- Top of page
- Methods and materials
Recent studies of plant–pollinator interactions have found visitor assemblages that are taxonomically diverse and spatially and temporally variable (e.g. Herrera 1988; Fishbein & Venable 1996). Some have argued that such generalized interactions may be the rule (Ollerton 1996; Wilson & Thomson 1996). However, only a few species of a diverse assemblage of flower visitors may actually pollinate the flowers (reviewed in Ollerton 1996; Johnson & Steiner 2000). Studies that characterize pollinator effectiveness of each member of a visitor assemblage are necessary to determine whether generalized pollination systems actually predominate. We know little of the pollination systems of many Cactaceae, plants that are abundant, widespread and ecologically important in arid and semiarid lands of the New World.
The flowers of many cacti are visited by bees that specialize on cacti (Simpson & Neff 1987). Bees are generalized to their nectar sources, capable of visiting many kinds of plants for nectar, but many if not most bees are pollen specialists (Wcislo & Cane 1996). Females visit a restricted subset of the plant species available for pollen, which they use as food for their larvae. Their preferences appear to be genetically based and unaffected by the abundance of potentially alternative floral resources available (Wcislo & Cane 1996).
The relationship between the degree of a visitor's specialization to its host plant and the location of that visitor along the mutualism–antagonism continuum has been little explored in pollination research. Pollen-specialist bees might be antagonists to the plants they visit, because they actively collect and sequester pollen for larval provisions, making that pollen unavailable for pollination (Harder & Barclay 1994). They can also be more efficient pollen harvesters than generalist bees (Strickler 1979). Conversely, pollen-specialist bees might be more likely to be mutualists than generalized visitors, because they may be adapted to the flowers they visit, and may be more abundant and predictable than other visitors (Cane & Payne 1993). Evidence to date on the relative value of specialist vs generalist visitors to flowers has been equivocal (Motten, Campbell & Alexander 1981; Neff & Simpson 1990; Keys, Buchmann & Smith 1995).
The behaviour of floral visitors at a given flower can be quite different depending on whether they are seeking nectar or pollen, and these differences in behaviour can affect pollinator effectiveness. Thus organizing visitors into functional groups according to the reward they seek may be a more appropriate way to categorize them than by their taxonomic affiliation (Bosch, Retana & Cerda 1997).
I studied the floral visitor assemblages of two species of barrel cacti: Ferocactus cylindraceus (Engelm.) Orcutt and F. wislizeni (Engelm.) Britt. and Rose. These plants are functionally outcrossers and have no form of vegetative reproduction, making them completely dependent on floral visitors for their reproduction (McIntosh 2002a,b). To determine the pollinator effectiveness of different visitors, the single-pollinator-visit method was used (Motten et al. 1981; Pellmyr & Thompson 1996), in which a single visit is allowed to each virgin flower, and the result of that visit (fruit set, seeds per fruit) is recorded. The quantity and quality components of pollinator effectiveness were estimated (Herrera 1987, 1989; Fishbein & Venable 1996), and the relative importance of different visits or visitors to the total realized fecundity of the plants was determined. Fecundity was measured as the total number of seeds produced by each group of visits or visitors (Fishbein & Venable 1996; Pellmyr & Thompson 1996). Evaluating pollinator effectiveness in terms of the total effect on fecundity is important because the effect of the quality components (e.g. fruit set or seeds per fruit) can easily be overwhelmed by the quantity component.
The questions addressed in this study were:
What is the taxonomic composition and relative abundances of floral visitors to F. cylindraceus and F. wislizeni at the study sites?
Do pollinators differ in their effectiveness?
Do nectar-collecting visits differ in quality from pollen-collecting visits?
Do visits by male bees differ in quality from visits by female bees?
Because of the generalized floral morphology of Ferocactus flowers (see Methods), it was expected that the visitor assemblage to these plants would be diverse in composition and variable in time. It was also expected that pollen-specialist bees would pollinate, but that they would comprise only a part of the pollinator assemblage. Finally, because different behaviours are required to collect pollen vs nectar from Ferocactus flowers, it was expected that pollen visits would significantly differ from nectar visits in per-visit quality.
- Top of page
- Methods and materials
For both Ferocactus species, pollinating visitors were three cactus bees (D. rinconis, S. duplocincta and A. opuntiae), Megachile spp. and Halictus spp. (Table 1). No fruits ever resulted from visits by small bees (body length ≤ 7 mm) in the family Halictidae (Augochlorella spp., N = 25 visits, and Lasioglossum spp., N = 8), the non-native honey-bee (Apis mellifera N = 5 visits), and bees of the tribe Anthidiini (N = 1). Visits by flies (Diptera; N = 10) and butterflies (Lepidoptera; N = 1) also failed to result in any fruit set.
Table 1. Composition of floral visitor taxa, by number of visits (relative abundance), 1996–98. Bold taxa are pollinators (see Methods). Taxa in italics are cactus specialists. Visits by some non-pollinating taxa were not recorded in 1998 (‘nr’). Differences in total visits recorded among years/plant species reflect differences in sampling effort, not visitation rates. Total visits are more than those shown in other figures and tables because this table includes multiple visits to single flowers and fruits that were destroyed by predaceous Lepidoptera. ‘F.cyl’ = F. cylindraceus, ‘F.wis’ = F. wislizeni. Bee taxa are ordered as in Michener (2000)
|Floral visitorsN = total visits recorded||F.cyl 1996 N = 36||F.cyl 1997 N = 91||F.cyl 1998 N = 127||F.wis 1996 N = 35||F.wis 1997 N = 144||F.wis 1998 N = 192|
| Halictus spp. (N = 2)|| 0|| 0|| 0|| 0|| 2|| 0|
| Lasioglossum spp. (N = 14)|| 0|| 2||nr|| 3|| 9||nr|
| Augochlorella spp. (N = 41)|| 1|| 0||nr|| 4||36||nr|
| Ashmeadiella opuntiae (N = 104)|| 0|| 6|| 0|| 8||28||62|
| Anthidiinae (N = 2)|| 2|| 0|| 0|| 0|| 0|| 0|
| Megachile spp. (N = 10)|| 1|| 0|| 0|| 0|| 3|| 6|
| Diadasia rinconis (N = 307)||25||70||113|| 2||36||61|
| Svastra duplocincta (N = 124)|| 1|| 4|| 14||18||29||58|
| Apis mellifera (N = 5)|| 0|| 1|| 0|| 0|| 0|| 4|
|Unidentified large bees (N = 3)|| 0|| 2|| 0|| 0|| 0|| 1|
|Unidentified small bees (N = 1)|| 0|| 1||nr|| 0|| 0||nr|
|Other: Diptera (N = 6)|| 6|| 0|| 0|| 0|| 0||nr|
|Other: Lepidoptera (N = 1)|| 0|| 0|| 0|| 0|| 1||nr|
For F. cylindraceus, most of the visits observed were by the three cactus bees (Table 1). Among these, visits by D. rinconis were by far the most frequent. For F. wislizeni, 65 (1997) to 80% (1996) of visits were received from cactus bees. For F. wislizeni, in 1996 visits by S. duplocincta were the most abundant (64%), whereas in 1997 and 1998, visits were almost exactly evenly divided among the three cactus bees.
For F. cylindraceus, the mean number of seeds that resulted from a single pollinator visit (among those visits that resulted in a fruit), was 149 (range: 11–896). For F. wislizeni, the mean was 148 (range: 2–951).
Visit type did not have a significant effect on fruit set for F. cylindraceus (likelihood ratio = 1·180, df = 3, P = 0·7579, N = 180). The effect of bee species was not tested because there were so few visits other than D. rinconis visits to F. cylindraceus flowers (D. rinconis: 181 visits, S. duplocincta: 16, A. opuntiae: 6). In a nominal logistic model fit test of the effects of bee species and visit type on fruit set in F. wislizeni (pooled across years), bee species was significant and visit type was not (Table 2). Visits by D. rinconis to F. wislizeni resulted in a greater fruit set than visits by either S. duplocincta or A. opuntiae (Fig. 1a).
Table 2. Effect of bee species (three cactus bees) and visit type on fruit set for F. wislizeni (N = 259 visits). Significant values are in bold type. Model R2 = 0·0531
| ||df||Wald χ2||P|
|Visit type||3|| 2·4596858||0·4826|
Figure 1. Pollinator effectiveness, quality components, for the three cactus bees. Visits pooled across years and across Ferocactus species: (a) percentage fruit set; (b) mean seeds per fruit ± SE. Means with different letters were significantly different in a post-hoc Tukey–Kramer HSD comparison. A.opu = A. opuntiae, D.rinc = D. rinconis, S.dup = S. duplocincta.
Download figure to PowerPoint
Visit type did not have a significant effect on seeds per fruit for F. cylindraceus (χ2 = 1·3494, df = 3, P = 0·7174, N = 73). In Wilcoxon/Kruskal–Wallis tests of the effects of bee species and visit type on seeds per fruit in F. wislizeni (pooled across years), bee species was again the only significant factor (Table 3). A post-hoc Tukey-Kramer HSD comparison showed that visits by D. rinconis resulted in significantly more seeds per fruit than visits by S. duplocincta (Fig. 1b). Nectar visits were the most abundant visit type for all three cactus bees, and thus resulted in the most total seeds (data not shown).
Table 3. Wilcoxon/Kruskal–Wallis tests for effects on seeds per fruit, F. wislizeni. Fruits are used as the unit of observation (N = 60). Significant values are in bold
Visit abundance and per-visit pollinator quality can differ in the strength of their effect on total fecundity (percentage of all seeds produced by observed visits). Differences among the three cactus bees in visit abundance had a much greater effect on F. cylindraceus fecundity than did differences in per-visit effectiveness (Fig. 2a,b). In contrast, on F. wislizeni, visit abundances were almost exactly equal among the three cactus bees, but the greater per-visit effectiveness of D. rinconis visits, in terms of both fruit set and seeds per fruit, meant that D. rinconis visits accounted for most of the fecundity (Fig. 2c,d). Thus, with regard to visits by the three cactus bees pooled across years, although visits by D. rinconis accounted for only 33% of all visits to F. wislizeni, they resulted in 79% of all seeds.
Figure 2. Relative abundances and pollinator effectiveness of bee taxa visiting Ferocactus flowers. A.opu = A. opuntiae, D.rinc = D. rinconis, S.dup = S. duplocincta, other poll. = other pollinating taxa, non-poll. = non-pollinating taxa. In 1998 visits by small non-pollinators were not recorded; hence this category is not included: (a) 1996 & 1997 visits to F. cylindraceus; (b) 1998 visits to F. cylindraceus; (c) 1996 & 1997 visits to F. wislizeni; (d) 1998 visits to F. wislizeni.
Download figure to PowerPoint
Overall, female bees visited flowers more often than male bees, for both D. rinconis and S. duplocincta (data not shown). Female to male ratios at flowers ranged from 2·5:1 to 3·5:1 for D. rinconis, and from 2·2:1 to 3:1 for S. duplocincta. There was considerable among-year variation in these proportions. On F. cylindraceus, in 1996–97 only male S. duplocincta were observed, whereas in 1998 only female S. duplocincta appeared. On F. wislizeni, female:male ratios for D. rinconis were 0·56:1 in 1997, but 14:1 in 1998. Also on F. wislizeni, female and male S. duplocincta visits were about equal in 1996 and 1997, but the female to male visit ratio was 4·5:1 in 1998.
Male and female bees did not differ in per-visit fruit set, for either D. rinconis (χ2 = 0·086, df = 1, P = 0·7697, N = 261 flowers) or S. duplocincta (χ2 = 0·004, df = 1, P = 0·9473, N = 105 flowers). Male and female bees also did not differ in per-visit seeds set per fruit, for either D. rinconis (Kruskal–Wallis test, χ2 = 0·2701, df = 1, P = 0·6033, N = 105 fruits) or S. duplocincta (Kruskal–Wallis test, χ2 = 0·3888, df = 1, P = 0·5329, N = 16 fruits). However, because female bees visited more than male bees, female visits were responsible for a higher proportion of total seeds produced (data not shown).
- Top of page
- Methods and materials
This study provides a counter-example to the popular notion that plant-pollinator systems are mostly generalized. The visitor assemblages of both species of Ferocactus were dominated by three species of cactus-specialist bees (a functional group), and these bees were virtually the only pollinators of these plants.
Some cacti may have more diverse assemblages than those observed here. Mandujano, Montana & Eguiarte (1996) found that 89% of all visits recorded to Opuntia rastrera were by cactus-specialist bees. However, Johnson (1992) found that only 30% of all visits to Echinomastus were by cactus bees. Other pollination studies of cacti have not reported visit numbers broken down by taxa.
The introduced honey-bee (Apis mellifera), which is commonly found on Prosopis, Cercidium and Carnegiea flowering at the same time and near F. cylindraceus plants (M. E. McIntosh, unpublished observation), rarely visited Ferocactus flowers (Table 1). This is in agreement with other studies that found few honey-bees on cacti (Schmidt & Buchmann 1986; Osborn, Kevan & Lane 1988; McFarland, Kevan & Lane 1989). Also absent were the generalist bees Bombus (bumble-bees) and Xylocopa (carpenter bees), both common in the Sonoran desert in Arizona.
The cactus-specialist bees were even more dominant among the pollinating visitors than they were among visitors in general. Of visits that resulted in a mature fruit, 99% (F. cylindraceus) and 94% (F. wislizeni) were by cactus bees. Other visitors were either effective pollinators but rare (Megachile and Halictus spp.), or were relatively common but never pollinated the flowers (small halictids such as Augochlorella spp.; Table 1). However, although some types of visitors did not pollinate flowers with a single visit, it is quite possible that multiple visits by such visitors will actually lead to fruit set and seed production. Thus at times or in areas where the cactus bees observed in this study do not occur, the plants might still be able to reproduce.
A single visit to a F. cylindraceus flower resulted in a mean of 149 seeds; mean seeds per fruit for open-pollinated flowers was 575 (McIntosh 2002b). Thus, from one visit, F. cylindraceus flowers received roughly 26% of the pollen needed to set the average number of seeds in an open-pollinated fruit. For F. wislizeni, the mean number of seeds that resulted from a single was 148; mean seed set for open-pollinated flowers was 724. Thus, F. wislizeni flowers received roughly 20% of the pollen needed to match the seed set of open-pollinated flowers. Because visitation rates are relatively high (a visit usually occurred within 15 min of removing the cover), and the flowers are open for several days, it is probable that these flowers receive all the pollen they need to achieve an average seed set in the first day of opening. Data from a previous study demonstrated that these plants are not pollen-limited (McIntosh 2002b).
Previous work on the pollination biology of Ferocactus species is limited to two observational studies of visitors to F. wislizeni that do not report abundances or visitation rates (Grant & Grant 1979; Simpson & Neff 1987). These studies recorded species of Lithurge, Diadasia, Megachile and Perdita echinocacti as visitors to F. wislizeni (all but Megachile are cactus specialists). Other studies of the pollination of cacti with large, open-bowl flowers have not directly measured pollinator effectiveness (e.g. Mandujano et al. 1996).
Previous studies comparing the pollination effectiveness of specialist vs generalist bees on a single plant species have found no difference (Motten et al. 1981; Neff & Simpson 1990; Keys et al. 1995). It is likely that the cactus bees observed in this study were effective pollinators because they are relatively large and hairy, not because they are specialists. Likewise, the fact that most of the (generalist) halictids observed here were small and relatively hairless is probably what prevented them from being effective pollinators. If they had been small, relatively hairless cactus specialists, they would probably still be ineffective pollinators.
This study found that of the two factors contributing to the total realized fecundity of Ferocactus plants, the abundance (quantity) component was the most important. Nectar visits were more abundant than any other visit type, for all three cactus bees, suggesting that the cost of producing nectar might be offset by the gain in quantity of pollinator visits.
Because visitors must use different behaviours to collect pollen vs nectar in Ferocactus flowers, it was expected that pollen-collecting visits would be significantly different in per-visit quality than nectar-collecting visits. This was not the case: there were no significant differences between pollen-collecting visits and nectar-collecting visits. Previous studies examining the benefits to the plant of nectar visits vs pollen visits have reached contrasting conclusions (e.g. Conner, Davis & Rush 1995; Gomez & Zamora 1999; Cane & Tepedino 2001). It is likely that effectiveness of visits is affected more by species-specific interaction characteristics (such as floral morphology) than by the type of reward collected.
These results are in accord with the recent finding that the structure of flower/flower–visitor interaction webs are often nested, with specialized organisms interacting primarily with a subset of their potential partners that are more generalized (Bascompte et al. 2003; Dupont, Hansen & Olesen 2003). In this case, specialized bees visit a generalized plant that hosts many visitors. However, these results also demonstrate the well-documented observation that not all visitors are pollinators – a fact that is overlooked by many studies of plant-‘pollinator’ (actually plant–visitor) interaction webs (Dicks, Corbet & Pywell 2002; Bascompte et al. 2003).
This study has several limitations. The single-pollinator visit method characterizes pollinator effects on only the female reproductive success of plants, and also does not address the conditionality of pollinator effectiveness (Thompson & Pellmyr 1992; Thomson & Thomson 1992). This method does, however, distinguish outcross from self pollen, because in these plants self pollen will result in almost no seeds. Also, because these plants have a very high rate of fruit set under natural conditions, if a single visit does not result in a fruit, it is likely because of deficiencies in the quantity or quality of pollen, not from postpollination effects such as resource limitation. These plants produce very large numbers of seeds under natural conditions, but they will also set a fruit with a relatively small number of seeds, hence the single-pollinator visit methods enables one to measure the quantitative effect of a visit (seed number), not just the qualitative effect (fruit set). Finally, this method provides a common basis for comparing diverse visitors, and because it is commonly used, it facilitates comparisons among studies.
I did not directly measure the abundance or visitation rates of different taxa, and visits were recorded only between 09·00 and 13·00 hours. This limited observation period probably sampled the vast majority of floral visitors, however (as was also found by Mandujano et al. 1996; Minckley et al. 1999). Only one population of each plant species was studied, and it is probable that the visitor assemblages and pollinator assemblages of these plants are geographically variable. A recent study of visitors to F. wislizeni flowers at a location partway between the two sites in this study found a markedly different taxonomic composition of the visitor fauna (dominated by small halictid bees; A. Eaton-Mordas, personal communication). Studies of floral visitors to Ferocactus in other localities are needed to determine if the dominance of specialist cactus bees observed here is prevalent elsewhere.
The effects of different groups or types of visits on the postdispersal success of plant offspring, effects which can potentially swamp predispersal effects, were not measured (Gomez 2000; Herrera 2000). A holistic approach that evaluates the overall effect of visit type or visitor group on the lifetime reproductive success of the is obviously preferable to a reductionist examination of slight differences in fruit or seed set due to visitor identity or visit types (Zamora 2000).