Slug herbivory as a limiting factor for the geographical range of Arnica montana


  • Helge Bruelheide,

    1. Albrecht-von-Haller-Institute for Plant Sciences, Department of Ecology and Ecosystem Research, Untere Karspüle 2, D-37073 Göttingen, Germany
    Search for more papers by this author
  • Ulrich Scheidel

    1. Albrecht-von-Haller-Institute for Plant Sciences, Department of Ecology and Ecosystem Research, Untere Karspüle 2, D-37073 Göttingen, Germany
    Search for more papers by this author

, H. Bruelheide (fax + 49 551 393556; e-mail:


1  We tested whether slug herbivory is a factor restricting the rare perennial Arnica montana to high elevations in the Harz mountains, Lower Saxony.

2  In one experiment we artificially increased the mollusc population density in plots containing native Arnica montana populations. Leaf loss and damage to Arnica increased significantly, whilst damage to other plant species in the same plots was unaffected by mollusc density.

3  A second experiment examined damage by molluscs of transplanted Arnica montana plants at three different altitudes. Whilst damage to Arnica was negligible at 610 m a.s.l. (where natural populations occur), molluscs removed 8% of Arnica leaf area at 385 m a.s.l. and 75% at 180 m a.s.l. At the two lower sites, protective caging of Arnica plants significantly reduced the amount of leaf tissue consumed by molluscs. The impact of mollusc herbivory on Arnica montana therefore appears to increase with decreasing altitude.

4  Despite the weak relationship between leaf damage to Arnica montana and mollusc abundance, periods of peak mollusc activity may well coincide with those phases in Arnica life history that are most sensitive to tissue removal by herbivores.

5  Our results support the hypothesis that polyphagous herbivores play a major role in limiting the distribution of preferred plant species. We discuss how selective herbivory may influence the distribution of Arnica montana populations in these grassland communities.


The overall limits to the geographical ranges of plant species are normally considered to be due to climatic factors. Models that predict range shifts as a response to anthropogenic climatic changes therefore tend to consider the direct effects of climate (Huntley 1991; Hendry 1993; Holten 1993; Huntley et al. 1995; Jäger 1995a,b), although the relationship between climatic parameters, such as temperature or precipitation, and plant performance is often far from obvious (Pigott & Huntley 1981; Schulz & Bruelheide 1999). For example, it is easy to cultivate some montane species, but not others, at lower elevations (Wocke 1940), suggesting that other factors are limiting plant ranges. At a local scale, all other organisms present at a site will influence an individual plant's behaviour. Plant ecologists tend to consider competition to be the most important biotic factor and to ignore other possible biotic interrelationships despite their known importance for plant distribution patterns (Jäger 1990). This paper concentrates on herbivory, which, although largely neglected, may be a significant factor limiting plant ranges and distribution.

Slugs and snails form an important part of the native herbivore fauna of central European meadows. Mollusc damage has been investigated from agricultural (Glen et al. 1991; Standell & Clements 1994; Frank 1998) and behavioural (Mellanby 1961; Grime & Blythe 1969; Pallant 1972) points of view, but studies focusing on botanical aspects are much rarer. Although damage to mature plants is rarely detrimental to plant populations (Ehrlén 1995), molluscs grazing on seedlings do impact plant community composition (Hanley et al. 1995a; Hulme 1996; Clear-Hill & Silvertown 1997). However, very few studies (Cates 1975; Reader 1992; Westerbergh & Nyberg 1995) have suggested explicitly that mollusc herbivory may be a primary cause of plant distribution patterns on a larger scale. In particular, it appears that the lower distribution boundaries of some montane species could be explained by altitudinal differences in the herbivore pressure (Galen 1990).

Previous field observations have indicated a high mollusc grazing pressure for several central European plant species, especially Compositae. One such species is Arnica montana L., which is found from Scandinavia to the Iberian peninsula and is generally restricted to nutrient-poor montane meadows, although it also occurs, much more rarely, in lowland bogs but does not colonize grasslands at lower elevations. In the Göttingen area (Lower Saxony, Germany) the only major populations of Arnica are in the Harz mountains (Garve 1994). Changes in land use over the last 50 years, either in the form of abandonment or intensification of fertilizer addition, mowing and grazing, have increased the threat to this species, which is also susceptible to atmospheric nitrogen deposition and toxification (Van Dam et al. 1986; Pegtel 1994). Arnica montana is therefore endangered (Korneck et al. 1996) and information regarding its population biology and the factors regulating its establishment in upland plant communities is urgently needed.

Observations during the unusually rainy growing seasons in 1994 and 1995, when innumerable slugs were found grazing on leaves and flowering stems of plants in the Harz mountains, led to the hypothesis that Arnica montana is restricted to higher elevations because herbivore pressure is less than at lower levels. This hypothesis was supported by food choice experiments that clearly demonstrated that Arnica leaves and seedlings form a preferred food source for several slug species occurring in montane meadows (Scheidel & Bruelheide 1999a).

We attempted to show how mollusc herbivory may limit the natural range of Arnica montana in the Harz mountains by (i) examining the effects of artificially increased slug populations on leaf damage to Arnica montana and (ii) analysing how mollusc damage to transplanted Arnica plants varies with altitude.

Materials and methods

Experiment 1

Slug densities were artificially increased by releasing specimens of Arion lusitanicus (Mabille) into native populations of Arnica montana in a montane meadow, dominated by Agrostis tenuis and Nardus stricta. The study site has a population of several thousand individuals of Arnica montana and is located near Braunlage in the Harz mountains, Lower Saxony, Germany, at an elevation of about 600 m a.s.l.

At the beginning of May 1997, when Arnica leaves began to emerge, 30 plots, 1 m × 1 m, were established, each containing 10–125 basal rosettes. Twenty of them were enclosed with polyethylene netting (mesh width 1.6 mm, to a height of 45 cm and buried 5–10 cm deep in the soil, fixed to wooden posts at the corners) to prevent intrusion or escape of slugs and snails. The other ‘uncaged’ 10 plots were simply marked with wooden rods.

On 21 May 1997, 20 immature individuals of the slug A. lusitanicus, which had been collected in a garden at an elevation of 165 m a.s.l., where herbivore pressure is likely to be high, were released into each of 10 enclosed plots (‘caged, with slugs’). This procedure was repeated in the same plots on 8 July. The remaining 10 plots (‘caged, without slugs’) served as controls to determine a possible fence effect.

Damage to leaves of Arnica montana was assessed approximately every 9 days, until the meadows were mowed successively between the middle of July and September. Ten shoots in each of the 30 plots were randomly chosen to estimate the percentage of consumed leaf area of all non-senescent leaves. Leaves were assigned to classes with up to 10%, 25%, 50%, 75% or 100% of the total leaf area removed. Numbers within each class were tallied and multiplied by the mean value of each class to give a total for all the measured leaves within a plot without considering the absolute leaf area. This sum was then divided by the number of developed leaves on the 10 shoots (i.e. the first four small cataphyll-like leaves on each shoot were not considered). In addition, on the same dates the length and width of all the leaves on three shoots in each plot were measured and used to calculate total leaf area and relative growth rate. Mean relative growth rate of leaf area (RGRL) over the interval t1 to t2 was calculated by (RGRL) = (lnLA2 – lnLA1)/(t2 – t1) after Hunt (1990, p. 28), with LA1 and LA2 being leaf area at t1 and t2, respectively.

On each sample date the vegetation height of each plot was measured and leaves from five to 10 other species (10 leaves per species) were chosen at random to estimate their damage using the same scale as for Arnica. As species composition differed, different plant species contributed to the mean damage value of the total vegetation of different plots.

The densities of slugs active above ground were estimated in four 1-m2 plots of each treatment by examining the vegetation closely for about 5 minutes on a rainy day in August.

Experiment 2

Seeds of Arnica montana, collected at Braunlage, were sown in May 1996 in soil in a greenhouse at the New Botanical Garden in Göttingen, Germany. Single plants were transferred to 10-cm plastic pots containing a low-pH compost in October and, in November, these pots were transferred to three sites at different elevations. The highest site was a montane meadow at the same location as in experiment 1 (610 m a.s.l.), the second a meadow on a nutrient-poor soil near Bad Lauterberg on the southern border of the Harz mountains (385 m a.s.l.) and the third plot a meadow in the New Botanical Garden in Göttingen (180 m a.s.l.). A plot of 5 m × 6 m was established at each site to receive 30 pots placed in a regular 1 m × 1 m grid.

Almost all of the plants survived the winter, but during March and April 1997 many plants in Göttingen were destroyed by burrowing rabbits. Values for leaf area in April and May at the Göttingen plot were based on only the four undamaged plants. The lost plants were replaced in May by young Arnica individuals of approximately the same age collected from a montane population. Subsequently, the experiment was continued with 24, 23 and 23 plants in Braunlage, Bad Lauterberg and Göttingen, respectively.

In the middle of May 1997, 10 plants per plot were randomly selected and caged to exclude slugs. Polyethylene nets (as before, mesh width 1.6 mm, to a height of 45 cm, buried 5–10 cm deep in the soil) were fixed to three wooden rods to enclose an area around the single plants of about 30 cm diameter. Shoots of other plants growing within a radius of about 10 cm around the pots were repeatedly cut.

The plots were visited about every 10 days until the meadows were mowed in July, and the number of developed leaves on the main shoot of each plant was counted, together with the number of lateral shoots, each of which was considered to have four leaves. The length and width of the leaves of the main shoot were measured to calculate the total leaf area. The relative growth rate and leaf damage of each plant was assessed in the same way as in experiment 1.

In addition, the height of the surrounding vegetation was measured and, when weather conditions were appropriate for high slug activity, slugs active above ground were counted by examining the vegetation of four 1 m × 1 m areas within the plots as in experiment 1.

Statistical analysis

All statistics were performed using SAS 6.02 (SAS Institute 1987). Tests of departure from Gaussian distribution refer to Shapiro–Wilk (using proc univariate). Tests of sum of ranks were performed according to the Kruskal–Wallis method (proc npar1way) following manual calculation of single comparisons after Schaich & Hamerle (Bortz et al. 1990).


Experiment 1

The relative growth rate of Arnica montana, considering all leaves of the rosette and of the flowering stem, reached a maximum at 0.15–0.19 day–1 on 19 May and then decreased almost to zero in July. The total leaf area per plant in July, when growth had ceased, ranged from 100 to a maximum of almost 700 cm2 and was, on average, about 200 cm2. There were no statistically significant differences between the three treatments. The mean height of vegetation, which reached a maximum of 89 cm in the middle of July, did not differ significantly between the treatments.

The percentage of leaf area consumed (Fig. 1) remained below 1% in all three treatments up to the point where slugs were released, but immediately increased as slugs began to eat Arnica leaves.

Figure 1.

Mean percentage of Arnica montana leaf area consumed (experiment 1) from May to September 1997. Arrows indicate dates of slug release. Error bars indicate standard deviation.

The intensity of leaf damage depended on density of Arnica (Fig. 2), being particularly marked (and increasing over time) in plots with very low densities. At Arnica densities of only 10 shoots per m2 the loss of leaf area was about 20% at the end of June, whereas in the plot with the highest densities of Arnica (125 m–2) the damage remained below 5%. The mean percentage of leaf area eaten reached approximately 7% by the middle of June (Fig. 1), with the reduced rate of damage between 28 May and 6 June possibly due to the dry weather during this period. There was little further damage from the middle of June onwards and even a slight recovery when low slug activity coincided with continued growth and the appearance of senescent leaves, which were not considered in the damage assessment.

Figure 2.

Relationship between mean percentage of consumed leaf area and number of Arnica shoots per m2 in the 10 caged plots with slugs (experiment 1). Trend lines show linear regression for the three dates analysed.

At the end of June, A. lusitanicus individuals were no longer found on the plots. The slugs must either have escaped or have become a prey to the numerous crows and other slug predators in this locality. After releasing new slugs, the mean amount of damage increased to 10% in July and almost 12% in August (Fig. 1). In contrast, the loss of leaf area in the caged plots without A. lusitanicus and in the uncaged plots increased only slightly to 1% in the middle of July and to about 3% at the beginning of September.

The native slug fauna, examined in 12 of the plots in August, after individuals of the second release of A. lusitanicus had vanished, was dominated by Deroceras sp. A maximum surface density of 14 active individuals per m2 was found. Single individuals of A. subfuscus and the A. fasciatus group were also present. No significant differences in slug density between the treatments were detected by the Kruskal–Wallis test. Molluscs were the only herbivores observed to cause damage to the leaves of Arnica montana.

The statistical analysis (single comparisons subsequent to the Kruskal–Wallis test, P < 0.05) revealed significant differences in the percentage of total eaten leaf area between the plots with A. lusitanicus and the plots without this slug for all measuring dates from 28 May to 1 September. The caged plots without A. lusitanicus and the uncaged plots did not differ significantly.

The mean damage to other plant species remained below 2% in the majority of cases (Fig. 3); only Plantago lanceolata and Trifolium pratense suffered significant losses of leaf area (6% and 10%, respectively). No significant differences between plots with and without A. lusitanicus were detected for species other than Arnica montana. It is clear that Arnica leaves were the main living plant component of the diet of the released slugs.

Figure 3.

Mean damage to plant species other than Arnica montana in the plots (experiment 1): percentage of consumed leaf area from May to September 1997.

Experiment 2

The measurement of the height of the surrounding vegetation revealed delayed growth and reduced maximum height with increasing elevation (Fig. 4). The transplanted Arnica montana specimens started to grow earlier at lower elevations (Fig. 5), although growth at Göttingen was interrupted in May when replacements were introduced for the plants destroyed by rabbits. Subsequently, the plants in Göttingen suffered heavy damage, which resulted in reduced growth. The maximum relative growth rate in Göttingen was reached at the beginning of April (0.07 day–1), in Bad Lauterberg at the end of April (0.14 day–1) and in Braunlage in the middle of May (0.07 day–1). Unlike both native (see experiment 1) and transplanted Arnica in Braunlage, the plants in Göttingen and Bad Lauterberg continued to grow beyond July. In Göttingen a further peak of relative growth rate occurred in July (0.08 day–1).

Figure 4.

Mean height of vegetation in the plots (experiment 2) from May to July 1997.

Figure 5.

Mean leaf area per plant of transplanted and caged Arnica montana at the three sites in Braunlage, Bad Lauterberg und Göttingen (experiment 2).

Mucus trails and faeces around the damaged plants and the typical slug feeding pattern indicated that leaf damage was mainly caused by molluscs. No other herbivores were observed feeding on Arnica leaves.

In Göttingen, the percentage of leaf area consumed increased dramatically in 3 weeks from mid-May to mid-June. Little new damage was found after mid-June but levels eventually reached about 70% on uncaged and 35% on caged plants (Fig. 6). According to the Kruskal–Wallis test (P < 0.05), the amounts of damage differed significantly between the treatments on 26 May, 12 June and from 7 July to 29 July. Two of the nine caged and nine of the 14 uncaged plants died because of nearly complete consumption.

Figure 6.

Mean percentage of consumed leaf area of transplanted Arnica montana in Göttingen, Bad Lauterberg and Braunlage (experiment 2).

The amount of damage increased less rapidly and reached much lower levels at Bad Lauterberg. The maximum was reached on 9 July, by which time the uncaged plants suffered nearly 8% loss of leaf area compared with only about 3% for the caged plants. The treatments differed significantly from 23 May to the end of July. None of the plants died.

Little damage was observed on the transplanted Arnica in Braunlage; the loss of leaf area remained below 0.25%, all plants survived and the treatments did not differ significantly on any observation date.

The maximum slug number in Göttingen was found in June with 11 individuals of Deroceras sp. per m2 (Fig. 7). Single specimens of A. ater or A. lusitanicus were found on later dates. After July, a maximum of three specimens of Deroceras per m2 was observed. Similar maximum densities of Deroceras were observed at Bad Lauterberg, but they occurred later, in July. Arion ater and A. fasciatus could also be found regularly at Bad Lauterberg. In the plot in Braunlage, a maximum of only one specimen of Deroceras sp. per m2 and two of A. fasciatus was observed.

Figure 7.

Maximum slug numbers per m2 in the three plots (experiment 2), counted at four areas (A–D) inside each plot in June or July 1997.


The results of experiment 1 indicate that Arnica montana is more acceptable than other grassland species. Although acceptability in experiment 1 was tested with only one slug species (A. lusitanicus), results of food choice experiments using leaf discs of several grassland perennials and three slug species have revealed that Arnica montana is highly acceptable to all the slug species that were included in the test (Scheidel & Bruelheide 1999a).

Among all the slug species occurring at sites for experiment 1, A. lusitanicus contributed the most to the damage of Arnica leaves because of its large size and high abundance when present. Arion lusitanicus is still spreading through Europe and increasing levels of slug grazing in various ecosystems, and possibly severe problems in agriculture, may therefore be expected (Frank 1998). The native slug fauna, which did not differ between treatments, also prefers Arnica (in the field as shown in experiment 2, and in food choice experiments with Deroceras sp. in Scheidel & Bruelheide 1999a) but causes only little damage because of the smaller animals’ lower food demand and their lower densities.

The losses of leaf area of about 10% after artificially increased density of slugs (experiment 1) and 8% after transplantation of Arnica to Bad Lauterberg (385 m a.s.l., experiment 2) fall in the range of 5–10%, suggested by Crawley (1989) as an average rate of loss to herbivorous insects. In Braunlage (610 m a.s.l.), however, where there are native Arnica populations, damage to transplanted individuals remained much lower, at a rate similar to the native plants other than those subjected to artificially increased herbivore pressure (a maximum of 3% damage in summer). After the meadows were mown in summer, new growing shoots suffered damage of approximately 9% in September, when Deroceras sp. was most abundant.

Could this degree of herbivory affect plant fitness? In Göttingen (180 m a.s.l.), more than 70% of the total leaf area was removed by slugs and, as a result, about half the plants died. Such severe detrimental effects can be expected to prevent the survival of an Arnica population at this site but evaluation of the effects of lower levels of tissue removal may be much more difficult, as discussed in Begon et al. (1996, pp. 315–325). Although the direct loss of photosynthetically active tissue or of reproductive organs implies a reduction in fitness, the plant may, in the long term, respond to damage by activation of meristems, intensification of cell division and delay of senescence, which together result in compensatory growth (McNaughton 1983). Compensatory growth is connected with an increase of nutrient turnover and may, on the whole, be advantageous for the plant (Owen & Wiegert 1976). Further effects of herbivory may be a change in nutrient allocation within the plant (Trumble et al. 1993) or of the plant architecture, possibly leading to higher reproductive success (Inouye 1982; Escarréet al. 1996; Lennartsson et al. 1997). In our study, herbivory had neither positive nor negative effects on flower and seed production, but the observations are based on only one vegetation period, which may be insufficient to assess long-term effects (cf. Karban & Strauss 1993; John & Turkington 1997).

Nevertheless, significantly increased amounts of damage at lower elevations strongly indicate that mollusc herbivory exerts a varying impact on Arnica population performance at different altitudes. Assuming that environmental stress increases with increasing elevations, our results may also be consistent with the theory of Menge & Sutherland (1987), who suggested that the relative importance of predation (e.g. herbivory) in community regulation at low recruitment intensities decreases as the ‘environmental stress’ becomes higher. The authors explained this tendency by stating that more severe environmental conditions affect mobile organisms (herbivores) to a greater extent than sessile organisms (plants) and thus reduce the importance of herbivory. Unfortunately, empirical data on herbivory gradients in montane plant species or communities is nearly non-existent, although Galen (1990) did find that ungulate and aphid herbivory increased at the lower boundary of the subalpine range of Polemonium viscosum.

In our case, a negative correlation between slug density and elevation might be expected because lower temperatures, especially winter minima, may result in retardation of the slug's life cycle (South 1982; Baur & Raboud 1988). Further information of the variation of slug density with altitude is needed.

The most probable explanation for varying degrees of damage to Arnica is different slug densities. We found slug abundance to be negatively correlated to elevation, although methodological problems and the small number of observations prevented statistical validation. The clumped distribution and underground life of slugs demand special investigational procedures (Hunter 1968; South 1964; Mühlenberg 1989; Frank 1998). Our method of searching the vegetation is highly dependent on weather and more appropriate for assessing actual slug activity above ground rather than absolute abundance but, in the absence of more reliable methods, may give a reasonable estimate of the risk of herbivory.

The observed differences in slug density may be partly due to factors other than altitude, such as vegetation structure (Scheidel & Bruelheide 1999b). The primary productivity of the plots increased with decreasing elevation and the vegetation became more dense. Several studies (Reader 1992; Rees & Brown 1992; Bonser & Reader 1995) have found that transplanted individuals suffered greater tissue loss when vegetation density was higher, possibly because of the improved habitat quality for herbivores. In addition to this indirect effect, the surrounding vegetation may exert varying direct competitive effects on transplanted Arnica montana. However, we attempted to minimize these interactions by transplanting in pots and cutting the plants around Arnica.

Matters become even more complicated when the timing of the plant–herbivore interaction is also considered. Damage to plant populations will be at maximum when high slug activity coincides with the phase in the plant's life cycle that is most sensitive to herbivory. As for most perennial plant species, the most vulnerable time for Arnica is in the spring when the young leaves emerge. At lower elevations, this phase starts earlier, as indicated in Fig. 5 for Göttingen compared with Braunlage, where growth starts later but the higher rates (as indicated in a steeper slope of the plant leaf area) result in plants having nearly the same leaf area by the middle of July. The sensitive period is therefore shorter at higher elevations. Additionally, maximum slug activity is earlier in Göttingen and thus more likely to coincide with the sensitive period than in Braunlage.

The preference of molluscs for young leaves may be due not only to a high protein content and a low content of sclerenchymatous tissue but also to low levels of deterrent chemicals. Unlike this study, Speiser & Rowell-Rahier (1991) observed increasing damage by snails upon Adenostyles alliaria leaves over the course of the growing season, as the content of chemical deterrents decreased. Some plant species are known to increase the production of such substances after damage, a mechanism known as induced resistance (Wratten et al. 1984; Van Dam & Vrieling 1994; Mutikainen et al. 1996). In food choice experiments, A. lusitanicus showed a significantly lower preference for Arnica leaves from plants that had been previously damaged than for those which had not (Scheidel & Bruelheide 1999a). Feeding on plants early in the growing season may therefore be advantageous for a slug because it avoids deterrent compounds. However, the annual cycle of sesquiterpene production, which is a possible deterrent agent in natural populations of Arnica montana, has not yet been investigated.

Cates (1975) emphasized the importance of the coincidence of high slug activity and the productive phase for Asarum caudatum. The chemically palatable ecotype of Asarum escapes slug grazing pressure by having its maximum growth period earlier, when slugs are not yet active, than the unpalatable ecotype. Temporal aspects of interactions were also emphasized by Westerbergh & Nyberg (1995), who found that mollusc grazing on Silene dioica was most intensive during the first part of the growing season.

Phenological differences between plants have been proposed to be of general importance in plant–herbivore interactions (Collinge & Louda 1989). Different timing of development has also been found to be responsible for sun/shade distribution patterns of herbivory on several plant species (Maiorana 1981; MacGarvin et al. 1986; Louda & Rodman 1996). Global warming might be expected to change such phenological interrelationships, particularly in montane and subalpine plant communities (Price & Waser 1998), and, therefore, influence plant–herbivore interactions.

The effects of timing are not restricted to the coincidence of high herbivory pressure with emerging new leaves, but also with the seedling stage. Molluscs generally prefer plants at the seedling stage (Byers & Bierlein 1982) and palatability may change with seedling age (Hanley et al. 1995b). Whereas mature plants are rarely killed by slugs, the effects of seedling damage by molluscs are normally lethal and slugs may therefore exert their greatest impact on plant population dynamics by grazing on the seedling stage. Clear-Hill & Silvertown (1997) found that seedling survival following germination from sown seeds was reduced by releasing slugs into grassland plots. Marked effects of timing were found by Hanley et al. (1996), who determined the impact of slugs on seedlings germinating from natural seed banks in grassland gaps. The results obtained in their spring and autumn experiments differed in seedling emergence, slug numbers and slug impact on species composition.

In greenhouse experiments Arnica seedlings have been found to suffer severe damage by slugs selectively grazing on them (Scheidel & Bruelheide 1999a), and further experiments will examine slug damage to Arnica montana seedlings in the field. Seedling mortality might be a bottleneck factor of Arnica population dynamics at submontane altitudes, where damage to mature plants seems to exert no long-term lethal effects. At lower elevations, the damage to both adult plants and seedlings caused by slugs can be expected to have more than additive effects and thus be a key factor in the population biology of Arnica.


We would like to express our warm thanks to P.J. Grubb, who enthusiastically encouraged us to attempt the first experiments with Arnica, to Anne Theenhaus, who kindly provided literature and advice on the zoological aspects, and to all our colleagues and friends who never became bored discussing the subject, especially Thomas Flintrop and Ute Jandt. We are also grateful to M. Runge who provided an ecological niche for our research in his department. We thank L. Haddon, M. Fenner, and two anonymous referees for very helpful suggestions to improve the quality of the paper, and D.F. Whybrew for the final polishing of the manuscript.

Received 9 July 1998revision accepted 18 March 1999