Consequences of plant population size and density for plant–pollinator interactions and plant performance


  • Kaisa Mustajärvi,

    Corresponding author
    1. Department of Biological and Environmental Science, University of Jyväskylä, PO Box 35, FIN-40351 Jyväskylä, Finland; and
      Kaisa Mustajärvi, Department of Biological and Environmental Science, University of Jyväskylä, PO Box 35, FIN-40351 Jyväskylä, Finland (
    Search for more papers by this author
  • Pirkko Siikamäki,

    Search for more papers by this author
    • *

      Present address: Oulanka Biological Station, University of Oulu, Liikasenvaarantie 134, FIN-93999, Kuusamo, Finland.

  • Saara Rytkönen,

    1. Department of Biological and Environmental Science, University of Jyväskylä, PO Box 35, FIN-40351 Jyväskylä, Finland; and
    Search for more papers by this author
  • Antti Lammi

    1. Department of Biological and Environmental Science, University of Jyväskylä, PO Box 35, FIN-40351 Jyväskylä, Finland; and
    2. South-west Finland Regional Environment Centre, PO Box 47, 20801 Turku, Finland
    Search for more papers by this author

Kaisa Mustajärvi, Department of Biological and Environmental Science, University of Jyväskylä, PO Box 35, FIN-40351 Jyväskylä, Finland (


  • 1 Habitat fragmentation and the resulting decline in the local abundance of plant species can affect biological interactions. We examined the effects of abundance on plant–pollinator interactions by observing the pollinator service and subsequent reproductive output of a mostly outbreeding, but self-compatible, plant, Lychnis viscaria, in experimental populations of different sizes (number of individuals) and densities (distance between individuals).
  • 2 Bumblebees, the main pollinators of L. viscaria, preferred larger populations, but visitation rates were higher in sparser populations. Pollinators were attracted to the larger inflorescences in sparse populations, which were also more visible due to their larger area for a given size.
  • 3 Bumblebees probed more flowers within plants in sparse populations, probably due to the larger inflorescences and longer flight distances between individuals.
  • 4 Subsequent reproductive success (capsule production) was higher in sparse populations, due to differences in pollination success and resource competition, and their interaction. In self-compatible species, such as L. viscaria, reproductive success may be determined more by resource availability, whereas self-incompatible plants may be more sensitive to changes in pollinator abundance.
  • 5 We conclude that plant–pollinator interactions are sensitive to changes in both the size and spatial arrangement of plant populations, which can affect their demography and genetics. In this study, species density had a greater effect than size and the unexpectedly beneficial effects of low density may be due to greater resource availability.


Local abundance of species, which can be represented by either the distance between neighbouring individuals (density) or the number of individuals (population size) (Kunin 1997) tends to decrease when human activities lead to habitat fragmentation and destruction. Changes in both size and spatial structure may have profound effects on ecological interactions and population dynamics.

Stochastic processes will, for example, have more influence on the dynamics and genetic composition of small populations (e.g. Lande 1988; Barrett & Kohn 1991; Ellstrand & Elam 1993; Schemske et al. 1994), while important ecological interactions, in particular the mutualistic relationship between plants and pollinators, are likely to be disrupted in small and isolated populations (McKey 1989; Rathcke & Jules 1993; Aizen & Feinsinger 1994a).

Larger populations of plants are likely to be more attractive to pollinators resulting in higher visitation rates and therefore pollination success (Sih & Baltus 1987; Ågren 1996), whereas small populations may suffer from insufficient pollen transfer and consequently lower seed set (e.g. Jennersten 1988a; Lamont et al. 1993; Ågren 1996; Fischer & Matthies 1998). In addition, the level of inbreeding may be higher in small, isolated populations (e.g. Barrett & Kohn 1991; Falconer & Mackay 1996) because of the higher rate of selfing and more frequent matings between close relatives. The resulting inbreeding depression can reduce the fitness of these plants compared with those in larger populations (Menges 1991; Aizen & Feinsinger 1994a; Ouborg & Van Treuren 1994; Heschel & Paige 1995).

The spacing of individuals within a population may also affect visitation rates and the foraging behaviour of pollinators and consequently, the level of out-crossing and inbreeding (reviewed in Handel 1983). Previous studies indicate that higher plant density is associated with higher visitation rates of both hummingbirds (Feinsinger et al. 1991) and insect pollinators (Kunin 1993), and this is likely to have a greater impact for self-incompatible species which cannot compensate for lower pollinator abundance by selfing. Moreover, plant density probably affects the behaviour of pollinators which are more likely to move between individuals in dense populations, where flight distances are shorter, than in sparse populations where increased visits within a plant may favour within-plant pollen transfer (geitonogamy, reviewed in de Jong et al. 1993).

Both the size and the density of a population are known to affect pollination and subsequent reproductive performance (e.g. Sih & Baltus 1987; Feinsinger et al. 1991; Kunin 1993, 1997; Bosch & Waser 1999), but strong correlations between these two factors (Ågren 1996) mean that experimental manipulations are needed to separate their effects. Unlike earlier studies of population density (e.g. Kunin 1997; Bosch & Waser 1999), we allowed population area to vary as well as size and density of populations, because the area occupied by a plant population of a particular size will increase when the distance between individuals increases. Specifically, we address the following questions:

(1) Does the pollinator community and the rate at which it visits a plant population dependent on size and density?

(2) Do the size and density of a plant population affect the behaviour of its pollinators? and

(3) Do the size (number of individuals) and density (distance between individuals) of a population affect the reproductive success of its component plants?


The sticky catchfly, Lychnis viscaria L. (Caryophyllaceae) is an animal-pollinated, self-compatible perennial herb, occurring in sunny open habitats such as dry meadows and rocky cliffs. The plant occurs throughout northern and central Europe in fairly distinct patches from a few to several thousand individuals. Seed was collected from a population of c. 150 individuals, on a sandy roadside at Korpilahti, central Finland (62°15′ N, 25°30′ E, for details see Siikamäki & Lammi 1998; Lammi et al. 1999). The distances between neighbouring L. viscaria plants in 11 natural populations in Central and Southern Finland range from 0 to 200 cm (personal observations).

L. viscaria overwinters as a green rosette which usually produces between 1 and 40 flowering stems (for this population in field conditions, mean = 28 stems), each bearing about 25 flowers. The flowers are protandrous; the anthers have dehisced by the time the stigma becomes receptive and are pollinated mainly by bumblebees with some lepidopteran species (Jennersten 1988b; Wilson et al. 1995).

Establishment of the artificial populations

Seed collected during summer 1995 was sown in 4 × 4 cm pots (five seeds per pot, thinned to one seedling after emergence). Seedlings were transferred to the field in May 1996 after 2 months growth under glasshouse conditions. The Laukaa Research and Elite Plant Station (62°15′ N, 25°30′ E) comprised c. 100 ha of open, cultivated and fertilized field and we used an uncultivated 5 m wide ‘buffer zone’, situated between cultivated fields and a ditch running along a private management road. Patches of L. viscaria, were established on the relatively homogenous (and favourable) soil (mean ± SE; pH = 5.69 ± 0.21, P = 14.00 ± 3.66 mg L−1, K = 92.86 ± 8.75 mg L−1). The remainder of the buffer zone was mowed throughout the growing season, to reduce interspecific competition for light.

Two levels of size and density were used and three replicates of each population type; small dense (total area 0.84 m2), small sparse (4.64 m2), large dense (6.40 m2) and large sparse (54.88 m2) were randomly assigned to sites within the buffer zone, separated by at least 200 m (range 200–250 m). Unfortunately, one of the large dense populations was later destroyed, leaving only two replicates of this treatment.

A layer of sand and a plastic film was spread over the soil to prevent interspecific competition and seedlings were then planted in four rows of 25 (large populations) or in two rows of five (small populations) with individuals separated in both directions by 80 cm (sparse) or 20 cm (dense). The population size and distance between plants were within the range found in natural populations (personal observations).

The surrounding vegetation consisted of cultivated Phleum pratense, Trifolium hybridum, Hordeum vulgare, Brassica rapa, as well as naturally occurring meadow species but the closest natural population of L. viscaria is 15 km away. Its relatively early flowering period meant that there were few other food plants available for nectar feeding insects.

Observations of pollinator service

We studied overall pollinator visitation rates and diversity of pollinator fauna for groups of 10 plants. All the individuals in small populations were included, whereas a line of five consecutive plants was selected at random along one long edge of each large population, and these together with the adjacent five from the next row in were used for all observations. During each 15-min observation period we recorded all flower-visiting insects and classified them as bumblebees (Bombus spp.), honey bees (Apis mellifera), nectar robbing bees, syrphid flies (Syrphidae) or others. Visitation rates were calculated for all pollinators and for bumblebees alone.

As the diversity of visitors in plant populations has been reported to decrease with decreasing plant population size, we also calculated the Shannon-Wiener diversity index (H) for each patch at each observation period.

inline image

where Pi is the proportion of species i in a sample. This index is the most widely used measure of species diversity and is thus probably the most useful for comparisons.

Bumblebees are probably the main pollinators of L. viscaria (Kwak & Jennersten 1986; Jennersten 1988b) and we therefore followed the movements of individual bumblebees within populations. Observations were made at various times to control for fluctuations in visitation patterns. One bumblebee was selected at the start of the observation period and was followed until it left the Lychnis population. The next pollinator to arrive was followed for the duration of its visit, and so on until the end of the 15-min period. If the visits started within the observation period it was still followed until the pollinator left even if the 15 min had elapsed (which was rare). We recorded the number of individual plants and the number of flowers per plant visited and the total duration of foraging by each pollinator.

Both types of observation were carried out at least once a day in each population for 8 days between 18th June, when flowering started, and 1st July, when most of the flowers had started to produce capsules. Observations were made between 09.00 and 16.00, when the relatively sunny, clear and calm conditions favoured pollinator activity. Several populations were observed at the same time and the order in which they were studied was randomised daily to minimize any bias due to daily patterns and differences in conditions.

Reproductive success of the plants

Reproductive success was determined as the percentage of flowers that developed into capsules, mean seed weight, number of seeds set per capsule and total seed production per plant. The two longest flowering stems on each of the 10 individuals in the observed patches were examined after the flowering season to determine the final flower number and number of developed capsules as a percentage of total flowers produced (capsules plus dehisced flowers). Five mature unopened capsules were collected from each of the observed plants and weighed. A random subset of 20 seeds from each plant was weighed to the nearest 0.001 mg. Seed set per capsule was calculated by dividing the average mass of five capsules from a plant by the average mass of the seeds they contained. Total seed yield was estimated by multiplying the number of seed capsules on the two longest flowering stems by the mean capsule mass.

Statistical analysis

Pollinator data were analysed using observation periods in which all visits to a patch or single pollinator visits (second type of observations) were recorded as data points. Reproductive data were analysed using plants as data points within populations.

Data were analysed using anova models in which populations were nested within size and density treatments. Size and density were used as fixed factors and population as a random factor (Zar 1999), with degrees of freedom assigned as in Table 1. The degrees of freedom were adjusted for unbalanced design by SPSS (version 8.0), which uses Satterthwaite’s method. Interactions between treatments whose effect was clearly non-significant (P > 0.300) in the full model, were subsequently excluded to increase the power of the test, as was population when its effect had P > 0.300. When the raw data were not normally distributed, the data were transformed (either log or square root) to meet the assumptions for parametric tests. Flower number was used as a covariate in the analysis of reproductive success, to separate any density effects that were not related to inflorescence size.

Table 1.  Design of anova models. The random factor (population, c) is nested within fixed factors (size, a and density, b) (Zar 1999). However, all the designs are unbalanced and the final degrees of freedom are therefore adjusted using Satterthwaite’s method
Source of variationd.f.FNumerator d.f.Denominator d.f.
Sizea − 1MSSize/MSPopulationdfSizedfPopulation
Densityb − 1MSDensity/MSPopulationdfDensitydfPopulation
Size*Density(a − 1)(b − 1)MSSize*Density/MSPopulationdfSize*DensitydfPopulation
PopulationAb (c − 1)MSPopulation/MSErrordfPopulationdfError
Errorabc (n − 1)   


Pollinator service

The artificial populations were visited by total of 1726 insects during 142 observation periods (12–14 periods per patch). Of these, 590 (34.2%) were syrphid flies, 380 (22.0%) bumblebees, 104 (6.0%) honey bees, 27 (1.6%) nectar robbing bees and 625 (36.6%) were miscellaneous insects, mostly dipteran (flies) but with occasional lepidopteran and hymenopteran.

The diversity of the insects visiting a patch of 10 individuals during a 15-min observation period (Fig. 1a) tended to be higher in large populations (Shannon-Wiener index, nested anova: population: F8,131 = 3.32, P = 0.002; size: F1,8 = 3.67, P = 0.092; density: F1,8 = 1.505, P = 0.255). When all visiting insects were included, the sparse populations had significantly more visitors per plant than the dense populations, but the size of a population had no effect (population: F8,131 = 1.62, P = 0.124; size: F1,8 = 1.49, P = 0.256; density: F1,8 = 8.59, P = 0.019; Fig. 1b).

Figure 1.

Mean (+ SE) Shannon-Wiener diversity indexes (a), total number per plant (b), bumblebee number per plant (c) for visitors to in different types of populations.

The analysis was then repeated including only the main pollinators, bumblebees. Plants were visited more frequently by bumblebees in the larger populations than in the small ones, and more bumblebees per plant were observed in sparse than in dense populations (nested anova: population, F1,131 = 2.14, P = 0.036; size: F1,8 = 16.807, P = 0.003; density: F1,8 = 9.706, P = 0.014; Fig. 1c). Pollen-consuming flies, which probably contribute little to pollination and whose behaviour is difficult to interpret (Jennersten 1988b; our observations), showed no preference for different population sizes and densities (population, F8,125 = 2.70, P = 0.009; size: F1,8 = 0.847, P = 0.384; density: F1,8 = 0.084, P = 0.779).

Only bumblebees, the main pollinators of L. viscaria, were followed for the analysis of behaviour. A total of 186 visits was recorded (mean per population = 17 ± 6). The duration of bumblebee visits did not differ between population types (Table 2, Fig. 2a). There was, however, a tendency for the bumblebees to visit more individuals in larger and denser populations (Table 2, Fig. 2b) but to probe more flowers within an inflorescence when plant density was low, irrespective of population size (Table 2, Fig. 2c).

Table 2.  Results ( anovas) on the behaviour of the bumblebees in artificial plant populations (square root transformed data). The interaction term had P > 0.300 in all analyses and was therefore excluded from the model. The degrees of freedom are adjusted for the unbalanced design
 Visitation timeVisited plantsFlowers/plant
  • **

    P < 0.01,

  • *

    P < 0.05,

  • P < 0.1.

Size1, 102.891, 103.781, 151.74
Density1, 120.091, 113.611, 2211.61**
Population8, 1751.898, 1752.32*8, 1290.68
Figure 2.

Mean (+ SE) duration of visits (a), number of plants visited (b), and number of flowers probed per plant (c) by bumblebees foraging in different types of populations.

Some of the variation in pollinator behaviour may, however, have been due to the significant difference in the flower number between different density treatments (population: F8,94 = 2.92, P = 0.006; size: F1,8 = 0.06, P = 0.818; density: F1,8 = 13.73, P = 0.006; Fig. 3). Unfortunately, we did not have the required data to add flower number as a covariate in the analysis and thus separate the effects of resource availability. However, the behaviour observations allowed us to calculate how many times a pollinator visited particular plant and to correlate this estimate with flower number, albeit only for small populations. In sparse populations, bumblebees probed more flowers within a plant when they visited larger inflorescences (rs = 0.500, P = 0.009) and tended to visit plants with such inflorescences more frequently (rs = 0.658, P = 0.054), but no correlations were found in dense populations (visitation rate: rs = − 0.073, P = 0.841; number of visited flowers: rs = 0.289, P = 0.129).

Figure 3.

Mean flower number of the two longest flowering stems of the plants in different treatments.

The contribution of inflorescence size to the effects of density treatment on pollinator visitation rates and number of visited flowers, was analysed further (Fig. 4a,b) via one-way anova of the population means for these variables with population size as factor followed by a one-way ancova of the residuals with density as factor and flower number as covariate. Large populations had significantly higher mean visitation rates ( anova, F1,9 = 9.959, P = 0.012), but population density had no effect when flower number was added as a covariate ( ancova, density: F1,8 = 0.787, P = 0.148; flower number F1,8 = 2.563, P = 0.148). Neither population size ( anova, F1,8 = 0.332, P = 0.579) nor density had an effect on the number of flowers probed per plant ( ancova, density: F1,8 = 1.003, P = 0.615; flower number F1,8 = 0.615, P = 0.455).

Figure 4.

Population means for flower number plotted against mean visitation rates (a), and mean number of probed flowers within an inflorescence (b) for the surviving 11 artificial populations.

Reproductive success

As plants in the sparse population had larger inflorescences, we used flower number as a covariate to separate those effects due to inflorescence size from those related to other aspects of density.

Seed production per capsule did not differ significantly with population size, but was higher in dense populations (Table 3, Fig. 5a). Plants in smaller and sparser populations produced significantly heavier seeds (Table 3, Fig. 5b). The percentage of flowers setting capsules was significantly higher in sparse populations (Table 3, Fig. 5c) and covaried significantly with flower number. Population size had no effect on capsule production. The larger inflorescences in sparse populations led to higher total seed production (Fig. 5d), but neither density nor size affected total seed yield (Table 3) after the effect of flower number was removed.

Table 3.  The results of the statistical analysis ( anova model with populations nested within random factors size and density (Zar 1999) on plant reproductive success)
 No. of seeds/capsuleSeed weight% of developed fruitsTotal seed production
  • ***

    P < 0.001,

  • **

    P < 0.01,

  • *

    P < 0.05,

  • 0.05 < P < 0.1.

Flower no.1, 842.2911, 936.79*1, 934.41*1, 8438.41***
Size1, 80.401, 80.901, 71.101, 80.60
Density1, 154.411, 125.31*1, 119.62**1, 121.64
Population8, 841.458, 931.767, 933.538, 832.67*
Size*Density1, 71.50
Figure 5.

Reproductive success of Lychnis viscaria individuals (mean + SE) in the artificial populations. (a) Number of seeds produced per capsule, (b) mean seed weight (mg), (c) percentage of developed capsules of the total flower production, and (d) total seed production (mean mass of capsules × number of developed capsules in the two longest flowering stems).


Pollinator service

Both the size and density of the plant population affected plant–pollinator interactions. As expected, the main pollinators, bumblebees, made more visits to larger populations but, in contrast to earlier studies (Feinsinger et al. 1991; Kunin 1997; but see Bosch & Waser 1999), the number of visits per plant was higher in sparse population.

Pollinators may have been attracted to the sparse populations by their larger inflorescences (as shown by Klinkhamer et al. 1989; Klinkhamer & de Jong 1990; and references therein). Pollinators have also been reported to probe more flowers in larger inflorescences (Klinkhamer et al. 1989; Klinkhamer & de Jong 1990; Ohashi & Yahara 1998) and in sparse populations here. Population density had no effect on population mean visitation rates or number of probed flowers, when flower number was a covariate, further indicating that the larger inflorescences played an important role in attracting pollinators to sparse populations. However, the power of an anova for population means as data points with a covariate is quite low, compared with the nested model used for single observations, and it can only indicate the relative importance of different factors: separating the effects of flower size and density would require addition of the flower number as a covariate in all analyses. Under our experimental conditions low individual density may lead to more resources being available and thus to increased flower number, which is therefore more than a simple covariate. Such resource-mediated density effects may also influence pollinator behaviour in natural populations with few interspecific competitors (e.g. L. viscaria in cracks of rocky cliffs, or pioneer species). The beneficial effects of spacing do not apply, however, to situations in which the sparse spacing is in fact caused by lack of resources. Instead, many fragmented, sparse and small plant populations grow in unsuitable, deteriorating habitats and are likely to have small inflorescences. Our results also suggest, that small inflorescence size, along with size and density, is likely to be a very important factor in reducing the visitation rates to these populations.

The increased flower number could not, however, completely explain the success of the sparse populations. Optimal foraging theory (Charnov 1976) predicts that in sparse populations, pollinators will switch between plants less often (as shown, for instance by Beattie 1976) and visit more flowers on a plant when plant density is low (Heinrich 1979; Zimmerman 1981; Klinkhamer & de Jong 1990; Cresswell 1997). Here too, when the interplant distance is long, it appears to be more profitable for a pollinator to probe more flowers within an individual than to switch between plants.

Another factor may be that increasing either the distance between individuals or their total number increases the area, and thus the visibility, of a plant population. Pollinator abundance has been shown to decrease with decreasing habitat area (Jennersten 1988a; Aizen & Feinsinger 1994b). Unlike these studies, we examined the effect of area by increasing distance between a constant number of individuals. Visitation rates still increased indicating that area, as well as numbers and plant density, may influence plant–pollinator relationships in small populations. An experiment to separate the three aspects of local abundance (number, distance and area) was beyond the resources available but the results do suggest that the effects of area should be considered.

As in earlier studies (e.g. Aizen & Feinsinger 1994b), pollinator diversity increased with increasing habitat area (Fig. 1a). The Shannon-Wiener diversity index is logarithmic and therefore reacts slowly at low species levels, suggesting that the effect may have been underestimated.

Plant reproductive output

Density affected the allocation of resources to reproduction. Plants in sparse populations produced fewer, albeit heavier, seeds but their overall reproductive success, measured as total seed production (Fig. 5d), total capsules and even ratio of capsules to flowers, was higher. Population size had no effect on plant reproductive success, although plants were visited more often in large populations.

The reproductive success of a plant is determined by: (i) the availability of pollen for fertilization of ovules, and (ii) resources available for seed production, i.e. light, nutrients and water (Zimmerman & Pyke 1988). Although, population size effects are mostly mediated through pollinator visitation rates (aspect i), plant density affects both and we suggest that the resources available for reproduction are of particular importance for self-compatible L. viscaria. In this design, resource competition and pollinator service are linked because sparse populations have larger inflorescences and they cannot be separated completely. Our assumption that populations with identical density experienced the same level of resource competition is supported by similar flower numbers in small and large populations. Visitation rates differ between large and small populations, but reproductive success does not, suggesting that it was probably determined more by available resources. Pollination-related size and density effects on reproductive success appear to depend on mating system, with self-incompatible species being more susceptible than self-compatibles (such as L. viscaria) to low plant density and small population size (Feinsinger et al. 1991; Kunin 1993; but see Karoly 1992).

Density effects have previously been tested in experimental plants growing among other species (see Feinsinger et al. 1991; Kunin 1993). Pollen lost to the background and clogging of target species stigmas by heterospecific pollen may therefore have contributed to the observed density effects in earlier studies but not in our competitor-free design. L. viscaria is a weak competitor (Jennersten et al. 1988) and, in Finland, usually occurs in relatively open environments (e.g. thin soils of rocky cliffs, roadsides), where intraspecific competition is stronger than interspecific competition. Therefore, although the beneficial effect of sparse spacing may apply to some natural situations, it may not be more widely generalizable.

Although visitation, and thus pollination, rates were higher in sparse populations geitonogamy will predominate (Van Treuren et al. 1993; Karron et al. 1995). Dense populations might therefore experience some benefit from the greater cross-pollination rate, even though the total visitation rates were lower. The effect of density on allocation of resources to seeds might reflect the lower quality of self-pollen. Self-fertilized ovules may have been selectively aborted and resources available in sparse populations allocated to seed quality rather than quantity.

Implications for conservation

Sparse populations do not necessarily suffer reduced pollination or reproductive success and, if the populations consist of vigorous self-compatible plants with few interspecific competitors, increased spacing may be beneficial. This does not apply, however, to small, sparse populations growing in unsuitable habitats that usually consists of small individuals. The small size of individuals in these populations is likely to be an important factor, along with size and density, to reduce pollinator visitation rates.

L. viscaria is self-compatible and pollinator behaviour promotes geitonogamy. However, in several self-incompatible plant species the outcrossing rates observed are clearly related to pollen service (see Horovitz & Harding 1972; Karoly 1992). In Salvia pratensis the outcrossing rate depended on density rather than on population size (Van Treuren et al. 1993). Both spatial structure and population size shape plant–pollinator relationships. In self-incompatible plants, the effect is mediated by reduced pollen flow and consequent lower seed set (Kunin 1997; references therein); in our study the situation is more complex, resource availability is important, but changed pollinator behaviour may have an effect particularly in species sensitive to inbreeding.


We thank P. Aronta, I. Eriksson, M. Hovi, T. Savolainen and J. Tissari for their assistance the field. We are grateful to R. V. Alatalo, E. Mattila, and especially P. G. L. Klinkhamer whose comments significantly improved the manuscript. We are thankful to Konnevesi Research Station, and Laukaa Research and Elite Plant Station for all types of support. A special thanks to K. Nissinen for statistical advise and editor L. Haddon for revision of the language. This study was supported financially by the Academy of Finland (to PS, KM, AL and R. V. Alatalo) and Finnish Cultural Foundation (to AL) and is a part of the Finnish Biodiversity Research Programme (FIBRE) 1997/99.

Received 23 September 1999 revision accepted 6 July 2000