No evidence of senescence in a 300-year-old mountain herb

Authors


Correspondence author. E-mail: ehrlen@botan.su.se

Summary

1. Understanding how vital rates and reproductive value change with age is fundamental to demography, life history evolution and population genetics. The universality of organism senescence has been questioned on both theoretical and empirical grounds, and the prevalence and strength of senescence remain a controversial issue. Plants are particularly interesting for studies of senescence since individuals of many species have been reported to reach very high ages.

2. In this study, we examined whether the herb Borderea pyrenaica, known to reach ages of more than 300 years, experiences senescence. We collected detailed demographic information from male and female individuals in two populations over 5 years. An unusual morphological feature in this species enabled us to obtain exact age estimates for each of the individuals at the end of the demographic study.

3. We used restricted cubic regression splines and generalized linear models to determine nonlinear effects of age and size on vital rates. We then incorporated the effects of age and size in integral projection models of demography for determining the relationship between age and reproductive value. As the species is dioecious, we performed analyses separately for males and females and examined also the hypothesis that a larger reproductive effort in females comes at a senescence cost.

4. We found no evidence for senescence. Recorded individuals reached 260 years, but growth and fecundity of female and male individuals did not decrease at high ages, and survival and reproductive value increased with age. The results were qualitatively similar also when accounting for size and among-individual vital rate heterogeneity, with the exception that male flowering probability decreased with age when accounting for size increases.

5.Synthesis. Overall, our results show that performance of both male and female plants of B. pyrenaica may increase rather than decrease at ages up to several centuries, and they support the notion that senescence may be negligible in long-lived modular organisms. This highlights the need to explore mechanisms that enable some species to maintain high reproductive values also at very high ages and to identify the evolutionary reasons why some organisms appear to experience no or negligible senescence.

Introduction

The question why many organisms show evidence of senescence, defined as a decrease in performance with age, has occupied evolutionary biologists for decades. Classical evolutionary explanations of senescence are based on the notion that selection on a trait will be weaker the later it is expressed in life (Medawar 1952). As a consequence, accumulation of mutations or antagonistic pleiotropy may result in senescence (Williams 1957; Hamilton 1966). However, it has recently been argued that senescence is not an inevitable consequence of Medawars’s insight (Baudisch 2005; Steinsaltz, Evans & Wachter 2005). Theoretical models of senescence based on resource allocation have been used to show that the separation of soma and germ lines can make degradation, rather than reparation, of somatic cells optimal with increasing age (Kirkwood 1990; Kaplan & Robson 2009). This explanation of senescence should, however, not apply to organisms lacking an early separation of cells, like many modular organisms. For modular organisms such as plants, it has also been argued that the performance of individuals may improve with age if vital rates covary closer with size than with age and if average size increases with age (‘negative senescence’; Vaupel et al. 2004). In addition, if individuals experience senescence in terms of a decrease in the reproductive value with age (Fisher 1930), this may be masked at the population level by differences among individuals (Vaupel et al. 1998; van de Pol & Verhulst 2006). For example, if frail individuals with lower vital rates are gradually removed from a population owing to their lower survival probabilities, then the resulting increases in mean individual vital rates may counterbalance the effects of age per se and result in increases in average reproductive value with increasing age at the population level (Vaupel et al. 1998).

The empirical evidence for senescence is equivocal and differs among organism groups. While senescence seems to be common in some animal groups, including mammals (Partridge & Barton 1996), this does not appear to be true for several other groups (Finch 1990; Vaupel et al. 1998, 2004). For plants, relatively little information about senescence of individuals is available (Monaghan et al. 2008; Munné-Bosch 2008). Evidence of senescence mainly comes from short-lived species (Roach, Ridley & Dudycha 2009). Studies with long-lived plants have often not found clear evidence of senescence (reviewed in Roach 1993). In the extremely long-lived Bristlecone pine, detailed physiological studies corroborate observations that no senescence seems to occur (Lanner & Connor 2001). Records of very long life spans for other plant species, mostly trees, also seem to suggest an apparent lack of senescence (e.g. Watkinson & White 1986; Bond 2000). However, high mean survival rates and long life spans in some plant species do not necessarily imply a lack of negative effects of ageing. Presence and intensity of senescence should be judged on the basis of age-related changes in vital rates and reproductive values rather than on average rates.

In this study, we collected demographic field data of the dioecious herb Borderea pyrenaica and investigated how performance changes with age for individuals of up to 260 years. To the best of our knowledge, detailed demographic data have not previously been collected from individuals of such high known ages. Borderea pyrenaica is particularly suited for studies of senescence not only because it is unusually long-lived, but also because it is non-clonal and has morphological features that enable easy and precise age determination (Garcia & Antor 1995a). Because Bpyrenaica is dioecious we could also examine whether observed differences in reproductive effort between sexes are associated with differences in senescence. We monitored female and male plants in permanent plots in two populations over 5 years to examine how size, survival, growth and reproduction differ among individuals of different age. In a previous study, with this species, we used females from one of the populations to evaluate different methods of determining nonlinear effects of age and size on vital rates, for incorporation into demographic models (Dahlgren, Garcia & Ehrlen 2011). In this study, we use the method found to most reliably identify nonlinear relationships, restricted cubic regression splines, to investigate whether there is evidence of senescence in Bpyrenaica. For each sex, we examined how vital rates changed with age and size. We then incorporated the nonlinear age- and size-dependent vital rate models into integral projection models (IPMs) of demography to study the total effects of age on reproductive value. We address the following specific questions: (i) How are growth, survival and fecundity influenced by increasing age? (ii) Are negative direct effects of age hidden by positive indirect effects via size? (iii) Do relationships between age and vital rates differ between sexes? (iv) Does the reproductive value decrease or increase with age, as a result of the combined effects of age and size on all vital rates? Overall, our results provide no evidence for senescence in B. pyrenaica. On the contrary, we conclude that performance of both male and female plants may remain constant, or increase, at ages up to several centuries, also when accounting for size.

Materials and methods

Study System

Borderea pyrenaica Miégeville (Dioscoreaceae) is a relict dioecious plant restricted to screes of the central Pyrenees, usually at altitudes higher than 1800 m a.s.l. Vegetative growth and flowering start almost simultaneously in June. Pollinators include flies, ants and lady beetles, and the seed-to-ovule ratio is high, usually above 80% (García, Antor & Espadaler 1995). Fruits reach their final size by early July, remain greenish until late August, and seeds are released in September. This small geophyte has an unusual feature: a distinct scar is left on the tuber each year when the annual stem dies back in September (Garcia & Antor 1995a). This makes age determination of the plant by morphological inspection possible. Available data show that Bpyrenaica has one of the longest life spans recorded for a herb, over 300 years (Garcia & Antor 1995a). Males start reproduction at younger ages than females, at 10–20 vs. 15–35 years of age (Garcia & Antor 1995b). They show a larger floral display, flower more frequently (almost yearly) and grow faster than females. As a result, population functional sex ratios are male biased (Garcia & Antor 1995a). The earlier and more frequent flowering in males is probably the result of a lower reproductive effort (Garcia & Antor 1995b).

Data Collection

Two populations were used for the study. The Pineta population (42° 41′ N, 0° 06′ E; 2000 m a.s.l.) is located on a steeply sloping scree with a thick layer of stones, and B. pyrenaica densities may reach a few hundred plants per square metre. The Ordesa population (42° 39′ N, 0° 01′ E; 2100 m a.s.l.) is located on a flatter area with fewer stones, and B. pyrenaica density is less than one hundred plants per square metre. Both populations comprise several thousand individuals. Six permanent plots in Pineta and 15 in Ordesa were established for monitoring. Plant individuals were followed over 5 years in Pineta and 4 years in Ordesa and censused in August each year. For each individual, we recorded the presence or absence, sex (male, female or vegetative), number of leaves, length of the largest leaf, number of flowers, number of fruits in females and number of seeds in each fruit. The determination of the sex of an individual is not possible in the field unless the plant produces flowers. Absence of an individual at one recording could be due to either death or dormancy. Based on previous knowledge of the system and the infrequency of dormancy over more than one season in this data set, individuals that did not produce above-ground parts in 2 consecutive years were considered dead from the first of these 2 years. Individuals reappearing the year after absence were considered dormant in the first year and assigned the size observed after reappearance. In the last year of monitoring, tubers of both living and dead plants were excavated. Biomass was determined for tubers dried at 40 °C to constant weight. Above-ground size was calculated as log(number of leaves × length2). This measure was chosen based on regression models of biomass on different combinations of leaf number and stem length using data from a smaller set of plants of different sizes. Plant ages were estimated by counting the number of scars on 748 excavated tubers.

Data Analysis

We estimated effects of age on above-ground size, tuber biomass, growth, fecundity and survival using regression models. Growth was defined as the change in above-ground size between two subsequent years. Fecundity was divided into two components: the probability of flowering and the number of seeds in females or the number of flowers in males.

The effects of age on size and vital rates were modelled with ‘5-knot restricted cubic regression splines’ to allow nonlinear relationships (see Dahlgren, Garcia & Ehrlen (2011) for details of this method). Restricted cubic spline functions are constituted by cubic piecewise regressions with the tails of the curve restricted to be linear so as to make them more reliable for predictions (Harrell 2001). The significance of age was determined by testing the null hypothesis that all spline components equalled zero. Significance tests of individual spline components are not meaningful, since they depend strongly on the other splines, and we were mostly interested in the curves for high ages. We, therefore, plotted the estimates for models with statistically significant effects of age. Based on 95% confidence intervals around the curves, we then determined whether there were clearly significant positive or negative trends for advanced ages. The curves estimated using the parametric restricted cubic spline functions were compared with curves produced by the nonparametric loess method (Cleveland 1979; using default settings in the R function ‘loess’) and with ‘semiparametric’ methods of penalized smoothing splines (P-splines; e.g., Lee & Oh 2007), in the r packages pspline (using default settings in the sm.spline function) and mgcv using the ‘gam’ function. Overall conclusions regarding senescence remained the same regardless of method, despite P-spline curves being in some cases unrealistically wiggly (not shown). In the remaining cases, the knot number estimated for optimal smoothing in the gam function (based on generalized cross-validation or maximum likelihood and using the default type of smoothing splines, which are very similar to restricted cubic splines) was fewer than five. This should entail that our a priori choice of five knots is sufficient. Results of patterns for old ages were also in agreement with preliminary analyses where linear regression lines were fit to the data of the 20–40 oldest individuals (not presented). We pooled the data for all years, but fitted models separately for the two populations Pineta and Ordesa, because they are known to differ in several ways (Garcia & Antor 1995a). The design package (Harrell 2001) for r 2.9.1 (R Development Core Team 2009) was used to fit all models as well as for plots and inference.

The effects of age on above-ground size and tuber mass were modelled with ordinary least-squares regression. Despite the fact that individual plants were measured every year in the field, mean above-ground sizes over the entire study interval were used for these analyses because changes in relative size between years were small, and tuber mass could be measured only once at the end of the study. Above-ground size and tuber mass were included in models as their natural logarithms. The effects of age on above-ground growth were modelled using a multiple regression of size on size the previous year and age.

The effects of age on probabilities of survival and flowering were modelled using logistic regression. Age estimates were possible to obtain for almost all large individuals that died during the study interval, but only for a subset of the smaller individuals (42 of 119 recorded deaths in Pineta and 17 of 45 deaths in Ordesa). To be able to include the effects of both size and age in the same models, we imputed most probable age estimates of all non-aged individuals based on size, using a 5-knot restricted cubic spline model of size predicting age (cf. Harrell 2001). Trends in models fitted to data containing imputed age estimates and models fitted to data from only individuals with observed age were always the same (not presented).

The effects of age on seed number were modelled using quasi-Poisson generalized linear models (i.e. Poisson regression with standard errors adjusted to account for overdispersion). The number of male flowers (log-transformed) had an approximately Gaussian distribution, and age effects were modelled with ordinary least-squares regression since overdispersion was very high (Young, Campbell & Capuano 1999).

We investigated the potential effects of among-individual heterogeneity in vital rates by adding individual as a random factor in models of growth and fecundity, using the lmer function in the lme4 package in r, and compared these models with the models without random effects (Cam et al. 2002; van de Pol & Verhulst 2006). To test the hypothesis that correlations of vital rates with age were caused by indirect effects of size, we tested for age effects in models including 5-knot restricted cubic spline functions of both age and size.

To examine whether males and females exhibited different relationships between age and vital rates, we tested the effect of sex and its interaction with age in all models, prior to the other analyses. Since the relationship between flowering probability and age differed substantially between sexes (see Results), it was modelled separately for each sex. To avoid a biased sample by ignoring individuals that did not flower during the entire interval (most probably females because of the frequent and early onset of flowering in males), we included vegetative individuals in models of both sexes and calculated weights representing the estimated probability of an individual of a given size and with an unknown sex to be female (vs. male). We used the size and sex of vegetative individuals that started to flower during the study period to calculate these weights as the ratio of the predicted probability of not flowering for females to the sum of the predicted probabilities of not flowering for males and females.

Demographic Modelling

To examine how reproductive values changed with age, we specified IPMs using the nonlinear regression functions obtained from the statistical analyses. Parameterization using regression models and implementation of the models using r followed the methods in Ellner & Rees (2006). Seed survival in the seed bank and seedling establishment rates were calculated based on data from a sowing experiment (M. B. García, unpublished data). IPMs are an extension of matrix models (Easterling, Ellner & Dixon 2000), allowing continuous states. Restricted cubic regression splines have been found to reliably incorporate complex nonlinearities in IPMs (Dahlgren, Garcia & Ehrlen 2011). We used the IPMs to investigate how the reproductive value, which is a measure of how much an individual of a given age should contribute to future population growth (or fitness; Fisher 1930), changes with age. Reproductive values (V; as functions of size and age) for each sex (here for males, represented by the subscript m) can be calculated based on the equality:

image(1)

where λ is (total) population growth rate, fm is the fecundity function (based on the regression models of flowering probability and reproductive output of flowering individuals) and Pm is the survival–growth transition kernel (based on the regression models of survival and growth); inline image is a constant that does not affect the relative distribution among individuals. Therefore, eqn 1 can be divided by inline image, and if we let vm Vm/inline image, then the solution for vm is (Steve Ellner, pers. comm.):

image(2)

In male models, the observed ratio of the number of seeds to the number of male flowers was used to convert number of male flowers to number of seeds. Vital rates were specified as nonlinear functions of both age and size unless these parameters did not clearly affect a vital rate (P > 0.6). Other (less) nonsignificant variables were kept, because models including also nonsignificant variables generally produce more reliable predictions (Harrell 2001). As a result, only age and nonlinear size effects on growth in the model of males in Pineta as well as all effects on seed number in the model of females in Ordesa were removed. We also simplified models by optimizing the Akaike information criterion (AIC). Reproductive value curves calculated with IPMs based on these vital rate models were almost identical to the curves from IPMs based on the regression models described earlier and are not presented. In accordance with the statistical analyses, models of the different sexes did not differ for survival, and for growth, they differed only in the Pineta models. Given the low number of very old individuals, absorbing age classes including all individuals over 200-year old were used. Relationships of age with reproductive values as well as stable age distributions depend on the population growth rate in demographic models. To examine relationships corresponding to stable population size and the assumed long-term average population growth rate, we used models where individual growth intercepts were increased to satisfy λ = 1.000. IPMs based on observed vital rates predicted growth rates very close to, but slightly below, unity (λ = 0.999 and 0.988 for females in Pineta and Ordesa). Models with observed and increased growth intercepts resulted in qualitatively similar results, and only results for models with growth intercepts corresponding to stable population size are presented later.

Results

Age vs. Growth, Survival and Fecundity

We were able to determine the age of 424 of 518 recorded individuals in Pineta, with ages spanning from 1 to 260 years. In Ordesa, we determined the age of 324 of 368 recorded individuals, with ages spanning from 1 to 175 years. Above-ground size and tuber size were closely linearly related (R2 = 0.88 and 0.77 in Pineta and Ordesa, respectively), and both were significantly and asymptotically related to age (R2 = 0.42 and 0.62). Maximum size was approached at an age of about 100 years in Pineta (Fig. 1a), but already at about 50 years in Ordesa (Fig. 1b). Size changed only little between years in both populations and plant size in a given year was close to linearly related to size the previous year in regression models (Pineta: R2 = 0.88, Ordesa: R2 = 0.83).

Figure 1.

 Relationship between age (years) and above-ground size in Borderea pyrenaica of the Pineta population (a), and the Ordesa population (b). Solid lines are 5-knot restricted cubic spline regression lines. Dashed lines are 95% confidence intervals. Size was calculated as log(number of leaves × length2).

Annual growth was not affected by age in Pineta (P = 0.52). For the Ordesa population, there was a weak (P = 0.10) trend of increasing growth with age, but only up to about 70 years. There was a significant difference in growth between sexes only in Pineta (P = 0.001), mainly in terms of a faster growth of males, but growth of neither sex was affected by age. In regression models of growth over all study years, practically zero variation was explained by allowing intercepts to vary among individuals, resulting in the same estimates of age effects. This suggests that the lack of negative effects of age on growth was not caused by among-individual heterogeneity. Size was slightly nonlinearly related to survival and fecundity in models including also age (a representative example in Fig. 2a).

Figure 2.

 Survival over size (a) and age (b) in population Ordesa (5-knot restricted cubic spline regression lines with 95% confidence intervals), showing the effects of the variables on log-odds of survival adjusted to the mean of the other variable (mean size = 7.81 and mean age = 50.5 years). Size was calculated as log(number of leaves × length2). Note the different y-axis scales in panels a and b.

Survival was very high overall and increased with age for old individuals in both populations, after a dip in the curves at ages of around 60 years (P < 0.001; Fig. 3). There was no evidence of differences in the survival curves between sexes. The significant relationships between survival probability and age, with increasing survival probability at old ages, remained in both populations in models also including size (P < 0.001 in Ordesa and P = 0.030 in Pineta, Fig. 2b).

Figure 3.

 Survival probability over age (years) in the Ordesa population (5-knot restricted cubic spline regression line with 95% confidence interval). In cases when age was not directly recorded, it was estimated based on size (grey triangles).

The positive effects of age on flowering probability at low ages levelled off at intermediate ages in both Pineta (Fig. 4) and Ordesa (P < 0.001 for both sexes). Trends at high ages differed between sexes and populations, but slopes were not significant at high ages apart from a decrease with age for male flowering probability in Ordesa (not shown). Seed number showed similar, albeit not significant (P = 0.13 in Pineta), age patterns as flowering probability, i.e., with initial increases that levelled off at intermediate ages. In males, flower number increased with age in both of the populations (P < 0.01), levelling off at advanced ages in Pineta, but increasing also at advanced ages in Ordesa (not shown).

Figure 4.

 Effects of age (years) on flowering probability in the Pineta population (P < 0.001 for both sexes). Relationships were modelled with 5-knot restricted cubic regression splines. Light grey crosses, males; dark grey triangles, females; black circles, juveniles of unknown sex; 95% confidence intervals (not included for clarity) show that the slopes at high ages do not differ significantly from zero.

Relationships between age and fecundity parameters were largely unaffected by including random individual intercepts, suggesting that among-individual heterogeneity does not obscure negative effects of old age. The inclusion of size in the models showed that the observed trends of positive effects of age on fecundity for young ages seemed to be partly caused by size. Male flowering probability decreased significantly with age in both populations when including size in the regression models (P < 0.01). Moreover, male flower number and female flowering probability were no longer significantly positively related to age when including size in models [although a positive trend for male flower number (P = 0.080) remained in Ordesa].

Age-Related Changes in the Reproductive Value

Using the statistical models as components of IPMs, we found that reproductive values increased with age over the entire observed age span for females and males in both populations (Fig. 5a,b). The relationship between reproductive value and age at advanced ages largely reflected that of survival and age, as reproduction had only small effects on reproductive values.

Figure 5.

 Reproductive value over age for females (black) and males (grey) in Pineta (a) and in Ordesa (b) (for λ = 1.000, see Materials and methods). Age zero = seeds in the seed bank.

Discussion

In our field study of B. pyrenaica, we were able to precisely age and quantify all vital rates for plants of up to 260 years of age. Detailed demographic studies including individuals of such high ages have, to be the best of our knowledge, not been reported previously. The study was designed to detect signs of senescence in a herb that is one of the oldest ever recorded, with life spans of over 300 years. The long life span together with morphological features enabling accurate ageing of individuals in this species implies that our possibilities to detect consistent trends in vital rates with age should be relatively good. Still, the overall pattern of constant or even improved performance with age in both males and females provides no evidence of senescence in B. pyrenaica.

Mean individual size and fecundity increased asymptotically with age up to ages of around one-third of the maximum age observed per population. Survival increased with age in both populations, and no deaths were observed for the oldest individuals. However, despite the fact that increases were statistically significant, estimates of survival at high ages rest on few observations and should be taken with greater caution than results for growth and fecundity. Constant fecundity and survival (mortality plateaus) with increasing age at high ages have been found also in other plant species as well as in some animals (e.g. Finch 1990; Ehlers & Olesen 2004; Horvitz & Tuljapurkar 2008). However, in general, such a lack of negative effects of high age does not necessarily imply a lack of senescence. In short-lived species, a cross-sectional study such as ours could not be used to distinguish age effects from environmental differences experienced by different cohorts, but such cohort effects should not cause long-term trends in our data. Still, senescence might be masked by effects of size or by vital rate heterogeneity and successive removal of frail individuals (Vaupel et al. 1998, 2004).

Vital rate heterogeneity could result in increases in mean vital rates with increasing age if frail individuals within populations have intrinsically lower survival, growth or fecundity rates and are gradually removed because of lower survival probabilities. In the present study, two observations suggest that vital rate heterogeneity was not the main cause of the lack of negative effects of old age. First, there were no qualitative changes in relationships between age and vital rates when allowing intercepts to vary among individuals in the regression models (cf. Cam et al. 2002). The observed relationships (or lack thereof for high ages) may thus be the result of changes within individuals over time, although the fact that our study duration was relatively short compared with the potential life span individuals implies that our ability to detect effects of heterogeneity should be relatively low in this study. Second, for older individuals, the increase in survival probability with increasing age and the usually constant mean fecundity and size are unlikely to be the results of vital rate heterogeneity because at the extreme ages reached by this plant, individuals with inherently lower vitality should, arguably, to a large extent already have disappeared from populations. Still, it is possible that part of the early increases in vital rates with age were caused by vital rate heterogeneity and the gradual disappearance of less fit individuals.

Constant or increasing vital rates with increasing age have been suggested to partly depend on increases in average size with age (Harper 1977, Vaupel et al. 2004). In B. pyrenaica, much of the observed increases in vital rates with age for younger plants appeared to be caused by such a positive relationship between size and age. However, statistical analyses accounting for size failed to identify consistent signs of senescence, in terms of overall decreases in vital rates with age or decreases at high ages. Indeed, the increase in survival probability appears to be a sign of ‘negative senescence’ (cf. Vaupel 2004). Increases in vital rates despite halted growth, as in B. pyrenaica, may suggest that old plants use resources to increase robustness or vitality, but not size. In contrast to the lack of overall negative effects of age, vital rates in young adults sometimes decreased with age, despite increases in size, reached a minimum at ages of about 50–60 years and then increased at higher ages (Figs 2 and 3). Similar patterns have been observed for vital rates and related traits also in other species and seem to coincide with life history stage transitions (Vaupel et al. 1998; Bond 2000; Low & Pärt 2009). In B. pyrenaica, such negative effects of age in younger individuals may be the result of a higher proportional reproductive effort in young adults (Garcia & Antor 1995b). However, the decrease occurred in both sexes in spite of considerably higher reproductive investments, slower growth and lower flowering probability in females than in males (Garcia & Antor 1995a, this study). Overall, relationships between age and all vital rates were similar between sexes in B. pyrenaica.

The presence of senescence should ultimately be judged based on the relationships between age and reproductive value, as reproductive values integrate information from all vital rates for current and future ages. In Bpyrenaica, reproductive values increased with age in both males and females in both populations, mainly as a result of nonlinear age and size effects on survival. This result suggests that there is no overall senescence but that the potential contribution of individuals to population growth tends to increase the older they get. Accordingly, in a physiological study of B. pyrenaica, neither hormones related to senescence nor free radicals increased with age (M. Oñate, M.B. García and S. Munné-Bosch, unpublished data). In that study, it was suggested that the lack of senescence may be related to the fact that B. pyrenaica individuals alternate meristems during their life span. Also in other investigated long-lived iteroparous herbs, negative effects of age have usually not been found (Roach 1993). Detailed studies of senescence in long-lived plants as well as for other types of organisms are still too few to allow generalizations about when and where senescence is most important. Although our study with B. pyrenaica only followed individuals for a very short portion of their potential life span, the results suggest that the negative effects of age can be very weak or absent in some organisms, enabling not only a very long life span but also a high performance at ages of several centuries. This emphasizes the need for further studies, investigating both the physiological and morphological mechanisms that enable some species to maintain high reproductive values also at very high ages and the reasons why evolution has resulted in widely different degrees of senescence in different organisms.

Acknowledgements

We are grateful to R. Antor, D. Guzmán and D. Goñi for field assistance, Carlos M. Herrera for technical support during age dating, the National Park of Ordesa (project 018/2008) and the Regional Government of Aragón (LIFE project B4-3200/96/503) for supporting and funding to M.B.G., The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) for funding to J.E. and J.P.D., and The Swedish Research Council to J.E.

Ancillary