 individuals started breeding at 1 year of age, 95 individuals at 2 years of age and 46 after 2 years of age
Age of first breeding
Early-life group size
Early-life territory quality
Early-life year quality
Age × Age first breeding
Age × Sex
Age × Early-life group size
Age × Early-life territory quality
Age × Early-life year quality
Summaries were derived from binomial response, linear mixed models with survival to the next year as dependent variable (see Methods for details). Variables included in the final models are indicated in bold.
There was considerable variation between individuals in the conditions encountered during the first 3 years of their life (territory-level food availability (mean ± SD=0.08 ± 0.46, range=−0.73–1.66), population-level food availability (mean ± SD=78.84 ± 10.16, range=54.5–115.0) and group size (mean ± SD= 2.78 ± 0.63, range=2.00–5.33). The onset of survival senescence was not associated with early-life territory quality, early-life population-level food availability and early-life group size (Table 1). Also, the rate of the age-dependent decline in survival was not related to any of the three measures of early-life condition, as indicated by the absence of significant interaction effects with age (Table 1).
Age of first breeding increased with natal territory quality (mean ± SE=0.14 ± 0.05, t =2.82, P <0.01), tended to increase with natal population-level food availability (mean ± SE=0.004 ± 0.002, t =1.93, P =0.06), but was not related to natal group size (mean ± SE= −0.05 ± 0.05, t =−1.18, P =0.24). Age of first breeding did not differ between males and females (mean ± SE=0.03 ± 0.07, t =0.34, P =0.73), but differed between cohorts (=13.45, P <0.01).
When considering dominant individuals of all ages, reproductive lifespan was similar for individuals with different ages of first reproduction (β = 0.01, SE=0.03, z =0.11, P =0.78). Males and females showed similar lifespans (β = −0.08, SE=0.05, z =−1.58, P =0.11). The model-predicted reproductive lifespans for individuals that first bred at 1, 2 or after 2 years of age were 6.08, 6.12 and 6.25 years, respectively.
Late-life annual survival probabilities declined progressively with age in Seychelles warblers, thus providing clear evidence for survival senescence in this natural population. That similar declines in survival were observed in males and females is consistent with predictions about sex-specific senescence. Evolutionary hypotheses of senescence predict higher rates of survival senescence in the sex with higher annual mortality (Williams, 1957; Charlesworth, 1994; Ricklefs, 1998), and similar rates of senescence when sexes have similar annual mortality and adopt similar strategies to maximize fitness (Promislow, 2003; Clutton-Brock & Isvaran, 2007). The latter is the case in the Seychelles warbler, where individuals have similar annual mortality (Brouwer et al., 2006), form long-term pair-bonds, defend stable territories over many years, and both sexes invest extensively in most aspects of parental care (Komdeur, 1991, 1994b).
The disposable soma hypothesis predicts that individuals that start breeding at a later age should show reduced survival senescence because lower reproductive investment during early life has allowed them to allocate more resources to somatic maintenance and long-term survival (Kirkwood & Rose, 1991). That a higher age of first breeding was associated with delayed senescence in the Seychelles warbler supports this hypothesis. For example, a 13-year-old individual that became a breeder when it was 5 years old, and therefore invested less in reproduction and territorial defence during early life, had a similar survival probability to a nine-year-old individual that started reproducing when it was 1 year of age.
Although the disposable soma and antagonistic pleiotropy hypotheses predict similar trade-offs between early-life reproductive effort and senescence, these trade-offs are caused by different mechanisms. In this study, disposable soma is the most likely mechanism to explain the delay in senescence for individuals that first bred independently at a later age because in the Seychelles warbler a delayed onset of independent breeding, and the associated reduction in resources allocated to reproduction during early-life, is most probably due to the limited availability of high-quality independent breeding opportunities, rather than a genetic mechanism (Komdeur, 1992). To confirm this, it would be informative to explore the physiological mechanisms that are affected by early-life reproductive effort and predict mortality. One mechanism that may contribute to this pattern might be the build-up of senescent cells within organs as a result of telomere shortening because of oxidative stress; previous work has shown that telomere dynamics predict mortality in the Seychelles warbler (Barrett et al., 2013). Future studies might profitably investigate how age of first breeding, early-life reproductive investment, early-life conditions and survival senescence are related to telomere dynamics. In contrast, the antagonistic pleiotropy hypothesis could be tested formally by looking at genetic correlations between early-life reproductive effort and late-life survival. A negative genetic correlation between early-life reproductive effort and late-life survival would support the antagonistic pleiotropy hypothesis. Alternatively, antagonistic pleiotropy might be investigated by identifying specific loci that have opposite effects on early- and late-life fitness by mapping quantitative trait loci.
Several other studies in wild populations have also shown that increased survival senescence was associated with investment in early reproduction, or an earlier age of first breeding, (McCleery et al., 1996; Orell & Belda, 2002; Reid et al., 2003; Reed et al., 2008). However, these studies were conducted in open populations, in which individuals could move freely in and out of the study area, and so emigration and death could not be distinguished; therefore, they were measuring apparent, rather than actual, survival. In our study, survival estimates reflect true survival as individuals remain in the study population until they die (Komdeur et al., 2004). In addition, because adult predation is absent on Seychelles warblers, it is likely that most individuals succumbed to intrinsic, senescence-related mortality, rather than dying from extrinsic causes unrelated to senescence. Comparative studies across species also found longer lifespans or a later onset of survival senescence in species with a higher age of first breeding (Blumstein & Møller, 2008; Péron et al., 2010). However, other studies found no relationship between survival senescence and early-life reproductive effort (Gaillard et al., 2003), or that lower late-life survival was associated with higher age of first reproduction (Nussey et al., 2007; Aubry et al., 2011). Potential explanations for these contrasting results include the possibilities that individuals may have started breeding at a higher age because they experienced harsh environmental conditions during early life (e.g. Nussey et al., 2007), or that they were of inferior phenotypic quality (e.g. Aubry et al., 2011). There is some evidence that in the Seychelles warbler a later onset of independent breeding is related to individual quality; Richardson et al. (2007) found that individuals that had been a subordinate previously (and thus had a later onset of independent breeding) had a higher chance to be deposed from their dominant breeding position. Alternatively, investment in early-life reproduction might be state-dependent, such that individuals in better physiological condition are able to produce more offspring, at the same cost, as individuals in worse physiological condition that produce fewer offspring (McNamara & Houston, 1996).
Intriguingly, the reproductive lifespans of individuals started breeding at different ages were similar. Apart from reducing the costs of reproduction and territory defence, and thereby delaying survival senescence, a later onset of independent breeding in the Seychelles warbler has other fitness benefits. For example, subordinates have been shown to benefit through the accumulation of breeding experience, which is subsequently associated with increased reproductive output during the individual's first independent breeding attempt (Komdeur, 1996d). In addition, male subordinates may increase their lifetime reproductive success by budding off part of their resident territory (Komdeur & Edelaar, 2001). Finally, female subordinates may gain indirect (kin-selected) benefits from helping (Richardson et al., 2003), or direct benefits (maternity) by laying eggs within the dominant's nest (Richardson et al., 2002). Therefore, because reproductive lifespans were similar for individuals with a different age of first breeding, other sources of benefit may lead to higher fitness of individuals with a later onset of independent breeding in this species. A previous study suggested that this may be the case in the Seychelles warbler as 1-year-old Seychelles warblers may achieve greater lifetime reproductive success when they remain as a subordinate helper in a territory for several years than when they would acquire an independent breeding position immediately (Komdeur, 1992). However, to compare the fitness of individuals with different ages of first reproduction as a dominant territory holder, genetic data on lifetime reproductive success is required for each individual (including shared breeding before the onset of independent reproduction, and extra-pair offspring). Because this data is not yet available, we were not able to estimate lifetime reproductive success for the individuals included in this study. A genetic pedigree is currently under construction which may be used to investigate fitness of individuals with different ages of first reproduction as dominants.
The lack of a direct relationship between late-life survival and early-life conditions in our study is contrary to what we expected based on the disposable soma hypothesis. Other recent studies of early-life conditions and survival senescence in wild populations also failed to detect a correlation between these two factors (Descamps et al., 2008; Millon et al., 2011). One other study showed an indirect relationship between the rate of survival senescence and poor conditions during early life, which occurred through their mutual linkage with age of first breeding (Nussey et al., 2007). Individuals that experienced poor conditions during early life started breeding at a later age, and subsequently showed increased rates of survival senescence (Nussey et al., 2007). Although in our study, survival senescence was not directly related to early-life conditions, age of first breeding increased with natal food availability (see also Komdeur, 1992), and survival senescence was delayed in individuals that started independent breeding at a later age. Our results therefore also suggest an indirect association between survival senescence and early-life food conditions through age at first breeding. A possible explanation for the absence of a direct relationship between survival senescence and early-life conditions might be that Seychelles warblers adopt a conservative reproductive tactic, in which survival is favoured over reproduction. In long-lived species, a long remaining reproductive life is more valuable than high current reproductive investment that might result in the death of an individual. This is especially so in species that experience low adult extrinsic mortality, variable juvenile survival and prolonged periods of parental care, in which parents cannot predict the survival of their offspring (Hirshfield & Tinkle, 1975). Indeed, the life history of the relatively long-lived Seychelles warbler is characterized by low extrinsic mortality and considerable variation in juvenile survival across years (Brouwer et al., 2006). In addition, adults may not be able to predict adequately the conditions that promote offspring survival, as the optimal timing for nest initiation that will result in the highest juvenile survival is 2 months before the peak in food availability (Komdeur, 1996b). Consequently, individuals should reduce their reproductive effort under harsh conditions in order to maintain their survival (e.g. Hamel et al., 2010; Martin & Festa-Bianchet, 2010). Such a scenario might mask a direct relationship between late-life survival and early-life conditions. Conversely, it is possible that periods of lower food availability (dietary restriction) decelerate senescence, which is well established in laboratory organisms (e.g. Weindruch & Walford, 1982; Nakagawa et al., 2012), but has not yet been investigated in wild vertebrate populations.
Nature Seychelles kindly allowed us to work on Cousin Island Nature Reserve, and we thank them and the Cousin Island staff for providing accommodation and facilities during our visits. The Department of Environment and the Seychelles Bureau of Standards gave permission for fieldwork and sampling. We thank Karl Phillips, Joost van den Heuvel and an anonymous reviewer for helpful comments on the manuscript. This study would not have been possible without the help of countless fieldworkers during the whole study period, supported by grants from the UK Natural Environment Research Council and the Netherlands Organisation for Scientific Research. MH was supported by a grant to JK, DSR and TB from the Netherlands Organisation for Scientific Research (NWO-ALW1PJ/07080).