Mechanisms of hummingbird-mediated selection for flower width in Ipomopsis aggregata

Ecology, July, 1996 by Diane R. Campbell, Nickolas M. Waserm, Mary V. Price

INTRODUCTION

Angiosperm flowers exhibit remarkable diversity in form. Floral morphology often serves to distinguish species (Grant 1949), and appears to correlate well in many cases with the morphology of prominent flower visitors (e.g., Straw 1956, Pijl 1961, Grant and Grant 1965, Macior 1983). The apparent morphological "fit" between flowers and pollinators is thought by many workers to reflect natural selection to exclude ineffective pollinators and to increase pollen transfer by effective ones (Trelease 1881, Muller 1883, Grant and Grant 1968, Stebbins 1970, Macior 1971, Armbruster 1988).

Such selection might occur through two basic mechanisms. First, floral morphology can influence access to floral rewards for animals such as bees and birds, and the efficiency with which these animals can extract the rewards (Wolf and Hainsworth 1972, Inouye 1980, Harder 1986, Grant and Temeles 1992). Access and efficiency in turn can influence which animals choose to visit a flower and how often they do so (Waser and Price 1981, 1983, 1985, Pleasants and Waser 1985, Duffield et al. 1993). Second, floral morphology can influence how much conspecific pollen is applied to and removed from the bodies of visitors, and thus the amount of useful pollen transfer per visit (Paton 1990, Campbell et al. 1991, Stanton et al. 1991). These two mechanisms of selection on morphology, through visit frequency and pollen transfer per visit, are analogous to Herrera's (1987, 1989) "quantity" and "quality" components of overall pollinator "effectiveness" (for other views of quantity vs. quality in pollination services see Schemske 1983, Waser 1983, Campbell and Motten 1985, Ayre and Whelan 1989, Waser and Price 1990, Kunin 1993).

The potential for pollinator-mediated natural selection on floral morphology appears to be great. Quantitative variation in characters such as flower size and shape, placement of sex parts, and nectar secretion rate is commonly observed in natural plant populations (e.g., Clausen 1951, Schoen 1982, Waser and Price 1984, Campbell 1989, 1992, Galen 1989, Real and Rathcke 1991). In at least a few cases this variation is partly heritable (Schwaegerle and Levin 1991, Mitchell and Shaw 1993, Robertson et al. 1994, Stanton and Young 1994, 1996, Campbell 1996, Galen 1996), indicating the potential for evolutionary response to pollinator-mediated selection. Furthermore, recent studies have documented phenotypic selection on morphometric floral characters under natural conditions (Nilsson 1988, Campbell 1989, Galen 1989, Schemske and Horvitz 1989, Johnston 1991, Herrera 1993).

Despite this progress, we remain largely ignorant of the actual mechanisms of pollinator-mediated selection on floral characters. We know even less, in any critical sense, about the role of morphological fit between pollinator and flower, whether mediated by an effect on the quantity, or the quality, of pollination services. Our intent here is to explore these issues, using as our example of morphology the width of the flower corolla tube in scarlet gilia, lpomopsis aggregata (Polemoniaceae). Corolla width varies quantitatively in natural populations of this montane perennial herb, and Campbell (1989, 1991) found that the per-flower success in exporting pollen to surrounding plants increased with width. We subsequently proposed a number of possible mechanisms [ILLUSTRATION FOR FIGURE 1 OMITTED] that could contribute to this observed phenotypic selection (Campbell et al. 1991). We were able to show that wider flowers received more visits by hummingbirds in nature ("visits/flower" on the second line of Fig. 1; hereafter "visitation rate") and thus a greater quantity of pollination services. Another mechanism, apparently involving quality of pollination services, led to even stronger selection: hummingbirds exported more pollen to receptive stigmas from wide flowers than from narrow ones, following a single visit ("pollen exported/visit" on the second line of Fig. 1; hereafter "visit effectiveness"). Since corolla width correlates positively in these populations with corolla length, stamen length, and nectar production rate (Campbell 1996), these relationships could in principle result from direct selection or indirect selection of correlated characters. One prominent approach to separating these possibilities employs multiple regression (Lande and Arnold 1983). Using that approach we found that the apparent selection based on visitation rate can indeed be explained by phenotypic correlations with other traits, coupled with hummingbird preference for nectar-rich flowers. On the other hand, visit effectiveness appeared to be influenced only by corolla width and not by the correlated character of corolla length (Campbell et al. 1991, Campbell 1996). Furthermore, experiments designed to separate direct and indirect selection have failed to detect effects of stamen length and of nectar volume on the amount of pollen exported per visit (Mitchell and Waser 1992, Campbell et al. 1994). Thus the evidence to date suggests that the relationship between corolla width and visit effectiveness is a direct one.