Pat Monaghan, Division of Environmental and Evolutionary Biology, Institute of Biomedical and Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK. E-mail: P.Monaghan@bio.gla.ac.uk
1The consequences of environmental variability for life-history evolution are predicted to depend on the pattern of covariation amongst life-history traits. Using data from a 20-year study of individually marked red-billed choughs, we investigate the short- and long-term life-history consequences of population-wide variation in reproductive conditions, and demonstrate clear among-cohort variation and covariation in life-history parameters.
2The mean number of offspring fledging per breeding event varied among years, and was correlated with environmental conditions (temperature and rainfall) during the months preceding breeding. As the variance in breeding performance did not differ among years and choughs did not miss breeding seasons, variation in environmental conditions affected the whole breeding population. Thus the quality of the chough's breeding environment varied amongst years.
3Juvenile survival, the probability of recruitment to the breeding population and breeding longevity varied amongst cohorts, and these were positively correlated with the quality of the cohort's natal environment. Offspring fledging under good conditions were more likely to survive to breeding age and recruit, and had longer breeding lives than offspring fledging under poor conditions.
4Age at first breeding varied amongst cohorts, and increased with population size at maturity rather than natal conditions.
5The total number of offspring that recruits ultimately fledged varied primarily with breeding longevity rather than recruitment age. Thus, the consistent positive covariation amongst life-history traits meant that the total number of offspring fledged by recruits during their breeding life varied amongst cohorts, and was correlated with the quality of a cohort's natal conditions. Choughs fledging under good conditions themselves ultimately fledged more offspring.
6Such environmentally determined variation in offspring fitness is expected to influence optimal patterns of parental investment. We discuss the predictions that environmental variability should select for investment in adult survival and for reduced reproductive effort in poor years.
The premise that life-history traits covary rather than fluctuate independently is central to life-history theory. Traits are hypothesized to be linked by physiological trade-offs, where allocating limiting resources to one activity reduces the resources available for investing elsewhere. Investment levels are consequently expected to be negatively correlated (Roff 1992; Stearns 1992). However, as individuals vary in their ability to invest simultaneously in competing activities, positive correlations among phenotypic traits are frequently observed in analyses across individuals (van Noordwijk & de Jong 1986; McNamara & Houston 1996). Such variation in investment ability, and thus positive life-history covariation, can arise because individuals differ in genetic quality (Partridge & Harvey 1988), or as a function of their staggered progression through a stage-structured life-history (McNamara & Houston 1996). However, individual investment ability can also be greatly influenced by current environmental conditions (e.g. Gosler & Carruthers 1999; Moss & Croft 1999; Weimerskirch, Zimmermann & Prince 2001), and by conditions that individuals or their ancestors encountered previously (Mousseau & Fox 1998). Recent studies suggest that the environment experienced during early development can have particularly profound consequences for future phenotypes and investment ability (Lindström 1999; Metcalfe & Monaghan 2001; Lummaa & Clutton-Brock 2002). Individual variation in early nutrition can give rise to ‘silver-spoon’ or ‘lead-spoon’ effects, causing positive covariation among multiple future life-history traits (Grafen 1988; Madsen & Shine 2000). At the population level, evidence that widespread variation in the natal environment can influence the life-histories of entire cohorts of offspring is also now emerging (Lindström 1999). However, such population-wide synchronization of traits can itself have implications for the overall correlation structure of individual life-histories. For example, the performance of cohort members born under ‘silver spoon’ conditions may subsequently be constrained by density-dependent effects operating in the aftermath of a successful breeding season (e.g. Clutton-Brock et al. 1991; Rose, Clutton-Brock & Guinness 1998). Thus the extent to which large-scale variation in the natal environment can create consistent positive covariation amongst individual life-history traits is not immediately apparent. Further empirical information describing the life-history, fitness and population consequences of population-wide environmental variation is required (Benton & Grant 1996; Saether & Bakke 2000; Beckerman et al. 2002).
Investigating the extent and pattern of among-cohort variation in life-histories requires long-term study of marked individuals. Here, we present data from a 20-year study of individually marked red-billed choughs Pyrrhocorax pyrrhocorax Linnaeus in an isolated population on the Scottish island of Islay. We show that mean chough breeding success varied among years in relation to weather conditions, and thus that the quality of the breeding environment varied among years. We investigate the extent to which post-fledging life-history traits varied amongst cohorts, and demonstrate marked positive covariation between the quality of a cohort's natal environment and juvenile survival, recruitment and breeding longevity. We consequently suggest that the total number of offspring that a chough ultimately fledges varies with its own natal conditions, and discuss the implications of such pervasive cohort-level environmental effects for optimal patterns of parental resource allocation.
Materials and methods
Choughs have a very restricted distribution in Scotland, being almost entirely confined to the Inner Hebridean islands of Islay and Colonsay. Islay (55° N, 6° W) lies 25 km west of the Argyll coast. Its 45 breeding pairs of choughs (1998 census, Cook et al. 1999), down from a peak of approximately 78 pairs in 1986 (Monaghan et al. 1989), represents approximately three-quarters of the total Scottish population. The Scottish Chough Study Group (SCSG) has studied these birds continuously since 1981, as well as the smaller adjacent population on Colonsay (56° N, 6° W, 14 breeding pairs in 1998, D.C. Jardine pers. comm.). Although Colonsay is only 10 km from Islay, choughs move between the islands relatively rarely (6 exchanges of ringed birds in 20 years, SCSG data). Further, despite good observer coverage and the chough's conspicuous calls and aerial displays, choughs are very rarely observed on other Hebridean islands or on the Scottish mainland (Argyll Bird Club pers. comm.). Islay's chough population is thus relatively discrete, making it a valuable system for studying population-scale processes (Perrins, Lebreton & Hirons 1991).
Choughs on Islay breed once each year, nesting in cavity sites that have been accurately mapped (Bignal, Bignal & McCracken 1997). Each year from 1981 to 2000, 50–60% of occupied sites were visited towards the end of the chick-rearing period, and the number of offspring fledging recorded. Clutch size was recorded from 1981 to 1996. To allow cohort members to be followed and future life-history variation to be investigated, fledglings were individually marked using unique combinations of three colour rings. A total of 992 fledglings were colour-ringed from 1981 to 2000 (mean of approximately 50 per year, ranging from 22 in 1982 to 93 in 1985); thus roughly half the population has generally been individually identifiable in the field. Almost 9000 reliable resightings of these birds were documented during the study period.
To investigate whether post-fledging survival varied amongst cohorts, mark–recapture models were used to estimate cohort-specific survival probabilities (program MARK, White & Burnham 1999). The resighting period was defined as May 1st to July 1st each year, encompassing the period when fledglings were ringed and breeding adults resighted. Individuals were recorded as present (observed) or absent (not observed) within this period, with multiple resightings of an individual equating to single encounters. Choughs that were known to have emigrated from Islay were excluded from the analysis. Few choughs were ringed or resighted in 1982, so survival analyses were based on data from 1983 to 2000. As at least one resighting occasion subsequent to a survival period is necessary to calculate separate survival and resighting probabilities (Lebreton et al. 1992), parameters pertaining to the 1983–98 cohorts were estimable. Few birds with rings missing have been observed during the study (SCSG unpublished data); such ring loss as does occur generally involves old birds that are typically established breeders who can continue to be recognized by their remaining rings. In the few cases where ring loss meant that an individual's identity was uncertain, all possible encounter histories were excluded from the analysis. Given the discrete nature of the population, variation in apparent survival probability is likely to primarily reflect variation in mortality rather than dispersal.
A fully time-dependent Cormack-Jolly Seber model (CJS) was initially fitted (Lebreton et al. 1992). Bootstrap goodness-of-fit tests showed that this model fitted the data (P = 0·53). The variance inflation factor (ĉ) which quantifies data overdispersion was calculated as the observed model deviance divided by the mean bootstrapped deviance (method a), and as the mean observed ĉ (model deviance divided by the deviance degrees of freedom) divided by the mean bootstrapped ĉ (method b, Cooch & White 1998). For the general CJS model, ĉ was 1·04 (method a) and 1·09 (method b) indicating minimal overdispersion (a ĉ of 1·00 indicates perfect fit, Cooch & White 1998). This model was then constrained to investigate age, time and cohort-specific variation in resighting and survival probabilities. Akaike's Information Criterion (AIC) was used to identify the most parsimonious model from each candidate set. AIC values were adjusted to allow for the extent of overdispersion measured by ĉ (Cooch & White 1998). Recent literature increasingly advocates the use of AIC values as standard model selection procedure (Burnham & Anderson 1998; Anderson & Burnham 1999). Using information criteria to select amongst candidate models obviates problems associated with multiple testing in classical statistics, and allows comparison of non-nested models (Burnham & Anderson 1998). Likelihood ratio tests were additionally used to assess whether differences in fit between nested models were statistically significant. Further bootstrap goodness-of-fit tests were used to assess the fit of the most parsimonious model at each stage.
To investigate whether the proportion of fledglings that recruited to the breeding population, age of first breeding or breeding longevity varied amongst cohorts, the identities of breeding choughs were recorded across Islay during each breeding season. Sexes were distinguished by size and behaviour (Tella & Torre 1993; Cramp & Perrins 1994; Bignal et al. 1997). The proportion of fledglings recruiting and mean recruitment age were estimated for cohorts up to 1996, the last from which all potential breeders recruited by 2000. Mean breeding longevity was estimated for cohorts up to 1992. Although three old breeders remain from earlier cohorts (singles from 1985, 1987 and 1988) these individuals are towards the end of their potential lifespans (the 1985 individual is the oldest on record). The addition of further years to their longevity did not qualitatively alter the results, and if anything exacerbated observed trends. The few individuals whose first or last breeding year was uncertain were excluded from the analyses.
To investigate whether the total number of offspring that a recruit ultimately fledged (total offspring fledged, TOF) varied amongst cohorts, we summed TOF directly. However, as breeding success was not recorded at all sites in all years, complete reproductive histories were not available for all recruits. We used the available complete data to calculate the proportion of variation in TOF that was accounted for by variation in breeding longevity and mean breeding success, respectively, and thus considered the extent to which among-cohort variation in these traits might result in among-cohort variation in total productivity.
To investigate whether life-history parameters varied with environmental conditions, we investigated relationships with weather variables and population size. As the chough's British range is coincident with the relatively warm and dry O1H3T1 bioclimatic subregion, temperature and rainfall are likely to be key environmental variables affecting performance (Monaghan et al. 1989). Accordingly, daily rainfall and temperature data for Islay were obtained from the UK Meteorological Office British Atmospheric Data Centre. Mean rainfall and maximum temperature were calculated for early summer (April–June, covering the breeding season), late summer (July–September), early winter (October–December) and late winter (January–March) for each year. The number of breeding pairs of choughs on Islay was censused completely in 1982, 1986, 1992 and 1998 (Monaghan et al. 1989; Bignal et al. 1997; Cook et al. 1999). Partial census data (Bignal, Bignal & Easterbee 1988) and demographic models were used to estimate population size in intervening years. The eight weather variables and population size were not significantly correlated with each other (late winter temperature and rainfall P = 0·09; all other pairs P > 0·1), or with time across the study period (all P > 0·1). Univariate and stepwise linear regression models were used to test whether population size and weather during breeding, or weather during the three preceding seasons explained significant proportions of among-year variation in mean population breeding success. As survival estimates generated by mark–recapture models are not independent, relationships among survival probabilities and environmental variables cannot be investigated similarly (Lebreton et al. 1992). Instead, time-dependent survival probabilities were constrained as linear functions of environmental variables using logit link functions within survival models (White & Burnham 1999). AIC values and likelihood ratio tests were used to assess whether constraints improved model fit and thus whether survival probabilities covaried with environmental parameters. The proportion of variation in survival accounted for by each covariate was calculated as the deviance difference between models with and without the covariate, divided by the total associated deviance (Gaillard et al. 1997; White & Burnham 1999). General linear models were used to assess whether future life-history parameters varied with sex, cohort or environmental conditions. Statistical analyses were performed in SPSS (version 10·0). Minimum acceptable models are presented, followed by the partial probabilities associated with retained terms. For non-significant terms, the probability associated with their addition to this model is quoted. Data were log or arcsine transformed where necessary, and means are quoted ± 1SE.
From 1981 to 2000 the number of offspring fledging per chough breeding attempt averaged 2·0 ± 0·1 but varied amongst years, from 1·3 ± 0·3 in 1994 to 2·5 ± 0·3 in 1984 (1-way anova, F19,604 = 1·8, P = 0·04, CV = 0·16). However, the variance in breeding success did not differ among years (Levene's test F19,585 = 1·0, P = 0·45), and was not correlated with mean success across years (r18 = 0·24, N = 20, P = 0·31). Further, in breeding pairs where both adults were colour ringed and breeding success was recorded in both particularly good and particularly poor seasons, pairs performed relatively consistently across seasons (correlation between a pair's breeding success in good and poor seasons: r11 = 0·77, N = 13, P = 0·002). Thus mean population breeding success varied among years because performance varied across the whole population rather than because subsets of pairs performed well or poorly in different years. Mean population breeding success in one year was not correlated with mean success the following year (r18 = 0·14, N = 20, P = 0·57), or two or three years subsequently (r17 = 0·03, N = 19, P = 0·89 and r16 = −0·06, N = 18, P = 0·83, respectively), and autocorrelation analysis showed no significant periodicity in the data (P > 0·5 for all lags).
Across years, mean breeding success increased with late summer temperature and decreased with late winter rainfall during the seasons prior to breeding, both highly significant effects (Fig. 1). Mean breeding success did not vary with population size (P = 0·84) or with other weather variables (all P > 0·1). Thus the quality of a breeding season varied among years, in relation to pre-breeding weather conditions; the effects of this on the quality of the breeding environment are reflected in the population mean breeding performance. Consequently, to examine the extent of covariation among life-history traits and investigate whether the quality of the natal season had long-term life-history consequences for cohort members, we used mean population breeding success as a yearly index of natal conditions, and investigated whether among-cohort variation in future life-history traits was related to this index.
Resighting probability varied among years (P < 0·001, Table 1a). Whilst likelihood ratio tests suggested that resighting probability tended to differ between young and adult age classes (P = 0·06), there was greatest statistical support for retaining the age-independent variable (model 1, Table 1a).
Table 1. Mark–recapture models of year- and age-dependence in resighting probability (a and d), age-dependence in survival probability (b) and year-dependence in survival probability (c). The most parsimonious model in each section is indicated in bold. Definitions: ϕ = survival probability; p = resighting probability, a1, a2, a3 and ad indicate first, second and third year and adult age-classes; t and c indicate time (year) and cohort dependence in these age-classes, respectively. Superscripts identify the comparative models used in likelihood ratio tests
Likelihood ratio test
Age & year in resighting probability (p)
= 96·7, P < 0·0012
1st year & adult
= 24·9, P = 0·071
1st + 2nd year & adult
ϕ(t)p(a1 & 2t,adt)
= 26·4, P = 0·061
Age in survival probability (ϕ)
Single age class
1st year & adult
= 245·7, P < 0·0011
1st & 2nd year & adult
= 39·9, P < 0·0015
1st, 2nd & 3rd year & adult
= 23·3, P = 0·086
1st & 2nd year, adult & 12 +
ϕ(a1t,a2t,adt,12 + t)p(t)
= 6·6, P = 0·256
Year in survival probability (ϕ)
= 20·2, P = 0·096
= 16·6, P = 0·289
Adult & 2nd year constant
= 25·4, P = 0·059
= 51·3, P < 0·00111
Age in resighting probability (p)
Single age class
1st year & adult
= 23·1, P = 0·0611
1st + 2nd year & adult
= 14·2, P = 0·5112
Comparison of age-structured survival models showed that the probability of surviving from one breeding season to the next differed among first-year (fledging to age one), second-year (age one to age two) and adult (all older) birds (P < 0·001, Table 1b). There was no support for further subdivision, and thus the best supported age-structured model incorporated first-year, second-year and adult age classes in survival probability (model 6, Table 1b).
To investigate whether survival probabilities varied amongst cohorts, models including cohort-specific survival within each age class were compared with constant survival models. First-year survival varied markedly with year and therefore cohort (P < 0·001, Table 1c). Although likelihood ratio tests suggested that second-year survival also varied amongst cohorts (P = 0·05), AIC values did not support this conclusion (Table 1c), suggesting that the available data were insufficient to accurately quantify temporal variation in this parameter (Lebreton et al. 1992). Models that included constant adult survival were best supported (Table 1c), although too few choughs reached adulthood for some cohort-specific survival probabilities to be estimated, precluding any more detailed investigation of variation in adult survival (for example, using multistate mark–recapture models). Further modelling at this stage did not support the inclusion of age-specific resighting probabilities (Table 1d). Thus the most parsimonious model comprised cohort-specific first-year survival, constant second-year and adult survival and year-specific, but age-independent, resighting probability (model 11, Table 1). This model provided a good fit to the data (Bootstrap goodness-of-fit test, P = 0·33, ĉ = 1·04a and 1·08b), and was used to investigate whether survival probabilities varied with current environmental conditions, or with the quality of a cohort's natal year.
When the index of natal conditions and environmental variables describing the first year of life were introduced as covariates of first-year survival, there was support for the retention of natal conditions, and late summer and late winter rainfall explaining 49%, 54% and 49% of the variation, respectively, and 91% in total (Table 2a). Fledglings survived their first year well when born under good conditions ( = 27·9, P < 0·001), and when the summer and winter following fledging were dry ( = 30·7, P < 0·001 and = 27·8, P < 0·001, respectively). These covariates explained a similar proportion of the variance in first-year survival to the time-dependent model (= 5·1, P = 0·97). First-year survival did not vary with population size ( = 0·2, P = 0·66, Table 2a) or other weather variables (all P > 0·1).
Table 2. Mark–recapture models of relationships between environmental covariates and (a) first-year survival and (b) second-year survival. The most parsimonious model in each section is indicated in bold. Covariate codes: sr = late summer rainfall during the first year of life; wr = late winter rainfall during the first year of life; natal = index of conditions during a cohort's natal year; next = index of conditions during the breeding season following birth; popsize = estimated size of the breeding population. See Table 1 for key to other notation
1st year survival covariates
Natal summer rain
1st winter rain
2nd year survival covariates
1st summer conditions
Although the evidence that second-year survival varied among cohorts was initially ambiguous, there was support for models that constrained second-year survival as a function of conditions during both the natal and the subsequent breeding season, explaining 19% and 48% of variation, respectively, and 53% in total (Table 2b). Choughs survived their second year well when conditions were good during both their natal season ( = 4·3, P = 0·04) and the breeding season that coincided with the start of their second year of life ( = 13·4, P < 0·001). These covariates explained a similar proportion of the variance in second-year survival to the time-dependent model ( = 14·3, P = 0·43). The most parsimonious model provided a good fit to the data (Fig. 2). Thus the probability of survival through the first year of life improved with the quality of a cohort's natal conditions (Fig. 3).
Overall, 13·5% (116/856) of fledglings colour-ringed by 1996 recruited to the breeding population. The proportion of fledglings that recruited varied amongst cohorts, from 0·04 in 1989 (2/55) to 0·30 in 1983 (15/50, = 28·3, P = 0·02), and increased with the quality of the cohort's natal conditions (Fig. 4).
age at first breeding
Age at first breeding varied with cohort and sex (2-way anovaF16,105 = 2·8, P = 0·001, R2 = 0·22, partial probabilities: Cohort P = 0·009, Sex P = 0·001, Interaction P = 0·73). Mean recruitment age varied from 2·0 ± 0·1 for the 1993 and 1994 cohorts to 3·0 ± 0·2 in 1984 and 1988. Males recruited slightly younger than females (mean ages of 2·5 ± 0·1 and 2·9 ± 0·1, respectively). To relate among-cohort variation in recruitment age to environmental variables, the cohort factor in this model was replaced by the index of natal conditions and population size during the first possible breeding year (2 years after birth). Age at first breeding increased with population size during the first possible breeding year (GLM F2,105 = 9·9, P < 0·001, R2 = 0·15, partial probabilities: Sex P = 0·002, Population size P = 0·002) but not natal conditions (P = 0·68). Thus, among-cohort variation in recruitment age was related to population size at maturity rather than the quality of the natal environment.
On Islay, established breeders virtually never missed breeding seasons. The number of breeding events that recruits survived to make varied with cohort and sex (2-way anova on log transformed data, F14,81 = 2·4, P = 0·009, R2 = 0·20, partial probabilities: Cohort P = 0·04, Sex P = 0·04, Interaction P = 0·84). Mean breeding longevity varied from 1·0 attempt for the single 1982 recruit to 7·6 ± 1·6 for the 1984 cohort. Males survived for more breeding years than females (modes 7 and 4, medians 6 and 4, respectively). When the cohort factor was replaced with environmental covariates, breeding longevity varied with the quality of a cohort's natal conditions, but not with age at first breeding. Recruits born under good conditions had longer breeding lives (Fig. 5).
As chough breeding success varies markedly with age (SCSG unpublished data) and breeding longevity varied amongst cohorts, we could not simply use a recruit's mean breeding success to compare productivity amongst cohorts. Instead, we investigated whether productivity varied among cohorts within single age-classes of recruits. As most choughs recruited aged three and recruits from some cohorts made few further attempts, there were sufficient data to investigate among-cohort variation in productivity in 3-year-old breeders only. To maintain data independence in these analyses, indices of current breeding conditions were corrected by excluding the productivity of each focal bird. After accounting for breeding conditions during the season in which breeding was attempted, 3-year-old productivity increased with natal conditions in males (F2,28 = 4·3, P = 0·03, R2 = 0·19, partial probabilities: current conditions P = 0·04, natal conditions P = 0·03) but not in females (current conditions F1,31 = 5·6, P = 0·02, R2 = 0·17, natal conditions P = 0·35). Males born under good conditions appear to have higher annual productivity, at least when aged three.
total offspring fledged
Variation in breeding longevity explained 88% and 82% of the variation in the total number of offspring that recruits ultimately fledged for males and females, respectively. Across individuals for whom reproductive histories were complete, TOF varied among cohorts (anova on log-transformed data, F15,53 = 2·4, P = 0·01, R2 = 0·29, Sex P = 0·12) and was related to the quality of a cohort's natal conditions but not age at first breeding (Fig. 6). Recruits born under good conditions ultimately fledged more offspring.
On Islay, mean chough breeding success varied amongst years. This was because performance varied across the population as a whole rather than because subsets of pairs performed particularly well or poorly in different years, or did not breed at all. Mean success was related to weather conditions prevailing during the months preceding breeding: parents fledged more offspring when the previous late summer had been warm and the preceding late winter had been dry. Thus, the quality of the breeding environment varied amongst years. Temporal variation in vertebrate productivity has been widely linked to variation in food availability (e.g. Brinkhof & Cave 1997; Bergallo & Magnusson 1999; Oro, Pradel & Lebreton 1999; Weimerskirch et al. 2001), and breeding success is likely to have varied in choughs because weather affected foraging conditions. Choughs feed on surface-active and dung-associated invertebrates, and on soil invertebrates such as leatherjackets Tipula spp. L. (McCracken et al. 1992; McCracken & Foster 1993; Bignal et al. 1996). Warm summer temperatures can increase invertebrate abundance (Pritchard 1982; Gittings & Giller 1999). Heavy winter rainfall can reduce leatherjacket densities through flooding and interfere with beetle flight and dung colonization (Pritchard 1982), or may adversely affect the foraging behaviour of the birds themselves. Such effects may influence parental provisioning of chicks and thus breeding success directly, or influence success indirectly via intergenerational effects of prebreeding environments on parental condition and investment patterns (Mousseau & Fox 1998).
Post-fledging life-history traits varied among chough cohorts and were consistently correlated with the quality of a cohort's natal season, as indicated by the overall breeding performance of the population. Choughs fledging in years when breeding conditions were good were more likely to survive their first and second years of life, to recruit to the breeding population, and subsequently to make more breeding attempts. First-year survival may covary with natal conditions because environmental effects operating during the breeding season persist into the first year of life. However, as environmental conditions did not remain constant across consecutive years, future life-history traits could not have covaried with natal conditions because those conditions persisted throughout a cohort's lifetime. Further, as established breeders rarely missed breeding attempts and contributed consistently to fledgling cohorts, it is unlikely that life-history traits varied amongst cohorts because cohorts differed in genetic quality. There is increasing evidence that conditions experienced early during life can profoundly affect an individual's future life-history (Lindström 1999; Metcalfe & Monaghan 2001). Such ‘silver-spoon’ effects linking individual variation in early nutrition with variation in future performance have previously been demonstrated in birds (Haywood & Perrins 1992; Sedinger, Flint & Lindberg 1995; Merilä & Svensson 1997; Birkhead, Fletcher & Pellatt 1999). Our results suggest that large-scale variation in the natal environment can result in major and persistent differences in life-history parameters amongst entire cohorts of offspring.
Choughs showed consistent positive covariation amongst life-history traits, relating natal conditions with juvenile survival, recruitment and breeding longevity. Thus, as the total number of offspring ultimately fledged varied primarily with longevity in choughs (as in other birds, Newton 1989), total productivity is likely to have varied markedly among cohorts. We could only assess inter-cohort variation in annual productivity in 3-year-old breeders. Three-year-old males born under good conditions showed higher annual reproductive success, an effect which will contribute to intercohort differences in total productivity. Whilst the productivity of 3-year-old females did not improve with natal conditions, there was no evidence that females born in poor seasons compensated for a decrease in longevity by increasing their fecundity. Although density-dependent effects on vertebrate life-histories have been widely demonstrated (Clutton-Brock et al. 1991; Saether 1997), recruitment age was the only trait that varied with population size in choughs, at least over the current range of data. Both males and females recruited later when the population was large during their first possible breeding year, an effect that may be due to limited territory availability (Bignal et al. 1997). As breeding longevity did not vary with recruitment age, this constraint on recruitment may not have greatly influenced the total number of offspring fledged. Further, across individuals for whom reproductive histories were complete, the total number of offspring varied with natal conditions but not recruitment age. Thus in choughs, the relationship between natal conditions and the total number of offspring fledged is clear, suggesting that individual fitness varies with the quality of the natal environment, which is in turn related to the weather conditions prevailing during the months prior to its birth.
Whilst mean population breeding success varied less in choughs than in systems where breeding occasionally fails completely (examples in Clutton-Brock 1988), the consistent positive covariation amongst life-history traits resulted in considerable among-year variation in the fitness return from breeding. The mean number of offspring breeding attempts ultimately resulting per egg laid (estimated as the mean number of offspring fledging per egg multiplied by the proportion of fledglings recruiting and the mean breeding longevity of a cohort recruit) varied 55-fold, from 0·02 in 1989 to 1·10 in 1983. As evidence of physiological trade-offs can be masked at the population level by individual variation in investment ability (McNamara & Houston 1996), the observation that life-history traits were positively correlated across chough cohorts does of course not preclude the existence of underlying trade-offs at the individual level. Given a trade-off between current reproductive effort and residual reproductive value (Stearns 1992), the occurrence of such marked among-year variation in reproductive returns is likely to influence optimal patterns of parental investment. Selection for investment in self-maintenance and thus for increasing iteroparity is predicted (Orzack & Tuljapurkar 1989; Benton & Grant 1996). Indeed, adult survival was relatively high in choughs (Fig. 2), suggesting that adults may invest in maintaining their own condition (Gaillard, Festa-Bianchet & Yoccoz 1998). Parents might also be predicted to modulate their reproductive effort in response to environmental conditions, reducing their investment when conditions mean that fitness returns are likely to be poor (Clutton-Brock 1991; Erikstad et al. 1998; Festa-Bianchet & Jorgenson 1998; Weimerskirch et al. 2001). Such a strategy would exacerbate initial environmentally determined fitness variation amongst cohorts. However, in the absence of an independent measure of parental investment it is not possible to distinguish the relative roles of constraint and reproductive restraint in driving among-cohort fitness variation in choughs. As such cohort effects can affect population stability (Forchhammer et al. 1998, 2001; Benton et al. 2001; Beckerman et al. 2002)), this provides an intriguing route whereby interactions between environmental variation and adaptive patterns of parental resource allocation could influence population dynamics.
We thank Earthwatch, Merial Animal Health, Nature Conservancy Council, Royal Society for the Protection of Birds, Scottish Executive Environment and Rural Affairs Department, Scottish Natural Heritage and WWF-UK for financial support during the study, and RSPB and SNH for funding the establishment of a data base. Islay farmers kindly allowed access to nest sites. Many people assisted with data collection at various points, including Martin and Robin Bignal, John and Pamela Clarke, David Jardine, Clive McKay, Neil Metcalfe, Allen Moore, Malcolm Ogilvie, Elizabeth Still, Judy Stroud, Paul Thomson and RSPB, SNH and Islay Natural History Trust staff. Lukas Keller advised on analyses, and Peter Arcese and Jon Wright commented on manuscript drafts. Nest visits were licensed by the Nature Conservancy Council and Scottish Natural Heritage.