Senescence and age-related reproduction of female Columbian ground squirrels

Authors

  • D. R. Broussard,

    Corresponding author
    1. Department of Biological Sciences, 331 Funchess Hall, Auburn University, Auburn, AL 36849–5414, USA; and
      David Broussard, 331 Funchess Hall, Department of Biological Sciences, Auburn University, Auburn, AL 36849-5414, USA. Tel. (334) 844 9252; Fax: (334) 844 9234; E-mail: brousdr@auburn.edu
    Search for more papers by this author
  • T. S. Risch,

    1. Department of Biological Sciences, 331 Funchess Hall, Auburn University, Auburn, AL 36849–5414, USA; and
    Search for more papers by this author
    • Present address: Department of Biological Sciences, Arkansas State University, State University, AR 72467, USA.

  • F. S. Dobson,

    1. Department of Biological Sciences, 331 Funchess Hall, Auburn University, Auburn, AL 36849–5414, USA; and
    Search for more papers by this author
  • J. O. Murie

    1. Department of Biological Sciences, University of Alberta, Edmonton, Alberta, T6G 2E9, Canada
    Search for more papers by this author

David Broussard, 331 Funchess Hall, Department of Biological Sciences, Auburn University, Auburn, AL 36849-5414, USA. Tel. (334) 844 9252; Fax: (334) 844 9234; E-mail: brousdr@auburn.edu

Summary

  • 1Three hypotheses have been proposed to explain age-structured patterns of reproductive investment and somatic investment: residual reproductive value, senescence and evolutionary restraint. We evaluated these hypotheses for female Columbian ground squirrels (Spermophilus columbianus) by examining age-related patterns of somatic and reproductive investment. Females were designated as successful (those that weaned litters) and unsuccessful (those that did not wean litters).
  • 2Somatic investment varied among both successful and unsuccessful females of different ages, with yearlings having the highest investment. Considering all females, reproductive investment varied among age classes with yearlings and the oldest (6–9-year-olds) having the lowest investments. However, when only successful females were considered, reproductive investment was lowest in the yearlings and not significantly different among older females.
  • 3The highest proportion of successful females occurred in the middle adult age classes, while yearlings and the oldest females displayed the lowest proportion of successful females. During the breeding season, somatic investments of successful and unsuccessful females differed significantly only in the yearling age class, with unsuccessful females having the highest investment.
  • 4Evolutionary restraint or constraint explained patterns of reproduction in the yearling age class, where both reproductive investment and proportion of reproductive females were low. There was evidence for senescence of reproduction by some of the oldest females.

Introduction

Allocation of resources to reproduction (reproductive investment) and growth and maintenance (somatic investment) is an important aspect in an organism's life history (Stearns 1976, 1992; Morris 1987; Roff 1992). Trade-offs occur between these aspects of life histories because individuals cannot allocate simultaneously maximal amounts of resources to both producing offspring and to growth and maintenance. Furthermore, as females age, investment of resources to survival and maintenance vs. reproduction may change, based on the expectation of future offspring (Pianka 1988). Reproductive output of most birds and mammals usually increases, and then declines or remains constant over their lifetime (Clutton-Brock 1984, 1988; Derocher & Stirling 1994; Pärt 1995). Three hypotheses have been proposed to explain these patterns of lifetime reproduction. They include the residual reproductive value hypothesis, evolutionary restraint hypothesis and the senescence hypothesis.

The residual reproductive value hypothesis (Pianka & Parker 1975) suggests that as females age, reproductive investment in current offspring should increase as residual reproductive value decreases. Residual reproductive value is the age-specific expectation of all future offspring beyond those presently at stake (Williams 1966). Because survivorship is assumed to decrease with age, each offspring should become more valuable to females as they age. Assuming that individuals have a fixed energy budget throughout life, an individual with a high probability of future reproductive success should be less likely to invest highly in current reproduction at a cost to somatic investment than an organism with a lower probability of future reproductive success. Therefore, older individuals should display high reproductive investment compared to younger individuals, and a concomitant decrease in somatic investment (Pianka 1988).

The evolutionary restraint hypothesis suggests that the youngest females restrain their reproductive investment to avoid a possible cost to future reproduction and/or survival (Curio 1983; Marchetti & Price 1989; Rohwer 1992; Blums et al. 1996). According to this hypothesis, primiparous females should allocate more resources to somatic needs and less to reproduction compared to older females, to avoid this possible cost. Therefore, primaparous females should have lower reproductive investments and higher somatic investments compared to older females.

The senescence hypothesis (Williams 1957; Partridge 2001) addresses biological processes underlying the marked decline in age-specific fitness components, such as fertility, as an organism ages. Individuals are senescing if they exhibit a generalized deterioration of many physiological functions, including declining reproduction and annual survival. Senescence is based on an evolutionary trade-off between traits that enhance reproduction early in life and declining condition of an organism later in life (Kirkwood & Rose 1984; Rose 1984). Senescence predicts that older individuals will show an age-specific decline in reproductive investment and somatic investment and an increase in age-specific mortality.

The purpose of our study was to evaluate the residual reproductive value hypothesis, the evolutionary restraint hypothesis, and the senescence hypothesis with an extensive data set on a large population of known-aged Columbian ground squirrels (Spermophilus columbianus, Ord). We evaluated these hypotheses using litter mass as a reflection of reproductive investment and change in female mass from emergence from hibernation to litter emergence as a reflection of somatic investment during the breeding season (see Millar 1977; Tuomi 1980; Ebenhard 1990). Thus, we measured reproductive and somatic investments during the same time period in a hibernating species with a short active season (Dobson 1988). The three hypotheses are not mutually exclusive, but they differ in the age-specific predictions. We applied these predictions to young, middle-aged and older female Columbian ground squirrels. Our goal was to find evidence for the most likely of the hypotheses, and the most parsimonious explanation of age-structured reproductive and somatic investments.

Materials and methods

study organism

Columbian ground squirrels are hibernating, semifossorial rodents whose reproductive ecology and life history have been well studied. They provide an excellent model for evaluating hypotheses about life-history evolution because much is known about their population biology (e.g. Boag & Murie 1981; Zammuto & Millar 1985a,b). Population size is regulated by food resources (Dobson & Kjelgaard 1985a; Dobson 1995; Dobson & Oli 2001) and life histories are extremely plastic and influenced by individual body condition (Dobson & Kjelgaard 1985b; Dobson & Murie 1987; Dobson 1988; King, Festa-Bianchet & Hatfield 1991; Dobson 1992; Neuhaus 2000). Adult ground squirrels emerge from their hibernation burrows in spring and mate. Once the young are weaned in early summer, the adult females must then prepare for an 8-month period of hibernation, which begins in late summer (Murie & Harris 1982). Therefore, Columbian ground squirrels have a short amount of time in which they must allocate resources to both reproductive and somatic needs (Dobson, Badry & Geddes 1992). Due to the long hibernation season and short active season, trade-offs between reproduction, survival and daily metabolic maintenance should be pronounced. Additionally, Columbian ground squirrels are long-lived for rodents, and senescence is most evident in long-lived mammalian species (Kirkwood & Rose 1984).

data collection

Field studies were conducted from 1983 to 1990 at the Turnbull National Wildlife Refuge, 35 km south-west of Spokane, Washington, USA (47°26′ N, 117°36′ W, elevation 695 m). The 9-ha study site contains a large meadow surrounded by an open woodland of Ponderosa pine (Pinus ponderosa) and rocky outcrops. A large wetland bounded the site on the south, and was an unsuitable habitat for ground squirrels. Ground squirrels were trapped within a day after emergence from hibernation in early to mid-March Captured individuals were weighed to the nearest 5 g, examined for reproductive condition and marked uniquely with numbered metal ear tags and dye markings on their pelage. New ear tags and dye markings were reapplied as needed on recapture.

Juvenile ground squirrels were captured within 1–2 days after emergence from natal burrows (in early to late May). Emergence date could be anticipated because date of breeding was known or could be estimated within a few days. Mating dates were determined by observing above-ground copulations, or if mating was not observed, from the external appearance of the female's genitalia, behavioural evidence of underground copulation (including oral cleaning of the groin area by both sexes), detection of a copulatory plug or the presence of sperm in vaginal smears. Dates of litter emergence were estimated by adding 51 days to the estimated or observed mating date: 24 days from mating to parturition and 27 days from parturition to first emergence of the litter (Shaw 1925; Murie & Harris 1982; Murie 1992). The site of emergence was anticipated for most litters. Natal burrows could be identified from observations of females stocking them with nesting material before young were born, or by observing females emerge from or enter them during the lactation period.

During the period of litter emergence, the study site, especially the locations of natal burrows, were checked one to three times daily for newly emerged juveniles. Juveniles were captured in wire-mesh traps that were placed over natal burrows, or in live traps that surrounded the burrows. Most juveniles were captured, weighed to the nearest 5 g and marked within the first 3 days of emergence. The lactating female that associated with a natal burrow was assumed to be the mother of the litter that emerged from that burrow. Female Columbian ground squirrels are highly philopatric and settle as adults near the burrows in which they were born (Murie & Harris 1984). Therefore, female age was known from date of birth for most females in our data set (161 of 229). However, females during the first year of the study were of unknown age. These females, along with a few rare females that dispersed into the study site, were assigned a minimum known age of 2 years (68 of 229). Exclusion of these females changed none of the age-structured patterns of reproductive investment and somatic investment and, hence, changed none of our interpretations of these patterns. Therefore, these females were included to accumulate a substantial sample of the oldest females, for examination of the proportion of successful individuals in each age group (see Results). As the age assigned to these females is the minimum possible, their inclusion in the oldest age groups is a conservative treatment of the data.

Data on reproductive investment and somatic investment were collected each time a female weaned a litter. We obtained two measures of investment.

  • 1Reproductive investment: total litter mass at first emergence from the natal nest.
  • 2Somatic investment: change in a female's mass between her emergence from hibernation and emergence of her litter.

Total litter mass was calculated as the sum of juvenile mass in each litter near the time of weaning. Because lactation had virtually subsided at this time, the majority of direct energetic investment into the offspring was considered complete (Mattingly & McClure 1982; Kenagy, Sharbaugh & Nagy 1989; Michener 1989). We assumed that litter mass at weaning is an accurate reflection of maternal energy investment, in part because litter size at weaning was not significantly different from litter size at conception or birth in Columbian ground squirrels (Murie, Boag & Kivett 1980). Because this is a wild population of ground squirrels where rearing of juveniles takes place below-ground, we were unable to assess reproductive investment for those females that lost litters during lactation. Females were designated as successful (those that weaned litters) and unsuccessful (those that did not wean litters). Unsuccessful females include those females known to have mated but did not lactate, those that mated, subsequently lactated, but lost their litter some time during lactation, and those that did not mate. For unsuccessful females, somatic investment was measured as the mass gained from her emergence from hibernation to 51 days after mating, the time at which her litter would have emerged. Some litters were killed by badgers (Taxidea taxus) and other litters may have suffered partial predation by badgers during lactation (Murie 1992). These litters were eliminated from data analyses (60 litters).

All analyses were performed using SAS (1990) statistical software. Analyses of variance (GLM procedure) were performed to test for the effect of age on reproductive investment and somatic investment. Data for age groups 6, 7, 8 and 9 were pooled due to low sample sizes in those age groups. Unless stated otherwise, P-values ≤ 0·05 were considered significant.

Some maternal and reproductive characteristics varied among years. To control for this variation, a two-way analysis of variance was used to search for differences among these variables with age and year as class variables. No interactions between age and year were found for any maternal or reproductive variable (all Ps > 0·05).

Results

age-structured variation

Analyses of variance and Duncan's multiple range tests were used to search for differences in maternal and reproductive variables among females of different age classes (Table 1). Among successful females aged 2 years and older, reproductive and somatic investments did not differ among these age classes. Thus, we pooled data on older females for some of our analyses. Successful yearling and older females (aged 2–9 years) showed different patterns in many maternal and reproductive characteristics. Older females had significantly higher body masses at spring emergence from hibernation and at litter emergence, lower gain in mass during the breeding season, larger litter sizes and earlier dates of litter emergence when compared to yearling females.

Table 1.  Reproductive and maternal variables for successful female Columbian ground squirrel by age class (± 2 standard errors of the mean). Juvenile mass is average mass of offspring at the end of lactation. Mass at litter emergence is female mass at or near the end of lactation. Data for older females (2–9 years old) are pooled for comparison with yearlings using Student's t-test
VariableAge class (years)
23456–9Older 2–9Yearling 1tP
  • *

    SE in days.

Mass at emergence  from hibernation (g)324 ± 11·8
(43)
333 ± 13·0
(39)
334 ± 16·8
(18)
357 ± 30·2
(13)
323 ± 22·8
(9)
332 ± 3·72
(122)
251 ± 6·50
(40)
10·80·001
Mass at litter  emergence (g)417 ± 9·40
(43)
412 ± 12·0
(39)
428 ± 9·30
(18)
438 ± 25·8
(13)
405 ± 35·4
(9)
418 ± 3·49
(122)
375 ± 5·11
(40)
 7·00·001
Somatic investment (g)
94 ± 5·2
(43)
79 ± 6·4
(39)
94 ± 8·6
(18)
80 ± 11
(13)
82 ± 9·8
(9)
87 ± 3·4
(122)
124 ± 7·37
(40)
 4·60·001
Reproductive investment (g)377 ± 35·1
(43)
363 ± 35·6
(39)
400 ± 50·1
(18)
412 ± 55·1
(13)
398 ± 55·7
(9)
379 ± 9·90
(122)
284 ± 13·9
(40)
 5·00·001
Litter size
3·98 ± 0·38
(43)
3·74 ± 0·40
(39)
4·53 ± 0·80
(18)
5·23 ± 1·04
(13)
4·10 ± 0·80
(9)
4·12 ± 0·13
(122)
2·95 ± 0·16
(40)
 4·70·001
Date of litter
 emergence*
29 April ± 2
(40)
1 May ± 2
(37)
30 April ± 3
(17)
25 April ± 5
(12)
29 April ± 3
(9)
29 April ± 1
(115)
3 May ±± 1
(40)
 3·90·001
Juvenile mass (g)98 ± 7·0
(43)
100 ± 6·1
(39)
97 ± 12
(18)
83 ± 10
(13)
93 ± 13
(9)
96 ± 2·0
(122)
101 ± 3·8
(40)
 1·10·25

To analyse annual survival and frequency of successful and unsuccessful reproduction in all females, age classes from 1 to 5 and pooled 6–9 were examined (Table 2). The oldest age classes were pooled because of low sample sizes among these females. Only half of yearling females weaned litters, while a vast majority of females aged 2–5 years were reproductively successful. In age classes 6–9, however, only 60% of the females weaned litters. Thus, there was an uneven distribution of successful vs. unsuccessful females (inline image = 28·85, P < 0·001). The females in the 2, 3, 4 and 5 age classes had a significantly higher proportion of success at weaning litters when compared to the yearling and older age classes (yearling: inline image = 40·2, P < 0·001; older: inline image = 24·04, P < 0·001). Annual survival did not differ significantly among age classes ( inline image = 0·835, P = 0·66).

Table 2.  Percentage of female Columbian ground squirrels by age class that weaned a litter and percentage of females that survived to the next year. Sample size for weaning is the number of females alive at the expected time of litter emergence for whom weaning status was known and sample size for survival is the number of females of known age at the time of litter emergence
Age (years)WeanSurvive
% (n)%(n)
1 49 (84)74 (84)
2 83 (54)83 (53)
3 93 (44)82 (43)
4100 (19)68 (19)
5100 (13)69 (13)
6–9 60 (15)73 (15)
All 73(229)77(227)

reproductive and somatic investments

When all (successful and unsuccessful) females were considered, somatic investment and reproductive investment differed among age classes (Fig. 1a; somatic investment: r2 = 0·27, F5,223 = 16·36, P < 0·001; reproductive investment: r2 = 0·15, F5,223 = 16·30, P < 0·001). A Duncan's multiple range test revealed that somatic investment was highest in the yearlings while litter mass was highest in the 5-year-olds. On average, yearling females had lower reproductive investment than females of prime breeding ages (viz. 2–5-year-olds). Also, on average, among the oldest females (6–9-year-olds), reproductive investment was significantly lower compared to females of prime breeding age (t120 = 2·07, P < 0·05). Among successful females, somatic investment and reproductive investment varied among age classes. Somatic investment was highest in the yearling age class (Fig. 1b; r2 = 0·17, F5,156 = 6·23, P < 0·001) while reproductive investment was highest in the 5-year-olds and lowest in the yearlings (Fig. 1; r2 = 0·15, F5,156 = 5·61, P < 0·001).

Figure 1.

Reproductive investment (black) and somatic investment (white) of female Columbian ground squirrels (a) all females (successful and unsuccessful); (b) only successful females. Data are presented as mean ±2 standard errors. Sample sizes appear above each bar. Age group 6 + represents age groups 6–9.

The effects of maternal mass at emergence from hibernation on reproductive investment and somatic investment were analysed. Using residuals of regression, we repeated the above analyses of searching for differences in reproductive investment and somatic investment among age classes with both variables statistically controlled for female spring emergence mass. Analyses were run first for all females, and then using only successful females. We found that controlling for spring emergence mass eliminated the significant differences among age classes when all females were considered (somatic investment: r2 = 0·02, F5,223 = 1·09, P = 0·37; reproductive investment: r2 = 0·06, F5,223 = 2·86, P = 0·02). The value for litter mass was considered biologically insignificant because age explained only 6% of the variation in litter mass. Controlling for spring emergence mass eliminated significant differences among age classes when only successful females were considered (somatic investment: r2 = 0·03, F5,156 = 1·08, P = 0·37; reproductive investment: r2 = 0·02, F5,156 = 0·86, P = 0·51). Therefore, spring emergence mass explained 14% of the variation seen in somatic investment among age classes and 12% of the variation seen in reproductive investment among age classes. Additionally, among successful yearling and older females, mass at emergence from hibernation was significantly negatively correlated with somatic investment (yearlings: r = −0·74, n = 41, P < 0·001; older: r = −0·51, n = 121, P < 0·001). Among older successful females, reproductive investment was significantly positively correlated with spring emergence mass (older: r = 0·29, n = 121, P < 0·001). However, among successful yearling females, reproductive investment was not significantly correlated with spring emergence mass (yearlings: r = 0·06, n = 41, P = 0·71).

Among yearling Columbian ground squirrels, female mass at emergence from hibernation differed between successful females (mean = 252 g ± 6·2) and unsuccessful females (mean = 229 g ± 5·6) (t-test: t82 = 2·67, P = 0·009). Among older females, there was no difference in mass at emergence from hibernation for successful (mean = 330 ± 3·6) and unsuccessful females (mean = 322 ± 8·4) (t-test: t143 = 0·80, P = 0·43). Somatic investment was compared using t-tests. Somatic investment of successful and unsuccessful females differed significantly only for yearlings (Fig. 2; t82 = 2·42, P = 0·02). However, when this analysis was repeated, controlling for mass at emergence from hibernation, successful and unsuccessful yearling females did not differ with respect to somatic investment (t82 = 0·73, P = 0·50). Among females older than yearlings there was no significant difference in somatic investment (Fig. 2; t143 = 0·25, P = 0·80).

Figure 2.

Somatic investment of successful (black) and unsuccessful (white) female Columbian ground squirrels. Data are presented as mean ±2 standard errors. Sample sizes appear above each bar. Age group 6 + represents age groups 6–9.

Discussion

We used our estimates of reproductive and somatic investments to evaluate three hypotheses that have been proposed to explain age specific patterns of resource investment during reproduction of female Columbian ground squirrels. The residual reproductive value hypothesis predicted that reproductive investment should increase with age, while somatic investment should decrease with age. The lower values of litter mass and litter size observed in the yearling age class seem consistent with the residual reproductive value hypothesis (Pianka & Parker 1975). However, the lack of increased reproductive investment by the oldest females allows us to reject the residual reproductive value hypothesis as an explanation of reproductive patterns seen in our study. Also, investment in body mass during reproduction did not decrease with age for reproductive females older than yearlings as the residual reproductive value hypothesis predicted.

The evolutionary restraint hypothesis states that reproductive investment of the youngest females is restrained by a cost to future reproduction and/or survival (Williams 1966; Curio 1983; Marchetti & Price 1989; Rohwer 1992; Hepp & Kennamer 1993; Blums et al. 1996). Consistent with this hypothesis was a lower proportion of yearling females producing a litter as compared to older females and higher somatic investments of yearling females compared to older females. In our study, a yearling females’ ability to successfully wean a litter appeared limited by low mass at emergence from hibernation. Therefore, a high proportion of these individuals might have refrained from attempting to raise a litter compared to older adult females. This appeared to be due to yearling females completing growth, as unsuccessful yearling females weighed significantly less at spring emergence than successful yearling females. Additionally, unsuccessful yearling females also gained significantly more mass during the reproductive period than successful females.

Evolutionary restraint provides an adaptive alternative to the hypothesis of developmental constraint. If growth to full adult size was not possible during the first year of life, then lowered reproductive investment might be constrained by the need for structural growth. Ground squirrels have deterministic structural growth (Dobson 1992; Dobson & Michener 1995), and thus restraint and constraint both seem reasonable explanations for lower reproductive investment in younger female Columbian ground squirrels. However, both hypotheses make similar qualitative predictions concerning patterns of reproductive and somatic investments, as well as the influence of spring emergence body mass. Both restraint and constraint predicted that lighter, primiparous female Columbian ground squirrels would forego or limit their first reproduction to complete growth and invest resources into themselves. Thus, we could not distinguish between these two possibilities.

We examined the senescence hypothesis (Williams 1957; Kirkwood & Rose 1984) by examining patterns of reproductive and somatic investment over the lifetimes of females, by quantifying the proportions of successful vs. unsuccessful females in each age class (Berube, Festa-Bianchet & Jorgenson 1999), and by examining survival of females by age class. When all females (successful and unsuccessful) were considered, reproductive investment declined significantly in the oldest females, because the oldest females (aged 6–9 years) contained the largest proportion of unsuccessful individuals among females older than yearlings. Senescence of reproduction in the some of the older females is similar to senescence of reproduction in older bighorn ewes (Berube et al. 1999), where the oldest age class of ewes had a relatively lower reproductive output compared to other adult age classes. Examinations of somatic investment between successful and unsuccessful older adults show that, unlike unsuccessful yearlings, the oldest (aged 6–9 years) unsuccessful females were not able to increase somatic investment. Millar (1994) and Morris (1996) found evidence of senescence in Peromyscus maniculatus and P. leucopus, respectively, where declining body condition of older females resulted in decreased reproductive success. In our study, senescence was evident primarily when examining the proportion of successful females in the oldest age group. Studies that investigate reduced reproductive investment, measured by decreased litter size and/or litter mass, may fail to detect senescence, as would studies that attempt to demonstrate senescence from a lower survival rate for older age classes (Slade 1995). Some of the oldest females showed senescence by the failure to wean a litter, but other older females showed similar reproductive and somatic investments as compared to younger females. Some older females exhibited reproductive senescence, while others did not.

Identification of senescent declines in reproductive and somatic investments, and survival, are difficult to identify. Although declines in reproduction and survival have generally been interpreted as evidence for senescence, Blarer, Dobeli & Stearns (1995) demonstrated that such patterns may be produced by optimal life histories, without physiological deterioration of individuals at advancing age. In addition, interspecific studies of senescence have assumed that evidence of senescence should begin at or near the age when individuals first reproduce (e.g. Promislow 1991; Gaillard et al. 1994). Columbian ground squirrels did not exhibit patterns of investments and survival that would indicate agreement with the above works, as we discuss below. Thus, we feel that the most reasonable conclusion is that our population actually exhibited senescent declines in reproductive investment by adult females at relatively old ages.

Blarer et al. (1995) showed that if survival declined due to continued growth throughout life and size-dependent mortality, then decreasing fecundity and survival could be produced by an optimal life history, in the absence of senescence. Columbian ground squirrels did not fit this scenario. First, structural growth, although continuing throughout life, occurs at an extremely low rate, around 1% per year, so that growth to adult size appears deterministic (Dobson 1992; Dobson, Risch, & Murie 1999). In our study, spring body mass, which probably reflects stored resources that females have at the start of the year, did not increase significantly in the oldest females, but instead exhibited a non-significant decline. Finally, the decline in mean reproductive investment through failure to produce litters occurred only in the oldest age class of adult females, rather than gradually as trade-off models generally predict. Thus, the basic conditions for applying Blarer et al.'s (1995) model to Columbian ground squirrels were not evident.

The lack of consistent directional changes in reproductive and somatic variables was also inconsistent with the way that senescence has been identified in interspecific comparisons. Promislow (1991) and Gaillard et al. (1994), for example, used the rate at which survival declines as evidence of how rapidly senescence occurs. In our study, survival was slightly higher in 2- and 3-year-old females, but variations among age classes were slight and not significant. Changes in somatic investment during the reproductive period were also essentially constant among females older than yearlings. The proportion of females that produced litters successfully was extremely high through 5 years of age, and then fell dramatically. Thus, the abrupt decline in mean reproductive investment among the oldest females appeared out of context with the rest of the age-structured life history, even taking trade-offs into account. Senescence appears to provide the best explanation for the reproductive decline.

Female mass at emergence from hibernation affected patterns of age-specific reproductive and somatic investment. Previous work has shown that maternal mass can influence litter size, average mass of offspring, or both (McClure 1981; Dobson & Myers 1989; Michener 1989; Hoogland 1995; Derocher & Stirling 1998; Festa-Bianchet & Jorgensen 1998; Millesi et al. 1999). Older adult females that emerged from hibernation at a heavy mass were able to invest more resources into reproduction while investing less into themselves (see also Dobson et al. 1999). Yearling females that emerged at a relatively high mass were able to wean litters, while those that emerged at a lighter mass completed growth instead of investing resources into producing a litter.

Reduced reproductive output of primiparous (first-time breeding) females (smaller litters and/or reduced proportion of individuals reproducing successfully) compared to multiparous females, as observed in this study, is consistent with other studies of vertebrates (e.g. Festa-Bianchet & King 1991; Dobson & Michener 1995; Risch, Dobson & Murie. 1995; Becker, Boutin & Larsen 1998; Huber et al. 1999; Millesi et al. 1999). The cessation of reproduction by older females has also been shown in other vertebrate species (e.g. Promislow 1991; Derocher & Stirling 1994; Gaillard et al. 1994; Packer, Tater & Collins 1998). The oldest females of these species show either reduced reproductive output or stop reproducing altogether. In our study, older females either reproduced at a ‘normal’ level or were more likely to fail completely. On average, the oldest female Columbian ground squirrels showed a significant decline in reproductive output compared to younger females because they were less likely to wean a litter compared to middle-aged females.

In conclusion, we found that some young female Columbian ground squirrels were limited by low spring emergence masses and therefore were constrained or restrained in their energetic investment by delaying production of litters, while completing growth during their first full year after birth. We also found that the oldest female age classes exhibited evidence of senescence of reproduction, and were the least likely to give birth to litters among adults older than yearlings. We concluded that evolutionary restraint or constraint best explained patterns of reproduction seen in the yearling females, while the senescence hypothesis best explained reproductive patterns seen in the oldest females. Due to the lack of increased reproductive investment by the older females, we found no support for the residual reproductive value hypothesis.

Acknowledgements

For assistance with fieldwork, we thank M.J. Badry, C. Cassady-St Clair, I. Cote, C. DeVries, H. Dundas, J. Hare, S. Hatfield, J. Hines, W.J. King, A. Knight, R. Lewis, B. MacWhirter, M. Ramsay, J. Rieger, J. Waterman, J.D. Wigginton and P. Young. We appreciate access to the study area granted by personnel of the Turnbull National Wildlife Refuge and thank R. Carr and D. Nichols for use of facilities at the Turnbull Laboratory for Ecological Research. A. Badyaev, S. Boback, C. Guyer, G.R. Hepp, D. Morris, P. Neuhaus, C. Osenberg and J. Styrsky made very helpful comments on the manuscript. Fieldwork was funded by a Natural Sciences and Engineering Research Council of Canada grant to J.O.M. Data analysis and manuscript preparation was funded by a National Science Foundation grant to F.S.D. (no. DEB-0089473).

Ancillary