|Home | About | Journals | Submit | Contact Us | Français|
Spatial variation in pollinator composition and abundance is a well-recognized phenomenon. However, a weakness of many studies claiming specificity of plant–pollinator interactions is that they are often restricted to a single locality. The aim of the present study was to investigate pollinator effectiveness of the different flower visitors to the terrestrial orchid Eulophia alta at three different localities and to analyse whether differences in pollinator abundance and composition effect this plant's reproductive success.
Natural pollination was observed in vivo, and manipulative experiments were used to study the pollination biology and breeding system of E. alta at three sites near Manaus, Brazil. To gain a better understanding of the underlying mechanisms of pollinator attraction, nectar composition and secretion patterns were also studied, floral scent composition was analysed and a bioassay was conducted.
Flower visitors, pollinator composition, pollinia transfer efficiency of particular pollinator species and natural fruit set differed among the investigated populations of E. alta. Flowers were self-compatible, partially autogamous and effectively pollinated by five bee species (four Centris species and Xylocopa muscaria). Visiting insects appeared to imbibe small amounts of hexose-rich nectar. Nectar sugar content was highest on the third day after flower opening. Floral fragrance analyses revealed 42 compounds, of which monoterpenes and benzenoids predominated. A bioassay using floral parts revealed that only floral tissue from the labellum chamber and labellum tip was attractive to flower visitors.
The data suggest that observed differences in reproductive success in the three populations cannot be explained by absolute abundance of pollinators alone. Due to behavioural patterns such as disturbance of effective pollinators on flowers by male Centris varia bees defending territory, pollinia transfer efficiencies of particular pollinator species also vary between study sites and result in differing reproductive success.
Plant species visited by similar functional pollinator types (pollinator groups sensu Fenster et al., 2004) exhibit suites of floral characters that researchers attribute to convergent floral adaptations to the behaviour, morphology and physiology of pollinators (Johnson and Steiner, 2000; Fenster et al., 2004). Plant–pollinator interactions are frequently regarded as tightly coevolved and highly specialized, and there has been a long-held belief that general evolutionary trends in pollination systems tend towards increased specialization; this concept was enunciated by Stebbins' most effective pollinator principal (MEPP), in which he stated that ‘…the characteristics of flowers will be moulded by those pollinators that visit it most frequently and effectively’ (Stebbins, 1970: 318).
However, over the last decade, as part of the debate about the underlying principles of specialization in pollination systems, this concept has been re-evaluated (Ollerton, 1996; Waser et al., 1996; Johnson and Steiner, 2000; Aigner, 2001; Fenster et al., 2004; Waser and Ollerton, 2006). Data analysis of flower visitors and their occurrence on flowers in plant communities has shown that highly specialized systems (wherein a plant species is associated with a single flower visitor species) are actually rather rare, and that generalized relationships are most common (see Ollerton, 1996; Waser et al., 1996). Running counter to this general phenomenon, most Orchidaceae are considered highly specialized with respect to their flower–visitor interactions (Tremblay, 1992) and coevolution in this family has resulted in many interesting and unusual pollination syndromes (van der Pijl and Dodson, 1966; Faegri and van der Pijl, 1979; Johnson et al., 1998; Pauw, 2006). An analysis of published data on orchid pollinators by Tremblay (1992) revealed that there are on average about three pollinator species per orchid species.
Specialized plant–pollinator interactions are particularly vulnerable to pollen and pollinator limitation, which can affect reproductive success (Burd, 1994; Ashman et al., 2004; Burd et al., 2009). In a review of the reproductive success of nectariferous and nectarless orchids, Neiland and Wilcock (1998) noted that with respect to fruit production, tropical orchid species are on average only one-third as successful as their temperate counterparts, irrespective of whether a species produces nectar or not. Their tropical orchid data set is biased in that it is drawn from Central American epiphytic orchids, a natural consequence of the meagre data available on tropical terrestrial species. Aware of this lacuna, Neiland and Wilcock (1998) write that it is unclear if the commonly encountered low reproductive success of tropical orchids is the result of (1) regional or growth habit phenomena, (2) dissimilarities of temperate versus tropical orchid population structures (as reported by Ackerman 1986), (3) a lack of suitable pollinators in the tropics or (4) increased competition for fewer pollinators. Clearly, it is possible that any combination of the aforementioned factors may be involved and that different factors may be involved depending upon the species and its biological context. As Neiland and Wilcock (1998) recognize, what is needed are more studies on the reproductive biology of tropical terrestrial orchids.
The present study investigates the pollination biology, breeding system, nectar production and floral scent composition of Eulophia alta, a perennial, tropical, terrestrial orchid. Dressler (1981) understood that only medium-sized to large bees could pollinate the large ‘gullet’ flowers typical of Eulophia and he considered Xylocopa bees to be the usual pollinators of this genus. According to the MEPP, floral evolution at the population level is driven by the most effective pollinator (Stebbins, 1970). However, as pointed out by Waser et al. (1996), one of the weaknesses of reproductive biology studies claiming specificity of plant–pollinator interactions is that they are often very restricted temporally and spatially. It is common practice for studies to focus on pollinator interactions of a single population per plant species. However, spatial variation in pollinator composition and abundance is now a well-recognized phenomenon (e.g. Campbell, 1987; Ackerman et al., 1997; Parra-Tabla et al., 2000), and hence it is important to study a species at more than one site whenever possible. With this in mind, the pollination biology and reproductive success of three E. alta populations in Amazonia were studied: (1) on well-drained ground in a ‘campina sombreada’, (2) a steep slope (likewise ‘campina sombreada’) and (3) a dry, sandy locality (‘campina aberta’). To characterize the importance of each flower visitor the relative abundance of the observed flower visitors at each site was evaluated, and additionally, pollinia transfer efficiency of each visitor species was measured by counting the number of flower visits leading to deposition of pollinia on stigmas and/or removal of pollinaria from flowers.
The pantropical genus Eulophia R.Br. ex Lindl. as presently conceived belongs to sub-tribe Cyrtopodiinae (tribe Cymbidieae) and comprises about 230 species that grow in tropical and southern Africa, south-west Arabia, Madagascar, the Mascarenes, tropical and subtropical Asia, Southeast Asia, Australasia and tropical America (Thomas, 1998). Thomas further points out that no satisfactory treatment of Eulophia exists; many of the species are widespread and highly variable, and there are even saprophytic forms (Dressler, 1981). A given Eulophia species can grow in widely different habitats from swamps, to forests, sandy beaches, mountain grasslands and even semi-desert regions. Eulophia alta (L.) Fawc. & Rendle is the sole representative of Eulophia in the New World, probably originating in the Old World, where it grows in tropical sub-Saharan Africa. In the New World it grows from Georgia and Florida (USA) south through the Caribbean and from Mexico south to Paraguay and Argentina. The plant produces an erect, loose raceme bearing about 80–120 medium-sized to large distally purple proximally green flowers (Fig. 1). The stature (erect leaves) of the study plants varied from 0·5 to 1·5 m tall and inflorescences were similar in size. Vouchers of E. alta were deposited in the herbarium of the Universidade Federal do Amazonas, Manaus, Brazil, and ULM.
The pollination biology of E. alta was studied from April to July 1999 in a region 17 km north of Manaus, Brazil. For the comparative study, three different sites on terra firme (non-inundated land) were chosen. E. alta was the only orchid species growing at each site. Distances between populations ranged from 3·5 to 5 km. The plant typically grows on well-drained soil principally in ‘campina sombreada’ (shaded campina) or in ‘campina aberta’ (open campina; Braga and Braga, 1975). Campina is a special, relatively low, woody Amazonian vegetation; it occurs on white sand patches nested within Amazonian forest. The edaphic and floristic composition of campina is distinct from Amazonian forest and other vegetation types. The first study site was a campina sombreada called ‘sitio’, which had been burned 10–15 years ago. The site covered about 2500 m2 and the vegetation was dominated by E. alta and shrubby Piper spp. Weekly phenological estimates during 3 months of investigation showed that 4313 ± 315 flowers were open daily. The second campina sombreada site was a steep (30°) NNW-orientated slope, hereforth called ‘slope’. This site covered about 2650 m2 and had experienced earth slides and other forms of erosion associated with the laying of a pipe several years prior to this study. Vegetation composition here was similar to ‘sitio’ but less dense, and the area was partly shaded by Cecropia sp. trees (a plant indicative of disturbed sites). Daily flower number of E. alta was about half of that at ‘sitio’ (2263 ± 172 open flowers per day). The third site, here named ‘sand’, covered about 2600 m2 and was situated at a dry, sandy locality (‘campina aberta’), which up to about 10 years prior to the study had been a sand mine. At this site, vegetation cover was only herbaceous and plant diversity was low. Although E. alta plants were frequent here, there were on average only 1596 ± 155 flowers in bloom during the study period and therefore flower density was lower than at the other two sites.
Morphometric measurements were made on at least 20 randomly collected flowers. Width of the labellum chamber entrance, length and width of the dorsal sepal, lateral sepal, lateral petals and labellum were recorded (see Fig. 2). Presence and site of osmophores and nectaries were detected by staining whole flowers with neutral red (according to Vogel, 1962). The flower parts were then preserved in 70 % ethanol for further analysis. Preserved material was further dehydrated, embedded in paraffin and sections cut using a microtome. Thin sections were stained with Toluidine O (following Gerlach 1984) and viewed with a light microscope. For scanning electron microscopy (SEM) study of surface morphology, flower parts stored in 70 % ethanol were placed in FDA (formaledhyde dimethyl acetate, 95 : 5) for at least 24 h, dehydrated further, critical-point dried, sputter-coated with gold and then viewed at 5 kV using a Zeiss DSM 942 scanning electron microscope.
Flower longevity is known to be influenced by pollination (e.g. Arditti, 1976; Gori, 1983; Primack, 1985; Proctor and Harder, 1995). In subtribe Cyrtopodiinae, two hard pollinia with a stipe and viscidium form a single pollinarium and the anther sits like a cap on the column (Dressler, 1981). Normally, the anther caps of E. alta fall off during pollinaria removal, or in unpollinated flowers they eventually dry and fall off. To evaluate the influence of pollination events on flower longevity in E. alta, the following six treatments were applied to each of a set of 15 mature flowers: removal of anther cap, pollinarium intact = (A − P + ); anther cap intact, removal of pollinaria = (A + P − ); anther cap and pollinaria removed = (A − P − ); anther cap and pollinaria intact = (A + P + ); pollinated (outcrossed), own pollinaria removed = (Poll + P − ); pollinated (outcrossed), own pollinaria intact = (Poll + P + ). Anther caps and pollinaria were removed on the first day of anthesis, and pollination was carried out on the third day after flower opening. The time until the resupinate flowers either (1) wilted (indicating unfertilized flowers) or (2) became upright as fruits developed (successful pollination) was recorded. To determine the period of stigma receptivity, 368 flowers were marked at the beginning of anthesis, and in the morning and at midday on each day after the onset of anthesis (here defined as flower opening) at least eight flowers were manually pollinated with conspecific pollinia from neighbouring plants. Fruit set was recorded.
In order to investigate the breeding system of E. alta, fruit set of variously manipulated flowers was counted and compared with reproductive success under natural conditions. When pollinator exclusion was necessary, bags were made from nylon stockings pulled over wire cages which were big enough to prevent contact between flowers and nylon mesh. The cages were placed at appropriate height over the inflorescences and tied to the inflorescence stems and stabilizing bamboo sticks that were fixed to the ground. All covers were removed when flowering was finished and inflorescences were tagged with coloured threads until fruit harvest.
To test whether apomixis occurs, 154 flowers (one inflorescence at each site: 79 flowers at ‘sitio’, 37 flowers at ‘slope’ and 38 flowers at ‘sand’) were emasculated and then bagged. Self-compatibility in E. alta was evaluated by ‘hand’-pollinating using forceps. At ‘sitio’, to test for autogamy, five flowers were hand-pollinated with their own pollinia on each of five bagged inflorescences and, to test for geitonogamy, on another five inflorescences five flowers were pollinated with pollinia from other flowers of the same inflorescences. To test for xenogamy, a similar experimental protocol was carried out with pollinia from inflorescences of plants at least 5 m distant. The capacity of E. alta for spontaneous auto-pollination was evaluated at each locality by bagging 12 inflorescences in bud with fine mesh nylon.
To determine the degree of pollinator-mediated outcrossing, all open flowers on 12 inflorescences at each site were emasculated. These inflorescences were left unbagged until fruits matured and then total fruit set was recorded. Additionally, the occurrence of pollinator-mediated geitonogamy was investigated. At each site, 20 flowers on each of eight inflorescences, which were distinct both spatially and temporally, were emasculated; then, after removal of anther caps, the pollinaria of all other flowers on these inflorescences were stained with a 1 : 10 000 solution of neutral red. Pollen transfer of stained pollinia to emasculated flowers was monitored and recorded daily.
To determine the reproductive success of E. alta under natural conditions, fruit set of in total 87 unmanipulated inflorescences at ‘sitio’, ‘slope’ and ‘sand’ was counted. Furthermore, in order to measure the degree of possible pollen limitation, induced cross-pollinations were performed for comparison with natural fruit set. Twelve inflorescences bagged in bud stage at ‘sitio’ were allowed to come into flower then hand-pollinated, using forceps, with pollinia from plants at least 5 m distant.
Besides fruit set, other reproductive modalities that might differ due to different pollination modes, such as outcrossing versus auto-pollination, are seed set and seed viability. However, as there are millions of small seeds per fruit in orchids, accurate counts of seed numbers are difficult to obtain (Proctor and Harder, 1994). Based on data from Proctor and Harder (1994), Neiland and Wilcock (1995), and Nazarov and Gerlach (1997), Johnson and Edwards (2000) found that in orchids, ovule number correlates with the number of pollen grains typically deposited onto a stigma after an effective pollinator visit; therefore, the expected pollen–ovule ratio should be about one if entire solid pollinia are deposited on stigmas, as borne out by the study of Nazarov and Gerlach's (1997) for Coryanthes senghasiana. Given this, we assumed that in E. alta, each pollination event resulting in fruit set is sufficient to produce full seed set because whole pollinia are transferred. For this reason, we decided to measure seed quality, here treated as the number of seeds containing an embryo, of fruits produced by different hand-pollination modes. Orchid seeds have no endosperm; they consist mainly of the embryo and a translucent testa. Therefore, embryos are readily visible when seeds are viewed under a light microscope (Leitz Ortholux II). Ten flowers on each of ten bagged-in-bud inflorescences were subjected to the following treatments: hand-pollinated with pollinia from donor plants (at least 5 m distant), pollinated with pollinia from other flowers of the same plant, and self-pollinated with their own pollinia. For each of the treatments, ten fruits (one per plant) were randomly chosen and from each fruit 500 seeds were taken and checked for the presence of an embryo. The percentage of seeds with an embryo (hereafter ‘embryo content’) was calculated per fruit and averaged per treatment.
Flower position in an inflorescence can be a factor in fruit set and seed number as a result of resource allocation (Corbet, 1998; Medrano et al., 2000; Vallius, 2000). To study the effect of flower position with respect to sexual reproduction, fruit set and embryo content of flowers from the lower, middle and upper third of ten inflorescences were evaluated after hand-pollination (outcrossed, minimum 5 m distant) of all flowers. Total fruit set was recorded and ten randomly chosen fruits of the lower, middle and upper third of each inflorescence were examined for embryo content.
In order to assess the composition and abundance of flower visitors a total of 229 h were spent observing the plants at the three sites combined. Observations were made at single inflorescences or small groups of inflorescences during 2–4-h periods at different times of the day, from early morning (0630 h) until late afternoon (approx. 1800 h). Observed flower-visiting insects were counted and their behaviour was noted. Most of the insect species were identified on the wing. Specimens of each species were caught for identification. All flower visitors collected were sent to specialists at the Universidade Federal do Paraná, Curitiba, Brazil, for identification, and are deposited in the entomological collections of INPA, Manaus and in the insect collection of the Herbarium ULM at the University of Ulm, Germany.
Relative abundance (RA) is given as the percentage of observed flower visits of each insect species at a particular site (100 % = all observed flower visits of all insect species). Pollinia transfer efficiency (PE) of each pollinator species at a particular site is given as the percentage of observed ‘successful’ flower visits (visits with deposition of pollinia on stigmas and/or removal of pollinaria from flowers) divided by all observed flower visits of that insect species. The resulting total pollinia transfer effectiveness (TE) of pollinators at a particular site was calculated as the number of ‘successful’ flower visits of a particular pollinator species over a period of 100 h of observation.
At each site, reproductive success of male and female function was evaluated on randomly chosen inflorescences. In order to consider only outcrossing pollinia transfers, pollinaria of a chosen inflorescence were stained with a neutral red solution (1 : 10 000) after removal of the anther cap. Following Singer and Cocucci (1997), donor efficiency, i.e. the ratio between male and female function, was calculated by dividing female reproductive success (i.e. percentage of flowers pollinated with unstained pollinaria) by male reproductive success (i.e. percentage of flowers having stained pollinaria removed). At each site throughout the flowering season, a total of eight inflorescences of differing flowering times were examined.
Flowers of E. alta have no spurs and no large nectar droplet was visible inside or outside the flowers. However, most flower visitors were observed to search for something at the labellum tip, and traces of fluid were found here, but never inside the chamber. Nectar production and composition over time were investigated by rinsing the labellum tip (the chamber could not be rinsed as this caused too much tissue damage) of 96 flowers with 70 % ethanol and then collecting the liquid for later analysis. Samples were taken from bagged 1–12-d-old flowers. Nectar samples were analysed by high performance liquid chromatography (HPLC) as described in Witt et al. (1999) using a Waters High Performance Carbohydrate Column supplemented with a complementary Sentry Guard precolumn, a 510 pump and a 717+ autosampler. Before injection all samples were dried in a vacuum centrifuge (Savant Speed Vac SC 100) and diluted with 200 µL water. Injection volume was 10 µL and samples were eluted with an acetonitrile–water mixture (71 : 28) at a flow rate of 1·4 mL min−1 and temperature of 35 °C. Glucose, fructose and sucrose were detected using a Waters 410 refraction index detector and quantified by means of Waters ‘Millenium’ software.
Floral scent was collected following the method of Jürgens et al. (2000). To prevent flowers being visited and pollinated by insects, five individuals were bagged (as described above for evaluation of the breeding system) when flowers were still in bud. The volatile compounds of a single flower per inflorescence were collected using a headspace technique. To accumulate the floral scent and prevent wind drift, each flower was covered with a glass jar (10 cm in diameter and 10 cm long) and the scented air surrounding the flower within the jar was pumped (200 ml min−1 for 2 h, using a battery-operated membrane pump, G12/01 EB, Gardner Denver Thomas GmbH, Inc., Puchheim, Germany) into a narrow glass tube packed with 150 mg adsorbent [TenaxTA (2,6-diphenyl-p-phenylene oxide), mesh size 60–80 and Carbotrap™, mesh size 20–40]; thereafter, the adsorbed volatiles were extracted with 1 mL of acetone into glass vials, which were then hermetically sealed. The adsorbent tubes were preconditioned by washing with acetone and dried at 250 °C.
GC-MS analyses were performed using an ion trap instrument wherein both ionization and mass analysis occur in the same chamber. The GC-MS (Saturn 2000, Varian, Walnut Creek, CA, USA, and 8200 CX auto injector) was equipped with a fused-silica capillary column (CP-Sil 8CB-MS, 0·25 µm film thickness, 30 m long, 0·25 mm inner diameter, Varian). The 1-μL samples were introduced using a 1079 injector; an injection split ratio of 20 was applied. The temperature of the column was programmed to rise from 60 to 260 °C, at a rate of 8 °C min−1; the temperature of the injector, transfer line and ion trap was held at 200, 175 and 200 °C, respectively. Settings of the MS were: mass spectra 70 eV (in EI mode) and scan range was 40–650 amu at a scan rate of 1 scan min−1. The GC-MS data were processed using the Saturn Software package version 5·2·1. Component identification was processed using the NIST 05 mass spectral database (NIST algorithm) and confirmed by comparison of retention times with published data (Jennings and Shibamoto, 1980). Identification of some compounds was also confirmed by comparison of mass spectra and retention times with those of authentic standards.
While preparing flower parts for preservation in alcohol (at a distance of at least 20 m from a Eulophia population), 13 bees (three Centris rubella, five C. varia and five unidentified individuals) were observed to be attracted exclusively to the labellum tissue, especially to parts of the labellum chamber (ten of 13 bees). To confirm this chance observation, the labellum tip (Fig. 3A), labellum base (Fig. 3B), column, sepals and petals were tested separately for their potential attractiveness to flower visitors in the field. Flower parts cut from ten flowers were put on Petri dishes and the open dishes were placed 100 cm above ground at a distance of 40 cm from each other in a natural population of E. alta. Distance to the nearest growing Eulophia was 2·6 m and the experiment was carried out for 2 d between 1100 and 1400 h as prior observation of the daily activity of flower visitors had shown that all species visited flowers more or less around midday.
Data sets were tested for normal distribution and homogeneity of variances and further tests were applied accordingly. Data on flower longevity and embryo content of seeds produced from different pollination experiments were submitted to ANOVA and least significant difference (l.s.d.) tests for post-hoc comparison of means. A Kruskal–Wallis test was used to analyse variation of seed embryo content with respect to position in inflorescences. Comparison of inflorescence flower numbers among treatments and populations was done with ANOVA followed by Tukey's HSD test for unequal numbers. For fruit set, Kruskal–Wallis ANOVA was used to test for pair-wise differences among treatments and populations, followed by a Tukey–Kramer test for non-parametric data as a post-hoc test. Male and female function and the resulting donor efficiencies of inflorescences were analysed with MANOVA and l.s.d. tests for a post-hoc comparison between sites. All statistical analyses were performed with STATISTICA 5·1.
The erect racemose inflorescences of E. alta bear about 100 flowers (average of sample sets ranged from 86 to 123 per inflorescence, Table 1). The three sepals (19–24 × 7–8 mm) and two petals (15–17 × 5·5–7 mm) are nearly equal in size and shape. The free labellum (19·5–22 × 10·5–12 mm) is trilobate; the central lobe serves as a landing platform for flower visitors. Together with the two lateral petals, the two lateral lobes of the labellum embrace the column, forming a labellum chamber, into which flower visitors may enter. Inside the labellum chamber is a ‘narrow pass’ approx. 8·5–9 mm wide and approx. 6 mm high. However, the lip is to some degree flexible and the weight of flower visitors may depress it, thereby increasing the size of the labellum chamber entrance. With the exception of the labellum, perianth segments are pale purplish pink. Most parts of the labellum are also pale pink, but the tip and base are dark purplish pink. The colouration of the column is likewise pale: thus, the labellum contrasts sharply with the surrounding, less conspicuous, floral parts. Neutral red staining of flowers revealed no (additional) colour marks on sepals and lateral petals, suggesting that either flowers of E. alta have no osmophores or osmophores occur in the naturally dark-coloured parts of the labellum and are therefore not apparent by this technique. Anatomical study of the labellum shows that this appendage has three epidermis cell types. This is in contrast to the sepals and lateral petals, which have only one. Inside the labellum chamber at the base of the lip, besides the commonly encountered flat epithelial cells, there are papillose, tube-like cells with thin cell walls (Fig. 3B), which may function as osmophores (compare, for example, Vogel, 1962; Davies and Winters, 1998; Teixeira et al., 2004). Furthermore, the adaxial surface of the lip is covered with cone-like, well-cutinized cells. Stomata-like openings were found near the vascular bundles at the labellum tip (Fig. 3A); these openings might indicate the presence of intrafloral nectaries (but see Daumann, 1970).
Unpollinated flowers having an intact pollinarium and anther cap began wilting on the 16th day of anthesis, and lasted at most for 21 d before finally abscising. Experimental manipulation of flowers showed that pollinaria removal and pollination had the most significant effect on flower longevity, followed by anther cap removal. Flowers pollinated on the third day of anthesis, stopped nectar secretion and scent production within 24 h, and wilting of petals and initiation of fruit development were visible by the fifth day, independent of pollinaria removal (Table 2). It took up to 69 d for fruits to mature. Differences in flower longevity between all groups were highly significant (ANOVA: F5,84 = 1398·77, P < 0·001; l.s.d. test: P < 0·001) except between pollinated flowers with and without their own pollinaria (P = 0·42). Stigma receptivity was similar to that of whole flower longevity of unpollinated flowers. Stigmas were receptive up to the 17th day of anthesis and set 100 % fruits after pollination (284 flowers between 1 and 17 d old). Thereafter, stigma receptivity decreased steadily to 12·5 % fruit set for flowers pollinated on the 21st day of anthesis (84 flowers between 18 and 21 D old).
None of the emasculated bagged flowers set fruit, whereas all flowers that were hand-pollinated with pollinia of the same flower or flowers of the same inflorescence set fruit, as did flowers pollinated with foreign conspecific pollinia, indicating that our study plants are not apomictic but fully self-compatible. E. alta was found to be partially autogamous. Fruits developed in 4·4 % of the bagged flowers at ‘sitio’, 4·9 % at ‘slope’ and 6·1 % at ‘sand’, confirming that some fruit production takes place without a vector when anther caps dry and fall off, and pollinaria bend or fall down onto the stigma of the same flower; hence, auto-pollination or vectorless self-pollination (Catling, 1990) occurs. Furthermore, variation in auto-pollination among sites was statistically significant (Kruskal–Wallis ANOVA: H2,36 = 10·03, P < 0·01).
All hand-pollinated flowers used in experimental manipulations (xenogamy, geitonogamy and autogamy) for seed quality studies set fruit. Nevertheless, seed quality, as defined by seed embryo content, differed significantly between treatments (ANOVA: F2,27 = 132·31, P < 0·001). Differences between selfing (autogamy = 85·4 %, geitonogamy = 86·8 %) and outcrossing (98·2 %) were highly significant (l.s.d. test: P < 0·001). After pollination of all flowers with outcrossed pollinia, seed quality decreased significantly (Kruskal–Wallis test: H2,30 = 23·39, P < 0·001) in an acropetal direction (Table 3). There was no fruit abortion on the lower third of the inflorescence, and the further from the base the greater the degree of abortion, culminating in 24 % abortion in the upper third of the inflorescence. Seed size showed a similar trend with seeds in lower fruits being larger than seeds in fruits at the top of the inflorescence (data not shown). Furthermore, in the self-compatibility experiment, when only 25 flowers of an inflorescence were outcrossed by hand, they all set fruit (100 % fruit set), whereas when all flowers in 12 inflorescences were outcrossed by hand, fruit set was lower (93·3 %, Table 1).
Eulophia alta is exclusively melittophilous. At the present sites anthophorid bees are the main flower visitors and most effective pollinators of the many-flowered inflorescences. The highest flower visitor frequencies were recorded on approx. 3–5-d-old flowers. In total, 19 species representing six Hymenopteran and two Lepidopteran families were observed visiting the flowers (Table 4). The relative abundance of flower visitors and pollinators, and pollinator pollinia transfer efficiencies at the different localities are given in Table 4. The bee species Centris inermis and Centris bicornuta could not be distinguished from each other by eye in the field, and for this reason the data in Table 4 give the sum of the two species. Species richness of flower visitors at the three sites was quite different. At ‘sitio’ 19 species of flower visitors and pollinators were observed, whereas there were nine species at ‘slope’ and five at ‘sand’. Field observations showed that flowers are visited until the 14th day of anthesis. Most individuals of insect species recorded as ‘nectar-seeking’ (Table 4), for example all anthophorid bees, were observed extending their proboscis and probing extensively at the labellum tip. Regularly, anthophorid bees were observed secreting a substance onto the labellum tip and thereafter imbibing a fluid. However, not all nectar-seeking insects entered the labellum chamber after probing the labellum tip. In contrast to the often time-consuming probing at the labellum tip, when insects ventured into the labellum chamber, they remained there for only a few seconds. As entering of the labellum chamber is required for effective pollen transfer, insect species that were never seen to enter the labellum chamber are treated as flower visitors, but not pollinators. For example, C. flavifrons was unable to enter the chamber due to its relatively large body size and weight, and although members of the genera Euglossa, Eulaema, Eudynereus and Ceratina probed at the labellum tip, they never entered the labellum chamber. Only insects with an appropriate body size (body length between 10 and 20 mm) are able to fit into the labellum chamber; upon entry they made contact with the stigma and anther (Figs 1 and and2,2, Table 4). When appropriately sized flower visitors entered the labellum chamber, they usually were effective pollinators (i.e. deposited pollinia and/or removed pollinaria). In the case of two small bees, Megachile sp. and Centris sp., some visits into the labellum chamber were ineffectual in pollinating the flower. Pollinaria deposition on insects was usually on the posterior portions of the thorax, especially the scutellum and/or metanotum; placement varied slightly among species depending on body size and shape. When pollinators backed out of the labellum chamber, pollinia carried by the bees were transferred from their body to the stigma, and subsequently a pollinarium from the anther was attached to the insect. This sequence ensured that pollinator-mediated selfing (autogamy) by deposition of pollinia onto the stigma of the same flower during a single visit did not occur. Moreover, pollinator-mediated geitonogamy was rare (see Table 1), yet features known to prevent geitonogamous pollination, such as protandry, bending of pollen stalk or anther cap retention (see Catling and Catling, 1991), were not observed in E. alta.
Relative abundance of flower visitors differed between sites (Table 4); on average, species with the highest relative abundance were five Centris species (C. minuta, C. flavifrons, C. rubella, C. varia and C. sp.), and Xylocopa muscaria, followed by C. spilopoda and Megachile sp. Summing the total number of effective visits (TE/100 h) over all sites, C. minuta proved to be the most important pollinator followed by C. rubella, C. varia, X. muscaria and C. spilopoda. The strong territorial behaviour of male C. varia individuals (which also visited flowers effectively) at ‘sitio’ was striking. In the early morning, these males established their territory in parts of the Eulophia population where there was high flower density; during the day, they attacked all flower visitors within their territory except for female C. varia bees. Conversely, flower-visiting C. varia females were disturbed by ‘mating attacks’ by conspecific males. The territorial behaviour of C. bicornuta and C. inermis was less ferocious and these bees often lost in competitions with C. varia for territory. Other Centris species such as C. minuta, C. spilopoda and C. sp. were observed to establish their territories at the margins of or outside the Eulophia population, on plants of E. alta and an unidentified Piper species.
Orchid populations growing in a campina sombreada (‘sitio’ and ‘slope’) received more flower visits per hour than those in campina aberta (‘sand’; Table 4). Relative pollinator abundance, however, was lowest at ‘sitio’, where the highest proportion of non-pollinating insects was found. Number of effective pollinator visits was highest at ‘slope’ (60·5 effective flower visits per 100 h) and lowest at ‘sand’ (33·9 effective flower visits per 100 h). Natural levels of fruit set showed a significant variation among populations (Kruskal–Wallis ANOVA: H2,77 = 49·93, P < 0·001) with 10·9 % at ‘sitio’, 19·8 % at ‘slope’ and 14·7 % at ‘sand’. Pollinator-mediated outcrossing reached 11·6 % at ‘slope’, and 5·6 and 4·1 %, respectively, at ‘sitio’ and ‘sand’ (Kruskal–Wallis ANOVA: H2,36 = 22·93, P < 0·001). In contrast, experimental outcrossing by hand-pollinations at ‘sitio’, the only site studied, produced 93·3 % fruit set (Table 1).
Differences in flower number in inflorescences used for evaluation of donor efficiencies in the three populations proved to be insignificant (ANOVA: F2,21 = 1·21, P = 0·32). Variation of male and female reproductive success and donor efficiency of tested inflorescences was highly significant between sites (MANOVA: Wilks' λ = 0·29, Rao's R6,38 = 5·50, P < 0·001). Differences in subsequent tests were found to be significant for male (F2,21 = 9·58, P < 0·01) and female reproductive success (F2,21 = 8·34, P < 0·01), but not for the resulting donor efficiency (F2,21 = 1·45, P = 0·26). In the investigated inflorescences, most pollinaria were removed at ‘slope’, followed by ‘sitio’ and ‘sand’ (Table 5; 24·6, 21·0 and 12·8 %, respectively). Female reproductive success showed a similar ranking order, i.e. 5·6 % pollinated flowers at ‘slope’, 3·0 % at ‘sitio’ and 2·4 % at ‘sand’. On average, inflorescences having the highest donor efficiencies were at ‘slope’ (0·23); at ‘sand’ and ‘sitio’ donor efficiencies of inflorescences were lower (0·16). The overall donor efficiency of 0·23 at ‘slope’, followed by ‘sand’ and ‘sitio’ with 0·19 and 0·14, respectively, parallels ranking of natural fruit set. At all three sites, there were clearly more pollinaria removals than effective pollinations.
Concomitant with anthesis, E. alta produces small amounts of nectar on the labellum tip. The nectar seems to dry immediately after secretion and is visible only under a microscope where it appears as a thin film over the labellum tip. To imbibe this film of dried sugar, insects usually lick at the labellum tip, or as suggested by the present observations, they first secrete a substance onto the surface then remove the liquid with their proboscis. In bagged flowers, nectar was found up to 11 d from the start of anthesis, with the highest sugar content on the third and fourth days of anthesis (Fig. 4). Composition analysis showed that the nectar is dominated by glucose and fructose, and can be classified as ‘hexose-rich’ according to Baker and Baker (1983).
Flowers emit what was to our nose a sweet/floral scent from the base of the labellum. Scent emission begins when a flower opens and peaks on the ninth day of anthesis. Even on the 17th day, a weak scent was detectable. A summary of the floral scent chemistry showing the relative amounts of volatiles, and their distribution among the main chemical compound classes, i.e. fatty acid derivatives, benzenoids, phenyl propanoids, isoprenoids and nitrogen-containing compounds, is given in Table 6. Floral fragrance analysis of E. alta yielded 40 compounds, 33 of which were identified. The floral scent of E. alta is predominantly composed of monoterpenes (40·4 %) and benzenoids (38·7 %). The main volatile compounds were jasminaldehyde (16·5 %), followed by cis-β-ocimene (13·4 %) and benzyl acetate (9·8 %).
The bioassay experiment revealed that only floral tissue from the labellum chamber and labellum tip was found to attract flower visitors. During a total of 6 h of observation, 18 insects were recorded landing on the tissue of the labellum chamber (four X. muscaria, seven C. minuta, six C. varia and one C. sp.), and three were observed to land on the labellum tip (one X. muscaria, one C. minuta, one C. varia). No insects were observed to land on tissue from the column, sepals or petals.
Throughout its range, Eulophia alta is a weedy, colonizing plant of disturbed environments. Despite being a colonizing species, it is functionally specialized and at the sites detailed here is seemingly dependent on a particular functional pollinator type. Although E. alta was studied at three localities that varied in degree of perturbation, they all fall into the general habitat variation found in its range. The flower in E. alta has a form that restricts entry to insects of a distinct body size and form. Even so, 19 different insect species belonging to Hymenoptera and Lepidoptera were observed visiting E. alta flowers. Of these, only nine bee species had a body size that allowed them to fit between the column and labellum, and thus potentially to act as pollinators. Except for one megachilid species (of minor importance), in E. alta all effective pollinators were anthophorid bees. With respect to the most important pollinators, at each site slightly differing subsets of 3–5 anthophorid bee species accounted for more than 90 % of the effective visits. If the combined data from all three sites are averaged, four Centris bee species (C. minuta, C. varia, C. rubella and C. spilopoda) and X. muscaria account for more than 80 % of the effective visits over 100 h of observation. A recent phylogenetic interpretation (Straka and Bogusch, 2007) places Centris and Xylocopa bees in different subfamilies of the Apoidea (Apinae and Xylocopinae, respectively; see also Michener, 2000), suggesting that these genera are not closely related; they do, however, form a guild united in having a functional type fitting the morphological and behavioural requirements suitable to the pollination of E. alta. Eulophia is an Old World genus that has its centre of diversity in Africa. The populations of E. alta in the New World represent the only expansion of the genus into the Americas; it is probable that they have spread from Africa as Centris bees are restricted to the New World (Michener, 2000) and together the pollinating species overlap much of the New World range of E. alta. We consider it likely that some of the study Centris species pollinate E. alta outside of the study area, and possibly other Centris species elsewhere in the Americas. Unlike Centris, Xylocopa is found on most continents and achieves maximum species diversity in Africa (Leys et al., 2002) and may pollinate E. alta there. The fact that Centris and Xylocopa pollinate E. alta in the New World is probably related to their morphological similarity to the originally pollinating bees in the Old World (likely species of Xylocopa, Anthophora and Amegila). Moreover, it is thought that African and South American subgenera of Xylocopa are not particularly close (Leys et al., 2002), strengthening the argument that the relationship of the pollinating bee genera is one of form and behaviour rather than close phylogenetic affinity.
Considering effective visits (TE; Table 4), it would seem that E. alta is rather more specialized with respect to pollinators than might be concluded by casual inspection of data on insect visitors and number of pollinator species. Similarly, Lock and Profita (1975) found that a West African species of Eulophia, E. cristata, a common terrestrial orchid, is visited by butterflies, flies, wasps and bees; however, only one of two pollinating Xylocopa species, X. olivaceae, was a regular visitor and appeared to be the usual pollinator.
Few studies on floral biology in Orchidaceae have focused on more than one population per species (but see, for example, Borba and Semir, 2001; Brys et al., 2008); we find that not only local pollinator spectra, but also pollen donor efficiency and pollinia transfer efficiency of particular pollinator species (see Table 4; e.g. C. varia, C. spilopoda and X. muscaria) vary between study sites and result in differing reproductive success (Table 1). Due to their abundance and efficiency, the most important pollinators at ‘sand’ were C. minuta, C. varia and C. sp., whereas at ‘slope’ C. minuta, C. rubella, C. spilopoda and X. muscaria played the most important roles, and at ‘sitio’ the C. bicornuta/inermis complex, C. varia, C. rubella, C. minuta, C. spilopoda and X. muscaria were most effective with respect to total pollina transfer. Our results show that flowers at the sites ‘slope’ and ‘sand’ are visited by a smaller spectrum of insects than flowers at ‘sitio’. The sheltered environment and the more favourable conditions at this site, which result in a highly diverse and relatively dense vegetation cover, may be the reason for the generally high abundance and diversity of insects, and in turn high species richness of flower visitors and most effective pollinators observed at this site. Many insect species show limited foraging ranges and are sensitive to changes in vegetation structure, habitat type, size and connection to other areas (Janzen, 1971; Powell and Powell, 1987; Johnson and Bond, 1992; Parra-Tabla et al., 2000). Disturbance may result in the loss of alternative food sources, nesting habitats and refuges against predators for pollinators (Aizen and Feinsinger, 1994). Differences in pollinator availability between populations may lead to variations in reproductive success (Campbell, 1987). It is also possible that differing flower numbers and densities at the three sites (sitio > slope > sand; see site descriptions) might affect flower visitor frequencies per flower and resulting reproductive success. Bullock et al. (1989) showed that due to differences in plant and flower number, abundance and diversity of pollinators differ significantly between sites separated by no more than 300 m.
If the number of flower visits by pollinator species translated directly (or nearly so) into greater fruit set, then we would expect the highest natural fruit set to occur at ‘sitio’, where 549·8 of the observed flower visits (per 100 h) were made by pollinator species; at ‘slope’ 447·1 visits were from potential pollinators and 197·5 visits at ‘sand’. In fact, highest natural fruit set was found at ‘slope’, and lowest in ‘sitio’. It would appear that relative pollinator abundance, which was lowest at ‘sitio’ (67·4 %, and where there was the highest proportion of non-pollinating insects) and highest at ‘sand’ and ‘slope’ (97·5 and 91·5 %, respectively), also plays a role. With regard to absolute effectiveness, the locality having the highest number of pollinia transfers was ‘slope’ (60·5 per 100 h). This result accords well with donor efficiency (DE), which is of particular importance in explaining differences between sites as natural fruit set parallels DE ranking of sites. If we relate for pollinator species the number of effective visits per 100 h to the number of observed flower visits per 100 h it becomes apparent that pollinators at ‘sand’ were most efficient as 15·2 % of their visits (33·9 of 197·5 pollinator visits) led to successful pollination. Pollinator species at ‘sitio’ were least successful with 47·8 of 549·8 visits (6·8 %) being effective for pollinia transfer (for comparison: ‘slope’ 60·5 of 447·1 visits = 10·3 %). The present data suggest that differences in reproductive success in the investigated E. alta populations cannot be explained by the absolute abundance of pollinators alone. From the field observations it was clear that disturbance of potentially effective pollinators during flower visits was one reason for differing pollinia transfer efficiencies. This was particular the case at ‘sitio’, where after potential pollinators had landed and fed for a short time on labellum tip nectar, they often fled from flowers before entering the labellum chamber because of harassment by male C. varia bees. It is known that male Centris bees defend territories around food sources using pheromones and show aggressive behaviour (Frankie et al., 1980; Vinson et al., 1995). The observed territorial behaviour of male C. varia individuals at ‘sitio’ can explain the relatively low number of effective visits (TE per 100 h) of most other pollinating bee species at this site as compared with the other sites. At the other two sites, C. varia showed no clear territorial behaviour, possibly due to less favourable conditions, such as less vegetation cover, fewer total flowers and lower flower density, which rendered establishment and defence of territories untenable.
Overall, the present results are compatible with the recognized trend among tropical orchids to produce little fruit; commonly less than 50 % of flowers produce fruit (Neiland and Wilcock, 1998). We found that after pollination, there was a rapid cessation of nectar and flower scent production followed by wilting; this phenomenon, besides being energetically economical, may prevent a waste of limited pollinator services to already pollinated flowers. Casper and La Pine (1984) propose that when pollen is presented in pollinia, an individual flower benefits little in either receiving or donating pollen by a visit from a second pollinator for the simple reason that the first pollinator to visit the flower is likely to remove all pollen and deposit sufficient pollen for full seed set. Initially, however, when a potentially effective pollinator visits a flower, there is no deposition, only attachment; but as an orchid pollinator usually visits many conspecific flowers, this initial effect is swamped out. The pollination-induced wilting observed in E. alta tends to support the hypothesis of Casper and La Pine (1984) and correlates with findings in many other orchids (e.g. Luyt and Johnson, 2001, and references therein). The low donor efficiency ratios at the three study sites indicate that there were distinctly more pollinarium removals than pollinations. Relatively low pollinator transfer efficiency and an extremely low donor efficiency ratio help to explain the overall low reproductive success of E. alta. Pollinaria losses from pollinators can occur in a number of ways, the most obvious being when bees remove them by grooming (see also Janzen, 1980).
Eulophia alta is highly self-compatible but not apomictic, and auto-pollination as defined by Catling (1990) occurs. This finding confirms the personal communications of Salazar and Ackerman given in Catling and Catling (1991); auto-pollination in E. alta was described to be by bending of the caudicle, stalklett or pollen mass (see Salazar pers. comm. in Catling, 1990). In the present study, the nylon covers may have reduced the incidence of self-pollination because they shade the flowers and may retard anther cap drying and abscission. This would explain why combined experimental fruit set by autogamous, geitonogamous and xenogamous pollination, especially at ‘sand’, were markedly lower than natural fruit set (Table 1). At ‘sand’, shading trees and shrubs were absent and under natural conditions the proportion of auto-pollination at this site might well be higher than that recorded by the simulations; additionally, auto-pollination might contribute to the relatively high natural fruit set at this site. The auto-pollination mode observed in E. alta differs from that described by Williamson (1984) for eight African Eulophia species. In the mostly small, inconspicuous-flowered species which he studied, Williamson found that auto-pollination occurred by stigma contact with outgrowths of the pollinia.
With respect to reproductive success, the present data strongly suggest that pollinator-mediated outcrossing (11·6 % at ‘slope’, and 5·6 and 4·1 % at the other sites) cannot solely account for observed natural fruit set, which was markedly higher (19·8 % at ‘slope’, 10·9 % at ‘sitio’, 14·7 % at ‘sand’). Yet, although capable of auto-pollination, E. alta did not seem to be especially adapted to selfing. Fruit set was relatively low compared with typically pollinator-independent orchids (see Neiland and Wilcock, 1998, and references therein) and the resultant seeds were significantly less viable than outcrossed seeds (similar results were reported for E. cristata by Lock and Profita, 1975). Moreover, fruit set in E. alta following hand-pollination (93·3 %) was much higher than fruit set under natural conditions (maximum of 19·8 %).
Although E. alta produces minute amounts of nectar, the quality and quantity is sufficient to be attractive to flower-visiting insects, and to promote pollination. The fact that there are many flowers per population may compensate somewhat for the ‘poor’ nectar reward of individual flowers. Moreover, individual plants as well as entire populations are a reliable nectar source for months. During the 3 months of observation, the E. alta populations studied were in flower constantly. Individual inflorescences had about 100 flowers, the last floral buds opening on average 28 d after the first. Due to extreme flower longevity single inflorescences presented open receptive flowers for about 47 d. Thus, the flowering phenology of E. alta can be classified as ‘steady state’ (Gentry, 1974), wherein a plant produces a few flowers per day over an extended period of time (usually a month or more).
It has been argued that absence or low levels of floral rewards in many orchid species might be a means of reducing geitonogamy (see Johnson and Nilsson, 1999, and references therein). The minute nectar reward of individual E. alta flowers in combination with age-based differences in quality and quantity, therefore, might be a way of avoiding excessive pollinator-mediated geitonogamy by encouraging movement to other inflorescences for optimal nectar rewards while at the same time maintaining low-level attractiveness of unpollinated flowers.
Neiland and Wilcock (1998) reviewed the literature and showed that average fruit set of rewarding orchids is higher than that of deceptive ones. Ackerman et al. (1994) have shown that even a small nectar reward in Comparettia falcata enhances pollinator attraction and fruit set. Neiland and Wilcock (1998) suggest that the evolution of nectar production within Orchidaceae has been the most frequent means of escaping reproductive limitation of low pollinator frequencies. In Eulophia cristata, Lock and Profita (1975) reported flower visits by nectar-seeking bees and butterflies although the flowers are scentless to the human nose and offer no free nectar in the spur. They assumed that these visits are stimulated by the promising appearance (colour, shape) of the flowers; notwithstanding this, overall only relatively few flowers are visited by pollinating bees, because bees quickly learn that nothing is to be gained from such visits. In E. alta a nectar reward is offered, but the low fruit set of E. alta indicates that this still does not translate into high fruit set, probably because nectar is exposed on the labellum so that imbibing does not equate with pollination. In this connection it should be mentioned that Kullenberg (1961, p. 281) discovered that the terrestrial African species, E. horsfallii, which exposes no superficial nectar, was visited by food-linked Xylocopa bees that tear holes with their mandibles into the labellum base and lick. Such observations are highly suggestive and may, in part, explain the selective forces that have led to the evolution of superficial nectar in E. alta. It is still unclear why flower visitors of E. alta enter the labellum chamber at all if nectar is only offered on the labellum tip. It is possible that the similarity of colours (both the tip and the base of the labellum are deep purplish pink) and odour emission at the labellum base repeatedly deceive them into entering the labellum chamber in search of more nectar. It seems that pollinators recognize and are attracted to flowers signalling higher nectar content. There was a tendency among observed nectar-seeking insect species to visit flowers most frequently on the third day of anthesis (our unpubl. res), the day when nectar sugar amount in bagged flowers is at a maximum. After the third day the amount of sugar in bagged flowers decreased, indicating resorbtion of nectar. However, we do not know how nectar consumption by visitors affects nectar secretion in E. alta. As in other species, it may well be that visited flowers secrete nectar for much longer than bagged flowers. Decreasing nectar sugar content and amount in older flowers may direct pollinators to younger flowers, which are less likely to have donated pollen or received pollinia, or to different inflorescences. Further studies are needed to clarify how flower visitors perceive floral signals and if they recognize corresponding reward quality from a distance.
In many orchids, floral fragrances are a major attractant for pollinators (van der Pijl and Dodson, 1966). Besides being a reward in and of itself (for perfume-collecting male euglossine bees; see, for example, Williams and Whitten, 1983; Gerlach and Schill, 1991; Gerlach, 1995), floral fragrances may influence pollinators in several ways: they may attract from a distance, they may trigger search behaviour, and they may act as cues for alighting or probing, or as nectar guides (Faegri and van der Pijl, 1979). Several of these functions may be involved in the case of E. alta. As summarized by Dobson (1994) for many bees, visual cues operate primarily at long range (within several metres; but see Chittka and Raine, 2006) whereas flower volatiles are generally considered effective in orientating bees at short range within 1 m or less. However, for bees such as male Euglossines, which can trace odours over considerable distances (up to 1 km, Dressler, 1982), it is well known that fragrances may be more important for pollinator attraction over long distances than visual cues (Bergström, 1978; Williams, 1983). The many-flowered inflorescences of E. alta are certainly visually attractive but the attraction of anthophorid bees to seemingly inconspicuous cut flower parts at a distance of 20 m from the E. alta population strongly suggests that floral scent plays an important role in attraction over long distances. The floral scent of E. alta consists largely of aromatic alcohols, aromatic esters, phenyl propanoids and monoterpenes. Many of the floral scent compounds found in this species have previously been reported as floral constituents in other orchids (see Knudsen et al., 2006) and many are known from euglossine-pollinated species (Williams and Whitten, 1983). Some of the floral scent compounds (alpha-pinene, benzyl acetate, benzyl benzoate, eugenol, 2-phenylethyl alcohol) have been shown to elicit an electroantennography response in experiments with male Euglossine bees (Eltz and Lunau, 2005). Most of these compounds are even known to elicit behavioural responses in Euglossine males (see, for example, Williams and Whitten, 1983). However, in contrast to the information that is available for Euglossine bees, odour data or observations on Centris bee-associated flowers is scarce. The sweet floral fragrance of Passiflora alata, a species pollinated by different Centris and Xylocopa species, is dominated by aromatic alcohols (e.g. benzyl alcohol, 4-methoxy phenol, 2-phenylethyl alcohol) and monoterpenes (e.g. nerol, geraniol) (Varassin et al., 2001). Varassin et al. (2001) observed that bees visited the flowers 1 h after anthesis, after the beginning of scent emission. This supports the idea that scent alone might be an important attractant for Centris bees. More research is needed to understand the role of floral fragrances in Anthophoridae bee attraction.
Floral traits that increase attractiveness to pollinators are predicted to evolve through selection on male rather than female function (see Vaughton and Ramsey, 1998, and references therein). Several authors have found a relationship between visitation rate or male fitness curve (MFC) and flower number or floral display (e.g. Rademaker and DeJong, 1998; Ohashi and Yahara, 1998; Vaughton and Ramsey, 1998), suggesting that visitation is often proportional to the number of flowers per plant. In general, visitation rate or MFC seems to be a decelerating function of floral display. In the special case of orchids, Maad (2000) found for Platanthera bifolia that both male and female fitness were highest in plants with many flowers and concluded that a large inflorescence attracts more pollinators. This correlates with findings of other authors who have found that increasing flower numbers per plant contribute to pollinator attraction, thereby increasing the probability of reproduction (Nilsson, 1992, and references therein; Rodriguez-Robles et al., 1992). As E. alta flowers remain receptive for about 16 d and floral scent of unpollinated flowers increases until the ninth day of anthesis, they contribute to display and long-distance attraction as well as retain the possibility of becoming fertilized. Thus, increased attractiveness by surplus flowers should increase both male and female fitness. This effect together with pollinator limitation could be large enough to explain the differences between 93 % fruit set in hand-pollinated inflorescences and a maximum of 19·8 % natural fruit set in E. alta.
Pollen limitation (PL) is a concept derived from sexual-selection theory as applied to plant reproductive ecology and evolution, and refers to the observed phenomenon that inadequate pollen quantity or quality can translate into reduced seed quantity or quality, i.e. plant reproductive success (Ashman et al., 2004). The most common interpretation of PL is that of Burd and his co-workers (Burd, 1994; Ashman et al., 2004), who deduce that pollination environment varies randomly, and therefore that stochastic models best describe observed phenomena. These authors note (Ashman et al., 2004) that ‘stochastic variation among flowers in pollen receipt is a bet-hedging strategy that commonly leads to low seed set (Stephenson, 1981) or PL (Burd, 1995).’ There has been an over-simplification of this concept, in that experimental supplemental pollinations are routinely used to test for the presence of PL and whenever the results of these exceed natural seed set the system is branded ‘pollen-limited’ or pollinator-limited, and the researchers then attempt to explain the discrepancy. The theoreticians instrumental in developing and popularizing this concept understand that ‘…the standard empirical approach for detecting PL remains problematic for several statistical…and biological reasons’ (Ashman et al., 2004); this last point is often forgotten. We suggest that in our system there is no reason to suppose that the ideal of 93 % (as realized by supplemental hand-pollination) represents an optimum, often if ever achieved in nature. It is hard to imagine, even under the best of conditions, that pollinators in natural richness and abundance will visit each receptive orchid flower nor that there will be a nearly one-to-one correspondence between pollinator removal of pollinaria and deposition on receptive flowers. Relative fruit or seed set per se does not determine the survival of a population or species but rather it is whether the seeds produced are sufficient in quality and quantity to sustain the population through generations. It is possibly less a question of pollinator limitation leading to high PL, than the intrinsic “fuzziness” of this orchid system. The widespread distribution of this colonizing terrestrial orchid points to the success of E. alta despite the findings here that individuals typically set less than 20 % fruit. It is misleading to take an unnatural optimum (outcrossed, hand-pollinated flowers) as the standard for a natural system that depends upon a subtle interplay of many interrelated factors, and which has a certain amount of built-in ‘play’ (bet-hedging of some authors) to accommodate variable (stochastic) conditions. Moreover, the performance of a pollinator assemblage at a given locality is more than the sum of the performances of a single pollinator species. Species-specific pollinator performance is dynamic and influenced at each locality (or in each community) by the dynamics peculiar to that pollinator community. In a recent paper by Burd et al. (2009), they extend their earlier theoretical framework to ovule number and consider pollination success at the floral level in many plants to be stochastic, the inevitable consequence of the hazards of pollination service by biotic and abiotic vectors. In turn, this acts as a source of selection, and they find that there is a positive correlation between stochastic variation in ovule fertilization opportunity and ovule number (Burd et al., 2009). According to them, this explains the commonly observed phenomenon that flowers receiving experimental supplemental pollination frequently have elevated seed production. Their analyses lead them to conclude that ‘…on average, flowers contained ovules that would not have been fertilized in the absence of hand pollination’ (Burd et al., 2009); and incidently, thereby begging the question of the utility of the supplemental pollination standard. Burd et al. (2009) predict that those plants experiencing greater uncertainty of pollination attributable to whatever factor (e.g. reliance on specialist pollinators, clumped pollen deposition) will have many ovules and they cite orchids as a possible example of this. This explanation is harmonious with our results and we concur with Burd et al.'s (2009) assessment that ‘overproduction of ovules makes evolutionary sense in the light of stochastic disparity in mating success’. Such a view opens the door to the understanding that not every ovule produced is expected to be fertilized nor every flower to be pollinated.
We have demonstrated that performance of a pollinator assemblage at a given locality is more than the sum of the standard performances of single pollinator species. Species-specific pollinator performance is not a fixed mathematical expression, but is influenced at each locality (or in each community) by the interactions peculiar to that pollinator community, which has its own dynamics. This study has highlighted some of the factors that need to be considered when studying natural orchid populations and underlines the need for multiple study sites to gain a deeper appreciation of orchid reproductive dynamics. In the case of E. alta, the typical orchid bauplan of its flowers and their size restricts entry to insects of a distinct body size, shape and behaviour, and only those visitors which fulfil these requirements can act as successful pollinators. Eulophia alta flower form varies little throughout its wide range, and hence pollinator type probably remains much the same although pollinator species may and does vary. This underlines the basic resilience of functional pollinator types whereas species, in this case pollinating bees, are more fluid in space and time.
We thank Professor Pe. J. S. Moure, Curitiba, for the identification of insects, and two anonymous reviewers for helpful criticisms and suggestions.