EVOLUTION OF FLORAL TRAITS IN A HERMAPHRODITIC PLANT: FIELD MEASUREMENTS OF HERITABILITIES AND GENETIC CORRELATIONS

Genetic variances, heritabilities, and genetic correlations of floral traits were measured in the monocarpic perennial Ipomopsis aggregata (Polemoniaceae). A paternal half-sib design was employed to generate seeds in each of four years, and seeds were planted back in the field near the parental site. The progeny were followed for up to eight years to estimate quantitative genetic parameters subject to natural levels of environmental variation over the entire life cycle. Narrow-sense heritabilities of 0.2–0.8 were detected for the morphometric traits of corolla length, corolla width, stigma position, and anther position. The proportion of time spent by the protandrous flowers in the pistillate phase (“proportion pistillate”) also exhibited detectable heritability of near 0.3. In contrast, heritability estimates for nectar reward traits were low and not significantly different from zero, due to high environmental variance between and within flowering years. The estimates of genetic parameters were combined with phenotypic selection gradients to predict evolutionary responses to selection mediated by the hummingbird pollinators. One trait, corolla width, showed the potential for a rapid response to ongoing selection through male function, as it experienced both direct selection, by influencing pollen export, and relatively high heritability. Predicted responses were lower for proportion pistillate and corolla length, even though these traits also experienced direct selection. Stigma position was expected to respond positively to indirect selection of proportion pistillate but negatively to selection of corolla length, with the net effect sensitive to variation in the selection estimates. Anther position also was not directly selected but could respond to indirect selection of genetically correlated traits.     

MEASUREMENTS OF SELECTION IN A HERMAPHRODITIC PLANT: VARIATION IN MALE AND FEMALE POLLINATION SUCCESS

I measured phenotypic selection of floral traits through both male and female functions of the hermaphroditic flowers of Ipomopsis aggregata (Pursh) V. Grant subsp. aggregata (Polemoniaceae). Fluorescent powdered dyes were used to track movement of pollen by hummingbirds and to measure pollen delivery to individual plants as well as pollen receipt. A phenotypic selection analysis revealed that selection due to male-male competition during pollination was capable of delaying flowering date and widening corolla tubes by 0.22 and 0.24 standard-deviation units, respectively, in a single generation. Several floral traits were highly correlated with each other. Multivariate selection analysis suggested that selection through male function directly favored late flowering as well as a sexual expression characterized by a short pistillate phase and long corollas. Selection intensities through male and female functions were of similar overall magnitude during the pollination stage of the life cycle, but different traits were favored, and selection sometimes acted in opposing directions. In 1985, selection through female function favored increased time spent in the pistillate phase and exserted stigmas (unlike selection through male function). As a result, individual plants varied greatly in functional gender. Plants that had exserted stigmas and narrow corollas and that spent a disproportionately long time in the pistillate phase achieved greater pollination success as females, while plants with the opposite traits achieved greater success as males. Moreover, female pollination success tended to increase, and male pollination success to decrease, with time spent in the pistillate phase, supporting a critical assumption of sex-allocation theory. Selection in the populations studied fluctuated from year to year and was highly sex-specific.     

POLLEN TRANSFER BY NATURAL HYBRIDS AND PARENTAL SPECIES IN AN IPOMOPSIS HYBRID ZONE

Models of hybrid zones differ in their assumptions about the relative fitnesses of hybrids and the parental species. These fitness relationships determine the form of selection across the hybrid zone and, along with gene flow, the evolutionary dynamics and eventual outcome of natural hybridization. We measured a component of fitness, export and receipt of pollen in single pollinator visits, for hybrids between the herbaceous plants Ipomopsis aggregata and I. tenuituba and for both parental species. In aviary experiments with captive hummingbirds, hybrid flowers outperformed flowers of both parental species by receiving more pollen on the stigma. Although hummingbirds were more effective at removing pollen from anthers of I. aggregata, hybrid flowers matched both parental species in the amount of pollen exported to stigmas of other flowers. These patterns of pollen transfer led to phenotypic stabilizing selection, during that stage of the life cycle, for a stigma position intermediate between that of the two species and to directional selection for exserted anthers. Pollen transfer between the species was high, with flowers of I. aggregata exporting pollen equally successfully to conspecific and I. tenuituba flowers. Although this study showed that natural hybrids enjoy the highest quality of pollinator visits, a previous study found that I. aggregata receives the highest quantity of pollinator visits. Thus, the relative fitness of hybrids changes over the life cycle. By combining the results of both studies, pollinator-mediated selection in this hybrid zone is predicted to be strong and directional, with hybrid fitness intermediate between that of the parental species.     

'Anti-bee' and 'pro-bird' changes during the evolution of hummingbird pollination in Penstemon flowers

Floral phenotypes may be as much the result of selection for avoidance of some animal visitors as selection for improving the interaction with better pollinators. When specializing on hummingbird-pollination, Penstemon flowers may have evolved to improve the morphological fit between bird and flower, or to exclude less-efficient bees, or both. We hypothesized how such selection might work on four floral characters that affect the mechanics of pollen transfer: anther/stigma exsertion, presence of a lower corolla lip, width of the corolla tube, and angle of flower inclination. We surgically modified bee-pollinated P. strictus flowers changing one trait at a time to make them resemble hummingbird-pollinated P. barbatus flowers, and measured pollen transfer by bumblebees and hummingbirds. Results suggest that, apart from 'pro-bird' adaptations, specific 'anti-bee' adaptations have been important in shaping hummingbird-flowers. Moreover, some trait changes may have been selected for only if changing in concert with other traits.     

Functional Significance of Interspecific Variation in Burmeistera Flower Morphology: Evidence from Nectar Bat Captures in Ecuador

What causes flowers to diverge? While a plant's primary pollinator should strongly influence floral phenotype, selective pressures may also be exerted by other flower visitors or competition with other plants for pollination. Species of the primarily bat‐pollinated genus Burmeistera (Campanulaceae) frequently cooccur, with up to four species in a given site, and broadly overlap in flowering phenology, typically flowering throughout the year. The genus displays extensive interspecific variation in floral morphology in the degree that the reproductive parts (anthers and stigma) are exserted outside of the corolla, and species can be roughly classified as either long or short‐exserted. I tested two hypotheses regarding the functional significance of such variation: (1) exsertion lengths correspond to pollination by bat species of different sizes; and (2) variation serves to partition pollinator's bodies spatially and thus reduces interspecific pollen transfer. I captured bats in Ecuador to evaluate the identity and location of the Burmeistera pollen they were carrying. Results show that exsertion does not correspond to specialization on different pollinators; different bat species carried pollen of both flower types just as frequently. In support of the second hypothesis, pollen from flowers of different exsertion lengths was found to occur on different regions of bats' heads. This may serve to reduce competition for pollination among coexisting Burmeistera.     

Pollen transfer by hummingbirds and bumblebees, and the divergence of pollination modes in Penstemon

We compared pollen removal and deposition by hummingbirds and bumblebees visiting bird-syndrome Penstemon barbatus and bee-syndrome P. strictus flowers. One model for evolutionary shifts from bee pollination to bird pollination has assumed that, mostly due to grooming, pollen on bee bodies quickly becomes unavailable for transfer to stigmas, whereas pollen on hummingbirds has greater carryover. Comparing bumblebees and hummingbirds seeking nectar in P. strictus, we confirmed that bees had a steeper pollen carryover curve than birds but, surprisingly, bees and birds removed similar amounts of pollen and had similar per-visit pollen transfer efficiencies. Comparing P. barbatus and P. strictus visited by hummingbirds, the bird-syndrome flowers had more pollen removed, more pollen deposited, and a higher transfer efficiency than the bee-syndrome flowers. In addition, P. barbatus flowers have evolved such that their anthers and stigmas would not easily come into contact with bumblebees if they were to forage on them. We discuss the role that differences in pollination efficiency between bees and hummingbirds may have played in the repeated evolution of hummingbird pollination in Penstemon.     

The effect of wild radish floral morphology on pollination efficiency by four taxa of pollinators

The effects of floral morphology on rates of pollen removal and deposition by different pollinators in generalist plant species are not well known. We studied pollination dynamics in wild radish, Raphanus raphanistrum, a plant visited by four groups of pollinators: honey bees, small native bees, butterflies, and syrphyd flies. The effects of anther position and other factors on pollen removal during single visits by all four pollinator taxa were measured. Flowers with high anther exsertion (i.e., anthers placed higher above the opening of the corolla tube) tended to have the highest numbers of pollen grains removed, but this effect was strongest for honey bees and butterflies. For all pollinator taxa, pollen removal increased with the number of pollen grains available on a flower and whowed a positive, decelerating relationship with the duration of the visit. The effects of stigma position and other factors on pollen deposition during single visits by honey bees and butterflies were also studied. The nectar-feeding butterflies had a higher pollination efficiency (percentage of pollen grains removed from anthers that were subsequently deposited on a stigma) than the nectar- and pollen-feeding honey bees. Flowers with intermediate stigma exsertion had the highest numbers of pollen grains deposited on their stigmas by butterflies, but stigma exsertion had no effect on deposition by honey bees. For both butterflies and honey bees, pollen deposition on the recipient flower increased with the amount of pollen removed from the donor flower, and there was a positive, decelerating relationship between deposition and time spent at the flower; these results are analogous to those for pollen removal. The effects of anther and stigma exsertion on pollen removal and denosition did not fit predictions based on patterns of floral correlations, but results for morphology, pollen availability, time spent per visit, and pollinator efficiency are in broad agreement with previous studies, suggesting the possible emergence of some general rules of pollen transfer.     

High-resolution mapping and functional analysis of se2.1: a major stigma exsertion quantitative trait locus associated with the evolution from allogamy to autogamy in the genus Lycopersicon

The degree to which stigmas are exserted above the stamen in flowers is a key determinant of cross-pollination (and hence allogamy) in many plant species. Most species in the genus Lycopersicon are obligate or facultative outcrossers and bear flowers with highly exserted stigmas. In contrast, the cultivated tomato (Lycopersicon esculentum) bears flowers with flush or inserted stigmas promoting self-fertilization. It has been observed that a major QTL, se2.1, on chromosome 2 is responsible for a large portion of phenotypic variation for this trait and that mutation(s) at this locus were likely involved in the evolution from allogamy to autogamy in this genus. To understand the genetic and molecular basis of stigma exsertion, we have conducted a high-resolution mapping at the chromosome region harboring the se2.1 QTL. The results indicate that this is a compound locus, comprising at least five tightly linked genes, one controlling style length, three controlling stamen length, and the other affecting anther dehiscence, a taxonomic character used to distinguish Lycopersicon species from other solanaceous species. This cluster of genes may represent the vestiges of an ancient coadapted gene complex in controlling mating behavior.     

Reproductive ecology of distylous Palicourea Padifolia (Rubiaceae) in a tropical montane cloud forest. I. Hummingbirds' effectiveness as pollen vectors

The adaptiveness of distyly has been typically investigated in terms of its female function, specifically pollen receipt. However, pollen loads on stigmas can only provide moderate support for Darwin's hypothesis of the promotion of legitimate crosses. To determine the effectiveness of hummingbirds as pollen vectors between floral morphs and the consequences in terms of male (pollen transfer) and female function (pollen receipt) in Palicourea padifolia (Rubiaceae), floral visitors, their foraging modes, and temporal patterns of floral visitation were observed and documented. Differences in pollen and stigma morphology, pollen flow, rates of pollen deposition, and/or stigmatic pollen loads were then evaluated for their contribution toward differences in reproductive output between floral morphs. A pollination experiment with stuffed hummingbirds that varied in bill size was done to evaluate the contribution of bill variation toward differences between floral morphs in pollen receipt and pollen transfer and female reproductive output. Anthers of long-styled flowers contained significantly more and smaller pollen grains than those of short-styled flowers, independently of corolla and anther lengths. The shape and orientation of the stigma lobes differed between morphs and were significantly longer among short-styled flowers. Hummingbird visitation rates did not differ significantly between floral morphs, and foraging movements from focal plants towards neighboring plants were independent of floral morph. Stigmatic pollen loads under field conditions and those after controlled hummingbird visitation, along with rates of pollen accumulation through the day indicated that stigmas of short-styled flowers receive proportionately more legitimate (intermorph) pollen grains than did those of long-styled flowers. However, the species of hummingbird was marginally significant in explaining variation in pollen deposition on stigmas. Lastly, intermorph pollinations of P. padifolia resulted in significant differences in fruit production between floral morphs, independent of pollination treatment and pollinator species; short-styled flowers proportionately developed almost twice the number of fruits developed by long-styled flowers.     

The mean and variability of a floral trait have opposing effects on fitness traits

Background and Aims Floral traits are essential for ensuring successful pollination and reproduction in flowering plants. In particular, style and anther positions are key for pollination accuracy and efficiency. Variation in these traits among individuals has been well studied, but less is known about variation within flowers and plants and its effect on pollination and reproductive success.Methods Style deflexion is responsible for herkogamy and important for pollen deposition in Passiflora incarnata. The degree of deflexion may vary among stigmas within flowers as well as among flowers. We measured the variability of style deflexion at both the flower and the plant level. The fitness consequences of the mean and variation of style deflexion were then evaluated under natural pollination by determining their relationship to pollen deposition, seed production and average seed weight using structural equation modelling. In addition, the relationship between style deflexion and self-pollen deposition was estimated in a greenhouse experiment.Key Results We found greater variation in style deflexion within flowers and plants than among plants. Variation of style deflexion at the flower and plant level was positively correlated, suggesting that variability in style deflexion may be a distinct trait in P. incarnata. Lower deflexion and reduced variation in that deflexion increased pollen deposition, which in turn increased seed number. However, lower styles also increased self-pollen deposition. In contrast, higher deflexion and greater variability of that deflexion increased variation in pollen deposition, which resulted in heavier seeds.Conclusions Variability of style deflexion and therefore stigma placement, independent from the mean, appears to be a property of individual P. incarnata plants. The mean and variability of style deflexion in P. incarnata affected seed number and seed weight in contrasting ways, through the quantity and potentially quality of pollen deposition. This antagonistic selection via different fitness components may maintain diverse style phenotypes.     

Shortened anther-stigma distance reduces compatible pollination in two distylous Primula species

Aims Distyly has been regarded as an adaptation to improve compatible pollination between two floral morphs with reciprocal herkogamy. The hypothesis that the different positions of anthers and stigmas within flowers as well as their reciprocal position between morphs, reduce the probability of self pollination raised by Darwin has been rarely tested. In this study, we measured stigmatic pollen loads in response to reduced reciprocal herkogamy in two Primula species.Methods To see whether reciprocal herkogamy can increase compatible and/or reduce incompatible pollen deposition, thus promoting compatible pollination, we shortened the distance between anthers and stigmas within the flowers by changing the position of the corolla tube, to which the anthers were fused, i.e. reduced herkogamy in natural populations of Primula secundiflora and P. poissonii and quantified stigmatic pollen loads in the field over 2 years.Important findings In both species, stigmatic pollen loads were significantly higher in the long-styled (L-morph) than in the short-styled morph (S-morph) in both control and manipulated flowers, but percentage of compatible pollen in S-morph were higher. Flowers manipulated to halve the anther–stigma distance showed a similar pattern for 2 years: total pollen grain counts on stigmas did not differ significantly but compatible pollen grains in L- and S-morphs were significantly decreased in both species. The percentage of compatible pollen loads was decreased by 68.7% in P. secundiflora and 65.3% in P. poissonii in L-morphs, while it decreased by 30.6% and 2.9% in S-morphs, respectively. Our manipulation of the relative position of anthers and stigmas in the two distylous species indicated that a lower degree of herkogamy reduced compatible but incompatible pollen transfer was likely to increase. The higher proportion of compatible pollen in the S-morph than in the L-morph in the two Primula species could be attributed to the accessibility of two-level sexual organs, floral orientations and pollinator behaviors. This is a first attempt to manipulate intraflower herkogamy for understanding adaptation of heterostyly, shedding insights into how the reciprocal herkogamy promotes compatible pollination.     

水稻籼粳交DH群体花器性状的遗传分析

水稻花器性状是影响杂交稻制种异交结实率的重要因素之一。利用一个水稻籼粳交(窄叶青8号/京系17)来源的DH群体对水稻柱头外露率、花柱长、柱头长、柱头宽、花药长、花药宽、颖花长、颖花宽和颖花长宽比9个花器性状和穗抽出度进行遗传分析。不论性状在2个亲本间的差异显著与否,在DH群体中10个研究性状在基因型间的差异均达到极显著水平(P<0.01),并表现为连续变异和明显的超亲分离。除花药宽的遗传力较低(50%)、且估计的基因数多达13个之外,其余9个研究性状的遗传力均较高,为68%～93%,控制这些性状的基因数估计为7～10个。这些结果表明,10个研究性状均为受多基因控制的数量性状,其增效和减效基因在2个亲本中均有分布,通过基因重组可产生正向和负向两个方向的超亲基因型。除了柱头长、柱头宽和穗伸出度3个性状未发现有基因间的上位性互作外,其余7个性状均检测到显著的互补性互作。性状相关和通径分析的结果显示,与柱头外露率关系最密切的花器性状为柱头长、柱头宽和颖花长宽比,颖花长和颖花宽主要通过影响颖花长宽比来对柱头外露率产生影响。其次,与柱头外露率关系较密切的为花柱长。柱头外露率高的基因型通常表现为长柱头、长粒型和长花柱。花药长、花药宽与上述花器性状间的相关性较弱。在DH群体中,穗伸     

Experimental studies of pollen carryover: Hummingbirds and Ipomopsis aggregata

Summary We present results of experiments designed to identify floral characteristics that influence patterns of pollen carryover by hummingbirds visiting Ipomopsis aggregata flowers. We used fluorescent dye powders as pollen analogues. For all four experimental treatments considered, amounts of dye deposited on recipient stigmas declined linearly as a function of flower position in a visitation sequence. The decline was significantly steeper when recipient flowers had pollen-carrying anthers than when they did not; whereas degree of stigma clogging and presence or absence of empty anthers did not influence carryover. From this we conclude that presence of pollen on recipient flowers significantly reduces the average number of subsequent flowers reached by donor pollen. We discuss mechanisms for this effect and its significance for the evolution of floral structure.     

Why do stigmas move in a flexistylous plant?

Flexistyly is a recently documented stylar polymorphism involving both spatial and temporal segregation of sex roles within hermaphroditic flowers. Using the experimental manipulation of stigma movement in self-compatible Alpinia mutica, we tested the hypothesis that selection for reducing interference between male and female function drives the evolution and/or maintenance of stigma movement. In experimental arrays, anaflexistylous (protogynous) flowers served as pollen donors competing for mating opportunities on cataflexistylous (protandrous) flowers. The pollen donors were either manipulated so their stigmas could not move or were left intact, and their success was determined using allozymes to assess the paternity of recipient seeds. We found that manipulated flowers sired a significantly smaller proportion of seeds, showing that stigma movement in unmanipulated plants increased male fitness. This result was strongest under conditions in which pollen competition was expected to be highest, specifically when pollinators visited multiple donor plants before visiting recipient flowers.     

The pollination biology and breeding system of Monarda fistulosa (Labiatae)

Summary Successful cross-pollination of Monarda fistulosa is the result of a complex interaction among flower opening, the pollen-bearing areas of the pollinators and/or their behavior, and the maturation of the stigmas. The flowers open continuously from 0800–2000 h providing a temporally predictable rich source of nectar and pollen. Recently opened flowers may reduce the ability of bees to discriminate between resource rich and poor patches and encourage systematic foraging within patches. The continuous opening of flowers coupled with protandry also results in some flowers of most capitula being in the staminate and others in the pistillate phase. Autogamy is highly unlikely due to strong protandry and the spatial separation of anthers and stigmas. Geitonogamy, at least that mediated by Bombus is unlikely because the pollen is spread over a relatively large area of the wings, which reduces the likelihood of a stigma contacting just deposited pollen. Because pollen is transferred from the much smaller coxal area of Anthophora and other bees that mistake the stigmas of early pistillate phase flowers for stamens some geitonogamy seems inevitable. However, the delayed receptivity of young stigmas to self-pollen decreases the likelihood of self-pollen germinating on such stigmas. Older stigmas are equally receptive to self- and cross-pollen and the number of pollen grains germinating and pollen tubes reaching the base of the style increases with flower age.     

Performance of Apis mellifera, Bombus impatiens, and Peponapis pruinosa (Hymenoptera: Apidae) as pollinators of pumpkin

Pollination services of pumpkin, Cucurbita pepo L., provided by the European honey bee, Apis mellifera L., were compared with two native bee species, the common eastern bumble bee, Bombus impatiens (Cresson), and Peponapis pruinosa Say, in New York from 2008 to 2010. Performance of each species was determined by comparing single-visit pollen deposition, percentage of visits that contacted the stigma, flower-handling time, fruit and seed set, and fruit weight per number of visits. Fruit yield from small fields (0.6 ha) supplemented with commercial B. impatiens colonies was compared with yield from those not supplemented. A. mellifera spent nearly 2 and 3 times longer foraging on each pistillate flower compared with B. impatiens and P. pruinosa, respectively. A. mellifera also visited pistillate flowers 10-20 times more frequently than B. impatiens and P. pruinosa, respectively. Yet, B. impatiens deposited 3 times more pollen grains per stigma and contacted stigmas significantly more often than either A. mellifera or P. pruinosa. Fruit set and weight from flowers visited four to eight times by B. impatiens were similar to those from open-pollinated flowers, whereas flowers pollinated by A. mellifera and P. pruinosa produced fewer fruit and smaller fruit compared with those from open-pollinated flowers. Fields supplemented with B. impatiens produced significantly more pumpkins per plant than nonsupplemented fields. B. impatiens was a better pollinator of pumpkin than P. pruinosa and should be considered as a promising alternative to A. mellifera for pollinating this crop.     

Selection on floral design in Polemonium brandegeei (Polemoniaceae): female and male fitness under hawkmoth pollination

Plant-pollinator interactions promote the evolution of floral traits that attract pollinators and facilitate efficient pollen transfer. The spatial separation of sex organs, herkogamy, is believed to limit sexual interference in hermaphrodite flowers. Reverse herkogamy (stigma recessed below anthers) and long, narrow corolla tubes are expected to promote efficiency in male function under hawkmoth pollination. We tested this prediction by measuring selection in six experimental arrays of Polemonium brandegeei, a species that displays continuous variation in herkogamy, resulting in a range of recessed to exserted stigmas. Under glasshouse conditions, we measured pollen removal and deposition, and estimated selection gradients (β) through female fitness (seeds set) and male fitness (siring success based on six polymorphic microsatellite loci). Siring success was higher in plants with more nectar sugar and narrow corolla tubes. However, selection through female function for reverse herkogamy was considerably stronger than was selection through male function. Hawkmoths were initially attracted to larger flowers, but overall preferred plants with reverse herkogamy. Greater pollen deposition and seed set also occurred in reverse herkogamous plants. Thus, reverse herkogamy may be maintained by hawkmoths through female rather than male function. Further, our results suggest that pollinator attraction may play a considerable role in enhancing female function.     

FLORAL DISPLAY IN PHYLA INCISA: CONSEQUENCES FOR MALE AND FEMALE REPRODUCTIVE SUCCESS

The inflorescences of Phyla incisa consist of flowers in two phases: younger, nectar-containing flowers that have yellow corolla throats and older, nectar-lacking flowers that have dark purple corolla throats. Observations of pollinator visitation patterns to both natural and manipulated inflorescences were made to determine the role of each flower phase in pollinator attraction. The effect of older-phase flowers on male and female reproductive success was determined by comparing stigmatic pollen loads and estimates of pollen removal from inflorescences having different numbers of these flowers. The pollinators of Phyla selected larger inflorescences more often than expected based upon the size distribution of inflorescences available to them. Both younger- and older-phase flowers contributed to the attraction of pollinators, but the latter were less effective in this function. The presence of older-phase flowers significantly increased the visitation rate to inflorescences and the amount of pollen removed but had little effect on pollen deposition on stigmas. The lack of correspondence between pollen deposition and pollinator-visitation rate was not due to stigma saturation, since stigma loads varied greatly. The data indicate that the deposition of pollen on stigmas in this species is a relatively stochastic process, whereas pollen removal from inflorescences occurs at a much more regular rate. Old-phase flower retention appears to contribute to reproductive success through increased pollen donation when pollinator activity is high and may also increase the probability of seed set when pollinators are rare.     

COMPONENTS OF PHENOTYPIC SELECTION: POLLEN EXPORT AND FLOWER COROLLA WIDTH IN IPOMOPSIS AGGREGATA

In the hummingbird-pollinated herb Ipomopsis aggregata, selection through male function during pollination favors wide corolla tubes. We explored the mechanisms behind this selection, using phenotypic selection analysis to compare effects of corolla width on two components of male pollination success, pollinator visit rate and pollen exported per visit. During single visits by captive hummingbirds, flowers with wider corollas exported more pollen, and more dye used as a pollen analogue, to stigmas of recipient flowers. Corolla width was less strongly related to visit rate in the field, and had no direct effect on visit rate after nectar production and corolla length were controlled for. Moreover, the phenotypic selection differential was 80% higher for the effect on pollen exported per visit, suggesting that this is the more important mechanism of selection.