Do dichogamy and herkogamy reduce sexual interference in a self-incompatible species?

Authors


Correspondence author. E-mail: candai@virginia.edu

Summary

1. Dichogamy and herkogamy respectively represent the temporal and spatial separation of male and female reproductive functions. They are regarded as mechanisms to avoid selfing, to promote outcrossing and, particularly in self-incompatible plants, to reduce sexual interference. However, little is known about the extent to which these mechanisms reduce sexual interference, and whether this reduction impacts fitness.

2. We studied patterns of dichogamy and herkogamy, their influence on sexual interference, and the fitness outcome in self-incompatible Passiflora incarnata. We manipulated flowers to be adichogamous or unisexual to evaluate fruit production under increased and decreased sexual interference.

3. Incomplete protandry guaranteed that almost half of the pollen could be successfully exported without interfering with the stigmas, indicating dichogamy may facilitate male pollination success. We found no difference in pollen deposition between natural and emasculated flowers, suggesting that herkogamy does not decrease self pollination. Increased herkogamy resulted in reduced pollen deposition and lower probability of setting fruits, however, a higher seed number. No difference in female fitness was detected under experimentally increased or decreased sexual interference.

4. Taken together, our results suggest that dichogamy is mostly driven by the advantage to male fitness and herkogamy is chiefly determined by female fitness. The lack of difference in female fitness under varied levels of sexual interference indicates that male function is more likely to play a role in shaping floral traits that reduce sexual interference.

Introduction

Hermaphroditism is the predominant breeding system in flowering plants, with over 90% possessing both male and female reproductive organs (Renner & Ricklefs 1995). Having two sexes in one flower increases the efficiency of pollen removal and deposition by pollinators (Barrett 2002b) and reproductive assurance (Fenster & Marten-Rodriguez 2007). However, it is generally suggested that hermaphroditic plants are under selective pressure to separate male and female functions to avoid inbreeding depression (Charlesworth & Charlesworth 1987) and/or reduce sexual interference (Lloyd & Yates 1982). Separation of sexual functions by temporal (dichogamy, Lloyd & Webb 1986; Bertin & Newman 1993), spatial (herkogamy, Webb & Lloyd 1986) and physiological (self-incompatibility, Charlesworth 2006) approaches is common.

Sexual interference is generally viewed as a cost of hermaphroditism (Barrett 2002a). It occurs when the male and female functions of a plant interfere with one another, either physically or biochemically, hampering successful outcrossing. Sexual interference potentially reduces both male and female fitness. On one hand, if self pollen is deposited on stigmas, it decreases the quantity of pollen that can be carried away by pollinators to other individuals (i.e. pollen discounting, Harder & Wilson 1998). On the other hand, the self pollen will occupy the surface area of the stigma that would otherwise be available for outcross pollen. In self-compatible species, self pollination might result in a higher selfing rate and lower quantity/quality of offspring (e.g. Herlihy & Eckert 2007). In self-incompatible species, the self pollen may limit the available stigmatic area, or it may affect the siring ability of outcross pollen more directly. For example, in the self-incompatible Asclepias exaltata, self pollen repressed the siring ability of outcross pollen and greatly reduced fruit production (Broyles & Wyatt 1993). Several other studies have produced similar results, especially in some late-acting self-incompatibility systems with self-rejection in the ovary (Waser & Price 1991; Seavey & Carter 1994). This decreased female fitness due to self pollination is often considered ovule discounting (Barrett 2002a). In addition to the within-flower level, sexual interference could also happen at the inflorescence level, through geitonogamous pollen deposition.

Sexual interference is considered an important selective pressure in shaping floral traits (Barrett 2002a). In self-incompatible species, dichogamy and/or herkogamy is thought to have evolved in response to selection to reduce sexual interference (e.g. Waites & Agren 2006), since physiological constraints already prevent the plants from selfing and inbreeding (Richards 1997). Recently, a theoretical study also suggested that sexual interference can favour the evolution of dichogamy, even in the absence of inbreeding depression (Sargent, Mandegar & Otto 2006). However, in spite of the high co-occurrence of self-incompatibility and dichogamy (Bertin 1993), few studies have experimentally tested whether male–female interference is reduced through temporal or spatial sexual separation (but see Cesaro et al. 2004; Routley & Husband 2006).

Phenotypic manipulation is an emerging approach for determining whether and how traits are adaptive. Traits are altered experimentally and the relative fitness of altered and unaltered individuals is then compared (Wade & Kalisz 1990; Ketterson et al. 1996). This technique has been applied to investigate how floral phenology influences offspring traits (Galloway & Burgess 2009), how long-distance dispersal affects competitive ability (Lenssen et al. 2005) and the adaptive response of shade-avoidance (Schmitt, Dudley & Pigliucci 1999). This manipulative approach may provide insight into how sexual interference affects fitness. By artificially making flowers adichogamous or unisexual, and comparing their female fitness to control flowers, we can determine whether dichogamy and herkogamy reduce sexual interference via female function. In addition, all previous studies investigating the consequence of self pollination for self-incompatible plants have either hand-pollinated flowers first with self and then with outcross pollen or have simultaneously applied a mixture of both pollen types to stigmas (reviewed in Barrett 2002a). Compared with these approaches, experimentally altering floral traits has the advantage of evaluating the combined effect of floral traits and pollinator behaviour, both of which may contribute to the intensity of sexual interference.

In this study, we address whether dichogamy and herkogamy in passion flower (Passiflora incarnata) reduce within-flower sexual interference and evaluate their consequences for male and female pollination and reproductive success. In P. incarnata, style deflexion is a process in which the styles gradually bend down from near vertical to a position close to the anthers after flower opening. We evaluated patterns of style deflexion with the delay of deflexion representing dichogamy and the degree of deflexion representing herkogamy. We then studied the relationship of dichogamy and herkogamy with sexual interference, asking: (i) Does the delay in style deflexion reduce sexual interference by allowing for pollen export before the presentation of stigmas near the anthers? (ii) Does the degree of style deflexion result in different levels of self pollination, and hence sexual interference? (iii) How does the degree of style deflexion affect female fitness? (iv) Does fruit production depend on the intensity of sexual interference?

Materials and methods

Study species

Passiflora incarnata (Passifloraceae) is a perennial, self-incompatible vine that grows throughout the southeastern United States (May & Spears 1988). It blooms from July to September in central Virginia, and its primary pollinator is the common carpenter bee (Xylocopa virginica). Flowers open fairly synchronously between 11:00 h and noon, and remain functional for only one afternoon. There is usually only one flower open per plant per day. Solitary flowers exclude the possibility of geitonogamy; therefore, only within-flower sexual interference is examined in this study. The flower is composed of a petal platform (corona) on which pollinators land, five downward-facing anthers that form a plane parallel to the petal platform, and three styles from a superior ovary above the anthers. When a carpenter bee lands on the horizontal platform of petals, it pushes its head and thorax in-between the anther plane and the petal platform (Fig. 1). It then moves in a circle, foraging for nectar in the centre of the flower and passively accumulating pollen on the dorsal part of the thorax. The fruit of P. incarnata is a large, edible, unilocular berry with seeds that are enclosed by fleshy arils.

Figure 1.

 Style deflexion (mean ± SD) at times after the opening of Passiflora incarnata flowers. Schematic picture of a carpenter bee visiting a P. incarnata flower is on the top. The deflexion of the two styles shown is 0 mm.

The movement of styles has been observed in many species in the Passifloraceae of Central America (Janzen 1968) and in P. incarnata (May & Spears 1988). From our field observations, the styles with stigmas at the tip are initially positioned upwards when the petals first open. In this position, it is almost impossible for the stigmas to receive pollen via pollinator visitation. Over the next hour or so, the styles then deflex to a lower position, resulting in a higher possibility of contact with the pollen-covered dorsal thorax of the carpenter bees. Pollinators begin to visit the flowers as soon as the petals are open; therefore, flowers can receive several visits during the period of deflexion. Both the temporal pattern of style movement and the final degree of style deflexion vary among flowers. We measure style deflexion as the vertical distance between stigmas and anther plane. If the stigma is above the plane, the value of style deflexion is positive; if below, it is negative (Fig. 2).

Figure 2.

 The distribution of final style deflexion in Passiflora incarnata flowers. Schematic pictures of three deflexion groups (deflexion < 0; deflexion = 0; and deflexion > 0) are shown to illustrate the relative positions of stigmas and anthers.

Passiflora incarnata is functionally andromonoecious with both male and hermaphroditic flowers on the same plant (May & Spears 1988). However, only hermaphroditic flowers were included in experiments discussed here.

Experimental design and statistical analysis

All experiments were conducted in a natural population of P. incarnata at the Shadwell Preserve of the Jefferson Monticello Foundation in Shadwell Virginia, USA in the summers of 2007, 2008 and 2009. Passiflora incarnata plants have rhizomatous growth and branches emerging from the ground in a clump are generally from one genotype. Experiments and observations were performed at the level of individual flowers and we included only one flower per apparent genotype in each treatment daily. The natural population is large and apparent genotypes were not frequently resampled. All analyses were done using sas (SAS Institute, 2005). In all analyses, ‘day’ was initially included as a blocking factor. However, it was never significant and thus was dropped from the analyses and results presented here. Variables did not require transformation to meet analysis assumptions.

Patterns of dichogamy and herkogamy

To explore the patterns of dichogamy and herkogamy in P. incarnata, the deflexion of 42 styles on fourteen haphazardly chosen flowers was measured at repeated intervals in a 3-day period in July 2007. The measurements were taken every 15 min to 1 h from anthesis until about 5 h after flower opening. In 2008, the final degree of style deflexion was measured after 13:00 h on 1001 styles from 330 randomly chosen P. incarnata flowers (some flowers had four styles instead of three).

Dichogamy and sexual interference

We examined pollen removal before and after style deflexion under open pollination to infer whether dichogamy reduces sexual interference. We collected all five anther sacs from random flowers at four times: right before flower opening – 0 min (n = 27); 30 min after flower opening (n = 60); 90 min (n = 60); and 420 min (n = 27). After 420 min, which is approximately 19:00 h, there was rarely any pollinator visitation and flowers started to wither. Hence, the pollen remaining at the last collection was unlikely to function in reproduction. The anther sacs were collected on 4 days during the flowering peak in July 2009 and every one of the above time intervals was included each day. All anther sacs were dried overnight at 50 °C and stored at room temperature before counting. All pollen from each flower was then rehydrated in 30 mL of 70% ethanol and loosened by sonication. The samples were counted using HIAC Model 9703 liquid particle counting system (Pacific Scientific, Rockford, IL, USA). To estimate the pollen remaining in each flower, the pollen count of nine subsamples of 1 mL each were averaged and multiplied by 30 (the total volume). The pollen remaining in the anther sacs at the four times was compared using analysis of variance (anova, proc glm) and pair-wise comparisons were made using Tukey.

Herkogamy and sexual interference

We investigated whether herkogamy reduces sexual interference via reduced self pollen deposition. We compared pollen deposition in emasculated flowers (no interference) with natural flowers to estimate the level of self pollination in relation to different degrees of style deflexion. Emasculation was done by clipping off all anthers prior to the opening of flowers. Natural flowers were left untouched.

This experimental design assumes that pollinator behaviour is not influenced by emasculation. To check this assumption, we observed pollinator visitation on natural and emasculated flowers. The observations were made daily between 11:30 and 11:45 h on three or four treatment pairs (six or eight flowers) for 10 days in late July 2008. The number of visits to every flower by carpenter bees was recorded. A paired t-test (proc ttest) was used to compare pollinator visitation rate between emasculated (n = 35) and natural flowers (n = 35).

To estimate the level of self pollination for different degrees of style deflexion, we compared pollen deposition on natural and emasculated flowers, under the assumption that the difference between the two was due to self-pollination. A total of 400 flowers (200 of each type) spread over a 13-day period from late July through early August 2008 were used in the study. We included the same number of natural and emasculated flowers each day. Final style deflexion was measured by a digital caliper after 13:00 h. We compared style deflexion between natural and emasculated flowers using Student’s t-test. Pilot experiments indicated that cutting the stigmas after 19:00 h had no effect on fruit development. Accordingly, stigmas of each flower were collected after 19:00 h and dried overnight in an incubator at 50 °C. Pollen deposition was counted on one stigma for each flower for a subset of emasculated and natural flowers, chosen to include an equal representation of style deflexion across its distribution. In total, there were 118 stigmas from emasculated flowers and 119 stigmas from natural flowers in each subset.

To count the pollen grains, one stigma from each flower in each subset was immersed in a drop of 8 mol L−1 NaOH on microscope slides overnight and then mounted with Permount (Fisher Scientific, Fair Lawn, NJ, USA). Extra NaOH was removed with filter paper before mounting. Under an Olympus (IX-70) inverted epi-fluorescence microscope, pollen grains glowed bright orange contrasting with pale stigmatic tissues (yellow filter cube with Ex510/20 nm, 525DCLP and Em590/35 nm). Images of every stigma were taken under 20× magnification using Orca-2 Hamamatsu CCD camera and stitched by Photoshop (Adobe Systems, 2002). The number of pollen grains per stigma was counted using imagej (Rasband 2009). The difference in pollen deposition between emasculated and natural flowers was tested with ancova, including style deflexion as the covariate, emasculation treatment as a fixed effect and their interaction (proc glm).

We then evaluated whether style deflexion affected female fitness. We followed all natural and emasculated flowers after pollination and recorded if they set fruit. All fruits (n = 237) were protected from frugivory by aluminium-screen bags and harvested after maturation. Seeds per fruit were counted and air dried. To address how the degree of style deflexion affected pollen deposition, we did regression analysis (proc reg). We then used logistic regression to ascertain whether style deflexion and pollen deposition (respectively and combined) affected the probability of setting fruits (proc logist). For flowers that set fruit, we tested whether style deflexion influenced seed number using regression. A multiple regression was also employed to explore the relative importance of style deflexion and pollen deposition on the seed production.

Increased and decreased sexual interference

We manipulated flowers to increase and decrease sexual interference and examined the effect on female fitness. To increase sexual interference, we made flowers effectively adichogamous by bagging to exclude pollinator visitation until 1–1·5 h after flower opening, at which point flowers had fully deflexed their styles. Untreated flowers acted as controls. We conducted this experiment on 4 days in July 2009, with 24–40 flowers per day equally split between the treatments, for a total of 70 flowers per treatment. We followed the flowers through fruit harvest. To decrease sexual interference, we followed the flowers emasculated for the Herkogamy study (above) through fruit production. Emasculated flowers were functionally only female and therefore not expected to show sexual interference. An equivalent number of natural flowers were used as controls (see Herkogamy and sexual interference). All fruits from the 400 flowers (200 in each treatment group) were harvested upon maturation. The weight of five random seeds per fruit was measured from both studies. A chi-square test was used to compare the fruit-set between the control and treatment group (proc freq, sas). The difference in seed number per fruit and seed weight was evaluated for each study by a Student’s t-test.

Results

Patterns of dichogamy and herkogamy

Incomplete dichogamy and variation in herkogamy are found in P. incarnata. It took from 30 to 90 min for flowers to deflex their styles to a lower position after opening (Fig. 1). The timing of deflexion differed among flowers. During the process of style deflexion, frequent pollinator visits were observed, resulting in pollen removal prior to female function. Because pollen is removed throughout the day and there is some overlap in the presentation of pollen and stigmas, we consider that the delay of style deflexion creates incomplete protandry (Lloyd & Webb 1986). There was also variation in the degree of final style deflexion (Fig. 2). The final style deflexion constitutes a continuous spectrum ranging from −4·93 to 5·00 mm, and implies different degrees of herkogamy. Some stigmas are below the anther plane; some around the same level with anthers; and others above (Fig. 2). The range of mean deflexion per flower was slightly less, −3·42 to 4·92 mm.

Dichogamy and sexual interference

The delay of style deflexion in P. incarnata may decrease sexual interference. There was less pollen remaining at each successive interval (Fig. 3, F3,170 = 170·12, < 0·0001, all pair-wise comparisons were significant at α = 0·05). Approximately one-fifth of all pollen that could be removed was gone by 30 min after flower opening and almost a half of pollen had been taken at 90 min, when most styles just finished their deflexion process (Fig. 1).

Figure 3.

 Pollen remaining (mean ± SE) in Passiflora incarnata flowers at different times after flower opening under open pollination. After 420 min (approximately 19:00 h), there is rarely any pollinator visitation and flowers start to wither.

Herkogamy and sexual interference

We found no evidence that herkogamy decreases sexual interference. After controlling for style deflexion, there was no sign of self pollination when comparing pollen deposition between natural and emasculated flowers (treatment: F1,233 = 0·19, = 0·67). Further, we found no evidence that the emasculation treatment affected pollinator visitation rate or style deflexion in P. incarnata. The pollinator visitation rate per flower per 15 min was very similar between emasculated (mean ± SE: 8·7 ± 0·8) and natural (8·6 ± 0·8) flowers (paired = −0·17, d.f. = 34, = 0·87). Also, there was no significant difference in the final style deflexion between natural and emasculated flowers (= 0·39, d.f. = 235, = 0·70). Therefore lack of difference in pollen deposition between the natural and emasculated flowers is unlikely due to pollinator service or floral morphology, but rather the lack of male–female interference.

However, style deflexion affected pollen deposition, the probability of setting fruit, and the seed number per fruit. The degree of style deflexion had a strong effect on pollen deposition (F1,233 = 152·31, < 0·0001) and was not affected by emasculation (treatment × style deflexion: F1,233 = 0·57, = 0·45). Pollen deposition decreased with increasing style deflexion (Fig. 4a, Pollen deposition = −155·8 × Style deflexion + 1114·5, R2 = 0·39, < 0·0001). Nearly 40% of the variation in pollen deposition was explained by the degree of style deflexion. The comparison between flowers that set fruit and those that did not indicated that both style deflexion and pollen deposition played an important role (Fig. 5). A smaller style deflexion and a higher pollen deposition respectively resulted in a greater probability of setting fruits (Style deflexion: χ2 = 10·32, n = 237, = 0·001; Pollen deposition: χ2 = 20·95, n = 237, < 0·0001). The multiple logistic regression revealed that style deflexion acted via pollen deposition to affect fruiting probability (Style deflexion: χ2 = 0·29, n = 237, = 0·59 and Pollen deposition: χ2 = 11·38, n = 237, = 0·0007). A smaller style deflexion also lead to a reduced seed production (Fig. 4b, Seed number = 3·13 × Style deflexion + 71·9, R2 = 0·09, = 0·0004) whereas there was not a significant relationship between pollen deposition and seed production (= 0·53, = 0·60). When combined in a multiple regression, both style deflexion and pollen deposition contributed to determine seed production (Seed number = 4·49 × Style deflexion + 0·01 × Pollen deposition + 57·0, R2 = 0·14, < 0·0001), which indicates that style deflexion may influence the seed production through other routes besides affecting pollen deposition. Natural and emasculated flowers were combined in these tests because emasculation did not influence pollen deposition.

Figure 4.

 Pollen deposition on stigmas (a) and seed number per fruit (b) of Passiflora incarnata flowers that differ in the degree of style deflexion. A single stigma was used to estimate flower-level pollen deposition (also Fig. 5).

Figure 5.

 Style deflexion and pollen deposition (mean ± SE) of Passiflora incarnata flowers that set fruits and that did not.

Increased and decreased sexual interference

The probability of setting fruit, seed number per fruit and mean seed weight did not change when sexual interference was artificially increased by presenting pollen and stigmas at the same time or when sexual interference was artificially decreased by emasculation (Table 1).

Table 1.   Fruit set, and seed number and seed weight (mean ± SE) in Passiflora incarnata treated to: (a) increase male–female interference and (b) decrease male–female interference. In (a), the treatment was bagging flowers for 1–1·5 h. In (b), the treatment was emasculation before flowers opened. χ2 and Student’s t-tests were employed to compare the control and treatment groups. None of the comparisons were significant at < 0·05 level
Variable(a) Increased male–female interference
ControlBaggedχ2 or t
Fruit-set (Fruit number)76% (53)74% (52)0·04
Seed number78·7 ± 3·179·7 ± 3·70·21
Seed weight (mg)168·9 ± 2·1167·8 ± 2·3−0·35
 (b) Decreased male–female interference
ControlEmasculated 
Fruit-set (Fruit number)42% (84)45% (90)0·16
Seed number73·5 ± 2·276·1 ± 2·5−0·76
Seed weight (mg)151·0 ± 2·0151·2 ± 2·0−0·05

Discussion

Dichogamy: delay of style deflexion

In P. incarnata, most style deflexion finished by 1·5 h after flower opening, when almost half of the pollen had been removed, indicating a partial separation of male and female function. This separation may be advantageous for pollen dispersal because before styles deflex to a lower position where they can make contact with pollinators, there is no sexual interference to impede successful pollen export. When we experimentally presented pollen and stigmas at the same time, we found no impact on female fitness. Thus, we infer that the dichogamy in P. incarnata is likely to be under selection to reduce sexual interference and promote outcrossing via male function.

The idea that protandry may increase male fitness is not new (Lloyd & Webb 1986), and has been supported experimentally (e.g. Jersakova & Johnson 2007). It has been suggested that protandry provides a competitive advantage for early pollen and prolongs the duration of male function (Bertin 1993). However, whether protandry benefits female function of a flower has rarely been tested. Our study is among the few that manipulated flowers to be adichogamous and found protandry did not increase female fitness. Similar to the present study, in self-compatible Chamerion angustifolium, protandrous plants had little change in seed set relative to adichogamous plants, but higher outcross siring success (Routley & Husband 2003). Likewise to protandry, protogyny was found to have little effect on female fitness. In Aquilegia, flowers rendered protogynous did not have increased outcrossing when compared to adichogamous flowers (Griffin, Mavraganis & Eckert 2000). Overall, these studies suggest that dichogamy might be selected through male fitness, rather than female fitness, as a mechanism to reduce sexual interference.

Variation in herkogamy: degree of style deflexion

We observed a wide range of herkogamy among P. incarnata flowers as measured by style deflexion. The degree of style deflexion not only affected the proximity between stigmas and anther, thus the sexual interference, but also the contact between pollinator and stigma, thus the pollination success. If herkogamy reduces sexual interference, we predicted that emasculated flowers would show decreased interference due to the lack of self pollination. Counter to our prediction, emasculated flowers received the same amount of pollen as natural flowers. Regardless of the emasculation treatment, successful pollen deposition and fruit production was associated with smaller style deflexion. However, smaller style deflexion caused a reduced seed set per fruit.

Our results do not support the conventional idea that the smaller the separation between stigmas and anthers, the greater the level of self-pollination (Webb & Lloyd 1986), and therefore pollen discounting (Harder & Wilson 1998) and ovule discounting (Verdu et al. 2006). However, it may be a bit hasty to infer that different degrees of style deflexion have no effect on self-pollination. All studies, like the one presented here, that have evaluated the natural level of self-pollination have compared hermaphroditic and emasculated flowers and employed the difference in pollen deposition as evidence of self pollination [Routley & Husband (2006) used a slightly different approach]. Some studies found self pollination because intact flowers had higher pollen deposition (Donnelly, Lortie & Aarssen 1998; Rodet et al. 1998; Routley & Husband 2006; Castro, Silveira & Navarro 2008), whereas others failed to detect it (Nishihiro & Washitani 1998; Eckert 2000; Cesaro et al. 2004). In P. incarnata, the circular movement of pollinators, which are constantly in touch with pollen-bearing anthers or receptive stigmas when stigmas bend low enough, suggests that self-pollination should occur, especially when style deflexion is near or below zero.

The absence of apparent self-pollination may largely be due to pollinator behaviour. On average, a passion flower can receive up to 40 pollinator visits in a day and the stigmas are often covered with pollen (C. Dai, unpublished data). In this case, the deposition of self pollen on natural flowers could be easily overshadowed by the excessive outcross pollen from pollinators to emasculated flowers. Therefore, the amount of pollen deposition may not be affected by emasculation, but instead the stigmatic surface area and/or the style deflexion. Indeed, from Fig. 4 it is clear that pollen deposition increases as the style deflexion decreases. Self pollination may be more prominent in a pollinator-limited environment. In a future study, fluorescent dye could be used as a pollen substitute to confirm the potential self pollination (Dejong et al. 1992).

In P. incarnata, seed set increased with increasing herkogamy. In self-incompatible Polygala vayredea, the same pattern of increased female fitness (using number of pollen tubes as a proxy) with increased herkogamy was found (Castro, Silveira & Navarro 2008). In self-compatible Aquilegia, selfing rate was negatively associated with herkogamy, suggesting increased herkogamy may be selected to reduce the production of low-quality seed (Herlihy & Eckert 2007). Nevertheless, the outcrossing rate in Narcissus failed to reveal a monotonic increase with increasing herkogamy (Medrano, Herrera & Barrett 2005). It is worth mentioning that in our study, the pattern of more seeds with higher style deflexion is the same in both natural and emasculated flowers. Thus, it is unlikely that the effect on seed set is from cryptic self-pollination. In addition, in contrast to the pattern found for seed set, increased style deflexion reduced the probability of setting fruit. As a consequence of the opposite relationship between seed and fruit production and style deflexion, total predicted seed production per plant is relatively similar across a range of style deflexion (flowers on the same plant have similar degrees of style deflexion, C. Dai unpublished data). The association of more fruit per plant with fewer seeds each or fewer fruit with more seeds each may reflect plant-level resource limitation of total seed production. Therefore, in P. incarnata, the opposite fitness effects of style deflexion at the fruit and seed level might play a role in maintaining the variation observed in style deflexion. In addition, plant level control over seed production may also explain the significant effect of style deflexion on seed number beyond its contribution to pollen deposition.

Increased and decreased sexual interference

Increasing sexual interference had no impact on fruit and seed production. In P. incarnata, increased sexual interference was achieved by delaying floral presentation to pollinators until styles finished deflexion. Nearly a dozen studies on self-incompatible species, all of which employed hand pollination, found that female fitness was reduced by adding self pollen (e.g. Bertin & Sullivan 1988; Dinnetz 1997; Kawagoe & Suzuki 2005). It is believed that the decrease in female fitness is due to physical pollen clogging and/or chemically late-acting self-incompatibility (Barrett 2002a). However, several investigations found little influence of self pollination on female fertility (Shore & Barrett 1984; Barrett & Glover 1985; Scribailo & Barrett 1994; Cesaro et al. 2004). Interestingly, these exceptions have all been either distylous or tristylous. Only recently, a study on Raphanus raphanistrum, a homomorphic species with sporophytic self-incompatibility, also found no effect of self pollination on seed set (Koelling & Karoly 2007). In P. incarnata, self-incompatibility is likely sporophytic (Do Rego et al. 1999; Muschner et al. 2003). When the rejection of pollen occurs at the stigmatic surface, there is little chance for ovule discounting, and correspondingly less chance of decreased seed-set due to self pollination. Another possible explanation is that while we artificially postponed the flowering time by 1·5 h, other natural flowers opened normally and were visited by pollinators. At the time we presented the adichogamous flowers, there was already a great amount and diversity of pollen carried by pollinators. The negative effect of increased sexual interference might have been cancelled out by the positive effect of more successful pollination.

We applied emasculation to get rid of within-flower sexual interference and found no change in female fitness. Emasculation has frequently been employed to examine whether autonomous selfing occurs in self-compatible species (e.g. Davis & Delph 2005; Ruan et al. 2008; Kosinski 2010). In these studies, emasculation typically results in a decrease in female fitness. However, in P. incarnata, we anticipated that emasculation would result in a higher fruit and seed set due to a reduction in self-incompatible pollen and therefore a lack of sexual interference. Indeed, this pattern is seen in self-incompatible Bulbine (Vaughton & Ramsey 2010) and a different species of Passiflora (Snow 1982). However, our finding of no association between sexual interference and female fitness supports results in Narcissus (Cesaro et al. 2004). Given the relatively stable fruit and seed output under increased and decreased sexual interference scenarios, it is possible that the female fitness in P. incarnata is limited by resources (C. Dai, unpublished data) rather than mate opportunity. Reduced importance of sexual interference in the evolution of traits that influence female fitness is in agreement with Bateman’s principle (Bateman 1948).

Acknowledgements

We thank Q.F. Wang and K. Hastings for field assistance, D. Roach and T. Roulston for sharing their equipment, W.M. Keck Center (University of Virginia) for providing imaging facilities, and an International Doctoral Fellowship to CD from American Association of University Women (AAUW) Educational Foundation for financial support. We appreciate the comments of three anonymous reviewers. Special thanks to CD’s dissertation committee for their help throughout the work, especially Janis Antonovics for his suggestion on statistical analysis. This work was also supported by NSF DEB-0316298 to LFG.

Ancillary