SELECTION ON POLEMONIUM BRANDEGEEI (POLEMONIACEAE) FLOWERS UNDER HUMMINGBIRD POLLINATION: IN OPPOSITION, PARALLEL, OR INDEPENDENT OF SELECTION BY HAWKMOTHS?
2013; Oxford University Press; Volume: 67; Issue: 8 Linguagem: Inglês
10.1111/evo.12102
ISSN1558-5646
AutoresMason W. Kulbaba, Anne C. Worley,
Tópico(s)Ecology and Vegetation Dynamics Studies
ResumoEvolutionVolume 67, Issue 8 p. 2194-2206 ORIGINAL ARTICLEFree Access SELECTION ON POLEMONIUM BRANDEGEEI (POLEMONIACEAE) FLOWERS UNDER HUMMINGBIRD POLLINATION: IN OPPOSITION, PARALLEL, OR INDEPENDENT OF SELECTION BY HAWKMOTHS? Mason W. Kulbaba, Mason W. Kulbaba umkulbam@cc.umanitoba.ca Department of Biological Sciences, University of Manitoba, Winnipeg, Manitoba, R3T 2N2 CanadaSearch for more papers by this authorAnne C. Worley, Anne C. Worley Department of Biological Sciences, University of Manitoba, Winnipeg, Manitoba, R3T 2N2 CanadaSearch for more papers by this author Mason W. Kulbaba, Mason W. Kulbaba umkulbam@cc.umanitoba.ca Department of Biological Sciences, University of Manitoba, Winnipeg, Manitoba, R3T 2N2 CanadaSearch for more papers by this authorAnne C. Worley, Anne C. Worley Department of Biological Sciences, University of Manitoba, Winnipeg, Manitoba, R3T 2N2 CanadaSearch for more papers by this author First published: 11 March 2013 https://doi.org/10.1111/evo.12102Citations: 22AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract Particular floral phenotypes are often associated with specific groups of pollinators. However, flowering plants are often visited, and may be effectively pollinated by more than one type of animal. Therefore, a major outstanding question in floral biology asks: what is the nature of selection on floral traits when pollinators are diverse? This study examined how hummingbirds selected on the floral traits of Polemonium brandegeei, a species pollinated by both hummingbirds and hawkmoths. In array populations of P. brandegeei, we measured pollen movement, and female (seeds set) and male (seeds sired) fitness under hummingbird pollination. We then compared the patterns of selection by hummingbirds with our previous study examining selection by hawkmoths. We documented contrasting selection on sex organ positioning through female function, with hummingbirds selecting for stigmas exserted beyond the anthers and hawkmoths selecting for stigmas recessed below the anthers. Furthermore, hummingbirds selected for longer and wider corolla tubes, and hawkmoths selected for narrower corolla tubes. Therefore, contrasting selection by hawkmoths and hummingbirds may account for variation in sex organ arrangements and corolla dimensions in P. brandegeei. We documented how floral traits under selection by multiple pollinators can result in either an intermediate "compromise" between selective pressures (sex organs) or apparent specialization (corolla tube length) to one pollinator. Floral designs are complex phenotypes composed of multiple traits that cumulatively influence interactions with pollinators. Traditionally, individual floral traits have largely been viewed as a response to selection by a single pollinator, or functional group of pollinators (Stebbins 1970). However, many plant species are pollinated by two or more functional groups that may each contribute to selection on floral traits (Waser et al. 1996; Ollerton 2006). The response to selection by multiple groups of pollinators can vary, depending on the strength and nature of selection by each pollinator (Aigner 2001; Sahli and Conner 2011). Therefore, characterizing selection by each functional group of pollinator is an essential first step toward understanding net selection exerted on floral traits, and thus the evolution of floral design. There exist few independent estimates of selection on floral traits by the diverse pollinators of a single plant species. Even fewer studies have described both female and male fitness under multiple pollinators (but see Sahli and Conner 2011). Aigner (2001) developed optimality models that predicted how floral traits could respond to a single versus multiple pollinators. Generally, the response of floral traits to selection by multiple pollinators depended on the strength and targets of selection, and could involve trade-offs in fitness (Aigner 2001, 2004). A fitness trade-off may occur when an increase in fitness under one pollinator corresponds to a decrease in fitness under another pollinator (e.g., Muchhala 2007). More recently, Sahli and Conner (2011) described four potential responses of floral traits to variable selection by two distinct pollinators. First, if pollinators exert contrasting selection on a trait, the mean trait value may be an "intermediate" or "compromise" phenotype between the optimum of each of the two pollinators. Second, if pollinators select similarly on the same trait, the mean trait value may represent an optimum for both pollinators. Third, if selection on a floral trait is much stronger from one of the two pollinators, the mean value of a floral trait may represent a specialization towards that pollinator. This response may occur when a particular trait is more critical for pollination to one pollinator, or the trait optimum for the second pollinator covers a wider range of phenotypes. Finally, the simultaneous presence of both pollinators may change the nature of selection. Therefore, selection may be nonadditive such that the impact on the floral trait cannot be predicted by selection from each pollinator acting singly. These four outcomes may differ among floral traits so that flowers may have a combination of intermediate phenotypes and / or a mosaic of phenotypes apparently specialized to different pollinators (cf. Aigner 2001). The surprisingly few studies that have empirically examined selection on a single species by multiple pollinators found instances of the first three potential outcomes (Castellanos et al. 2004; Muchhala 2007; Sahli and Conner 2011). However, most of these studies examined selection on artificial flowers, or on a small number of floral traits (but see Sahli and Conner 2011). Additional studies are needed to develop a general picture of the frequency of each outcome. Temporal and spatial variation in the abundance of different pollinators can affect both trait means and genetic variation in floral traits (Forrest et al. 2011). For example, variation in the pollinator assemblage of Calathea ovandensis contributed to increased variation in corolla length (Schemske and Horvitz 1984). In contrast, parallel selection on a floral trait by multiple pollinators is expected to reduce the genetic variation in that trait. However, the prevalence of parallel versus contrasting selection by distinct pollinators is unknown (Sahli and Conner 2011). Regardless of pollinator diversity, the response of heritable traits to selection may be constrained by gender conflict or genetic correlations (Roff 1997). First, the majority of flowering plants are hermaphroditic, and can function as both maternal and paternal parents. As a consequence, contrasting selection through female and male functions has been predicted to be common in hermaphrodites (Morgan 1992). However, few studies have compared gender-specific selection on floral traits because male fitness (seeds sired) is difficult to assess and therefore seldom directly measured, especially in the presence of multiple pollinators. The limited evidence available suggests that conflicting selection among gender functions is rare (Ashman and Morgan 2004; Sahli and Conner 2011). Second, genetic correlations caused by pleiotropy can also constrain the response to selection, irrespective of gender function or differing targets of selection among pollinators (Conner, 2002; Delph et al. 2004). Hummingbirds and hawkmoths are important pollinators that have influenced the floral design of many North American plant species (Fægri and van der Pijl 1972; Grant 1983, 1985). A common feature of plant taxa pollinated by hummingbirds is the presence of a tubular corolla of similar length to the bird's bill (Grant and Grant 1968). The correspondence between the corolla and bird's bill lengths enforces contact between plant sex organs and the feathered regions of the bird's head where pollen is carried (Lertzman and Gass 1983). The exsertion of sex organs beyond the opening of the corolla tube further reinforces contact between sex organs and the hummingbird. In contrast, plants pollinated primarily by hawkmoths often display flowers with narrow corolla tubes (Nilsson 1988; Alexandersson and Johnson 2002; Brunet 2009; Kulbaba and Worley 2012). Furthermore, plants pollinated by hawkmoths may also display stigmas recessed below the anthers. The presentation of a recessed stigma is thought to increase the proximity of the female sex organs to the slender pollen-bearing moth proboscis (Webb and Lloyd 1986; Barrett 2002). Finally, plants pollinated by hummingbirds often produce large volumes of relatively dilute nectar compared to hawkmoth-pollinated plants (Baker 1972). However, the production of dilute nectar may have evolved more as a deterrent to bee visitation and overly vigorous bird visitation, rather than as a bird attractant (Cronk and Ojeda 2008). We examined selection by two major pollinators of the sub-alpine perennial Polemonium brandegeei. Natural populations of P. brandegeei occur along the Rocky Mountain of North America from New Mexico to southern Montana and appear as far east as south western South Dakota (Davidson 1950). The flowers of P. brandegeei display traits associated with hummingbird and hawkmoth pollination, and both animals are confirmed pollinators (Kulbaba and Worley 2008). Overall, P. brandegeei flowers exhibit tubular corollas of dimensions that are typical of hummingbird-pollinated taxa (Kulbaba and Worley 2008). However, the cream-white coloration of the corolla, and the strong, heavy sweet smell emitted by the flowers are traits associated with hawkmoth pollination (Grant 1983, 1985). The flowers of P. brandegeei display a high level of continuous and heritable variation in style length, resulting in a range of approach (stigmas exserted beyond anthers) to reverse (stigmas recessed below anthers) herkogamy (Kulbaba and Worley 2008). Here we describe selection on floral traits by hummingbirds in experimental populations, hereafter "arrays." We then compare these results to our previous study examining selection by hawkmoths on P. brandegeei floral design, where we documented selection for reverse herkogamy through female function, and selection for narrow corolla tubes and high nectar sugar concentration through male function (Kulbaba and Worley 2012). Measurement of hummingbird foraging behavior, pollen deposition, and removal, in addition to the number of seeds set and sired by each plant, allowed us to compare various indices of reproductive success. We explored the following specific hypotheses. (1) We predicted that increased corolla diameter (as a measure of flower size), or nectar reward (volume, sugar concentration), would result in more visits from hummingbirds and / or increased handling time of flowers. In contrast to floral rewards, we did not expect herkogamy or floral dimensions other than corolla diameter to affect hummingbird behavior, but we did expect these traits to influence female and male fitness through their effects on pollen deposition and removal, respectively. In particular, (2) we expected plants displaying flowers with approach herkogamy to receive more outcrossed pollen, and therefore set more seeds. Finally, (3) we predicted that plants displaying flowers with longer corolla tubes would have more pollen removed from their anthers, and sire more seeds than plants with shorter corolla tubes because the corolla tubes of P. brandegeei flowers are generally shorter than the hummingbird's bill (Kulbaba and Worley 2008). Materials and Methods GENERAL DESIGN To measure selection by hummingbirds on the floral traits of P. brandegeei, we used a similar approach to our previous study that measured selection by hawkmoths (Kulbaba and Worley 2012). We presented arrays of plants to the ruby-throated hummingbird (Archilochus colubris), and measured pollen movement (pollen removal and deposition), and selection gradients on floral traits through female (seeds set) and male (seeds sired) fitness. Although, A. colubris is not a confirmed pollinator of P. brandegeei, it is similar in behavior and morphology (most importantly bill length: 13.4–19.0 mm) to the confirmed hummingbird pollinator, Selasphorus platycercus (16.0–20.3 mm; Pyle 2001). We grew experimental plants from seed collected at Taylor Canyon, Colorado (39°34′33″N 104°22′26″W) under greenhouse conditions in 800 mL conical Deepots® with Premier Pro Mix® medium and Osmocote Plus® slow-release fertilizer (10 : 10: 10). A total of five arrays were used in our experiment. Each array consisted of 12–14 plants (total of 62 individuals) that were chosen based on their floral phenotypes. As one of our major goals was to examine how hummingbirds selected on variable stigma-anther separation (herkogamy), we ensured an equal number of approach and reverse herkogamous plants in each array. Plants were arrayed in a circular arrangement as indicated in figure 1 of Kulbaba and Worley (2012). Such an arrangement of plants included alternating approach and reverse herkogamous individuals, which provided an even distribution of herkogamous phenotypes and an equal probability of hummingbirds encountering either phenotype upon entering the array. Finally, we standardized all experimental inflorescences to seven freshly opened flowers that remained open for the duration of the experiment. The range of herkogamy and corolla trait variation in our arrays was similar to variation observed in natural populations (Kulbaba and Worley 2008). We conducted the experiments in two locations in rural Manitoba (approximately 174 km separation), to ensure that different birds were involved in the pollination of experimental plants. The first location was in an oak / aspen woodland north of Anola, Manitoba (49°56.355′N 96°36.734′W). Our second location was west of Overton, Manitoba (51°00.719′N 98°46.664′W), in another oak / aspen clearing. In May 2010, we presented ruby-throated hummingbirds with feeders consisting of 30% sucrose solution, to entice hummingbirds to our study sites. Feeders remained at the sites for the duration of the experimental period, but were emptied and thoroughly washed before arrays were assembled. Hummingbirds visited each array for a total of two consecutive days (approximately 7:00 a.m. to 8:30 p.m.). We presented the arrays to hummingbirds between 18 July 2010 and 27 August 2010 when weather and the availability of floral phenotypes permitted. The empty and cleaned feeders remained in the center of the arrays, because the hummingbirds had been habituated to their presence. An observer was present for the entire duration of each experiment, to discourage any nonbird visitors from foraging on the arrays. After dusk, the array plants were placed indoors to prevent nocturnal visitors from foraging. HUMMINGBIRD BEHAVIOR All hummingbird foraging bouts were recorded with a video camera to document foraging behavior on P. brandegeei flowers. For each foraging bout, we calculated the total number of flowers visited during the bout, the total number of plants visited, and the number of flowers visited per plant. We calculated handling time for visits to individual flowers by dividing the number of frames the visit lasted by the video frame rate (29 frames/sec). Mean handling times were calculated for individual plants by averaging the individual handling times of flowers on each array plant. We also examined the sequence of plants visited by hummingbirds, within foraging bouts. Finally, we examined the movement of hummingbirds within an inflorescence by comparing the number of upward versus downward movements within an inflorescence. The pattern of intraplant movement has been shown to affect the segregation of sex functions within inflorescences (Harder et al. 2004). FLORAL TRAITS, POLLEN MOVEMENT, AND FITNESS We measured floral dimensions and nectar properties (volume and sugar concentration) before birds visited each array. Two to three freshly opened flowers from each plant were destructively measured for floral traits and nectar properties before each array, while consistently maintaining seven open flowers per plant. Floral trait measurements and nectar properties were averaged for each experimental plant. Measured traits included: corolla width (distance from the tip of one petal to the opposing petal tip), corolla tube length, corolla tube width at the apex and base of the flower, sex organ separation (herkogamy), presentation of sex organs relative to the base of the flower, and flower mass (dry weight of corolla, calyx, androecium, and gynoecium). Herkogamy was measured as the distance between the stigma and the closest anther. Plants with recessed stigmas had a negative value of herkogamy and plants with exserted stigmas had positive values of herkogamy. See figure 2 of Kulbaba and Worley (2012) for precise location of measurements. Nectar sugar concentration was determined with the aid of a calibrated handheld refractometer (Fisher Scientific), with a temperature correction. We corrected our values with the conversion table in Kearns and Inouye (1993; Table 52, p. 172) because small volumes of nectar can introduce error when estimating the concentration of sucrose equivalents. Finally, 11 plants were reused in three replicate arrays, but flowers on different inflorescences were used. Therefore, we resampled the nectar properties to determine if nectar volume or sugar concentration changed, and obtained new floral measurements prior to assembling each array. Furthermore, we performed a repeated measures analysis of variance on the number of seeds set and seeds sired, to explore the consistency of female and male fitness in these plants. We measured pollen deposited on stigmas and pollen remaining in anthers of P. brandegeei plants after pollination by hummingbirds. We allowed 24 h in the greenhouse for fertilization, because a trial experiment with hand-pollinated plants determined this time period to be sufficient for ovule fertilization (Kulbaba and Worley 2012; M. Kulbaba, unpubl. data). Stigmas were removed from five of the experimental flowers and mounted on fuchsin jelly slides to stain pollen grains (Beattie 1971). We quantified pollen deposition by manually counting the pollen grains on the stigma from digital images (Image-pro express; Media Cybernetics Inc., Rockville, MD). As we were unable to distinguish between self and outcross pollen, pollen deposition measures reflected both. All five anthers from five of the experimental flowers were collected and stored in 70% ethanol. The number of pollen grains remaining in anthers after hummingbird visitation was assumed to indicate pollen removal. Pollen remaining in anthers was quantified with a Multisizer 3 particle counter (Beckman-Coulter, Fullerton, CA), and averaged over the five flowers per plant. Fitness was assessed through direct measures of female, male, and total fitness. Female fitness was the number of seeds set per plant, and male fitness was the number of seeds in an array sired per plant. Mature seeds from array plants were collected, and germinated under the same conditions as the experimental array plants. Offspring were raised to the seedling stage, when leaf tissue was collected and dried in silica gel. Whole DNA from up to five random offspring per individual array plant was extracted from dried leaf tissue, and genetically screened with six polymorphic microsatellite loci as described in Kulbaba and Worley (2011). We manually scored microsatellite profiles with genemapper 4.0 (Applied Biosystems, Foster City, CA), and then analyzed the profiles with the parentage program Cervus v 3.03 (Kalinowski et al. 2007). Cervus 3.03 determines the most likely paternal parent with a maximum-likelihood approach (Meagher 1986). Mendelian segregation probabilities were first simulated based on 10,000 cycles, complete sampling of 11–13 individuals (reflecting array size minus the known maternal parent, because P. brandegeei is self-sterile). The simulation included 0.750 as the proportion of loci typed, and a typing error of 0.01. We chose a strict confidence level of 0.95, and a relaxed level of 0.80 because we did not have to consider pollen outside the array populations (cf. Nishizawa et al. 2005; Hodgins and Barrett 2008). STATISTICAL ANALYSIS We examined associations between floral traits and (1) pollinator behavior, (2) pollen removal and deposition, and (3) female and male fitness. We used the same general analysis of covariance (ANCOVA; Proc GLM, Proc GENMOD) approach that we used in the hawkmoth selection study (Kulbaba and Worley 2012). Array number and site were included as fixed effects in all analyses. Hummingbird visitation behavior included handling time, number of flowers visited per plant, and the number of times a plant was visited over all foraging bouts ("plant visits"). These variables were each used as dependent variables with floral traits used as explanatory variables (covariates). We assessed female and male function by analyzing each of pollen deposition, pollen removal, number of seeds set and number of seeds sired per plant as dependent variables with floral traits, and variables describing visitation behavior as explanatory variables (covariates). Initial models included all two-way interactions between covariates and fixed effects, as well as two-way interactions among covariates. We removed nonsignificant interactions and covariates in a reverse step-wise manner (Sokal and Rohlf 1995). All analyses were performed in SAS 9.1.2 (SAS Institute 2004). Standardized selection gradients (β) were estimated as the coefficient between standardized fitness measures and floral traits. We standardized fitness measures (seeds set and seeds sired) across all arrays by dividing individual fitness measures by mean fitness (Lande and Arnold 1983). Finally, we calculated total relative fitness as the average of relative female and relative male fitness to prevent larger values of female fitness from overwhelming total fitness. We chose two techniques to standardize our estimates of selection gradients. First, we standardized the traits by dividing individual trait values by the mean trait value calculated over all arrays, such that the mean trait value was 1 (Hereford et al. 2004). We used this approach because we were particularly interested in how pollinators selected on a highly variable floral trait, stigma–anther separation (herkogamy) relative to other less variable traits (e.g., tube diameter). Second, to compare our results to the majority of other studies measuring selection on floral traits, we also employed the variance-standardized method of Lande and Arnold (1983). Trait values were standardized to a mean of 0 and a variance of 1 (Lande and Arnold 1983). Variance-standardized traits were calculated as the difference between individual trait values and trait means, divided by the trait standard deviation across all arrays. Linear (β) and quadratic (γ) selection gradients were examined for all explanatory floral traits, and were both included in the analyses. Slope estimates (b) are from analyses of transformed data, but selection gradients are reported from analyses of standardized but untransformed data, because estimates from transformed fitness measures can be biased (Lande and Arnold 1983). Finally, we compared the magnitude of selection gradients between female and male fitness, under a given pollinator with a Welch's t-test, assuming unequal variances. To test for potential multivariate or correlated selection, we analyzed nonlinear multidimensional fitness surfaces. We used the projected pursuit regression approach described in Schluter and Nychka (1994) to fit spline curves to our fitness data without making assumptions about the shape of the fitness function. Projected pursuit regression provides a more powerful technique to detect correlated selection than the multiple regression approach. Furthermore, determining the number of effective parameters of spline curves can objectively determine how linear or nonlinear a given fitness function is. We analyzed fitness surfaces for relative female, male, and total fitness. Results Our analyses provided the opportunity to detect both linear and nonlinear selection though analysis of covariance, and multivariate projected pursuit regression. All interaction and quadratic terms in the ANCOVA models were nonsignificant, and removed from the models. Therefore, the effects of covariates in the present analyses were linear, and independent of other covariates. These results were reflected in the projected pursuit regressions. Selection gradients from the two standardization techniques were very similar, with selection gradients from mean standardized traits being slightly smaller in magnitude than selection gradients estimated using variance standardized traits. We therefore present selection gradients from mean standardized analyses, as a conservative estimate of selection on floral traits. Furthermore, the effects of different sites (Overton and Anola) did not affect female (F1,52 = 0.28; P = 0.599) or male fitness (Wald's χ2 = 0.19; P = 0.665). Therefore, analyses presented below do not include site effects. POLLINATOR FORAGING BEHAVIOR Ruby-throated hummingbirds visited an average of 5.8 plants, 3.3 flowers per plant, and an average total of 17 flowers per foraging bout. Visits to individual flowers were very brief, with a mean handling time of 0.74 sec (range: 0.34–1.33 sec). Although all array plants were visited in each of the replicate arrays, a small fraction of the available plants were visited in each bout. We observed a total of 863 movements between flowers, within an inflorescence. A total of 405 movements occurred in an upward direction, whereas the remaining 458 proceeded in a downward direction; this did not correspond to a significant difference (χ2 = 2.48 for expectation of equal number of upward and downward movements P = 0.214). A summary of arrays and bird behaviors is presented in Appendix S1. The volume of nectar produced by P. brandegeei flowers affected two aspects of hummingbird foraging behavior. First, hummingbirds visited fewer flowers per plant on plants with larger volumes of nectar per flower. Second, hummingbirds spent longer handling flowers with larger volumes of nectar (Table 1), but this relationship was marginally nonsignificant. The effect of array was significant in both analyses, and may have resulted from different numbers of visiting birds and bird behaviors across arrays. Across all arrays, we observed an average of 35.8 foraging bouts, but the number of bouts ranged from 17 (Array 1) to 52 (Array 4; Appendix S1). Table 1. Analyses of Archilochus colubris handling time and the number of flowers visited per plant on arrays of Polemonium brandegeei. Handling time was log-transformed and the number of flowers visited per plant was square root transformed before analysis. Slope estimates b (SE) are in bold face. Initial models included all floral measurements and nonsignificant covariates were deleted using backwards elimination Effect Number of flowers visited per plant Handling time per flower Array F4, 56=10.85****P < 0.05, **P < 0.001, ***P < 0.0001, †P < 0.075. F4,56=9.54****P < 0.05, **P < 0.001, ***P < 0.0001, †P < 0.075. Nectar volume F1, 56=6.07***P < 0.05, **P < 0.001, ***P < 0.0001, †P < 0.075. F1, 56=3.78†*P < 0.05, **P < 0.001, ***P < 0.0001, †P < 0.075. b (SE) −0.337 (0.14) 0.012 (0.01) R2 of model 0.47 0.46 *P < 0.05, **P < 0.001, ***P < 0.0001, †P < 0.075. As we predicted, hummingbirds did not respond to variation in corolla tube dimensions or sex organ positioning. Unexpectedly, the sequence of plants visited by hummingbirds, and the total number of times an individual plant was visited did not vary with corolla diameter or nectar characteristics. FEMALE FUNCTION—POLLEN DEPOSITION AND SEEDS SET UNDER HUMMINGBIRD POLLINATION Basal-tube diameter (hereafter tube diameter) affected pollen deposition. Plants displaying flowers with relatively wide tubes had more pollen deposited on their stigmas after hummingbird visitation (Table 2; Fig. 1). The relationship between seed set and corolla-tube diameter was marginally nonsignificant (P = 0.057; Table 2). However, the relationship between tube diameter and total fitness was significant (see Total Fitness below). In addition, the number of pollen grains deposited on stigmas was significantly correlated with seed set (r = 0.266; P < 0.05). Finally, the plants that were reused in multiple arrays did not set significantly different numbers of seeds across the arrays, as determined through repeated measures analysis of variance (F2,20 = 0.03; P = 0.970). Table 2. Analyses of female function (pollen deposition and seed set) in arrays of Polemonium brandegeei after pollination by Archilochus colubris. Slope estimates, b (SE), are in bold and are from log-transformed data. Selection gradients, β (SE), are in italics and are based on untransformed data standardized to a mean of 1 (Hereford et al. 2004). Initial models included all floral measurements and visits per plant. Nonsignificant covariates were removed using backward elimination Effect Pollen deposition Seeds set Array F4, 56=0.17 F4, 56=9.74* Tube diameter F1, 56=30.86*
Referência(s)