HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (18) Citing Articles via Google Scholar Google Scholar Articles by Smithson, A. Search for Related Content PubMed Articles by Smithson, A. Agricola Articles by Smithson, A.

The consequences of rewardlessness in orchids: reward-supplementation experiments with Anacamptis morio (Orchidaceae)¹

School of Biological Sciences, University of Exeter, Hatherly Laboratories, Prince of Wales Road, Exeter EX4 4PS UK

Received for publication November 28, 2001. Accepted for publication May 3, 2002.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Pollinators are expected to respond to low reward availability in an inflorescence by visiting fewer flowers before departure, thus potentially causing reduced visitation, but also reduced geitonogamous selfing. I tested this hypothesis using Anacamptis morio, an orchid that does not reward its pollinators. Supplementation of inflorescences with artificial nectar did not result in an increase in fruit set on supplemented inflorescences compared to control inflorescences and tended to reduce pollinia removal. Supplementation resulted in reduced fruit quality, but there was no evidence that this was as a result of inbreeding depression. Behavioral experiments showed that pollinating bumble bees, as predicted, visited more flowers on supplemented inflorescences. Bumble bees also deposited more self-pollen on supplemented inflorescences, but this was marginally significant. Bumble bee queens removed significantly more pollinia from control inflorescences, while Bombus terrestris and B. lucorum workers did not. I conclude that while pollinators behaved as predicted, there was weak evidence that pollinia removal, pollen deposition, and fruit set followed the predictions of the hypothesis. I argue that this was probably because some pollinators were more efficient at removing and depositing pollen on control inflorescences, while others were not.

Key Words: Anacamptis morio • Bombus • France • geitonogamy • nectarless flowers • Orchidaceae • pollinator behavior • pollinia

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Pollinator responses to the amount of reward produced by a flowering plant are expected to have major selective effects on plant mating systems and consequently on floral evolution. If an inflorescence has a reduced nectar production rate compared to conspecifics or does not produce any reward at all, pollinators visit far fewer flowers on that inflorescence (e.g., Zimmerman, 1983 ; Cibula and Zimmerman, 1987 ; Johnson and Nilsson, 1999 ; Smithson and Gigord, 2001 ). Such pollinator behavior is predicted by the marginal value theorum: animals are expected to leave a patch when the average rate of energy intake falls to the mean rate for all patches (Stephens and Krebs, 1986 ). Thus each flower will receive fewer visits from pollinators, and overall visitation rates to a given plant will decrease. With fewer visits, however, pollinators are expected to carry less self-pollen from one flower to the next within an inflorescence, potentially resulting in both reduced geitonogamous selfing (de Jong, Waser, and Klinkhamer, 1993 ; Klinkhamer and de Jong, 1993 ) and increased amounts of pollen available for outcrossing (pollen discounting; Holsinger, Feldman, and Christiansen, 1984 ). If selfing is detrimental to a plant because it causes inbreeding depression among offspring, the advantages of reduced geitonogamous selfing and increased seed paternity may outweigh the disadvantages of reduced visitation rate, and, therefore, reduction in reward may become selectively advantageous (Dressler, 1981 ; Nilsson, 1992b ; Klinkhamer and de Jong, 1993 ; Johnson and Nilsson, 1999 ). Significant inbreeding depression has been found in a wide range of plant species (Charlesworth and Charlesworth, 1987 ).

Can the geitonogamy hypothesis explain why so many species of orchids produce no nectar or other reward for their pollinators (Dressler, 1981 ; Ackerman, 1986 ; Dafni, 1987 ; Nilsson, 1992b )? Approximately one-third of all orchid species are thought to be rewardless (van der Pijl and Dodson, 1966 ; Gill, 1989 ). These species are unable to produce nectar, and their pollen also usually cannot be a reward because in the Orchidaceae, pollen is clumped into pollinia that are affixed to the pollinator (Johnson and Edwards, 2000 ). Pollination of rewardless orchids appears to be effected by naive animals that, during exploration of their environment, sample a small number of rewardless inflorescences before switching to more profitable food sources (Heinrich, 1975 ; Nilsson, 1992a ; Smithson and Gigord, 2001 ).

The evolution of rewardlessness has been regarded as paradoxical because many rewardless orchids appear to have extremely low fruit set compared to their rewarding congeners (Gill, 1989 ; Neiland and Wilcock, 1998 ; but see Larson and Barrett, 2000 ). Mimicry is unlikely to explain the evolution of rewardlessness, as this has occurred comparatively rarely in the evolution of the Orchidaceae (Dafni, 1984 , 1987 ; Johnson, 2000 ). The pollinators of rewardless species not only learn quickly to switch to a plant species with a reward, but visit few flowers per inflorescence on rewardless plants and also a short sequence of inflorescences before departing the patch (Nilsson, 1980 ; Dressler, 1981 ; Dafni, 1987 ; Smithson and Macnair, 1997 ). Therefore several authors have argued that reward loss could be an extreme development of the geitonogamy-avoidance hypothesis (Dressler, 1981 ; Nilsson, 1992b ; Johnson and Nilsson, 1999 ).

There have been few specific tests of the geitonogamy hypothesis, particularly in pollinia-based systems. Directly testing the geitonogamy hypothesis in an orchid population with both rewardless and rewarding morphs is not possible, as no such species is known. The presence or absence of nectar production appears to be fixed in all species of orchids studied so far. Johnson and Nilsson (1999) found that, by supplementing the rewardless orchid Anacamptis morio (L.) R. M. Bateman, Prigeon, and M. W. Chase (= Orchis morio L.) with artificial nectar, pollinium removal and deposition increased in one of two populations tested. However, supplementation was done only once, and thus the results may not be truly representative of natural patterns of reward presentation. Smithson and Gigord (2001) found that supplementing the rewardless orchid Barlia robertiana (Loisel.) Greuter inflorescences with artificial nectar over 20 d significantly decreased pollinia removal, while pollen deposition and fruit set remained unaffected by manipulation; no geitonogamous pollen depositions were recorded. Additionally, Salguero-Faria and Ackerman (1999) found that adding artificial nectar to Comparettia falcata, an orchid that produces small amounts of nectar, affected neither pollinia removal, fruit set, nor geitonogamous pollen depositions.

I tested the geitonogamy hypothesis using Anacamptis morio, a completely rewardless terrestrial orchid. In this species, calculations from pollinia bending rates (Johnson and Nilsson, 1999 ) and observations of pollinator behavior (A. Smithson, unpublished data) indicated geitonogamous pollen depositions were likely to occur. I carried out reward-supplementation experiments to test the effects of reward on pollinator behavior, pollen transport within inflorescences, and plant reproductive success.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
The study system
Anacamptis morio is a rewardless orchid species common throughout most of Europe; its reproductive ecology has been studied (Nilsson, 1992a ; Johnson and Nilsson, 1999 ). It does not specifically mimic any other rewarding plant. Anacamptis morio has rose- to purple-colored flowers and is found principally in meadows and dry grasslands on neutral to alkaline soils at a range of altitudes. Each flower has two pollinia that are removed by visiting pollinators either singly or in pairs. Pollen is deposited gradually from the pollinaria onto plant stigmas.

I studied three populations of A. morio in the Corbiéres and Minervois regions of southern France during 1999 and 2000. They were located in Davejean (42°57.495' N, 2°35.913' E), Velieux Tour (43°23.408' N, 2°45.994' E), and Velieux (43°23.558' N, 2°45.205' E). I selected the study populations specifically to maximize the range of population sizes covered in the region, as population size is expected to influence a number of factors, such as pollinator visitation rates and selfing rates. Flowering population sizes were 2658 (Davejean), 866 (Velieux Tour), and 338 (Velieux) plants, and average densities were 0.682, 0.577, and 0.141 flowering plants per square meter, respectively. Flowering occurs in these populations during April and early May.

Pollinator visits to study populations were rare. Beeflies (Bombylius sp.) were frequently observed visiting plants, but were never observed removing or carrying pollinia. During 1999 and 2000 only bumble bees were observed removing pollinia. I recorded Bombus lapidarius (L.) queens (one visitation sequence), B. terrestris (L.) or B. lucorum (L.) queens (three) and workers (one), and B. pascuorum (Scopoli) queens (one) during a minimum of 100 h of observations.

In each of the study populations I found that both pollinia removal (population means from 16.4 to 48.1%) and fruit set (14.6–24.4%) were significantly limited by access to pollinators when compared with hand-pollinated plants (N = 20 per population). Plants were self-compatible (full fruit set for hand-pollinated selfed inflorescences, N = 20 per population), but did not automatically self (enclosed unpollinated inflorescences, N = 20 per population) (A. Smithson, unpublished data).

Population supplementation experiments
During 1999, before the orchids flowered, I randomly selected (using the coordinates of plant positions within populations, ensuring spreading of pairs within populations) approximately 20 pairs of inflorescences in each population (Davejean: 20 pairs, Velieux Tour and Velieux: 21 pairs). Members of a pair were 0.3–2 m apart and were measured and matched for the following traits: total number of flowers produced, inflorescence height (to the nearest 5 mm), lip length and width and spur length and width (to the nearest 0.1 mm).

In the early morning of every day I supplemented every open flower on one member of each pair with "nectar" consisting of 2 µL of 30% sucrose solution using a 10-µL Hamilton microsyringe. This volume and concentration chosen was to allow comparison with natural nectar production rates in a related species (Anacamptis fragrans (Pollini) R. M. Bateman, Prigeon, and M. W. Chase; A. Smithson, unpublished data). As a control measure the microsyringe was inserted into every open flower of the other member of the pair but no sucrose was dispensed. The sucrose was placed into the tip of the flower spurs; to ensure that pollinia turnover dynamics was modified only by bees' behavior in response to reward in the spur tip I did not dispense sucrose onto any other parts of the flower. Nectar was not allowed to build up within flowers and any remaining nectar was removed during the next supplementation. Experimental and control inflorescences were enclosed in green mesh bags to exclude pollinators until manipulations began. That occurred when a minimum of five flowers were open on inflorescences of both members of a pair. I had found in previous experiments that ants tended to consume supplemented sucrose, so that the stem bases of each pair was surrounded with fluon GP1 (Whitford Plastics, Runcorn, UK), which deterred ants but had no observable effect on the inflorescences.

Inflorescences were supplemented for 20 d or until all flowers on an inflorescence had withered if that happened first. Small numbers of inflorescences were lost to predation during the experiment. Total pollinium removal was recorded at the end of the experiment, and plants were once again enclosed with protective green mesh bags. Fruits were counted after 2–3 wk and were collected when fully ripe 3–4 wk later. I attempted to collect three fruits per plant—from the bottom, middle, and top of each inflorescence—although low fruit set often limited collection. Fruit length and width were measured in the field to the nearest 0.1 mm. Total seed mass per fruit was obtained in the laboratory to the nearest 0.00001 g, after fruits had been dried in silica gel at room temperature and seeds removed.

To test whether treatment-induced changes in fruit production resulted in differences in the likelihood of flowering the following year, I marked experimental plants during flowering and recorded their positions to the nearest 20 cm. During 2000 all plants that survived to fruiting in 1999 were found again and their status in 2000 (flowering, not flowering, or not emergent) noted.

The numbers of pollinia removed and fruits set, as well as fruit length, fruit width, and seed mass were compared with a two-way analysis of variance after log₁₀ transformation. The effects of both treatment and population were analyzed as fixed factors, since populations were selected to maximize diversity in population size. Because inflorescence position was expected to influence pod size, I compared fruit length, fruit width, and seed mass using a partially hierarchical four-way ANOVA, with treatment, population, plant, and position within plant as factors and with plants nested within population. All factors were again fixed, as each level was selected nonrandomly from those available. Interaction terms in the model were all nonsignificant, so they were removed from the model. To ensure that results were not artifactual, I tested for any covariance between floral traits and treatment on pollinia removal, fruit set, and fruit size using a two-way analysis of covariance and also for any differences in total flowering time between treatments using two-way analysis of variance. The sample size for fruits was considerably lower than for pollinia removal and fruit set because of the low number of fruits formed and the occasional loss of plants. To test whether the probability of flowering in the experimental plants changed from 1999 to 2000, I compared their statuses in the first year with those in the second year using a G test.

Statistical analysis was performed with SPSS version 9 for Windows (SPSS, Chicago, Illinois, USA). When repeated statistical tests were carried out, these were corrected using the sequential Bonferroni method (Rice, 1989 ).

Pollinator behavior experiments
Pollinator behavioral responses to supplemented and control orchids were studied in detail in one population (Davejean) during 2000 using a "bee interview" technique similar to that of Thomson (1988) and Johnson and Nilsson (1999) . Designated pairs of plants (approximately 40 in all) were potted up in natural soil after their precise location was marked. Plant pairs were used only if they had the same number of open flowers, and the inflorescences were measured and matched as closely as possible for other traits (inflorescence height, lip length and width, spur length and width). One member of each pair was supplemented with 2 µL of 30% sucrose solution placed into the tip of the spur of each open flower, while the other had a control manipulation. Each pot was then affixed side by side to the end of a long cane and gently offered to a bumble bee while it was foraging on a rewarding plant species. Bumble bees were not scared away by this action, and the inflorescences remained stationary during presentation. Coin tosses were used to decide the identity of the rewarding inflorescence and the side of the cane on which each plant was placed. Only inflorescences without previous pollinia removals or depositions on open flowers were used; thus pairs were kept covered with green mesh bags before and after experimental sessions, and inflorescences were reused only if pollinators failed to remove or deposit pollinia or if a sufficient number of new flowers opened. After experiments, nectar was removed from all supplemented flowers to ensure that no nectar carried over to subsequent experiments. Experiments were conducted over a 4 wk period. Afterwards all plants were replaced in their original positions in the study area.

Individual bumble bees, both queens and workers of a range of species, were offered a pair of potted A. morio. I recorded the species and caste of the bee, along with any active rejections of A. morio (defined as obvious deviation of the bee's flight path or speed but subsequent avoidance of the inflorescences) and counted any pollinaria the bee was carrying from other A. morio plants. I grouped B. terrestris and B. lucorum (hereafter denoted "B. terrestris-or-lucorum") because it was difficult to distinguish the two species in the field. For bees that chose to visit one or more inflorescences, I recorded the number and sequence of flowers visited on the inflorescence and whether pollinia were removed or pollen deposited. Additionally the time spent on each flower was noted to the nearest 0.1 s. After a visit to inflorescences, the bee was caught and its thorax marked with solvent-free correction fluid to prevent it being tested again. Occasionally, pollinators visited more than one of the inflorescences offered to them. In data analysis, I used only the first inflorescence, as I expected that pollinator behavior would change in subsequent visits.

Because of the low numbers of bumble bees in the study area—most were observed merely overflying the area and occasionally sampling A. morio—I located a range of rewarding plant species that attracted numerous foraging bumble bees in the surrounding habitat. These rewarding species were principally rosemary (Rosmarinus officinalis L.) and red deadnettle (Lamium purpureum L.) and to a lesser extent early-flowering trees (Prunus sp.) and tree heath (Erica arborea L.).

I used t tests after log₁₀ transformation to test for differences between the two inflorescence types in mean number of flowers visited per inflorescence by bees, mean time spent per flower, mean number of pollinia removed per inflorescence visit, and mean number of stigmas per inflorescence visit that received pollen depositions. I also tested for differences in the probability of pollinium removal and deposition by dividing the number of depositions or removals by the number of flowers visited and compared the result using t tests after arcsine transformation. This measure allows differences in the mean efficiency of pollen transport by bees over all the flowers they visited to be quantified. Differences in geitonogamous pollen depositions between treatments were tested by comparing the numbers of depositions onto stigmas made by pollinators that were not carrying pollinaria from any previously visited inflorescences. For this analysis, I excluded data from pollinators that did not remove any pollinia while on an inflorescence. I tested for any effects of pollinator species and caste on each data set using analysis of variance.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Population-supplementation experiments
Figure 1 compares mean pollinia removal, fruit set, fruit size, and seed mass for supplemented and control inflorescences in each population tested. Supplementation had no significant effects on pollinia removal (Table 1; F_1,116 = 1.46, P > 0.1), although the mean number of pollinia removed from control inflorescences is greater than for supplemented ones across all three populations (Fig. 1; overall mean number of pollinia removed ± 1 SE for control inflorescences = 3.80 ± 0.44, for supplemented inflorescences = 3.15 ± 0.43). The mean number of pollinia removed also varied significantly across populations, being greatest at Velieux and tending to increase with decreasing population size (Table 1). Neither treatment nor population had significant effects on fruit set (Table 1); in fact, mean fruit set was identical for the two treatments (overall mean fruit set ± 1 SE for control inflorescences = 1.12 ± 0.19, for supplemented inflorescences = 1.12 ± 0.22). A significant effect of treatment on fruit length was detected (Table 2), in that fruits from control plants were longer than those from supplemented ones (Fig. 1), but no significant effect of treatment on fruit width or seed mass was found (Table 2). The large mean and variance in fruit mass for Davejean (Fig. 1) reflected both the small sample size (fruit set was particularly low) and the biases in fruit positions at this site.

View larger version (26K): [in this window] [in a new window]   Fig. 1. Results of experiments with Anacamptis morio in which 2 µL of 30% sucrose solution was added to all open flowers on selected inflorescences in three populations (shaded bars), and resultant reproductive success was compared with controls (open bars). Bars indicate means + 1 SE

Of 118 experimental plants that survived to set fruit in experiments in 1999, 112 were relocated the following year. In 2000 72.3% flowered, 17.0% did not flower, and 10.7% did not emerge. I found no significant differences in status in 2000 between treatments (G test of independence: G_{adj 2} = 0.06, P > 0.05).

Pollinator-behavior experiments
Seventy-two successful trials were completed. The majority of visits were made by B. terrestris-or-lucorum (58%), followed by B. pascuorum (29%). The remaining 13% were made by four species (B. pratorum (L.), B. hortorum (L.), B. lapidarius, B. wurfleinii Radoszkowski). Most B. terrestris-or-lucorum that visited inflorescences were workers (81%) while, conversely, most bees of other species were queens (67%). Most bees were not scared away by having the inflorescences presented to them and continued foraging on the rewarding species, clearly rejecting the A. morio inflorescences when they passed by them. The proportion of acceptances varied significantly among bumble bee species (acceptance rates were 20.0% for B. pascuorum, 6.0% for B. terrestris-or-lucorum and 10.8% for other bees; G test of independence: G_{adj 2} = 19.78, P < 0.001) and among castes (acceptance rates were 29.2% for queens and 3.7% for workers; G_{adj 2} = 81.33, P < 0.001), but not across the different coflowering plant species (G_{adj 2} = 5.44, P > 0.05). There was no indication that bees could discriminate between rewarding and empty inflorescences before visiting them (they visited empty inflorescences first 35 times and rewarding ones 37 times).

Data analysis for both bee behavior and pollen removal and deposition data was complicated by significant heterogeneity of results across bee species and caste. Preliminary analysis testing these factors revealed that by far the highest proportion of the variation in each data set was explained by differences between B. terrestris-or-lucorum workers and all other bees combined. Therefore in data analysis I contrasted these two groups (hereafter termed a "group" effect in analysis of variance).

Results (Fig. 2) showed that bumble bees, as predicted, visited significantly more flowers on supplemented inflorescences compared to control inflorescences. The number of flowers visited on controls (mean ± 1 SE = 2.06 ± 0.35 flowers) was similar to those made by pollinators observed foraging on unmanipulated inflorescences in natural populations (mean ± 1 SE = 3.11 ± 0.63 flowers; 9 visits observed). An analysis of variance found not only a significant effect of treatment (F_1,68 = 13.24, P = 0.001), but also a significant effect of group (F_1,68 = 4.57, P < 0.05). Figure 2 shows that B. terrestris-or-lucorum workers visited fewer flowers on supplemented inflorescences compared to other bees and indeed did not visit a significantly greater number of flowers on supplemented inflorescences compared to control inflorescences.

View larger version (32K): [in this window] [in a new window]   Fig. 2. Results of pollinator behavior experiments with Anacamptis morio. Supplemented (shaded bars) and control (open bars) inflorescences were offered in pairs to foraging bumble bees (Bombus sp.). Bars on the left of each graph show the results for all data combined, while the two sets of bars on the right show the results of splitting data into two categories (B. terrestris and B. lucorum workers, all other bees). Bars indicate means + 1 SE. Significant differences between supplemented and control inflorescences, as determined by t tests after Bonferroni correction, are indicated above the bars. (*) 0.1 < P < 0.05, * P < 0.05, ** P < 0.01, *** P < 0.001

Pollinia removal did not differ significantly overall between the two treatments (Fig. 2). This result was, however, an artifact of combining the two groups of bees that had very different pollinia removal characteristics: B. terrestris-or-lucorum workers tended to remove slightly but not significantly more pollinia on supplemented inflorescences. In contrast, all other bees removed significantly more pollinia on control inflorescences (Fig. 2). Analysis of variance revealed a significant effect of treatment by group interaction (F_1,68 = 5.67, P < 0.05), highlighting the reversal in treatment effects between the two groups of bees.

Pollen deposition data was difficult to analyze due to the low numbers of pollen depositions recorded overall and because only five bees were found carrying pollinia before visiting an inflorescence. Therefore I analyzed only self-depositions made by bumble bees. The data set was further reduced by removing bees that did not remove pollinia on an inflorescence, leaving a sample size of 29 bees. In contrast to pollinia removal data, self-pollen deposition was consistently greatest on supplemented inflorescences, although this effect was marginally significant (Fig. 2; t₂₇ = 1.85, P = 0.08). While both groups of bees showed the same trend, the pattern was strongest for B. terrestris-or-lucorum workers (t₉ = 1.57, P = 0.15; Fig. 2). Analysis of variance, however, revealed marginally significant effects of treatment (F_1,25 = 3.40, P = 0.08) and no significant effects of group (F_1,63 = 0.04, P > 0.05) or treatment by group interaction (F_1,56 = 0.19, P > 0.05).

Analysis of the probability of pollinia removal on a per-flower basis by bees showed that B. terrestris-or-lucorum workers removed pollinia equally efficiently on supplemented and control flowers (Fig. 2). In contrast, all other bees were significantly more efficient at removing pollinia from control flowers (Fig. 2). Analysis of variance found significant effects of both treatment (F_1,68 = 6.60, P < 0.05), group (F_1,68 = 4.65, P < 0.05), and treatment by group interaction (F_1,68 = 6.26, P < 0.05). The probability of pollen deposition showed similar trends to those of pollinia removal: B. terrestris-or-lucorum workers were only depositing pollen on supplemented flowers and other bees were equally efficient on both flower types. However, there were no significant effects, either in t tests (Fig. 2) or in analysis of variance.

Interestingly, I found that the likelihood of pollen deposition on a flower increased with the total amount of time spent on an inflorescence (r = 0.43, P = 0.02, N = 27), but not the amount of time spent on a flower (r = –0.06, P > 0.05, N = 27). In contrast, the likelihood of pollinia removal increased with the time spent on a flower (r = 0.37, P = 0.002, N = 64), but not the total amount of time spent on an inflorescence (r = –0.08, P > 0.05, N = 64).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
With these experiments I aimed to test the hypothesis that pollinators would respond to the absence of reward on an inflorescence by visiting only a few flowers before departing and that this behavior would reduce both visitation and geitonogamous selfing when compared with an inflorescence that provides abundant reward (Dressler, 1981 ; Nilsson, 1992b ; Johnson and Nilsson, 1999 ). I found that supplementing A. morio inflorescences with artificial nectar caused pollinating bumble bees to visit significantly more flowers per inflorescence, as predicted by the hypothesis. Bumble bees deposited self-pollen on more flowers per inflorescence on supplemented plants, although the effect was only marginally significant, indicating that increased geitonogamous selfing was possible. Further, the quality of fruits produced from supplemented inflorescences, as quantified by fruit length, was significantly lower than for controls, suggesting that increased geitonogamous selfing could have led to inbreeding depression in supplemented fruits. However, other results did not support the geitonogamy hypothesis. Firstly, mean fruit set on supplemented and control inflorescences in plant populations was identical, not higher for supplemented inflorescences as predicted. Fruit set on both supplemented and control inflorescences remained highly pollinator-limited, so there was no possibility that maximal fruit set or pollinia removal had been reached. Secondly, there was a trend for increased pollinia removal from control inflorescences in plant populations, even though the geitonogamy hypothesis predicts higher pollinia removal from supplemented inflorescences. Behavioral experiments showed that queens, which were the most common pollinators in the field, removed significantly more pollinia from control inflorescences. Note that this was not so for B. terrestris or B. lucorum workers. Thirdly, a comparison of inflorescences that were experimentally selfed with those that were outcrossed revealed no significant difference in numbers of fruits set, fruit length, or fruit width (A. Smithson, unpublished data). These findings argue against the hypothesis that the decrease in fruit quality observed on supplemented inflorescences was due to inbreeding depression. The increase in reproductive effort of control inflorescences through producing longer fruits also did not impact reproduction in the following year. Overall my results found little evidence that control inflorescences had a disadvantage in total reproductive success compared to supplemented ones, as some researchers had predicted (Neiland and Wilcock, 1998 ; Johnson and Nilsson, 1999 ).

How can the contradictions between pollinator behavioral, pollen transfer, and inflorescence reproductive success data be explained? One possibility is that my results were, wholly or partially, an artifact of the experimental conditions. I found that long-term supplementation with sucrose solution had a significant negative effect on flower longevity, which could have resulted in artificially reduced fruit set and pollinia removal in the population supplementation experiments. Similar negative effects of supplementation have been observed previously (Ackerman, 1981 ). However, I observed that nearly all pollinia removals and pollen depositions occurred before supplemented inflorescences had started to wither; thus, it seems unlikely that the small observed differences in floral longevity would have affected the results of the population supplementation experiments. In addition, sucrose solution did not contaminate other flower structures and was unlikely to have directly affected pollinia removal or deposition other than through differential pollinator behavior. Pollinators behaved normally on inflorescences, and there was no indication that offering them pairs of inflorescences caused any adverse avoidance reactions that could have lead to abnormal behavior. Bees did not assess the presence of nectar present in flowers visually, nor did floral traits significantly interact with treatment effects. When foraging on supplemented flowers, bees probed fully to the bottom of the spurs and drained nectar completely. Old nectar was not allowed to accumulate in flowers, so could not have caused any avoidance reactions by pollinators. I therefore conclude that my experimental results were unlikely to be artifactual.

The results of pollinator behavior experiments clearly showed that, while pollinators visited more flowers on supplemented inflorescences as predicted, this behavior did not always translate simply into higher pollinia removal and pollen deposition on supplemented inflorescences. I argue that my results differed from the predictions of the geitonogamy-avoidance hypothesis because of the mechanics of pollen transfer in A. morio and because different groups of pollinators did not transfer pollen in the same way. My results showed that pollinia removal was more likely the longer that a pollinator spent on a flower, while this was not the case for pollen deposition. In plants with diffuse pollen, an increase in time spent on a flower by pollinators frequently increases the amount of pollen transferred within that flower (Galen and Stanton, 1989 ). The total amount of time spent on an inflorescence increased the likelihood of pollen deposition, while it did not increase the likelihood of pollinia removal. In the Orchidaceae, the pollinaria are known to bend forward at a certain rate after removal from a flower, which increases the chances of successful pollen deposition (Johnson and Edwards, 2000 ), suggesting that pollinaria bending led to the increased chances of pollen deposition with inflorescence residence time. I found that B. terrestris-or-lucorum workers spent significantly more time on both supplemented flowers and on supplemented inflorescences compared to controls. Such behavior would be expected if empty flowers were quickly rejected and more time was spent extracting nectar from full flowers. In contrast, queens spent the same amount of time on flowers of both treatments and on inflorescences of both treatments. This behavior is more unexpected and suggests that queens spent a significant amount of time searching for nectar on control flowers and inflorescences. This suggestion is supported by the significantly greater efficiency in removing pollinia from control flowers and by similar behavioral patterns that I observed in previous experiments using another rewardless orchid (Smithson and Gigord, 2001 ). I argue that it is this crucial difference in searching behavior that leads to different patterns of pollen transfer. B. terrestris-or-lucorum workers spent more time on flowers and inflorescences when these were supplemented, but only visited a slightly greater number of flowers on supplemented inflorescences; therefore they tended to remove slightly more pollinia from and deposit more self-pollen on supplemented inflorescences. Queens spent similar amounts of time on both flower types and were more efficient at removing pollinia from control flowers. Therefore, despite the greater number of flowers that queens visited on supplemented inflorescences, they removed significantly more pollinia in total from control inflorescences. While it was difficult to see patterns in pollen deposition due to small sample size, queens did show a reduced tendency to deposit more self-pollen on supplemented inflorescences compared to B. terrestris-or-lucorum workers: in fact, queens deposited self-pollen on control inflorescences while B. terrestris-or-lucorum workers did not. Differences in pollen turnover caused by different pollinator types have been found by other researchers (e.g., Schemske and Horvitz, 1984 ; Stanton et al., 1991 ).

I argue that it is these differences in way that the two pollinator types remove and deposit pollen that has resulted in the final reproductive success of supplemented and control A. morio inflorescences not following the predictions of the geitonogamy hypothesis. In the field, queen bumble bees made most visits to A. morio (5), and a minority of visits were made by B. terrestris-or-lucorum workers (1). While a small sample size, this result is similar to other studies where the pollinators of A. morio were also bumble bee queens (Nilsson, 1992a ; Johnson and Nilsson, 1999 ). If field populations of A. morio were pollinated mainly, but not entirely, by queen bumble bees that removed and deposited pollen as described above, greater pollinia removal would have been expected on control inflorescences, and similar levels of fruit set would have been expected on both inflorescence types. These were the results observed. Greater fruit size on control inflorescences could be explained not only by inbreeding depression, but also by queen bumble bees depositing pollen more efficiently on control stigmas. Unfortunately, low sample sizes prevented me analyzing the amount of pollen deposited on the stigmas of A. morio in behavior experiments.

If my hypothesis is correct, why did the searching behaviors of B. terrestris-or-lucorum workers and queens differ? B. terrestris and B. lucorum are early-emerging species, and their worker bees are likely to have gained more foraging experience compared to both queens, since these forage only before their workers have hatched and must undertake other duties such as nest-finding and egg-laying, and also compared to workers of other species like B. pascuorum and B. lapidarius, which emerge later. Naive bees may have a higher expectation for finding a reward on a novel flower than experienced bees: they may make more effort to find nectar when they do not locate it initially and thus stay longer on control flowers. The suggestion that B. terrestris-or-lucorum workers were more experienced is supported by their significantly lower acceptance rates for A. morio.

Can the geitonogamy-avoidance hypothesis explain the maintenance of rewardlessness in A. morio populations? My data are equivocal. My results do not show strong support for the geitonogamy-avoidance hypothesis (Dressler, 1981 ; Nilsson, 1992b ; Johnson and Nilsson, 1999 ): although higher self-pollen deposition was probably found on supplemented inflorescences in behavior experiments, fruit set and pollinia removal in the field did not significantly differ between supplemented and control inflorescences. Final conclusions can only be reached, however, after testing whether selfing rates of progeny from supplemented inflorescences are higher than fruits from control inflorescences and also after determining whether there is significant inbreeding depression in A. morio. I am currently testing both these hypotheses. My data also gave no strong support for the pollinia-removal advantage hypothesis (Smithson and Gigord, 2001 ): significantly higher pollinia removal was observed on control inflorescences by the predominant pollinators in behavior experiments, but there was only a trend for higher pollinia removal in the field. Overall I suggest that my results may reflect a combination of selection patterns resulting from visitation by two types of pollinators that differed in the ways that they turned over pollen within inflorescences. My results confirm earlier work however (Smithson and Gigord, 2001 ): there is not necessarily a simple link between pollinator visitation of flowers within inflorescences, pollinia removal, pollen deposition, and fruit set in the Orchidaceae. Experiments studying pollen turnover dynamics in rewarding and rewardless orchids, comparing experienced and naive pollinators, are essential to explore further the evolution of rewardlessness.

In my experiments, I supplemented a small number of inflorescences spread within populations of a rewardless orchid. Would my results have been the same had I supplemented a higher frequency of orchids? This question is important, because a rewardless morph could only spread to fixation within a population if it has a fitness advantage, relative to a rewarding morph, regardless of frequency. Clearly it would be difficult to test this hypothesis in natural populations due to the time involved in the supplementation of large numbers of inflorescences. I suggest, however, that pollinator responses to rewarding and rewardless morphs in a hypothetical polymorphic population may hinge on the relationship between the frequency of a morph in a population, flowering time, and pollinator experience. Naive pollinators visiting the population early in the flowering season might respond as bumble bee queens did here: by removing more pollinia from rewardless inflorescences, which could result in a fitness advantage for rewardless morphs through increased pollinia removal and thus seed paternity. Experienced pollinators visiting later in the flowering season might respond as B. terrestris-or-lucorum workers did here: by depositing less self-pollen on rewardless morphs. Rewardless morphs might then have either a fitness disadvantage relative to rewarding ones through decreased fruit set or a fitness advantage through decreased selfing of fruits and increased seed paternity. If a pollinia-removal advantage is combined with early flowering, it is easy to predict how rewardlessness reaches fixation, since if there is a sufficiently large pool of naive pollinators I expect that the fitness advantages for rewardless morphs would occur regardless of morph frequency. For the geitongamy-reduction hypothesis, it is more difficult to explain how rewardlessness reaches fixation: as rewardlessness becomes common, pollinator visitation rates would decrease, pollinator-limitation would increase, and selection might be expected to favor increased fruit set on rewarding morphs at some point even if selfing rates were high (Nilsson, 1992b ; A. Smithson and L. D. B. Gigord, unpublished data). Experiments utilizing artificial arrays with varying frequencies of rewarding and rewardless orchids could be used to test these hypotheses. Interestingly, both geitonogamy-reduction and pollinia-removal advantage hypotheses predict higher male fitness for control inflorescences, although the mechanisms behind these predictions differ. Further experiments quantifying the relative fitness of supplemented and control inflorescences through offspring paternity are, I suggest, likely to prove important in determining the reasons why so many orchid species are rewardless.

	FOOTNOTES

 
¹ The author thanks Richard Gianfrancisco, Lucinda Healy, Colin Lee, Sheena McKendrick, and Derek Nichols for assistance with field experiments; Luc Gigord for assistance and much discussion; Genevievre and Christian Blé and all the landowners for permission to conduct this research and for subsequent assistance; and James Ackerman and two other anonymous reviewers for their comments on an earlier draft of the manuscript. The author was supported by Natural Environment Research Council (UK) research fellowship GT5/98/12/TS.

² A.Smithson{at}exeter.ac.uk

	LITERATURE CITED

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Ackerman J. D. 1981 Pollination biology of Calypso bulbosa var. occidentalis (Orchidaceae): a food-deception system. Madroño 28: 101-110

Ackerman J. D. 1986 Mechanisms and evolution of food-deceptive pollination systems in orchids. Lindleyana 1: 108-113

Charlesworth D. B. Charlesworth 1987 Inbreeding depression and its evolutionary consequences. Annual Reviews of Ecology and Systematics 18: 237-268[CrossRef][ISI]

Cibula D. A. M. Zimmerman 1987 Bumble bee foraging behavior: chances in departure decisions as a function of experimental nectar manipulations. American Midland Naturalist 117: 368-376

Dafni A. 1984 Mimicry and deception in pollination. Annual Reviews of Ecology and Systematics 15: 259-278

Dafni A. 1987 Pollination in Orchis and related genera: evolution from reward to deception. Orchid Biology: Reviews and Perspectives 4: 79-104

de Jong T. J. N. M. Waser P. G. L. Klinkhamer 1993 Geitonogamy: the neglected side of selfing. Trends in Ecology and Evolution 8: 321-325

Dressler R. L. 1981 The orchids: natural history, and classification. Harvard University Press, Cambridge, Massachusetts, USA

Galen C. M. L. Stanton 1989 Bumblebee pollination and floral morphology: factors influencing pollen dispersal in the alpine sky pilot, Polemonium viscosum (Polemoniaceae). American Journal of Botany 76: 419-426[CrossRef][ISI]

Gill D. E. 1989 Fruiting failure, pollinator inefficiency, and speciation in orchids. In D. Otte and J. A. Endler [eds.], Speciation and its consequences, 458–481. Sinauer, Sunderland, Massachusetts, USA

Heinrich B. 1975 Bee flowers: a hypothesis on flower variety and blooming times. Evolution 29: 325-334[CrossRef][ISI]

Holsinger K. E. M. W. Feldman F. B. Christiansen 1984 The evolution of self-fertilization in plants: a population genetic model. American Naturalist 124: 446-453[CrossRef][ISI]

Johnson S. D. 2000 Batesian mimicry in the non-rewarding orchid Disa pulchra, and its consequences for pollinator behaviour. Biological Journal of the Linnean Society 71: 119-132

Johnson S. D. T. J. Edwards 2000 The structure and function of orchid pollinaria. Plant Systematics and Evolution 222: 243-269

Johnson S. D. L. A. Nilsson 1999 Pollen carryover, geitonogamy, and the evolution of deceptive pollination systems in orchids. Ecology 80: 2607-2619[CrossRef][ISI]

Klinkhamer P. G. L. T. J. de Jong 1993 Attractiveness to pollinators: a plant's dilemma. Oikos 66: 180-184[CrossRef][ISI]

Larson B. M. S. C. H. Barrett 2000 A comparative analysis of pollen limitation in flowering plants. Biological Journal of the Linnean Society 69: 503-520

Neiland M. R. M. C. C. Wilcock 1998 Fruit set, nectar reward, and rarity in the Orchidaceae. American Journal of Botany 85: 1657-1671[Abstract/Free Full Text]

Nilsson L. A. 1980 The pollination ecology of Dactylorhiza sambucina (Orchidaceae). Botaniska Notiser 133: 367-385

Nilsson L. A. 1992a Animal pollinators adjust plant gender in relation to floral display: evidence from Orchis morio (Orchidaceae). Evolutionary Trends in Plants 6: 33-40

Nilsson L. A. 1992b Orchid pollination biology. Trends in Ecology and Evolution 7: 255-259

Rice W. R. 1989 Analyzing tables of statistical tests. Evolution 43: 223-225[CrossRef][ISI]

Salguero-Faria J. A. J. D. Ackerman 1999 A nectar reward: is more better?. Biotropica 31: 303-311[CrossRef][ISI]

Schemske D. W. C. C. Horvitz 1984 Variation among floral visitors in pollination ability: a precondition for mutualism specialization. Science 225: 519-521[Abstract/Free Full Text]

Smithson A. L. D. B. Gigord 2001 Are there fitness advantages in being a rewardless orchid? Reward supplementation experiments with Barlia robertiana. Proceedings of the Royal Society of London Series B Biological Sciences 268: 1-7

Smithson A. M. R. Macnair 1997 Negative frequency-dependent selection by pollinators on artificial flowers without rewards. Evolution 51: 715-723[CrossRef][ISI]

Stanton M. H. J. Young N. C. Ellstrand J. M. Clegg 1991 Consequences of floral variation for male and female reproduction in experimental populations of wild radish, Raphanus sativus L. Evolution 45: 268-280[CrossRef][ISI]

Stephens D. W. J. R. Krebs 1986 Foraging theory. Princeton University Press, Princeton, New Jersey, USA

Thomson J. D. 1988 Effects of variation in inflorescence size and floral rewards on the visitation rates of traplining pollinators of Aralia hispida. Evolutionary Ecology 2: 65-76

van der Pijl L. C. H. Dodson 1966 Orchid flowers: their pollination and evolution. University of Miami Press, Coral Gables, Florida, USA

Zimmerman M. 1983 Plant reproduction and optimal foraging: experimental nectar manipulations in Delphinium nelsonii. Oikos 41: 57-63[CrossRef][ISI]

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (18) Citing Articles via Google Scholar Google Scholar Articles by Smithson, A. Search for Related Content PubMed Articles by Smithson, A. Agricola Articles by Smithson, A.