• In plant populations where reproductive output is limited by pollinator visitation, plants with attractive floral displays should have a selective advantage. We examined the effect of inflorescence height on pollination success in Primula farinosa, which is dimorphic for scape length.
• To test the hypothesis that fruit and seed initiation are more strongly pollen-limited in the short-scaped than in the long-scaped morph, and that this difference is affected by spatio-temporal variation in pollen limitation, we conducted a hand-pollination experiment in four populations over 2 yr.
• Pollen limitation of fruit initiation varied among populations and years, and was stronger in the short-scaped than in the long-scaped morph.
• The results suggest that interactions with pollinators will need to be considered for a full understanding of the maintenance of this striking polymorphism. The study also shows that, although pollen limitation is likely to vary in space and time in many plant species, such variation is not necessarily associated with variation in selection on floral characters.
Plant–animal interactions have important effects on community structure, population dynamics and trait evolution in plants. Pollen availability often limits seed production (Burd, 1994; Ashman et al., 2004) and also has consequences for the evolution of traits contributing to floral display (Kingsolver et al., 2001; Fenster et al., 2004). For traits associated with pollinator attraction, selection should often be affected by the level of pollen limitation (Ashman & Morgan, 2004). Spatio-temporal variation in pollen limitation has been demonstrated in several study systems (Campbell, 1987; Dieringer, 1992; Ågren, 1996; Alexandersson & Ågren, 1996; Dudash & Fenster, 1997; Baker et al., 2000; Goodwillie, 2001; Price et al., 2005), and may result in spatio-temporal variation in the average performance and population dynamics of plants. Spatio-temporal variation in interactions with pollinators may also influence trait-fitness relationships and translate into spatio-temporal variation in selection. While spatial variation in plant–pollinator interactions could promote divergent evolution of floral characters, temporal variation in interactions would limit the potential for divergent evolution (Caruso et al., 2003). This stresses the need for multipopulation studies over several years, as a single-year multipopulation study may emphasize the potential for divergent evolution, whereas a single-population multiyear study may emphasize the opposite. Spatio-temporal variation in selection on floral traits has sometimes been reported (Schemske & Horvitz, 1989; Caruso, 2000; Maad, 2000), but the causes of this variation have rarely been demonstrated experimentally (Fenster et al., 2004; but cf. Miller, 1981).
Primula farinosa is a small perennial herb with a basal leaf rosette. It usually has a long-scaped inflorescence. However, in seminatural grasslands on the large island of Öland, south-east Sweden, populations are polymorphic for scape length and also include a short-scaped morph (Fig. 1). As a result, plants have their inflorescence displayed either well above the soil surface, or very close to the ground. This difference in scape morph may influence interactions with grazers, pollinators and seed predators, and variation in the intensity of these interactions may affect the relative performance of the two scape morphs.
Severed scapes and fruit capsules with exit holes are readily linked to grazing or seed predation, respectively, and the magnitudes of these interactions are therefore straightforward to monitor. However, to infer pollination success from patterns of fruit initiation requires an experimental approach, as low levels of fruit initiation can also be attributed to a lack of resources. In a previous study of a single P. farinosa population on Öland, the short-scaped morph was more pollen-limited than the long-scaped morph, and the difference was larger in high than in low vegetation (Ehrlén et al., 2002). However, it is not known how far variation in pollination intensity affects among-population variation in fruit and seed initiation in the two scape morphs. To assess spatio-temporal variation in selection exerted by pollinators on scape length, it is necessary to link pollen limitation with scape length in multiple populations over several years.
We conducted a supplementary hand-pollination experiment in four populations in 2 yr to determine the extent to which variation in fruit and seed initiation can be attributed to interactions with pollinators. We asked: (1) Are fruit and seed initiation more strongly pollen-limited in the short-scaped than in the long-scaped morph? (2) Does pollen limitation vary between populations and years? (3) If present, does such spatio-temporal variation in pollen limitation translate into spatio-temporal variation in pollinator-mediated selection, or is the short-scaped morph consistently more pollen limited?
Materials and Methods
Primula farinosa L. (Primulaceae), bird's-eye primrose, is a hermaphroditic, self-incompatible, distylous, perennial herb (Hambler & Dixon, 2003). It is distributed in Europe from central Sweden and Scotland to central Spain and Bulgaria (Tutin et al., 1972). The long-scaped morph occurs throughout this area, whereas a short-scaped morph occurs on the large Swedish island of Öland (Lagerberg, 1957). Crossing experiments suggest a simple Mendelian inheritance of scape morphology, with a dominant allele for short scape (J.Å. and J.E., unpublished data). Primula farinosa is found in lime-rich, moist meadows, and its persistence at a given site is favoured by disturbances such as grazing (Sterner & Lundqvist, 1986; Lindborg & Ehrlén, 2002). It produces leaves in a basal rosette and normally three to 12 pink flowers in a single umbel. The number of inflorescences per basal rosette is usually one but occasionally up to three. Rosettes sometimes divide, but do not form extensive clones. Flowering takes place in May. In the study area, P. farinosa is pollinated mainly by butterflies (especially Pyrgus malvae) and solitary bees (especially Osmia bicolor). The fruit is a multiseeded capsule which matures in early to mid-July. Entire inflorescences are sometimes grazed or damaged by trampling by livestock (cattle, sheep and horses). The initiated fruits are often attacked by the larvae of the small tortricid moth Falseuncaria ruficiliana. Fruits may also be infected by the smut fungus Urocystis primulicola.
The field experiment was conducted in the northernmost part of Stora Alvaret on southern Öland, Sweden. Within a 4 × 10-km area we have located 52 P. farinosa populations with both short-scaped and long-scaped plants. The area is characterized by shallow soils on limestone rock, and has a shrub layer of varying density that consists mainly of Potentilla fruticosa and Juniperus communis. Almost the entire area is grazed, although the intensity and timing of grazing as well as the grazing animals (cattle, sheep or horses) differ among populations. The present study was conducted in two consecutive years (2001 and 2002) in four large P. farinosa populations (Tranekärr, Lenstad, Dröstorp and Ekelunda), where both scape morphs were represented by several hundred plants.
To quantify the degree of pollen limitation in the two scape morphs, we conducted a hand-pollination experiment. In each year we marked 56–69 short-scaped and 60–88 long-scaped plants in each population with small sticks. We marked more long-scaped than short-scaped individuals because of the expected greater incidence of grazing damage in the former morph. Only long-styled plants were used, as short-styled plants are difficult to pollinate by hand without damaging the flowers. Fecundity did not differ between style morphs in a previous study (Ehrlén et al., 2002). Half the plants were randomly assigned to the open-pollinated control group, while the others received supplementary hand-pollination. The experimental plants were pollinated with compatible pollen (from short-styled plants) by brushing dehiscing anthers across the receptive stigmas. Pollen was collected within a radius of 50 m from the focal plant. Each flower received supplementary hand-pollination once during the period of stigma receptivity.
During flowering, we recorded the number of flowers produced by each plant included in the experiment. In late June and early July we counted all initiated fruits. Grazed plants were excluded from the analyses. We included only plants with single inflorescences in our study (in the study populations, 7% of the plants produce more than one inflorescence). All enlarged ovaries were counted as initiated. In a few unclear cases, fruits were opened and those containing enlarged ovules were counted as initiated. From all four populations in 2001 and from Lenstad in 2002, two fruits per plant were collected, and initiated seeds and noninitiated ovules were counted in the laboratory. The seeds from Lenstad in 2002 were also weighed. For each plant we then calculated mean seed mass by dividing total seed mass by the number of seeds in collected fruits.
We used a four-factor anova to analyse the effects of scape morph, pollination treatment, population and year on seed initiation (number of seeds initiated per ovule). Seed initiation was arcsine square-root transformed before analysis. To examine whether fruit initiation was negatively related to seed initiation, we included fruit initiation as a covariate in an ancova.
For fruit initiation (number of fruits initiated per flower) it was not possible to normalize residuals through transformation. Instead, we used a generalized linear model with a quasibinomial distribution and logit link to analyse the effects of scape morph, pollination treatment, population and year on fruit initiation. To test specific factors we used analysis of deviance (anodev; Crawley, 2002). Higher-order interactions were removed from the generalized linear models if they were not significant at α = 0.25. To study how fruit initiation varied naturally among scape morphs, populations and years, we performed separate analyses for open-pollinated controls.
We used a two-factor anova to analyse the effects of scape morph and pollination treatment on mean seed mass in Lenstad in 2002. To examine whether seed size was negatively related to seed initiation, we included seed initiation as a covariate in an ancova.
The number of flowers may influence pollinator visits per flower and the proportion of flowers initiating fruits (Schemske, 1980; Klinkhamer et al., 1989). However, ancova indicated that the number of flowers did not affect fruit and seed initiation significantly, or qualitatively alter the effects of other factors on these response variables. Thus we present here only models without number of flowers. Statistical tests were performed with the software packages statistica (StatSoft, 2005) and r (R Development Core Team, 2005).
Among open-pollinated controls, differences in fruit initiation between scape morphs varied significantly among populations (scape morph × population interaction, anodev: F3,414 = 3.11, P = 0.026). Fruit initiation was consistently high in the long-scaped morph (>82%), but more variable in the short-scaped morph (40–56% in Lenstad and Ekelunda populations compared with 67–87% in Tranekärr and Dröstorp populations; Fig. 2a). Fruit initiation was c. 15–20% higher in 2001 than in 2002 in the Tranekärr and Ekelunda populations, but did not differ between years in the Lenstad and Dröstorp populations (year × population interaction, anodev: F3,414 = 4.54, P = 0.0038).
The number of fruits initiated per flower was more strongly pollen-limited in short-scaped than in long-scaped plants, as indicated by the significant scape morph × pollination interaction (Table 1; Fig. 2a). However, fruit initiation was also lower in short-scaped plants after supplementary hand-pollination (anodev: F1,422 = 17.8, P < 0.0001; Fig. 2a). Pollen limitation of fruit initiation varied spatio-temporally, as indicated by a significant population × year × pollination interaction (Table 1). There was no evidence that the relative degree of pollen limitation in short- and long-scaped plants varied among years or populations (these interaction terms had P > 0.25 and were removed from the model).
Table 1. Effects of scape morph (short vs long), population, year and pollination (open-pollinated control vs supplementary hand-pollination) on fruit initiation (number of fruits initiated per flower) in Primula farinosa, analysed with a generalized linear model
Study conducted in four populations in 2001–02 on the island of Öland, south-east Sweden. Models were built with quasibinomial distribution and logit link.
Scape × pollination
Scape × population × year
Pop × year × pollination
The number of seeds initiated per ovule was c. 14% higher in long-scaped than in short-scaped plants, and supplementary hand-pollination increased seed initiation by c. 10% (Fig. 2b; Table 2). Pollen limitation of seed initiation did not vary significantly among scape morphs or populations (no significant interactions in anova; Table 2), nor did it vary among years in the 2-yr study in the Lenstad population (Table 3).
Table 2. Effects of scape morph (short vs long), population and pollination (open-pollinated control vs supplementary hand-pollination) on arcsine square-root seed initiation (number of seeds initiated per ovule) in Primula farinosa, analysed with three-way anova
Study conducted in four populations on the island of Öland, south-east Sweden in 2001.
Scape × population
Scape × pollination
Population × pollination
Scape × population × pollination
Table 3. Effects of scape morph (short vs long), year, and pollination (open-pollinated control vs supplementary hand-pollination) on arcsine square-root seed initiation (number of seeds initiated per ovule) of Primula farinosa, analysed with three-way anova
Study conducted in the Lenstad population in 2001–02.
Scape × year
Scape × pollination
Year × pollination
Scape × year × pollination
There was no evidence of a trade-off between the proportion of flowers initiating fruit development and the proportion of ovules initiating seed development in individual fruits. Fruit initiation was not related to seed initiation among hand-pollinated plants (F1,200 = 1.04, P = 0.31), and was positively rather than negatively related to seed initiation among control plants (parameter estimate extracted from ancova: F1,180 = 4.78, P = 0.030).
In Lenstad in 2002, seed mass was negatively related to seed initiation (parameter estimate extracted from ancova: F1,52 = 18.3, P < 0.0001), and plants that received supplementary hand-pollination produced c. 20% smaller seeds than open-pollinated controls (Table 4). Seed mass did not differ between long- and short-scaped plants.
Table 4. Effects of scape morph (short vs long) and pollination (open-pollinated control vs supplementary hand-pollination) on seed mass in Primula farinosa, analysed with two-way anova
Study conducted in the Lenstad population in 2002.
Scape × pollination
This study has documented considerable spatio-temporal variation in pollen limitation, and has shown that fruit initiation is overall more strongly pollen-limited in the short-scaped than in the long-scaped morph of P. farinosa. However, the relative pollen limitation of long- and short-scaped plants did not vary across years or populations. Seed initiation was lower in the short-scaped than in the long scaped morph, and seed initiation increased after supplementary hand-pollination. However, the effect of hand-pollination on seed initiation did not differ between scape morphs. Morph-specific differences in pollen limitation were thus expressed mainly in terms of differences in fruit initiation. Several other studies have also shown that pollen limitation is a more important factor for fruit initiation than for seed initiation (reviewed by Burd, 1994).
Despite considerable spatio-temporal variation in pollen limitation, we did not detect any significant variation in pollinator-mediated selection on scape morph. This is surprising, especially given that we did find among-population variations in the relative fruit initiation between morphs. Apparently the link between spatio-temporal variations in selection in a pollen-limited plant and spatio-temporal variations in pollinator-mediated selection is not straightforward.
The long-scaped morph was, overall, less pollen-limited than the short-scaped morph. The most plausible explanation for this is that the pollinators of P. farinosa use visual cues to find their flowers, and that the long-scaped inflorescences are detected more easily (Omura & Honda, 2005). In addition, insect pollinators, including bees, butterflies and syrphids, may adjust their behaviour to avoid predators when foraging (Dukas, 2001; Munoz & Arroyo, 2004). Pollinators may avoid the short-scaped morph if landing close to the ground increases the predation risk. Plant height also affects plant fitness through attractiveness to pollinators in other plant species (Peakall & Handel, 1993; Verbeek & Boasson, 1995; O’Connell & Johnston, 1998; Totland, 2001). However, plant height is usually strongly affected by environmental factors and plant age, and the evolutionary outcome of this phenotypic selection may be weak. In P. farinosa, phenotypic selection on scape morph is likely to correspond well with genetic selection, as our crossing experiments suggest a strong heritability.
Pollen limitation did not explain all the difference in fruit initiation between the two scape morphs. The long-scaped morph also outperformed the short-scaped morph after supplementary hand-pollination. This could reflect morph-specific differences in resources available for fruit maturation. The developing fruits on short-scaped plants are likely to be more shaded by surrounding vegetation and may therefore have a lower photosynthetic capacity than fruits on long-scaped plants. The leaf rosette of the short-scaped morph might also be shaded by its own infructescence. Alternatively, resource division between fruit development and other functions may differ inherently between the two scape morphs. These nonpollinator-related factors could be responsible for the spatial variation in relative fruit initiation between the morphs, a pattern that pollen limitation did not explain.
Seed mass was negatively related to the proportion of ovules initiating seed development. Baker et al. (1994) have shown that larger P. farinosa seeds germinate better and more quickly, and produce larger seedlings. Plants with a low seed set may thus produce seedlings with a competitive advantage, which could compensate for the low seed number. However, despite a 12% lower seed initiation, the short-scaped morph did not produce larger seeds than the long-scaped morph.
Pollen limitation is likely to vary in space and time in many plant species. As demonstrated by the present study, such variation is not necessarily associated with variation in selection on floral characters. Additional experimental studies exploring the relationship between pollen limitation and selection on floral traits are clearly needed. Nevertheless, this study has established that fruit set is, overall, more strongly pollen-limited in the short-scaped than in the long-scaped morph of P. farinosa. Interactions with pollinators are thus likely to be important for the relative fecundity of the two scape morphs and the maintenance of this striking polymorphism. A full understanding of the dynamics of this polymorphism will require the incorporation of spatio-temporal variation in interactions with other biotic agents that are likely to affect selection on scape length, such as seed predators and grazers. With the present study in mind, the link between spatio-temporal variation in, for example, grazing intensity and spatio-temporal variation in grazing-mediated selection for prostratedness would also require attention.
We are grateful to Sofia Käck and Saskia Sandring for assistance in the field, and to Petra Korall for seed counting. The study was conducted at the Ecological Research Station, Ölands Skogsby, and was supported financially by FORMAS (The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning, to J.Å. and J.E.).