Pollinator-mediated isolation in sympatric milkweeds (Asclepias): do floral morphology and insect behavior influence species boundaries?


Author for correspondence: Susan Kephart Tel: +1 503 370 6481 Fax: +1 503 375 5425 Email: skephart@willamette.edu


  • We explored whether mechanical or ethological differences provide pollinator-mediated floral isolation capable of reinforcing existing species barriers among sympatric Asclepias with divergent floral morphologies: A. incarnata, A. verticiallata and A. syriaca.
  • In a common garden, we quantified pollinator visitation and flight patterns, differences in corporal attachment of pollinia to insects, and the potential outcome of putative floral barriers for interspecific pollination and fruit set.
  • We detected significant variation in the importance, constancy, and behavior of major pollinators on sympatric asclepiads, including Bombus, Xylocopa and large sphecid wasps. Pollinia attach differentially to the arolium on insect legs for A. syriaca, but to the tarsal hairs in other asclepiads. Fruit-set was lower in mixed than unispecific patches of Asclepias.
  • We detected mechanical isolation between A. syriaca and its congeners and a tendency toward wasp pollination in A. verticillata. All three species appear to show some specialization for long-tongued hymenoptera and lepidopterans. Pre–mating barriers provide a potentially effective means of reducing interspecific pollination, but more study is needed in species visited by generalists.


Pollinators are a potentially powerful, yet little understood, selective force in plant speciation. For over two centuries, scientists have linked differences in visitor behavior or morphology to variation in floral characters, eventually invoking a key role for insects in angiosperm diversification (Sprengel, 1793; Darwin, 1876, 1877; Robertson, 1895; Grant & Grant, 1965; Stebbins, 1970; Kiester et al., 1984; Crepet & Friis, 1987). Recent molecular and morphological analyses continue to detect floral radiations involving specific pollinator guilds (Johnson & Steiner, 1997, 2000; Hodges, 1997; Dodd et al., 1999; Goldblatt et al., 2001), but many plants use a broad array of ‘generalist’ visitors, confounding our understanding of their value as pre-mating barriers during primary speciation or as reinforcement after secondary contact (Herrera, 1996; Waser et al., 1996; Johnson & Steiner, 2000). Spatio-temporal variation in pollinator faunas also creates fluctuating selective regimes that may favor flexibility over specialization (Fenster & Dudash, 1991; Ollerton, 1996; Waser, 1998; Levin, 2000). Thus, an important question remains: Can differing floral displays elicit responses from ‘generalist’ animal vectors that are consistent enough to function as pre–mating barriers to hybridization?

Variable floral traits can potentially maintain or reinforce species boundaries through prezygotic mechanisms, by inducing differential visitation or constancy among visitors (Wallace, 1889; Grant, 1949; Waddington, 1983; Broyles et al., 1996; Schemske & Bradshaw, 1999; Ramsey et al., 2003), partitioning the placement of diverse pollen types on insects (Macior, 1965, 1977; Grant, 1994), or establishing novel pollination environments for hybrid species (Straw, 1956, but see Wolfe et al., 1998). Ethological and mechanical differences (i.e. floral isolation), by acting at pollination, can reduce pollen wastage, stigma or stylar clogging, or the production of inviable or unfit offspring, even in species with strong post-mating barriers to hybridization (Levin, 1971; Grant, 1992).

Theoretically, in both specialist and generalist foraging environments, selection should favor floral traits and pollinator behaviors that increase fitness by enhancing conspecific pollen transfer or reducing interspecific pollination. Multiple studies confirm that pollinators affect the magnitude and direction of selection on reproductive characters, from flowering time to floral size, scent, and shape (e.g. Ashman & Stanton, 1991; Galen, 1989, 1996; Robertson & Wyatt, 1990; Campbell et al., 1991). However, while Grant (1994) reviewed 29 cases of pollinator-mediated floral isolation, it is studied mostly in species visited by distinct, and often specialized groups of foragers (e.g. hummingbirds or hawkmoths in Ipomopsis, Aquilegia; Grant et al., 1993). Similarly, in Mimulus, hummingbird and bumblebee pollinators provide strong premating isolation linked to QTLs with major effects on floral traits (Bradshaw et al., 1995; Ramsey et al., 2003). By contrast, pollinator-mediated reproductive isolation is relatively uncommon between plant species visited by ‘generalist’ pollinators or species with strong post–zygotic, genomic barriers (Grant, 1994; Levin, 2000). Analysis of prezygotic factors is also hindered by insufficient data on interspecific flights, rates of pollen transfer, and the extent of mixed loads in sympatric populations (Campbell et al., 2002). Thus, despite intensive study of diverse pollinator communities (Heinrich, 1975; Kalin-Arroyo et al., 1982; Herrera, 1996; Potts et al., 2001), an understanding of premating isolation in highly generalized pollination environments will require more quantitative data relating visitor behavior to specific floral types in natural and manipulated populations.

Study system

Milkweeds (Asclepias sp.) provide an excellent case for assessing the roles of floral morphology and pollinator specificity in maintaining species boundaries. Diversification is extensive: over 100 species exist in the Americas alone (Woodson, 1954; Wyatt & Broyles, 1994; Fishbein, 2001). The flowers and pollination mechanism are specialized but attract a diverse fauna of ‘generalist’ visitors (Roberston, 1981; Macior, 1965; Kephart, 1983). During pollination, insects probe open floral hoods for nectar, and transfer paired, saccate pollinia en masse to the stigmatic chambers of subsequent flowers (Woodson, 1954). During transport, the pollinia rotate 90°, facilitating correct insertion of the margin from which pollen tubes emanate. Tracking pollen movement is also feasible in populations: a grooved corpusculum links each pair of pollinia (Fig. 1), and remains attached to tarsal hairs or other appendages of insects long after pollination, allowing estimates of the number of pollinia carried and inserted by polytropic visitors (Kephart, 1983; Broyles et al., 1996; Fishbein & Venable, 1996a; Ivey et al., 2003).

Figure 1.

Floral structure of Asclepias flowers, showing stigmatic chambers between five saccate hoods, and pollinaria of each species, including a black, grooved corpusculum attached to two saccate pollinia. Drawings by Butch Colyear.

In addition, evolutionary biologists (Grant, 1949; Stebbins, 1974) once plausibly surmised that the complex pollination mechanisms, and the floral size differences among species, provided a lock and key barrier to interspecific pollination. Further study instead revealed strong postzygotic isolation and ineffective mechanical barriers (Kephart & Heiser, 1980; Broyles et al., 1996; Hatfield & Kephart, 2003). However, many species are closely sympatric, and selection for premating floral isolation might reduce interspecific insertion and gametic wastage, reinforce existing reproductive barriers among species (Wallace effect, sensuGrant, 1966, 1981) or limit competition for pollinators via niche partitioning or character displacement (Levin & Anderson, 1970; Kephart, 1983; Waser, 1983; Armbruster & Herzig, 1984; Levin, 2000). Among A. syriaca and A. exaltata, which hybridize in nature, pollinator constancy appears to limit F1 hybrid formation (Broyles et al., 1996), but little is known of the role of premating isolation in species with post–zygotic barriers and generalized pollination. Overall, the floral complexity of asclepiads led Grant (1994) to cite the family as a key, unexplained omission in a list of taxa with floral isolation, and also begs the question: What advantage, if any, does a specialized pollination mechanism confer on plants that employ such a generalized array of polytropic visitors?

We evaluated floral isolation in natural and experimental populations of three perennial asclepiads native to prairie remnants throughout the US: A. incarnata, A. verticillata (Series Incarnatae), and A. syriaca (Series Syriacae) (Woodson, 1954). All three are sympatric in diverse habitats of varied moisture content, yet syriaca (common milkweed) has the broadest habitat range, verticillata (horsetail milkweed) is more prevalent on dry slopes, and incarnata (swamp milkweed) predominates in seasonally wet environments (Curtis, 1955; Kephart & Heiser, 1980; Deam, 1984). Except in four states, geographic sympatry of all species is complete for the entire eastern to mid-western US. Although data on floral biology exist for multiple populations of Asclepias (reviewed in Wyatt & Broyles, 1994, see also Morse, 1994; Morgan & Schoen, 1997; Lipow & Wyatt, 2000; Broyles, 2002; Ivey et al., 2003), most field studies of pollinators have been limited to wild populations of a single species (e.g. syriaca: Jennersten & Morse, 1991; incarnata: Ivey et al., 2003; verticillata: Willson et al., 1979; tuberosa: Fishbein & Venable, 1996a). Thus, we investigated pollinator behavior in relation to floral divergence for sympatric populations at two trophic levels: foraging insects and multispecies assemblages of Asclepias. We asked: Do the differing floral displays of co-occurring milkweeds lead to differential pollinator visitation or positioning of pollinia on insects? Are the ethological or mechanical differences contributing to premating reproductive isolation? Prior research suggested that insects visiting sympatric ascelpiads might differ in relative frequency of visitation, fidelity, and placement of pollinia on their bodies (Macior, 1965; Kephart, 1983). If floral divergence arose during the initial evolutionary radiation of Asclepias, we predicted that species in different clades would differ most in insect pollination. Alternatively, selection to reduce competition or hybridization might lead to stronger floral isolation among A. verticillata and A. incarnata: both species coflower in mid-summer and occur in the same clade (Incarnatae; Fishbein, unpublished). Analysis of floral visitors can provide new insights on the strength of pre–mating reproductive barriers.

Materials and Methods

We limited our study of mechanical or ethological barriers to putative pollinators, that is, floral visitors known to carry or insert pollinia based on direct observation, insect capture, and video recording. First we assessed whether pollinators differed in visitation rates, behavior, and flight patterns, or in the corporal attachment of pollinia. We also determined the effectiveness of reproductive isolation at the pollinator–floral interface by quantifying interspecific pollen deposition and fruit production.

Experimental study population

We established and studied a common garden population of same-aged plants for three years. Plants of all species grew in equal densities (3 × 3 m area plots) in 6 × 3 arrays with two replicates each of unispecific plots (18 plants/plot), 2-species plots (9 plants/species), and 3-species plots (6 plants/species). Plots were located in a rectangular grid within a single 1 ha site within flying distance of known pollinators, as the intent was to simulate unispecific and multispecies patches within a single community. The garden site supports a rich insect fauna on flowering trees and forbs; it occurs near grassland remnants in the vicinity of Bloomington, Indiana, USA.

Floral displays vary among species in morphology and phenology. Flowering extends from June to August; syriaca flowers first but overlaps with the concurrent seasons of incarnata and verticillata (Kephart, 1987). Flowers of these species vary in size, shape, and color (Fig. 1, Table 1) and in the arrangement of hemispheric to globose umbels (i.e. inflorescence unit sensuFishbein & Venable, 1996b). Pollinia and stigmatic chambers differ significantly in size among species (Kephart & Heiser, 1980). All three milkweeds supply large quantities of nectar to insect foragers (Willson & Bertin, 1979; Willson et al., 1979; Ivey et al., 2003).

Table 1.  Floral displays of sympatric Asclepias
  1. Modified from Woodson, 1954. l, length.

Syriaca (Syriacae)globose; few on 2–12 cm peduncles; stems unbranchedPale pink, on 2–5 cm flexible pedicels4–5 mm w. lobed margins; l > gynostegiumincurved; l < hood
Incarnata (Incarnatae)hemispheric; many 1.5–7 cm peduncles; stems branch, form flat-topped displayDeep pink, on 1–1.5 cm erect pedicels1.5 mm, entire; l ≅ gynostegiumincurved; l > hood
Verticillata (Incarnatae)hemispheric; many on 1.5–3 cm peduncles; stems unbranchedGreen-white, on 0.6–0.8 cm flexible pedicels1.5 mm, entire; l < gynostegiumincurved; 2x > hood

Differential pollinator visitation and behavior: ethological factors

We studied insect visitation and behavior in year 3 when plants were well established and flowered abundantly. Observations began in June on syriaca and totaled over 100 h on 43 d of flowering overlap among species (4 July–22 August). We tallied the abundances of each pollinator as instantaneous densities/plot for each asclepiad in survey walks along a specific path (1–4×/sampling date; 08:00–21:00 h). Walks occurred within 1 m of all plants, allowing detection of insect sizes (≥ 8 mm) known to carry pollinia. For each pollinator, we computed the percentage of instantaneous plot surveys in which that insect species occurred at least once. The product of this frequency (f) and insect density/plot (d) were summed for each major pollinator (> 5% frequency) to give its proportional relative importance (RI) in the fauna on each plant species. We also reanalyzed insect capture data from eight mixed and unispecific populations near Bloomington, IN, including four sites of sympatry (Table 1, Kephart, 1983). We used SPSS statistical programs to analyze pollinator data; we computed the proportional similarity of pollinators on each plant species as in Kephart (1983).

We conducted three studies of insect fidelity (i.e. constancy of visitation to plant species), standardizing observation periods to 20 min intervals. In each study we recorded the number and sequence of insect flights between plants and/or plots, number of umbels visited, and species of plant and insect, using an elevated field chair placed 2 m from the edge of each plot. In the common garden, we recorded insect flights in all plots starting in June on syriaca and on all species for multiple observation periods on 16 d of flowering overlap of Asclepias (17 July−22 August). Close inspection confirmed insect identity if needed. Unknown insects were given descriptive names, and later captured and compared to insect collections identified by affiliates of USDA laboratories. Bombus was the most frequent visitor to syriaca and incarnata in the common garden; thus, in a second study there, individual bees were marked once during each of two weeks in June, and followed on multiple days for constancy. In July we evaluated insect preference and constancy in coflowering incarnata and verticillata, by placing single stems of both species alternately in a 4 × 5 array of 20 glass bottles; pruning them to provide equal stem heights, umbel density, and flower number; and replenishing water and flowers over 2 d. Foragers were presumed naïve to these asclepiads as the urban site in Bloomington, IN was approx. 4 km from known populations.

On 15 d from June to August, we filmed over 365 floral visitors representing 16 putative species, and quantified the behavior and movements of major pollinators. Each insect was filmed until it left the common garden (or for ≤ 15 min) in the order observed along survey routes. Frame analysis in slow motion enabled us to determine if presumed major pollinators transfer pollinia, and to detect behaviors leading to pollinium removal and insertion (e.g. foraging vs grooming). We predicted that the floral display might affect the direction of approach to flowering umbels, or the principal body contact of insects on flowers (i.e. the most common of all visible contacts of a forager on each floral display). Filming occurred under clear and overcast skies when temperature and precipitation levels were suitable for insect flight. Plant and insect vouchers are deposited with Indiana and Willamette University herbaria and entomological collections and at USDA laboratories, Beltsville, MD, USA.

Pollinium placement: mechanical factors

Mechanical and ethological factors interact in pollination, resulting in variation in the body part to which milkweed pollinia attach and from which pollinia are transferred to subsequent flowers. To evaluate both factors, over 1000 floral visitors were captured during three flowering seasons in the garden and natural populations. Crepuscular and nocturnal visitors were infrequent, so we sampled noctuid, sphingid, and geometrid moths using conventional light traps. All insects were identified to species or genus, and examined under magnification (10–30 X) for the position and number of pollinia (0–2) associated with each attached corpusculum. These data provide indirect estimates of pollen load (PL, or number of pollinia initially removed) and the proportion of these pollinia (PT) that were either lost or transferred to other flowers. Also, the combination of data on pollinium position, interspecific foraging flights, and interspecific insertion allowed a more definitive test of the hypothesis by Robertson (1887) and Macior (1965) that the dispersion of pollinia on different parts of an insect constitutes a mechanical barrier to hybridization (Robertson, 1887; Macior, 1965). Strong support for mechanical barriers exists if pollinia of different asclepiads vary significantly in placement on insects and if interspecific pollinium insertions are absent or reduced in frequency relative to the frequency of foraging flights between focal species.

Forager-dependent mechanical effects are also implied if the principal contact of an insect's body on Asclepias flowers on videotape differs from the actual body positions at which pollinia of an asclepiad attach. Such discrepancies help us infer whether floral structure differentially affects pollinium attachment to insects, irrespective of visitation preferences. Mechanical factors might also be implied if guilds of pollinators with high PLs or presumed rates of transfer (PT) are clustered by size instead of taxonomic affinity.

Effectiveness and fitness advantage of floral isolation

Both flight directionality and plant species influence the potential fate of pollinia in ways that may differentially contribute to reproductive isolation. Thus, we evaluated the potential selective advantage of floral isolation by species, partitioning the inconstant foraging flights of major pollinators into two components. We computed indirect estimates of the amount by which interspecific flights reduced both male fitness (i.e. gametic wastage via pollen dispersal) and female fitness (i.e. loss of or competition for insertion positions caused by foreign pollinia). For each species i, we estimated the potential percentage loss of its pollinia to other asclepiads caused by infidelity as 100 • (# of interspecific flights from species i/total flights from species i). Similarly, potential losses in female fitness of species i caused by infidelity were estimated by: 100 • (# of interspecific flights to species i/total flights to species i).

We also tested the effectiveness of floral isolation by direct determination of interspecific pollination levels in mixed and unispecific plots in the common garden. We collected 10–12 senescent umbels per week in unispecific and mixed plots of each species, except where few umbels were in flower. Inserted pollinia were extracted from the stigmatic chambers of all flowers and identified to species. We marked and counted flower number in umbels of nearby plants weekly to determine the percentage of total flowers initiating and maturing fruit throughout the season.


Divergence in pollinator visitation

The taxonomic composition of insect visitors was similar across milkweed species in that Bombus griseocollis and its congeners emerged as the most frequent taxon observed in field surveys (19–32% of taxa; Table 2). On verticillata, digger wasps (Sphex, 22%) assumed a major role, whereas wasps were superceded by carpenter bees (Xylocopa virginica) and Bombus on incarnata, and by Bombus and Apis on syriaca. Butterflies were 2× more common on incarnata and syriaca (22–24%) than on verticillata (10%), and their relative importance also varied (Table 2): three species of nymphalids prevailed on syriaca (e.g. Epygaeus clarus, Vanessa atalanta) while the pipevine swallowtail (Battus philenor) and other papilionids preferentially visited incarnata. Chaulignathus was the only consistent pollinia-carrying Coleopteran visitor.

Table 2.  Percent composition of pollinators comprising ≥ 2.5% of fauna on Asclepias in a common garden. Entire season (June–August) and flowering overlap (4 July–23 August) are shown for early flowering syriaca. Also depicts mean number of insects recorded for proportion of N survey walks in which a given insect was detected at least once
Insect TaxonAsclepias syriacaAsclepias incarnataAsclepias verticillata
Entire seasonDuring overlapInsects/ walkEntire seasonInsects/ walkEntire seasonInsects/ walk
N946 867726 366 
Hymenoptera (bees) 58.159.3 7.0 ± 0.32 59.12.6 ± 0.15 45.61.6 ± 0.09
Bombus sp. 28.829.5 4.3 ± 0.28 32.23.8 ± 0.25 19.11.2 ± 0.06
Apis mellifera 27.427.910.4 ± 0.54  4.11.9 ± 0.51 12.32.3 ± 0.20
Xylocopa virginica  1.2 1.3 1.0 ± 0.00 20.21.7 ± 0.10  9.81.1 ± 0.04
Hymenoptera (wasps)  3.5 2.9 1.1 ± 0.07 14.41.2 ± 0.04 36.31.3 ± 0.07
Sphex ichneumoneus  1.3 0.6 1.4 ± 0.25  6.61.2 ± 0.06 16.11.6 ± 0.14
S. pennsylvanicus  0.1 0.1 1.0 ± 0.00  1.51.2 ± 0.12  6.31.0 ± 0.04
Cerceris sp.  0.7 0.8 1.1 ± 0.14  0.71.0 ± 0.00  4.91.0 ± 0.00
Bembix sp.  0.5 0.5 1.0 ± 0.00  4.31.1 ± 0.06  1.61.0 ± 0.00
Myzinum sp.  0.0 0.0 0.0  0.61.3 ± 0.33  4.61.0 ± 0.00
Lepidoptera (butterflies) 22.122.2 1.1 ± 0.03 24.11.2 ± 0.04 10.41.0 ± 0.03
Battus philenor  1.0 1.0 1.0 ± 0.00  7.61.5 ± 0.11  0.00.0
Epygaeus clarus  3.7 3.9 1.1 ± 0.06  0.11.0 ± 0.00  0.00.0
Vanessa atalanta  3.5 3.7 1.3 ± 0.11  0.00.0  0.31.0 ± 0.00
Speyeria cybde  2.6 2.3 1.0 ± 0.00  1.01.0 ± 0.00  0.00.0
Danaus plexippus  2.7 2.9 1.0 ± 0.00  2.91.1 ± 0.04  0.00.0
Colias philodice  0.7 0.7 1.0 ± 0.00  2.91.0 ± 0.05  3.81.1 ± 0.07
Lepidoptera (moths)  5.1 5.3 1.2 ± 0.06  1.71.0 ± 0.00  1.91.0 ± 0.00
Coleoptera 10.7 9.9 2.6 ± 0.23  0.71.0 ± 0.00  5.51.8 ± 0.36
Chaulignathuspennsylvanicus 10.710.0 2.6 ± 0.23  0.31.0 ± 0.00  5.51.7 ± 0.10
Diptera  0.4 0.5 2.5 ± 1.5  0.00.0  0.31.0 ± 0.00

When insect frequencies and densities on flowers were compared in a single index of importance (RI, Fig. 2), the bee domination of larger-flowered syriaca and incarnata is marked (> 75% of major pollinators), relative to an even distribution of major hymenopteran visitors on verticillata. The distribution and densities of the four major taxa of native pollinators (Bombus, Xylocopa, Sphex and butterflies) differed significantly for each pairwise combination of Asclepias (2 × 4 contingency tables, χ2 > 130, P < 0.001). Proportional similarities (PS) among species combinations based on insect densities were: 78.3% for syriaca, incarnata; 61.7% for incarnata, verticillata; and 43.8% for syriaca, verticillata.

Figure 2.

Depicts relative importance of major pollinators on each species of Asclepias. Depicts percent abundances of pollinators based on frequency and density of each insect taxon on flowers during survey walks throughout flowering overlap.

Capture data from parapatric and sympatric populations in Indiana revealed similar visitation patterns, but with some variation among populations (Table 3). Bombus was the most common visitor to syriaca and incarnata at all field sites except where outnumbered by honeybees (MMO). In contrast, except where dominance was shared by Sphex and Bombus at PAY, wasps were more common on verticillata including Tiphiidae (Myzinum) and Vespidae (Polistes fuscatus) as well as Sphecidae. In sympatric populations, butterflies were again more common on incarnata and syriaca than on verticillata (Table 3). Except for Cisseps fulvicollis, diurnal and nocturnal moths were infrequent at most sites, rare during survey walks in our study, and carried few pollinia. Moths varied in importance in other studies of these asclepiads also (Frost, 1965; Willson et al., 1979; Morse, 1981).

Table 3.  Abundances (%) of major pollinators captured in unispecific (*) and sympatric populations of Asclepias syriaca (S), incarnata (I), and verticillata (V). Localities PAY/CAS and COL/STM occur in close proximity, within 10 km of common garden EXP
  1. Computed from Kephart (1983). Bold depicts two most common hymenopteran taxa for each asclepiad; bold-underline highlights plant species with highest percentage of lepidoptera relative to a sympatric congener. Totals are < 100% (i.e. other hymenoptera, hemiptera).

Bombus33.8 27.712.926.7 8.227.4 10.725.066.759.452.814.00.0
Apis28.2 20.911.426.713.116.4  6.518.712.5 0.011.532.00.0
Xylocopa 0.0  7.2 4.3 0.0 4.9 4.1  7.120.8 0.0 9.4 0.0 0.00.0
Sphex 4.3  7.6 8.623.321.326.0  3.612.5 6.3 6.2 1.1 8.021.7
Polistes 2.8  5.320.0 0.0 1.611.0  3.6 6.3 0.0 0.0 0.0 0.0 0.0
Myzinum 0.0  1.5 7.2 0.0 0.0 0.0 15.5 0.0 0.0 6.3 0.0 0.026.1
LEPIDOPTERA15.5  8.1 1.413.329.510.9 23.3 6.3 8.412.529.830.017.4
COLEOPTERA12.7  1.514.3 6.7 3.2 0.0  7.7 8.3 2.1 0.0 0.016.0 0.0
DIPTERA 0.0  0.8 2.9 3.3 0.0 0.0  0.6 0.0 0.0 3.1 3.4 0.0 8.7

Foraging behavior and fidelity

Bombus and other ‘generalists’ could effect ethological isolation if shared pollinators are highly constant. However, individually marked bumblebees visiting syriaca and incarnata for 9 d in June (n = 478 flights) were remarkably inconstant: 37.8% of all flights were interspecific and only 14.3% of individual bees were constant to a single plant species over 3 d. By contrast, inexperienced Bombus working experimental arrays of verticillata and incarnata showed high constancy (89% flight fidelity), as did other foragers (Hymenoptera: Apis 97%, Sphex 82%; Lepidoptera: Speyeria cybde 100%). Pollinators (n = 226 flights) visited incarnata significantly (1.5×) more often than verticillata both initially (χ2 = 4.5, P < 0.05, df = 1), and after 2 d (χ2 = 4.8, P < 0.05); mean foraging time/visit was marginally greater for incarnata (2.7 ± 0.3 m) than verticillata flowers (2.2 ± 0.4 m; P= 0.07).

Major pollinators on all three species in common experimental plots had fidelity rates that varied with both insect and plant taxa (Fig. 3). For native pollinators, constancy ranged from 40 to 57% (Sphex and lepidopterans on syriaca) to 89–91% (Sphex on verticillata, lepidopterans on incarnata). For Bombus alone, the potential reductions in pollen dispersal (male fitness) caused by interspecific flights to other asclepiads were 18.2–25.7% (Fig. 3).

Figure 3.

Pollinator constancy of each major pollinator as the percentage of total flights from a focal species i that were intraspecific.

Based on videotapes, insects also differed in their approach (n = 271) and behavior (n = 167) on flowers of varied milkweeds. Pollinators uniformly approached the largely flat-topped floral displays of incarnata from above (Fig. 4; χ2 = 71.6, P < 0.001, df = 2) whereas all visitors except Sphex approached the vertically borne umbels of syriaca and verticillata laterally or from below (χ2 = 7.6,10.0; P < 0.05. df = 2). Videotaping of visitor behavior also showed that the arolium between the claws of insect legs was consistently the principal contact on flowers of all milkweeds (χ2 = 4.3, P < 0.05, df = 1) by Apis (100% of principal contacts), Bombus (73100%) and lepidopterans (67–100%). Xylocopa exhibited similar behaviors and aroliar contacts on verticillata (58%) and syriaca (67%), whose flexible pedicels deflect readily under its weight, but tarsal hairs were the main contact on erect pedicels of incarnata by Xylocopa (67%). Tarsal hairs were also the key contact point for Sphex on incarnata (85%; χ2 = 7.4, P < 0.05, df = 1) and verticillata (97%; χ2 = 15.6, P < 0.001). During foraging, wasps seemed to reach stability by placing their leg tarsi on multiple outlying flowers with their bodies above the inflorescence, while bees placed their legs well below the hoods, with their bodies spanning one or two adjacent flowers.

Figure 4.

Percentages of insects observed on videotape that approached flowering umbels from above, below, or laterally relative to the inflorescence unit.

Pollinium placement: mechanical factors

The interaction of pollinator behavior and plant/insect morphology also led to varying patterns of pollinium deposition on insects. All major hymenopteran and lepidopteran taxa visiting syriaca carried significantly more pollinia on the arolium (a pad between claws of an insect leg) than on any other body part (82–92%, n= 4 taxa compared for arolia vs other; Table 4a, χ2 = 21.0, P < 0.001, d.f. = 1). By contrast, whereas the arolium was also the principal body contact for Apis and Bombus on other asclepiads, tarsal hairs formed 82–98% of attachments for native bees with pollinia of smaller-flowered incarnata and verticillata (Table 4b, n = 5 taxa compared for tarsal hairs vs other, χ2 = 46.2, P < 0.001, df = 1). Macior (1965) also noted the aroliar attachments of syriaca, but he did not include quantitative data for corpusculae attached to leg hairs of most insects, and implied erroneously that claws were the primary attachment position for pollinaria of incarnata. In addition, the dispersion of the remaining pollinia of smaller-flowered Asclepias on claws and arolia varied by plant and insect taxa, with relatively more incarnata pollinia on tarsal claws of Apis and lepidopterans (14–26%), and fewer on Sphex (3.2%) than for pollinia of verticillata on the same insect species (Table 4b). Otherwise, our observations extend and corroborate Macior's detailed descriptions of insect behavior.

Table 4a.  Distribution of pollinia on major pollinators of Asclepias syriaca reported as percentage of total distributed pollinia
Insect Body Appendage Apis melliferaBombus sp.Sphex sp.Lepidoptera
 Tarsal hair 0.48 5.13 5.83 0.24
Front LegClaw 0.12 0.24 0.97 0.24
Tarsal hair 0.24 1.71 0 1.94
Middle LegClaw 0.12 0 0 1.21
Tarsal hair 0.60 7.82 0 4.60
Hind LegClaw 0 0.49 0 1.69
Arolium23.621.8 9.7123.0
Mouthparts  7.52 1.96 0.97 7.51
Table 4b.  Distribution of pollinia on major pollinators of Asclepias incarnata (I) and A. verticillata (V) reported as percentage of total distributed pollinia
Insect Body Appendage Apis melliferaBombus sp.Xylocopa virginicaSphex sp.Lepidoptera
 Tarsal hair11.620.423.934.49.4324.844.148.27.3832.7
Front LegClaw 9.02 0.99  4.37 2.06 2.12 0.38 1.52 4.93 7.63 0
Arolium 6.14 9.16  1.86 2.21 0.11 0 2.27 5.53 0 0
Tarsal hair20.528.924.730.833.937.323.523.939.230.6
Middle LegClaw 9.21 1.83  1.45 2.65 1.51 0.29 0.54 2.40 6.11 0
Arolium 7.4910.8  0.72 0.59 0 0 0.49 2.76 0 2.04
Tarsal hair18.522.135.524.249.936.024.89.7438.234.7
Hind LegClaw 7.32 0.99  1.72 0.88 1.07 0 1.19 0.96 0.51 0
Arolium 3.94 3.52  0.68 0.29 0.03 0.10 0.77 0.96 0 0
Tarsal hair50.671.484.189.493.298.192.481.884.898
All LegsClaw25.6 3.81  7.54 5.59 4.47 0.67 3.25 8.2914.25 0
Arolium17.623.48  3.26 3.09 0.14 0.1 3.53 9.25 02.04
Mouthpart  6.31 1.27  5.14 1.92 1.50 0.38 0.88 0.60 1.02 0

Potential effectiveness of ethological and mechanical differences

Potential losses in pollen to other asclepiads as a result of pollinator infidelity were direction-dependent, and greatest for flights from syriaca to incarnata recipients (19.7% of all flights leaving syriaca), and from verticillata to incarnata recipients (19.1%; Table 5). However, no insertions of syriaca pollinia into other flowers occurred, and interspecific insertions exceeded 1% of total insertions only for incarnata and verticillata interactions (Table 5). Reductions in female fitness based on the flights received by these taxa were potentially high during flowering overlap as well: 17.4% of the pollinia-carrying visits to verticillata flowers were interspecific (Table 6). Estimated percent losses in female fitness caused by the receipt of pollinia from congeneric species were low for incarnata, which received many more total, and largely conspecific, flights than its congeners.

Table 5.  Interspecific flights of pollinators (%) from Asclepias species, and percentage interspecific insertions of pollinia in flowers
Pollen DonorN1Interspecific flights (%)N2Interspecific insertions in recipient chamber (%)
  1. N1, number of all flights from pollen donor during period of overlap. N2, number of pollinia scored that season in random flowers of asclepiads.

A. syriaca  2009 
 S → I 30419.7 0.0
 S → V 256 4.7 0.0
A. incarnata  1789 
 I → S 864 6.5 0.82
 I → V2286 5.6 1.61
A. verticillata  683 
 V → S 572 1.7 0.41
 V → I 86319.1 3.52
Table 6.  Percentages of interspecific flights received by focal Asclepias species i relative to pollinium insertion in its stigmatic chambers (estimates pollen receipt as female). Fruit initiation and fruit production for marked flowers in 2-species mixed and unispecific plots in garden
Interaction on focal species% of flights to focal speciesPollinia inserted/Flower% fruit Initiation (×100)% fruit Production (×100)
  1. N-values in last columns are the numbers of initiated and matured fruits, respectively, from all flowers produced in random umbels marked at first flowering.

A. incarnata(N) UniMixUniMixUniMix
 V → I6.90.72 ± 0.210.62 ± 0.2016.114.11.851.44
 S → I2.80.00.0    
N2382 flights1392 flwrs992 flwrs3532162722
A. verticillata
 S → V15.
 I → V2.10.26 ± 0.160.27 ± 0.11    
N839 flights1283 flwrs1240 flwrs13549175

Although the potential reductions in female fitness resulted in insignificant differences in pollinium insertions into stigmatic chambers of these species (P > 0.05), the number and percentage of fruits initiated and matured were higher in unispecific plots of incarnata and verticillata (Table 6), and a significantly larger proportion of initiated fruits reached maturity in unispecific plots of both species (χ2 = 19.2, P < 0.01, df = 1). Pollinia received by syriaca from other species were small relative to the stigmatic chamber size and produced short pollen tubes; also, as donors, the large syriaca pollinia do not insert in the small chambers of congeners in natural populations (Kephart & Heiser, 1980).


Experimental gardens have provided an effective, classical tool for understanding variation among plants (Turresson, 1925; Clausen et al., 1940). We asked if a shared community of generalist pollinators might be differentially affected by the varying floral displays of three sympatric asclepiads, and whether mechanical or behavioral traits (i.e. floral isolation) could maintain or reinforce existing post-mating barriers in Asclepias.

Differential pollinator visitation

Floral isolation of plant species can be mediated by generalist pollinators if the foraging behavior of individual insects is highly constant and consistent over time, or if plants specialize on a subset of generalist visitors. Overall species richness totaled 177 species observed or captured with pollinia in experimental plots and natural populations of syriaca (55), incarnata (108) and verticillata (153). The greater pollinator diversity on incarnata and verticillata likely represents higher among population variability relative to syriaca, as the common garden plots yielded 30–32 insect taxa on all species. However, pollinators representing three insect orders clearly discriminated among sympatric asclepiads. Overall, Xylocopa was most abundant on incarnata, and Bombus and Apis preferentially visited syriaca and incarnata over verticillata. By contrast S. ichneumoneus and other wasps were the primary floral visitors to verticillata, and a beetle, Chaulignathus pennsylvanicus, selectively visited verticillata and syriaca over incarnata. Butterflies also differed in their prevalence on syriaca (Nymphalidae), incarnata (Papilionidae), and verticillata (Pieridae, Colias philodice). Such discrimination by pollinators among the floral displays of Asclepias is not surprising, and corroborates recent evidence that pollen-foraging insects use search images to distinguish similar flowers and reduce flight times (Goulson, 2000).

Across geographic regions, strong similarities exist in the pollinator faunas on individual species of Asclepias, suggesting potential consistency in major pollinators. B. griseocollis was the most abundant native bee on all three species in this study, and in studies in Wisconsin and Illinois (Macior, 1965; Willson et al., 1979). However, only on verticillata did sphecid, vespid, and tiphiid wasps outnumber bees at these sites (e.g. 75% of captured hymenopterans compared to 13–20% on incarnata and syriaca; Macior, 1965). On syriaca, bumblebees constituted 72% of diurnal visitors in Maine (Morse, 1981), and Apis 29–49% of the fauna in Illinois (Willson & Bertin, 1979). This geographic similarity suggests either modest selective pressure for divergence in visitation among generalist visitors, inherent differences in pollinator attraction to varied asclepiads, or multiple instances of similar shifts in behavior as insects encounter and respond to differing floral morphologies or locally variable resources within communities. In our array experiment, naïve bees on Asclepias visited significantly more incarnata than verticillata flowers. These data, and the high rates of visitation to the former species in garden plots indicate a strong preference of many insects for incarnata.

Despite regionally similar faunas, the prevalence and inconstancy of Bombus and other pollinators in this study seem to preclude divergence in pollinator visitation as an effective pre-mating barrier. Many of the insects captured in our experimental plots carried two or more pollinium types, and Macior (1965) also noted pollinia from diverse species on a single insect. Significant variability in the major pollinators of incarnata (Ivey et al., 2003) also reinforces the inadequacy of differential visitation as an isolating mechanism in Asclepias: in 1995, carpenter bees were the most abundant visitor along with three genera of sphecid and tiphiid wasps whereas the highest visitation rates in 1996 were by lepidopterans and a single, dominant sphecid (Tachytes). By contrast to our study, Bombus had a minor role (Ivey et al., 2003). Robertson (1891) also recorded numerous wasps as well as bees on all three species.

Mechanical factors and pre-mating isolation

Differential corporal attachment of pollinia remains a possible means of pollinator-mediated floral isolation in selective environments of high sympatry, if competition for pollinators exists or if interspecific insertions result in fitness reductions leading to selection against gametic wastage. In analyses of over 1200 records of the positions of pollinaria on captured pollinators, we detected definitive evidence of mechanical barriers only between syriaca and its congeners. In syriaca, the distal pad-like arolium garnered the majority of pollinia, and is readily engaged by the large groove in the copusculae of syriaca pollinaria. Medium-sized bees also draw their legs upward along the large interhood space, guiding the arolium centrally across both stigmatic chambers and corpusculae. This specificity in pollinium placement, and the high efficiency or apparent pollen transfer in syriaca also suggest morphological specialization (Kephart et al., unpublished). By contrast, attachments to tarsal hairs prevailed for pollinaria of incarnata and verticillata, even when arolia formed the principal floral contact. The corpuscular grooves of these species may be too small to consistently engage the arolium, or the narrower interhood spaces may interfere with secure attachment of this structure. Ivey et al. (2003) also detected differences in the behavior or morphology of incarnata pollinators that may influence pollen transfer efficiency.

Overall, mechanical differences in the fit between insect appendages and floral structures in Asclepias impose a consistent physical barrier that reinforces isolation between syriaca and the two smaller-flowered species. This floral isolation arises from the combination of several factors: early flowering; repeated attachment of pollinaria to the distal arolium of insect legs; and large floral size, which prevents insertions into foreign stigmatic chambers (thereby avoiding potential reductions in male fitness by retaining pollinia on foragers who may visit conspecific flowers). The large floral size also limits access to its ovules by smaller pollinia whose pollen tubes may fail to form pollen tubes of sufficient length for fertilization (Kephart, 1981; Hatfield & Kephart, 2003). The barrier is effective in that interspecific insertions into syriaca flowers were rare during flowering overlap (1.2% of insertions in experimental plots). In natural populations, foreign insertions into its flowers exceed 20% only during late-season flowering, or at high floral densities of other asclepiads (Kephart & Heiser, 1980). Broyles et al. (1996) also found that pollinators inserted 40% fewer of large syriaca pollinia into A. exaltata flowers than expected based on mixed pollinium loads.

Mechanical and ethological factors failed to isolate incarnata and verticillata when grown in sympatry, but populations of these species diverged more in pollinator visitation than expected. Except for Bembix americana, the wasp densities on verticillata flowers were 2–4× greater than for incarnata, on which carpenter bees were 5× more abundant. This suggests the possibility of selection for divergent guilds of generalist visitors in habitats with rich pollinator faunas. Differential visitation clearly does not prevent interspecific visitation by individual foragers, but it may reduce the likelihood of interspecific pollen transfer. Attracting relatively constant foragers (e.g. 89% fidelity of S. ichneumoneus) might be advantageous, given the potential for loss of verticillata pollen (19.1%) to incarnata flowers. As a pollen recipient, verticillata is also at greater risk for stigma clogging by the large incarnata pollinia whose longer pollen tubes easily traverse its styles (Kephart, 1981). By contrast, although bees and butterflies were more constant on incarnata relative to the same insects on verticillata, the potential fitness gains from specialization or higher constancy are lower for incarnata. A higher ratio of interspecific insertions/flower in mixed: unispecific plots of verticillata (1.7) relative to mixed: unispecific plots of incarnata (1.1) also supports this contention. Fruit initiation and maturation were low overall for verticillata, and fruits initiated in mixed plots had a significantly lower probability of maturation than in unispecific plots. Theoretical and empirical data show that rare flowers are at a reproductive disadvantage and thus benefit more from flower-constant pollinators (Levin & Anderson, 1970; Goulson, 1994; Kunin & Iwasa, 1996), but a similar argument might also be made for the flowers least preferred by pollinators even if present at high floral densities.

Floral specialization and generalist pollinator-mediated isolation

Grant (1994) highlighted the absence of floral isolation in the specialized flowers of Asclepiadaceae, and he rightly predicted the need for further study. Our study implies some ethological and mechanical divergence in pollination for sympatric Asclepias, which have highly specialized flowers yet depend on insect ‘generalist pollinators’. Mechanical isolation was effective only across clades, between syriaca and all other congeners. This premating isolation may simply be a byproduct of the initial radiation of these species. However, the broad ecological tolerance of syriaca (Curtis, 1955) may also create more opportunities for interspecific hybridization historically and today, and thus a greater potential selective advantage for floral divergence. Pre-mating reinforcement of cross–incompatibility barriers between close congeners incarnata and verticillata is also possible if a consistent and extensive use of relatively constant wasp pollinators emerges in verticillata. Unfortunately, published data on pollinators exist for only 5.8% of the family, including specialized pollination of A. woodii by fast-flying beetles (Scarabaeidae) that are purported to occupy the niche of large bees in South African grasslands (Ollerton, 1997; Ollerton et al., 2003).

For a vast majority of Asclepias species, however, an important question still remains: why use generalist pollinators? From a plant perspective, attracting a generalist fauna can assure reproduction and reduce competition by broadly exploiting the most abundant visitors (Motten, 1986; Rathcke, 1988). Because pollen as well as resource limitation of fruit set occurs variably in Asclepias, competition for pollinators is potentially important (Kephart, 1983; Morse & Fritz, 1983; Broyles & Wyatt, 1997). Inconstant foraging on a sympatric, and morphologically similar guild of plants (e.g. Asclepias) can also enhance the resources available to foragers as well as their efficiency on the specific floral structures involved (Rathcke, 1988; Gross, 1992). These arguments establish a reciprocal basis for evolution, yet it remains largely untested. Slight support for facilitation exists in syriaca: during flowering overlap with verticillata and incarnata, that is, late-season flowers supported higher densities of Bombus and produced more fruit (Kephart, 1987). Ethological divergence to attract a large, but specific subset of generalist pollinators (e.g. wasps) while retaining access to a more diverse assemblage might also be advantageous for species that are less preferred by the most abundant flower visitors within a community. We challenge others to continue to pursue in-depth studies of pollinator-mediated floral isolation in sympatric species and to augment field and molecular studies of natural populations with common garden plots. Although molecular studies have clarified putative or contested cases of hybrid floral isolation in Penstemon (Wolfe & Elisens, 1993; Wolfe et al., 1998), pollinator-mediated isolation in Aquilegia (Hodges & Arnold, 1994; Hodges, 1997), and homoploid hybrid speciation in Helianthus (Rieseberg, 1991; Gross et al., 2003), in each case the studies were effective because of prior or subsequent field study of the interacting populations. In addition, the relative contributions of pre- and postzygotic factors to reproductive isolation are rarely quantified, making evolutionary trends difficult to detect (Ramsey et al., 2003).


The authors would like to thank Willamette University students S. DeMay and A. Henderson for contributions to literature search, data entry, and preliminary analyses of videotapes and B. Colyear for his exceptional illustrations. S. Broyles, J. Butler, C. Ivey, C. Fenster, V. Grant, H. Lindon, C. Longnecker, J. Pariera, P. Vitt and several anonymous reviewers also gave helpful comments related to this paper. The Atkinson Fund of Willamette University and Indiana University provided logistical and monetary support for the research and data analysis.