Pollination and insect visitors to the putatively brood-site deceptive endemic spurred helmet orchid, Corybas cheesemanii

Abstract

Deceptive orchids typically employ food or sexual deception of male insects. Brood-site deception, when flowers are pollinated by female flies fooled into attempting oviposition, is less well characterized for orchids but is probably common among New Zealand terrestrial orchid species. Helmet orchids (Corybas, Nematoceras and related genera) are considered specialist brood-site deceivers that mimic mushrooms and are pollinated by ovipositing female fungus gnats (Mycetophilidae). We monitored pollination in the endemic spurred helmet orchid, Corybas cheesemanii, and trapped insects visiting orchids and sympatric mushrooms. Flowering occurred from mid-May to early July, each flower lasted 23.14 d, 78.5% of the population flowered (147/194 plants), and c.25% of these set seed naturally (50% of bagged orchids set seed via autonomous self-pollination). Traps caught mostly flies (Mycetophilidae, Anisopodidae and Lauxaniidae), but orchid, mushroom and control traps did not attract significantly different fauna. Seven mycetophilids (of both sexes) visited both orchids and mushrooms, but we found no compelling evidence that C. cheesemanii was pollinated by these. Only some mycetophilid individuals caught above orchids were small enough to enter orchids, but none carried orchid pollinia. If C. cheesemanii is fungus gnat-pollinated, the range of species and sexes attracted suggests a more generalist pollination strategy than assumed.

Keywords:

• brood-site deception
• Corybas cheesemanii
• fungus gnat
• helmet orchid
• Mycetophilidae
• phenology
• pollinators
• self-pollination

Introduction

Deceptive pollination systems attract pollinators without providing rewards (Jersáková et al. 2006). Pollinators may be attracted by signals that mimic resources such as food or mates, or those that exploit pollinator preferences for certain features, e.g. ultraviolet reflectance (Schaefer & Ruxton 2009; Gaskett 2013). Brood-site deceptive pollination relies on insects such as Diptera and Coleoptera that usually oviposit in decaying organic matter, and occurs in at least 10 angiosperm families (Urru et al. 2010, 2011; Policha & Roy 2012). Pollinators are lured with signals such as odours that mimic oviposition substrates including carrion or fungi (Jürgens et al. 2006; Kaiser 2006; Urru et al. 2011; van der Niet et al. 2011). The flowers can be very large (e.g. titan arum Amorphophallus titanum) and generally dark coloured, often with a contrasting light-coloured trap chamber (Beaman et al. 1988; Urru et al. 2011).

Brood-deceptive orchids

Orchids are well-known for their unusual deceptive pollination systems. The best studied systems are food deception, when nectarless orchids are pollinated by foraging insects, and sexual deception, when orchids are pollinated by male insects fooled into sexual behaviour (recent reviews include Phillips et al. 2009; Schiestl & Schlüter 2009; Vereecken et al. 2010; Gaskett 2011).

Brood-site deception is less common among orchids than food or sexual deception, but is reported or suspected in several groups. There is strong evidence for brood-site mimicry in the foul-smelling South African terrestrial endemic orchid Satyrium pumilum (Diseae: Satyriinae; van der Niet et al. 2011). This species is pollinated by female flies from the family Sarcophagidae that usually mate and oviposit on carcasses. Neoptropical Dracula orchids (Epidendreae: Pleurothallidinae) were suspected brood-site mimics because their red-brown colours, fungal aromas and gill-like structures resemble mushrooms (Kaiser 2006; Dentinger & Roy 2010; Endara et al. 2010). Observations confirm that at least two species, Dracula felix and Dracula lafleurrii, are predominantly visited by drosophilid flies that usually reproduce in fungi (Endara et al. 2010). Charles Darwin noted preliminary evidence for brood-site mimicry in another pleurothallid orchid. He observed numerous insect eggs laid in flowers in a Kew hothouse (in Zootrophion atropurpureum, as Masdevallia fenestrata; Darwin 1885).

Odours are not mandatory for brood-site deception. Southeast Asian Paphiopedilum (Cypripedioideae: Cypripedieae) are faintly scented or scentless. However, as is typical for brood-site deceivers, they are mostly pollinated by female insects (Syrphidae; hoverflies) that sometimes lay eggs in the flowers (Bänziger 1996; Shi et al. 2007, 2009; Bänziger et al. 2012). Pollinator attraction may be visual in Paphiopedilum with some species exploiting insects' innate attraction to yellow, whereas others may attract insects with dark spots that mimic aphids (Shi et al. 2009).

Fly-pollinated orchids

Most orchids are pollinated by insects. After Hymenoptera, the Diptera (true flies) are the most important group (Christensen 1994). Flies are vital and sometimes sole pollinators for a range of rewarding, food-deceptive and sexually deceptive orchids. For example, Chinese Cypripedium pollinated by scathophagids and drosophilids (Li et al. 2012), various South African species pollinated by long-tongued flies (Nemestrinidae; e.g. Anderson & Johnson 2009) and South and Central American species whose pollinators include tachinids, sciarids, chloropids and phorids (e.g. Borba & Semir 2001; Singer et al. 2006). Fly pollination is particularly important in New Zealand, which has a wide diversity of flies and fly-pollinated plants, and lacks diversity in some important pollinator guilds such as social hymenoptera and butterflies (Newstrom & Robertson 2005). Many New Zealand orchids are likely to be fly-pollinated, e.g. rewarding epiphytes Earina and Winika (Lehnebach & Robertson 2004) and deceptive terrestrials such as the greenhoods Pterostylis and Diplodium and related and synonymous genera (Lehnebach et al. 2005).

New Zealand and Australian Helmet orchids (Corybas, Nematoceras, Singularybas and related genera) are widely believed to attract female fungus gnats as pollinators (Mycetophilidae: Diptera) via brood deception and mimicry of fungal scents and colours (Jones 1970; Fuller 1994; Jersáková et al. 2006; Scanlen 2006; Clements et al. 2007; St George 2007). Chance sightings indicate that fungus gnats do visit, remove pollinia from, and sometimes oviposit in, the flowers from at least some helmet orchid species (Miller 1918; Fuller 1994; Scanlen 1996, 2006, 2008; St George 2001). However, fungal mimicry is untested, and it is not known whether only females are attracted. Furthermore, taxonomic revisions mean that these reports are for species that are now mostly accepted as Nematoceras rather than Corybas (for old and new names, see Clements et al. 2007). After revisions, New Zealand retains just one Corybas species, the endemic spurred helmet orchid Corybas cheesemanii (Hook. f. ex Kirk). This species is presumed to be brood-deceptive and pollinated by fungus gnats, but spectral and gas chromatography–mass spectroscopy analyses reveal strong ultraviolet reflectance, and no evidence of the fungal mimicry expected of a brood-site mimic (Kelly & Gaskett unpublished data).

Mycetophilid fungus gnats (Nematocera: Diptera) are traditionally considered inefficient pollinators in comparison to more recent fly groups that tend to be larger-bodied and more densely haired (Mesler et al. 1980; Okuyama et al. 2004). However, fungus gnats are known to be efficient and specialist pollinators of small, dark, sometimes foul-smelling flowers, including the orchid Listera cordata (Mesler et al. 1980; Goldblatt et al. 2004; Okuyama et al. 2004). They may well pollinate New Zealand's unrewarding Corybas and Nematoceras orchids, which are similarly small and dark, but rarely scented. New Zealand has a diverse range of endemic fungus gnats available to act as pollinators (Tonnoir & Edwards 1927; Jaschhof & Didham 2002; Toft & Chandler 2004). Some families are relict lineages from the Jurassic, such as Ditomyiidae and the recently discovered endemic family Rangomaramidae. Also present are the globally abundant families Keroplatidae and Mycetophilidae, which probably diverged more recently but remain basal to many other Dipteran taxa.

There are currently no published data on whether fungus gnats do pollinate New Zealand's only Corybas species, C. cheesemanii, and some report that it is predominantly self-pollinating (Molloy 1990). Here we survey the phenology and fruit-set of C. cheesemanii, test for autonomous self-pollination and evaluate evidence for brood-site deception of fungus gnats by comparing insects trapped above C. cheesemanii flowers and sympatric mushrooms, quantifying the dimensions of the orchids and the sizes of the potential pollinators, and searching for evidence of oviposition in flowers. If C. cheesemanii is a classic brood-site deceiver, we predict that orchids and mushrooms will be visited by the same species of fungus gnats, that there will be more female than male fungus gnats attracted, that insects attracted will be small enough to fit within orchids to access pollinia, and that orchids will contain insect eggs.

Materials and methods

Study species and site

Corybas cheesemanii grows in lowland to montane scrub or forest shady leaf litter below Kunzea ericoides (kānuka), Pinus radiata, Beilschmiedia tarairi (taraire) and Nothofagus (beech) spp. (Smith 2009; Scanlen & St George 2010). Each plant produces only a single flower comprising a pearlescent ultraviolet-reflective helmet-like dorsal sepal (Kelly & Gaskett, unpublished data) and small lateral petals above a white labellum with a pair of nectarless spurs (de Lange et al. 2007; Scanlen & St George 2010). Corybas cheesemanii is not considered threatened and is distributed widely across the Three Kings, North, South and Chatham Islands (de Lange et al. 2006). Our study site (c.210 m²) was in Oratia, southwest of Auckland, New Zealand (latitude – 36.919499, longitude 174.611635). The orchids grew close to a stream, in a gully, in leaf litter below Cyathea dealbata (silver fern), Myrsine australis (red matipou), Coprosma lucida (shiny karamu) and Hedycarya arborea (pigeonwood).

Phenology and floral measurements

The study site was comprehensively surveyed for C. cheesemanii plants from April until mid-December 2011. We searched the entire site with care during the lead up to flowering and are confident that we found every individual plant that produced an above-ground leaf during the season. Plants in 10 clusters or subsites were labelled individually and monitored before, during and after flowering. Developmental stages recorded were: leaf only, emerging stem (Fig. 1A), bud (Fig. 1B), flowering imminent (Fig. 1C), open flower (Fig. 1D), mature flower (purple/brown in colour; Fig. 1E), shrivelled flower, senesced, long stem with swelling ovary (indicating fruit set) and seed set. Surveys were conducted once a week before flowering commenced (April–May), then twice a week during flowering (May–June), and then once every second week during fruit and seed set (August–December). Measurements of flower length and width, and stem height and width (after floral senescence to assess ovary swelling and fruit set; Fig. 1F,G) were taken every 1–2 wk over the flowering season for 20 individually tagged orchids. Measurements of stem width and height continued after floral senescence until seed set for those flowers that had an elongating stem that indicated fruit set. We dissected 16 mature flowers to measure the width of the narrowest internal diameter a pollinator would have to pass to access the pollinia and stigma. We also checked for the presence of nectar or exudate by probing the nectary-shaped spurs at the labellum base.

Figure 1  Stages of floral development in Corybas cheesemanii. A, emerging flower stem; B, bud; C, flowering imminent; D, open flower; E, mature open flower. Phenological measurement locations for Corybas cheesemanii flowers: F, FL, flower length; SH, stem height; SW, stem width; and G, FW, flower width.

Pollinator exclusion

To test for self-pollination, enclosures made from tulle netting (2-mm mesh), wire and garden staples were placed over 18 C. cheesemanii orchids as soon as flower buds appeared to prevent pollinator access to flowers. The nets were checked regularly for damage or holes. After flower senescence, the plants were assessed regularly for fruit set by monitoring stem length and ovary width.

Insect trapping

Insects visiting C. cheesemanii flowers and any sympatric mushrooms (Basidiomycetes) growing within 100 m of flowers were sampled with traps constructed from plastic bottles cut in half with the neck then inverted inside the base to form a funnel. The trap was then placed above the specimen with the bottom of the bottle pointing upwards, and held in place 10 cm above the ground using metal stakes (as per Marino 1991). Trapping commenced in April 2011, before C. cheesemanii began to flower, when traps were placed above mushrooms (n=10) or above nearby leaf litter (mushroom controls, n=10). As each mushroom senesced, usually within 4 d, all traps were moved to new mushrooms and a total of 43 individuals were surveyed. Once flowering began in mid-June, additional traps were set above mature C. cheesemanii flowers (n=10) and nearby C. cheesemanii leaves, i.e. plants that were not flowering (orchid controls, n=10), giving a total of 40 traps. Before the flowering season, traps were checked and emptied weekly. During peak flowering, traps were emptied every 1–2 d for 2 wk. As flowers senesced or disappeared, traps were relabelled and moved to new flowers until the end of the flowering period (n=17 in total). Insects were taken to the laboratory to be pinned and identified and checked for orchid pollinia. Insect specimens were first identified to order, and then all flies were identified to at least family level. Mycetophilidae were identified to species and gender. The width of each insect caught was measured using callipers. The composition of insects visiting the four treatments was compared with non-parametric analyses of similarity (Hammer et al. 2001). Analysis of variance was used to compare floral dimensions of orchids that did and did not set seed (SPSS 20). The observed and expected numbers of male versus female mycetophilids visiting mushrooms and orchids were compared with chi-square test. Values are presented as mean±SEM.

Results

Phenology

For all the subsites combined, there were 194 C. cheesemanii plants, 75.8% of these produced a flower. Floral buds appeared during early May, with just over 100 orchids in bud. The flowers commenced opening 24.36 d later and each flower remained open for 23.14 d. Peak flowering occurred in early June when c.62% of the orchids were in flower, and all flowering had finished by early July. Death or disappearance of flowers (perhaps by herbivory) occurred throughout flowering and fruiting, with the majority gone by mid-December. Of the 147 orchids that flowered, 43 (29.25%) commenced post-pollination stem elongation, and of those 37 (25.17%) were confirmed to have set seed.

Floral measurements

Corybas cheesemanii flowers were 9.55±0.23 mm long and 5.92±0.18 mm wide. After flower senescence, stem width was not significantly different between flowers that were pollinated and set fruit (2.55±0.20 mm) and those that did not set fruit (2.48±0.10 mm; ANOVA d.f.=1, F=3.029, P=0.10). However, stem elongation differed markedly between flowers that set fruit (53.69±7.58 mm, max.=172 mm, n=42) and flowers that did not (7.73±0.33 mm, n=17; ANOVA d.f.=1, F=9.536, P<0.005). The width of the flower opening in a mature flower was 1.81±0.09 mm and the narrowest part of the flower was 0.84±0.04 mm. No nectar was found in the orchid spurs. No insect eggs were found in any of the flowers.

Pollinator exclusion

Of the 18 bagged flowers, 12 developed a long stalk with a swollen ovary, indicating pollination, and of those, nine (50%) were confirmed to have set seed. The remaining six orchids either died before flowering finished, or their flowers senesced without setting seed.

Insect trapping

Across all the treatments, Diptera were the most commonly trapped insect group (93.3% of all specimens). The most commonly caught Diptera families were Anisopodidae, Lauxaniidae and Mycetophilidae (Fig. 2). An analysis of similarity found no significant difference in the composition of the fly families caught by the different trap treatments (orchid, orchid control, mushroom, mushroom control; R=−0.0008, P=0.46; Fig. 2). The orchid traps did catch a subset of those mycetophilids caught above mushrooms, especially Mycetophila colorata, Mycetophila fagi, Mycetophila filicornis, Mycetophila nr. subspinigera and Mycetophila vulgaris sp. group (Fig. 3), but overall the composition of mycetophilids was not significantly different between the traps (R=0.06, P=0.0596).

Figure 2  Number of flies per fly family trapped above mushrooms, near mushrooms (mushroom controls), above flowering Corybas cheesemanii (orchids) and above non-flowering C. cheesemanii plants (orchid controls).

Figure 3  Flies of the genus Mycetophila trapped above mushrooms, mushroom controls, Corybas cheesemanii orchids and orchid controls.

Male Mycetophila were slightly more abundant than females in both mushroom and orchid traps (Fig. 4), but this difference was not statistically significant when compared between these treatments for all Mycetophilidae (χ ²=1.261, d.f.=1, P=0.26) or each Mycetophila species (all P values>0.05). Measurements of flowers (see data above) and trapped insects indicate that very few insects were small enough to be able to enter a flower and access the pollinia and stigma (Table 1). Microscope examinations found no pollinia on any of the insects trapped.

Figure 4  Number of male and female Mycetophila trapped above mushrooms, mushroom controls, Corybas cheesemanii orchids and orchid controls.

Table 1  Dimensions inside Corybas cheesemanii orchids, and the widths of insect taxa trapped above orchids; insects were measured at the widest part of the thorax.

Specimen	Width±SEM (mm)	n	No. of individuals < 1 mm
Orchid's narrowest point	0.84±0.04	16
Insects trapped^1
Anisopodidae	1.31±0.02	74	3
Lauxaniidae	1.89±0.04	69	0
Tipulidae	0.92±0.05	28	12
Trichoceridae	0.73±0.07	7	6
Mycetophilidae	1.19±0.03	78	14
Mycetophila colorata
female	1.20±0	2	0
male	0.97±0.12	3	2
M. fagi
female	1.39±0.06	18	0
male	1.27±0.08	13	2
M. filicornis
female	1±0.04	6	1
male	1.14±0.08	11	2
M. nr. subspinigera
female	1.1±0.1	2	0
male	1.05±0.06	6	1
M. vulgaris (sp. group)
female	1.07±0.13	3	1
male	1±0.12	4	1
^1Heleomyzidae not included. The single specimen trapped above an orchid was too damaged for measurements.

Discussion

Although a subset of the mycetophilids visiting mushrooms also visited orchids, the range of insects attracted, the lack of significant differences between control and treatment traps, and the lack of pollinia observed on any insect, prevents any conclusions being drawn about the identity of the pollinator or pollinators of C. cheesemanii. There may have been no C. cheesemanii pollinator species at our study site, or insect pollination may be extremely rare for this orchid. Self-pollination is estimated to be present in c.60% of New Zealand orchids, compared with c.31% of orchids internationally (Molloy 1990; Peter & Johnson 2009). However, C. cheesemanii is unlikely to be exclusively self-pollinated because the fruit-set rates we observed in the wild (c.25%) and bagged individuals (50%) are much lower than expected for obligate selfing in orchids. Autonomous self-pollination in orchids results in almost complete pollination success, e.g. c.95% for Australian Epipogium roseum (Zhou et al. 2012) and 100% for South African Eulophia clavicornis var. nutans (Peter & Johnson 2009). However, the pollination rates we observed in C. cheesemanii are still higher than expected for completely insect-pollinated deceptive orchids. The global mean for rewarding orchids is 52.9%, for all deceptive orchids it is 25.03% (Neiland & Wilcock 1998), and for only sexually deceptive orchids it is 18.4% (Gaskett 2011). We suggest that C. cheesemanii has a mixed strategy involving both insect pollination and autonomous self-pollination. This strategy tends to result in intermediate pollination rates similar to those we observed, e.g. 31.5% for Satyrium pumilium (van der Niet et al. 2011); 26.3% for Paphiopedilum barbigerum and 58.5% for Paphiopedilum dianthum (Shi et al. 2009). A mixed strategy prioritizing insect pollination but allowing self-pollination in the absence of pollinators is likely to be adaptive when pollinators are rare or orchid populations are isolated (Molloy 1990; Lehnebach et al. 2005; Bernhardt & Edens-Meier 2010).

Interestingly, twice as many C. cheesemanii set seed when bagged than in natural conditions (50% versus 25%). This may be because pollinators were visiting unbagged flowers, but so infrequently that pollinia were removed more often than they were deposited. As orchids have their pollen bundled into pollinia, one pollinator visit is sufficient to remove the entire pollen load (Johnson & Edwards 2000). Orchids that had pollinia removed but not deposited would therefore be unable to self-pollinate. Unfortunately, the tiny, enclosed flowers of C. cheesemanii meant that we were unable to check for pollinia removal in unbagged flowers without permanently damaging the flowers.

The low pollination rates typical for orchids are generally attributed to pollinator limitation, i.e. the absence or rarity of pollinators (Wilcock & Neiland 2002; Tremblay et al. 2005). The long-lived flowers typical of orchids and observed here for C. cheesemanii (23 d) may be an adaptation to counteract this (Primack 1985; Neiland & Wilcock 1998; Ashman 2004). Corybas cheesemanii flowers in June when there are reduced numbers of flying insects, but this does coincide with peak abundances of many mycetophilid species (Tonnoir & Edwards 1927; Toft & Beggs 1995). In contrast, most brood-site deceptive flowers are ephemeral and last only a few days (Stensmyr et al. 2002; Urru et al. 2010). The flowers of the brood-deceptive orchids Satyrium pumilum and Dracula spp. are similarly surprisingly short-lived for orchids, despite abundant pollinators (c.11 d; Endara et al. 2010; van der Niet et al. 2011). Flowers of the brood-deceptive orchid Paphiopedilum barbigerum last for a similar period when pollinated (c.13 d), whereas unpollinated flowers last for a similar period to those reported here for C. cheesemanii (c.21 d).

Although we were unable to demonstrate any clear differences between the treatments, the trapping method proved effective for sampling mycetophilids and other flies generally associated with leaf litter and fungi (Peterson 2007; Evenhius & Okadome 2011; Matile 2011). The surprisingly large number of insects caught in mushroom control traps perhaps suggests that they were set too close to mushrooms and we would design any further studies to avoid this issue. However, the smaller numbers of insects caught in orchid control traps compared with the other treatments suggests that the insects were not completely ubiquitous at all locations in the field site.

If C. cheesemanii does require insect pollinators, it may have several pollinator species. Although sexually deceptive orchids are usually highly species specific in their pollinator attraction, food and brood-site deception can involve more generalist pollinator attraction (for deceptive orchid pollinator lists, see Cozzolino et al. 2005; Singer et al. 2006; Phillips et al. 2009; Johnson 2010; Gaskett 2011). For example, brood-deceptive Cretan Arum cyrenaicum and Arum concinnatum attract 11 Diptera families, with flies of several species appearing in just a few individual flowers (Urru et al. 2010). A wide diversity of flies pollinate brood-deceptive stapeliads, and many appear to share pollinator species (Meve & Liede 1994; Jürgens et al. 2006). In contrast, both Rafflesia pricei and the dead horse arum, Helicodiceros muscivorus, predominantly attract flies of just two calliphorid species (Beaman et al. 1988; Stensmyr et al. 2002). Among brood-deceptive orchids, Satyrium pumilium is pollinated exclusively by sarcophagid flies (especially Sarcophaga spp.; van der Niet et al. 2011), Paphiopedilum barbigerum by the syrphid Episyrphus balteatus (Shi et al. 2009) and Dracula by drosophilids (especially Zygothrica spp.; Endara et al. 2010). This variation in the degree of generalism or specialism is consistent with the fungal preferences of mycetophilids; while some are generalists, others are much more host specific in their choice of oviposition sites (Jakovlev 2012).

Surprisingly, we caught equal numbers of males and females of all mycetophilid taxa in control traps and above mushrooms and orchids. Most flowers confirmed to be brood-site deceptive attract predominantly female insects (Urru et al. 2011), but attraction of both sexes is reported for some fungi (Jakovlev 2012). Our study also suggests mushrooms attract both sexes of fungus gnats, perhaps indicating that the use of mushrooms in fungus gnat reproduction is not limited solely to oviposition by females. Data about mating, courtship and oviposition, and the specific role of fungi in these behaviours, are surprisingly limited for any species of Mycetophilidae (Okuyama et al. 2004), although there are many reports of mycetophilid larvae in fungi (Jakovlev 2012). Correspondingly, it is difficult to predict the role of a putative fungal mimic such as C. cheesemanii in the mating system of adult fungus gnats that may be potential pollinators. However, if C. cheesemanii is pollinated by mycetophilids (as suggested by anecdotal observations), it may be via the provision of a rendezvous site where both sexes congregate rather than by mimicking a simple oviposition site for females only. Pollination attraction by the provision or mimicry of rendezvous sites may well pre-date the evolution of floral rewards (Fishman & Lilach 2013). It is still present in some extant species such as Dracula lafluerrii orchids, which mimic fungi and are pollinated by courting and mating female and male drosophilids, and red Israeli poppies and anemones, which are pollinated by mating glaphyrid beetles (Scarabaeoidea; Endara et al. 2010; Keasar et al. 2010). The presence of fly eggs in other New Zealand helmet orchids (Nematoceras spp.; Scanlen 2006) suggests that the full repertoire of meeting, courtship, mating and oviposition can indeed take place on orchids. Whether these behaviours result in orchid pollination is yet to be confirmed.

Despite the inconclusive nature of our data, we are still able to recommend that a broader definition of brood-site deception be considered when testing for this phenomenon in the New Zealand flora. Bänziger et al. (2012) suggest that two forms of brood-site deception exist in Paphiopedilum, classic chemical attraction of female insects with entrapment, oviposition and pollination, and a perhaps more basal form in which visual attractants replace chemical lures, some food deception occurs and both female and male insects are attracted. Similar variation may occur in New Zealand where mixed and generalist pollination strategies are common (Newstrom & Robertson 2005).

Acknowledgements

We thank Oratia Church Trust and Geoff Davidson for access to the study site. Field assistance and support was provided by John and Linda Kelly and Velimir Gayevskiy.