RESEARCH ARTICLE

AMERICAN JOURNAL OF BOTANY

Floral function: eff ects of traits on pollinators, male and female pollination success, and female fi tness across three of milkweeds ( ) 1

R a ffi ca J. La Rosa 2,3,4 and Jeff rey K. Conner2

PREMISE OF THE STUDY: Central questions in reproductive ecology are whether the functions of fl oral traits in hermaphrodites create confl ict between sexes that could slow evolution, and whether individual fl oral traits function in pollinator attraction, effi ciency, or both. We studied how fl oral traits aff ect pollinator visitation and effi ciency, and how they aff ect male and female function and female fi tness within and across three Asclepias species that diff er in fl oral morphology.

METHODS: Using separate multiple regressions, we regressed pollen removal, deposition, and fruit number onto six fl oral traits. We also used path analyses integrating these variables with pollinator visitation data for two of the species to further explore fl oral function and its eff ects on fruit production.

KEY RESULTS: Most traits aff ected male pollination success only, and these eff ects often diff ered between species. The exception was increased slit length, which increased pollinia insertion in two of the species. There were no interspecifi c diff erences in the eff ects of the traits on female pollination success. All traits except horn reach aff ected pollination effi ciency in at least one species, and horn reach and two hood dimensions were the only traits to aff ect pol- linator attraction, but in just one species.

CONCLUSIONS: Traits tended to function in only one sex, and more traits aff ected function through pollinator effi ciency than through attraction. There was no signifi cant link between female pollination success and female fi tness in any of the three species; this pattern is consistent with fruit production not being limited by pollen deposition.

KEY WORDS fl oral function; natural selection gradient; path analysis; plant-pollinator interactions

Flowers of animal-pollinated function to manipulate polli- Most fl owering plants are hermaphroditic, so fl oral traits can nators into transporting pollen among fl owers. Floral traits attract function to increase pollen export (male function) and/or pollen pollinators (e.g., Mitchell, 1994; Conner and Rush, 1996; Schemske receipt (female function). Individual fl oral traits of hermaphroditic and Bradshaw, 1999 ; Johnson et al., 2003 ; Hansen et al., 2012 ), of- fl owers can act in harmony (Delph and Ashman, 2006) by aff ecting ten reward them (e.g., Real and Rathcke, 1991 ; Silva and Dean, male and female function in the same direction (common adaptive 2000 ), and promote effi cient pollen transfer (e.g., Nilsson et al., peaks; Sahli and Conner, 2011) or by aff ecting only one sexual 1987 ; Fulton and Hodges, 1999 ; Conner et al., 2009 ). If increased function (specialization). Alternatively, traits may be in confl ict or pollination success results in increased fi tness, selection will act on experience tradeoff s by aff ecting male and female function in op- those fl oral traits that infl uence pollination. posing directions. Functional confl ict between the sexes can slow the evolution of a trait, but in most species studied to date, trait 1 Manuscript received 8 September 2016; revision accepted 19 December 2016. function is in harmony or is specialized (Maad and Alexandersson, 2 W. K. Kellogg Biological Station, Department of Plant Biology, and Program in Ecology, 2004 ; Sahli and Conner, 2011 ; reviewed by Delph and Ashman, Evolutionary Biology and Behavior, Michigan State University, 3700 E. Gull Lake Dr., 2006 ). One way harmony can be achieved is through greater polli- Hickory Corners, Michigan 49060 nator visitation ( Delph and Ashman, 2006 ), which should increase 3 Current address: Department of Ecology and Evolutionary Biology, University of Colorado Boulder, Boulder, CO 80309 female function until pollen is no longer limiting and increase male 4 Author for correspondence (e-mail: raffi [email protected] ) function until no more unfertilized ovules are available in the pop- doi:10.3732/ajb.1600328 ulation. However, in most studies that have examined trait eff ects

150 • AMERICAN JOURNAL OF BOTANY 104 (1): 150 – 160 , 2017; http://www.amjbot.org/ © 2017 Botanical Society of America JANUARY 2017 , VOLUME 104 • LA ROSA AND CONNER—MILKWEED FLORAL TRAIT FUNCTION • 151

on both male and female function, data on natural pollinator visita- et al., 2006 ; Maad, 2007 ). Milkweeds also have a history of provid- tion rates were not included (but see Galen and Stanton, 1989; ing insight into the eff ects of fl ower number on male and female Mitchell, 1994 ; Gómez et al., 2006 ; Campbell et al., 2014 ). function ( Willson and Rathcke, 1974 ; Willson and Price, 1977 ), but Hypotheses of fl oral function can be tested by collecting data on there are fewer examples of fl oral trait function. pollinator behavior to tease apart fl oral traits into those that act to To our knowledge, only two studies ( Morgan and Schoen, 1997 ; attract pollinators (e.g., fl oral color, Stanton et al., 1986; corolla Caruso et al., 2005 ) have measured phenotypic selection on fl oral size, Fenster et al., 2006 ; fl oral scent, Raguso, 2008 ), and those that morphology in A. syriaca, using the pollination measures of polli- function to eff ectively transfer pollen to and from the pollinator narium removal and pollinium insertion as estimates of male and (e.g., stigma and anther exertion, Conner et al., 1995 ; anther color female fi tness, respectively; however, neither study considered the and position, Ushimaru et al., 2007 ; spur length, Sletvold et al., specifi c functions of individual traits for pollinator attraction or ef- 2010 ). Studies integrating traits, measures of pollination and fi t- fi ciency. Our study integrates data on fl oral traits, pollinator visita- ness, and pollinator visitation allow us to identify these relation- tion, pollen removal and deposition, and female fi tness (fruits ships ( Gómez et al., 2006 ). Traits that act to attract pollinators have produced) in three locally co-occurring species of milkweeds with in some cases been shown to similarly aff ect both sexes ( Baranzelli divergent pollinators and fl oral traits. et al., 2014 ), and traits aff ecting the effi ciency (i.e., proportion of Th e three species in our study, the closely related Asclepias syri- successful pollination per visit) of pollinators, such as fl oral tube aca and A. tuberosa L. (Fishbein et al., 2011), and A. viridifl ora Raf., length, can aff ect both sexes (e.g., Nilsson, 1988 ), or sometimes be share many of the same fl oral traits unique to milkweeds, but the specialized for one sex (e.g., Sahli and Conner, 2011; reviewed in fl owers of these species are diverse in size, shape, and color ( Fig. 2 ). Delph and Ashman, 2006 ). While they are oft en visited by honeybees, bumblebees, and wasps, Understanding how traits function in one species may enable the composition of pollinators diff ers between species, as shown in predictions about function in other species, especially with knowl- other studies ( Kephart, 1983 ; Betz et al., 1994 ; Fishbein and Venable, edge about trait distributions and selective agents in each species. 1996 ; Ivey et al., 2003; Stoepler et al., 2012). We measured six fl oral For example, predators may select similarly on insect wing dimen- traits that might interact with pollinators to aff ect pollination suc- sions or coloration ( Svensson and Friberg, 2007 ), or long tongued cess ( Fig. 2 ). All but one (gynostegium width) were measured in the nectar-feeders can select for convergence or divergence of fl oral two prior studies on A. syriaca (Morgan and Schoen, 1997; Caruso tube or nectar spur length ( Johnson and Steiner, 1997 ; Manning et al., 2005 ). and Goldblatt, 1997 ; Whittall and Hodges, 2007 ). In this paper, we Our study addresses the following three questions: (1) Are there address fl oral function across sexes in three species of milkweeds. confl icts between male and female fl oral trait function? (2) Are Milkweeds (Asclepias L.) are that have unusual her- there diff erences in function and selection on the same traits in maphroditic fl owers ( Fig. 1 ) and show convergence of fl oral struc- closely related species, and if there are diff erences, can they be ex- tures only with the distantly related orchid family (Orchidaceae). plained by selective agents and/or interspecifi c mean diff erences in Pollen grains are clustered into pollinia, which enable more precise the traits? (3) Do individual fl oral traits function more for attract- estimates of pollination success than is possible in most angio- ing pollinators or the effi ciency of pollen transfer? Seed production sperms with loose pollen. By scoring the number of pollinaria (at- of milkweeds is oft en resource limited ( Willson and Price, 1980 ; tached pairs of pollinia) removals and insertions, orchids and Caruso et al., 2005 ), which can reduce or eliminate the relationship milkweeds are ideal for studies of fl oral function through both between pollen deposition and female fi tness, weakening selection sexes. Orchids have been used to study the eff ects of traits and on fl oral traits that aff ect pollen deposition. By focusing our com- fl ower number on male and female function and female fi tness parison between sexes on pollination success, rather than seed set, (e.g., Nilsson, 1988 ; O’Connell and Johnston, 1998 ; Benitez-Vieyra we should be able to detect eff ects of traits on the pollination suc- cess of each sex to understand function. We predict that the traits that aff ect pollinator at- traction will infl uence male and female polli- nation success similarly, and that traits that aff ect pollination effi ciency will more oft en be specialized (aff ecting just one sex) or be in confl ict (Delph and Ashman, 2006). Th ese predictions broadly apply to all three Ascle- pias species, although pollinator composition, the extent to which populations are well adapted to their pollinators, and the current distribution of each trait will play an impor- tant role in the relationships between traits and fi tness components across species.

MATERIALS AND METHODS

Study system— Asclepias syriaca, A. tuberosa, FIGURE 1 Side and top view of the reproductive whorls of an Asclepias fl ower with fl oral struc- and A. viridifl ora can sometimes be found tures labeled (see fi gure key). The petals are excluded. within a single hectare, and may overlap in 152 • AMERICAN JOURNAL OF BOTANY

12.5 km. In 2008, we sampled 45 Asclepias syriaca plants at the Pond Laboratory Facility of Michigan State University’s Kel- logg Biological Station (42.410 ° N, 85.392 ° W). We selected individ- uals that were likely to be geneti- cally distinct by only choosing one ramet (stalk) per cluster of likely clones; Asclepias syriaca has the ability to spread vegeta- tively by rhizomes, and Kabat (2010) found genets (genetic in- dividuals) were made up of ra- mets that covered 1–30 m2 . We chose ramets from clusters that were separated by two or more meters and had fl oral traits that looked distinctively diff erent in shape and color; ramets within a cluster typically had visibly simi- lar fl oral characteristics (La Rosa, personal observation). We studied all 51 A. tuberosa plants in a clearing at Fort Custer Recreation Area (42.326 ° FIGURE 2 Top and side views of (A) Asclepias syriaca, (B) A. tuberosa, and (C) A. viridifl ora fl owers showing the six N, 85.331° W) in 2008, and fl oral traits measured: gy, gynostegium width; hl, hood length, hh, hood height; hr, horn reach; sl, slit length; sampled 212 out of more than gw, gap width. Note that hoods are greatly reduced and horns are absent in A. viridifl ora. Two and a half hoods one thousand A. viridifl ora plants were removed for the side views in (A) and (B) and two hoods were removed from the side view in (C). The pet- at the Fort Custer Training Cen- als are excluded. Bar = 2 mm. ter (42.306° N, 85.333° W) in 2009 by walking transects spaced their pollinator composition, despite their diverse fl oral size and 6 m apart, and fl agging the nearest plant within a 3 m radius ev- shape (Table 1, Fig. 2). We studied one naturally occurring popula- ery 6 m along each transect. Genets in these last two species are tion of each of the three perennial milkweed species in southwest well defi ned because they arise from a single root crown ( Wilbur, Michigan; the maximum distance between the populations was 1976 ).

TABLE 1. Descriptive statistics for the six fl oral traits (mm), display size, fl ower number, and fi tness estimates. Means and phenotypic coeffi cients of variation

(CVP = s.d./ ¯ x *100) were calculated for each species. The geometric mean of the fl oral traits (excluding gap width and horn reach) is an estimate of overall fl ower size. Relative size was calculated within each species by dividing each fl oral trait by the geometric mean. Flowers, pollinaria removals and insertions, and fruits were counted on a single ramet. Pearson product-moment correlations are given between pollinaria removals per fl ower and insertions per fl ower. Pollinium insertions per fruit were calculated for each ramet by dividing the number of pollinium insertions per fl ower sampled by the number of fruits per fl ower. The one-way ANOVA tests compare the relative sizes of each fl oral trait, and letter superscripts within a row are from Tukey HSD tests. Diff erent letters indicate values that are signifi cantly diff erent. The df = 2 for all ANOVA tests except for horn reach where df = 1. A. syriaca ( n = 42–45) A. tuberosa ( n = 40–46) A. viridifl ora ( n = 205–212) ANOVA

Variable Mean (SD) Rel. size CV P Mean (SD) Rel. size CV P Mean (SD) Rel. size CV P F-ratio Gynostegium width 2.19 (0.12) 0.76a 5.4 1.53 (0.05) 0.64b 3.5 2.74 (0.12) 1.67c 4.3 4000.5* Hood length 3.24 (0.42) 1.12a 12.9 2.58 (0.37) 0.98b 14.1 0.40 (0.09) 0.24c 23.1 5268.6* Hood height 5.08 (0.44) 1.76a 8.9 5.93 (0.36) 2.25b 6 4.74 (0.31) 2.89c 6.5 861.8* Horn reach 2.04 (0.34) 0.71 16.5 1.74 (0.25) 0.66 14.3 — — — 2.3 Slit length 1.90 (0.12) 0.66a 6.1 1.88 (0.08) 0.71b 4.2 1.40 (0.09) 0.85c 6.2 302.4* Gap width 0.61 (0.12) 0.21a 19.2 0.43 (0.09) 0.16b 21 0.62 (0.08) 0.38c 13.4 454.8* Geometric mean 2.88 a 2.64 b 1.64c 2592.2* Display size 109.51 (68.70) 62.7 66.04 (43.76) 66.3 110.46 (52.56) 47.6 Total fl ower number 189.03 (111.21) 58.8 182.27 (179.52) 98.5 185.11 (107.51) 58 Pollinaria removals per fl ower 2.02 (1.21)a 59.8 1.11 (0.67)b 60.2 2.32 (0.75)a 32.3 45.2* Pollinium insertions per fl ower 0.42 (0.23)a 56.0 0.18 (0.11)b 62.5 0.30 (0.17)c 57.2 21.7* Removal-insertion correlationr = 0.23* r = 0.06r = 0.37* Fruit number 5.95 (5.60)a 94.2 3.18 (2.27)b 71.6 3.42 (2.52)b 73.6 12.4* Pollinium insertions per fruit 24 (5.1) 14 (3.5) 20 (1.3)

* P < 0.0001. JANUARY 2017 , VOLUME 104 • LA ROSA AND CONNER—MILKWEED FLORAL TRAIT FUNCTION • 153

Trait measurements and fi tness components— In milkweeds, each fl ower has fi ve stigmatic openings in the gynostegium wall leading to a pair of ovaries ( Woodson, 1941 ). Th e pollinia are produced on either side of each stigmatic slit within the wall of the gynostegium ( Fig. 1 ). Each pair of pollinia is connected by a corpusculum, which has a small groove that catches the hairs or claws of pollinators,

resulting in passive removal. Eff ective pollination requires that a pollinium is inserted into another stigmatic slit for fertilization to owernumber Fruit occur ( Wyatt, 1978 ). Milkweeds also have a corona, consisting of a ora ring of hoods that surround the gynostegium. Th e hoods contain per fl Pollinia inserted Pollinia nectar, which is the only pollinator reward; the size of hoods varies A. viridifl greatly between species (Woodson, 1954). In many species, a horn forms within each hood, and sometimes protrudes from the hood, reaching over the gynostegium (Fig. 2A, B). We measured six fl oral ower traits: gynostegium width, hood length, hood height, horn reach, all the traits with explained by tness component

slit length, and gap width ( Fig. 2 ). per fl To measure fl oral traits, we photographed 2–3 fresh fl owers from each plant from the top and side (oft en with 2.5 hoods re- removed Pollinaria −0.02 (0.02) −0.06 (0.04) −0.01 (0.05) moved as in Fig. 2) using a Canon (New York, New York, USA) † digital single-lens refl ex (SLR) camera and 60 mm macro lens, with calipers set to 5 mm in each photo for calibration. We fi rst deter- mined landmarks that were identifi able across the range of fl oral variability ( Fig. 2 ), then used these landmarks to make calibrated −0.16 (0.11) −0.01 (0.02) −0.02 (0.04) 0.03 (0.05) linear measurements from the digital photographs using ImageJ 0.10 (0.11) 0.05 (0.02)* 0.03 (0.04) −0.01 (0.05) † † is the percent variance in that fi variance is the percent 2

R

version 1.49 ( Rasband, 1997–2012 ). Th e measurement error (see 0.37* 0.08** 0.05 0.14** † methods in Appendix S1; see Supplementary Data with this article) owernumber Fruit using landmarks for each trait was less than 5% for most traits ex- cept for gynostegium width in A. tuberosa (7%) and A. syriaca per fl Pollinia inserted Pollinia A. tuberosa (11%), and slit length in A. viridifl ora (41%). Among-plant varia- The species. tion was greater than within-plant variation (within-plant trait re- −0.16 (0.09) −0.16 peatability), except for gap width for two species (Appendix S2). † Annual fi tness component estimates were taken from one ramet Asclepias ower per genet, because it is diffi cult to identify entire genets in A. syri-

aca, and we sampled similarly across species. Th e A. viridifl ora gen- per fl

ets typically had only one ramet (mean = 1.19, s.d. = 0.43) and A. removed Pollinaria tuberosa oft en had several ramets in a single well-defi ned cluster 0.06 (0.10) −0.04 (0.09) −0.23 (0.11)* — — — (mean = 6.7, s.d. = 6.73, median = 5). Because fl owers open asyn- † chronously within umbels and remain continuously open for sev- eral days ( Wyatt, 1981 ; Kephart, 1987 ), we used a hand lens to score pollinium insertion and pollinarium removal on every fl ower from umbels that had all fl owers open and with some fl owers senescing, in three size oral traits and display

TABLE 2. Pollinator visits to Asclepias populations. Visits to A. syriaca were owernumber Fruit recorded from 9 hours and 29 min of video from 2008, and visits to A. tuberosa

were recorded from 8 hours and 10 min of video from 2008. Visitation to A. per fl A. syriaca viridifl ora was surveyed in the population in 2014 (see Methods). The mean inserted Pollinia (SEM) number of visits per minute for A. syriaca and A. tuberosa is calculated from 10 or more minutes of video of each focal ramet. Number of visits (%) ower PollinatorA. syriaca A. tuberosa A. viridifl ora A. mellifera (honey bee) 86 (64.7%) 74 (93.7%) per fl

Bombus spp. 5 (3.8%) 1 (1.3%) 30 (100%) removed Pollinaria Wasps 9 (6.8%) 0 < 0.10. P

Cantharidae sp. 17 (12.8%) 0 † Lepidoptera 6 (4.5%) 4 (5.1%) Diptera 6 (4.5%) 0 < 0.05, Hemiptera 4 (3.0%) 0 P of the whole model. cance fl (SE) for selection estimates gradient Standardized

Total number of visits 133 79 30 0.35* 0.10 0.40* 0.15 0.26 2 Mean no. visits per min 0.25 (0.28) 0.17 (0.24)† < 0.01, * P

† Trait TABLE 3. TABLE Gynostegium widthGynostegium −0.10 (0.09) −0.03 (0.10) 0.24 (0.14) −0.09 (0.10) −0.16 (0.10) the signifi Hood lengthHood height sizeDisplay 0.27 (0.10)*n R 0.28 (0.11)*Total 0.18 (0.11) 0.19 (0.12) 0.07 (0.09) −0.07 (0.17) −0.04 (0.17) −0.01 (0.10) 0.14 (0.09) −0.16 (0.09) 45 0.34 (0.14)* 0.12 (0.09) −0.06 (0.09) 0.07 (0.10) 45 0.13 (0.09) 0.01 (0.02) 0.17 (0.10) 42 0.00 (0.04) 0.07 (0.02)** 0.16 (0.05)** 0.05 (0.04) 51 0.19 (0.05)** 51 44 212 212 206 ** Slit lengthGap width −0.09 (0.09) −0.14 (0.09) −0.01 (0.10) −0.09 (0.09) 0.03 (0.14) −0.11 (0.14) 0.05 (0.09) 0.13 (0.10) 0.22 (0.09)* 0.07 (0.09) −0.14 (0.10) 0.17 (0.10) −0.01 (0.02) 0.09 (0.04)* 0.04 (0.05) Honey bee visits only. Horn reach −0.20 (0.12) −0.21 (0.13) 0.39 (0.20) 154 • AMERICAN JOURNAL OF BOTANY

maximizing the number of days during which fl owers within the fi tness data collection in 2009, that is, that this species was visited umbel were accessible to pollinators. To sample the most fl owers almost exclusively by bumblebees. per plant, we scored insertions and removals, and quantifi ed dis- play size of each ramet on the day when the maximum number of Data analysis— Each of the three fi tness component estimates umbels were mature. We scored an average of 100 (s.d. = 50.1) (male or female pollination success or female reproductive success) fl owers per A. syriaca ramet, 52 (s.d. = 17.3) per A. viridifl ora ra- was regressed separately on fl oral traits and display size using stan- met, and 58 (s.d. = 36.8) per A. tuberosa ramet; thus, we scored dard linear multiple regression. We also ran separate multiple re- pollination success for an average of 53% of all A. syriaca fl owers gressions including quadratic terms for each fl oral trait, but only and 28–32% of all fl owers on the other two species (cf. Table 1 ) for one was signifi cant (see Results) so we focus mainly on the linear a total of 18 413 fl owers scored in the entire study. terms. All of the independent variables and estimates of pollination Our estimates of plant size were display size and total fl owers success were continuous and normally distributed. Female repro- produced by one ramet. We measured display size as all of the ductive success was also continuous, because maximum fruit num- fl owers open on the ramet on the day we sampled, which included ber across species ranged from 10–28, but was oft en right skewed every fl ower sampled for pollinium insertions and pollinaria re- (i.e., there was a preponderance of small values). Log transforma- movals, plus all open fl owers from partially blooming umbels; tion produced a normal distribution, but did not qualitatively thus, display size was always equal to or greater than the number change the results; thus, we used untransformed fi tness data for our of fl owers sampled. We estimated total fl ower number per ramet analyses. Th e sample sizes for two of the species are small for the by counting every umbel at the end of the season and multiplying number of variables fi t in the regressions and path analyses (below), by the mean number of fl owers per sampled umbel for each but note that all individuals in the A. tuberosa patch, and nearly all individual. individuals in the A. syriaca patch, were included. Fitness compo- Our dependent variables included measures of pollination func- nent estimates were relativized by dividing by the mean, and the tion and female reproductive success. We estimated male and traits were standardized to a mean of zero and standard deviation female pollination success as the average number of pollinaria of one. removed per fl ower and the average number of pollinia inserted per We used the geometric mean of the four traits that represent a fl ower, respectively. Th e per-fl ower averages were continuous, rang- dimension of a fl oral structure (excluding horn reach and gap ing from 0–5, because each fl ower had a maximum of fi ve pollinaria width) to calculate an estimate of total size. We divided all of the and fi ve stigmatic slits into which a pollinium could be deposited. traits by the geometric mean to calculate a relative size of each trait, We estimated the total number of pollinaria removed or inserted then we compared these between species using one-way analyses of per ramet as the product of the average per-fl ower removals or in- variance (ANOVA), with post hoc Tukey Honest Signifi cant Dif- sertions and the total ramet fl ower number. We measured annual ference (HSD) tests when the ANOVA results indicated at least one female reproductive success as the number of mature fruits on each species was signifi cantly diff erent from the others. Comparisons of focal ramet. Fruits were counted in September, once they had ma- horn reach were between only two species because A. viridifl ora tured, because milkweeds can abort developing fruits using a late- does not have horns. Th e ANOVA analyses were performed using acting self-incompatibility system ( Gibbs, 2014 ). In all 203 fruits JMP, version 10.0 ( SAS, 2012 ). from 39 of the 45 A. syriaca plants, we found that fruit and seed To compare selection between species and between sexes, we number were highly correlated (r = 0.96, p < 0.001). Fruit and seed used analyses of covariance (ANCOVA). We compared the eff ects number were also highly correlated (r > 0.97) in two other Asclepias of traits on each sex by fi tting three ANCOVA models, one for species, A. exaltata L. and A. incarnata L. (La Rosa, 2015). Th ree to each species, with relative pollination success as the response seven plants from each species died or were damaged by deer or variable, fl oral traits and sex as main fi xed eff ects, plant ID as a rodents before the fruits could be counted, so they were excluded random eff ect, with a trait-by-sex interaction. We also fi t three from analyses that included fruit number. models with each relative fi tness component as a response variable, fl oral traits and species as main eff ects, and all trait-by-species in- Pollinator visitation— To relate fl oral traits to pollinator visitation, teractions. Because there could be no missing data, we fi t these we collected pollinator data from each A. syriaca and A. tuberosa latter three models without horn reach and again with only horn individual using high-defi nition Canon (VIXIA HF10, New York, reach. A signifi cant interaction term indicated that selection on New York, USA) digital video cameras. Ten or more minutes of video a trait diff ered between sexes or species, respectively. Regression was taken of all open fl owers on each focal ramet immediately aft er and ANCOVA analyses were performed using JMP, version 10.0 the pollinium insertions and removals had been counted. Th e order (SAS, 2012). that we visited plants on a given day was determined by plant ID Path analyses complement selection gradients by testing func- and not by any plant traits. For both A. syriaca and A. tuberosa , tional relationships between traits, selective agents, fi tness compo- videos were recorded between 10:00 and 15:00. From the video, we nents, and total fi tness, uncovering potential causes of selection recorded the number of pollinator visits, the duration of each visit, ( Conner, 1996 ) and correlations between sexes. We used a struc- and pollinator taxon to order (subdivided further for Hymenoptera tural equation modeling framework (Mitchell, 1992) to fi t path and Coleoptera). An insect taxon was considered to be a potential analyses that integrated fl oral traits, pollinator visitation (A. syriaca pollinator if at least one individual of that taxon had been captured and A. tuberosa only), male and female pollination success, and fe- or seen carrying pollinia. We did not collect pollinator data on A. male reproductive success. The pollinator data for A. syriaca viridifl ora in 2009, but visitation to this population was later ob- included a combination of several orders of insects, whereas A. served over 3.5 hours in 2014 by walking through the population tuberosa included only non-native honey bees—the dominant pol- and recording all insect visitors that had pollinia on their bodies. linator to this population—and a common, and sometimes domi- Results confi rmed our anecdotal observations during the trait and nant, pollinator of A. tuberosa in other states including Arizona, JANUARY 2017 , VOLUME 104 • LA ROSA AND CONNER—MILKWEED FLORAL TRAIT FUNCTION • 155

TABLE 4. ANCOVA results comparing the eff ect of traits on (A) sex (male and female pollination success) for the three species, Asclepias syriaca, A. tuberosa, and A. viridifl ora , and (B) species for three fi tness component measures. Trait main eff ects representing selection averaged over sexes or species were included in the model, but are not shown for simplicity. Trait-by-sex and trait-by-species interactions test for diff erences in selection among sexes and species, respectively. Values in bold are signifi cant at P < 0.05. A. syriaca A. tuberosa A. viridifl ora (A) Source dfF-ratio P-value dfF-ratio P-value df F-ratio P-value Gynostegium width × sex 1, 37 0.49 0.487 1, 43 0.44 0.510 1, 205 0.06 0.801 Hood length × sex 1, 37 0.57 0.455 1, 43 0.03 0.857 1, 205 0.15 0.704 Hood height × sex 1, 37 0.55 0.462 1, 43 0.00 0.992 1, 205 0.25 0.616 Horn reach × sex 1, 37 0.01 0.935 1, 43 0.75 0.391 — Slit length × sex 1, 37 0.54 0.469 1, 43 2.21 0.1441, 205 5.87 0.016 Gap width × sex 1, 37 0.31 0.582 1, 43 0.27 0.604 1, 205 1.18 0.278 Display size × sex 1, 37 0.66 0.423 1, 43 3.11 0.085 1, 205 0.53 0.466 Model n = 45 R2 = 0.68 n = 51 R2 = 0.51 n = 212 R2 = 0.55

Pollinaria removed per fl ower Pollinia inserted per fl ower Fruit number (B) Source dfF-ratio P-value dfF-ratio P-value df F-ratio P-value Gynostegium width × species 2 0.82 0.444 2 1.19 0.305 2 2.22 0.110 Hood length × species 2 4.30 0.015 2 1.07 0.346 2 0.53 0.587 Hood height × species 2 7.19 0.001 2 2.75 0.066 2 0.26 0.775 Horn reach × species 1 0.19 0.660 1 0.68 0.4121 17.48 <0.0001 Slit length × species 2 1.61 0.203 2 2.12 0.121 2 0.69 0.504 Gap width × species 2 4.66 0.010 2 1.46 0.235 2 1.68 0.189 Display size × species 2 2.14 0.119 2 0.92 0.400 2 1.00 0.368 Model fi t (excluding horn)n = 308 R2 = 0.16<0.01 n = 308 R2 = 0.09 0.11n = 292 R2 = 0.20<0.01 Model fi t (horn only)n = 96 R2 = 0.01 0.90n = 96 R2 = 0.02 0.54n = 86 R2 = 0.18<0.01

USA (Fishbein and Venable, 1996) and Wisconsin, USA (Betz χ 2 = 34.49, df = 25, P = 0.098, RMSEA = 0.04), indicating that they et al., 1994 ). were acceptable fi ts to the data. Th e full model for A. tuberosa had We fi t hypothesized causal relationships through simple paths a signifi cant χ 2 goodness-of-fi t value ( χ2 = 80.99, df = 36, P < 0.001, from each fl oral trait and display size to pollinaria removed per RMSEA = 0.18), but when we split the model into male-only and fl ower (male pollination success) and pollinia inserted per fl ower female-only models, both reduced models were nonsignifi cant (female pollination success), and through compound paths leading and fi t acceptably (χ 2 = 26.40, df = 17, P = 0.067, RMSEA = 0.122, to pollination success, but passing through the number of pollina- and χ 2 = 30.27, df = 23, P = 0.142, RMSEA = 0.092, respectively). tor visits and mean visit duration for A. syriaca and A. tuberosa . We Th e coeffi cients and their signifi cance did not qualitatively change hypothesized that increased display size in a single day would in- between the full model and the within-sex models, so we present crease total fl ower number, and that total fl ower number would the full model here for ease of comparison between species. Sam- directly infl uence the total number of pollinaria removed and pol- ple sizes for two of the species were small given the number of linia inserted on the ramet, as well as fruit number. Additionally, paths fi t, but the signifi cant partial regression coeffi cients re- we fi t a compound path from fl ower number to fruit number pass- mained signifi cant aft er reducing the models from six to four in- ing through the total number of pollinium insertions. We did not dependent variables, and the regression and path analysis results fi t paths between the fl oral traits and total fl ower number, because were consistent with each other. All path analyses were per- we found no trade off (negative correlation) between fl ower size formed in R v. 3.1.1 ( R Core Team, 2014 ) using the lavaan ( Rosseel, and total fl ower number. 2012 ) package. Initially, the model for each species had a signifi cant χ2 good- ness-of-fi t value, indicating they were not acceptable fi ts to our data and that something was likely missing from our model. To RESULTS remedy this, we used modifi cation index values, which calculate the decrease to the χ2 value if a path is added; this identifi es statis- Trait and pollinator comparisons between species— Asclepias tically infl uential pairwise relationships that may not have been syriaca had signifi cantly larger fl owers than A. tuberosa , and both included in our original model (Grace, 2006). From these values, were signifi cantly larger than A. viridifl ora ( Table 1 ; Fig. 2 ). Relative we chose to only add additional parameters that were biologically sizes (aft er correcting for fl ower size) of all fl oral traits except plausible, including a correlational (double-headed) arrow be- horn reach diff ered signifi cantly among the three species (Table tween the residual variation of total removals and insertions and 1). Correlations among the six fl oral traits were low and mostly causal (single-headed) arrows from gynostegium width, hood nonsignifi cant, especially in A. viridifl ora and A. tuberosa; the length, horn reach, and insertions per fl ower leading to fruit only correlations above 0.45 were between horn reach and the two number. Th ese were added to all three models, with the exception hood traits in A. syriaca (Appendix S3). Display size was not sig- of horn reach in the A. viridifl ora model. Two of the three fi nal nifi cantly correlated with the six fl oral traits, except for a signifi - path models had nonsignifi cant χ 2 goodness-of-fi t values ( A. syri- cant positive correlation with hood height in A. syriaca ( r = 0.34; aca , χ 2 = 47.69, df = 36, P = 0.092, RMSEA = 0.09; A. viridiflora , Appendix S3). 156 • AMERICAN JOURNAL OF BOTANY JANUARY 2017 , VOLUME 104 • LA ROSA AND CONNER—MILKWEED FLORAL TRAIT FUNCTION • 157

Asclepias tuberosa and A. viridifl ora were each visited almost ex- relative to the number of independent variables necessary when es- clusively by honeybees (Apis mellifera) and bumblebees (Bombus timating quadratic coeffi cients. spp. ) , respectively, whereas the pollinators of A. syriaca were more diverse ( Table 2 ). Still, two-thirds of the visitors to A. syriaca were Trait function: attraction and effi ciency— Increased hood height honeybees, with the next most common being soldier beetles (Can- and decreased horn reach signifi cantly increased the number of tharidae) at 13%. Bumblebee and honeybee visitors to these Asclepias pollinator visits to A. syriaca (Fig. 3A), and the increased visita- species consistently had pollinia on their bodies, perhaps because of tion led to increased pollinia insertions. However, there was no the abundance of hairs and spines and their tendency to frequently evidence that any of the individual fl oral traits functioned for pol- grip the fl owers between the hoods as they forage for nectar (La linator attraction in A. tuberosa . Instead, larger daily fl oral dis- Rosa, unpublished video observations). Th e other pollinators with plays increased both the number and duration of honeybee visits smoother appendages, such as fl ies and soldier beetles, tended to in A. tuberosa ; pollinator duration was also increased by larger carry fewer pollinia (La Rosa, personal observation) that were at- daily displays in A. syriaca. In all other cases, traits directly af- tached only to the ends of their tarsi, and never on their tibiae or fected pollinia insertions and removals, rather than aff ecting them femurs (La Rosa, unpublished data from A. incarnata ). through pollinator visitation, suggesting they were effi ciency traits (Fig. 3). Increased hood height signifi cantly increased removals in Trait eff ects on male and female function— Th e only signifi cant A. syriaca and A. viridifl ora and decreased removals in A. tuberosa diff erence between sexes was for slit length in A. viridifl ora (Ta- ( Fig. 3 ), and increased hood and slit length and decreased gynoste- bles 3 and 4A ), but even for this trait there was no evidence for gium width increased insertions in A. tuberosa (Fig. 3B). confl ict. Rather, increased slit length increased insertions, but had little eff ect on removals ( Table 3 ). Th e general pattern in the other Eff ects on fruit production— Each species produced a mean of about traits was for traits to aff ect the two sexual functions in the same 185 fl owers per ramet ( Table 1 ), for a potential of approximately 370 direction or only specializing through one sex, primarily through fruits, yet each ramet produced only three to six fruits on average, pollinarium removals ( Table 3 ). Th is lack of confl ict is consistent with <2% of their potential production. Fruit set did not appear to be lim- the moderately positive correlations between pollinaria removal ited by pollen receipt, because plants averaged 14–24 pollinia in- and pollinium insertions per fl ower in two of the species ( Table 1 ). serted for every fruit they produced ( Table 1 ). Nevertheless, pollinium insertions per fl ower had an eff ect on fruit number in A. viridifl ora — Interspecifi c diff erences in the eff ects of traits on pollination— they had no eff ect on fruit number in the other two species ( Fig. 3 ). Th ere were no signifi cant diff erences among the three species in Display size, on the other hand, had a positive eff ect on female fi tness how the traits aff ected female pollination success, but the eff ects of in all three species ( Tables 3 and 4B ). both hood dimensions and gap width diff ered signifi cantly for pollinarium removals ( Table 4B ). Pollinarium removal per fl ower increased with increasing hood height in A. syriaca and A. viridifl ora , DISCUSSION and decreasing hood height in A. tuberosa ( Table 3 ). Th is is consis- tent with A. tuberosa having the tallest hoods but the same or We found no confl ict of trait function between the sexes, consistent smaller sized pollinators than the other two species. Similarly, nar- with many previous studies on other taxa (e.g., Sahli and Conner, rower gaps increased removals in A. syriaca (which has wide gaps, 2011 ; reviewed by Delph and Ashman, 2006; but see Morgan and Table 1) and wider gaps increased removals in A. tuberosa (which Schoen, 1997 ). Th e only signifi cant trait by sex interaction was for has narrower gaps), suggesting that a trait size intermediate be- slit length in A. viridifl ora , which had a positive eff ect on insertions tween these two species functions best for honeybees, the predomi- but no eff ect on removals; a pattern also found by Morgan and nant pollinator in both species. Increased hood length increased Schoen (1997) in A. syriaca. Given that we have evidence that all pollinarium removal in A. syriaca and A. tuberosa ( Table 3 ), with but one of the traits we measured aff ect effi ciency of pollination, little to no eff ect in A. viridifl ora ( Table 3 ). Th is represents a diff er- this means that the lack of confl ict between sexes is not limited to ent scenario than hood height and gap width, because A. viridifl ora attraction traits as suggested by Delph and Ashman (2006) . already has the shortest hoods of the three species. Eff ects on male function diff ered signifi cantly between species Horn reach did not directly aff ect pollinarium removal or pol- for hood length, hood height, and gap width. If these functional linium insertion (Table 3; Fig. 3), but we found there was a signifi - relationships translate into male fi tness diff erences, our results sug- cant negative quadratic eff ect of horn reach on pollinia insertions gest the possibility of further divergence between species for hood only in A. tuberosa ( γ = −0.30, P = 0.04), suggesting that intermedi- length and convergent evolution for the other two traits. Measures ate values maximized insertions. We caution against putting too of selection on these traits based on number of seeds sired are nec- much emphasis on this result, however, given the low sample size essary to test these hypotheses.

← FIGURE 3 Path analyses relating traits, display size on a single day, pollinator visitation over 10 min, total fl ower number, and components of annual fruit production for (A) Asclepias syriaca , (B) A. tuberosa , and (C) A. viridifl ora . Blue arrows are paths infl uencing male function (pollinaria removals per fl ower), red arrows are paths infl uencing female function (pollinium insertions per fl ower) and fi tness (fruit number), black arrows aff ect both, and black arrows with no variable at their origin show residual variation unexplained by the model. Blue, red, and black arrows are P < 0.10 and gray are P ≥ 0.10. Single-headed arrows are hypothesized causal relationships and double-headed arrows are correlations. Arrow thickness represents the magnitude of the path coeffi cient, and dashed lines represent negative coeffi cients. Gray boxes are variables not included in the selection gradient analyses. Correlations between residual variation of endogenous variables are not shown for simplicity. *p < 0.05, ** P < 0.01, *** P < 0.001. 158 • AMERICAN JOURNAL OF BOTANY

We found limited support for the measured traits functioning to 1990 ). Because resource limitation is expected to infl uence male fi t- attract pollinators in the two species in which we collected pollina- ness less than female fi tness, the relative strength of selection on tor data; as is the case with many species, the main attractant may fl oral traits through male and female fi tness could be infl uenced by be display size (e.g., Willson and Bertin, 1979 ; Benitez-Vieyra et al., resource limitation as predicted by Bateman (1948). Bateman's hy- 2006 ). Increased hood height and decreased horn reach did in- pothesis has been tested in milkweeds (e.g., Queller, 1983 ; Caruso crease visitation to A. syriaca , so taller hoods may act as visual at- et al., 2005 ), because of the relative ease of estimating male fi tness tractants and shorter horns may interfere less with pollinators. It is through pollinaria removed and the large number of fl owers rela- interesting to note that increased visitation increased pollinarium tive to fruits set. However, without a clearer understanding of the insertions but not removals in this species. eff ect of pollinaria removed on fruits sired, it is diffi cult to make a We found much more evidence that the traits directly aff ected robust test of Bateman. In this study, we compared the functional male and female function, that is, they functioned as effi ciency eff ects of two sexes at the stage of pollination, but without male traits. Most studies of effi ciency traits in other species focus on fi tness (e.g., fruits sired or seeds sired), we are unable to make simi- traits like anther and stigma position and corolla tube or nectar lar comparisons between selection on traits through male and fe- spur length (e.g., Conner et al., 1995 ; Ushimaru et al., 2007 ; Sletvold male fi tness. Th is, however, is an important comparison to make in et al., 2010 ). Because milkweeds lack any analogs to these traits, it is future studies. We found that several fl oral traits function to in- diffi cult to compare our results to the broader literature. Still, all of crease pollen removal and deposition and are likely adaptive for the traits we measured would aff ect the position of pollinator body pollination success in these species. However, we have yet to un- parts, especially legs, relative to the position of the corpusculum cover what factors determine when pollination success infl uences and stigmatic slit, so it is not diffi cult to see how they might aff ect reproductive success to cause adaptive evolution of fl oral traits in pollinarium removals and pollinium insertions. Asclepias . Th e path analyses detected no positive eff ect of total pollinia in- In summary, we found no evidence for confl ict between sexes, serted on fruit number for any of the species (Fig. 3). In A. tuberosa, despite the fact that the traits we studied functioned more in polli- the strong path between total insertions and fruit number had a large nation effi ciency than attraction. In most of the traits examined, standard error and was therefore not statistically signifi cant; the sig- there were signifi cant diff erences among species in their function. nifi cant path in A. viridifl ora was negative. A disconnect between Additional studies integrating pollinator behavior, trait function, pollen received and fruit production suggests some combination of and male and female fi tness estimates will improve our under- three possible mechanisms. Resource limitation of fruit production, standing of fl oral adaptation. as Caruso et al. (2005) found in A. syriaca, can limit the number of fruits a plant can produce to fewer than the number of pollinia it re- ACKNOWLEDGEMENTS ceives. Incompatible self-pollinia, and ineff ective insertions can also aff ect this relationship. We discuss these two below, while noting that Th e authors thank I. Dworkin, J. Lau, and D. Schemske for their our data cannot distinguish between these three mechanisms. insightful comments, the reviewers for their thoughtful and In A. syriaca, Howard and Barrows (2014) found that 88% of pol- thorough comments, and D. Kilmartin, M. Lucke, and L. Porcher linia inserted by honeybees were selfed, and Pleasants (1991) calcu- for help in the fi eld. Th is research was funded in part by the National lated that 71% of the pollinia that honeybees inserted came from Science Foundation Graduate Research Fellowship Program, a within 1 m of the recipient fl ower, increasing the likelihood of self- Michigan State University Plant Sciences Fellowship, a grant from pollination in large plants with multiple ramets. Conversely, only Prairie Biotic Research, Inc., a T. Wayne and Katherine Porter about half of insertions by native bumblebees involved self-pollen Fellowship, a G. H. Lauff Research Award, the W. K. Kellogg (Howard and Barrows, 2014). Milkweeds employ late-acting abor- Foundation, and National Science Foundation Grant DBI-0638591. tion of developing fruits (Gibbs, 2014), and self-incompatibility can Th ey also thank the Michigan Department of Natural Resources, be one cause of fruit abortion, and can further draw resources from the Fort Custer Training Center, and the Kellogg Biological Station outcrossed fruits. With very low levels of self-compatibility in A. tu- for access to fi eld sites. Th is is KBS contribution no. 1978. berosa (Wyatt, 1976), large numbers of pollinia inserted would not necessarily translate to high fruit set, and could even have a negative DATA STORAGE eff ect (as seen in A. viridifl ora , Fig. 3C ). Pollinia that are ineff ectively inserted may prevent pollen tubes Data available from the Dryad Digital Repository: http://dx.doi. from reaching the ovaries. A pollinium has the advantage over org/10.5061/dryad.n81g1 (La Rosa and Conner, 2017). loose pollen of potentially delivering a saturating amount of pollen in a single pollinator visit (e.g., Bookman, 1984 ; Queller, 1985 ). LITERATURE CITED However, the pollen tubes grow from only one edge of the pollin- ium ( Wyatt, 1976 ), so if the pollinium is inserted backward or is not Baranzelli , M. C. , A. N. Sérsic , and A. A. Cocucci . 2014 . Th e search for Pleiades fully inserted, and the edge where the pollen tubes emerge is not in trait constellations: functional integration and phenotypic selection in enveloped in nectar, they will not germinate, preventing fertiliza- the complex fl owers of Morrenia brachystephana (). Journal of tion (Kevan et al., 1989; Sage and Williams, 1995). For those pol- Evolutionary Biology 27 : 724 – 736 . linia that are outcrossed and properly inserted, resource limitation Bateman , A. J. 1948 . Intra-sexual selection in Drosophila. Heredity 2 : 349 – 368 . Benitez-Vieyra , S. , A. M. Medina , and E. Glinos . 2006 . Pollinator-mediated se- can then weaken or eliminate the eff ect of pollinium insertions on lection on fl oral traits and size of fl oral display in Cyclopogon elatus , a sweat fruits, reducing selection on fl oral traits through female fi tness. bee-pollinated orchid. Functional Ecology 20 : 948 – 957 . We also know very little about the eff ect of pollinaria removals Betz , R. F. , R. D. Struven , J. E. Wall , and F. B. Heitler . 1994 . Insect pollina- on seeds sired, apart from one published study in A. exaltata , where tors of 12 milkweed (Asclepias ) species. Th irteenth North American Prairie the eff ect was positive, but weak (R 2 = 0.22; Broyles and Wyatt, Conference 45 – 60 . JANUARY 2017 , VOLUME 104 • LA ROSA AND CONNER—MILKWEED FLORAL TRAIT FUNCTION • 159

Bookman , S. S. 1984 . Evidence for selective fruit production in Asclepias. Kephart , S. R. 1983 . Th e partitioning of pollinators among three species of Evolution 38 : 72 – 86 . Asclepias. Ecology 64 : 120 – 133 . Broyles , S. B. , and R. Wyatt . 1990 . Paternity analysis in a natural population of Kephart , S. R. 1987 . Phenological variation in fl owering and fruiting of Asclepias exaltata : multiple paternity, functional gender, and the “Pollen- Asclepias. American Midland Naturalist 118 : 64 – 76 . Donation Hypothesis” . Evolution 44 : 1454 – 1468 . Kevan , P. G. , D. Eisikowitch , and B. Rathwell . 1989 . Th e role of nectar in the Campbell , D. R. , M. Forster , and M. Bischoff . 2014 . Selection of trait combina- germination of pollen in Asclepias syriaca L. Botanical Gazette 150 : 266 – 270 . tions through bee and fl y visitation to fl owers of foliosissimum . La Rosa , R. J. 2015 . Floral evolution in milkweeds: Evidence for selection past Journal of Evolutionary Biology 27 : 325 – 336 . and present. Ph.D. dissertation, Michigan State University, East Lansing, Caruso , C. M. , D. L. D. Remington , and K. E. Ostergren . 2005 . Variation in re- Michigan, USA. source limitation of plant reproduction infl uences natural selection on fl oral La Rosa , R. J. and J. K. Conner . 2016 . Data from: Floral function: eff ects of traits of Asclepias syriaca. Oecologia 146 : 68 – 76 . traits on pollinators, male and female pollination success, and female fi tness Conner , J. K. 1996 . Understanding natural selection: an approach integrating across three species of milkweeds (Asclepias). Dryad Digital Repository. selection gradients, multiplicative fi tness components, and path analysis. doi:10.5061/dryad.n81g1. Ethology Ecology and Evolution 8 : 387 – 397 . Maad , J. 2007 . Phenotypic selection in hawkmoth-pollinated Platanthera bifo- Conner , J. K. , R. Davis , and S. Rush . 1995 . Th e eff ect of wild radish fl oral mor- lia : targets and fi tness surfaces. Evolution 54 : 112 – 123 . phology on pollination effi ciency by four taxa of pollinators. Oecologia 104 : Maad , J. , and R. Alexandersson . 2004 . Variable selection in Platanthera bifo- 234 – 245 . lia (Orchidaceae): phenotypic selection diff ered between sex functions in a Conner , J. K. , and S. Rush . 1996 . Eff ects of fl ower size and number on pol- drought year. Journal of Evolutionary Biology 17 : 642 – 650 . linator visitation to wild radish, Raphanus raphanistrum. Oecologia 105 : Manning , J. C. , and P. Goldblatt . 1997 . Th e Moegistorhynchus longirostris 509 – 516 . ( Diptera: Nemestrinidae ) pollination guild: long-tubed fl owers and a spe- Conner , J. K. , H. F. Sahli , and K. Karoly . 2009 . Tests of adaptation: functional cialized long-proboscid fl y pollination system in southern Africa. Plant studies of pollen removal and estimates of natural selection on anther posi- Systematics and Evolution 206 : 51 – 69 . tion in wild radish. Annals of Botany 103 : 1547 – 1556 . Mitchell , R. J. 1992 . Testing evolutionary and ecological hypotheses using Delph , L. F. , and T.-L. Ashman . 2006 . Trait selection in fl owering plants: how path analysis and structural equation modelling. Functional Ecology 6 : does sexual selection contribute? Integrative and Comparative Biology 46 : 123 – 129 . 465 – 472 . Mitchell , R. J. 1994 . Eff ects of fl oral traits, pollinator visitation, and plant size Fenster , C. B. , G. Cheely , M. R. Dudash , and R. J. Reynolds . 2006 . Nectar on Ipomopsis aggregata fruit production. American Naturalist 143 : 870 – 889 . reward and advertisement in hummingbird-pollinated Silene virginica Morgan , M. T. , and D. Schoen . 1997 . Selection on reproductive characters: (Caryophyllaceae). American Journal of Botany 93 : 1800 – 1807 . fl oral morphology in Asclepias syriaca. Heredity 79 : 433 – 441 . Fishbein , M. , and D. L. Venable . 1996 . Diversity and temporal change in the Nilsson , L. A. 1988 . Th e evolution of fl owers with deep corolla tubes. Nature eff ective pollinators of Asclepias tuberosa. Ecology 77 : 1061 – 1073 . 334 : 147 – 149 . Fishbein , M. , D. Chuba , C. Ellison , R. J. Mason-Gamer , and S. P. Lynch . 2011 . Nilsson , L. A. , L. Johnsson , L. Ralison , and E. Randrianjohany . 1987 . Phylogenetic relationships of Asclepias (Apocynaceae) inferred from non- Angraecoid orchids and hawkmoths in central Madagascar: Specialized pol- coding chloroplast DNA sequences. Systematic Botany 36 : 1008 – 1023 . lination systems and generalist foragers. Biotropica 19 : 310 – 318 . Fulton , M. , and S. A. Hodges . 1999 . Floral isolation between Aquilegia formosa O’Connell , L. M. , and M. O. Johnston . 1998 . Male and female pollination suc- and Aquilegia pubescens . Proceedings. Biological Sciences 266 : 2247 – 2252 . cess in a deceptive orchid, a selection study. Ecology 79 : 1246 – 1260 . Galen , C. , and M. L. Stanton . 1989 . Bumble Bee Pollination and Floral Pleasants , J. M. 1991 . Evidence for short-distance dispersal of pollinia in Morphology: Factors Infl uencing Pollen Dispersal in the Alpine Sky Pilot, Asclepias syriaca L. Functional Ecology 5 : 75 – 82 . Polemonium viscosum (). American Journal of Botany 76 : Queller , D. C. 1983 . Sexual selection in a hermaphroditic plant. Nature 305 : 419 – 426 . 706 – 707 . Gibbs , P. E. 2014 . Late-acting self-incompatibility—the pariah breeding sys- Queller , D. C. 1985 . Proximate and ultimate causes of low fruit production in tem in fl owering plants. Th e New Phytologist 203 : 717 – 734 . Asclepias exaltata. Oikos 44 : 373 – 381 . Gómez , J. M. , F. Perfectti , and J. P. M. Camacho . 2006 . Natural selection on R Core Team . 2014 . R: A language and environment for statistical comput- Erysimum mediohispanicum fl ower shape: Insights into the evolution of zy- ing. R Foundation for Statistical Computing, Vienna, Austria. URL: http:// gomorphy. American Naturalist 168 : 531 – 545 . www.R-project.org/ . Grace , J. B. 2006 . Part IV Applications and illustrations. In Structural equation Raguso , R. A. 2008 . Wake up and smell the roses: Th e ecology and evolution of modeling and natural systems, 206–288. Cambridge University Press, New fl oral scent. Annual Review of Ecology Evolution and Systematics 39 : 549 – 569 . York, New York, USA. Rasband , W. S. 1997–2012 . ImageJ, U.S. National Institutes of Health, Hansen , D. M. , T. Van der Niet , and S. D. Johnson . 2012 . Floral signposts: Bethesda, Maryland, USA. URL: http://imagej.nih.gov/ij/ . testing the signifi cance of visual “nectar guides” for pollinator behaviour Real , L. A. , and B. J. Rathcke . 1991 . Individual variation in nectar production and plant fi tness. Proceedings of the Royal Society B: Biological Sciences 279 : and its eff ect on fi tness in Kalmia latifolia. Ecology 72 : 149 – 155 . 634 – 639 . Rosseel , Y. 2012 . lavaan: An R package for structural equation modeling. Howard , A. F. , and E. M. Barrows . 2014 . Self-pollination rate and fl oral-display Journal of Statistical Soft ware 48 : 1 – 36 . size in Asclepias syriaca (common milkweed) with regard to fl oral-visitor Sage , T. L. , and E. G. Williams . 1995 . Structure, ultrastructure, and histochem- taxa. BMC Evolutionary Biology 14 : 144 . istry of the pollen tube pathway in the milkweed Asclepias exaltata L. Sexual Ivey , C. T. , P. Martinez , and R. Wyatt . 2003 . Variation in pollinator eff ective- Plant Reproduction 8 : 257 – 265 . ness in swamp milkweed, Asclepias incarnata (Apocynaceae). American Sahli , H. F. , and J. K. Conner . 2011 . Testing for confl icting and nonadditive Journal of Botany 90 : 214 – 225 . selection: fl oral adaptation to multiple pollinators through male and female Johnson , S. D. , and K. E. Steiner . 1997 . Long-tongued fl y pollination and evo- fi tness. Evolution 65 : 1457 – 1473 . lution of fl oral spur length in the Disa draconis Complex (Orchidaceae). SAS . 2012 . JMP , Version 10.0. SAS Institute Inc., Cary, North Carolina, USA. Evolution 51 : 45 – 53 . Schemske , D. W. , and H. D. Bradshaw Jr . 1999 . Pollinator preference and the Johnson , S. D. , C. I. Peter , L. A. Nilsson , and J. Ågren . 2003 . Pollination success evolution of fl oral traits in monkeyfl owers (Mimulus ). Proceedings of the in a deceptive orchid is enhanced by co-occurring rewarding magnet plants. National Academy of Sciences, USA 96 : 11910 – 11915 . Ecology 84 : 2919 – 2927 . Silva , E. M. , and B. B. Dean . 2000 . Eff ect of nectar composition and nectar con- Kabat , S. M. 2010 . Genetic structure of the common milkweed, Asclepias syri- centration on honey bee (Hymenoptera: Apidae) visitations to hybrid onion aca L. Master's thesis, University of Michigan, Ann Arbor, Michigan, USA. fl owers. Journal of Economic Entomology 93 : 1216 – 1221 . 160 • AMERICAN JOURNAL OF BOTANY

Sletvold , N. , J. M. Grindeland , and J. Ågren . 2010 . Pollinator-mediated selec- Willson , M. F. , and R. Bertin . 1979 . Flower-visitors, nectar production, and infl o- tion on fl oral display, spur length and fl owering phenology in the deceptive rescence size of Asclepias syriaca. Canadian Journal of Botany 57 : 1380 – 1388 . orchid Dactylorhiza lapponica. Th e New Phytologist 188 : 385 – 392 . Willson , M. F. , and P. W. Price . 1977 . Th e evolution of infl orescence size in Stanton , M. L. , A. A. Snow , and S. N. Handel . 1986 . Floral Evolution: Asclepias (Asclepiadaceae). Evolution 31 : 495 – 511 . Attractiveness to Pollinators Increases Male Fitness. Science 232 : 1625 – 1627 . Willson , M. F. , and P. W. Price . 1980 . Resource limitation of fruit and seed pro- Stoepler , T. M. , A. Edge , A. Steel , R. L. O’Quinn , and M. Fishbein . 2012 . duction in some Asclepias species. Canadian Journal of Botany 58 : 2229 – 2233 . Differential pollinator effectiveness and importance in a milkweed Willson , M. F. , and B. J. Rathcke . 1974 . Adaptive design of the fl oral display in ( Asclepias, Apocynaceae) hybrid zone. American Journal of Botany 9 9 : Asclepias syriaca L. American Midland Naturalist 92 : 47 – 57 . 448 – 458 . Woodson , R. E. Jr . 1941 . Th e North American Asclepiadaceae. I. Perspective of Svensson , E. I. , and M. Friberg . 2007 . Selective predation on wing morphology the genera. Annals of the Missouri Botanical Garden 28 : 193 – 244 . in sympatric damselfl ies. American Naturalist 170 : 101 – 112 . Woodson , R. E. Jr . 1954 . Th e North American Species of Asclepias L. Annals of Ushimaru , A. , T. Watanabe , and K. Nakata . 2007 . Colored fl oral organs in- the Missouri Botanical Garden 41 : 1 – 211 . fl uence pollinator behavior and pollen transfer in Commelina communis Wyatt , R. 1976 . Pollination and fruit-set in Asclepias : A reappraisal. American (Commelinaceae). American Journal of Botany 94 : 249 – 258 . Journal of Botany 63 : 845 – 851 . Whittall , J. B. , and S. A. Hodges . 2007 . Pollinator shift s drive increasingly long Wyatt , R. 1978 . Experimental evidence concerning the role of the corpusculum nectar spurs in columbine fl owers. Nature 447 : 706 – 709 . in Asclepias pollination. Systematic Botany 3 : 313 – 321 . Wilbur , H. M. 1976 . Life history evolution in seven milkweeds of the genus Wyatt , R. 1981 . Th e reproductive biology of Asclepias tuberosa. II. Factors de- Asclepias. Journal of Ecology 64 : 223 – 240 . termining fruit-set. Th e New Phytologist 88 : 375 – 385 .