Geranium sylvaticum increases pollination probability by sexually dimorphic flowers

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Geranium sylvaticum increases pollination probability by sexually dimorphic flowers © 2022 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. Published version Soininen, Jaakko O. S.; Kytöviita, Minna‐Maarit Soininen, J. O. S., & Kytöviita, M. (2022). Geranium sylvaticum increases pollination probability by sexually dimorphic flowers. Ecology and Evolution, 12(12), Article e9670. https://doi.org/10.1002/ece3.9670 2022 Ecology and Evolution. 2022;12:e9670.   | 1 of 12 https://doi.org/10.1002/ece3.9670 www.ecolevol.org Received:25November2022 | Accepted:10December2022 DOI:10.1002/ece3.9670 RESEARCH ARTICLE Geranium sylvaticum increases pollination probability by sexually dimorphic flowers Jaakko O. S. Soininen | MinnaMaarit Kytöviita This is an open access article under the terms of the CreativeCommonsAttribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. ©2022TheAuthors.Ecology and Evolution published by John Wiley & Sons Ltd. DepartmentofBiologicaland Environmental Sciences, Faculty of Mathematics and Science, University of Jyväskylä, Jyväskylä, Finland Correspondence JaakkoO.S.Soininen,Departmentof BiologicalandEnvironmentalSciences, Faculty of Mathematics and Science, UniversityofJyväskylä,PObox35,FI- 40014, Jyväskylä, Finland. Email: [email protected] Funding information EmilAaltonenFoundation,Grant/Award Number: 220210 N Abstract Sexualdimorphismisexpressedasdifferentmorphologiesbetweenthesexesofa species.Dimorphismispronouncedingynodioeciouspopulationswhichconsistof female and hermaphrodite individuals. The small size of female flowers in gynodioeciousspeciesisoftenexplainedbyresourcere-allocationtoseedproductioninstead of large flowers. However, pollinator attraction is critical to female fitness, and factors otherthanresourcesavingsareneededtoexplainthesmallsizeoffemaleflowers.We hypothesized that the floral size dimorphism in the perennial gynodioecious Geranium sylvaticum (L.) is adaptive in terms of pollination. To test this “pollination hypothesis,” we video recorded the small female and large hermaphrodite G. sylvaticum flowers. We parameterized floral visitor behavior when visiting a flower and calculated pollination probabilities by a floral visitor as the probability of touching anther and stigma withthesamebodypart.Pollinationprobabilitydifferedintermsofflowersexand pollinatorspecies.Bumblebeeshadthehighestpollinationprobability.Thesmallfemale flowers were more likely to receive pollen via several pollinator groups than the large hermaphrodite flowers. The pollen display of hermaphrodites matched poorly withthestigmadisplayofhermaphrodites,butwellwiththatoffemales.Althoughthe smallsizeoffemaleflowersiscommonlyexplainedbyresourcere-allocation,weshow thatsexualdimorphisminflowersizemayincreasethemainreproductivefunctions ofthefemalesandhermaphrodites.Dimorphismincreasespollinationprobabilityin females and fathering probability of the hermaphrodites likely driving G. sylvaticum populations towards dioecy. KEYWORDS disruptive selection, flower size, Geranium sylvaticum,gynodioecy,pollination,sexual dimorphism TAXONOMY CLASSIFICATION Behaviouralecology,Botany,Entomology,Evolutionaryecology,Populationecology 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 2 of 12 | SOININEN and KYTÖVIITA 1 | INTRODUCTION Sexual dimorphism in plants refers to the morphological differences between sexes. Sexual differences in vegetative traits are usuallyabsent.Althoughsomeintrinsicdifferencesarepresentin theprimarysex organs,sexualdimorphismrefersto,andismost pronounced in, differences in flower or inflorescence morphology (Ashman, 2005). Sexual dimorphism is most notable in dioecious andgynodioeciousspecies.Indioeciousspecies,thetwosexesare expressedindifferentplantindividuals.Gynodioeciouspopulations consist of female individuals that bear flowers with only the female function, and hermaphrodite individuals with both the female and male function (Ågren & Willson, 1991; Eckhart & Chapin, 1997; Miller & Venable, 2003).Approximately6%ofangiospermsaredioecious (Renner & Ricklefs, 1995)andgynodioecyispresentin2.2%of angiospermfamilies,while0.5%ofdicotspeciesaregynodioecious (Godin&Demyanova,2013).Gynodioeciousspeciesareproposed to arise as cytoplasmic determinants followed by mutations that causethelossofthemalefunctioninhermaphroditeflowers(Budar et al., 2003; Schnable & Wise, 1998). Despitetheunderstandingofthemechanismsofhowgynodioecymayarise,itischallengingtoexplainformanyreasons.Theloss of male function entails that females lose half of the reproductive fitnessofthehermaphroditesassociatedwithpollen.Duetothisinborn disadvantage of the females, females must compensate for the lost half of their reproductive fitness in comparison to hermaphrodites (Lewis, 1941; Lloyd, 1976).Intheabsenceofalleviatingfactors, the increased contribution to the gene pool of the offspring should at least account for the fitness derived from pollen. The female compensation in fertility is usually less than the required compensation whichinsomecasescanbeexpectedashighas200%,althoughthe compensationdependsonthesexratioofthepopulationandthe mechanism of male sterility (Charlesworth & Charlesworth, 1978; Lewis, 1941). Increased seed viability and increased offspring fitness resulting from avoidance of inbreeding depression in females havebeensuggestedtoreducefemaledisadvantagefurther(Dufay &Billard,2012; Puterbaugh et al., 1997). Females may gain benefitsbyavoidinginbreedingdepression(Baker,1959; Charlesworth & Charlesworth, 1978; Sakai et al., 1997), but the resulting benefit is difficulttoevaluate.Assomegynodioeciousspeciesshowlittleinbreedingdepression(Mutikainen&Delph,1998), female advantage bycross-pollinationmaynotbeuniversal. The consequent loss of fitness along the male function is not the only problem posed by females in gynodioecious species. Fitness in females is critically dependent on pollinators visiting female flowers after visiting the pollen-bearing hermaphrodite flowers. As a generalrule,insexuallydimorphicspecies,thefemaleflowersare significantly smaller than those of the larger, showier hermaphrodite flowers (Ågren & Willson, 1991;Barret&Hough,2013; Miller & Venable, 2003). Female flowers may also provide less nectar to thepollinators(Delph&Lively,1992; Klinkhamer et al., 1991; Varga, Nuortila, & Kytöviita, 2013) and intrinsically lack pollen. Because pollinators strongly discriminate between flowers and prefer large andshowy(Bond&Maze,1999; Martin, 2004), symmetric flowers (Moller, 1995)withamplerewards(Delph&Lively,1992; Varga & Kytöviita, 2010), hermaphrodite flowers are predicted to be selected for these traits in promotion of their male function (Vaughton & Ramsey, 1998).Inlinewiththeshowinessandrewards,insectsvisit hermaphrodite flowers more frequently than those of the females in mostgynodioeciousspecies(Asikainen&Mutikainen,2005a; Cuevas et al., 2008; Van Etten & Chang, 2014; Varga & Kytöviita, 2010). Furthermore,manypollinatorsexhibitflowerconstancy,i.e.,behavior where the pollinator learns fidelity toward a specific rewarding plant species or morph (Waser, 1986). Flower constancy is proposed tobebasedonthehandlingskillsrequiredtoaccessrewards(Ishii & Kadoya, 2016),visualappearance(Gegear&Laverty,2005;Ishii & Masuda, 2014), and olfactory cues (Laska et al., 1999; Wright & Schiestl, 2009) of the flower that the pollinator learns to favor. Flower constancy is considered an important aspect of the evolutionary ecology of plant– pollinator interactions as it improves the pollination services received by the plant. For instance, it reduces the probability of clogging the stigma with the pollen of other species (Morales & Traveset, 2008; Muchhala & Thomson, 2012).On the other hand, it reduces the amount of wasted pollen in terms of transport to intraspecific recipient flowers (Schmid et al., 2016). The flower constancy and consequent potential passing over the females by the pollen carriers are aggravated by the fact that there areusuallyfewerfemalesinagynodioeciouspopulation(Asikainen & Mutikainen, 2003; Chang, 2006). This often leads to minority disadvantage (Levin, 1972) and females receive less visits by pollinators which mainly forage the most common morphs (Levin, 1972; Van Etten & Chang, 2014). Females cannot equal hermaphrodites in frequency (Charlesworth & Charlesworth, 1978) to escape minority disadvantage (Levin, 1972), but females could attract pollinators moreefficiently(Glaettli&Barrett,2008) and counteract the minority disadvantage by increased floral attraction. Furthermore, female flowers may compensate for smaller flower size by remaining in thereceptivephaselonger(Ashman&Stanton,1991).Despitethese potential counteractive measures, females have been frequently shown to receive fewer pollinator visits than hermaphrodites or males(Asikainen&Mutikainen,2005a;Bond&Maze,1999; Cuevas et al., 2008; Van Etten & Chang, 2014; Varga & Kytöviita, 2010) although not universally in all studies (e.g., Cervantes et al., 2018). In hermaphrodite flowers, the male function may pose different evolutionary selection pressures on floral morphology than the female function (Barret, 2002). Hermaphrodites are subject tothecostofincreasedinbreedingdepressionresultingfromself- pollination (Charlesworth & Charlesworth, 1987;Varga,Vega-Frutis, & Kytöviita, 2013).Arisingfromthedifferentevolutionarilystable strategies in thesexes,pollinator-limitedmalesare also proposed toallocateonfloraldisplayandreward(Thomson&Brunet,1990). In gynodioecious populations, hermaphrodites gain most of their fitness through the male function due to the presence of females (Charlesworth, 1981; Lloyd, 1976;Vamosi&Otto,2002). This should 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 3 of 12 SOININEN and KYTÖVIITA select for larger floral displays and pollen production in hermaphrodites (Vaughton & Ramsey, 1998) because the male function is promoted by pollen export and thus ultimately attractiveness to pollinators. Moststudiesexplainsexualflowersizedimorphismingynodioecious species by different aspects of resource allocation and trade-offs(e.g.,Ashman,1992, 1994; Delphetal.,1996; Miller & Venable, 2003). Seed production demands a substantial portion of plantresources(Ashman,1992).Forexample,Ashman(1992) found that Sidalcea oreganaplantsallowedtomakeseedsallocated20% less biomass to floral structures, and in turn, plants that were not, produced 40%morefloral biomassthe next yearthan the plants that were allowed to produce seeds the first year. The higher allocation in seed set in females vs. hermaphrodites has been suggested to be possible via enhanced resource allocation to female function (Ashman,1994; Chang, 2006). The decreased size of the corolla as well as the loss of stamens in females may leave more resources for seedproduction(Ashman,1994; Eckhart, 1992). We argue that the benefitgainedfromre-allocatingfloralbiomasstoseedmassisinadequategiventhatthesmallflowersizehandicapspollination(Bond& Maze, 1999; Martin, 2004).Itwouldbemoreeconomicfortheplant tore-allocateresourcestoseedsfromlesscriticalsourcessuchas older parts of foliage or roots rather than the critical floral display. Thedifferenceinflowersizebetweensexesisageneralphenomenon, and we propose that factors other than resource savings are neededtoexplaintheapparentmismatchbetweencostsandbenefits of the smaller flower size in females in gynodioecious plant populations. Inthiswork,weexploreanalternative,butnotnecessarilyexclusivehypothesistoexplainsexualdimorphism.WefocusonGeranium sylvaticum,agynodioeciousperennialplantwithsexuallydimorphic populations consisting of female and hermaphrodite individuals. The femaleflowersaresmallerthanthehermaphroditeones(Asikainen & Mutikainen, 2005a; Varga & Kytöviita, 2010), provide less nectar (Varga, Nuortila, & Kytöviita, 2013), and naturally no pollen as a reward for pollinators. The female flowers are visited less frequently by insect visitors (Asikainen & Mutikainen, 2005a; Varga & Kytöviita, 2010). We hypothesize that the small size of female flowers in G. sylvaticum is adaptive because it increases pollination probability in females and thus the fitness gained by female function in females and male function in hermaphrodites. We test this “pollination hypothesis” by comparing the probability of pollen transport fromanthertoreceptivestigma(I)betweenhermaphroditeflowers and(II)betweenhermaphroditeandfemaleflowers.Supportforthe hypothesisthatsexualdimorphismisadaptivewillbeevidencedif (I)issmallerthan(II).Furthermore,wecomparetheprobabilityof pollen transport from an anther to a stigma by the most common floral visitors of G. sylvaticum. We hypothesize that the small size of female flowers in G. sylvaticum is an adaptation to pollination bybumblebeesandexpectthatbumblebeesratherthantheother common floral visitors are responsible for pollen transport between flowers. Each insect visitor has characteristics that determine its specific pollination efficiency (Motten, 1986). These are how frequently and how faithfully the insect visits a given host, how much pollen it carries during visits, and how the visitor morphology and foragingbehaviormatchwiththeflowermorphology.Inthepresent work, we investigate the latter point related to visitor behavior and howitmatchesthemorphologyofthetwosexesofG. sylvaticum. 2 | MATERIALS AND METHODS 2.1 | Study organism Geranium sylvaticum(L.)isaself-compatibleperennialwithEurasian distribution (Stroh, 2014). Geranium sylvaticum is common in meadows but thrives also in shade (Korhonen et al., 2004), in particular when nutrient availability is high (Hokkanen, 2003). The plant is gynodioecious,andtheproportionoffemaleplantsvariesbetween0% and23%betweenpopulations(Asikainen&Mutikainen,2005a;M.- M.Kytöviita,personalobservations).Bothfemaleandhermaphrodite flowers offer nectar as a reward for pollinators (Varga, Nuortila, & Kytöviita, 2013).Thefruitmaturesin3 weeksafterfertilization andisaschizocarpwithfiveloculesandthemaximumnumberof seeds per fruit is five. 2.2 | Field measurements We estimated pollen transport probabilities by quantifying floral visitors and their behavior in detail in video recorded G. sylvaticumplants.Theplantsweregrowinginanexperimentalsiteofthe University of Jyväskylä established in an old field year 2008 at KonnevesiFinland(62°35′17.4″N26°14′03.2″E).Altogetherseven female plants and 13 hermaphrodite plants were video recorded when in full bloom between June 14 and 18, 2021. The plants were of the same age and size and were composed of 34 floral shoots on average.Alternatingbetweenrandomfemaleandhermaphrodite plants, a portion of the inflorescences were recorded on average in 40-min intervals. Multiple cameras ensured that the temporal variability in insect activity did not affect behavior in the sexes differently. The plants were video recorded during the most active period of insects(9:00 a.m.to6:00 p.m.).Theimagewasfocusedsothat10–15 fully open flowers in a plant could be followed simultaneously with a sufficient accuracy to distinguish the pollinator behavior (Figure 1). ThehardwareusedforrecordingincludedCanonEOS550Ddigital camerawith55–250 mmobjectivesetto250 mm,aswellasportable computer-runcameraswiththeuseoftheapplicationOBSstudio ver.26.1.1.(64bit).Altogether30 hofvideodataweregatheredon the20plantindividuals,ofwhichatotalof13 hweregatheredon the hermaphrodite plants, and 17 h on the female plants. Female plantswererecordedmoretocompensatefortheexpectedlesser visitation rates in females versus hermaphrodites. 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4 of 12 | SOININEN and KYTÖVIITA 2.3 | Visitation parametrization Theflower-visitinginsectswereassignedtosevengroups(hereafter visitor groups) which consisted of bumblebees in the genus Bombus (hereafter Bombus), honeybees Apis mellifera (L.) (hereafter Apis), hoverflies of the family Syrphidae (hereafter Syrphidae), and solitary Hymenoptera,Diptera,Hemiptera,andColeoptera.Syrphidaewere intentionallyseparatedfromDipteraingeneralduetotheirdistinct behavior and abundance, and Apis mellifera from other eusocial bees because Apis mellifera is farmed in Finland and does not occur naturally. The behavior of an insect was parameterized so that a contact with the reproductive surface of an anther or stigma in the flower was noted along with the body part of the insect that had made the contact. The body parts were classified as follows: headventral,headdorsal,foot,leg,thoraxventral,thoraxdorsal, abdomen ventral, and abdomen dorsal. We only report visitations where it was possible to distinguish the movements of an insect within a flower and whether it had contacted the floral reproductiveorgans.Inadditiontothemovementsoftheinsect, thestigmaphase(receptive/non-receptive)andthetimeofvisit were recorded. Visitation frequencies by insect group per hour wereextrapolatedbyfollowingtheflowersvisibleonthescreen forthelengthofthevideo.Thesedataarenotbasedonthesexual organ contact data as all visitations were usable to estimate the data on frequency. WhenthefruitswereripeinAugust,theschizocarpswerecollectedanddried(60°C,12 h).Basedontheseedscarsintheschizocarps, the average seed production per flower, total seed production perplant,andtheratioofundeveloped-to-developedschizocarps were counted. 2.4 | Data analysis Intheprobabilityestimationsandstatisticalanalyses,weonlyused the contacts with the ventral side of the insect's body (i.e., ventral sideofhead,thorax,andabdomen).ThisisbecauseG. sylvaticum flowers are sternotribic (Kozuharova, 2002), and dorsal contacts by theinsectswereineligible.Dorsalcontactswouldnottransmitpollen; although an insect could touch anthers with dorsal side, it could not land upside down on the stigma. Correspondingly, there were a few dorsal anther contacts (mainly with the head), but no dorsal stigma contacts in the video material. FIGURE 1 Examplesofthevideomaterialillustratingthebehaviorofdifferentinsectpollinatorsandthedifferentsexesoftheplant Geranium sylvaticum. 1. Apis melliferavisitinganon-receptivehermaphroditeflowertouchingtheantherswiththehead.2.ASyrphidae resting in a receptive hermaphrodite flower. The fly slips under the anthers and makes little contact with reproductive structures. 3. Bombus pratorumvisitingareceptivehermaphroditeflower.Thebeehasclimbedoverthereproductivestructuressothatthethoraxcontactsboth the anthers and the stigma. 4. Apis melliferavisitingafemaleflower.Duetothesmallsizeoftheflower,thebeereachesthenectariesover thereceptivestigmaandtouchesthestigmasurfaceswithventralsideofthethorax.5.ASyrphidaevisitingafemaleflowermakingcontact with the stigma while reaching the nectaries across the stigma. 6. Bombus soroeensisonafemaleflower.Duetothesmallsizeoftheflower, thebeehasplanteditselfovertheflowerforeasiestaccesstonectaries.Ittouchesthestigmawithitsventralsideofthorax. 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 5 of 12 SOININEN and KYTÖVIITA The probability of pollination was calculated and defined as the probability of an insect contacting an anther with a certain part of the body and contacting the stigma of a flower with the same body part. The probabilities of pollen transport between the compared groups were calculated with the basic formula of the probability of two independent events: whereP(A)istheprobabilityofanthercontacts(anthercontacts/visits) inthevisitorgroup,andP(B)istheprobabilityofcontactingstigmain the receptive phase (stigma contacts/visits) in the same visitor group. This method of calculating pollination probability focuses on insect behavior when visiting a flower but does not estimate the holistic pollinator efficiency (Motten, 1986). The probabilities of pollination were calculated separately in the twoplantsexes(femalesandhermaphrodites)andthedifferentpollinator groups, so that ventral contacts to anthers and stigma were takenintoaccount,respectively,forallbodyparts(head,thorax,and abdomen) and summed to get the final pollination likelihood. The data were statistically analyzed as follows. The data on anther or stigma contacts were analyzed with generalized linear regressionmodel(GLM,withthelogitlinkfunctionandbinomialfamily distribution).IntheGLM,thefrequencyofventralstigmaoranther contact(yes/no)wassetastheresponsevariableandtheplantsex (not in the model for anther contact) and visitor group as predictor variables. The visitor group analysis was repeated by setting each visitor group as the reference level to compare the visitor groups with eachother.Intheanalyses,visitsbyDiptera,Coleoptera,Hemiptera, and solitary Hymenoptera were not tested as there were not enough visits to draw reliable conclusions. For visitations per flower per hour frequency data, negative binomial generalized linear model with log link function was used, with visitation rate per flower per hour set astheresponsevariableandsexandthenumberoffloralshoots wereusedaspredictorterms.AnalysesforeffectsonmeanoroverallseedproductionwereconductedwithGLMusingtheGaussian family distribution with logit link function for the response variable, which was either the mean seed production per flower in a plant or total seed production. Models with the ratio of undeveloped to developed schizocarps as the response variable were conducted with quasibinomialdistributionfamilyandlogitlinkfunction.Aspredictor terms,visitationratesbyinsectgroupsperflowerinanhour,sex, and number of floral shoots were used, depending on the optimal modeldeterminedbyAICvaluesand/orthedistributionofresiduals. Within each insect group, the differences in frequencies in stigma contacts between plant sexes were analyzed with two- sample Z-testforprobabilitiesusingasubsetofdataatatimecontaining only one visitor group. To test the statistical significance of the differences in the pollination probabilities between the plant sexes, we resampled the data by randomly selecting pairs of anthercontact(0/1)andstigmacontact(0/1)iterated5000times.If there was a contact (1) on both anther and stigma, it was taken as a probable pollination event. These resampled data conformed to the aforementionedgroupingsothatallcombinationsofsex,bodypart, andvisitorclasswerepresent.Therandom-pairdatawereusedin two-sampleZ-testtoanalyzethedifferencesintheanther–stigma random frequencies of successful pollinations between the female and hermaphrodite groups in the respective body parts (head ventral,thoraxventral,andabdomenventral)andintherespectivevisitor groups (Syrphidae, Bombus and Apis). The data were analyzed using the statistical programming software R, ver. 4.0.2. (64 bit). 3 | RESULTS Altogether, we recorded 536 insect visits in the study plants. Of the visits, 406 were observed in flowers with receptive stigma. In G. sylvaticum flowers, the stigma lobes are closed when non- receptive and open in the receptive female phase (Varga, Nuortila, & Kytöviita, 2013).Therecordedinsectsbelongedtoarangeoftaxa: bumblebees (Bombus pratorum (L.), B. soroeensis (Fabricius), B. lucorum coll. (L.), B. sporadicus (Nylander), B. pascuorum (Scopoli), B. hypnorum (L.), B. terrestris (L.), B. bohemicus (Seidl), and B. lapidarius (L.)), honey bee (Apis mellifera (L.)), syrphid flies (e.g., Sphaerophoria scripta (L.), Syrphus ribesii (L.), Microdon sp., Cheilosia sp., and Helophilus pendulus (L.)), and solitary Hymenoptera (Lasioglossum sp. and Corynis obscura(F.))accompaniedwithvariousspeciesofDiptera;various Brachycera and some small Nematocera. At the site, occasionally Coleoptera such as Zacladus geranii (Paykull), Corizus hyoscyami (L.), Coreus marginatus (L.), Pentatoma rufipes (L.), and Dolycoris baccarum (L.) were observed as well as some butterflies (Vanessa atalanta (L.), Anthocharis cardamines (L.), Aglais io (L.), and Araschnia levana (L.)) and some moths (Geometridae). In terms of the plant sexes (hermaphrodite/female), we recorded 68/163 visits by Bombus, 63/88 by Syrphidae, 9/1 by Solitary Hymenoptera, 45/32 by Apis, and 14/15byDiptera,respectively.Itshouldbenotedthatthesedata are simply the recorded visitations that were distinguished in the videos, and do not represent differences in frequency of visitations betweensexes,butthenumberofvisitswevideorecorded(sizeof data) and on which the contact probability calculations are based on. 3.1 | Visitation frequency Pooling all of the insect groups, the female plants received 12.5 visits per flower per hour and hermaphrodites received 20.14 visits per hour. However, these overall visitation rates were not statistically significantlydifferentbetweenthesexes(df=45,AIC= 354.58, Estimate = 0.48, z = 1.26, p = .208).Visitationfrequenciesbythe three focal insect groups in females/hermaphrodites were 5.44/9.33 in Bombus (df = 45, AIC = 981.11, Estimate = 0.53, z = 4.88, p < .001),1.48/2.61inApis (df =45,AIC= 351.13, Estimate = 0.56, z = 2.69, p = .007),and3.8/4.08inSyrphidae(df=45,AIC= 218.43, Estimate = 0.07, z = 0.126, p = .899).VisitationratesbyBombus and P(A∩B)=P(A)×P(B) 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6 of 12 | SOININEN and KYTÖVIITA Apis, but not by Syrphidae, were statistically significantly smaller in females than in hermaphrodites. 3.2 | Anther contacts Overall,theanthersinhermaphroditeflowershad.34probabilityto be contacted during a visit. The probability to contact anthers with the ventral side of the body of an insect during a floral visit was .78 in Bombus, .07 in Apis, and .06 in Syrphidae. The probability to contact anthers with ventral side by members of Syrphidae did not differ from that of Apis (df =69,AIC= 52.1, Estimate = 0.48, z = 0.38, p = .70)butwaslowerthanthatofBombus (df =69,AIC= 52.1, Estimate = 4.36, z = 3.74, p < .01).Theprobability to contact anthers with ventral side by Apis was inferior to that by Bombus (df =69,AIC = 52.1, Estimate 3.88, z = 4.20, p < .01).Ventralanthercontacthierarchywasthusestablishedas Bombus > Syrphidae,Apis. The visitor probabilities to contact anthers are visualized in Figure 2. 3.3 | Stigma contacts The stigma contacts were influenced by plant sex. Females had an overall higher probability of receiving a contact to the receptive stigma by a floral visitor (p = .72)thanhermaphrodites(p = .16) (df =367,AIC= 331.75, Estimate = 2.50, z = 6.15, p < .01).The probability to contact a female/hermaphrodite receptive stigma with the ventral side in the main visitor groups was .87/.63 in Bombus (df =180,AIC= 154.25, Estimate = 1.37, z = 2.60, p = .01),.34/<.001 in Apis (df =58,AIC= 49.213, Estimate = 0.34, z = 3.76, p < .001), and .70/<.001 in Syrphidae (df =108,AIC= 120.98, Estimate = 0.7, z = 7.18, p < .001). Between visitor group comparisons revealed that in both sexes the stigma contact probability in the group Syrphidae was greater than in Apis (df = 367, AIC = 331.75, Estimate = −1.49, z = −3.59,p < .01)butlessthaninBombus (df =367,AIC= 331.75, Estimate = 2.94, z = −6.95,p < .01).Also,Apis had smaller probability to contact the stigma than Bombus (df =367,AIC= 331.75, Estimate = 2.94, z = 6.95, p < .01).Stigmacontacthierarchywasthus established as Bombus > Syrphidae > Apis. Visitor probabilities to make stigma contacts in female and hermaphrodite flowers are illustrated in Figure 3. 3.4 | Pollination probability The probability of transporting pollen from the anthers of a hermaphrodite plant to the stigma of hermaphrodite plant by a specific pollinator group during a single visit was .47 in Bombus and <.001 in the case of Apis and Syrphidae. The probability to make contact with the anthers of a hermaphrodite plant and then make a contact with the stigma in a female plant was .68 in terms of Bombus, .04 of Syrphidae, FIGURE 2 Averageanthercontact probabilities by different visitor groups (Bombus, Apis, and Syrphidae) in the hermaphrodite Geranium sylvaticum flowers (N =199).Anthercontactsare binary (yes, no), the contacts represent ventralthoraxcontactsonly. 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 7 of 12 SOININEN and KYTÖVIITA and .017 of Apis.Thedifferencebetweentheplantsexeswasstatistically significant in Bombusheadventral,thoraxventral,andabdomen ventral pollination probabilities (Table 1). Similarly, the pollination probability with Apisheadventral,thoraxventral,andabdomenventraldifferedstatisticallysignificantlybetweenthesexes(Table 1).In Syrphidae, pollination probabilities differed significantly between thesexesinheadventralandthoraxventralcontacts(nodatawere recorded on abdomen ventral contacts; Table 1).Ingeneral,thehermaphrodite plants had inferior probability to be pollinated by any pollinator group compared to that of females. The statistical significances and test values between the different visitor groups and body parts in female/hermaphrodite plants are shown in Table 1. Withoutsexdiscrimination,thelikelihoodtocontactanthers and then any stigma was .64 in Bombus, .04 in Syrphidae, and .01 in Apis. The possibility for autogamous pollination occurred only in the visitor group Bombus.In16.2%ofvisits,Bombus touched both the anthers and the stigma with the same body part, but taking the insect behavior within the flower into account it was estimated that only5.4%ofBombus visits in the hermaphrodites could have potentially led to autogamous pollination. According to the calculated probability values, the pollination probabilitiesinbothplantsexesrankasBombus > Syrphidae> Apis in the three main visitor groups. FIGURE 3 Averagestigmacontact probabilities by the different visitor groups (Bombus, Apis, and Syrphidae) in the hermaphrodite and female Geranium sylvaticum flowers. Stigma contacts are binary (yes, no), the contacts represent ventralthoraxcontactsonly. TABLE 1 Statistics,probabilities,andp-valuesfromatwo-sampleZ-testforequalityofproportionstestdepictingthecomparisons between the probability to pollinate female (F) versus hermaphrodite (H) Geranium sylvaticum flowers in different visitor groups (rows) and their respective ventral side body parts (columns). VISITOR CLASS Head Thorax Abdomen Bombus, F vs. H plants χ2 = 807.48,df= 1 p(F) = .15,p(H) < .001 p < .01 χ2 = 252.96,df= 1 p(F) = .64,p(H) = .49 p < .01 χ2 = 9.1,df= 1 p(F) < .001p(H) = .002 p < .01 Apis, F vs. H plants χ2 = 2,df= 1 p(F) < .001,p(H) < .001 p = .13 χ2 = 58.36,df= 1 p(F) = .012,p(H) < .001 p < .01 χ2 = 13.08,df= 1 p(F) = .003,p(H) < .001 p < .01 Syrphid, F vs. H plants χ2 = 5.15,df= 1 p(F) = .001,p(H) < .001 p = .02 χ2 = 168.84,df= 1 p(F) = .34,p(H) < .001 p < .01 NA Note: Syrphidae abdomen ventral contact probabilities were not comparable as no contacts occurred within this group. 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8 of 12 | SOININEN and KYTÖVIITA 3.5 | Seed production Hermaphroditesproducedonaverage1605 ± 996seedsperplant and females 2654 ± 402. The difference between the sexes was statistically significant (df =12,AIC= 240.28, Estimate =−758.44, t = −2.16,p = .05).Femalesproducedstatisticallysignificantlymore seedsalsoperflower(3.51 ± 0.47seedsperflower)thanthehermaphrodites(1.81 ± 0.40seedsperflower)(df=12,AIC= 23.12, Estimate = −1.66, t = −0.10, p < .0001).Sex affected the ratio of wilted flowers to schizocarps (distinguished from flowers by missing seeds or swollen ovaries and elongated stigma), between females (6.8%ofallflowersdidnotdevelopintoschizocarps)andhermaphrodites(26%ofallflowersdidnotdevelop)(df= 12, Estimate = 1.45, t = 2.76, p = .02). Bumblebeevisitationrateperflowerperhourwaspositivelyrelated to the mean number of seeds produced per flower in a plant (df =10,AIC= 18.9, Estimate = 0.02, t = 2.29, p = .05).Visitation rate by Syrphidae was also positively related to the mean production of seeds per flower (df = 10, AIC = 19.03, Estimate = 0.03, t = 0.01, p = .05),butApis visitation rate did not have any statistically significant relationship (df =10,AIC= 24.123, Estimate = 0.01, t = 0.70, p = .50).Thebumblebeevisitationratereducedtheratio of undeveloped flowers to developed schizocarps marginally significantly (df = 10, Estimate = −0.04, t = −2.34, p = .04). Also, Syrphidae visitation ratio had a statistically significant negative effect on the ratio of undeveloped flowers to developed schizocarps (df = 10, Estimate =−0.05,t = −3.0,p = .01),butApis had no effectontheundeveloped-to-developedschizocarpsratio(df= 10, Estimate =−0.03,t = −1.4,p = .18). 4 | DISCUSSION Althoughsexualdimorphismhasariseninseveraldistinctgenera (Miller & Venable, 2003;Thomson& Brunet,1990), the underlying evolutionary effectors are not clear (Charlesworth, 1981; Delphetal.,1996;Thomson&Brunet,1990).Toexplaindimorphismbetweenthesexes,argumentsfornon-adaptive(reviewed by Delph et al., 1996), anti-selfing (Baker, 1959; Kawagoe & Suzuki, 2003; Wilmer, 2011), and resource allocation hypotheses (Ashman,1994; Chang, 2006;Delphetal.,1996; Eckhart, 1992) have been forwarded. The morphology of a flower is a compromisebetweendifferentselectionpressures(Galen,1999). Larger flowers often receive higher visitation rates and have been proposed to evolve due to directional selection promoting increased floral attraction (Martin, 2004; Stanton & Preston, 1988). Visitation rates have been frequently shown to be positively linkedwithflowersize(Bond&Maze,1999; Martin, 2004; Van Etten & Chang, 2014). However, the reverse has not been documented previously: how small, visually unattractive flowers could make up for the loss of visitation rates. Thesizeofthesexorgansintheflowerplaysacrucialrolein the pollination probability. Due to developmental constraints, corolla size in a flower increases in size in symmetry with the other partsofaflower(Moyroyd&Glover,2017; Paterno et al., 2020). In G. sylvaticum flowers, the smaller petal size is associated with smallerstylelengthandlargerpetalswithlongerstyles(Asikainen & Mutikainen, 2005a). When the stigma is in the receptive phase, the style is typically longer in hermaphrodite flowers than in females (Asikainen&Mutikainen,2005a). The long style length in hermaphrodite flowers has positive and negative effects on reproduction. The hermaphrodite stigma in the receptive phase protrudes over the anthers. This has positive effects as a means of prevention of autogamy in Geranium species (Konarska & Mazierowska, 2020; Philipp, 1985) in addition to the partial protandry in this species (Asikainen&Mutikainen,2005a; Varga, Nuortila, & Kytöviita, 2013). However, the long style length has negative effects on the female function in hermaphrodites as it reduces the probability of pollen transfer on stigmas by pollinators as is demonstrated in this study. The style length has been shown to have a relatively narrow optima for pollen deposit and pollinator contact probability in Brassica napus flowers (Cresswell, 2000). Larger flowers are advantageous in male function in the way of pollen transport from hermaphrodite flowers (Ashman, 1992). In agreementwithourstudy,Ashman(1992) found that, although longerpetalscontributedtoabetterpollenexport,thepetallengthwas a poor predictor of pollen deposition. Concluding from the contacts to G. sylvaticum reproductive organs in our study, pollen display in hermaphrodites matched stigma display in hermaphrodite flowers poorly.Incontrast,femaleflowermorphologywasabettermatch tothehermaphroditepollendisplay.Accordingly,differentaspects ofmorphologypromotedifferentsexualfunctions.Hermaphrodite morphologyisadaptedtopollenexport(Ashman,1992;Asikainen & Mutikainen, 2005a;Bond&Maze,1999).Infemales,flowermorphologythatmaximizespollenreceiptonstigmaaccordingtohermaphrodite pollen display should be selected because it is the sole function of the female flowers. Thesmallsizeoffemaleflowersisoftenexplainedbytheresourcere-allocation hypothesisstatingthatthe energyandnutrient investment difference between hermaphrodite and female flowers may be allocated to seed production (Ashman, 1994; Chang, 2006; Eckhart, 1992).Inthiswork,wechallengethenon- adaptiveandresourcere-allocationhypothesesinexplainingthe floral dimorphism in gynodioecious plants. We specifically tested the “pollination hypothesis” that flower size variation in G. sylvaticum is adaptive because it enhances probability of a visitor contacting stigma, and thus promotes pollination probability in females. We stress that we did not measure pollen deposition, but probability of pollen deposition. We base this estimate on the assumption that only when an insect makes a ventral contact with thereceptivestigmalobes,pollenisdeposited.Inthecasewhen the receptive stigma lobes are not contacted, pollen cannot be transmitted.Insupportofthepollinationhypothesis,thestigmas in the small female flowers were more likely to be contacted by visitors than the stigmas in hermaphrodite plants. This indicates that the balance between visitor attraction and consequent pollen 20457758, 2022, 12, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ece3.9670 by University Of Jyväskylä Library, Wiley Online Library on [28/12/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License