Toni Laaksonen, Section of Ecology, Department of Biology, University of Turku, FIN-20014 Turku, Finland. Fax: + 358 2333 6550; E-mail: email@example.com
1Two main mechanisms have been proposed to account for the changes in breeding performance of long-lived animals with age: progressive appearance or disappearance of different quality individuals within a cohort and within-individual improvements with age.
2We studied age-related breeding performance of vole-eating Tengmalm’s owls (Aegolius funereus, L.), whose prey populations fluctuate in a cyclic manner in western Finland. In years of low vole abundance, most yearling owls were not able to breed. In years of initially intermediate but increasing vole abundance, clutch size increased with female age and partners of old males initiated egg-laying earlier than those of young males. In years of initially high but decreasing vole abundance, the differences in laying dates were still detectable, whereas age-related differences in clutch size of females tended to disappear.
3Within-cohort analyses did not indicate quality differences between surviving and dying males of the same age in years of increasing vole abundance, whereas in years of decreasing vole abundance early breeding males were more likely to survive than late-breeding males. This novel finding indicates that quality differences between individuals may induce observed age-related differences in only some years, and especially that differential mortality may occur in a different year when the difference in breeding performance is detectable.
4Within-individual analyses showed that individual males advanced their nest initiation with age, and there was a trend that females laid larger clutches with increasing age. In nests of long-lived males, a significant decline in the clutch size from mid-age (2–4 years) to older age (6–10 years) was found.
5We conclude that environmental variation may mask age-related differences in breeding performance at the population level, since the differences appeared to emerge more clearly in poor and intermediate food conditions. Some evidence for the differential mortality-hypothesis was found for males, which are mainly responsible for subsisting their families during the entire breeding season. Substantial improvements in competence or increased reproductive effort from 1 to 3 years of age, and deteriorated skills due to senescence at older ages may account for within-individual changes in breeding performance of males.
Four primary hypotheses that are not mutually exclusive and can operate concurrently within a population have been proposed to explain the observed age-related patterns in breeding performance. The differential mortality- and delayed breeding-hypotheses suggest progressive disappearance or appearance of different-quality individuals within an age-cohort at the population level, and the constraint- and restraint-hypotheses consider changes between breeding attempts of an individual (Forslund & Pärt 1995). The differential mortality-hypothesis suggests that individuals performing poorly in breeding are of lower phenotypic quality than better performing individuals and, hence, also more prone to die or disperse before the next breeding season (Smith 1981; Nol & Smith 1987). Thus, when proportionally more poor-quality than high-quality individuals die between breeding seasons, the mean reproductive performance of a cohort improves in successive years because the relative proportion of high-quality individuals in it increases every year. According to the delayed breeding-hypothesis, the mean performance of a cohort at later age increases if high-quality individuals delay their first breeding attempt and, therefore, recruit in the breeding population later than lower-quality individuals (Hamann & Cooke 1987).
We studied age-related breeding performance of Tengmalm’s owls (Aegolius funereus), a small nocturnal bird of prey that nests in tree-holes and nest-boxes in the northern coniferous forests (Mikkola 1983). Tengmalm’s owls have distinct intersexual duties during the breeding season: while females stay at the nest until the mid-nestling stage, males provide food for their mates and offspring throughout the breeding season (Korpimäki 1981). Most females change breeding site each year (Korpimäki, Lagerström & Saurola 1987; Korpimäki 1993), whereas males stay resident at the site after their first breeding attempt, and their lifetime breeding success can be estimated (Korpimäki 1988c, 1992). An earlier study indicated differences in breeding performance between age classes of Tengmalm’s owls at the population level (Korpimäki 1988b), but the mechanisms leading to age-related differences remained unclear. Here we investigate whether temporally varying levels of natural food abundance determine the occurrence of age-related differences in breeding performance. We use long-term data on laying dates and clutch sizes to investigate age-related changes in breeding performance. We first compare individuals of different ages cross-sectionally at the population level, then examine within-cohort variation between individuals of the same age and finally, investigate consecutive breeding attempts of individuals. Specifically, the following questions were addressed: (1) do the effects of parental age on breeding performance differ at varying levels of natural food? (2) Are age-related differences in performance due to within-cohort changes in the quality of breeding individuals or within-individual changes in competence or reproductive effort? (3) Does the breeding performance of long-lived individuals decline in later years of life?
The data were collected in the Kauhava region (c. 63°N and 23°E), western Finland, during 1984–99. The study area covers c. 1300 km2 and has c. 500 nest boxes and 30 known natural cavities suitable for Tengmalm’s owls. Methods for finding nests and determining laying dates and clutch sizes are given by Korpimäki (1987).
Annually 70–90% of parent owls were trapped from their nests in the early nestling period (Hakkarainen et al. 1996) and aged to four age classes: 1-, 2-, 3- and +3-year-old individuals (Table 1). Age determination was based on recaptures of ringed individuals or on the moult pattern of the primary feathers (see details on the method in Glutz von Blotzheim & Bauer 1980; Hörnfeldt, Carlsson & Nordström 1988). The real age in years was calculated for birds that had been ringed as nestlings and for birds with multiple records if they had been captured for the first time as yearlings or 2-year-olds. This was not carried out for birds that were captured for the first time when they were > 2 years old because the reliability of the age determination is lower in older age classes (Hörnfeldt et al. 1988). All the second or third nests of polygamous individuals were excluded from this study because breeding success is lower in the second nests of the season (Carlsson, Hörnfeldt & Löfgren 1987; Korpimäki 1988d, 1989b).
Table 1. The number of Tengmalm’s owl nests from which at least one parent was trapped, and the number of different aged parent owls trapped at these nests during 1984–99. Phases of the 3-year vole cycle are presented during 1984–93, after which they have been less clear and have not been classified (see Material and methods)
Territory quality for most nest sites was graded from 1 to 5 based on the number of breeding attempts at the site during a 10-year period (1977–86 for 133 territories and 1980–89 for 28 territories) (Korpimäki 1992). The proportion of poor habitats is higher and the abundance of main food (voles) and alternative prey (small birds) is lower on low-quality territories than on high-quality territories of Tengmalm’s owls (Korpimäki 1988c; Hakkarainen et al. 1997).
To account for the high among-year variation, the original values of laying date and clutch size were standardized by subtracting the annual population mean from the parameter value and dividing this value with standard deviation. Negative standardized values for laying date mean that laying began before the annual mean and positive values indicate laying after the mean date. For clutch size, negative values indicate lower and positive values higher than the annual mean performance. In low vole years, no yearling males and only few yearling females were nesting, whereas in years of decreasing vole abundance they formed the majority of breeders (Table 1). This means that the standardized values will be calculated with means derived from old birds in low vole years and with means mainly derived from yearling birds in the decrease vole years. Considering the possible existence of age-related differences in performance this could lead to an over- or underestimation in the standardized values of older birds when comparing between years with or without yearling breeders. Furthermore, the reproductive effort of parent owls may differ according to the phase of the vole cycle (Hakkarainen & Korpimäki 1993, 1994). For these reasons, the cross-sectional analyses were done separately for each phase of the vole cycle.
For analyses of senescence in males and within-individual changes in breeding performance of individual females, longitudinal data from all study years was used and standardized values were calculated with means and deviations derived from older than 1-year-old birds. This was performed to minimize the variation caused by varying proportions of yearling breeders (see Table 1). On the other hand, this value is calculated from a smaller number of nests and is more prone to random error.
The data were divided into two parts for a population level analysis: a cyclic period during 1984–93 consisting of four low, three increase and three decrease years according to the direction of change in local vole abundance during the breeding season of owls, and a weakly cyclic period from 1994 to 1999 (Fig. 1 and Table 1). Cross-sectional analyses were performed in the cyclic period separately for different phases of the cycle by combining data from years in the same cycle phase. In the weakly cyclic period all 6 years were pooled. Interactions between years and the factors of the model were checked before pooling. To avoid pseudoreplication one randomly selected observation of each individual was used in each analysis. Male and female ages were factors in a two-way anova-model with laying date as dependent variable, and in an ancova-model where clutch size was the dependent variable and laying date was as a covariate to account for seasonal decline in clutch size (Korpimäki & Hakkarainen 1991). For significant results an explanation was sought in territory quality, which was added into the model.
Males were considered to have survived or died after a breeding attempt according to whether or not they were recorded in the area in following years. Some of the males may have dispersed outside the study area or bred in unknown localities, but they should be few (Korpimäki 1993; Hakkarainen et al. 2001). Connection between breeding performance and survival was analysed separately in the increase and decrease phases of the vole cycle. Survival was entered as the dependent variable (dying = 0 and surviving = 1) in a binary logistic regression where male age was a categorical independent variable and standardized laying date and clutch size were continuous independent variables. The model contained, first, all variables and interactions and was then reduced by leaving, first, non-significant (P > 0·05) interactions and then non-significant variables out of the model. For analysis of delayed breeding and breeding experience, the breeding performance of males breeding for the first time was compared with t-tests to males with earlier breeding experience at the same age. Comparison was performed between 3-year-olds in the increase and decrease phases and between 2-year-olds in the decrease phase. In these groups all the males should have had an earlier breeding opportunity in a moderate or good vole year.
Breeding performance of the same individual males was followed in successive increase and decrease years of the vole cycle. The included males were breeding as 1-, 2- or 3-year-olds in the increase phase and 2-, 3- or 4-year-olds in the decrease phase, respectively. Similarly, two breeding attempts of individual females were followed pooling females that were breeding for the first time as 1- or 2-year-olds, and then for the second time at the age of 3–5 years. Within-individual changes were also examined for a group of long-lived males that were found breeding at the middle age (2–4 years) and then again at an older age (6–10 years). If there were many observations from the same individuals, 3-year-olds were preferred to others and 4-year-olds to 2-year-olds in the middle age class. In the older age class the observation at the oldest age was used. Significances of within-individual changes were analysed with paired t-tests.
All statistical tests were two-tailed with significance level P < 0·05 and performed with SAS- (ed.6·12, SAS Institute 1996, Cary, North Carolina, USA) and SPSS (version 9·0 SPSS Inc., Chicago, Il, USA) statistical software.
Because only few yearling females and no yearling males were capable of breeding in the low phase of the vole cycle (Table 1), most of the population level analyses were performed for the data from the increase and decrease phases of the vole cycle. In the increase and decrease phases of the vole cycle, the partners of yearling males had the latest and the partners of 3-year-old males the earliest laying date when the age of the female was controlled for [F3,155 = 3·90, P = 0·01 (Bonferroni-corrected 6× contrasts: P = 0·006) in the increase phase and F3,256 = 2·64, P = 0·05 in the decrease phase, no interactions with female age, Fig. 2]. Male age was not associated with the laying date of their partners in the low phase of the vole cycle (F2,33 = 0·56, P = 0·58, Fig. 2) or during the weakly cyclic period (F3,59 = 1·19, P = 0·32). Laying date was not related to female age (F3,33 = 1·00, P = 0·40 in the low, F3,155 = 0·39, P = 0·76 in the increase, F3,256 = 0·81, P = 0·49 in the decrease and F3,59 = 0·16, P = 0·92 in the weakly cyclic phase of the vole cycle). Territory quality did not have obvious effect on laying date when it was added into the anova-model as a factor (for territory quality, F4,122 = 0·99, P = 0·42 in the increase phase and F4,219 = 1·46, P = 0·22 in the decrease phase, no interaction with male or female age).
Three-year-old females had the largest and 1-year-old females the smallest relative clutch size in the increase phase of the vole cycle (F3,149 = 4·05, P = 0·008 (Bonferroni-corrected 6× contrasts: P = 0·006), but not in the low and decrease phases (F3,31 = 0·24, P = 0·86 and F3,246 = 0·79, P = 0·50, respectively, Fig. 3) or during the weakly cyclic period (F3,65 = 0·61, P = 0·61). There were no differences in clutch sizes between the partners of different-aged males (F2,31 = 0·53, P = 0·60 in the low, F3,149 = 0·41, P = 0·74 in the increase, F3,246 = 0·09, P = 0·96 in the decrease and F3,65 = 0·13, P = 0·94 in the weakly cyclic phase of the vole cycle). Territory quality had no obvious effect on the clutch size in the increase phase (for territory quality F4,118 = 1·65, P = 0·17, no interaction with male and female age).
In the decrease phase of the vole cycle, laying date predicted survival significantly (logistic regression: B = −0·84, Wald statistic = 10·83, d.f. = 1, N = 266, P = 0·001) in all age classes (interaction P = 0·13). The partners of surviving males appeared to initiate their egg-laying earlier than partners of dying males, as survival decreased with later laying date (Fig. 4). There was no obvious difference in clutch sizes of these males (B = −0·18, Wald statistic = 0·95, d.f. = 1, N = 256, P = 0·33, no interaction with male age). In the increase phase of the vole cycle, none of the entered variables was associated with male survival (logistic regression: total model χ2 = 8·47, d.f. = 11, N = 162, P = 0·67, all interactions and single variables in reduced models P > 0·3). There were no obvious differences in the performance between males with different breeding experience (Table 2).
Table 2. Mean (± SE) laying dates and clutch sizes of partners of experienced and inexperienced breeding males. Differences between males with and without breeding experience were analysed with t-test or Wilcoxon rank sum-test if the requirements of t-test were not fulfilled
Three-year-olds in the increase phase
−0·24 ± 0·28
−0·16 ± 0·11
0·19 ± 0·25
0·03 ± 0·16
Two-year-olds in the decrease phase
0·16 ± 0·22
−0·13 ± 0·09
−0·16 ± 0·22
0·05 ± 0·15
Three-year-olds in the decrease phase
−0·33 ± 0·21
−0·48 ± 0·13
0·10 ± 0·28
0·16 ± 0·40
The males breeding as yearlings in the increase phase and then as 2-year-olds in the decrease phase appeared to advance their nest initiation (paired t-test, P = 0·05, Fig. 5), but in other age classes similar advancement was not detected (P > 0·5). Clutch sizes of partners of these males did not change from the increase to the decrease phase in any age class (paired t-tests, all P > 0·5). There was a marginally significant trend for females to increase their clutch size with age (Table 3). Long-lived males had smaller clutches in their old days than at middle age (paired t-test, P < 0·01, Fig. 6), but no obvious change was found in the laying date of their partners (P > 0·5, Fig. 6)
Table 3. Within-individual changes in the standardized mean (± SE) laying dates and clutch sizes of females. Data were pooled so that the first breeding attempt includes 1- and 2-year-old females breeding for the first time in the area and the second breeding attempt includes a later attempt of the same individual at the age of 3–5 years. Paired t-tests were performed to determine whether the change between two different breeding attempts was significant
−0·20 ± 0·14
−0·38 ± 0·14
−0·24 ± 0·14
0·40 ± 0·24
Age-related differences in the breeding performance of Tengmalm’s owls were found in years of low vole abundance when no yearling males and only few females were capable of breeding, and in years of increasing vole abundance when food supply was intermediate at the early stages of breeding and young birds performed poorly compared to older ones. In accordance, the differences either became less clear or tended to disappear in the decrease years when the vole abundance was high to intermediate at the early stages of the breeding season. The few studies that have examined age-related breeding performance in relation to estimated environmental variation have concluded similarly that differences become more evident in demanding and stressful conditions (Hamer & Furness 1991; Sydeman et al. 1991; Ratcliffe et al. 1998).
the differential mortality hypothesis
The partners of yearling males began to lay eggs later than the partners of older males in the increase phase of the vole cycle, but our analysis did not indicate within-cohort quality differences between surviving and dying males in this phase. Delayed breeding was not the explanation either, as there were no obvious differences in the laying dates of partners of experienced and inexperienced males. Interestingly, there seemed to be differential mortality of early and late breeding males in the decrease phase of the vole cycle. The relatively good conditions at the early stages of breeding in the decrease years may have given a breeding opportunity to most individuals, whereas in poor or intermediate conditions the lower quality individuals were not able to breed or survive. Thus, in intermediate conditions there may not have been detectable selection because there were no quality differences among breeding individuals of the same age. Instead, since the early breeding 3-year-old males in the increase phase could have been selected during two preceding winters of relatively low food abundance, they may on average have been of better quality than the younger individuals in the increase phase. In the discussion above, we consider survival and timing of breeding to be characters measuring the quality of an individual. On the other hand, decreased survival may reflect a direct cost of late breeding. However, this should only emphasize the importance of early breeding in those years and not change the conclusions above.
Some earlier studies have found support for the differential mortality-hypothesis (Nol & Smith 1987; Gehlbach 1989; Wooller et al. 1990; Espie et al. 2000). However, none of these studies has considered the possible effect of yearly variation in food abundance on the occurrence of selection. Our results indicate that quality differences between individuals may induce observed age-related differences in only some years and, especially, that differential mortality may occur in a different year when the difference in breeding performance is detectable. Nevertheless, as in the earlier studies, we consider that differential mortality alone is not an adequate explanation for age-related differences in the breeding performance of Tengmalm’s owls in the increase phase of the vole cycle.
hypotheses on within-individual changes
The yearling males beginning late in the increase phase of the vole cycle advanced their clutch initiation to the same level with others in the next breeding season, which shows that they were not late breeders consistently. Female age had a positive association with clutch size in the increase phase of the vole cycle, and there was a within-individual trend for clutch size to increase with female age.
The poor performance of yearling owls of both sexes in the low and increase phases and the within-individual improvements could be explained by constraints in foraging skills. The young males may have had difficulties in providing food for their mates during the courtship period early in the increase years because they were not as efficient in foraging as the older males. The finding that young Tengmalm’s owl males had smaller prey caches in their nests than older males (Korpimäki 1988b) supports this view. In species where males provide food for their mate and brood, male age is expected to be more important than female age in determining breeding success (Newton 1979; Korpimäki & Hakkarainen 1991; Taylor 1994; Korpimäki & Wiehn 1998; Ratcliffe et al. 1998). However, clutch size increased with female age independent of male age, indicating that the differences were apparently due to events before the courtship-feeding period. Many studies have shown that maternal condition has an effect on clutch size, and food is probably the restricting factor behind it (e.g. Perrins 1970; Boutin 1990; Korpimäki & Wiehn 1998). Tengmalm’s owl females responded to food supplementation by laying larger clutches even when the natural food levels were high during the egg-laying period (Korpimäki 1989a). This lets us suspect that the young females have been in poorer initial body condition because they were less sufficient foragers than the old females, and they had not been in plentiful food conditions during their earlier life.
Constraint is hard to discern from restraint, and both could be working simultaneously (Curio 1983; Forslund & Larsson 1992). With the present data we cannot distinguish between constraint- and restraint-hypotheses, as reproductive effort at young age may have costs in terms of future survival and breeding success (e.g. Clinton & Le Boeuf 1993; Lunn et al. 1994). Most of the earlier studies, however, have supported the constraint-hypothesis, and only few have found support for the restraint-hypothesis (reviews in Saether 1990; Forslund & Pärt 1995; Martin 1995). Future work should take an experimental approach to distinguish between restraint- and constraint-hypotheses, possibly by offering food supplements and simultaneously measuring reproductive effort of individuals of different age in years of low or intermediate natural food abundance. Data on age-specific survival is also essential to determine whether restraint-hypothesis could work for Tengmalm’s owls.
Tengmalm’s owls showed age-related differences in breeding performance depending on varying environmental conditions. It seems that good food conditions during the preceding year and in the beginning of the breeding season can mask some age-related differences, while they may come up in more demanding conditions. We also found that quality differences between individuals of same age may be detectable in only some years. These results point out the importance of quantifying environmental variation and using long-term data when studying age-related breeding performance. The poor performance of yearling Tengmalm’s owls of both sexes in the low and increase phases of the vole cycle could probably be explained by constraints in foraging and other skills, but could also be the outcome of refrained reproductive effort. Deteriorated skills due to senescence at older ages may account for within-individual changes in breeding performance of long-lived Tengmalm’s owl males.
We are grateful to Vesa Koivunen, Petteri Ilmonen, Mikko Hast, Timo Hyrsky and Mikko Hänninen for assistance in the collection of these data, and to Esa Lehikoinen, Pia Mutikainen, Mia Rönkä, and two anonymous referees for improving the manuscript with their comments. The Academy of Finland financially supported this study (grants to E.K. and H.H.).