Why do flowers show variety?

The remarkable diversity of flowers has a profound impact on humankind. Flower diversity inspires masterpieces of art and literature, fuels highway beautification schemes, gives amateur gardeners moments of great joy and pride, and nurtures industries of bulb growers, seed companies, and horticultural supply houses. Botanists, for their part, have pondered the processes driving diversification in flower form for over two centuries (Sprengel [1793] 1972). Nearly all botanists give animal pollinators credit for the vast smorgasbord of flower shapes and sizes observed in nature. This conclusion is intuitively satisfying in that pollinators have the motive (energy and nutrition gain) and the means (via pollen transfer, a key step in plant sexual reproduction) to exert selection on the floral features of their host plants. Moreover, the taxonomic diversity of pollinators (insects, rodents, lizards, fish, primates, and birds) accommodates a tremendous range of morphology, sensory modality, and behavior to match the variation in flower form.

Phylogenetic studies have suggested that the acquisition of key floral characters, such as the spurs of columbine flowers, which foster interactions with novel kinds of pollinators, can launch the adaptive radiation of flowering plant groups Ultimately, the process of diversification in flower form among species or genera of plants must be fueled by genetic variation in floral traits at the intraspecific level. Indeed, within species and even populations, individual plants vary widely in flower size and form, and much of this variation appears to have a heritable basis . Are pollinators the sole driving force behind such variation? New evidence suggests that selection on flower form and size is a more pluralistic process, involving not only pollinators but also enemies and even aspects of the plant's abiotic environment. Indeed, several recent studies have taken issue with the notion that divergence in flower form solely reflects diversifying selection by pollinators

Flowers and the organs that compose them fulfill a number of ecological functions over their lifetimes

In this article, I advocate a more holistic view of the ecology of flower size and shape based on the general premise that the morphological features of an organism are subject to multiple and conflicting selection pressures. Specifically, I consider why pollinator-mediated selection does not fully explain variation in flower size within and among plant populations, drawing mainly from my studies of floral evolution in the alpine skypilot, Polemonium viscosum. When possible, I also draw on findings from other species to illustrate the general conditions under which ecological forces other than pollinators may drive the evolution of flower morphology.

In particular, I emphasize two neglected aspects of the functional ecology of flowers that bear on the evolution of flower size and form: interactions of plants with flower enemies and resource costs associated with floral display. My focus is on flower size in particular, although many of the points made in this article should also apply to the evolution of flower shape. Because my intent is to bring up new perspectives on the functional ecology of flowers, topics that have historically received more attention are not reviewed here. Two such topics that deserve special mention are the role of genetic constraints in the evolution of flower form and the interplay between mating system evolution and flower morphology.

The process of pollinator-driven diversification in flower form is one of evolutionary specialization. Plant populations or subpopulations are thought to diverge over time in response to directional selec

When enemies are not mobile over plant-to-plant distances, pollinators may provide a convenient vector, intimately linking the processes of pollination and parasitism ). For examplE he incidence of fungal infection correlates with bumblebee pollinator preference in populations of Dianthus Silvester (Shykoff et al. 1997). Conversely, in interactions involving pre-dispersal seed predators, pollination directly affects host-plant quality for enemies. In such cases, selection may favor enemies whose host preference converges on the floral preference of pollinators, because well-pollinated flowers provide greater food reserves than poorly pollinated ones. One of the best-studied interactions of this kind concerns hummingbird pollinators and Dipteran seed predators of I. aggregata. Both hummingbirds and flies discriminate among plants of I. aggregata on the basis of flower size (Brody 1992). Large flowers may have more ovules than small ones, and hummingbird preference ensures that a greater proportion of these ovules will develop into seed provisions for fly larvae.

The diverse behavioral mechanisms unifying the flower choice of pollinators and enemies suggest that the conflicting selection pressures of pollinator attraction and enemy avoidance, which are central to the escape hypothesis, may be widespread in animal-pollinated plant populations. Spatial models that explicitly incorporate enemy and mutualist abundances to predict variable selection on flower size would be valuable in exploring the generality of this idea. Another promising approach is illustrated by recent studies that contrast flower predator-mediated selection on sex morphs of dioecious or gynodioecious plant species (Puterbaugh 1998). For example, the escape hypothesis predicts that males, freed from selection for ovule protection, should show more exaggerated responses to pollinator-mediated selection on flower size than females of the same species. This idea offers an attractive hypothesis for the trend of greater allocation to corolla size in male flowers than female flowers of unisexual animal-pollinated plants, although alternative explanations for this trend are in no short supply (Delph 1996).

The resource-cost hypothesis: sky-pilot flowers and water use. The resource-cost hypothesis predicts that insofar as less conspicuous, smaller flowers require a lesser investment of essential resources from the plant than large, showy flowers, reduced flower size will be advantageous under resource-poor conditions. For example, for plants that, like P. viscosum, flower repeatedly over their lifetimes, pollinator-mediated selection should drive the evolution of large flower size if individuals have the resources to produce and maintain such flowers without compromising future survival and reproduction. Because resource heterogeneity within plant populations is a ubiquitous phenomenon (e.g., Bell and Lechowicz 1991, Reader and Bonser 1993, Stanton et al. 1994), the resource-cost hypothesis may have general applicability to the maintenance of genetic variation in flower size at the population scale. According to this hypothesis, resource heterogeneity acts as a selective template, allowing plants to respond evolutionarily to pollinator-mediated selection for large, showy flowers in areas of resource abundance but selecting against such responses in areas where resources are scarce (Figure 3).

• In this page
•  
• In a new window

• Download as PowerPoint Slide

Model showing how environmental forces may interact to influence the adaptive evolution of flower size. Possible (open arrows) and impossible (solid arrows) modes of flower evolution based on resource availability are indicated. Directions of arrows indicate increases (upward) or decreases (downward) in flower size. Evolution of small flowers may occur in patches of low (small ovals) or high (large ovals) resource availability. Pollinators exert selection for increased flower size in all patches, but such selection is constrained under low-resource conditions. Where high-resource conditions prevail during flowering, flower size may increase evolutionarily in response to pollinator-mediated selection.

Individuals may also adjust to differences in resource status through plastic changes in allocation to flowering. For example, flower size in alpine skypilot increases over the first several years of an individual's lifetime, as the plant grows in vegetative size (Galen 1996b). In other species, plants show reductions in corolla size after foliage is removed by herbivores (e.g., wild radish; Strauss et al. 1996). Such plastic adjustment of floral traits represents an alternative strategy for coping with resource heterogeneity that is distinct from the adaptive genetic response predicted by the resource-cost hypothesis.

A test of the resource-cost hypothesis must first elucidate whether the production or maintenance of large, showy flowers requires an increased investment of resources compared to investment in smaller, less conspicuous flowers. Although this assumption is intuitively appealing, evidence for it is often equivocal. For example, photosynthetic flower organs, including sepals, petals, and even pistils, contribute directly to energy for reproduction in several species (e.g., strawberry [Fragaria virginiana], Jurik 1985; orange [Citrus sinensis], Vu et al. 1985; petunia [Petunia bybrida], Weiss et al. 1990; buttercup [Ranunculus adoneus],Galen et al. 1993; white campion [Silene latifolia], Laporte and Delph 1996; and orchids [Spiranthes cernva], Antlfinger and Wendel 1997), reducing the demand for photosynthate from vegetative organs. In such species, increases in the photosynthetic capacity of flower parts could compensate for respiration by the larger attractive organs, if organ sizes scale allometrically (Galen et al. 1993). Nutrients allocated to flower tissues may be resorbed when organs senesce, again reducing the cost of floral structures on a whole-plant basis (Ashman 1994). The cost of investment in floral attractiveness is clearly “currency,” or resource, dependent and should be greatest for resources that can be neither manufactured in nor resorbed from flower parts. Water represents the quintessential resource of this kind. Not surprisingly, water loss through flower transpiration has been implicated as a major cost of reproduction in arid habitats (Nobel 1977).

In the alpine habitat of P. viscosum, water availability serves as a potent limiting factor for plant growth (Peterson and Billings 1982, Enquist and Ebersole 1994). To address whether large, showy flowers require more water over their lifetimes than smaller ones, I performed a simple potometer experiment (Galen et al. 1999). Immature flowering stems were excised from plants in the field and placed into test tubes filled with water to a constant volume. A stopper tightly fitted around each flowering stem minimized water loss aside from that drawn into the inflorescence. Subtending leaves were trimmed, as were the buds, to provide a constant number of five flowers on each stem. Cut pedicel surfaces were coated with petroleum jelly to minimize the effects of trimming on water loss.

Depletion of water in the tubes was then monitored under laboratory conditions for inflorescences during bud expansion and again at female receptivity, when most of flower growth has been completed. Linear regression was used to test whether the amount of water taken up by developing buds was related to the volume of the fully expanded corollas. Here, volume was used as a proxy for corolla size because organs grow in three dimensions. Corolla surface area was used in a similar regression analysis of water uptake by receptive flowers. (Corolla dimensions were analyzed because the corolla represents the attractive portion of the flower.) Corolla size was found to strongly influence water uptake by inflorescences, explaining approximately 60% of the variation in rates of water removal from tubes during each phase of flower development (Figure 4; Galen et al. 1999). These laboratory results suggest that increases in corolla size may incur physiological costs for skypilot plants flowering under dry conditions. However, for more meaningful comparisons of the cost of large versus small flowers, the water relations of intact plants should be studied in their natural environments.

• In this page
•  
• In a new window

• Download as PowerPoint Slide

Relationship of water uptake rate per flower to mature corolla size of Polemonium viscosum as measured for inflorescence stems inserted into sealed tubes of water under laboratory conditions in 1997. Water uptake rates at the time of bud expansion are plotted as a function of corolla volume (a), and water uptake rates for inflorescences composed of fully expanded flowers are plotted as a function of corolla surface area (b). Best-fit regression lines are shown (n = 7 plants, P < 0.05 for both stages of flower development). From Galen et al. (1999). These results suggest that plants with large flowers use more water during bud expansion and flower maintenance than plants with smaller flowers.

In the Rocky Mountains, precipitation over the course of the growing season follows a strong seasonal rhythm, with drought conditions prevailing during June and monsoon rains bringing extensive moisture in July and early August (Oberbauer and Billings 1981, Enquist and Ebersole 1994). Afternoon showers during the monsoon season reduce evapotranspiration and insolation, ameliorating water stress for alpine plants (Peterson and Billings 1982). At the field site on Pennsylvania Mountain, average precipitation during July is more than twice that in June (Western Regional Climate Center data from Leadville, Colorado, 1948–1997; National Climatic Data Center 1998). Flowering of skypilot plants on Pennsylvania Mountain takes place in early to mid-June in the krummholz (approximately 3500 m), whereas higher on the tundra slopes (approximately 3700–4000 m) flowering begins in July, concomitantly with seasonal monsoon rains.

Because plants in krummholz and tundra habitats differ in the synchrony of flower presentation with resource (water) availability, it is possible to test whether, as the resource-cost hypothesis predicts, the physiological cost of large, showy corollas depends on resource availability during flowering. To test this idea, I compared the water stress incurred in producing large, showy flowers for plants at low and high elevations. Plant leaf water potentials (ψ₁) can provide a sensitive index of the water stress associated with a given phase of growth or development. Leaf water potential reflects the affinity of water for leaf tissues. Under water stress, leaf water potentials become increasingly negative because the osmotic and physical properties of leaf cells enhance the tendency of water to remain within the leaf rather than diffusing to the atmosphere. If reproduction places a demand on the water balance of the plant, then leaf water potentials should drop (i.e., become more negative) during flowering or fruit set as water is diverted from vegetative to reproductive structures. As leaf water potential drops, stomatal closure occurs to maintain leaf turgor, and carbon assimilation ceases. For example, in the common alpine plant, Acomastylis rossii, rates of photosynthesis drop by 40% as leaf water potential declines from −0.8 MPa to −2.1 MPa (Bliss 1985).

By measuring leaf water potentials for krummholz P. viscosum flowering during the dry early portion of the summer and for tundra plants flowering during wetter midsummer conditions, it was possible to ascertain the extent to which water stress incurred by vegetative organs during flowering differs between the two habitats. As expected from seasonal precipitation patterns, results for 1997 and 1998 showed that, on average, leaf water potentials tended to be more negative for krummholz plants, which flowered in June, than tundra plants, which flowered in July (Figure 5; P < 0.08). However, the degree to which water stress differed between habitats depended strongly on the time of day (P ≤ 0.01).

• In this page
•  
• In a new window

• Download as PowerPoint Slide

Average leaf water potentials (ψ₁) for flowering Polemonium viscosum in June (krummholz) and July (tundra). Brackets show one standard deviation. In 1997 and 1998, 15 and 10 plants, respectively, were measured in each site both before dawn (solid bars) and at midday (hatched bars). In both years, differences in average leaf water potentials between sites were significant at midday (P < 0.0001; planned contrasts following analysis of variance) but not before dawn.

Before dawn, when leaf water potentials are at equilibrium with soil water potentials, little water stress was evident for krummholz or tundra plants over the 2-year study. In both years, mean pre-dawn water potentials were above −0.55 and did not differ significantly between sites (Figure 5). This result suggests that, despite seasonal rainfall differences, soils in both habitats supply water at similar rates to flowering skypilots. However, over the course of the morning, as leaf stomata opened to support photosynthesis, water potentials dropped to more negative values for plants flowering in the krummholz than those flowering on the tundra (Figure 5; P ≤ 0.0001). For example, in 1998, midday leaf water potentials averaged −0.86 ± 0.33 MPa (SD; n = 10) for krummholz plants in June and −0.61 ± 0.22 (SD; n = 10) for tundra plants in July.

This result is consistent with the idea that the monsoon precipitation pattern ameliorates water stress in skypilots mainly by increasing atmospheric humidity and reducing midday evaporative demand on vegetative tissues. Similar findings have been reported for a number of other Rocky Mountain plant species that co-occur with skypilots (Ehleringer and Miller 1975, Peterson and Billings 1982).

According to the resource-cost hypothesis, the same increase in corolla size should incur a greater physiological cost in the krummholz, where flowering plants are more water stressed, than on the higher tundra slopes. In 1998, I collected additional information to determine how flower size variation in each population affected the level of water stress experienced by vegetative organs. In the krummholz, midday leaf water potentials depended strongly on corolla surface area (Pearson's correlation coefficient, r = −0.85,n = 10, P < 0.002). The negative slope of this relationship (Figure 6) suggests that under dry conditions, the vegetative organs of plants diverting water to maintain large, showy flowers experience greater water stress than the vegetative organs of neighboring plants with smaller flowers. For tundra plants flowering after the onset of the monsoon rains, no such cost was revealed (Figure 6; r = 0.08, n = 10, ns). Other factors undoubtedly differ between the two habitats that could influence the relationship between flower size and water stress, including plant community composition and vegetation cover. However, seasonal drought provides the most parsimonious explanation for the negative relationship between flower size and plant water status in the krummholz.

• In this page
•  
• In a new window

• Download as PowerPoint Slide

Relationship of midday leaf water potential to corolla surface area for plants of Polemonium viscosum. Plants flowering in krummholz (a) and tundra (b) habitats were measured during the summer of 1998. Leaf water potential showed a significant relationship to corolla surface area in the krummholz habitat (regression line shown; P < 0.002, n = 10) but not in the tundra habitat (ns, n = 10).

These findings support one aspect of the resource-cost hypothesis: small, less conspicuous flowers indeed reduce the physiological stress associated with reproduction in times of resource shortage. The findings also shed new light on the smaller corolla size of P. viscosum near timberline: Under the dry conditions associated with the earlier flowering schedule of krummholz plants, smaller corollas may render plants less vulnerable, not only to flower enemies, but also to water stress during flowering (Figure 3). These advantages of reduced corolla size, coupled with the availability of alternative insect pollinators at low elevations, may help to offset the disadvantage that small-flowered skypilots incur in attracting pollinating bumblebees.

A definitive test of the resource-cost hypothesis must also assess the demographic consequences of producing large, showy flowers under resource-poor conditions. The hypothesis predicts that when resources are in low supply, stress associated with the production and maintenance of attractive flowers will bring about reductions in survival or reproduction. For example, Campbell (1997) found a negative genetic correlation between corolla width and survival rate in I. aggregata. This result suggests that a resource-based tradeoff between allocation to pollinator attraction and survival plays a role in the evolution of corolla size in this species.

In skypilots, the demographic consequences predicted by the resource-cost model could be studied by subjecting randomly chosen plants to either low or high resource availability during flowering and monitoring their future survival and reproduction. We are conducting a field experiment of this kind with skypilot plants, watering individuals differentially to approximate average rainfall schedules during either June (dry) or July (wet) on Pennsylvania Mountain. If the resource-cost hypothesis is correct, then the relationship between flower size and future fitness components should depend strongly on water availability, with a negative relationship manifest under the drought treatment. This prediction assumes that plants in low and high watering treatments are of a similar initial stage in terms of vegetative size and age. Otherwise, variation in flower size with stage could mask genetically based differences in flower size among individuals and limit the ability to detect a demographic cost of floral allocation.

Above the population scale, the resource-cost hypothesis for the maintenance of genetic variation in flower size is amenable to comparative approaches that contrast the floral traits of sister species or subspecies that occupy xeric and mesic habitats (i.e., eastern and western slopes of the Continental Divide). The hypothesis that resource heterogeneity provides a template mediating the efficacy of pollinator-mediated selection on flower size would be supported if lineages characterized by small, inconspicuous flowers were found to occupy habitats in which plants typically experience resource stress during flowering. In skypilot populations distributed longitudinally along the Rocky Mountains, flower size decreases from north to south along a gradient of increasing aridity from northern Colorado to Arizona (Galen et al. 1987). Similarly, populations of Clarkia xantiana (a wildflower of the southern Sierra Nevada) in hotter, more arid habitats have smaller flowers that are more likely to self-pollinate than those of populations in more mesic habitats (Eckhart et al. 1998, Runions and Geber 1998). Such an association of small flower size and drought is also found in a number of other annual plant species, in which it correlates with differences in mating system between populations of arid and mesic habitats (Guerrant 1989). Although other hypotheses cannot be ruled out, these geographic patterns suggest that among geographically isolated populations, the maintenance of variation in flower form and size is tied to resource heterogeneity.

It is not surprising that plant reproductive ecologists have, until recently, had a narrow view of floral evolution, focusing almost exclusively on the process of pollination. Pollination is, after all, an event unique to flowers and one that defines the flowering plants or angiosperms as a group. But focusing narrowly on sexual functions of flowers and on the role of pollinators in mediating sexual reproduction has two shortcomings. This focus overlooks both other important functions of floral organs and the physiological integration of flowers with the vegetative body of the plant.

Scientists studying the functional ecology of plant secondary chemistry came to a similar juncture 20 years ago, when the diversity of secondary chemicals was viewed only in relation to herbivore pressure (apparency theory; Feeny 1976, Rhoades and Cates 1976). Plant secondary chemistry was thought to reflect the ease with which plants could be found by herbivorous insects. Variation among habitats in the availability of resources such as carbon or nitrogen, which are the building blocks of secondary defenses, was ignored. Apparency theory nicely predicts interspecific variation in secondary chemistry in some plant communities but falls short of explaining patterns in others (Howe and Westley 1988). A sea change in that discipline occurred when Coley (1983) and Coley et al. (1985) published papers that emphasized theoretical and empirical connections between plant secondary chemistry and resource availability. In showing how plant defenses might evolve in relation to underlying resources as well as herbivore pressure, Coley's work provided a powerful springboard for understanding plant chemical diversification. Similarly, our understanding of floral evolution can benefit from a more holistic approach to flower function that incorporates not only plant-pollinator relationships but also resource requirements for flower development and maintenance.

Clearly, hypotheses about the functional significance of floral variation are not mutually exclusive; as illuminated by studies of P. viscosum, multiple ecological forces may act on flowers simultaneously. Pollinators, enemies, and resource heterogeneity appear to operate simultaneously in exerting selection on flower size and shape in this system and in at least one other system that has received comparable long-term study (I. aggregata; Brody 1992, Campbell 1997). Recently, O'Connell and Johnston (1998) also reported that the magnitude of selection on flower size from insect pollinators varies markedly among microhabitats in a population of the pink lady's slipper orchid, Cypripedium acaule. Although still few in number, these examples point to the need for a more pluralistic view of natural selection on floral traits.

There are promising signs that plant reproductive ecologists are embracing broader new perspectives: Recently, an issue of Ecology spotlighted connections between pollination and herbivory (Strauss and Armbruster 1997). Such studies may provide an ideal arena for addressing the resource-cost hypothesis because herbivores, by selectively reducing leaf area, essentially create resource (photosynthate) heterogeneity within plant populations. The abiotic environment is also gaining recognition as a source of selection on reproductive traits. For example, studies on the correlated evolution of physiological and reproductive characters are bringing new light to bear on the evolution of gender differences, no small part of which involves changes in flower morphology (Laporte and Delph 1996, Dawson et al. 1998, Dawson and Geber 1999).

In this article, I have suggested that resource availability may influence the efficacy of pollinator-mediated selection in bringing about increases in flower size. An alternative and little-explored way in which the abiotic environment may affect the evolution of flower size or shape is as a force of natural selection on the developmental processes that underlie the growth and differentiation of flowers. For example, small flowers may reflect a history of selection for abbreviated development time under growing seasons characterized by rapidly deteriorating environmental conditions (Geurrant 1989, Eck-hart et al. 1998, Runions and Geber 1998). By resolving how abiotic stress, as a source of selection on developmental processes, affects the evolution of flower shape and size, such new research may illuminate not only why, but how, flowers come to vary.