Delayed maturation in female bank voles: optimal decision or social constraint?


  • Anne-Caroline Prévot-Julliard,

    1. Division of Zoology, Department of Biology, University of Oslo, PO Box 1050 Blindern, N-0316 Oslo;
    2. Centre for Advanced Study, The Norwegian Academy of Science and Letters, Drammensveien 78, N-0271 Oslo, Norway;
    Search for more papers by this author
  • Heikki Henttonen,

    1. Centre for Advanced Study, The Norwegian Academy of Science and Letters, Drammensveien 78, N-0271 Oslo, Norway;
    2. Finnish Forest Research Institute, PO Box 18, FIN-01301 Vantaa, Finland; and
    Search for more papers by this author
  • Nigel G. Yoccoz,

    1. Centre for Advanced Study, The Norwegian Academy of Science and Letters, Drammensveien 78, N-0271 Oslo, Norway;
    2. Division of Arctic Ecology, Norwegian Institute for Nature Research, Polar Environmental Centre, N-9005 Tromsø, Norway
    Search for more papers by this author
  • Nils ChR. Stenseth

    1. Division of Zoology, Department of Biology, University of Oslo, PO Box 1050 Blindern, N-0316 Oslo;
    2. Centre for Advanced Study, The Norwegian Academy of Science and Letters, Drammensveien 78, N-0271 Oslo, Norway;
    Search for more papers by this author

Dr N. C. Stenseth, Division of Zoology, Department of Biology, University of Oslo, P.O. Box 1050 Blindern, N-0316 Oslo, Norway. Tel: +47-22854584/22854561. Fax: +47-22854605 (attn Stenseth). E-mail:


1. Density, maturation and survival of female bank vole (Clethrionomys glareolus) in the northern taiga of Finnish Lapland were studied using long-term capture–mark–recapture data from two large grids, one food-addition grid and one control grid, in 1982–94.

2. The density on the food grid was consistently higher than the density on the control grid.

3. Females born early in the breeding season usually matured, except at very high densities. Those born later in the summer season commonly delayed maturation to the following spring.

4. Winter survival of sub-adult (having delayed maturation) females was significantly higher than survival of adult (breeding) females. However, empirical values of sub-adult and adult survival, as well as difference between them, were not consistent with survival values assumed in theoretical models on optimal deferred breeding.

5. There was a density-dependent relationship between the maturation rate of young voles and the density of already established breeding females (both bank voles and all Clethrionomys together; C. rutilus and C. rufocanus occasionally occurred on the study grids). This density dependence was different for the two grids (weaker on the food-addition grid).

6. These findings are discussed within an evolutionary context: we have, on the basis of these findings, no evidence suggesting that the observed delayed maturation represents an evolutionary optimal strategy. Rather, there is evidence suggesting that the delay is due to social constraints.


Population fluctuations of voles and lemmings of the northern highly seasonal regions have interested ecologists ever since Elton (1924). Relatively little attention has, however, been devoted to the empirical study of life history tactics within fluctuating vole populations. On the other hand, this has been the subject of much theoretical work (see, for example, Schaffer & Tamarin 1973; Stenseth 1978; Stenseth & Framstad 1980; Stenseth & Ugland 1985; Stenseth et al. 1985; Ugland & Stenseth 1985; Gyllenberg, Hanski & Lindström 1996; Kokko & Ranta 1996). Among life history studies on periodically fluctuating rodents, some attention has been given to the study of the variation in the proportion of females reproducing and their age at first reproduction, as well as to how these traits are influenced by density and food supply (see, for example, Kalela 1957; Kalela 1962; Hansson & Henttonen 1985; Desy, Batzli & Liu 1990). Young individuals of periodically fluctuating rodents may either mature and reproduce in their year of birth or delay maturation until the following year (in which case they remain as non-breeding individuals during their year of birth, belonging to the functional category commonly named ‘sub-adults’; Kalela 1957; Myllymäki 1977a,b).

Two kinds of evolutionary scenario have been suggested to explain delayed maturation of young (cf. Stearns 1992). The first is based on optimal individual decision: if reproduction is costly (e.g. by reducing survival), then it may be better, evolutionarily speaking, to delay maturation so as to experience a higher survival during winter. In fluctuating rodent populations, delaying reproduction may be advantageous if the delay in reproduction is compensated for by increased survival (e.g. Mappes & Ylönen 1997). Differences in survival of sub-adults and adults may depend on the phase of the rodent fluctuations, with larger differences occurring at peak and decline phase when predation pressure is highest (Ylönen 1994). This scenario has, for fluctuating rodents, been studied through modelling (e.g. Kokko & Ranta 1996; Kaitala, Mappes & Ylönen 1997). The model by Kaitala et al. (1997) predicts optimal delayed reproduction only for a restricted set of survival values for both sub-adults and adults; however, this model did not rely on accurate empirical estimates.

The second scenario rests on social constraints relating to acquisition of breeding territories. For the territorial sex, acquisition of a territory is necessary before reproduction can be initiated. Delayed maturation may thus reflect the lack of vacant territories (e.g. territories may all be occupied by established breeders). Alternatively, there may be too many non-reproducing individuals attempting to acquire a territory with sufficient resources. Delayed maturation is then assumed to result from insufficient food and space. Social interactions among individuals may occur only within a given species or between species belonging to the same community.

In the taiga forests of our study area, the bank vole [Clethrionomys glareolus (Schreber 1780)] is the numerically dominant species within the Clethrionomys community. However, red voles (C. rutilus) and grey-sided voles (C. rufocanus) are commonly present (Henttonen et al. 1977; Henttonen et al. 1987; Henttonen & Hanski 1999). All Clethrionomys species seem to have a similar social organization, with breeding females defending territories (Viitala & Hoffmeyer 1985).

Here, we report on large scale and long-term data on maturation and delayed breeding, with additional experimentation on food supplementation, using the bank vole in northern Finland as the study organism. By using capture–mark–recapture (CMR) methodology (Lebreton et al. 1992) we estimated survival rates of females. We also used this methodology to estimate the recruitment rate of mature females (i.e. the proportion of newly matured females at a given trapping session; Pradel et al. 1997). As is common in the literature on microtine rodents, we refer to ‘sub-adults’ as reproductively inactive individuals (which on the basis of their age could breed), and to ‘adults’ as mature and reproductively active individuals (cf. Myllymäki 1977b). Recruitment of new adult females to a focal population may result either from maturation of young females or from immigration of adult females (Nichols & Pollock 1990).

To test the potential effect of food on recruitment and survival, we monitored two large grids (2·7 and 4·5 ha for the food-addition and control grids, respectively) over a long-term period (1982–94). The large size of the grids allowed us to minimize edge effect (Stenseth & Hansson 1979; Tanaka 1980). Regional dynamics of the forest rodent community in our study area are highly synchronized (Hanski, Henttonen & Hansson 1994, their Fig. 2). We considered each of the 13 years in the period 1982–94 as the unit for comparisons, since our primary objective was to compare recruitment in different years with different densities.

Figure 2.

Temporal variation in density (measured as the number of individuals per ha, ignoring possible edge effect) of bank voles Clethrionomys glareolus (broken line), and of all rodents species taken together (full line) on the control grid (upper panel) and on the experimental food addition grid (lower panel) from 1982 to 1994. The two periods during which food was added in the experimental grid are indicated by stars (*) and plus (+), respectively. The periods during which densities of bank voles were significantly different on the two grids are represented by the shadowed areas in the bellow line. Bank voles and red voles (C. rutilus) were present on both grids during the entire study period (bank vole being the most abundant species). From 1982 to 1984, grey-sided voles (C. rufocanus) and root voles (Microtus oeconomus) were also present in significant numbers on both grids. The abundance of the different species are detailed in lower graphs for this period. All densities have been estimated by capture–mark–recapture modelling of closed populations.

We attempt to discriminate between the two proposed evolutionary causes of delayed maturation – an evolutionary optimal decision or a constraint due to environmental conditions – by evaluating the quantitative assumptions underlying each scenario.

1. If delayed maturation is optimal, the reproductive value of females having delayed maturation should be higher than the reproductive value of females which matured during the summer of birth. In particular, delaying females could improve their survival by decreasing the risk of being preyed upon (since reproductive female are assumed to be more prone to being preyed upon, i.e. predation risk) or infected by parasites (Haukisalmi, Henttonen & Tenora 1988).

2. If females delay their reproduction because of the lack of food at a territory, maturation in the birth year may be enhanced by supplementing food. If maturation is delayed as a result of competition with reproducing adult females, then the level of maturation is, furthermore, expected to be lower when the number of established females is higher. In this context, the relative importance of adult bank vole females vs. entire adult Clethrionomys-community needs to be evaluated.

Materials and methods

Study site and data collection

The study was carried out in an old taiga forest at Pallasjärvi, Finnish Lapland (68°30′N, 24°09′E). For a detailed description of the study site, see Henttonen et al. (1987). For a comprehensive discussion on rodent dynamics and community in Finnish Lapland, see Henttonen & Hanski 1999). Two grids (one control grid and one experimental grid) were simultaneously studied from 1982 to 1994 (Fig. 1). The ‘control grid’ (4·5 ha) and the ‘experimental grid’ (2·7 ha; also called the ‘food grid’) were located in the same habitat, between a road and a lake, thereby reducing the edge effect on density estimation. The grids were separated by 200 m; very few individuals moved between the two grids [from the total of 2164 females, 40 (i.e. 2%) were observed on both grids].

Figure 1.

Study site in Pallas-Ounastunturi National Park in Finnish Lapland. The habitat between the lake and tree line is old growth forest dominated by spruce Picea abies, bilberry (blueberry) Vaccinium myrtillus in the field layer, and mosses Pleurozium, Hylocomium and Dicranum in the bottom layer. C is control grid, F is food grid. Terrain is flat close to the lake, but the slope is steep closer to the tree line.

The grids used in this study were larger and the time-scale longer than is common in studies using small mammals as experimental organisms. Small grid size may significantly constrain the numbers of animals included in such experiments: most commonly only around 50–100 individuals being involved (Boutin 1990). Also, the effects of immigration and emigration become relatively more important on small grids. When reviewing available studies on food-supplementation experiments, Boutin (1990) emphasized the need for larger scale and longer lasting studies; indeed, many of the conclusions reached on the basis of small scale food-addition experiments are rather unclear (see, e.g. Stenseth & Ims 1993).

Voles were caught using live-traps (the multiple capture Ugglan type). Both trapping grids consist of trapping-points located with 10-m spacing. During the snow-free periods (from late May to early October), a trapping session included checking traps at 6-h intervals for 4 days (traps were checked 16 times in total). Traps were, at any given point in time, placed at 20-m intervals (i.e. using only every second trapping point) and were moved every 24 h. During the snowy periods, permanent trap chimneys were located at 20-m intervals; winter trapping sessions consisted of 16–20 checks, usually four per day at 2–3-h intervals between morning and evening. Such a winter schedule was adopted to prevent trap mortality during low temperatures (as low as −35°C, even though the temperature on the ground level in the chimneys generally is about −15°C when it is −30°C on the snow surface). Bank voles are clearly day-active during winter.

Starting in the early summer of 1982 (after the first trapping session), oat seeds were added at each trapping point on the experimental grid. This was continued to late autumn 1986. Food was added after each trapping period. (To check for the effect of adding food on attractiveness of traps, trapping was repeated on the food grid immediately after feeding in February 1983: less than 40% animals known to be present at that instance were caught, suggesting that the addition of food profoundly reduces the attractiveness of the traps for few days.) Food was not added between late autumn 1986 and spring 1988; trapping continued, however, as on the control grid. From early summer 1988 (after the first trapping session in a snow-free period) to late autumn 1992 both oat seeds and mouse pellets were added on the food grid. During this period, food was added regularly so as to maintain it ad libitum. From 1993 to 1994, additional food was not provided on the experimental grid; however, trapping continued during the summer according to the standard schedule on both grids until late autumn 1994. The trapping schedules for the two grids are summarized in the Appendix.

Data from 1985 were not included in our analysis due to the very low number of individuals present on the control grid that year (Fig. 2).

At our study sites at Pallasjärvi, the territory size of breeding females is about the same for all Clethrionomys species (H. Henttonen, unpublished data). Territories of breeding females are also exclusive between species (Löfgren 1995b).

Three functional categories (‘juveniles’, ‘sub-adults’ and ‘adults’, sensuMyllymäki 1977b) of bank vole females were defined according to a combination of: (i) weight and breeding condition (open or closed vagina, vaginal plug, nipples); (ii) age-specific fur development (moulting) in Clethrionomys (Zejda & Mazak 1965; Viitala 1981); and (iii) home range behaviour. Females with juvenile fur (up to 4 weeks of age), closed vagina and weighing less than 18 g (usually 7–14 g) were classified as ‘juveniles’. Females with developing or fully developed post-juvenile or later stage fur (including winter fur), closed vagina, no visible nipples and weighing less than 20 g (usually 14–18 g depending on the season), were classified as ‘sub-adults’ (i.e. voles not reproducing even though their age would allow this; such individuals are the ones which experience delayed maturation). Females in post-juvenile or older fur with visible nipples were classified as ‘adults’. Some individuals were considered adults even though they had juvenile fur since their vaginas were open; such individuals weighed more than 18 g and were already sedentary within the grid. An individual remained in the adult category once it had been identified as such, even if it stopped reproducing.

Data analysis

Estimation of densities

Number of individuals present on a particular grid during a given trapping session was estimated by robust design using trap checking within that session (software capture, White et al. 1982). The densities are given by the number of individuals by a hectare (estimated numbers divided by grid area). Even though the grids were of different size, the edge effect is negligible here, since the grids were located between a road and a lake (Fig. 1). We tested the difference in bank voles densities between the two grids by comparing the 95% confidence intervals of densities provided by capture. This conservative procedure deals partly with the problem of multiple comparisons (Wainer 1996), but the strong asymmetry of confidence intervals prevented a simpler method based on using the standard errors.

Survival analysis

We estimated the survival probabilities of sub-adult and adult females on the control grid. An individual female could became an adult after being earlier recorded as a sub-adult. In that case, we split its capture history in two parts. In the sub-adult part, we considered that this individual was captured as a sub-adult and not released on the occasion following its last capture as a sub-adult. The adult part started with the first occasion this individual was seen as adult (see Julliard et al. 1999, for more considerations on this way of splitting). Adults of different ages were included in the same data set. We compared the survival probabilities of sub-adult and adult females on the control grid using the software surge 5·1 (Reboulet et al. 1999). We tested for any effects of year and month on survival of each female category. Then, we compared survival of the two categories both in snowy (October–May) and in snow-free (June–September) periods. In particular, we tested whether we could consider them as similar during the various periods. We considered capture–mark–recapture data from 1982 to 1992 for this survival analysis, because there were no more trapping sessions in winter in 1993 and 1994 (see Appendix).

Quantifying recruitment rates

The recruitment of adult females was quantified separately for each grid and year. By using recent developments of the capture–recapture methodology (Pradel 1996; Pradel et al. 1997), we estimated the proportion of new adults which became established in the population between two trapping occasions (denoted β see Pradel et al. 1997). This proportion was standardized to account for varying lengths of time intervals between trapping occasions. Therefore, we calculated βi, the proportions per day of adults which appeared in the adult component of the population between occasion i−1 and occasion i.

New adults could be individuals which spent the winter as a sub-adult (i.e. individuals which had delayed their maturation). For all years on both grids, these sub-adults matured in April–May in the year after their birth. The new adults could also be juveniles which did not delay maturation and which directly became adult. This is the group we were mostly interested in here. This category of new adults may appear at any time during the breeding season, approximately 1 month after the birth of young (Bujalska 1990). Finally, new adult females could have immigrated onto the grid. In this case, we were unsure of their prior breeding status. However, on the studied grids, reproducing females caught several times were trapped at 30-m intervals in average, suggesting that the dispersal of adult females (breeding dispersal; see Lambin 1997) is low. Hence, we assume that all new adults captured on the grids (corrected by capture probabilities), were actually breeding for the first time.

For each year, on each grid, we defined the level of recruitment as the highest value of the proportion of new adults for the year on that grid (excluding the beginning of the breeding season, when sub-adults matured to become adults). These peaks of recruitment are indicated by arrows in Fig. 3. We then modelled these recruitment levels βi,j,k (occasion i in year j, in grid k) as functions of density (i.e. number of individuals per ha) of different categories of individuals present at occasion i−1 of year j in grid k, using linear models (with S-plus 3·3, Venables & Ripley 1994): (i) all bank voles; (ii) all Clethrionomys individuals; (iii) established breeding females, either bank voles or all Clethrionomys. We also tested for a possible grid-effect. Since the recruitment levels were actually proportions, we preferentially used logit(βi,j,k). To account for the uncertainty in the estimates of βi,j,k, we weighted these estimates with var [logit (βi,j,k)], calculated as var [logiti,j,k)] = var (β,j,k)/{[βi,j,k*(l−βi,j,k)2]} (Burnham et al. 1987); var (βi,j,k) were obtained from the capture–mark–recapture estimates.

Figure 3.

Figure 3.

Recruitment patterns of adult C. glareolus females on the control and food grids from 1983 to 1994. X-axes represent the successive trapping sessions on a given year on each grid. The Y-axes represent the proportions of new adult females appearing in the adult component of the population per day during the interval between a given session and the previous session (see Methods). In 1985, there were too few females to obtain estimates of the proportion of new females on the grid.

Figure 3.

Figure 3.

Recruitment patterns of adult C. glareolus females on the control and food grids from 1983 to 1994. X-axes represent the successive trapping sessions on a given year on each grid. The Y-axes represent the proportions of new adult females appearing in the adult component of the population per day during the interval between a given session and the previous session (see Methods). In 1985, there were too few females to obtain estimates of the proportion of new females on the grid.


Variation in density of clethrionomys on the two grids

On both grids there were strong seasonal and between year variations in the density of all Clethrionomys-species from 1982 to 1994 (Fig. 2). Bank voles were clearly the most abundant species most of the time, although red voles were regularly present. From 1982 to 1984, the rodent community also included fairly high numbers of the grey-sided vole (C. rufocanus) and root (or tundra) vole (M. oeconomus), in different proportions on the two grids (Fig. 2). Bank vole densities, which were similar on both grids before the feeding experiment started, were not significantly different in 1982 (presumably because of the increase in other rodent species, Fig. 2). After this point, bank vole densities became significantly higher on the food grid until April 1987 (on average, density of bank voles was 2·1 times higher on the food than on the control grid during snow-free periods and 2·5 times higher during snowy periods). The densities reached the same level on both grids after a delay from the end of the feeding in late autumn 1986. When the feeding experiment was resumed in 1988, densities on the food grid became again significantly higher than that on the control grid. During this second part of the feeding experiment, the difference in densities was larger than during the first period (on average, density of bank voles was 3·7 times higher on the food than on the control grid during snow-free periods and 5·2 times higher during snowy periods), confirming that the food during the second period was of better quality for bank voles (pellets contain 22% protein, whereas oat seeds contain just 12–14% in Finland, Anonymous 1993). However, as soon as the experiment was terminated, the density on the food grid decreased to the same level as on the control grid (see Fig. 2). Similar densities previous to the food addition and after the termination of the food experiment suggest that there is no grid-effect on the density (as expected because the grids were located in the same habitat).

Variation in survival probabilities of sub-adult and adult females

Survival of sub-adults and adults did not show similar temporal variations during snow-free periods ([similar variations] vs. [different variations], χ217 = 48·07, P < 10−4). However, during snowy-periods, sub-adults and adults showed similar temporal variations ([similar variations] vs. [different variations], χ220 = 20·67, P = 0·42), sub-adult survival being consistently higher than adult survival ([no category effect] vs. [additive effect of category], χ21 = 4·74, P = 0·03).

Recruitment pattern on the two grids

The proportion of new adults among the category of adults are, for each year and both grids, summarized in Fig. 3. For most years, there were two peaks of appearance of new adults on a given grid: the first peak, in the beginning of the breeding season (end of May to mid-June, depending on the year), may be assumed to reflect the maturation of individuals having spent the winter as sub-adults. The second peak may be assumed to correspond essentially to the maturation of young females born earlier the same summer and maturing directly without any marked delay.

On both grids, the recruitment of adults was year-dependent. In some years there was a pronounced recruitment of new adults (e.g. 1989 and 1990 on the control grid; Fig. 3), whereas in other years, virtually no new adults appeared (e.g. 1983 on the food grid 1992 on the control grid; Fig. 3).

Recruitment of adult females and densities

We correlated the highest level of recruitment in any year (on a logit scale) with the density of: (i) all bank voles; (ii) all Clethrionomys individuals; (iii) established breeding females (either bank voles or all Clethrionomys) at the occasion preceding the peak of recruitment.

Among these available estimates of density, density of breeding Clethrionomys females gave by far the best relationship (R2 = 0·62), together with the density of already breeding bank vole females (R2 = 0·62); both total density of Clethrionomys (R2 = 0·44) and total density of bank voles (R2 = 0·38) exhibited a weaker relationship. The highest level of adult recruitment (on a logit scale) was thus correlated negatively with the density of already breeding Clethrionomys females on both grids.

The relationship between the recruitment rate and the density of breeding adults was weaker on the food grid than on the control grid (Table 1 and Fig. 4).

Table 1.  Relationship between the proportion of new adult bank vole females during the peak of adult recruitment and the density of adult Clethrionomys females already established. The values of the different effects are indicated under the model density*grid (R2 = 0·62; F3,17 = 10·52, P = 0·0004)
Standard error (SE) 0·27 0·07 0·270·07
P 0·0001 0·020·008
Figure 4.

Proportion of new adult bank vole females as a function of density of established Clethrionomys adult females (upper panel). In the lower panel this proportion is represented on a logit scale. Points and regression lines are represented for each grid (‘C’ and full line for the control grid, ‘F’ and dashed line for the food grid).

Since the Clethrionomys community primarily consisted of bank voles, we could not distinguish between the impact of adult bank vole females vs. all Clethrionomys females. However, in 1983 on the food grid, the recruitment of adult bank vole females seemed to have been prevented by the presence of adult red vole females, as well as maturing grey-sided vole females. This observation is consistent with the observed limitation of breeding opportunities of young bank vole females by grey-sided vole females in a similar habitat (Löfgren 1995a).

The proportion, β, of new adults can be approximated by the number of new adult females (nnew) divided by the total number of adult females (ntot = nnew + nad, where nad is the number of already breeding females). Two scenarios may therefore explain the observed negative relationship between β and nad:

1. The number of new adults is limited by the number of already established females. Then, the number of new adults actually decreases when the number of already established females increases.

2. The number of new adults is limited by some factors other than the number of already breeding females (e.g. competition among young females trying to acquire a breeding territory). In this case, the observed negative relationship between β and nad is achieved by nnew remaining constant as nad increases.

In order to discriminate between these two alternatives, we correlated the estimated number of established females (nad) at occasion i−1 with the number of new adult females captured at occasion i (the peak of maturation). There was no significant correlation between the density of new adult females and the density of established adult Clethrionomys females for either of the two grids (Table 2), suggesting that the recruitment of new adults was not exclusively limited by the densities of already breeding females.

Table 2.  Relationship between the density of new adult bank vole females at the peak of maturation (occasion i, corrected for the length between occasion i and occasion i−1) and the density of adult Clethrionomys females already established. The values of the different effects are indicated for the two grids separately
Treatment ValueSEt-valuePR2Fisher test
ControlIntercept 0·110·04  0·15F1,8 = 1·47
FoodIntercept 0·070·04  0·047F1,7 = 0·03
gridDensity 0·0030·006 0·590·58  


Effect of food addition on density

We compared the recruitment patterns of bank vole females on a control grid and on an experimental grid where food was regularly added. Food addition resulted in a higher density. Although several food addition experiments have been carried out previously on rodents, no consistent pattern regarding the influence of food on population dynamics and demographic parameters have been found. Most studies have explained increased density under food-addition by invoking immigration onto the food-supplemented grids (e.g. Gilbert & Krebs 1981 on C. rutilus;Löfgren, Hörnfeldt & Eklund 1996 on C. glareolus) and assuming survival to be unchanged (e.g. Saitoh 1989 on C. rufocanus;Doonan & Slade 1995 on Sigmodon hispidus). However, Banach (1986) noted a higher survival for C. glareolus when provided with supplemental food and Ford & Pitelka (1984) found a similar increase for M. californicus. Furthermore, Andrzejewski (1975) reported winter reproduction in C. glareolus resulting from food addition. In our food grid, oat seeds were not enough to induce winter breeding in C. glareolus, but when pellets were added, winter breeding occurred (H. Henttonen, unpublished data). While Gilbert & Krebs (1981) found no effect on the length of the breeding season and maturation rate of C. rutilus in response to food addition, Taitt & Krebs (1981) reported that Microtus townsendii matured earlier when food was added. For Apodemus and Peromyscus, food addition is, generally, seen to enhance breeding and increase reproduction (e.g. Flowerdew 1972; Taitt 1981). However, published results should be compared with great caution since they involve different experimental set-ups and conditions: (i) the type of food added has generally been different; (ii) the rodent species used exhibit different spacing systems and diets; and (iii) even in the same species, like the bank vole, the different experiments have been done in different geographical regions and habitat types, where feeding habits could differ (Hansson 1995). Nevertheless, our result, showing an increase in density with food addition, is, in general, consistent with what has been found in earlier studies (e.g. Boutin 1990). The increased density on the food grid may be due to decreasing territory size or larger overlap between neighbouring females when food is added (see Ostfeld 1985; Ims 1987) or a combination thereof.

Scenario for delayed breeding

Delayed breeding of rodent females may result from at least two different non-exclusive evolutionary scenarios. Either juvenile females delay maturation as a result of an optimal individual decision (for instance, to avoid predation, Korpimäki, Norrdahl & Valkama 1994; Ylönen 1994; but see Lambin et al. 1995; Mappes, Koskela & Ylönen 1998) or they delay maturation because of the lack of opportunities to breed. In the following, we discuss both these suggestions, on the basis of our results.

Optimal decision

From a theoretical point of view, delayed breeding should occur only if the residual reproductive value is high (e.g. Norrdahl & Korpimäki 1995). In particular, individuals should ‘gain’ in survival by delaying reproduction, as compared to reproductive individuals (e.g. by reducing predation risk). Notably, winter survival of individuals delaying maturation should exceed winter survival of reproducing individuals of same summer. This was what we found (Fig. 5): during winter, survival of sub-adults was significantly higher than survival of adults. On the basis of their model, Kokko & Ranta (1996) concluded that delayed maturation would be optimal if the difference between delayed and non-delayed female survival exceeded half of the expected number of mature offspring produced. In our study, the difference between sub-adult and adult female survival in snowy periods (i.e. October–May) was on average 0·14. The presence, in the data set of adult females, of old adults who have already survived a winter, could have lowered the survival estimate of adult females, compared to survival of adult females of the same age as sub-adults. However, even with this potential bias, our result implies that, for delayed reproduction to be optimal, adult females should not produce more than 0·3 mature offspring, which is very low (e.g. Ryszkowski & Truszkowski 1970). Since we potentially over-estimated the difference between survival of sub-adult and adult females, our result is conservative.

Figure 5.

Winter survival of sub-adult (open columns) and adult (solid columns) bank vole females from 1982 to 1994. These values were estimated by capture–mark–recapture under the model [φsnow month+year+category, φsnow-free (month+year)*category, pmonth+year+gr], and standardized as survival between 1 October and 30 May.

Instead of comparing survival values of sub-adults and adults, other authors have assumed particular values of survival of both reproducing and non-reproducing individuals when building models for evaluating the optimality of deferred breeding (e.g. Kaitala et al. 1997). In their model, Kaitala and co-workers assumed that adults (i.e. individuals which have already bred) did not survive through the winter and that 50% of sub-adults survived through the winter (from late summer to early spring). They assumed these values to be relevant for bank voles. However, in our study, survival of adult females during snowy periods (October–May, that is 8 months) always remained above zero (Fig. 5). In addition, winter survival (from October to May; i.e. 8 months) of both sub-adults and adults was generally less than 0·5 (varying from 0·04 to 0·64 for the different years).

Comparing our empirical estimates of winter survival of both reproducing and non-reproducing females, we conclude that there is no evidence for suggesting delayed breeding is optimal.


We found that recruitment of new breeding females was negatively correlated with the density of already established Clethrionomys females. This result is consistent with other correlative or experimental studies, which have demonstrated that adult Clethrionomys females limited maturation of young females, at intraspecific (Bondrup-Nielsen & Ims 1986; Gilbert et al. 1986) or interspecific (Löfgren 1995a) level. Breeding limitation could result from lack of space for breeding territories.

However, even though the density of breeding females has a clear impact on recruitment, it seems not to be the only factor influencing recruitment. Indeed, contrary to the proportion of new adult females, the number of new adult females was not negatively correlated with density of already breeding females. Competition among young females themselves may be such an additional factor. Likewise, as the negative correlation of the recruitment level and the number of breeding females was weaker on the food grid than on the control grid, food is clearly having an effect. More young could possibly obtain territories at a given density-level of breeding females, possibly because food addition allowed territory size to be reduced (Bondrup-Nielsen & Karlsson 1985). Thus, we are left with the conclusion that delayed maturation of bank voles females is partly due to social constraints. We do not have any evidence supporting the hypothesis that it could result from an optimal reproductive decision. However, we cannot distinguish between different density-dependent mechanisms leading to the observed patterns. Further experimental studies are needed to establish what this mechanism is.


This work was supported by grants from the Research Council for Natural Science in Finland, the Oskar Öflund Foundation and the Finnish Cultural Foundation. Among the students who have helped in the field, Pete Kuortti and Aarre Jortikka should be acknowledged in particular. The personnel at Pallasjärvi Station created the atmosphere where working at −35°C and in 1 m of snow was plain fun. Grants from the Centre for Advanced Study, the Norwegian Academy of Science and Letters (Oslo), to N.C.S., the Norwegian Science Council, to N.C.S., the University of Oslo, to N.C.S., and the French Ministry of Foreign Office (Lavoisier Program), to A.C.P.J. made the analysis possible. Financial support from UMR CNRS 5558, at the University of Lyon1 (France, to N.G.Y.) was also appreciated. We thank Yoseph Amaha for carefully transferring part of the data to a computer-readable format. We thank E. Johannesen, R. Julliard, K. McCoy and two anonymous referees for many useful comments on earlier versions of the manuscript.

Received 2 June 1998;revisionreceived 22 August 1998


Schedule of the trapping sessions on the control and food grids from 1982 to 1994. The broken lines join the trap sessions on the two grids which were close in time.

inline image

inline image