• Floral display;
  • herbivory;
  • herkogamy;
  • outcrossing rate;
  • self pollination;
  • water availability


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    If pollination is unpredictable, selection may favour the production of selfed seeds in the absence of pollen vectors, even in plant species with obvious adaptations for outcrossing. Pollination may be less predictable for plants growing in certain environments if environmental factors affect the floral phenotype. Through effects on flower morphology and the floral display, the environment may affect the outcrossing rate.
  • 2
    We manipulated two environmental factors, water availability and exposure to insect herbivores, in a common-garden experiment using a perennial herb, Datura wrightii. We measured herkogamy (the separation of anthers and stigmas within flowers), total flower length, and flower number, and used a single-gene trichome dimorphism as a marker to determine per-plant outcrossing rates.
  • 3
    The large amount of variation in herkogamy was affected by trichome type, irrigation and herbivory. In addition, watered plants had longer corollas, and plants attacked by herbivores had fewer open flowers. Thus environmental factors affect floral phenotype.
  • 4
    However, irrigation and herbivory did not directly affect outcrossing rate. There were indirect effects of these treatments on outcrossing because plants with increased herkogamy and fewer open flowers had higher outcrossing rates.
  • 5
      A greenhouse experiment showed that autonomous selfing is more likely when herkogamy is reduced, and can occur both as the flower opens and when the corolla is shed.
  • 6
      These experiments are among the first to show that within-population variation in the mating system can be due to environmentally induced variation in floral traits.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Whether or not self-pollination is beneficial should depend on the type and timing of selfing, and the environment in which selfing occurs. The type of selfing matters because geitonogamy, the vector-mediated movement of pollen within individuals, does not provide reproductive assurance (Lloyd 1992). Timing matters because selfing that occurs prior to opportunities for outcrossing may reduce the potential for outcrossed success (pollen and/or seed discounting; Holsinger, Feldman, & Christiansen 1984; Lloyd 1992). In contrast, delayed selfing (i.e. selfing after all opportunity for outcrossing has passed) should be favoured, because it will not increase pollen or seed discounting, and can provide reproductive assurance (Lloyd 1992; Lloyd & Schoen 1992). The environment matters because it can affect pollinator availability, the expression of inbreeding depression, and the floral phenotype, which may affect pollinator visitation and the resulting mating system.

Floral herbivory may directly reduce the attractiveness of flowers to pollinators (Adler 2000; Karban & Strauss 1993; Krupnick, Weiss & Campbell 1999). Foliar herbivory may also be important if it leads to decreased flower size, decreased pollen and nectar production, or changes in plant architecture and/or phenology which reduce plant attractiveness to pollinators (Agrawal, Strauss & Stout 1999; Lehtila & Strauss 1999; Mothershead & Marquis 2000; Quesada, Bollman & Stephenson 1995; Strauss, Conner & Rush 1997; Strauss et al. 1999). If herbivory decreases attractiveness, then it should decrease the outcrossing rate as well. While fitness of plants attacked by herbivores has repeatedly been shown to be less than that of plants protected from herbivores (reviewed by Marquis 1992), no study has compared outcrossing rates for plants exposed to and protected from foliar herbivory (but see Krupnick et al. 1999 for a discussion of floral herbivory and outcrossing rate).

Outcrossing may also be reduced in stressful environments. Selfing populations or subspecies are often found on the margins of species’ ranges (Schoen 1982; Stebbins 1957; Vasek 1964). Outcrossing rates and/or heterozygosity tend to be lower in populations from drier habitats in Clarkia (Holtsford & Ellstrand 1992), Lycopersicon (Rick, Fobes, & Holle 1977), Hordeum (Brown, Zohary & Nevo 1978), Gilia (Schoen 1982), and other plant species (reviewed by Hamrick, Linhart & Mitton 1979; Clegg 1980). Reductions in outcrossing rates were linked to reduced herkogamy (distance between male and female parts) in dry habitats in the Clarkia and Lycopersicon studies. Because proximity of the sexual parts increases the potential for autonomous selfing, the link between drought, morphology and outcrossing is an important one to pursue. Yet this link has primarily been explored in the context of genetic differentiation among subspecies or populations, not in the context of phenotypic plasticity within a single experimental population of plants.

The expectation of a link between environment, phenotype and mating system assumes that the alternative to vector-mediated outcrossing is autonomous selfing. If, however, flowers that are not outcross-pollinated are incapable of autonomous self-pollination, the alternative to outcrossing would be reduced reproductive success. Selfing is common in small-flowered species in which the sexual parts are likely to come in contact (Eckhart & Geber 1999; Johnston & Schoen 1996; Ritland & Ritland 1989). Some plant species with initial separation between the sexual parts have mechanisms allowing delayed selfing. For example, the anther filaments or the pistil may elongate during floral life, eventually leading to contact between them (Eckert & Schaefer 1998; Kalisz et al. 1999). Alternatively, anthers may brush against stigmas when corollas with attached anthers are shed as a unit (Dole 1990; Dole 1992). In any exploration of variation in the relative amount of selfing and outcrossing in a large-flowered, putatively outcrossing plant, it is essential that the potential for autonomous selfing be documented.

As part of a larger study of the importance of the fitness costs and benefits of insect resistance conferred by a trichome dimorphism in Datura wrightii, we undertook this examination of the mating system. We hypothesized that plants with greater herkogamy would have higher outcrossing rates, as would plants with more attractive floral displays (many, large flowers). We also hypothesized that attack by herbivores or drought conditions would reduce herkogamy, flower size and display size, and so reduce the outcrossing rate. These hypotheses assume that proximity of the sexual parts leads to higher rates of autonomous selfing, an assumption that we tested in the greenhouse. Our specific objectives were: (i) to document variation in floral phenotype and its relationship to experimentally imposed environmental variation in a common garden experiment; (ii) to estimate the minimum outcrossing exhibited by common garden plants and determine if outcrossing rate is related to floral phenotype or the experimental environment; and (iii) to establish whether selfing can occur without pollinator intervention, the timing of such autonomous selfing, and whether it is related to herkogamy.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Datura wrightii, Western Jimsonweed, is a perennial plant native to south-western North America. Flowers are white and trumpet-shaped, and average 17 cm long and 10 cm wide at the mouth. They produce copious nectar and scent, open at dusk, and wilt by the following noon, exhibiting the classic Hawkmoth-pollination syndrome (Baker 1961; Grant & Grant 1983). Under ideal conditions, individual plants can produce over 100 flowers in a single night over a flowering season lasting several months (Elle, van Dam & Hare 1999). The fruit is a spine-covered capsule containing several hundred seeds (Hickman 1993).

Two trichome phenotypes occur in this species. Plants either produce primarily (>95%) glandular trichomes (‘sticky’), or a similar proportion of non-glandular trichomes (‘velvety’), on leaves and stems. This difference is controlled by a single gene, with the sticky condition dominant (van Dam, Hare & Elle 1999). Glandular trichomes are hypothesized to increase plant resistance to herbivores; the costs and benefits associated with glandular trichome production by sticky plants have been examined elsewhere (Elle & Hare 2000; Elle et al. 1999).


The field experiment design is described in detail by Elle et al. (1999). Briefly, families expressing one of the two trichome types, sticky and velvety, were produced through a combination of self-pollinations and outcross-pollinations to randomize the genetic background within which trichome type was expressed, and to ensure that plants were homozygous at the trichome locus (four sticky and five velvety families). These plants were grown in a split-plot experiment initiated in 1997, in which the presence/absence of herbivory and the presence/absence of additional water were manipulated. Herbivory treatment was the main effect, and water treatments were replicated within herbivory treatment.

Herbivore-free conditions were maintained with weekly applications of the insecticide acephate; herbivore-attacked plants were exposed to the extant herbivore community. The primary mode of action of acephate is as an insect gut toxin, meaning it must be ingested for full efficacy. In combination with the timing of spraying (early morning, after the flowers had been open all night), we expected minimal effects of the insecticide on potential pollinators. Watered plants were furrow-irrigated for 8 h twice weekly during the growing season, while unwatered plants received only natural rainfall.

Sticky and velvety individuals were planted alternately within furrows in the experimental field. At the initiation of the experiment, 15 individuals from each family were present in each herbivory/irrigation treatment combination.


We measured floral traits on a subset of plants within each treatment (three per family in each irrigation/herbivory combination, 108 plants total) on two dates in 1998. Plants were randomly chosen prior to flowering from all surviving plants in the larger experiment; four of these plants (all herbivore-attacked) subsequently did not flower, and several plants in the herbivory treatment did not produce any seeds (see below). On each plant and date, herkogamy and total corolla length were measured on five flowers, and the total number of open flowers was recorded. Herkogamy was expected to be strongly correlated with outcrossing rate in D. wrightii, as in the related but smaller-flowered annual D. stramonium (Motten & Antonovics 1992; Motten & Stone 2000). Flower size and flower number are aspects of the floral display, which should affect pollinator visitation. In addition, if variation in herkogamy was primarily due to differences in flower size among treatments (i.e. increased floral expansion when watered), differences in corolla length, rather than herkogamy, would explain variation in the outcrossing rate. Occasionally fewer than five flowers were open on a given measurement date, especially on herbivore-attacked plants. At least four flowers were measured over the two dates and were used to calculate individual means for 97 out of 104 plants that flowered.

Herkogamy was measured as the smallest distance between the stigmatic surface and the closest end of a randomly chosen anther. Anther position (at the corolla mouth) appears constant in this species (E.E., personal observation). When the stigma was below the anther, measurement was from the lower end of the anther and resulted in a negative value. If the stigma and anther overlapped, the distance was scored as zero. If a capsule was formed on a measured flower, seed number was estimated by weighing 10 seeds to determine an average individual seed weight, and dividing the total weight of seeds produced in a capsule by this number (Elle et al. 1999).

All analyses were done with the SAS statistical package (SAS Institute 1988). We used analysis of variance (anova) to determine whether environmental (herbivory and irrigation treatments) or genetic (type and family effects) factors affected herkogamy, corolla length, or number of open flowers. Plant means were calculated for each variable prior to the analysis, and absolute values were used in the calculation of means for herkogamy so as not to underestimate the mean separation of the sexual parts when negative herkogamy was observed. Family was considered a random effect, and differences between the two trichome phenotypes were tested against variation among families nested within trichome type. Interactions between family within trichome type and other effects in the model were never significant, and so for simplicity the reduced model is presented here.

To explore the relationship between herkogamy and fruit and seed set, we performed additional analyses on individual flower/fruit combinations. Herkogamy varied greatly across plants and treatments (CV = 50·6%), compared, for example, to variation in corolla length (CV = 8·3%, n = 815 flowers). Thus using per-plant means to investigate the effect of variation in herkogamy on fruit and seed set could potentially obscure any relationship that might exist. Differences in herkogamy for flowers that did and did not form fruits were determined by t-tests. When fruit were produced, the extent to which seed number was affected by herkogamy, as well as environmental treatment effects, was determined by analysis of covariance (ancova), with herkogamy as the covariate.


We used the trichome locus as a genetic marker to estimate outcrossing rates of a subset of velvety plants, and to determine the relationship between variation in floral traits and outcrossing rate. Velvety parents were homozygous recessive at the trichome locus, so any sticky offspring produced by velvety parents were due to an outcrossing event. Because sticky parents were homozygous dominant at the trichome locus, we were unable to determine outcrossing rates for these plants. We collected a random sample of 50 ripe capsules for all plants in three velvety families for which floral phenotype was measured at the end of the 1998 growing season. The three families were chosen to represent the range of variation in herkogamy (see below) and viable seed production (Elle et al. 1999) in our experimental field. Of the 36 plants included in this survey (12 per family, three per herbivory/irrigation treatment combination), four plants, all herbivore-attacked, did not produce fruit and so could not be used in the analysis. Approximately 15–20 seeds per capsule were used to make a bulk seed batch for the individual; on average, a capsule contains 225 seeds (Elle et al. 1999). In the spring of 1999, 100 seeds from each of these velvety individuals were planted.

Outcrossing rate, t, of velvety parents was estimated as t = H/p, where H is the frequency of sticky offspring produced, and p is the frequency of the dominant (sticky) allele in the population or, in this case, each experimental plot (with or without herbivores). Incorporating the frequency of the sticky allele allowed us to account for the frequency of cryptic outcrossing among velvety plants. The frequency of the dominant allele was the frequency of homozygous sticky plants surviving in the field in 1998. For the herbivore-free field, 113 of 237 surviving plants were sticky (P = 0·477); in the herbivore-attacked field, 58 of 153 plants were sticky (P = 0·379). We assumed, when calculating this outcrossing rate, that there was no difference in the probability of mating within versus between trichome types. The two types were distributed evenly in the field to control for spatial effects, and there was no difference between the types in the trichome complement of the corolla (which could have affected pollinator behaviour). Because both sticky and velvety plants produce glandular trichomes when young (van Dam et al. 1999), offspring were not scored for trichome type until they had at least 25 leaves. If fewer than 50 offspring survived to be scored, additional seeds were germinated and plants were grown until they could be scored in the greenhouse. This additional germination was necessary for nine of the 32 parent plants, only one of which had a final sample size below 50 offspring; final sample sizes ranged from 47 to 82 offspring. Differences in sample size reflect variation in germination rate caused by maternal irrigation environment, which was not correlated with outcrossing rate (data not shown).

We used ancova to determine whether floral phenotype (herkogamy, corolla length and open flower number); environment (herbivory and irrigation treatments); or genes (family effects) affected the proportion of outcrossed progeny produced. Genes and environment may affect observed outcrossing rate either indirectly, through effects on phenotype, or directly. Direct effects could be caused, for example, by selective abortion of inbred fruit in stressed plants, and would be manifested by significant main effects in the model. Indirect effects would be manifested by significance of phenotypic covariates, but not main effects (if main effects are significant in analysis of floral variation). Because indirect effects proved important in our analysis (see Results), we also performed a standardized regression analysis including only the three phenotypic variables. Interactions with family were never significant in the ancova and so were removed from the final model presented here. The proportion of outcrossed progeny was subjected to the angular transformation prior to these analyses to conform to ancova assumptions (Sokal & Rohlf 1995).


To investigate the potential for and timing of autonomous selfing in D. wrightii, five pollination treatments were imposed on greenhouse-grown plants (Table 1). Plants were examined between 1400 and 1500 h daily during the experiment. At this time, flowers undergo anthesis (the beginning of pollen shedding) but corollas are still tightly furled; flowers open at dusk in summer (approximately 1930 hours). In the greenhouse, flowers remained open (no corolla wilting) until late afternoon of the day following anthesis.The corolla and attached stamens would usually be shed as a unit 2 days after anthesis.Herkogamy was measured at anthesis, except for the pre-anthesis (Control) treatment which was measured 1 day earlier. Post-anthesis emasculation was done prior to the corolla wilting. Any capsules that formed were collected and all seeds counted. A total of 30 plants were used in this experiment, 14 velvety and 16 sticky, 11 of which had two full sets of the five treatments imposed (41 sets total). Preliminary analyses indicated that neither trichome type nor the number of treated flowers on a plant affected the results, so all data were pooled for the analysis.

Table 1.  Pollination treatments used in the greenhouse experiment to evaluate the potential for autonomous selfing
NamePollen added from:Anthers removed?Corolla removed?Interpretation of numbers of seeds set
OutcrossDifferent plantAt anthesisWhen pollinatedMaximum possible in greenhouse
SelfSame flowerAt anthesisWhen pollinatedMaximum possible due to selfing
Prior/delayedNoneNoNoAutonomous selfing during entire floral lifetime
PriorNonePost-anthesis (1 day)Post-anthesis (1 day)Autonomous selfing as flower opens
ControlNonePre-anthesis (1 day)Post-anthesis (1 day)Apomictic or accidental seed set

Pollination treatments (Table 1) were applied randomly, subject to the following constraints. First, when individual plants did not have enough flower buds for all pollination treatments, the Prior/Delayed and Self treatments were performed first (in random order), because this contrast was considered potentially most interesting. In addition, if anthers and stigma were in contact at anthesis, the flower was not used for the Outcross treatment, but instead was randomly assigned to one of the other treatments. Even with this constraint, with the exception of immature flowers (Control treatment measured prior to anthesis), herkogamy did not differ among flowers assigned to the different pollination treatments (see Results).

The Prior and Prior/Delayed treatments are similar in that they indicate the number of seeds set due to autonomous selfing (Table 1). However, the Prior treatment indicates the number resulting from pollination (and subsequent fertilization) prior to or during the opening of the flower. The Prior/Delayed treatment includes pollination during the opening of the flower as well as pollination that occurs when the corolla wilts or during subsequent abscission (as a unit) of corolla and anthers. Both treatments may include autonomous selfing while the flower is open. The difference in seed set between the Prior or Prior/Delayed treatments and the Self treatment indicates the compensation through seed set that is possible without pollinator intervention at each stage of floral life. Comparing Prior and Prior/Delayed to each other indicates how much seed set is due to pollination during corolla abscission alone. Comparing the Self and Outcross treatments gives a preliminary indication of whether inbreeding depression occurs in D. wrightii, although this was not the focus of the current study. Seed production in the Control treatment indicates either apomixis or pollination through either greenhouse ventilation or accidental introduction of a pollinator.

To compare herkogamy, proportion fruit set, and seed number per fruit among pollination treatments, we performed one-way anovas with pollination treatment as the main effect; in all instances the model was highly significant (P < 0·0001), so we subsequently used the Ryan–Einot–Gabriel multiple range test (Ryan’s Q) to reveal differences among the treatments. This test performs stepwise comparisons of means, and controls for the experiment-wise type I error rate, which is appropriate when a collection of pairwise comparisons is made (Day & Quinn 1989). Proportion fruit set was subjected to angular transformation prior to the analysis (Sokal & Rolf 1995). Within each pollination treatment, differences in herkogamy for flowers that did and did not form fruits were determined by t-tests.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References


Velvety plants tended to have greater herkogamy than sticky plants, especially when unwatered and attacked by herbivores (Fig. 1). Overall, however, there was no difference in herkogamy between the two trichome types, although significant two- and three-way interactions between trichome type and irrigation and herbivory treatments (Table 2) indicate that the two trichome types respond differently to the different treatment combinations. Corolla length was reduced in unwatered blocks, and the number of open flowers was reduced when herbivores were present (Table 2; Fig. 1). In addition to these significant effects of environment on floral phenotype, there were significant family effects on all three traits, suggesting they have a genetic basis (Table 2).


Figure 1. Herkogamy, corolla length and number of open flowers (mean ± SE) for plants in an experimental field (n = 104 for herkogamy and corolla length; n = 108 for number of open flowers). Plants were grown with and without supplemental water and in the presence and absence of natural levels of herbivory.

Download figure to PowerPoint

Table 2. anova illustrating the effects of herbivory, water, family, and trichome type on herkogamy, corolla length and the total number of open flowers
 dfHerkogamyCorolla lengthOpen flowers
  1. Trichome type effects were tested against variation among families within types. Degrees of freedom differ for open flowers because several individuals never flowered, and so herkogamy and length were not measured on these individuals; however, these non-flowering individuals were included in the analysis of open flower number.

Herbivory 10·295 2·14 0·14710·217 0·350·553412218·3912·220·0007
Irrigation 10·129 0·93 0·33657·40812·070·00081 604·84 3·330·0711
Herbivory × Irrigation 10·004 0·03 0·86090·006 0·010·92301 63·04 0·350·5570
Trichome Type 12·010 0·89 0·37685·160 2·010·19891 397·84 0·420·5391
Family (Type) 72·25816·36 0·00012·563 4·180·00057 954·24 5·260·0001
Type × Herbivory 10·008 0·06 0·80380·854 1·390·24131 144·67 0·800·3743
Type × Irrigation 10·608 4·41 0·03860·473 0·770·38251 182·00 1·000·3192
Type × Herbivory × Irrigation 11·031 7·47 0·00760·510 0·830·36431 226·20 1·250·2671
Error880·138  0·614  93 181·48  

There was a strong negative relationship between herkogamy and fruit and seed set. Flowers forming fruits had significantly reduced herkogamy compared to those not forming fruits (1·45 versus 1·77 cm, t(813) = 4·3, P < 0·0001). In addition, seed number per fruit decreased significantly as herkogamy increased (Table 3; Fig. 2). This relationship did not vary among types or herbivory/irrigation treatments, but velvety plants produced more seeds when unwatered, and sticky plants when watered (type by herbivory interaction, Table 3).

Table 3.  Effect of herkogamy and environment on seed number per fruit
  1. This analysis is performed for individual flowers and fruits. Trichome type effects were tested against variation among families within types.

Herkogamy 193 36216·660·0001
Herbivory 116 801 3·000·0857
Irrigation 1 6 421 1·150·2863
Herbivory × Irrigation 1 8 294 1·480·2259
Trichome type 1  202 0·030·8586
Family (type) 7 5 918 1·060·3954
Type × Herbivory 116 365 2·920·0898
Type × Irrigation 129 316 5·230·0237
Type × Herbivory × Irrigation 112 467 2·230·1381
Error133 5 602  

Figure 2. Relationship between herkogamy and seed number per fruit in an experimental field (see Table 3; n = 149 flower/fruit combinations). Independent contrasts between the eight trichome type/herbivory/irrigation treatment combinations indicated that the slopes were homogeneous (F = 1·01, df 7, 133, P = 0·42), so all data are combined for this figure.

Download figure to PowerPoint


Outcrossing rates ranged from 0 to 67% for individual parent plants, and averaged 22–31% for the different herbivory and water treatment combinations. Plants with more open flowers had reduced outcrossing rates, but outcrossing did not vary among three velvety families or between irrigation and herbivory treatments (Table 4A). That is, there were no direct effects of genes or environment on outcrossing, although there were indirect effects of these factors through their effects on floral display. When the non-significant treatments (herbivory, irrigation and family) were removed from the ancova and a standardized regression analysis was performed, there was a significant positive effect of herkogamy on outcrossing rate, no effect of flower length, and a significant negative effect of open flower number (Table 4B; Fig. 3).

Table 4.  Effect of floral phenotype, environment and family on outcrossing rates of 32 velvety plants from three families
Herkogamy 10·045 1·880·183
Corolla length 10·037 1·550·226
Open flowers 10·25010·420·004
Herbivory 10·058 2·440·132
Irrigation 10·026 1·090·308
Herbivory × Irrigation 10·047 1·980·173
Family 20·008 0·320·728
  1. Outcrossing rate was estimated as the frequency of sticky offspring produced by velvety parents, divided by the frequency of sticky plants in the experimental field (see text). A, ancova, including phenotypic variables and all treatments; B, standardized regression analysis of outcrossing rate on phenotypic variables.

Herkogamy  0·0850·021
Corolla length−0·0200·540
Open flowers−0·1210·001
Total R2 (n = 32)  0·3480·007

Figure 3. Relationship between herkogamy, number of open flowers and outcrossing rate for 32 velvety plants in an experimental field. Outcrossing rates were determined using the trichome type locus as a genetic marker, and so could not be determined for sticky plants.

Download figure to PowerPoint


Greenhouse experiments indicated that autonomous self-pollination is possible in D. wrightii. Control flowers with anthers removed prior to dehiscence never set any fruit, indicating that neither apomixis nor vector-mediated pollen movement occurred in our greenhouse. The Outcross treatment always resulted in fruit set, whereas only some flowers in the remaining three treatments set fruits (Table 5). Fruit set was higher in treatments where pollen was added (Outcross and Self) than in the treatments testing for the presence of autonomous selfing mechanisms (Prior and Prior/Delayed; Table 5; Fig. 4), but as many as half the flowers measured in the latter treatments set fruits without manual pollen addition.

Table 5.  Herkogamy (cm) for flowers in different treatments of the greenhouse experiment (mean ± SE)
 All flowersNo fruit formedFruit formed
  1. Control flowers were measured 1 day prior to anthesis, when anther filaments were not fully elongated (see text). All flowers in the outcross treatment formed fruit, and all flowers in the control treatment did not.

Outcross251·26 ± 0·15 0251·26 ± 0·15
Self411·22 ± 0·16 51·83 ± 0·54361·13 ± 0·16
Prior/delayed411·29 ± 0·14201·60 ± 0·14210·99 ± 0·24
Prior361·12 ± 0·12241·42 ± 0·13120·52 ± 0·16
Control252·05 ± 0·13252·05 ± 0·13 0

Figure 4. Proportion of flowers that set fruit (open bars) and number of seeds per capsule when fruit were formed (shaded bars) for each pollination treatment in the greenhouse. Sample sizes are indicated within each bar. Different letters indicate differences in fruit or seed number determined by Ryan’s Q with α = 0·05. Treatments defined in Table 1. No capsules were formed in the Control treatment, which is not shown.

Download figure to PowerPoint

Herkogamy of flowers assigned to the different treatments did not differ, with the exception of the Control treatment. Flowers measured 1 day early for this treatment had significantly greater herkogamy than the other pollination treatments (Table 5). Anther filaments are still elongating at this stage in floral development (E.E., personal observation). For those pollination treatments where there was variation in fruit set, we compared herkogamy of flowers that did and did not form fruits using t-tests. For the Self treatment, there was no difference in herkogamy for flowers that did and did not form fruits (t39 = 1·5, P = 0·14), although the 36 flowers that formed fruits had reduced herkogamy compared to the five flowers that did not form fruits (Table 5). Within both autonomous selfing treatments, the probability of fruit set was increased with decreased herkogamy (Prior/Delayed: t39 = 2·2, P = 0·03; Prior: t34 = 4·1, P = 0·0002).

Seed number per fruit did not differ between the two pollen addition treatments (Self and Outcross), between the Self and the Prior/Delayed treatments, or between the Prior and Prior/Delayed treatments, although the Self and Prior treatments did differ (Fig. 4). Of those flowers that set fruit, the Prior/Delayed treatment had 72% seed set compared to the Self treatment; the Prior treatment had 46%. Despite this difference in magnitude, these two treatments were not significantly different from one another in seed number, in part due to large standard errors resulting from the variability in seed number per fruit (CV = 74·8% for Prior/Delayed treatment, CV = 84·7% for the Prior treatment; in contrast, CV = 43·6% for Self and 29·4% for Outcross). Yet these results indicate that about half the seeds produced by a flower may be due to selfing prior to floral opening. Aside from the relationship between herkogamy and fruit set, there was no further relationship between herkogamy and seed number in the Prior treatment (R2 = 0·10, P = 0·34). For the Prior/Delayed treatment, however, this relationship was strong and negative (R2 = 0·47, P = 0·0009). Thus, although overlap between the anther and stigmas is not necessary to ensure some self-pollination, reduction in herkogamy can increase the numbers of seeds produced due to corolla dragging at the end of floral life.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

In D. wrightii, variation in the floral phenotype depends on a combination of genetic and environmental factors. Although not designed to estimate heritability, our significant family effects indicate that there is a genetic component to floral traits in D. wrightii, as in other Solanaceous plants including D. stramonium (Motten & Stone 2000) and Nicotiana rustica (Breese 1959). Also important, however, are environmental factors, including the presence/absence of herbivory, and the presence/absence of supplemental water.

Studies that compared different populations or subspecies demonstrated that herkogamy is reduced in hot, dry habitats (Eckhart & Geber 1999; Holtsford & Ellstrand 1992; Rick, Fobes & Holle 1977; Schoen 1982). This study, which examined plastic responses to environmental factors within an experimental population, did not demonstrate a consistent relationship between water availability and herkogamy. Trichome types did respond differently to irrigation, however. Velvety plants tended to have greater herkogamy in unwatered blocks, while sticky plants tended to have reduced herkogamy when unwatered. Interpopulation studies predict reduced herkogamy and/or increased selfing in xeric habitats because such habitats are deemed more stressful for plants (Stebbins 1957), possibly selecting against full expansion of the flower when water is limiting. In the present study, although corollas were longer when plants were irrigated, corolla length and herkogamy were not correlated (R = –0·02, P = 0·64, n = 815 flowers). In addition, herkogamy of velvety plants may have responded contrary to predictions arising from interpopulation studies because unwatered conditions in Riverside are not ‘stressful’ for them, either in comparison to sticky plants or in the larger context of previous interpopulation studies. Velvety plants are more frequent than sticky plants in dry habitats throughout southern and central California (van Dam et al. 1999; Hare & Elle 2001). Sticky plants appear to be limited in distribution due to their production of a water-based trichome exudate (Hare & Elle 2001). A population survey of herkogamy in D. wrightii demonstrated that mean herkogamy of velvety plants growing in the Mojave desert was significantly greater than that of velvety plants growing in more mesic Riversidian scrub and coastal mountain habitats (2·30 versus 1·42 cm, t(111) = 3·97, P < 0·0001). A similar comparison cannot be made with sticky plants because they do not grow in the desert. But, in concert with the results of the present study, these data indicate that drought does not predictably reduce herkogamy in D. wrightii, in contrast to results found with other species.

Herbivory is also expected to affect floral phenotype. Foliar herbivory reduces flower size or pollen production relative to herbivore-free plants of various species (Agrawal et al. 1999; Lehtila & Strauss 1999; Mothershead & Marquis 2000; Quesada et al. 1995; Strauss et al. 1997; Strauss et al. 1999). In the current study, folivory reduced flower production but did not reduce total flower size. In addition, herbivory treatments interacted with the trichome type/irrigation interaction discussed above, with differences in herkogamy between the types and irrigation treatments magnified in herbivore-attacked plants. Reduction in flower number could be expected to reduce the attractiveness of D. wrightii to pollinators and, coupled with the changes in floral architecture documented here, to affect the mating system.

Despite the indication of a relationship between herkogamy and outcrossing rate in a related species of Datura (Motten & Antonovics 1992; Motten & Stone 2000), and in other species (Breese 1959; Brunet & Eckert 1998; Dole 1992; Hermanutz 1991; Holtsford & Ellstrand 1992; Karron et al. 1997; Motten & Antonovics 1992; Rick, Holle & Thorpe 1978; Ritland & Ritland 1989), there is, surprisingly, only a weak relationship in our experimental population of D. wrightii. This may be due, in part, to the potential for autonomous selfing in flowers with herkogamy as large as 0·5–1 cm (see below). However, the outcrossing rate was influenced strongly by the number of open flowers on a plant, such that plants with more open flowers had less outcrossing. This was probably due to increased geitonogamy, within-plant movement of self pollen via pollinators, on plants with many open flowers (de Jong, Waser & Klinkhamer 1993). This behaviour has been documented for Hawkmoth pollinators of D. wrightii (Grant & Grant 1983), and is especially interesting because herbivore-attacked plants had significantly fewer open flowers than herbivore-free plants. Although we found no direct effect of herbivory on outcrossing in this study, the indirect effect (through flower number) appears to be an enhancement of the outcrossing rate due to reduced pollen movement within individuals. This result is specific to the available pollinators in our experimental population during this study. Yet, if selfed and outcrossed progeny differ in quality, greater outcrossing rates in plants that produce fewer flowers may partially compensate for the reduced seed production we have previously documented for herbivore-attacked plants (Elle et al. 1999). Previous research on herbivory has primarily focused on the effect of herbivory on seed production (Marquis 1992; Strauss 1997). Recently, there has been interest in including components of male fitness (e.g. Agrawal et al. 1999; Delph, Johannsson & Stephenson 1997; Mutikainen & Delph 1996; Strauss et al. 1996). The results of the current study suggest that more intensive investigation of the link between herbivory and the mating system could be fruitful.

In addition to the geitonogamous selfing that may have occurred in our field experiment, our greenhouse experiment indicated that self-pollination can occur in D. wrightii without pollinator intervention. Autonomous selfing occurred in close to half of the flowers that were not hand-pollinated, and more seeds were produced through corolla shedding when herkogamy was reduced. Corolla shedding can lead to selfing in Mimulus (Dole 1990; Dole 1992) which, like D. wrightii, has stamens attached to the corolla tube. However, the importance of corolla shedding and other reproductive assurance mechanisms for selfing under field conditions is less clear (Eckert & Schaeffer 1998; Holsinger 1996), and would only properly be assessed via floral manipulations coupled with molecular marker studies (Schoen & Lloyd 1992). What is interesting about our results is that they indicate that autonomous selfing can occur without overlap of stigma and anthers, unlike in previous studies of Datura species (Motten & Antonovics 1992; Motten & Stone 2000; Snow & Dunford 1961). The negative relationship between herkogamy and autonomous seed set in the greenhouse also suggests a mechanism for the weak positive relationship between herkogamy and outcrossing in the field. If increased herkogamy reduces the potential for autonomous selfing in the absence of outcross pollination, plants with greater mean herkogamy may exhibit higher outcrossing rates.

About one-third of flowers can produce seeds autonomously in the greenhouse as they open and during the period when they are normally exposed to pollinators (the Prior treatment). Seed production occurs even in flowers without overlap between male and female parts, in which autonomous pollen transfer is unlikely once flowers are fully open. Pollen transfer may instead occur in these flowers as the pleated and tightly twisted corollas rapidly unfurl to a bell shape, often with a final ‘snap’ as the five points on the corolla limb separate (E.E., personal observation). Selfing prior to the opportunity for outcrossing is considered less advantageous than selfing after opportunities for outcrossing have passed, because it can lead to seed or pollen discounting (Holsinger 1996; Lloyd 1992; Wyatt 1983). The conditions allowing for prior selfing are fairly restrictive, and include low inbreeding depression, strong prepotency mechanisms, and pollinator limitation (Lloyd 1992). We have not yet explored these possibilities. But, the tendency for fruit formation to fail after the addition of self pollen to highly herkogamous flowers (Table 5) suggests that prepotency mechanisms may be important in D. wrightii.

In conclusion, we have shown that autonomous selfing is possible in D. wrightii, and that it is related to herkogamy. Floral phenotype varies with genetic and environmental factors. Despite obvious morphological adaptations for outcrossing, there is the potential for more than 70% of the seeds produced by D. wrightii to result from a combination of autonomous selfing and geitonogamy. If selfing rates within natural populations are similar to those in our experimental population, we hypothesize that interpopulation gene flow may also be restricted, especially if the limited pollinator movement associated with geitonogamous selfing is common. This could be important because populations of D. wrightii in southern and central California differ significantly from one another in the frequency of plants that produce glandular trichomes (Hare & Elle 2001; van Dam et al. 1999). While differences in costs of resistance and the localized herbivore community may explain some of the interpopulation variation in this resistance trait (Elle & Hare 2000; Elle et al. 1999), fixed differences may also have arisen through a combination of random colonization events and inbreeding if self-pollination is widespread.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The authors thank J. Hale, A. Jacques, A. Lacqui, R. Thomson and M. Valderrama for field and greenhouse assistance, and C. Eckert, P. McMillan, M. Neel, R. Sargent, and anonymous reviewers for helpful commentary on the manuscript. This research was supported by the US National Science Foundation.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Adler, L.S. (2000) Alkaloid uptake increases fitness in a hemiparasitic plant via reduced herbivory and increased pollination. American Naturalist 16, 9299.
  • Agrawal, A.A., Strauss, S.Y. & Stout, M.J. (1999) Costs of induced responses and tolerance to herbivory in male and female fitness components of wild radish. Evolution 16, 10931104.
  • Baker, H.G. (1961) The adaptation of flowering plants to nocturnal and crepuscular pollinators. Quarterly Review of Biology 16, 6473.
  • Breese, E.L. (1959) Selection for differing degrees of out-breeding in Nicotiana rustica. Annals of Botany 16, 331344.
  • Brown, A.H.D., Zohary, D. & Nevo, E. (1978) Outcrossing rates and heterozygosity in natural populations of Hordeum spontaneum Koch in Israel. Heredity 16, 4962.
  • Brunet, J. & Eckert, C.G. (1998) Effects of floral morphology and display on outcrossing in blue columbine, Aquilegia caerulea (Ranunculaceae). Functional Ecology 16, 596606.
  • Clegg, M.T. (1980) Measuring plant mating systems. Bioscience 16, 814818.
  • Van Dam, N.M., Hare, J.D. & Elle, E. (1999) Inheritance and distribution of trichome phenotypes in Datura wrightii. Journal of Heredity 16, 220227.
  • Day, R.W. & Quinn, G.P. (1989) Comparisons of treatments after an analysis of variance in ecology. Ecological Monographs 16, 433463.
  • Delph, L.F., Johannsson, M.H. & Stephenson, A.G. (1997) How environmental factors affect pollen performance: ecological and evolutionary perspectives. Ecology 16, 16321639.
  • Dole, J.A. (1990) Role of corolla abscission in delayed self-pollination of Mimulus guttatus (Scrophulariaceae). American Journal of Botany 16, 15051507.
  • Dole, J.A. (1992) Reproductive assurance mechanisms in three taxa of the Mimulus guttatus complex (Scrophulariaceae). American Journal of Botany 16, 650659.
  • Eckert, C.G. & Schaefer, A. (1998) Does self-pollination provide reproductive assurance in Aquilegia canadensis (Ranunculaceae)? American Journal of Botany 16, 919924.
  • Eckhart, V.M. & Geber, M.A. (1999) Character variation and geographic distribution of Clarkia xantiana A. Gray (Onagraceae): flowers and phenology distinguish two subspecies. Madrono 16, 117125.
  • Elle, E. & Hare, J.D. (2000) No benefit of glandular trichome production in Datura wrightii? Oecologia 16, 5765.
  • Elle, E., Van Dam, N.M. & Hare, J.D. (1999) Cost of glandular trichomes, a ‘resistance’ character in Datura wrightii Regel (Solanaceae). Evolution 16, 2235.
  • Grant, V. & Grant, K.A. (1983) Behavior of hawkmoths on flowers of Datura meteloides. Botanical Gazette 16, 280284.
  • Hamrick, J.L., Linhart, Y.B. & Mitton, J.B. (1979) Relationships between life history characteristics and electrophoretically detectable genetic variation in plants. Annual Review of Ecology and Systematics 16, 173200.
  • Hare, J.D. & Elle, E. (2001) Geographic variation in the frequencies of trichome phenotypes of Datura wrightii, and correlation with annual water deficit. Madrono 16, 3337.
  • Hermanutz, L. (1991) Outcrossing in the weed Solanum pytcanthum (Solanaceae): a comparison of agrestal and ruderal populations. American Journal of Botany 16, 638646.
  • Hickman, J.C. (1993). The Jepson Manual: Higher Plants of California. University of California Press, Berkeley, CA.
  • Holsinger, K.E. (1996) Pollination biology and the evolution of mating systems in flowering plants. Evolutionary Biology 16, 107149.
  • Holsinger, K.E., Feldman, M.W. & Christiansen, F.B. (1984) The evolution of self-fertilization in plants: a population genetic model. American Naturalist 16, 446453.
  • Holtsford, T.P. & Ellstrand, N.C. (1992) Genetic and environmental variation in floral traits affecting outcrossing rate in Clarkia tembloriensis (Onagraceae). Evolution 16, 216225.
  • Johnston, M.O. & Schoen, D.J. (1996) Correlated evolution of self-fertilization and inbreeding depression: an experimental study of nine populations of Amsinckia (Boraginaceae). Evolution 16, 14781491.
  • De Jong, T.J., Waser, N.M. & Klinkhamer, P.G.L. (1993) Geitonogamy: the neglected side of selfing. Trends in Ecology and Evolution 16, 321325.
  • Kalisz, S., Vogler, D., Fails, B. et al. (1999) The mechanism of delayed selfing in Collinsia verna (Scrophulariaceae). American Journal of Botany 16, 12391247.
  • Karban, R. & Strauss, S.Y. (1993) Effects of herbivores on the growth and reproduction of their perennial host plant, Erigeron glaucus. Ecology 16, 3946.
  • Karron, J.D., Jackson, R.T., Thumser, N.N. & Schlicht, S.L. (1997) Outcrossing rates of individual Mimulus ringens genets are correlated with anther-stigma separation. Heredity 16, 365370.
  • Krupnick, G.A., Weis, A.E. & Campbell, D.R. (1999) The consequences of floral herbivory for pollinator service to Isomeris arborea. Ecology 16, 125134.
  • Lehtila, K.P. & Strauss, S.Y. (1999) Effects of foliar herbivory on male and female reproductive traits of wild radish, Raphanus raphanistrum. Ecology 16, 116124.
  • Lloyd, D.G. (1992) Self- and cross-fertilization in plants. II. The selection of self-fertilization. International Journal of Plant Sciences 16, 370380.
  • Lloyd, D.G. & Schoen, D.J. (1992) Self- and cross-fertilization in plants. I. Functional dimensions. International Journal of Plant Sciences 16, 358369.
  • Marquis, R.J. (1992) Selective impact of herbivores. Plant Resistance to Herbivores and Pathogens: Ecology, Evolution and Genetics (eds R. S.Fritz & E. L.Simms), pp. 301325. University of Chicago Press, Chicago, IL.
  • Mothershead, K. & Marquis, R.J. (2000) Fitness impacts of herbivory through indirect effects on plant–pollinator interactions in Oenothera macrocarpa. Ecology 16, 3040.
  • Motten, A.F. & Antonovics, J. (1992) Determinants of outcrossing rate in a predominantly self-fertilizing weed, Datura stramonium (Solanaceae). American Journal of Botany 16, 419427.
  • Motten, A.F. & Stone, J.L. (2000) Heritability of stigma position and the effect of stigma-anther separation on outcrossing in a predominantly self-fertilizing weed, Datura stramonium (Solanaceae). American Journal of Botany 16, 339347.
  • Mutikainen, P. & Delph, L.F. (1996) Effects of herbivory on male reproductive success in plants. Oikos 16, 353358.
  • Quesada, M., Bollman, K. & Stephenson, A.G. (1995) Leaf damage decreases pollen production performance in Cucurbita texana. Ecology 16, 437443.
  • Rick, C.M., Fobes, J.F. & Holle, M. (1977) Genetic variation in Lycopersicon pimpinellifolium: evidence of evolutionary change in mating systems. Plant Systematics and Evolution 16, 139170.
  • Rick, C.M., Holle, M. & Thorp, R.W. (1978) Rates of cross-pollination in Lycopersicon pimpinellifolium: impact of genetic variation in floral characters. Plant Systematics and Evolution 16, 3144.
  • Ritland, C. & Ritland, K. (1989) Variation of sex allocation among eight taxa of the Mimulus guttatus species complex (Scrophulariaceae). American Journal of Botany 16, 17311739.
  • SAS Institute. (1988) SAS/STAT User’s Guide. Release 6·03. SAS Institute, Inc., Cary, NC.
  • Schoen, D.J. (1982) The breeding system of Gilia achilleifolia: variation in floral characteristics and outcrossing rate. Evolution 16, 352360.
  • Schoen, D.J. & Lloyd, D.G. (1992) Self- and cross-fertilization in plants. III. Methods for studying modes and functional aspects of self-fertilization. International Journal of Plant Sciences 16, 381393.
  • Snow, R. & Dunford, M.P. (1961) A study of interchange heterozygosity in a population of Datura meteloides. Genetics 16, 10971110.
  • Sokal, R.R. & Rohlf, F.J. (1995) Biometry: The Principles and Practice of Statistics in Biological Research, 3rd edn. W. H. Freeman, New York, NY.
  • Stebbins, G.L. (1957) Self fertilization and population variability in the higher plants. American Naturalist 16, 337354.
  • Strauss, S.Y. (1997) Floral characters link herbivores, pollinators and plant fitness. Ecology 16, 16401645.
  • Strauss, S.Y., Conner, J.K. & Rush, S.L. (1996) Foliar herbivory affects floral characters and plant attractiveness to pollinators: implications for male and female plant fitness. American Naturalist 16, 10981107.
  • Strauss, S.Y., Siemens, D.H., Decher, M.B. & Mitchell-Olds, T. (1999) Ecological costs of plant resistance to herbivores in the currency of pollination. Evolution 16, 11051113.
  • Vasek, F.C. (1964) The evolution of Clarkia unguiculata derivatives adapted to relatively xeric environments. Evolution 16, 2642.
  • Wyatt, R. (1983) Pollinator–plant interactions and the evolution of breeding systems. Pollination Biology (ed. L.Real), pp. 5195. Academic Press, Orlando, FL.