1. Understanding the degree to which reproductive success varies with an individual’s age and lifespan, and the degree to which population-level variation mirrors individual-level variation, is central to understanding life-history evolution and the dynamics of age-structured populations. We quantified variation in the survival probability of offspring, one key component of reproductive success and fitness, in relation to parent age and lifespan in a wild population of red-billed choughs (Pyrrhocorax pyrrhocorax).
2. On average across the study population, the first-year survival probability of offspring decreased with increasing parent age and lifespan; offspring of old parents were less likely to survive than offspring of young parents, and offspring of long-lived parents were less likely to survive than offspring of short-lived parents.
3. However, survival did not vary with parent age across offspring produced by groups of parents that ultimately had similar lifespans.
4. Rather, across offspring produced by young parents, offspring survival decreased with increasing parent lifespan; parents that ultimately had long lifespans produced offspring that survived poorly, even when these parents were breeding at young ages.
5. The average decrease in offspring survival with increasing parent age observed across the population therefore reflected the gradual disappearance of short-lived parents that produced offspring that survived well, not age-specific variation in offspring survival within individual parents.
6. The negative correlation between offspring survival and maternal lifespan was strongest when environmental conditions meant that offspring survival was low across the population.
7. These data suggest an environment-dependent trade-off between parent and offspring survival, show consistent individual variation in the resolution of this trade-off that is set early in a parent’s life, and demonstrate that such structured life-history variation can generate spurious evidence of senescence in key fitness components when measured across a population.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
However, given any positive or negative covariance between offspring quantity and quality, simply measuring the number of offspring reared to independence could provide a misleading picture of age-specific variation in a parent’s overall reproductive success and hence the magnitude of life-history variation and senescence (Kern et al. 2001; Priest et al. 2002; Descamps et al. 2008). In particular, offspring survival, one major component of offspring quality, can vary markedly in time and space and cause substantial variation in individual fitness and population growth rate (Gaillard, Festa-Bianchet & Yoccoz 1998; Reid et al. 2004; Ozgul et al. 2006). Evaluating offspring survival should therefore be integral to any study of life-history evolution or population dynamics. Since the conditions an individual experiences early in life can influence its subsequent life history (Lindström 1999; Kerr et al. 2007; Hamel et al. 2009), early environmental variation associated with parent age could potentially exert long-term effects that influence offspring survival after those offspring become independent from their parents and natal location (Hercus & Hoffmann 2000). Indeed, variation in offspring viability, maturation, survival or lifespan in relation to parent age has been documented in laboratory and domestic populations (e.g. Kern et al. 2001; Priest et al. 2002; Fox, Bush & Wallin 2003; Fuerst-Waltl et al. 2004; Benton et al. 2008). However, despite this clear expectation and the central role of offspring survival in driving life-history evolution and population dynamics, there are few rigorous studies relating offspring survival to parent age in wild populations (see Discussion).
The paucity of such studies reflects their inherent difficulty (Nussey et al. 2008). Offspring frequently disperse away from study areas, impeding accurate estimation of their subsequent survival. Even without dispersal, analyses must account for variation in the probability that a surviving offspring will be encountered. Long-term studies will be required to measure offspring survival across sufficient old parents to detect any senescence. This problem of small sample sizes will be exacerbated if old parents rear few offspring, further reducing the number of offspring whose subsequent survival can be measured. Finally, but critically, any such analysis must distinguish whether average variation in offspring survival with parent age observed across a population reflects age-specific variation within individual parents or changing population composition (Vaupel & Yashin 1985; Forslund & Pärt 1995; Cam et al. 2002; Reid et al. 2003a; Nussey et al. 2008). For example, average offspring survival could increase with increasing parent age if parents that can survive to old age also consistently produce offspring that survive well. Alternatively, average offspring survival could decrease with increasing parent age even in the absence of senescence if, given a trade-off between survival and reproduction, only parents that consistently invest less in offspring can survive to old age.
Statistical methods that aim to quantify age-specific variation in life history while controlling for such structured variation with lifespan have been developed, thereby allowing longitudinal and cross-sectional variation to be distinguished (Cam et al. 2002; van de Pol & Verhulst 2006; Nussey et al. 2008). However, these methods cannot necessarily be applied when offspring survival is the trait of interest, especially when the probability of encountering a surviving offspring is <1. Longitudinal data describing age-specific variation in offspring survival within individual parents will inevitably be sparser than data describing the number of offspring reared, as failed breeding attempts are no longer informative. Furthermore, the underlying offspring survival probability (or ‘frailty’), which is the trait of interest, cannot be observed directly but must be estimated from survival rates observed across groups of individuals (Vaupel, Manton & Stallard 1979; Fox et al. 2006). Given these circumstances, one approach is to quantify variation in offspring survival with parent age within groups of parents with similar lifespans, thereby minimizing the extent to which parents with different life-history strategies can be differentially represented at different ages.
Such structured variation in reproductive success with respect to parent lifespan is often viewed as a ‘nuisance’ factor that impedes studies of variation in reproductive success with age (van de Pol & Verhulst 2006; McCleery et al. 2008). However, variation in reproductive success with lifespan is itself of major interest in the contexts of life-history evolution and population dynamics (Westendorp & Kirkwood 1998; Descamps et al. 2006; Weladji et al. 2006; Ricklefs & Cadena 2007; Tuljapurkar, Steiner & Orzack 2009). Positive covariance between reproductive success and lifespan would indicate a skewed distribution of individual fitness and raise questions concerning the ecological and evolutionary causes and consequences of such variation. Negative covariance could indicate a trade-off between reproduction and adult survival, raise questions concerning the maintenance of different allocation strategies within populations, and potentially support evolutionary theories of senescence (Kirkwood & Rose 1991; Westendorp & Kirkwood 1998; Descamps et al. 2006; Nussey et al. 2008). Such structured life-history variation and covariation within and among individuals can also influence the magnitude of demographic stochasticity and hence population dynamics (Vindenes et al. 2008). Furthermore, just as variation in reproductive success with lifespan could confound estimates of variation with age, variation in reproductive success with age could confound estimates of variation with lifespan per se. For example, if offspring survival were to decrease with increasing parent age, then average offspring survival would decrease with increasing parent lifespan even given zero direct effect. Comprehensive understanding of life-history variation in relation to age and lifespan therefore requires parallel analyses that quantify variation in reproductive success with age while controlling for lifespan, and with lifespan while controlling for age. Such analyses, however, are rarely undertaken, particularly with respect to offspring survival.
We used long-term data from red-billed choughs (Pyrrhocorax pyrrhocorax L.) to quantify variation in first-year survival of offspring, a life-history trait that accounts for substantial variation in population growth rate (Reid et al. 2004), in relation to parent age and lifespan. First, we tested whether average offspring survival measured across the population varied with parent age or lifespan. Second, we tested whether offspring survival varied with parent age or lifespan after controlling for lifespan or age respectively, and hence whether population-level variation mirrored life-history variation in individuals. Finally, we tested whether relationships between offspring survival and parent age and lifespan varied with environmental conditions. In doing so we demonstrate substantial, environment-dependent and individually consistent variation and covariation in life history, and illustrate that such covariation can create spurious evidence of senescence when life-history variation is averaged across a population.
Materials and methods
The island of Islay, Scotland, holds a resident population of c. 45–70 breeding pairs of choughs that has been the subject of an individual-based study since 1981 (Bignal, Bignal & McCracken 1997; Reid et al. 2003b). These birds, which comprise the majority of the Scottish population, breed once each year and nest in traditional sites in caves and buildings. Females incubate the eggs and both parents provision chicks. Provisioning can continue for several weeks post-fledging (Bignal et al. 1997).
Each year, a sample of nest sites was monitored and the number of offspring fledged was recorded. Mean breeding success was 2·0 ± 0·2 (SE) fledglings per attempt (range 0–5, Reid et al. 2003b, 2004). Shortly before fledging, offspring at monitored sites were marked with unique combinations of coloured plastic rings to allow individual identification. During 1983–2007, c. 25–65% of all broods fledged in each year were colour-ringed (totalling 1214 offspring from 443 successful breeding attempts), and their parents checked for colour-rings. Over time, therefore, a sample of colour-ringed offspring produced by colour-ringed (and hence known age) parents has accumulated. However in most years, and particularly early study years, most parents were unringed and of unknown age (Reid et al. 2003a). Analyses of variation in offspring survival with parent age and lifespan therefore focused on the subset of colour-ringed fledglings that had colour-ringed (known age) parents.
Substantial resighting effort has provided >16 000 observations of colour-ringed choughs of all ages across Islay since 1983. Capture–mark–recapture (CMR) approaches can therefore be used to estimate annual apparent survival (ϕ) and resighting (p) probabilities, even though p was generally <1 (Reid et al. 2003b, 2004). Resighting data also suggest that dispersal from Islay is rare; no Islay-ringed choughs have been observed to settle elsewhere in >20 years despite searches of adjacent islands and regular monitoring of an adjacent population on Colonsay (c. 8 km away). Variation in ϕ is therefore likely to primarily reflect variation in mortality (Reid et al. 2003b, 2004). Mean age at first breeding is 2·5 ± 0·1 and 2·9 ± 0·1 years in females and males, respectively, and varies relatively little among individuals (range 2–4 years, Reid et al. 2003b). Breeding adults are highly philopatric and typically occupy the same territory throughout their reproductive lives. As adults breeding at monitored nest sites were systematically checked for colour-rings each year, the lifespan (age at death) of colour-ringed parents of colour-ringed offspring can be calculated directly with high confidence (annual resighting probability of monitored parents ≥0·96, estimated from multi-occasion CMR analysis). This system therefore provides a valuable opportunity to quantify variation in offspring survival in relation to parent age and lifespan.
Basic CMR models
Previous analyses of annual encounter histories of all 1214 fledglings colour-ringed during 1983–2007 showed that survival differed between first-year (ϕ1, fledging to age 1), second-year (ϕ2, age 1–2) and adult (ϕad, all subsequent ages) age-classes, and that ϕ1 and P varied among years (Reid et al. 2008). Among-year variation in ϕ1 was correlated with variation in environmental conditions. Specifically, a model that included three weather variables and tipulid larvae abundance (an important prey of choughs) explained c. 79% of observed variation in ϕ1 (Reid et al. 2008). Variation in ϕ1 in turn accounted for c. 27% of variation in population growth rate (Reid et al. 2004).
Some fledglings produced by parents of known age and lifespan were colour-ringed every year from 1987 to 2007 (see Results). Given this subset of fledglings, data were too sparse to model full among-year variation in ϕ1 or resighting probability (p), or therefore to model year by parent age or lifespan interactions. However, analyses needed to control for known among-year variation to ensure that relationships between ϕ1 and parent age and lifespan were not obscured or spuriously created. Each year was therefore categorized as high (>0·51), medium (0·29–0·42) or low (<0·21) with respect to population-wide ϕ1, and high (>0·87), medium (0·68–0·79) or low (<0·58) with respect to p based on analyses of all 1214 colour-ringed fledglings. These categories matched natural divisions in the data, and captured 84% and 90% of the total among-year variation in ϕ1 and p respectively (see Supporting Information). Variation in the first-year survival probability of offspring in relation to parent age and lifespan was then modelled within each of the three survival categories, while controlling for three categories of p. This structure allowed direct test of whether relationships between ϕ1 and parent age and lifespan varied among the three survival categories (i.e. between high, medium and low survival years), and hence for environmental dependence in life-history covariation with respect to the quality of an offspring’s natal year.
Separate CMR models were constructed for fledglings with fathers or mothers of known age or known lifespan (giving four sets of models). In each set, basic models included constant ϕ1, ϕ2 and ϕad within the high, medium and low survival categories (giving nine survival parameters) and constant p within the high, medium and low resighting categories (giving three resighting parameters). These models adequately fitted the four data sets (bootstrap goodness-of-fit tests: p =0·18–0·32) and the variance inflation factor (ĉ), estimated as observed deviance divided by mean bootstrapped deviance and as median c (Cooch & White 2006), indicated very little overdispersion (ĉ =1·04–1·10, median c =1·05–1·09). There was little support for models with other age structures in ϕ or p, or that included linear trends in ϕ1 during 1987–2007 (ΔqAICc > 2·2, see also Reid et al. 2003b, 2004, 2008). Models that included the high, medium and low categories in ϕ1 and p were much better supported than models in which ϕ1 or p were fully year-dependent or constant across all years (ΔqAICc > 18·0).
Constrained CMR models
To quantify relationships between parent age or lifespan and the first-year survival probability of offspring, basic CMR models were constrained by including paternal or maternal age or lifespan as individual covariates for each colour-ringed fledgling. Constrained models considered logit-linear, quadratic and cubic effects of parent age or lifespan on ϕ1 (although cubic models were not supported and are not reported). Initial models tested whether intercepts differed between high, medium and low survival categories, holding slopes constant. Slopes were then allowed to differ between survival categories, thereby testing for variation in the relationship between parent age or lifespan and ϕ1 given different environmental conditions.
While logit-linear models provide a valuable means of testing hypotheses relating ϕ1 to continuous variation in parent age and lifespan, such models inevitably constrain the shape of estimated relationships and may be most strongly influenced by parts of the distribution with most data (e.g. younger parent ages). Therefore, to further validate patterns of variation in ϕ1 estimated from logit-linear model, we ran additional ‘threshold’ CMR models where ϕ1 varied as step functions of parent age or lifespan (see Supporting Information).
Akaike’s Information Criterion, adjusted for small sample sizes and the degree of overdispersion estimated by ĉ (qAICc), was used to identify the best-supported model in each candidate set, with models considered better supported than rivals if the difference in qAICc (ΔqAICc) ≥2·0 (Burnham & Anderson 1998). As goodness-of-fit cannot be directly estimated for CMR models with individual covariates, values of ĉ estimated for basic models were used to adjust AIC for constrained models (thereby assuming that constrained models fitted the data as well as basic models, Cooch & White 2006). Back-transformed parameter estimates extracted from the best-supported models were used to reconstitute relationships between parent age and lifespan and ϕ1. In many cases, there was support for multiple candidate models. To incorporate uncertainty in model selection into parameter estimates and associated confidence intervals, standard model averaging procedures, where model coefficients estimated from competing models are weighted by each model’s AIC weight (AICw), were applied across all models where AICw≥0·01 (Burnham & Anderson 1998; Cooch & White 2006).
Final lifespans were unknown for colour-ringed parents that were still alive in 2008. To maximize sample sizes, individuals aged ≥13 years in 2008 were attributed their lifespan in 2008. These lifespans are therefore minimum estimates, but the degree of underestimation is likely to be small as choughs rarely survive beyond 14 years and these parents can already be considered ‘long-lived’ (see below). Parents that were still alive aged <13 years in 2008 were not attributed a lifespan. Sample sizes for models including parent lifespan are therefore smaller than for those including parent age.
Descriptive CMR models
Individual survival probabilities cannot be observed directly, but must be estimated across sufficiently large groups of individuals. Therefore, to help visualize patterns of variation in ϕ1 with parent age and lifespan, further CMR models were run to estimate ϕ1 for fledglings produced by parents of discrete age categories (2–3, 4–5, 6–7, 8–10 and ≥11 years) and lifespan categories (3–6, 7–8, 9–10, 11–13 and ≥14 years). These models were used for initial descriptive purposes rather than for hypothesis testing. Categories were defined to explore maximally variation in ϕ1 with parent age and lifespan while maintaining sufficient sample sizes of fledglings within each category to provide robust estimates.
Age vs. lifespan
To tease apart whether average variation in ϕ1 with parent age or lifespan observed across the population in fact reflected variation with parent lifespan or age, respectively, we quantified the degree to which ϕ1 varied with one variable while holding the other constant. As only long-lived parents can reach old ages, parent age and lifespan were inevitably correlated across the data set (r =0·56 and 0·60 for fathers and mothers respectively). We therefore did not attempt to distinguish their effects by including both within the same model (e.g. van de Pol & Verhulst 2006; McCleery et al. 2008). Instead, we first tested whether ϕ1 varied with parent age across groups of fledglings produced by parents with medium (8–12 years) or long (≥13 years) lifespans. For comparison, we also estimated mean ϕ1 across fledglings produced by parents with short lifespans (≤7 years), but did not estimate variation with parent age within this category due to the reduced variance. Secondly, we tested whether ϕ1 varied with parent lifespan across fledglings produced by parents aged 4–7 years. These restrictions were chosen to minimize variation in parent lifespan and age, respectively, while retaining sufficient sample sizes of fledglings and variation in parent age and lifespan respectively.
Analyses were run in program mark (White & Burnham 1999). The proportion of deviance explained by parent age and lifespan, and random effects of individual parents, were not estimable because fully parent-specific (or other individual-based) CMR models cannot be fitted. Effects of fledgling sex were not considered in detail because sexes were generally unknown prior to recruitment. However, sexes of recruited offspring did not vary with parent age or lifespan (p >0·2). Figures are curtailed at appropriate values of parent age and lifespan to avoid extrapolating relationships beyond ranges directly informed by the data. Results remained qualitatively similar when the few fledglings produced by parents aged ≥14 years were pooled into a single category.
Variation in parent age
Totals of 363 and 383 fledglings colour-ringed during 1987–2007 had known-age mothers and fathers respectively. Figure 1(a) shows the number of monitored fledglings attributed to parents of each age. Maternal and paternal ages tended to increase across fledglings colour-ringed during 1987–2007 (r =0·14 and 0·34 respectively). However, these correlations primarily reflected the absence of known old parents during early study years, and maternal and paternal ages did not increase markedly across fledglings colour-ringed during 1990–2007 (r = −0·02 and 0·12 respectively). The maximum observed maternal and paternal ages of 17 and 18 years occurred in 2002, while a maximum of 24 years could have been observed by 2007. Furthermore, during 1987–2007, observed maternal and paternal ages did not vary systematically with mean ϕ1 estimated across the whole population in a fledgling’s natal year (r = −0·06 for both sexes). Maternal and paternal ages were moderately positively correlated across 186 fledglings that had both parents of known age (r =0·26).
Offspring survival and parent age
Descriptive analyses suggested that on average across the population, fledglings produced by old parents, and to some degree by young parents, were less likely to survive to age 1 than fledglings produced by middle-aged parents (Fig. 2a).
Constrained models that included linear or quadratic effects of parent age on ϕ1 and interactions with survival category were slightly less well supported than models that included survival category but not parent age (Table 1). However, additive models that included linear, and to some degree quadratic, effects of parent age but no interaction with survival category were well supported (Table 1). These models show that on average, ϕ1 decreased with increasing maternal and paternal age irrespective of whether ϕ1 was high, medium or low across the population (Fig. 3a,b). Effects were substantial; fledglings produced by the oldest parents were on average c. 30–50% less likely to survive than fledglings produced by the youngest parents (Fig. 3a,b). Threshold models supported these conclusions (see Supporting Information).
Table 1. Models relating apparent first-year survival probabilities (ϕ1) of offspring to (a) paternal age, (b) maternal age, (c) paternal lifespan and (d) maternal lifespan across all monitored fledglings. SCat, PAge, MAge, PLife and MLife denote effects of survival category and paternal and maternal age and lifespan on ϕ1, and * and + denote interactive and additive effects respectively. Main effects were included for all interactions. All models included constant ϕ2 and ϕad within each survival category and constant P within each resighting category (p varied from 0·44 to 0·56, 0·79 to 0·81 and 0·96 to 0·97 for the low, medium and high categories respectively)
Model for ϕ1
qAICc, corrected quasi-likelihood Akaike’s Information Criterion.
(a) Paternal age
(b) Maternal age
(c) Paternal lifespan
(d) Maternal lifespan
Variation in parent lifespan
Totals of 342 and 378 fledglings colour-ringed during 1987–2007 had mothers and fathers with estimable lifespans. Figure 1(b) shows the number of monitored fledglings attributed to parents of each lifespan. As parents of any lifespan could be observed in all study years, maternal and paternal lifespans did not increase or decrease across fledglings colour-ringed during 1987–2007 (r = −0·08 and 0·08 respectively) and did not vary systematically with mean ϕ1 estimated across the whole population in a fledgling’s natal year (r = −0·06 and −0·08 respectively). Maternal and paternal lifespans were moderately correlated across 183 fledglings that had both parents of known lifespan (r =0·30).
Offspring survival and parent lifespan
Descriptive analyses suggested that on average across the population, fledglings produced by long-lived parents were less likely to survive to age 1 than fledglings produced by parents with shorter lifespans (Fig. 2b).
Constrained models that included effects of parent lifespan and additive or interactive effects of survival category were better supported than basic models that included survival category but not parent lifespan (Table 1). The best-supported model for paternal lifespan included a quadratic effect of lifespan and interaction with survival category (Table 1). Two models for maternal lifespan were approximately equally well supported: an additive linear effect and a quadratic effect with survival category interaction (Table 1). These models show that in years when survival was high or medium across the population, average ϕ1 decreased with increasing maternal and paternal lifespan (Fig. 3c,d). However, in years when survival was low across the population, fledglings produced by parents with medium lifespans were on average much more likely to survive to age 1 than fledglings produced by parents with long or very short lifespans (Fig. 3c,d). Estimated effects were again substantial (Fig. 3c,d), and threshold models supported these conclusions (see Supporting Information).
Offspring survival and parent age controlling for lifespan
Models in which ϕ1 varied with paternal or maternal age within the medium-lived (8–12 years) and long-lived (≥13 years) groups of parents were not strongly supported (Table 2). Model-averaged trends were weak, but suggested that ϕ1 tended to increase with increasing paternal age and decrease with increasing maternal age within groups of parents with similar lifespans (Fig. 4a,b). These models also showed that long-lived parents produced fledglings that were less likely to survive than fledglings produced by parents of the same age but with medium lifespans (Fig. 4a,b). Short-lived parents (≤7 years) produced fledglings that were more likely to survive than those produced by medium-lived or long-lived parents of the same age (Fig. 4a,b). There was little support for models that included interactive rather than additive effects of parent age and survival category on ϕ1 within the medium-lived or long-lived categories of parents (ΔqAICc > 2·1). Similarly, there was no support for threshold models of variation in ϕ1 with parent age (see Supporting Information).
Table 2. Models relating apparent first-year survival probabilities (ϕ1) of offspring to (a) paternal age and (b) maternal age across offspring of parents with medium (8–12 years) and long (≥13 years) lifespans, and to (c) paternal lifespan and (d) maternal lifespan across offspring of parents aged 4–7 years. SCat, LCat, PAge, MAge, PLife and MLife denote effects of survival category, lifespan category and paternal and maternal age and lifespan on ϕ1, and * and + denote interactive and additive effects respectively. Main effects were included for all interactions. All models included constant ϕ2 and ϕad within each survival category and constant p within each resighting category. ϕ1 was estimated as zero for offspring of long-lived parents in low survival years
Model for ϕ1
qAICc, corrected quasi-likelihood Akaike’s Information Criterion.
(a) Paternal age controlling for lifespan
(b) Maternal age controlling for lifespan
(c) Paternal lifespan controlling for age
(d) Maternal lifespan controlling for age
Offspring survival and parent lifespan controlling for age
Totals of 231 and 172 fledglings had fathers and mothers aged 4–7 years. Across these fledglings, models in which ϕ1 varied with parent lifespan were better supported than basic models that only included survival category (Table 2). The best-supported paternal models included additive linear and quadratic effects of paternal lifespan, and the best-supported maternal model included a linear effect of maternal lifespan and interaction with survival category (Table 2). These models show that when breeding aged 4–7, parents that ultimately had long lifespans produced fledglings that were less likely to survive than fledglings produced by similarly aged parents that ultimately had short lifespans (Fig. 4c,d). The decrease in ϕ1 with increasing maternal lifespan was steepest in years when ϕ1 was low across the population; long-lived mothers rarely produced offspring that survived through such years (Fig. 4c,d). Effects were substantial: fledglings produced by young parents that ultimately had long lifespans were c. 30–80% less likely to survive than fledglings produced by young parents that ultimately had short lifespans (Fig. 4c,d). Threshold models supported these conclusions (see Supporting Information).
Parent age, lifespan and offspring survival
On average across the population, the apparent first-year survival probability of fledgling choughs varied markedly with maternal and paternal age; fledglings produced by old parents were on average much less likely to survive than fledglings produced by younger parents. This pattern was consistent across years when environmental conditions were such that first-year survival was relatively high, medium or low across the population. It also broadly mirrors the average decrease in breeding success, measured as the number of offspring reared, that is often observed in old parents in wild populations (e.g. Reid et al. 2003a; Bowen et al. 2006; Nussey et al. 2006; Low et al. 2007; Keller et al. 2008; McCleery et al. 2008; Sharp & Clutton-Brock 2010), further reducing the average contribution of old age-classes to population growth rate.
Average fledgling survival also varied with maternal and paternal lifespan; parents that ultimately lived for many years on average produced fledglings that were less likely to survive. This pattern varied among years, however. In years when first-year survival was low across the population, only fledglings produced by parents with medium lifespans had any substantial probability of surviving to age 1.
Such variation in offspring survival with parent lifespan could create spurious patterns of variation with parent age that reflect the changing representation of parents with different life histories rather than age-specific variation within individuals (Cam et al. 2002; van de Pol & Verhulst 2006; Nussey et al. 2008). Indeed, within parents with medium (8–12 years) or long (≥13 years) lifespans, fledgling survival did not vary substantively with maternal or paternal age (and if anything tended to increase rather than decrease with increasing paternal age). In contrast, among fledglings produced by similarly aged parents (4–7 years), survival decreased markedly with increasing maternal and paternal lifespan. The substantial decreases in average fledgling survival with increasing parent age observed across the population therefore arose because long-lived parents (that reached old age) consistently produced fledglings that survived poorly while short-lived parents (that did not reach old age) produced fledglings that survived relatively well. There was little evidence of senescence with respect to offspring survival within groups of parents with similar lifespans.
We could not test for additive or interactive effects of maternal and paternal lifespan on offspring survival because too few offspring had both parents of known lifespan. However, across the available sample, maternal lifespan was only moderately correlated with paternal lifespan. This suggests that offspring survival may vary with the lifespan of both parents independently, rather than observed relationships with one parent’s lifespan entirely reflecting correlated variation with the lifespan of the other.
Structured variation in life history
van de Pol & Verhulst (2006) postulated that life-history variation might be structured such that long-lived and short-lived individuals have consistently low and high reproductive success respectively (their fig. 1E), but provided no empirical examples. Our data provide a fascinating example of this form of structured variation with respect to offspring survival.
Such negative correlations between parent lifespan and offspring survival are consistent with the existence of a trade-off between investment in self vs. offspring. It is particularly notable that long-lived parents already produced fledglings that survived poorly even when breeding at young ages. Furthermore, previous analyses showed that long-lived female choughs also lay smaller clutches and produce fewer fledglings when breeding at young ages than females with shorter lifespans (see Reid et al. 2003a). The fitness cost of poor survival of fledglings produced by long-lived females is not, therefore, mitigated by an increase in the number of fledglings produced. These data therefore suggest that individual choughs have consistently different life-history strategies, and that this variation in strategy is already set early in an individual’s breeding life. Individuals therefore appear to occupy different positions on a ‘live fast die young’ vs. ‘live slow die old’ continuum of life-history strategies (as suggested across species; Gaillard et al. 1989; Sæther, Ringsby & Roskaft 1996), rendering it difficult to estimate an individual’s contribution to population growth rate based on any single life-history component.
The existence of such life-history variation within a population raises intriguing questions of why the resolution of the putative trade-off between self and offspring varies among individuals, whether different strategies confer equal fitness or represent constraint, and what the population dynamic consequences might be. The relative fitnesses of the different observed combinations of parent lifespan and offspring survival are difficult to equate. This is because a parent’s observed lifespan represents the stochastic realization of its unobserved ‘frailty’ rather than itself being the underlying life-history trait (Vaupel et al. 1979; Fox et al. 2006; Tuljapurkar et al. 2009). Furthermore, the relative fitness of life-history strategies of inherently different lengths will depend on the degree of environmental variation experienced. Individual parents inhabiting variable environments might be expected to reduce investment in individual breeding attempts to maximize the probability of surviving until a productive year (Benton & Grant 1996). Such trade-offs might be particularly severe in poor environments where resources are scarce (Ricklefs & Cadena 2007). Indeed, offspring survival was most strongly negatively correlated with maternal lifespan in years when survival was low across the population, which is itself associated with poor environmental conditions (specifically, cold and wet weather and low prey abundance, Reid et al. 2008). The fledglings that survived through such years were almost exclusively produced by mothers with short lifespans. Producing offspring that can survive through poor environmental conditions may therefore have a substantial cost in terms of reduced maternal lifespan.
On Islay, first-year survival varies among choughs fledged from different territories (Reid et al. 2006, 2008) and adults typically remain highly philopatric to single territories throughout their breeding lives. It is therefore possible that conditions experienced on different territories prompt different life-history allocations and cause the observed covariance between parent lifespan and offspring survival. In choughs, most offspring mortality occurs after offspring are independent from their parents and natal territory (>90%, Scottish Chough Study Group, unpublished data). The association between low offspring survival and long parent lifespan is therefore unlikely to reflect curtailed parental care due to consistent pre-independence mortality of offspring fledged on particular territories (for example, due to high local predation rates). It could, however, reflect an adaptive increase in parental care in response to lower local adult survival rates (as shown by Ghalambor & Martin 2001 in relation to predation risk); such associations between life history and spatial variation in ecology require further investigation. Such consistent structured variation in life history could also influence the magnitude of demographic stochasticity and hence population persistence, although specific modelling is required to predict the precise population dynamic consequences of the observed structure (Vindenes et al. 2008).
Senescence and offspring survival
Previous studies that related measures of offspring survival to parent age, mostly in laboratory or domestic populations, have provided mixed results. Sons produced by older Callosobruchus maculatus mothers lived longer than those produced by younger mothers, but there was no such variation in daughters’ lifespan or with paternal age (Fox et al. 2003). Older mothers but not fathers produced shorter-lived daughters in five of six Drosophila strains (Priest et al. 2002). Daughters of older Simmental cows had longer lifespans (Fuerst-Waltl et al. 2004) while in humans, adult daughters but not sons of older mothers had shorter lifespans (Homo sapiens, Gavrilov et al. 1997). Larval or neonatal survival was lower in offspring of older Drosophila serrata (Hercus & Hoffmann 2000), but higher in offspring of older female black rockfish (Sebastes melanops, Berkeley et al. 2004) and painted turtles (Chrysemys picta, Paitz et al. 2007). In wild populations, the number of offspring recruited per offspring weaned or fledged, one measure of local survival, decreased with increasing maternal age in American red squirrels (Tamiasciurus hudsonicus, Descamps et al. 2008) but not stitchbirds (Notiomystis cincta, Low et al. 2007), while offspring survival increased with maternal age in Weddell seals (Leptonychotes weddellii, Hadley, Rotella & Garrott 2007). However, the latter five studies report population-level averages which, as evidenced by current analyses, may not reflect age-specific variation occurring in individual parents. Combined with the lack of evidence that fledgling survival varies directly with parent age in choughs, any general patterns of age-specific variation in offspring survival are not yet evident.
Survival of offspring produced by old parents might be expected to be particularly susceptible to the evolutionary processes that are hypothesized to cause senescence. This is because the reduced breeding success often observed in old parents would further reduce selection against any late-acting deleterious mutations or pleiotropic genetic effects on the survival of any offspring produced. This effect could, given certain life-history structures, be balanced by an increased sensitivity of fitness to variation in offspring survival as compared to offspring quantity, and hence stronger selection for increased or invariant offspring survival. Increased or stable survival of offspring of older parents might then reflect increased investment in a smaller number of higher quality offspring, or reflect terminal investment (Clutton-Brock 1984), further emphasizing that integrated fitness measures that incorporate all major fitness components are required to provide definitive evidence of senescence (Partridge & Barton 1996; Fox et al. 2003; Nussey et al. 2008).
An observation that offspring survival is largely independent of an individual parent’s age might, at first glance, appear to mean that age-specific variation in an individual parent’s overall reproductive success can after all be accurately measured simply by counting the number of offspring reared (as in many studies of wild populations). However, multiplying an age-dependent quantity of offspring by an age-independent offspring survival rate ϕ will reduce the absolute magnitude of age-specific variation in a parent’s overall reproductive success by a factor of ϕ below that estimated solely from offspring quantity. A ‘null’ relationship between offspring survival and parent age therefore still has major implications for the evolution and population dynamic consequences of age-structured life histories.
We thank the Islay farmers and landowners who kindly allowed access to nest sites, everyone who contributed to data collection on Islay, the Royal Society, Natural Environment Research Council, Scottish Natural Heritage and the Royal Society for the Protection of Birds for funding, and Jean-Michel Gaillard for helpful comments.