Pollen Donation Patterns in a Tropical Distylous Shrub (Psychotria suerrensis; Rubiaceae)
Author(s): Judy L. Stone
Reviewed work(s):
Source: American Journal of Botany, Vol. 82, No. 11 (Nov., 1995), pp. 1390-1398
Published by: Botanical Society of America
Stable URL: http://www.jstor.org/stable/2445865 .
Accessed: 02/03/2012 16:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
Botanical Society of America is collaborating with JSTOR to digitize, preserve and extend access to American
Journal of Botany.
http://www.jstor.org
AmericanJournalof Botany 82(11): 1390-1398. 1995.
POLLEN
DONATION
TROPICAL
PATTERNS
DISTYLOUS
IN A
SHRUB
(PSYCHOTRIASUERRENSIS;RUBIACEAE)l
JUDYL. STONE2
Departmentof Ecology and Evolution,State Universityof New York at Stony Brook, Stony Brook, New York 11794-5245
Selection on male functionhas been invoked to explain various floralfeatures,includingnumberof flowers,flowersize,
and flowercolor. Here I describe two experimentsdesigned to examine the efficiencyof distylyin promotingmale floral
function,as measuredby successfulpollen dispersalto stigmas.In both experiments,I performedemasculationsto control
the type of pollen locally available in a naturalpopulation of Psychotria suerrensis,a tropical shrub. In the "naturalrecipients"experiment,I allowed each floralmorphto donatepollen on alternatedays to emasculatedflowersof each morph.
In the "paired-recipients"experiment,I attachedpaired cuttingsof each morphto individualdonor plants. The resultsof
most efficiently
to low stigmas
bothmethodswere consistent.Pollen borne on low anthers(frompin plants) was transferred
(on thrumplants). Pollen borne on high anthers(fromthrumplants) was dispersed in equal amounts to flowersof both
morphs.The resultssuggestthatdistylyis only partiallyeffectivein achieving efficientpollen donation.Male functionof
but male functionof thrumsis not.A supplementalpollinationexperimentillustrates
pins is enhancedby the polymorphism,
thatseed set in this species is predominantly
pollen limited,reducingthe importanceof male function,in comparisonwith
species where seed set is primarilyresourcelimited.
Hermaphroditicflowersfulfillboth male and female
function.Female functionis relativelyeasy to monitor,
and has been the focus of most measuresof reproductive
success. However, male functionof flowersis undoubtedly important.A numberof workershave suggestedthat
the low fruit/flower
ratio common in floweringplants is
due to productionof excess flowers to enhance male
function (Willson and Rathcke, 1974; Willson, 1979;
Bawa and Webb, 1984; Sutherland and Delph, 1984;
Sutherland,1986). Male functionhas also been thought
to shape morphologyof individual flowers. In species
where seed set is limitedby resourcesand not by pollen,
selectionforpaternalsuccess could lead to morphological
changes thatare neutralwithrespectto seed set (Charnov,
1979; Willson, 1979; Queller, 1983; Bell, 1985; but see
Wilson et al., 1994). Various floral featureshave been
shown to affectprimarilymale function,as measuredby
pollen removal,pollen donation,or donationof fluorescent dyes used as pollen analogs. Among these are corolla color (Schoen and Clegg, 1985; Stanton,Snow, and
Handel, 1986), corolla size (Bell, 1985; Young and Stanton, 1990; Campbell et al., 1991), nectarproductionrates
be(Mitchell, 1993), and featuresreducinginterference
tween anthersand stigma(Lloyd and Yates, 1982).
A strikingmorphologicalfeaturethatalso may affect
male functionis the balanced polymorphismknown as
I Manuscriptreceived 17 October 1994; revision accepted 18 April
1995.
The authorthanksJ.D. Thomson foradvice duringthe course of this
work,and D. G. Lloyd forsharingunpublishedmanuscripts.The manuscriptwas improvedby commentsfromS. C. H. Barrett,K. S. Bawa,
B. L. Bentley,D. J. Futuyma,P. O'Neil, and J. D. Thomson. W. F.
Morris helped devise the maximum-likelihoodprocedure.The OrganizationforTropicalStudies providedlogisticalsupport.Financial support
was provided by the Organization for Tropical Studies, Sigma Xi
Grants-in-Aidfor Research, NSF Dissertation ImprovementAward
BSR91-22752, and a SUNY at Stony Brook GraduateFellowship.
2Current address:Departmentof Zoology, Duke University,
Durham,
NC 27708.
distyly.In distylous species, about half of the plants
(pins) have long styles and shortstamens,while the remainder (thrums)have short styles and long stamens.
Characteristically,the two morphological groups also
formmating groups, with each morph able to set seed
only when pollinated by the other morph. Distyly has
arisen independentlyat least 23 times in unrelatedtaxa
(Lloyd and Webb, 1992a); thereforethe adaptive forces
favoringit are presumablyimportantacross a spectrum
of ecological conditions. Tristyly,a balanced polymorphism with threestylarlengthmorphs,is less common,
but appears also to have several independentorigins,
since it is found in both monocots and dicots (Ganders,
1979). The term "heterostyly"encompasses both conditions.
The adaptivenessof distylyhas been investigatedalmost exclusivelyin termsof its female function,specifically pollen receipt:do flowersreceive morepollen from
plants of the othermorphthan fromplants of theirown
morph?The numerousattemptsto address this question
have failed to yield a clear result,partlybecause most
did not distinguishbetweenself and outcrosspollen loads
(reviewed by Ganders, 1979; Barrett,1990). Intrafloral
pollinationis not affectedby the existence of the polymorphism,but elevates theamountof same-morphpollen
foundon the stigma.Therefore,when workersneglectto
emasculate recipientflowers,theyare likely to conclude
thatthepolymorphism
is notefficient
in promotingpollen
transfer,since so much intramorphpollen is found on
stigmas.The few studiesthathave includedemasculation
of the recipient morph have found intermorphpollen
loads in excess of random expectation(Ganders, 1979;
Barrettand Glover,1985; Piper and Charlesworth,1986).
Distyly could, in fact, be neutralin termsof female
function,but be advantageousformale function.That is,
dointermorph
pollen donationcould surpassintramorph
nation,even if pollen receiptrevealed no disassortative
pollination.A recent model of the evolution of distyly
1390
November 1995]
STONE
POLLEN
DONATION
(Lloyd and Webb, 1992b; reviewed by Barrett,1990)
demonstratesthat variationin pollen donation could be
essentialin the originationand maintenanceof thisfloral
polymorphism.Earlier models had suggestedthatthe genetic matinggroups of distylywere likely to evolve before the morphological features (Charlesworth and
Charlesworth,1979). In contrast,Lloyd and Webb (following Darwin, 1877) suggestedthereverse.Theirmodel
assumes an "approach-herkogamous"ancestralstate,that
is, one in whichthe stigmais exsertedbeyondtheanthers
(Webb and Lloyd, 1986). A stylarlengthpolymorphism
could then occur if a mutantwith a shortstyle were to
become established.Conditionsfor a balanced polymorphism are based on pollen transferprobabilities.Their
model shows that if seed set is resource limited,it is
sufficientfor each morph to be more successful at donatingpollen to the opposite morphthanto flowersof its
own morph.If seed set is pollen limited,the sum of intermorphdonationand receiptmustbe greaterthantwice
In eithercase, pollen donationis
the intramorph
transfer.
important.
Althoughthe male reproductivesuccess of single into measure(reviewedby
dividuals is notoriouslydifficult
Snow and Lewis, 1993), assessing the overall success of
entireclasses of individuals can be less challenging.In
heterostylousspecies, pollen size polymorphismsoften
allow a directassessmentof the relative success of differentmorphs in donating pollen to stigmas (Darwin,
1877; Ganders, 1979; Barrettand Glover, 1985). Granted
the assumptionthatpollen donationis an importantcomponentof male fitness,the patternsof pollen transferbetween flowersof the same and different
morphscan be
used to assess the efficiencyof heterostylyin termsof
male function (Kohn and Barrett, 1992b; Lloyd and
Webb, 1992b).
The efficiencyof pollen transferin distyly,relativeto
pollen transferin a monomorphicpopulation,has been
measured in several ways (see Discussion). To consider
efficiencyfrom the perspective of male function,one
mustcompare the probabilityof a pollen grainproduced
by one morphto arrive at a legitimate(intermorph)vs.
illegitimate(intramorph)stigma. If the probabilityof a
pollen grainlandingon an intermorph
stigmaexceeds its
probabilityof landing on an intramorphstigma,thenthe
male functionof thatmorphis enhancedby distyly,since
more pollen is donatedthanwould be in a monomorphic
population. A findingthat distylous populations have
more efficientpollen donationthan(hypothetical)monomorphic ones would lend support to the Lloyd-Webb
model, which suggeststhatdistylycould be favoredeven
in the absence of diallelic matinggroups.
The objective of this studywas to assess the role of
distylyin promotingpollen donationbetween flowersof
a tropical distylous shrub, Psychotria suerrensis. This
species has a very slight pollen dimorphism(Stone,
1994), makingit difficultto determinethe morphof origin of pollen grainson stigmas.I thereforerestrictedthe
availabilityof pollen donationtype in the populationto
one morphor the otheron alternatingdays. The stigma
countsfromemasculatedflowersof bothmorphsallowed
me to compareratesof intramorph
and intermorph
pollen
transfer.A second objective was to determinewhether
seed set was limited by pollen availability or by re-
IN A DISTYLOUS
SHRUB
1391
sources,in view of thisissue's importancein the LloydWebb model. If resourceslimitseed set, thenpollen donation is of utmostimportancein the model, and pollen
receipt is omittedfromthe criticalinequalities. In contrast,if seed set is limited by pollen receipt,then the
model considers transferprobabilitiesfor both donation
and receipt.
MATERIALS AND METHODS
Studysite and species-The researchwas done in primary lowland rain forestat La Selva Biological Station,
in Heredia Province,Costa Rica. La Selva is located at
the transitionbetween the central volcanic mountain
chain and the Atlanticcoastal plain. Its climatefalls into
the tropicalwet forestlife zone of the Holdridge classification,receiving -4 m of rain a year,witha dryseason
duringFebruaryand March (McDade et al., 1994). Psychotria suerrensis(Rubiaceae) is a nonclonal understory
shrub.It occurs in low elevation wet forestsfromNicaragua to Panama (Taylor, 1991). At La Selva, it can be
found in relativelyisolated populations of 50-100 individuals on level residual soils on ridgetops,separated
fromotherpopulationsby low-lyingswamps. The population used for this studyis isolated fromotherpopulations by a distance of at least 50 m on all sides. Its
location is tied into La Selva's GIS data base, and the
locations of plants within the population are given in
Stone (1994).
Individuals of P. suerrensisnormallybear 3-15 inflorescences, each opening 0-4 flowersper day duringthe
blooming season, which extends frommid-Februaryto
mid-March.Flowers are 1 cm long withnarrow-tubular
corollas and symmetricalplacementof floralorgans betweenmorphs(Stone, 1994). Pin corolla tubes are slightly, but significantly,
longer than thrums(Stone, 1994).
The flowerslast a single day, openingjust beforedawn,
and wiltingat dusk. Anthersdehiscence depends on relative humidity.On dry days, anthersdehisce soon after
the floweropens. Antherson bagged inflorescencesdo
not normallydehisce for several hours afterdawn. The
flowerssecrete small amounts of nectar and are visited
and hummingbirds.
primarilyby small bees, butterflies,
Euglossine bees appear to be the most effectivepollinators (Stone, 1994). Psychotriasuerrensissets seed only
whenpollinatedby theoppositemorph(Bawa and Beach,
1983). It has two ovules per flower,and eithdrone or two
seeds per fruit.
Pollen counts/anther
were made to allow comparisons
of stigmaticpollen loads withotherpublishedstudies.In
1993, I collectedanthersfroma singleflowerof 12 plants
of each morph. Afterthe anthersdehisced, I added alcohol to the tubes. I countedthe numberof pollen grains
per flower using an Elzone particle counter at SUNY
Stony Brook.
Pollen movementto natural recipients: 1991 experiment-In this experiment,I compared pollen donation
capabilities of the two morphsby limitinglocal pollen
availabilityto one morphat a time.Stigmaticpollen loads
on emasculated recipientsof both morphsallow a comparison of pollen transferefficiencybetween each combination of morphs. In P. suerrensis,as in some other
1392
AMERICAN JOURNAL OF BOTANY
Rubiaceous species, pollen dimorphismis slight(Bir Bahadur,1968; Feinsingerand Busby, 1987; Stone, 1994).
I emasculated all individualsof one or the othermorph
on alternatingdays to restrictthe source of pollen on a
given day to the morphwith intactflowers.Plants with
more thanthreeinflorescenceswere used both as donors
and as focal recipients.Fourteenthrumplantsand 11 pin
plants in this population had displays of four or more
inflorescences.Late in the afternoonprecedingtreatment,
I bagged all inflorescenceson these 25 plants to prevent
pollinatoraccess. On plants withthreeor fewerinflorescences, I removed all large flowerbuds daily; therefore
these individualsneithercontributednor received pollen.
Pins and thrumsserved as donors on alternatingdays.
On days when thrumswere donors,I unbagged inflorescences in the early morningand emasculated flowersof
all pin plantsand flowerson half of the inflorescencesof
each thrumplant.Pollinatorsthereforemoved thrumpollen to stigmasof emasculatedflowersof bothmorphs.To
emasculate pin flowers,I pulled undehisced anthersup
throughthe corolla tube,withoutslittingthe tube. Emasculationswere normallycompletedby 7:00 a.m. On rare
occasions where antherswere noted to be already dehisced, I removed all of the open flowerson thatinflorescence. In the late afternoon,the stigma from each
flowerwas collected in a separate microcentrifuge
tube
and subsequentlymountedin fuchsinjelly. Stigmas were
assigned consecutivenumbers,and stigmaticpollen loads
were counted in a random order at 100X. On alternate
days, I emasculatedhalfof thepins and all of the thrums,
to give thedonationratesfrompin pollen to bothmorphs.
Each treatmentwas performedfor2 d.
I emasculatedhalf of the flowerson each plant of the
donor morphratherthan emasculatingall of the flowers
on half of the plants so thatI would not need to make
of donors
any assumptionsabout the spatial arrangement
and recipients.Note thatthis is a very conservativetest.
If intermorph
exceeds intramorph,
pollen transfer
thepatternmust be quite strong,since intramorphtransferincludes transferwithina single plant. On the otherhand,
if no differenceis found between intra-and intermorph
transfer,it is possible that intermorphtransferis more
efficientthan intramorph,
but thatthis differenceis obscuredby within-plant
movement.To examine thepotential effectsof geitonogamyon pollen flowpatterns,I calculated correlationcoefficientsbetween the number of
donor flowersopen on an individualon a given day and
the number of pollen grains received by emasculated
flowerson thatindividual.A high correlationwould indicate thatwithin-plant
pollen flow substantiallyincreases intramorph
transferestimates.
Plants' floraldisplays ranged fromone to 30 flowers
blooming on a given day. Thereforesample sizes would
be vastlyunequal if stigmaticpollen loads were used as
the unitof replication.In addition,stigmaticpollen loads
withina plant are not independent.Therefore,I used individual plants as the sampling unit, excluding individuals withfewerthanfourflowersper treatment.
I calculated the mean stigmaticpollen load for each plant by
donortype.These means were thenused in a Mann-Whitney U-testto compare pollen transferfor each combinationof donor and recipientmorph.
A controltreatment
consistedof emasculatingall flow-
[Vol. 82
ers in the populationin the morningand collectingtheir
stigmasfor pollen counts in the afternoon.This control
gives an estimateof the sum of two sources of contamination: self-pollinationresultingfromineffectualemasculation,and pollen flow fromoutside of the population.
These two sources of contaminationwould each affect
interpretation
of the resultsdifferently.
Pollen flow from
outside the populationwould be likely to accrue to both
morphsabout equally. Its most likely result,if undetected, would be failureto perceive an effectof the treatments. A Wilcoxon signed-rankstest for paired observations was used to compare plants' mean stigmaticpollen loads on controldays, when all flowerswere emasculated,to loads on experimentaldays, when some local
flowersserved as pollen donors. Contaminationby self
pollen is most importantif it causes systematicallyinflatedestimatesof pollen delivery to one morphor the
other.A Mann-WhitneyU-testwas used to compare differences between stigmatic pollen loads of the two
morphsforthe controland experimentaltreatments.
Pollen movementto paired recipients: 1992 experiment-In 1992, I used anotherdesign to reduce variabilityin depositiondue to the spatial arrangement
of the
two morphs.I attacheda pair of testtubes to the stemof
five large plants of each morph,which were to serve as
the primarydonors. On days when thrumswere to be
donors,I emasculatedall pin plantsin thepopulationand
placed paired emasculated cuttings,one pin and one
thrum,in the tubes of the designateddonor thrumplants.
On alternatingdays, thrumplants were emasculated and
paired cuttingswere placed in tubes on the pin donors.
This ensuredthatthe two types of recipientswere equidistantfromplants of the donor morph. Withina pair,
cuttingswere approximatelythe same size and had the
same numberof open flowers,rangingfromone to five.
Most inflorescenceshad eitherone or two open flowers.
Casual observationindicated thatpollinatorsvisited the
inflorescencesin the tubes as oftenas theyvisited other
inflorescenceson the plant. Thrumsserved as donors for
6 d, and pins served as donorsfor4 d. I collected stigmas
fromrecipientflowersin thelate afternoonand processed
them as before.I used a Wilcoxon signed-rankstest for
paired observations to compare mean pollen loads per
stigmaof the two recipientmorphs.
As a control,insteadof emasculatingall flowersin the
population, as I had done in 1991, I used a maximum
likelihood approach to compare the size-class distributions of the putative donors with that of pollen grains
receivedon stigmas.For thereferencedistribution,
I measured ten pollen grains fromeach of 16 self-pollinated
flowersof each morph.I divided the size-class distribution into seven classes and calculated the proportionof
pollen produced by each morphthat fell into each size
class. On 20 randomlychosen stigmas for each of the
fourintendedtransfertypes,I measuredall pollen grains
at 400X. I used the loss functionin SYSTAT's nonlinear
estimationprocedure(Wilkinsonet al., 1992) to estimate
a contaminationparameterthatmaximizedtheprobability
of generatingthe size-class distributionof pollen found
on stigmas for each of the four donor-recipient
combinations.The model used was:
November 1995]
STONE-POLLEN
DONATION IN A DISTYLOUS
obs = total-[(1 - rho)-ft+ rho.fp],
where obs = the observed numberof grains in a size
class on a stigma, total = the total number of grains
counted on thatstigma,ft = the fractionof thrumgrains
in thatsize class, fp = the fractionof pin grains in that
size class, and rho = the proportionof nonsourcepollen
(in this case, for thrumsas donors). For days when the
intendedtransferswere intramorph,contaminationestimates indicate levels of interpopulationpollen flow. In
contrast,for days when the intendedtransferswere betweenmorphs,contaminationestimatesinclude both selfpollinationand interpopulation
pollen flow,since boththe
nondonorand recipientmorphscome fromthe same pollen size distribution.
Pollen limitationof fruit set-Preliminary observationsindicatedthatfruitset among individualswas highly
variable. Fruitproductionvaries dependingon plant size,
light environment,and previous fruitinghistory(Stone,
unpublisheddata). Rather than tryingto match experimentaland controlplants for all of these factors,I used
experimentaland controlinflorescenceswithinplants.To
address the possibility that supplementallypollinated
flowersset seed at the expense of controlflowerswithin
a plant (see Zimmermanand Pyke, 1988), I compared
fruitset in controlinflorescenceswithdata on fruitset in
unmanipulatedplants.The data on unmanipulatedplants
were gatheredin two similarlysized populations at La
Selva during1991-1993 by D. Grahamas partof a longtermstudyon the dispersal and regenerationof P. suerrensis and otherunderstoryshrubs.
The flowersof P. suerrensis are sessile on terminal
inflorescences.It is very difficultto follow the fate of
individual flowersto the fruitingstage; thereforethe inflorescenceis theonly feasibleunitof replicationforsupplemental pollination studies. Three pairs of inflorescences were chosen on six individuals of each morph.
Each morningfor 15 d at the peak of the floweringseason, I hand-pollinatedflowers on one inflorescenceof
each pair and noted the number of flowerson control
inflorescences.Inflorescenceson three large shrubs of
each morph were bagged to provide anthersfor handpollinations. Within compatible type, no attemptwas
made to keep trackof donor identity.Pollen was applied
by sweeping a single antherfromthe compatiblemorph
across the stigmaticsurface.Hand-pollinationswere conducted between 6:30 and 8:00 a.m. Unripe fruitswere
collected in June.Flowers blooming before or afterthe
15-d period were not counted,so fruitset is recordedas
numberof fruitsper observed flower.Inflorescenceswith
fewerthan two flowersblooming duringthe observation
period were omittedfromanalysis.Fruitset betweennaturally pollinated and supplementallypollinatedinflorescences was comparedusing a Wilcoxon signed-rankstest.
Mean numberof seeds per fruitwas also comparedusing
a Wilcoxon signed-rankstest.
RESULTS
Emasculation reduced stigmaticpollen loads for both
pin and thrumflowers(Fig. 1). Across all treatments
excluding the control treatment,emasculated flowers received an average of 24.9 pollen grains(N = 469, SD =
1393
SHRUB
50
E
C'40
I
z
0
z' 30
zw
20
10
pin
thrum
RECIPIENT TYPE
TREATMENT
O intact
O emasculated
Fig. 1. Mean stigmaticpollen loads of intactand emasculatedflowers of individualplants.Intactpollen counts include intrafloral
and interfloraltransfersfromwithinthe same plant and otherplants of the
same morph.Emasculated counts include interfloraltransferfromthe
same plant and fromotherplants of both morphs.Values are means
fromall flowerscounted for the 11 thrumsand eight pins for which
paired comparisonswere possible. Verticallines indicate standarderrors.
39.69), while intact flowersreceived 44.0 grains (N =
316, SD = 66.47). This differenceis statisticallysignificant,based on a Wilcoxon signed-rankstest comparing
means for intact and emasculated flowersof individual
plants (P < 0.01, Ts = 25.5, N = 19). Based on differences between stigmaticpollen loads of emasculatedand
intactflowers,pin flowersappear to deposit more selfpollen than do thrumflowers(Fig. 1). A Mann-Whitney
U-test comparing the magnitude of individual plant's
mean differencesbetweenemasculatedand intactflowers
forthe two morphsreveals no significance(P > 0.05, Us
= 59, N = 11, N2 = 8); however,the sample size for
this comparisonis quite small. Geitonogamydid not aptransferlevels. The numberof
pear to inflateintramorph
donor flowers open on an individual on a given day
ranged fromzero to 13, with a median value of three.
The numberof donor flowerson a plant did not affect
the numberof pollen grainsreceived by emasculatedrecipientflowerson thatplant(Spearmankrankcorrelation,
r = 0.28, P > 0.05).
Pollen importfromotherpopulationswas substantial.
when all flowersin the local
For the controltreatment,
population were emasculated,stigmasreceived an average of 32.5 pollen grains(N = 181, SD = 36.41). During
1394
[Vol. 82
AMERICAN JOURNAL OF BOTANY
a.
0
40
30
-j
Mann-WhitneyU-teststatisticsbased on mean per-plantstigexperiment.On a given
maticpollen loads forthenatural-recipients
day, local pollen was only frompin or fromthrumplants, designated as "donors." All recipientflowerswere emasculated. Comparisonof the top two rows reveals the effectivenessof distylyfor
male function;theyindicatethe extentof disassortativepollen donation.The bottomtwo rows allow a comparisonof female function; theyindicatethe extentof disassortativereceipt.
TABLE 1.
Donors
wz20
0
Pins
Thrums
Pins, thrums
Pins, thrums
Recipients
N,, N2
Us
P
Pins, thrums
Pins, thrums
Thrums
Pins
7, 10
7, 7
7, 10
7, 7
59
25
30
7
0.019
0.949
0.626
0.025
SD = 30.8; Us = 30.8; N1 = 7, N2 = 9, P > 0.05).
However, when all local pollen of one morph was removed, pollen receiptpatternsdifferedbetweenpins and
410
thrums,as explained below.
Pollen donation abilitydifferedamong morphcombiexperiment(Fig. 2a).
nations for the natural-recipient
30
Pollen from pins was deposited significantlymore on
thrumstigmasthanon pin stigmas(Table 1). In contrast,
(0
pollen fromthrumswas deposited equally on stigmasof
both morphs(Table 1). In termsof pollen receipt,more
intermorph
pollen tendedto be depositedthanintramorph
pollen; however due to large variances this difference
40
was only significantforpins as recipients(Table 1). Pins
received significantly
morepollen fromthrumsthanfrom
t-op
t A0t
p -ot
p _)Ip
0)
pins.
Pollen donation abilityin the paired-recipients
experTYPE
andintamrphpolentrase TRANSFER
pficeny tll
rcpet'fowr* wr
iment confirmsthe 1991 natural-recipient
results (Fig.
0
1
2b). Pollen borne on low antherswas deposited preferentiallyon low stigmas(Ts = 29.5, N = 20, P < 0.001),
whereas pollen fromhigh anthersmoved equally to both
morphs(Ts = 192.5, N = 30, P > 0.05). Mean stigmatic
pollen loads were not significantlydifferentbetween
patternswere
years (t-test,P > 0.05), and pollen transfer
similar(Fig. 2). The only differencein relativemagnitude
of pollen transferwas fromthrumdonors to thrumrecipients. In the 1991 natural-recipients
experiment,thrums
appeared to receive morepollen frompins thanfromother thrums,althoughas mentionedabove, this difference
was not significant.In the 1992 paired-recipients
experiment,it appears thatthrumsreceived more pollen from
emasculated.Bars span the distancebetweenthe upper and lower quarotherthrumsthantheydid frompins (Fig. 2b), however
tile, with.the median included in the bar. (a) Natural-recipients
experia Mann-WhitneyU-testindicatesthatthisdifferencewas
ment.Values are medians and quartilesfromall flowerscountedforthe
Final interpretation
of resultsfromthe
also insignificant.
ten thrumsand seven pins that had more than four flowersin each
treatment.(b) Paired-recipientsexperiment.Values are medians and
1992 experimentdepends on pollen size distributions
quartilesfromall flowerscounted for each treatment.For pins as dofoundon stigmas,as explained below.
nors,N = 20. For thrumsas donors,N = 30.
Pin flowerscontained 1,987.4 pollen grains (N = 12,
SD = 399.97) and thrum flowers contained 1,870.8
grains (N = 11, SD = 365.15). There was no difference
in pollen productionbetween the two morphs (t-test,P
when some flowersin the local
experimentaltreatments,
> 0.05). Thrumpollen was, on average, largerthan pin
populationremainedintact,emasculatedflowersreceived
24.9 pollen grains (N = 469, SD = 39.69). This differ- pollen (47.4 pm vs. 39.6 Vm; t, = 16.9; df = 63; P <
enceis notstatistically
(Ts = 33,N = 16,two- 0.00 1), but had a largercoefficientof variation(0.104 vs.
significant
tailed P > 0.05). Population-level emasculation treat- 0.093; sv = 0.009; N = 60, P < 0.001). Figure 3 shows
the size-class distributionof pollen produced by each
ments did affectthe type of locally distributedpollen.
When all flowerswere emasculated,mean stigmaticpolmorph,along with the size-class distributionof grains
found on stigmas for each morphas the putativedonor.
len loads did not differbetween pins and thrums.Pins
For some transfers,the size-class distributionof pollen
received an average of 33.2 grains(N = 72, SD = 43.7),
measuredon stigmasappears the same as the distribution
and thrumsreceived an average of 32.1 arains (N = 109,
Ulo
November 1995]
STONE-POLLEN
DONATIONIN A DISTYLOUS SHRUB
PIN DONORS
O to pins
0.3 _
3 to thrums
E: production
0.2
0.1
0.0
1
2
3
4
.5
6
7
SIZE CLASS
THRUMDONORS
0 to pins
O to thrums
0.2 -
production
i
0.1
0.0
1
2
3
4
5
6
7
SIZE CLASS
Fig. 3. Pollen size-class distributionsfrom the 1992 experiment.
Size-class distributions
forpollen foundon stigmason days when pins
(top) or thrums(bottom)were designateddonors. White bars represent
grainsfoundon pin stigmas,graybars representgrainsfoundon thrum
stigmas,and hatchedbars representthe size-class distribution
of pollen
producedby each morph.
1395
producedby the donatingmorph(e.g., pin to thrum).For
other transfers,especially thrumto pin, the fitbetween
the size-class distributions
of pollen received on stigmas
and pollen productionis not good. These qualitativeobservationsare supportedby the maximum-likelihoodestimates.
The maximum-likelihoodestimationof contamination
fromthe 1992 experimentrangedfrom15 to 54%. Twenty-fourpercentof grainsin thrum-to-thrum
transfersand
29% of grains in pin-to-pintransferswere unlikely to
come fromeitherthe recipientor the putativedonor,and
can be attributed
to intermorph
pollen flow fromoutside
of the population.For pin-to-thrum
transfers,15% of the
pollen countedon stigmaswas most likelyto have come
fromthrums,eitherself or outside of the population.For
54% of thepollen countedon stigthrum-to-pin
transfers,
mas was estimatedto have come frompins, eitherself or
outside of the population. Therefore,the rate of thrumto-pintransferwas substantiallyoverestimatedby counts
of stigmaticpollen loads. This reinforcesthe conclusion
thatpollen is transferred
more efficiently
frompins than
fromthrums.
Althoughboth experimentsshow that pollen transfer
frompins is moreefficient
thantransfer
fromthrums,this
does not necessarilytranslateinto a higherlevel of compatible receipt by thrumsthan by pins, because more
thrumpollen may be dispersed overall. In 1991, it appears that pin-to-thrumtransfermay have exceeded
but this differencewas not signifithrum-to-pin
transfer,
cant (Mann-WhitneyU = 347; N1 = 20; N2 = 31; P >
0.05). In 1992, the two rates look equal, but the sizeclass distributionsof pollen grains on stigmas indicate
thatpin-to-thrum
transfer
may well have exceeded thrumto-pintransfer.
However, if legitimatetransferto thrums
exceeds legitimatetransferto pins, it is apparentlynot
translatedinto differentialfruitproduction. Unmanipulated individuals of the two morphs produced equal
numbersof fruitsper inflorescence(Table 2; t-testfor
differencesbetweenmeans,t = 0.8, df = 174, P > 0.05).
Inflorescenceswith hand-pollinatedflowersset more
fruitthaninflorescenceswithoutsupplementalpollination
(Table 2; Wilcoxon rank sum Ts = 59, N
28, P <
0.001). They also had more seeds per fruit(Table 2; Wilcoxon ranksum T, = 92, N = 30, P < 0.01). The control
inflorescenceson the plants given supplementalpollination had the same numberof fruitsper inflorescenceas
did unmanipulatedplants (Table 2; t-testfor differences
between means, t = 1.5, df = 186, P > 0.05).
2. Results of supplementalpollinationforinflorescencesof Psychotriasuerrensis.Fruitset is reportedper observed flowerbecause flowers
at the tails of the floweringseason were not counted.Values are means ? standarderrors.Sample sizes forthe manipulatedplantsare 30 pairs
of controlinflorescencesand supplementally-pollinated
inflorescencesacross 11 plants. Sample sizes forthe unmanipulatedplants are 87 pins
and 90 thrums,over two populationsin 3 yr (D. Graham,unpublisheddata).
TABLE
Fruits/observed
flower
Pollen added
Controlinflorescences
Unmanipulatedplants (pins)
Unmanipulatedplants (thrums)
1.5 ? 0.21
0.8 + 0.15
Seeds/fruit
1.6 ? 0.05
1.4 + 0.05
Fruits/inflorescence
11.1 ?
6.5 ?
5.2 ?
4.7 ?
1.20
0.75
0.49
0.49
AMERICAN JOURNALOF BOTANY
1396
POLLEN
DEPOSITION
POLLEN
PICK-UP
probabilities
Fig. 4. Possible mechanismforsuperiorpollen transfer
between lower level organs. The curved line in each flowerrepresents
a proboscisand the stippledbar to the leftof each flowerillustratesthe
portionof the proboscis thatis likely to contactthe anthers(leftpanel)
or stigma (rightpanel). The proboscis is likely to contact high-level
anthersover the entireportionthatentersthe corolla. In contrast,only
the lower portionof the proboscisis likelyto contactlow-level anthers.
Only the lower portionof the probosciswill contacta low-level stigma.
The proboscis may contact a high-level stigma over any part of its
lengththatentersthe flower,but, dependingon the orientationof the
bifid stigmaticlobes, it may also fail to contact the stigma. The net
resultof these contact probabilitieswould be accurate pollen transfer
between low organs and relativelyimprecise transferbetween upperlevel organs. Diagram adapted fromLloyd and Webb, 1992b.
DISCUSSION
patternsin P. suerrensiswere consistent
Pollen transfer
experimenover two floweringseasons and two different
tal techniques.In bothcases, distylyservedmale function
withonly partialefficiency.Pollen borne deep withinthe
to low stigmas,but
transferred
floraltube was efficiently
pollen borne towardsthe mouthof the corolla was redistributedequally on both floralmorphs.
What causes thispatternof pollen deposition?It seems
thatthe tips of pollinators' mouthpartsare placed fairly
preciselyin the bottomof the narrowfloraltube,causing
disassortativepollinationfromlow anthersto low stigmas
in much the way envisioned by Darwin (1877) for distylousplantsin general.Why is pollen producedon high
anthersdepositedequally on both low and high stigmas?
A possible explanationis thatthe proboscis contactspin
anthers only near the tip, but contacts thrumanthers
across the entireportionthatis insertedinto the flower
(Fig. 4, adapted fromLloyd and Webb, 1992b). Either
pin or thrumstigmas are then equally likely to contact
portionsof the proboscis thatbear thrumpollen. A similar idea was presentedby Ganders (1974), who postulated thatgreaterefficiencyin the pollinationof thrums
than pins is caused by constraintson pollinatororientation imposed by the corolla tube.
Pollen flow fromoutside of the population was substantial. During the 1991 control treatment,when all
flowersin the population were emasculated,stigmasreceived as much pollen as they did duringexperimental
treatments,when some flowers within the population
served as donors. However,patternsof pollen deposition
across morphtypesshow thattheexperimentaltreatments
did influencethe local pollen flow patterns.For the controltreatment,
pins and thrumsreceived equal quantities
of pollen. During experimentaltreatments,in contrast,
amountsof pollen, depins and thrumsreceiveddifferent
[Vol. 82
pendingon which morphin the local populationwas donatingpollen. Thus, althoughemasculationdid not affect
the total amountof pollen received, it did influencethe
type of pollen deposited on stigmas.
Results from the 1992 control treatmentsupportthe
impression that there were high levels of pollen flow
from outside of the population. Estimated interpopulational movementrangedfrom24 to 29% of pollen grains
so is
transfer,
received,but thisincludes only intermorph
Contaminationdue to selfpresumablyan underestimate.
ing did differbetween morphs. An estimated 54% of
came not from
transfers
grainsmeasuredin thrum-to-pin
thrums,but frompins. This value includes both self-pollination and pin pollen fromoutside the population. In
contrast,only 15% of pollen measured in intermorph
transfersto thrumscame from thrums.This value includes both selfingand pollen flow fromotherpopulations. Since interpopulationalflow was relativelyhigh,
we can conclude that contaminationof thrumsdue to
unsuccessfulemasculations was low. How do the conof the results?
taminationestimatesaffectinterpretation
Contaminationfromoutside the populationaffectedboth
morphs about equally, and thereforeshould not have
much effecton the relative transferrates measured for
the fourcombinationsof donor and recipient.Contaminationby incompleteemasculationsdoes change relative
transferlevels, however. Selfing by pins resultedin an
overestimateof transfersto pin flowers,implyingthat
lower than the
transferswere significantly
thrum-to-pin
measured values. The high degree of self-contamination
by pins does not, however, alter the conclusion that
transferexceeds pin-to-pintransfer,since
thrum-to-pin
is constantfor this comthe rate of self-contamination
parison.
Comparisons with other species-Asymmetrical pollen donationefficiencybetween morphsis consideredto
be common among heterostylousplants (reviewed by
Ganders, 1979; Barrettand Glover, 1985). Several authorshave concluded thatpin plantstypicallyreceive disproportionately more illegitimate pollen than thrum
plants(Ganders, 1979; Dulberger,1992). These estimates
include self-pollenin most cases, so it is impossible to
determinewhetherthe differing
proportionsof legitimate
pollination are simply due to differentself-pollination
rates. As Ganders (1974, 1979) emphasized, interfloral
ratesare criticalin evaluatingthe efficienpollen transfer
cy of the polymorphism.
Comparingpollen transferefficiencyacross species is
complicatedby the various ways thatefficiencyhas been
measured. In the simplesttype of measurement,pollen
grains of each type are counted on stigmas fromemasculatedflowersof each morph.The polymorphismis conif legitimate
sidered to serve female functionefficiently
(same-level) pollen loads exceed illegitimateloads. A refinementof this technique takes into account the frequencies of the morphs and theirdifferencesin pollen
production,so that stigmaticpollen loads are compared
withthoseexpectedunderrandomtransfer
(e.g., Ganders,
1974). A thirdmethoduses pollen productionlevels and
morphfrequenciesto calculate the probabilityof an individual pollen grainof one morphlandingon the stigma
of any othermorph(Lloyd and Webb, 1992b). These pol-
November 1995]
STONE
POLLEN DONATION IN A DISTYLOUS
approaches to measuringefficiencyof het3. Three different
erostyly.Data fromemasculatedrecipientsof Jepsonia heterandra
(Ganders, 1974). p to t indicatespin grainsfoundon thrumstigmas,
t to p indicatesthrumgrainson pin stigmas,etc. Method 1 simply
reportsthe mean numberof grains found for each transfertype.
For method 2, values are the observed numberof grains divided
by random expectationbased on morphfrequenciesin the populationand pollen productionrates.Method 3 gives probabilitiesthat
individualpollen grainsreach a stigmaof the specifiedtype (from
Lloyd and Webb, 1992b).
TABLE
Method 2
Method 1
Method 3
Transfer
type
Value
Rank
Value
Rank
Value
Rank
p to t
t to p
p to p
t to t
572
406
628
117
2
3
1
4
1.49
1.30
0.87
0.56
1
2
3
4
2.71
4.46
2.98
1.29
3
1
2
4
len transferprobabilitiescan be used to interpretboth
pollen donationand receipt.
Data fromemasculatedflowersof Jepsonia heterandra
(Ganders, 1974) illustratethat the three methods yield
(Table 3). The unmutuallyinconsistentinterpretations
adjusted stigmatic pollen counts indicate that distyly
functionsto delivermorelegitimatethanillegitimatepollen to thrumstigmas,but not to pins. When stigmatic
pollen loads are adjustedforpollen productionand morph
frequency,it appears thatboth morphsexperiencedisassortativepollination,but thatthe degree of disassortative
pollinationis higherforthrumsthanforpins. Finally,the
probabilitiesforindividualpollen grainslandingon compatible stigmasindicatethatthe highesttransferefficiency is fromthrumto pin. While the second two methods
agree thatdisassortativepollinationoccurs,theydifferas
to the directionof the asymmetry.
The data presentedin this experimentare most analto corogous to the thirdmethod.Ratherthanattempting
rect for differencesin morph frequencyor pollen production, I controlledpollen source for any given day.
Therefore,a comparisonof each morph'sabilityto donate
to both typesof flowerscan be made directlyfromcomparisonsof stigmaticpollen loads. Donation efficiencyis
dramaticallyhigherfor thrumsthan for pins. Based on
thepollen probabilitymethod,thedirectionof asymmetry
for this species is anomalous. A reanalysisof published
stigmaticpollen loads indicatesthatfor nearlyall distylous species, thrumpollen has a greaterprobabilityof
landing on a compatible stigma than does pin pollen
(Stone and Thomson, 1995).
A numberof studies with tristyloustaxa also support
the idea thatpollen flowbetween high-levelorgans usually exceeds transferbetween low-level organs. In Pontederia cordata, pollen producedby long anthershas the
highestprobabilityof encounteringa compatible stigma
(Barrett and Glover, 1985; Wolfe and Barrett, 1989;
Lloyd and Webb, 1992b). In experimentalpopulationsof
self-compatibleEichhornia paniculata, the short-styled
morphhad the highestmale fitnessand the long-styled
morphwas relativelyfemale (Kohn and Barrett,1992a).
The differencein functionalgender was directlyrelated
to differencesin pollen transferefficiency(Kohn and
among
Barrett,1992b). In Lythrumsalicaria, dye transfer
morphswas asymmetric,with more thanrandomexpec-
1397
SHRUB
4 Conditionsforinvasion of a monomorphicpopulationby a
style-lengthmutant(fromLloyd and Webb, 1992b). Mean pollen
to long-styledmorphis representedby
transferfromshort-styled
q,p,rom long-styledto long-styledby qpp,etc. Underparticularconditionsof limitationof seed set, both inequalitiesmustbe met for
a balanced polymorphismto become established.The last column
patternswould
indicateswhetheror not present-daypollen transfer
meet conditionsforinvasion.
TABLE
Invadingmorph
Limitation
of seed set
Short-styled
Long-styled
Short-styled
Long-styled
resources
resources
pollen
pollen
Necessary inequality
qtp>
qpt>
qtp+
qtp+
Upheld by experiments?
yes
qpp
1991 yes, 1992 no
qtt
qpt> 2qpp yes
qpt> 2qtt 1991 yes, 1992 no
morphsto mid- and
tation of transferfromshort-styled
long-styledmorphs (O'Neil, 1992). O'Neil (1992) also
pollen
showed differencesin seed set and morph-specific
limitation.In Psychotriasuerrensis,althoughpin pollen
more efficiently
thanthrumpollen, overwas transferred
all rates of disassortativepollinationto the two morphs
did not differ(because more thrumpollen is capturedby
stigmas), and fruitproductionby the two morphs was
equal.
Besides the directionof asymmetryin compatiblepollen transfer,
anotherapparentlyanomalous featureof P.
suerrensisis theequal pin: thrumpollen productionratio.
In most taxa surveyed,pin flowersproduce significantly
more pollen than thrumflowers(Ganders, 1979; exceptions listed by Dulberger, 1992). When pollen transfer
efficienciesare calculated on a per-pollengrainbasis, the
apparentlow efficiencyof pins as males in many species
may just be due to theirhigherpollen productionlevels,
and consequentlower per-grainredepositionrates.Alternatively,the higherproductionof pin than thrumpollen
in most species may be a resultof selectiondue to lower
per-grainredepositionrates of pollen fromlow anthers.
Implicationsfor evolutionof distyly-A recentmodel
of the evolutionof distyly(Lloyd and Webb, 1992b; reviewed by Barrett,1990) suggeststhatmale functionmay
morphinvade an
be importantin helping a short-styled
ancestralpopulation of long-styledflowers.Sets of inequalities concerningpollen transferprobabilitiesdefine
whetheror not a balanced polymorphismcould subsequentlybecome establishedaftersuch an invasion (Table
4). The sets of inequalitiesdifferdependingon whether
seed set is limitedby resourcesor by pollen. Male function is critical in situationswhere seed set is resource
limited.Where seed set is pollen limited,both male and
female functionare importantfor a balanced polymorphism to become established.
Presumingthatthe self-compatibleancestralformhad
patterns,theseresimilarmorphologyand pollen transfer
mutantbesults supportthe plausibilityof a short-styled
ing able to invade a long-styledpopulationdue to differwhetherseed set were
ential success in pollen transfer,
pollen limitedor resourcelimited.However,it is notclear
whetherconditionsfora balanced polymorphismare fulfilled.Results fromthe 1991 population-levelexperiment
suggestthata pin morphwould be favoredin a populationdominatedby thrumplants,butresultsfromthe 1992
paired-recipientsexperimentdo not. If conditionsprev-
1398
AMERICAN JOURNAL OF BOTANY
alent during1991 prevailed,a thrum-type
mutantwould
spread graduallythroughoutthe population. Fluctuating
selectionbetweenyearscould have assistedthe establishmentof the originalpolymorphism.
What does the fact thatfruitset is pollen limitedsay
about gender-specificselection pressureson Psychotria
suerrensis? As mentioned above, if fruitset were resource limited,the argumentcould be made thatselection
on floralmorphology,includingthe success of an invading mutant,would operate primarilythroughmale function. Since fruitset is pollen limited,it is reasonable to
conclude thatthe functionof the polymorphismin pollen
receipt is important.The fact that seed set, as well as
fruitset, increased with supplementalpollination indicates thatper-visittransferof pollen is important.Nevertheless,the dramaticincrease in fruitset resultingfrom
hand-pollinationsuggeststhattherewere numerousflowers that were never visited. It would be interestingto
know whetherfruitset would be affectedby supplemental pollinations applied only to flowersalso visited by
pollinators.Efficiencyof pollen capture by individual
flowersmay be less importantto female functionin this
species than overall pollinatorattraction.
LITERATURE CITED
BARRErT,S. C. H. 1990. The evolution and adaptive significanceof
heterostyly.
Trends in Ecology and Evolution 5: 144-148.
, AND D. E. GLOVER.1985. On the Darwinian hypothesisof the
adaptive significanceof tristyly.
Evolution39: 766-774.
BAWA, K. S., AND J. H. BEACH. 1983. Self-incompatibility systems in
the Rubiaceae of a tropicallowland wet forest.AmericanJournal
of Botany 70: 1281-1288.
, AND C. J. WEBB. 1984. Flower, fruit,and seed abortionin
tropicalforesttrees:implicationsforthe evolutionof paternaland
maternalreproductivepatterns.American Journal of Botany 71:
736-751.
BELL, G. 1985. On the functionof flowers.Proceedings of the Royal
Society of London, Series B 224: 223-265.
BIR BAHADUR.1968. Heterostylyin the Rubiaceae: a review. Journal
of Osmania University(Science) Golden Jubilee Special Volume:
207-238.
CAMPBELL, D. R., N. M. WASER, M. V. PRICE, E. A. LYNCH, AND R. J.
MITCHELL.1991. Componentsof phenotypicselection: pollen export and flowercorolla width in Ipomopsis aggregata. Evolution
45: 1458-1467.
CHARLESWORTH,
D., AND B. CHARLESWORTH.
1979. A model for the
evolutionof distyly.AmericanNaturalist 114: 467-498.
E. L. 1979. Simultaneoushermaphroditism
and sexual seCHARNOV,
lection. Proceedings of the National Academy of Sciences 76:
2480-2484.
DARWIN,C. 1877. On the different
formsof flowerson plants of the
same species. Murray,London.
R. 1992. Floral polymorphismsand theirfunctionalsigDULBERGER,
nificancein the heterostyloussyndrome.In S. C. H. Barrett[ed.],
The evolutionand functionof heterostyly,
41-84. Monographson
Theoreticaland Applied Genetics 15. Springer-Verlag,
Berlin.
FEINSINGER,
P, ANDW. H. BUSBY. 1987. Pollen carryover:experimental
comparisons between morphs of Palicourea lasiorrachis (Rubiaceae), a distylous,bird-pollinated,tropicaltreelet.Oecologia 73:
231-235.
F R. 1974. Disassortativepollinationin the distylousplant
GANDERS,
Jepsonia heterandra.Canadian Journalof Botany 52: 2401-2406.
. 1979. The biology of heterostyly.New Zealand Journal of
Botany 17: 607-635.
[Vol. 82
KOHN, J. R., AND S. C. H. BARRETT. 1992a. Experimentalstudies on
the functional significance of heterostyly. Evolution 46: 43-55.
. 1992b. Floral manipulationsreveal the cause of
, AND
male fitnessvariationin experimentalpopulations of Eichhornia
paniculata (Pontederiaceae).Functional Ecology 6: 590-595.
LLOYD, D. G., AND C. J. WEBB. 1992a. The evolution of heterostyly.
In S. C. H. Barrett[ed.], The evolutionand functionof heterostyly,
151-178. Monographs on Theoretical and Applied Genetics 15.
Springer-Verlag,
Berlin.
, AND
. 1992b. The selectionof heterostyly.
In S. C. H.
Barrett[ed.], The evolutionand functionof heterostyly,
179-207.
Monographs on Theoretical and Applied Genetics 15. SpringerVerlag,Berlin.
, ANDJ. M. A. YATES. 1982. Intrasexualselectionand the segregationof pollen and stigmasin hermaphrodite
plants,exemplified
by Wahlenbergiaalbomarginata (Campanulaceae). Evolution 36:
903-913.
McDADE, L. A., K. S. BAWA, H. A. HESPENHEIDE,AND G. S. HARTSHORN.
1994. La Selva: ecology and naturalhistoryof a neotropicalrain
forest.Universityof Chicago Press, Chicago, IL.
MITCHELL,R. J. 1993. Adaptive significanceof Ipomopsis aggregata
nectarproduction:observationand experimentin the field.Evolution 47: 25-35.
O'NEIL, P 1992. Variationin male and female reproductivesuccess
among floral morphs in the tristylousplant Lythrumsalicaria
(Lythraceae).AmericanJournalof Botany 79: 1024-1030.
PIPER, J., AND B. CHARLESWORTH. 1986.
The evolution of distyly in
Primula vulgaris. Biological Journal of the Linnean Society 29:
123-137.
QUELLER,D. C. 1983. Sexual selectionin a hermaphroditic
plant.Nature305: 706-707.
SCHOEN, D. J., AND M. T. CLEGG. 1985. The influence of flower color
on outcrossingrateand male reproductivesuccess in Ipomoea purpurea. Evolution39: 1242-1249.
SNOW, A. A., AND P 0. LEWIS. 1993.
Reproductive traits and male
in plants:empiricalapproaches.Annual Review of Ecology
fertility
and Systematics24: 331-351.
STANTON, M. L., A. A. SNOW, AND S. N. HANDEL. 1986. Floral evo-
lution: attractivenessto pollinatorsincreases male fitness.Science
232: 1625-1627.
STONE,J. L. 1994. Pollen transferand the evolutionof distyly:testsof
a model. Ph.D. dissertation,State Universityof New York. Stony
Brook, NY.
, AND J. D. THOMSON. 1995. The evolution of distyly: pollen
transferin artificialflowers.Evolution48: 1595-1606.
SUTHERLAND,
S. 1986. Patternsof fruitset: what controlsfruit-flower
ratiosin plants?Evolution40: 117-128.
, ANDL. F DELPH. 1984. On the importanceof male fitnessin
plants: patternsof fruitset. Ecology 65: 1093-1104.
TAYLOR,C. M. 1991. The vascular flora of the La Selva Biological
Station,Costa Rica: Rubiaceae. Selbyana 12: 141-190.
WEBB, C. J., AND D. G. LLOYD. 1986. The avoidance of interference
betweenthe presentationof pollen and stigmasin angiosperms.II.
Herkogamy.New Zealand Journalof Botany 24: 163-178.
WILKINSON,L., M. HILL, S. MICELI, P HOWE, AND E. VANG. 1992. SYS-
TAT, Version5.2. Systat,Inc., Evanston,IL.
WILLSON,M. F 1979. Sexual selection in plants.AmericanNaturalist
113: 777-790.
, AND B. J. RATHCKE. 1974. Adaptive design of the floral display
in Asclepias syriaca L. AmericanMidland Naturalist92: 47-57.
WILSON, P, J. D. THOMSON, M. L. STANTON, AND L. P RIGNEY. 1994.
Beyond floralBatemania: genderbiases in selectionforpollination
success. AmericanNaturalist 143: 283-296.
WOLFE, L. M., AND S. C. H. BARRETT. 1989. Patterns of pollen removal
and depositionin tristylous
Pontederiacordata L. (Pontederiaceae).
Biological Journalof the Linnean Society 36: 317-329.
YOUNG, H. J., AND M. L. STANTON. 1990. Influences of floral variation
on pollen removaland seed productionin wild radish.Ecology 71:
536-547.
ZIMMERMAN,M., AND G. H. PYKE. 1988. Reproduction in Polemonium:
assessing the factorslimitingseed set. American Naturalist 131:
723-738.