Gender and abiotic stress affect community-scale intensity of facilitation and its costs

Authors


Correspondence author. E-mail: brittany.cranston@otago.ac.nz

Summary

1. Facilitation allows some plant species to occupy environments that they are otherwise unable to inhabit, potentially leading to greater fitness, greater productivity or abundance, and range expansions. However, we know little about the costs incurred by facilitators or how variation in resource allocation by facilitators influences effects and costs.

2. Gynodioecy provides an opportunity to explore the effects of facilitation and potential costs because females and hermaphrodites differ in resource allocation patterns. We explored whether environmental stress and gender-specific facilitation influence species interactions in an alpine plant community. We investigated the degree of facilitation and correlative associated costs for the gynodioecious alpine plant, Silene acaulis, at two elevations (2317 and 2560 m) in the northern Rocky Mountains of Montana, USA.

3. Hermaphroditic Silene individuals supported a greater number of plant species and individual plants than female Silene (hermaphrodites = 4.2 ± 0.3, females = 3.5 ± 0.2; hermaphrodites = 11.5 ± 1.0, females = 9.1 ± 1.1). Facilitative effects on species richness significantly increased with elevation (2.1 ± 1.6 species at the low site vs. 3.2 ± 1.8 species at the high site), but abundance and percentage cover did not.

4. Silene fitness at the high elevation site (2560 m) was reduced compared to the lower site (2317 m) as measured by flower and seed production. Female flower production decreased by 40%, the number of seeds per fruit by 11.6% and leaf size by 24%.

5. Increases in the percentage cover of beneficiaries reduced the number of flowers per Silene plant, indicating a cost of facilitating other species, and this cost was slightly greater for females. However, seed size for hermaphrodites and leaf length for females increased with the percentage cover of beneficiaries, suggesting a mutual benefit of harbouring other species.

6.Synthesis. Our results show that gender can affect the balance between competitive and facilitative interactions. Stronger facilitation by hermaphroditic Silene acaulis, coupled with a lower cost of harbouring beneficiaries, suggests that life-history traits and related patterns of resource allocation can influence the facilitative effects of a species.

Introduction

Facilitation occurs when factors limiting the growth and survival of a plant are directly or indirectly alleviated by the presence of a neighbour (Bertness 1991). These positive interactions are often more pronounced in abiotically stressful environments such as deserts, arctic or alpine tundra and salt marshes (Hunter & Aarssen 1988; Bertness & Callaway 1994; Brooker et al. 2007; Callaway 2007). Facilitation allows some species to establish in environments that would otherwise be too stressful, potentially leading to range expansion (Bruno, Stachowicz & Bertness 2003), shifts in realized niches (Hacker & Gaines 1997; Choler, Michalet & Callaway 2001) and increased invasibility (Brooker et al. 2007). Thus, positive interactions can be a key process for understanding the distribution and abundances of species in plant communities (Hunter & Aarssen 1988; Brooker & Callaghan 1998; Klanderud & Totland 2005). For the purpose of this study, we define stress as a reduction in the fitness of an individual to sub-optimal levels as a result of an external force, such as resource limitation.

Generally, studies of facilitation have focused on the benefit provided by the facilitator and received by other species, or ‘beneficiaries’. However, facilitation commonly increases spatial proximity among species, which can have the concomitant effect of increasing competitive interactions (Callaway 1998), because plants that benefit from one another are also likely to simultaneously compete for limiting resources (Grime 1989; Callaway, Nadkarni & Mahall 1991; Holmgren, Scheffer & Huston 1997). Thus, beneficiaries have the potential to impose strong costs on their facilitators through their competitive effects. Such important negative effects of facilitation have seldom been explored in studies of plant interactions (McAuliffe 1984, 1988; Pugnaire, Haase & Puigdefabregas 1996; Holzapfel & Mahall 1999; Michalet et al. 2011). Yet, when costs are great, beneficiaries can eventually replace their facilitators, leading to dynamic changes in communities (McAuliffe 1988; Callaway & Davis 1993; Butterfield et al. 2010).

Gynodioecy provides a good opportunity to explore the effects and costs of various ecological processes like facilitation because females and hermaphrodites often differ in resource allocation patterns and responses to stress (Ashman & Schoen 1994; Alatalo & Molau 1995; Case & Ashman 2005). This sex-based variation in allocation and stress tolerance has the potential to alter competitive and facilitative interactions. Gynodioecious plant populations contain unisexual plants bearing male-sterile (female) flowers and hermaphroditic plants bearing perfect flowers with both male and female functions. Non-random spatial distributions of female and hermaphrodite individuals in several gynodioecious populations suggest that gender expression may be partially maternally inherited (Laporte et al. 2001; Klaas & Olson 2006), in that a female’s offspring may be either female or hermaphroditic, and that genders may respond in different ways to different ecological conditions. Gynodioecy is common in environmentally stressful environments where females are often overrepresented (Darwin 1877; Philipp et al. 1990; Alatalo & Molau 1995). This pattern suggests that females may allocate resources in ways that increase their fitness in stressful conditions more than hermaphrodites; however, the underlying mechanisms by which females achieve this are not clear (Delph 2003). The advantage of being female in stressful conditions may relate to the higher cost of reproduction for hermaphrodites that comes from producing both ovules and pollen.

Pollen accounts for a small proportion of the reproductive biomass in most species; however, it is richer in protein (Roulston & Cane 2002) and essential nutrients (N and P) than in both flowers and seeds (Ashman & Schoen 1994; Case & Ashman 2005). Pollen-producing plants also incur greater ‘opportunity costs’ than females because they invest resources in reproduction earlier in the season (Case & Ashman 2005) that might otherwise be available for vegetative growth. Such costs are measured as a reduction in reproductive fitness and can be more pronounced in nutrient-limited conditions, where females often occur at higher frequencies (Eckhart & Chapin 1997). Variation in the costs and benefits of these reproductive strategies provides unusual opportunities to examine how the allocation of resources to reproduction affects the ability of a plant to compete with or facilitate its neighbours.

We investigated the role of gender in facilitation by the cushion plant Silene acaulis (L.) Jacq var. subacaulescens (F. N. Williams) (Caryophyllaceae) at two elevations in alpine communities. Silene acaulis (hereafter referred to as Silene) is typically gynodioecious (Desfeux et al. 1996; Maurice et al. 1998), although individuals of indeterminate gender (those containing both female and perfect flowers) and hermaphrodites that produce no seed (functionally male) have also been observed (Shykoff 1988; Hermanutz & Innes 1994; Maurice et al. 1998). If environmentally stress-tolerant females in gynodioecious Silene populations have a lower competitive ‘response’ to their neighbours (see Liancourt, Callaway & Michalet 2005), they may be more susceptible to the negative effects of neighbour associations than hermaphrodites. Based on abiotic stress, females should occur at greater frequencies than hermaphrodites at high elevations (Darwin 1877; Philipp et al. 1990; Alatalo & Molau 1995). However, stress-tolerant females should pay a higher cost of facilitation in general, or potentially at higher elevations where resources are scarce (Liancourt, Callaway & Michalet 2005). Specifically, we tested four predictions using field measurements: (i) more severe environmental conditions will increase the frequency of female relative to hermaphrodite Silene, as gynodioecious females are often described as more stress tolerant than hermaphrodites and many studies have shown that female frequency increases with environmental severity (Darwin 1877; Philipp et al. 1990; Alatalo & Molau 1995); (ii) Silene will elicit stronger facilitative effects in conditions of higher abiotic stress, regardless of gender (see Bertness & Callaway 1994); (iii) hermaphrodite Silene will carry larger beneficiary loads than females (i.e. will be stronger facilitators) because they incur a greater allocational cost of producing pollen, and this reproductive cost could reduce the competitive effect of hermaphrodites on species harboured within the cushions; and (iv) female Silene growing at higher elevations will incur a greater cost of facilitation than hermaphrodites because stress-tolerant individuals are often more susceptible to competition (Grime 1979; Liancourt, Callaway & Michalet 2005).

Materials and methods

Field sites

Our study was conducted on Swiftcurrent Mountain, on the continental divide at two alpine elevations, low (2317 m) and high (2560 m), in Glacier National Park in northern Montana, USA, in July–August 2010. Tree line in this region occurs between 1900 and 2300 m (Fagre et al. 2000). The climate is a mix of semi-continental and semi-oceanic effects (Klasner & Fagre 2002). To the west of the continental divide, prevailing westerly winds and storms track inland from the Pacific Ocean. To the east of the divide, the climate is continental. Annual precipitation is approximately 200 cm and predominantly falls as snow (Klasner & Fagre 2002). Average July temperatures are 16 °C at 1830-m elevation (Finklin 1986).

Experimental design

To test whether Silene cushions facilitate the establishment of other plant species, 160 variable area plots (one cushion per plot, 80 cushion plots per site) of Silene cushions were established with the cushion area delineating the plot dimensions. Plots were located using a restricted random sampling design, controlling for variation in microhabitat and cushion size. Sampling was confined to cushions for which we could determine gender. Malleable wire rings, in the same dimensions as the cushions, were used to mark out a further 80 adjacent control plots per site (160 plots total per site) to the east or west of each cushion, avoiding any large rocks. Plots were measured at the high (2560 m) and low (2317 m) elevation alpine sites. Sampling was timed to coincide with flowering in Silene, so that individuals could be easily sexed. Silene in cushion plots were sexed, area estimated from the dimensions of the cushion [calculated as π(mean diameter/2)2] and the number of flowers counted. We counted the number of individuals of all plant species growing within cushion and control plots and estimated their percentage cover.

To determine the cost of facilitation, a further 16 female and 16 hermaphrodite cushions were selected at each site (64 cushions total). Cushions were measured on 16 August 2010. For each cushion, the total percentage cover of plant beneficiaries was estimated, 20 cushion leaves sampled and three fruits taken. Length, width and specific leaf area (SLA) were measured for each cushion leaf. For each fruit, the number of seeds was counted and the mean length from among 5–12 seeds was measured (depending on the number and condition of seeds in the fruit) using a lens micrometer. To measure the temperature difference between the high and low elevation sites, eight temperature loggers (ibuttons®; Maxim Innovation Delivered, Sunnyvale, CA, USA) were buried in pairs at each site. Four were buried beneath Silene cushions, and four were buried in the open, within 10 cm of a cushion containing a logger. Loggers were buried at a depth of approximately 8 cm.

Data analysis

We tested whether cushion gender varied with elevation using Pearson’s chi-square test. Linear mixed models (LMM) were used to determine whether species richness, abundance or percentage cover of beneficiaries were dependent on the elevation or the gender of the Silene host. We defined the degree of facilitation as the difference in beneficiary richness and abundance between cushion and control plots, so we specifically tested for an interaction between gender and plot type (cushion or control) and elevation and plot type. Cushion/control plot pairs were included as random intercept terms in the LMM. We tested for a cost of facilitation by modelling plant reproductive and morphological characteristics at the two elevations, as well as the gender and the percentage cover of beneficiaries. A cost was indicated by a significant negative effect of the percentage cover of beneficiaries on the plant trait (leaf size, seed size, number of seeds per fruit or number of flowers). The number of flowers per plant was modelled using an LMM with cushion size as a random effect to account for larger cushions bearing more flowers. Because cushions with more shoots can consequently bear more flowers, cushion size was used as a proxy for the number of shoots. The remainder of the reproductive measurements (seed size and number of seeds per fruit), as well as the leaf morphology measurements (length, width and SLA), was analysed using LMMs with each sample (leaf or seed) nested within plots included as random effects. All analyses were completed using r version 2.9.2 for Mac OS X (R Development Core Team 2009). All values are reported as mean ± SE.

Results

The high elevation site (2560 m) was cooler than the low elevation site (2317 m) with mean temperatures of 8.8 ± 0.1˚C and 10.4 ± 0.2˚C, respectively.

Female Silene cushions did not occur at greater frequencies than hermaphrodites at either the high (2560 m) or low (2317 m) elevation sites (χ2 = 0.26, d.f. = 2, = 0.61). The proportion of female plants was 42.5% at the low elevation site and 40% at the high elevation site. On average, female Silene cushions were larger (1219.7 ± 138.9 cm2) than hermaphrodites were (877.0 ± 97.2 cm2; F1,132 = 4.33 = 0.03). We identified 32 plant species in our plots. Potentilla nivea, Solidago multiradiata, Festuca idahoensis, Geum rossii and Dryas sp. were the most common plant species found in control plots, while Carex sp., Erigeron compositus, Sedum lanceolatum and Oxytropis campestris were the most common species in cushions. With few exceptions, species composition at the high elevation site was similar to the low elevation site. However, Sedum rosea and Polemonium viscosum were found only at the high elevation site, while Dryas sp., Potentilla fruticosa, Saxifraga lyallii, S. occidentalis and Picea glauca were only found at the low elevation site.

Beneficiary species richness was much higher in cushions (3.9 ± 0.2 species per cushion) than in the open control plots (1.2 ± 0.1 species per control plot; F1,314 = 206.71, < 0.001). Similarly, the abundance of beneficiaries was higher in Silene cushions (10.4 ± 0.8 beneficiaries per cushion) than in control plots (2.1 ± 0.2 beneficiaries per control plot; F1,314 = 124.22, < 0.001; Fig. 1). Percentage cover of beneficiaries was also greater in cushions (23.0 ± 2.1 % cover of beneficiaries) than in the open control plots (8.0 ± 1.3 % cover of beneficiaries; F1,314 = 37.81, < 0.0001; Fig. 1).

Figure 1.

 The increase in the (a) species richness of beneficiaries, (b) number of individual beneficiaries observed and (c) percentage cover of beneficiaries observed in 80 Silene cushions compared to 80 paired control plots (no cushion) at each of the low (2317 m) and high (2560 m) elevation sites at Swiftcurrent Mountain, Glacier National Park, Montana, USA.

Changes in the strength of facilitation with elevation depended on the specific measurement of the community. The effects of Silene on plant species abundance (frequency) were not significantly affected by elevation (F1,151 = 0.001, P = 0.983; Fig. 1b). However, cushions increased species richness by 2.1 ± 1.6 species at the low site vs. 3.2 ± 1.8 species at the high site (Fig. 1a: plot type × elevation interaction: F1,314 = 10.36, P = 0.002). We observed higher percentage cover of beneficiaries with elevation, but this trend was not statistically significant (F1,151 = 2.25, P = 0.136; Fig. 1c).

Hermaphrodite Silene were stronger facilitators than females as they supported more plant species than female cushions (hermaphrodites = 4.2 ± 0.3, females = 3.5 ± 0.2; F1,314 = 4.06, P = 0.04) and marginally greater abundances of beneficiaries (hermaphrodites = 11.5 ± 1.0, females = 9.1 ± 1.1; F1,314 = 2.75, P = 0.09; Fig. 1a,b). The percentage cover of beneficiaries was not affected by the gender of the facilitator (hermaphrodites = 24.2 ± 2.7%, females 21.1 ± 3.2%; F1,314 = 0.84, P = 0.36; Fig. 1c).

Silene reproductive and morphological traits differed over the ∼250-m elevational difference studied (Table 1). Silene at the high elevation site had 39.8% fewer flowers per plant (F1,52 = 13.5, P = 0.0001), 11.6% fewer seeds per fruit (F1,60 = 4.98, P = 0.03), 23.8% lower SLA (F1,60 = 143.58, < 0.0001) and 18.8% shorter leaves [F1,60 = 144.03, < 0.0001; although the negative effect of elevation on leaf length was greater for females (F1,60 = 8.29, P = 0.01)]. The effect of elevation on the number of flowers per plant was driven by female Silene (elevation × gender interaction: F1,52 = 4.63, P = 0.04); the number of flowers produced by hermaphrodites did not change with elevation (low = 160.1 ± 22.5, high = 159.9 ± 26.8). Seed size increased with elevation (F1,60 = 266.30, < 0.0001), although the increase was significantly greater for females that had 24.1% larger seeds at the high elevation site than hermaphrodites, whose seeds were only 15% larger (gender × elevation interaction; F1,60 = 18.96, < 0.0001). Hermaphrodite leaves were 9.1% wider than female leaves at the high elevation site (F1,60 = 8.83, P = 0.004), but leaf width was not affected by elevation (F1,60 = 0.02, P = 0.89).

Table 1.   Reproductive and morphological measurements for female and hermaphrodite Silene acaulis individuals at the high (2317 m) and low (2560 m) elevation sites at Swiftcurrent Mountain, Glacier National Park, Montana, USA
  d.f.Leaf widthLeaf specific leaf areaLeaf length
F P F P F P
Gender18.830.0040.030.861.810.18
Percentage cover112.890.0010.000.972.510.12
Elevation10.020.89143.58<0.001144.03<0.001
Gender × percentage cover10.380.540.630.4317.49<0.001
Gender × elevation10.020.991.280.268.290.01
Percentage cover × elevation10.480.490.780.3813.660.001
Gender × percentage cover × elevation11.290.261.440.230.110.75
 d.f.Number of seedsSeed sizeNumber of flowers
F P F P F P
Gender12.820.091.330.250.790.38
Percentage cover10.140.2962.82<0.0015.110.03
Elevation14.980.03266.30<0.00113.500.001
Gender × percentage cover10.300.591.160.290.840.36
Gender × elevation10.310.5818.96<0.0014.630.04
Percentage cover × elevation11.150.290.640.432.850.09
Gender × percentage cover × elevation10.480.491.380.250.640.42

The cover of plant species in Silene cushions correlated with changes in the reproduction and morphology of the Silene host, but in only one case did these changes differ with elevation. Increases in percentage cover of beneficiaries resulted in significant decreases in the number of flowers per host plant (F1,52 = 4.63, P = 0.03; Fig. 2a), although this effect was not strong, with the model accounting for only 18.1% of the variation in the data. Importantly, this cost did not vary between elevations or between genders. Neither the number of seeds per fruit (F1,60 = 1.44, P = 0.24; Fig. 2f) nor seed size (Fig. 2b) of Silene plants corresponded with increasing percentage cover of beneficiaries, indicating no cost for these traits. Seed size significantly increased with the percentage cover of beneficiaries (F1,60 = 62.82, < 0.001) harboured by hermaphroditic plants, suggesting a possible mutual benefit for hermaphrodites facilitating other species, but seed size for female plants decreased with percentage cover of beneficiaries, suggesting a cost of facilitation for this gender (F1,60 = 18.96, < 0.001). Because both flower number and seed size are negatively affected by percentage cover of beneficiaries in females only, they appear to suffer a greater cost of facilitation than hermaphrodites. SLA was not affected by percentage cover of beneficiaries (F1,60 = 0.00, P = 0.97; Fig. 2c), but leaf length increased with percentage cover at the low elevation site (F1,60 = 13.66, < 0.001; Fig. 2e). This effect of percentage cover on leaf length was driven by female Silene; percentage cover had no effect on the leaf length of hermaphrodites (gender × beneficiary cover interaction; F1,60 = 17.49, < 0.0001). Leaf width also significantly increased with percentage cover of beneficiaries for both genders (F1,60 = 12.89, P = 0.001; Fig. 2d).

Figure 2.

 Partial residual plot of (a) number of flowers per Silene cushion plant, (b) average seed size, (c) specific leaf area (SLA), (d) leaf width, (e) leaf length and (f) number of seeds per fruit by percentage cover of beneficiaries for Silene individuals at Swiftcurrent Mountain, Glacier National Park, Montana, USA. The partial residuals were calculated as the sum of the residuals from the fitted model and the estimated effects of interest in the plot. Differences between trends in hermaphrodite and female Silene are indicated where gender was a significant effect in the model.

Discussion

We found that the gender of a strong facilitator significantly affected the intensity of its facilitative effect and determined the level of some apparent costs or mutual benefits to the hosts for facilitating other species. Silene has been shown to be a strong facilitator of alpine plants in other studies (Griggs 1956; Jones & Richards 1962; Antonsson, Björk & Molau 2009; R.M. Callaway, unpublished data) as have other cushion species in other alpine systems (Cavieres & Badano 2009; Reid, Lamarque & Lortie 2010). Verdú, Villar-Salvador & García-Fayos (2004) and Montesinos, Verdú & García-Fayos (2007) also investigated the role of gender in facilitation by dioecious species and found that while females may be better facilitators during the seed-seedling stage, as beneficiaries grow female facilitators are more negatively impacted by neighbour associations than males. To our knowledge, however, no other studies have explored the ecological implications of gender for Silene or linked the gender of gynodioecious facilitators to their effect on beneficiaries.

Female frequency

Female frequency did not change with elevation, which is contrary to our prediction that harsh environmental conditions will favour higher frequencies of female Silene. In many gynodioecious populations, females are more prevalent in stressful environments (Philipp et al. 1990; Alatalo & Molau 1995), suggesting that they may have greater fitness than hermaphrodites in stressful sites. Indeed, we found that at the high elevation site, although female Silene cushions had fewer flowers per plant and smaller leaves, they also had significantly larger seeds than did hermaphroditic plants and so are more likely to germinate and produce seedlings (Eriksson 1999). This kind of apparent increase in female fitness relative to hermaphrodites is common in environmentally stressful environments (Pettersson 1992; Shykoff 1992; Delph & Carroll 2001; Delph 2003; Moles & Westoby 2004), and despite female Silene producing fewer flowers per plant at the high elevation site, increased fitness through greater seed size may help to explain the maintenance of female Silene in populations.

Facilitation intensity

‘Stress’ in plant ecology is difficult to define and can be ambiguous (Grime 1989; Körner 2003; Lortie et al. 2004). Nonetheless, despite a modest elevation difference relative to other studies of species interactions along alpine gradients (e.g. Callaway et al. 2002), we measured cooler temperatures, decreased fitness in Silene (40% fewer flowers per plant, 11.6% fewer seeds per fruit) and smaller leaves (24% smaller) at the high site in comparison with the low elevation site. Concomitant with these indicators of greater abiotic stress at the high site, and associated decrease in Silene performance, we found an increase in the facilitative effect of Silene on other plant species. This shift in facilitative effects with stress is consistent with a number of other studies in alpine systems (Choler, Michalet & Callaway 2001; Callaway et al. 2002; LeRoux & McGeoch 2008; Antonsson, Björk & Molau 2009; Cavieres & Badano 2009). Our interpretations of facilitation and competition are based on observations of pre-existing patterns, which are crucial for exploring whole-community patterns. We did not conduct experiments. However, experiments with cushion plants such as Silene and turf-forming alpine species have shown that similar spatial relationships are based to a large degree on facilitation (Choler, Michalet & Callaway 2001; Callaway et al. 2002; Klanderud & Totland 2005; Badano et al. 2006; Reid, Lamarque & Lortie 2010).

Hermaphrodites are stronger facilitators

We predicted that hermaphrodites would be stronger facilitators than females because they incur a greater allocational cost of producing pollen, and this reproductive cost could reduce the competitive effect of hermaphrodites on species harboured within the cushions. Our results supported this prediction. This interplay between gender costs and the strength of facilitation may have crucial effects on the ecological and evolutionary outcomes between Silene and other species (Brooker et al. 2007; Thorpe et al. 2011; Michalet et al. 2011).

Females incur a greater cost of facilitation

Facilitation is often conceptualized as a net effect derived from a balance between competitive and facilitative interactions (Callaway, Nadkarni & Mahall 1991; Callaway et al. 2002). In this context, facilitators have been shown to have competitive effects on beneficiaries despite a net facilitative effect (Muller 1953; Callaway, Nadkarni & Mahall 1991; Barton 1993; Franco & Nobel 1988, 1989). Although we know that competition and facilitation mechanisms can occur simultaneously as plants interact, the effect of beneficiaries on their facilitators has only rarely been considered. In our study, as the cover of beneficiaries increased, the number of flowers per cushion plant decreased, indicating a clear reduction in fitness of the facilitator. However, leaf size (width and length in females) and seed size for hermaphrodite Silene significantly increased with percentage cover of beneficiaries, while SLA remained constant. Silene, particularly females who do not need to produce pollen, may thus be allocating resources to leaf production, increasing photosynthetic area, to compensate for shading by beneficiaries at the expense of flower and seed production (Parkhurst & Loucks 1972).

For some gynodioecious species, females are more tolerant of environmental stress than hermaphrodites (Delph 2003; Ashman 2006). Therefore, we predicted that females would experience a greater cost of facilitation than hermaphrodites because of the trade-off between stress tolerance and competitive ‘response’ (Grime 1979; Liancourt, Callaway & Michalet 2005). However, our results suggest that females may be no more stress tolerant than hermaphrodites and thus female Silene may be allocating resources to avoid, or tolerate, competition (e.g. increasing leaf size), rather than in mechanisms for tolerating abiotic stress.

Conclusion

Facilitation has been well documented in alpine systems (Maillette 1988; Carlsson & Callaghan 1991; Jonasson 1992; Shevtsova et al. 1995; Choler, Michalet & Callaway 2001; Callaway et al. 2002), but we found that gynodioecy can substantially alter competitive and facilitative outcomes. Gynodioecy is a common breeding system in alpine environments, and our results suggest that gender of plant individuals should be taken into account in studies on facilitation. The higher reproductive costs of supporting beneficiaries suggest that females might be more easily displaced by beneficiaries than hermaphrodites and thus could result in life-history ratios in populations shifting towards the latter. Understanding the causes of the reduction in female fitness with elevation (i.e. the increase in facilitation with elevation, coupled with higher reproductive costs of supporting beneficiaries) will add substantially to our use of gynodioecious plants to explore competitive and facilitative interactions.

Acknowledgements

The authors would like to acknowledge Annalisa Ingegno and Giles Thelen who assisted in the field and in the laboratory. We thank C. Körner for his insightful comments on early versions of this manuscript. This work is part of a larger study was supported by the Miss E. L. Hellaby Indigenous Grasslands Research Trust (New Zealand; BHC), the Ecology Research Group (University of Otago, New Zealand; BHC), a University of Otago Prestigious Doctoral Scholarship (BHC), the University of Otago William Evans Fellowship (RMC) and the University of Montana International program (RMC). Thank you to Prof. Luise Hermanutz and KJMD’s laboratory group for thoughtful and detailed comments on early versions of the manuscript.

Ancillary