Proc. Nati. Acad. Sci. USAVol. 91, pp. 3-10, January 1994

Review

Modes and origins of mechanical and ethological isolationin angiospermsVerne GrantDepartment of Botany, University of Texas, Austin, TX 78713

ABSTRACT Mechanical and etholog-ical isolation between species is wide-spread in angiosperms with specializedanimal-pollinated flowers, being recordedin 29 species groups belonging to 27 generaand 16 families. Mechanical isolation oc-curs in two forms. (i) The common type,designated the Salvia type, operates whentwo or more species of flowers are adaptedfor different groups of pollinators withdifferent body sizes and shapes. (it) In thePedicularis type two flower species havethe same species of pollinator but pick uppollen from different parts of the pollina-tor's body. Four forms of ethological iso-lation are recognized. ('l In the Aqulegiatype, which is widespread, ethological iso-lation is a side effect of mechanical isola-tion. (u) The flower-constancy type, as thename suggests, is based on flower-constantforaging behavior. (ii) In the Ophrys type,floral scents attract male bees or waspsand play a role in their mating behavior;different species of flowers, often orchids,have different scents and attract differentsets of hymenopteran species. (iv) Themonotropy type occurs in plants pollinatedby hymenopterans with species-specific orgroup-specific flower preferences for nu-tritive purposes (monotropic and oligotro-pic bees and fig wasps). Three modes oforigin of floral isolation are confirmed byevidence: (') mechanical and ethologicalisolation arising as a by-product of abo-patric speciation, (fi) ethological isolationdeveloping by selection for reproductiveisolation per se, and (Wi) mechanical isola-tion arising as a by-product of characterdisplacement. Mode of origin i accountsfor the Salvia and Aquilegia types of iso-lation in nine known species groups andfor the Ophrys type in one group. Mode oforigin u accounts for the flower-constancytype of ethological isolation in two speciesgroups. Mode of origin ui explains me-chanical isolation in two groups. Sympat-ric origin of floral isolation by hybridspeciation and by flower constancy hasbeen proposed, but these modes are un-documented and improbable.

The possibility that differences betweenrelated plant species in flower structuremay function as a mechanical isolatingmechanism was suggested by Dobzhan-sky (1, 2) and Stebbins (3) in the period

1937-1950. Building on these earlyworks, I showed that the preconditionsfor reproductive isolation at the stage ofpollination are widespread in an-giosperms with complex floral mecha-nisms (4).My early paper (4) recognized etho-

logical as well as mechanical isolation inangiosperms. Mechanical isolation canoccur when two or more plant specieshave different flower structures that re-duce or prevent interspecific pollina-tion. Ethological isolation takes placewhen specialized flower-visiting ani-mals make preferential visitations to onespecies of flower which they recognizeby its specific shape, color, markings,and/or scent. Mechanical and etholog-ical isolation are likely to be combinedin actual cases, making it useful to groupthem in a collective mode, floral isola-tion.

After 1950 opinion regarding floralisolation was divided between support(5) and skepticism (6). It was obviouslynecessary to document floral isolationbetween sympatric species in nature.This was done by my students and my-self in the next few years (1952-1964) inAquilegia (7), Penstemon (8), Pedicu-laris (9), and Salvia (10). In these casesthe floral isolation is primarily mechan-ical but is supplemented by ethologicalisolation.

Since that period a much larger body ofevidence has been obtained from moreplant groups. One of the purposes of thispaper is to present the broader data basethat now exists. The broader data basemakes it possible to recognize differentmodes of mechanical and ethological iso-lation. A second purpose of this review isto outline these modes.A question which remains problemat-

ical is the origin of floral isolation. Thissubject was discussed in a preliminaryway in my early paper (4), where bothallopatric and sympatric models wereconsidered, but the models were theo-retical. What is needed is a modernanalysis of the problem on the basis ofour present expanded data base and inthe light of our current understanding ofspeciation. Such an analysis is pre-sented here.

The breakdown ofthe diverse phenom-ena of floral isolation into different formsis helpful in relation to a discussion of theorigin of floral isolation. We find, asmight be expected, that no one mode ofevolution will account for every form offloral isolation in every plant group. Thusthis paper will examine the diverse ori-gins of floral isolation.

Broad Evidence for Floral Isolation /

Examples of floral isolation in nAturehave now been reported in many plantgroups in many geographical areas. Thecases known to me are listed in Table 1.These cases occur in 29 species groups in16 families. They leave little doubt thatfloral isolation is a real barrier to geneexchange between species in many ani-mal-pollinated plant groups.

Floral isolation depends on floral spe-cializations. It will be noted that all but afew of the cases in Table 1 occur, aswould be expected, in genera and fami-lies with complex floral mechanisms. TheScrophulariaceae and Orchidaceae to-gether account for 10 of the 29 knownexamples.One family with complex flowers that

is missing from Table 1 is the Asclepi-adaceae. Years ago it was cited as agroup in which mechanical isolation waswell developed (3, 4, 47). But later stu-dents have downgraded mechanical iso-lation in the genus Asclepias to a pro-cess of secondary or negligible impor-tance (48, 49). This is a group in whichthe early predictions were plausible buthave not stood up under closer inspec-tion. Since an expectation ofmechanicalisolation in the Asclepiadaceae is stillwarranted, further search for valid ex-amples is desirable.Mechanical isolation is the main mode

in many of the examples in Table 1.Ethological isolation is often complemen-tary to mechanical isolation (e.g., in Aq-uilegia, Polygala, Penstemon, Pedicu-laris), confirming the usefulness of thecollective mode, floral isolation. Etho-logical isolation plays a primary role inseveral groups (Hedysarum, Cercidium,Phlox, Ophrys).The strength of the floral isolation

varies over a wide range. It is very

3

Proc. Natl. Acad. Sci. USA 91 (1994)

Table 1. Groups of related species which are mechanically (M) and/or ethologically (E) isolated in nature, listed by familyMode of

Species group Area Pollinators isolation Refs.Ranunculaceae

Aquilegia formosa subgroup and A. caerulea W. North Americasubgroup

BerberidaceaeEpimedium grandiflorum, E. sempervirens, and E. Japan

trifoliatobinatumPapaveraceaePapaver rhoeas, P. dubium, and three other species Great Britain

LeguminosaeCassia leiophylla and C. bicapsularis

Cercidium floridum and C. microphyllumHedysarum boreale and H. alpinum

PolygalaceaePolygala vauthieri and P. monticola

OnagraceaeFuchsia encliandra and F. parviflora

EuphorbiaceaeDalechampia brownsbergensis and D. fragrans

BalsaminaceaeImpatiens capensis and I. pallida

PolemoniaceaePhlox pilosa and P. glaberrimaPhlox drummondii and P. glaberrima

Ipomopsis aggregata and I. tenuituba

SolanaceaeSolanum grayi and S. lumholtzianum

ScrophulariaceaeMimulus cardinalis and M. lewisii

Diplacus puniceus, D. longiflorus, and D. calycinus

Penstemon centranthifolius, P. grinnellii, and P.spectabilis

Pedicularis groenlandica and P. attollensRhinanthus minor and R. serotinus

LabiataeSalvia apiana and S. mellifera

Monarda didyma and M. clinopodia

CampanulaceaeLobelia cardinalis and L. siphilitica

HaemodoraceaeAnigozanthos manglesii and A. humilis

MusaceaeHeliconia umbrophila, H. irrasa, and other species

OrchidaceaeOphrys insectivora, 0. speculum, and other speciesOphrys fusca and 0. lutea

Platanthera bifolia and P. chloranthaStanhopea tricornis and S. bucephalusAngraecum compactum and Neobathiea

grandidierana (closely related though placed indifferent genera)

Mexico

CaliforniaAlaska

Brazil

Mexico

Surinam

E. North America

Illinois and IndianaIllinois (synthetic

population)W. North America

Mexico

W. North America

California

California

CaliforniaEurope

California

E. North America

E. North America

Australia

Costa Rica

Hummingbirds (A. f.),hawkmoths (A. c.)

Bees

M, E 7,11,12

E 13

Bumblebees, honeybees E(inferred)

Bumblebees and other largebees (C. b.), smallerPtiloglossa bees (C. 1.)

BeesBumblebees, Megachile bees

Bees

14

M 15

E 16E 17

M 18

Hummingbirds (F. e.),bumblebees (F. p.), wheresympatric

Male euglossine bees

Hummingbirds (I. c.),bumblebees (I. p.)

ButterfliesButterflies

Hummingbirds (I. a.),hawkmoths (I. t.)

Large bees (S. l.), small bees(S. g.), where sympatric

Hummingbirds (M. c.),bumblebees (M. l.)

Hummingbirds (D. p., D. l.),hawkmoths (D. c.)

Hummingbirds (P. c.), carpenterbees (P. g.), wasps (P. s.)

BumblebeesBumblebees

Carpenter bees (S. a.),medium-sized and small bees(S. m.)

Hummingbirds (M. d.),bumblebees (M. c.)

Hummingbirds (L. c.),bumblebees (L. s.)

Wattle birds (Meliphagidae)

Hermit and nonhermithummingbirds

Europe and North Africa Male bees and waspsAlgeria Male Andrena bees

EuropeEcuadorMadagascar

M, E 19

E 20

M 21

E 22, 23E 23-25

M, E 12, 26

M 27

M

M(inferred)M, E

M, EM, E

28

29, 30

8

9, 31, 3233

M, E 10

M, E 34

M, E 35, 36

M, E 37

M 38, 39

M, EM, E

(inferred)M, EM, EM, E

MothsEulaema beesHawkmoths

40, 4142

4344, 4546

4 Review: Grant

Proc. Natl. Acad. Sci. USA 91 (1994)

strong in Ophrys (41), Mimulus (28), andPedicularis (9, 31, 32). In Aquilegia andIpomopsis, on the other hand, the floralisolation in itself brings about only areduction in interspecific gene ex-change; however, it does not act alonebut rather in combination with ecologi-cal and seasonal isolation, and the en-semble of all external barriers is highlyeffective (12).

Types of Mechanical Isolation

Mechanical isolation occurs in twoforms. In the first, which we will call theSalvia type for ease of reference, thecontrasting plant species have flowersadapted to different kinds of pollinatinganimals with bodies of different size andshape, such as bees vs. birds or large beesvs. small bees. In the second type, to bedenoted the Pedicularis type, the con-trasting plant species are pollinated bythe same species of animal, but the floralmechanism deposits and picks up thepollen on different parts of the pollina-tor's body. The floral differences respon-sible for mechanical isolation residemainly in the pollination mechanism inthe Pedicularis type. In the Salvia type,on the other hand, all or nearly all fea-tures of.the floral mechanism, includingthose involved in attraction, reward, andexclusiveness as well as pollen delivery,come into play.The Salvia type of mechanical isola-

tion is illustrated by Salvia apiana andSalvia mellifera. The flowers of thesespecies differ in size and conformation.S. apiana is adapted for and pollinatedby large carpenter bees (Xylocopa) andS. mellifera by various medium-sizedand small bees (Anthophora, Osmia,etc.). Each Salvia species is visited andpollinated by its normal set of bees withonly rare exceptions. Conversely, thefloral differences between the speciesset up a barrier to successful visitationand pollination of S. apiana by thesmaller bees and of S. mellifera by thelarge bees (10).Another example is the hummingbird-

pollinated Mimulus cardinalis and bum-blebee-pollinated Mimulus lewisii. Theseinterfertile species occur sympatrically insome sites in the Sierra Nevada, CA. Nonatural hybrids are found. Progeny testsof seeds collected on open pollinatedplants in sympatric colonies yielded nohybrids. Cross-pollination between thetwo species is also extremely rare in theexperimental garden (28).Other examples of the Salvia type of

isolation are found in Aquilegia, Fuchsia,Impatiens, Ipomopsis, Penstemon,Monarda, and Lobelia (Table 1).The Pedicularis type of mechanical

isolation is illustrated by the species pairPedicularis groenlandica and Pedicu-laris attollens. Both species have bum-

blebee flowers and are pollinated by thesame species ofBombus (B. bifarius andB. flavifrons). However, the differencesbetween the Pedicularis species in theirpollination mechanism are such as tobring about dorsal (nototribe) pollina-tion in P. groenlandica and foreheadpollination in P. attollens (9, 31, 32).The Pedicularis type is found also inPolygala (18), Rhinanthus (33), and Hel-iconia (38, 39).

The Aquilegia and Flower-ConstancyTypes of Ethological Isolation

Four forms of ethological isolation canbe recognized. First, ethological isola-tion may be a side effect of the Salviatype of mechanical isolation. If speciesA is adapted for hummingbird pollina-tion and species B for hawkmoth polli-nation, the birds may be unable to for-age effectively on species B. They as-sess the difficulties and shun species Bif an adequate supply of hummingbirdflowers is available. The hawkmothsmake a parallel adjustment in relation tospecies A. Examples of this type ofethological discrimination are found inspecies groups in Aquilegia and Ipo-mopsis which contaid hummingbird-pollinated species and hawkmoth-pollinated species (12, 50). For ease ofreference we will call this commonmode the Aquilegia type of ethologicalisolation.A second type of ethological isolation

is a product of flower constancy such asoccurs in bees, lepidopterans, and to alesser extent in long-tongued flies. Indi-vidual insects become fixed on one kindof flower, which they recognize by itsspecific color, shape, markings, orscent, and then forage preferentially onthat flower type during a succession ofvisitations. Flower-constant behaviorincreases the foraging efficiency of theinsects when visiting flowers with com-plex floral mechanisms, since once anindividual has learned how to work agiven floral mechanism it can workother flowers of the same type rapidly.As a corollary of this, flower constancyis a behavioral tendency, not an abso-lute rule. If it does not improve foraging,as when one type of flower is present ininadequate numbers, the insects switchto another and better source. The resultof interest to us here is that differentindividuals of the same bee species maybe foraging in the same two-species col-ony of plants, but some individual beesvisit exclusively or primarily one spe-cies of flower, while other individualbees visit the other flower species. (Formore on flower constancy see refs. 51-54.)Cases of ethological isolation based

on flower constancy are well known inthe experimental garden and were re-

ported in the early years in Antirrhinum(55), Gilia (4), Clarkia (56), Papaver(14), and Lantana (57), and more re-cently in artificial flowers (58). Flower-constant pollination separating speciesin nature has since been found in Epi-medium, Cercidium, Hedysarum,Phlox, Pedicularis, Rhinanthus, Platan-thera, and Angraecum/Neobathiea (Ta-ble 1).

Pedicularis groenlandica and P. at-tollens grow and flower together in thesame alpine meadows in the Sierra Ne-vada, CA. Both species are pollinatedby bumblebees but are mechanicallyisolated, as noted earlier. The mechan-ical isolation is reinforced by strongethological isolation based on flowerconstancy. Individuals ofBombus bifar-ius distinguish the two Pedicularis spe-cies by their floral features (Fig. 1).Some individual bumblebees visit P.groenlandica selectively, bypassingplants of P. attollens in their rounds,while other individual bees forage ex-clusively on P. attollens (9, 32). Lapsesof flower constancy in which an individ-ual bumblebee crosses over from onePedicularis species to the other occurbut are rare (9, 32).

The Ophrys and Monotropy Types ofEthological Isolation

In the third type of ethological isolation,the Ophrys type, floral scents attractmale bees or wasps and play a role intheir mating behavior. Two subtypes arerecognized.

In the genus Ophrys the flowers sim-ulate female hymenopterans in their vi-sual and olfactory features, and theysexually stimulate male bees (Andrena,Eucera, etc.) and wasps (Vespidae,Sphecidae, Scoliidae), which land onand attempt to copulate with the flow-ers. Each species of Ophrys has itsdistinctive floral markings and scent.Different species of Ophrys attractmainly different sets of hymenopteran

A B

I cmFIG. 1. Front view ofcorolla oftwo related

species of Pedicularis, showing lower lip, up-per hood, and beak extending outward andupward from hood. The stigma protrudesthrough the tip of the beak. The beak is dis-placed to the left in B to show the two purplespots. (A) P. groenlandica. (B) P. attollens.

Review: Grant

Proc. Natl. Acad. Sci. USA 91 (1994)

pollinator species. The males of one orseveral hymenopteran species have amating preference for one kind ofOphrys flower only, while males ofother pollinator species are attractedsexually to a different kind of Ophrysflower. The Ophrys flowers thus enlistthe species-specific mating preferencesof male hymenopterans in the service ofspecies-specific pollination. Some me-chanical isolation is associated with theethological isolation (41).Male bees of the genera Euglossa, Eu-

laema, etc. (Apidae, tribe Euglossini) inthe American tropics visit flowers of or-chids and other plant families for per-fume. They collect liquid droplets of ar-omatic substances (terpenes, etc.) fromthe scent glands of the flowers, storethem by capillarity on their hind legs, anduse them, apparently in a transformedstate, as sex pheromones (20, 45, 59-61).A given orchid species often attracts onlyone or a few species ofmale bees, and thesets of bee pollinators differ from oneorchid species to the next (44, 45, 60).Two sympatric species of the euphorbgenus Dalechampia attract different setsof male euglossine bees in nature. Thisdifferential attraction and pollinationholds up to a statistically significant ex-tent in an experimental mixture of theDalechampia species (20).The fourth type of ethological isola-

tion, designated the monotropy type, in-volves monotropy and related conditionsin hymenopterans. In many genera ofbees, especially among solitary bees, thespecies restrict their foraging to a singletaxonomic group of plants. The plantgroup may be a tribe, genus, speciesgroup, or species. Monotropic bee spe-cies collect nectar and pollen from asingle plant species; oligotropic bees col-lect flower food from a supraspecificgroup. Monolectic and oligolectic beesvisit a species or larger group, respec-tively, for pollen, but they may foragemore widely for nectar. Ten species ofPerdita (Halictidae) collect pollen fromCalochortus (Liliales) and mostly fromdifferent species or species groups. ThusPerdita calochorti collects pollen fromCalochortus nuttallii, Perdita leuco-stoma from Calochortus leichtlinii, andPerdita californica from Calochortussplendens, Calochortus kennedyi, andCalochortus concolor (62). Among otherbee genera with monotropy/oligotropyare Andrena (Andrenidae) and Diadasia(Anthophoridae) (62-66).The terminology of monotropy, oligot-

ropy, etc. is applied to bees, but perhapsthe best example of flower host specific-ity is provided by the fig wasps (Aga-onidae). Many species of Ficus have asingle species of wasp pollinator; someFicus species have two species of wasppollinators, and the wasp species usuallydiffer from one Ficus species to another

(67-69). One would predict ethologicalisolation based on the floral preferencesof the fig wasps.The relation between monotropy/

oligotropy in bees and ethological iso-lation in plants is complex and littlestudied. Monotropy and oligotropy ob-viously channel pollen distributionwithin taxonomic limits in a plant com-munity. However, this does not neces-sarily bring about ethological isolationbetween congeneric species. Oligotro-pic bees are group specific but not spe-cies specific, and they might or mightnot discriminate between related plantspecies in a sympatric colony. Further-more, monotropic bees are often not theprimary pollinators of the plants theyvisit, their host flowers being pollinatedmainly by other bees or other insects.The crucial factor is then the flower-visiting behavior of the primary pollina-tors. If these are not species-specific,floral isolation does not exist, andmonotropy in the secondary pollinatorsis more or less irrelevant. These arelimitations on the potential of monot-ropy/oligotropy for isolation.

After allowing for these factors, thepossibility remains that monotropy/oligotropy in bees may contribute posi-tively to ethological isolation in somesituations. Oligotropic Andrena speciesforaging in mixed colonies of Oenotherahave been observed to visit someOenothera species more often than oth-ers, but they also cross over from onespecies to another (66).The question of ethological isolation

by monotropic bees and fig wasps re-quires further study.

The Modes of Ethological IsolationSummarized

The phenomena of flower-visiting behav-ior that result in ethological isolation canbe grouped in several modes. It hasseemed best to me to group them in thefour modes presented above, but alter-native groupings into three or five typesare also possible. Each mode is charac-terized by a certain predominant patternof flower-visiting behavior. It may beuseful to summarize briefly these behav-ior patterns here.

In the Aquilegia type of ethologicalisolation, pollinators visit preferentially aflower species on which they can foragesuccessfully, but shun a plant species inthe same colony in which the floral re-ward is inaccessible or difficult of access.Put another way, the pollinators exhibitpreference for a plant species with aflower structure which permits a highratio of foraging benefits to foragingcosts, and they discriminate against aflower species with a low benefit-to-costratio in foraging. In the flower-constancymode of ethological isolation, individual

foragers of a polytropic insect speciesbecome fixed on one flower type in amixed plant colony, even though theycould forage successfully on the othertypes of flowers. In the Ophrys andmonotropy modes ofethological isolationa whole species of insect has a preferencefor a particular species of flower. Theinsects are foraging female hymenopter-ans in the monotropy mode, and sexuallyinterested male hymenopterans in theOphrys mode.

Origin of Floral Isolation as aBy-Product of Allopatric Speciation

Reproductive isolating mechanisms arisein two ways: as by-products of diver-gence and as products of selection forreproductive isolation per se. The firstprocess is general and ubiquitous. Thesecond process comes into play underspecial conditions as a reinforcement ofthe reproductive isolation produced bythe first mode. It affects mainly premat-ing barriers (5, 70, 71). Both processescan be expected to play a role in thedevelopment of floral isolation. We canexpect further to find cases in which thefirst process has acted alone and othercases in which both processes have beeninvolved.

Floral isolation between humming-bird-pollinated and hawkmoth-polli-nated species of Aquilegia and Ipomop-sis can be explained as a by-product ofdivergence without invoking selectionfor reinforcement (30). The floral andinflorescence characters which bringabout the floral isolation in these twogroups are adaptive in relation to onetype of pollinator or the other and aremaladaptive in the opposite flower-pollinator combination. There are nofloral features in these plant groups thatpoint to selection for isolation. In fact,this process can be ruled out. If selec-tion for isolation had played a role in thedevelopment of floral isolation in Aqui-legia and Ipomopsis, the floral isolationshould be enhanced in zones where theornithophilous and sphingophilous spe-cies hybridize as compared with hybrid-free areas. The actual pattern is theopposite; floral isolation is diminishedin the hybrid zones (30).The pattern in Diplacus is parallel in

most respects to that in Aquilegia andIpomopsis with one exception. We do nothave enough evidence to say whether theornithophilous and sphingophilous spe-cies of Diplacus are florally isolated toany significant extent. Otherwise thefacts and conclusions are the same as forAquilegia and Ipomopsis. The orni-thophilous and sphingophilous charac-ters ofcontrasting species inDiplacus areby-products of divergence. Selection forisolation has played no detectable role inthis development (30).

6 Review: Grant

Proc. Natl. Acad. Sci. USA 91 (1994) 7

The next question is the mode ofspeciation involved in divergence of thecontrasting ornithophilous and sphin-gophilous species ofAquilegia, Ipomop-sis, and Diplacus. Diverse modes ofspeciation are known in plants (71, 72).Most of these can be ruled out in thespecies groups in Aquilegia, Ipomopsis,and Diplacus (30). This reduces the pos-sibilities to allopatric speciation on thediploid level. The biogeographical pat-terns are consistent with a course ofdivergence by allopatric speciation (30).There are two modes of allopatric

speciation: geographical speciation andquantum (or peripatric) speciation. Inthe first mode the divergence passesthrough an intermediate stage of spa-tially isolated geographical races; in thesecond mode the intermediate stage is aperipherally isolated local race. The twomodes thus differ in the pathway fol-lowed at the racial level of divergence.In dealing with mature species groups,in which the speciation process has runto completion, it is usually difficult if notimpossible to determine which pathwaywas taken in the historically past stageof racial divergence. Consequently, inmost actual species groups, we do nothave sufficient evidence to distinguishbetween the two modes of allopatricspeciation (71).This is as true in Aquilegia, Ipomopsis,

and Diplacus as it is in most other groups.However, one factor favors the pathwayof geographical speciation in Aquilegia,Ipomopsis, and Diplacus. The contrast-ing ornithophilous and sphingophilousspecies in these genera are pollinated byhighly motile animals with substantialfood requirements and wide foragingranges. A small isolated local race willprobably be an inadequate resource basefor the development of either an ornitho-philous or sphingophilous pollinationsystem. However, geographical racescan become pollination races which areadapted for one set of pollinators in anarea where those pollinators are abun-dant and for a different set in anotherarea. Such geographical/pollinationraces are known (73). It seems likely thatthe ornithophilous and sphingophilousspecies in Aquilegia, Ipomopsis, andDiplacus went through an intermediatestage of geographical/pollination races(30).Comparable evidence for six other

groups in temperate North America-Impatiens, Mimulus, Penstemon, Salvia,Monarda, and Lobelia (Table 1)-suggests that floral isolation in thesegroups is also a by-product of allopatricspeciation. The speciation is on the dip-loid level in the first five groups, as it isin Aquilegia, Ipomopsis, and Diplacus,and is on a tetraploid homoploid level inMonarda.

The nine plant groups mentioned inthis section all have the Salvia type ofmechanical isolation combined with theAquilegia type of ethological isolation.

Origin of Floral Isolation in Ophrys

We can speculate about the origin offloral isolation in Ophrys. Ophrys racesand species living in different areasprobably evolve visual and olfactoryfloral features that are attractive tomales of hymenopteran species in thoseareas, and along with these, also evolvestructural characters fitting the bodyparts of the local hymenopteran pollina-tors. The particular floral charactersand the corresponding species of malehymenopterans originally have an allo-patric distribution, but may becomesympatric as a result of range exten-sions. The floral isolation would then bea by-product of allopatric divergence offloral characters in relation to the spe-cies-specific mating preferences of themale bees and wasps. This hypothesisstates in effect that the origin of floralisolation in Ophrys is a special case offloral isolation arising as a by-product ofallopatric speciation as, discussed in thepreceding section.

Selective Origin of Ethological Isolation

If interspecific hybridization betweentwo plant species results in wastage oftheir reproductive potential, and if thisloss of reproductive potential is selec-tively disadvantageous, selection is ex-pected to build up barriers to hybridiza-tion (5, 70-72). The conditions neces-sary for bringing selection forreproductive isolation into play are notpresent in many plant groups. Corre-spondingly, we find no evidence of theaction of selection for isolation in Aqui-legia, Ipomopsis, Diplacus, and othergroups, as pointed out previously. Insome groups, however, the essentialpreconditions may exist, and selectionfor reproductive isolation can be ex-pected to go into action.A barrier to hybridization that can be

built up by selection for isolation, if thepreconditions exist, is the flower-constancy type of ethological isolation.This mode of isolation is based on spe-cies-specific recognition features in flow-ers pollinated by flower-constant beesand lepidopterans. Visual and olfactoryfloral features that differ between plantspecies and are recognized by flower-constant bees and lepidopterans arewidespread (4). It is probable that someofthese recognition features are productsof selection for ethological isolation.However, we have little independent ev-idence to support this plausible sugges-tion.

I previously described the ethologicalisolation between Pedicularis groen-landica and Pedicularis attollens basedon flower constancy of bumblebees.The flower constancy is based in turn onfloral differences between the two spe-cies which the bees recognize, differ-ences that include size and shape offloral parts, markings, and scent (Fig.1). Some of these character differences,such as the length and shape of the beak,are functional in the pollination mecha-nism; these characters are set aside forthe present discussion. But other char-acters serve no apparent primary func-tion in pollination. P. attollens has twolarge purple spots in the upper front partof the hood; P. groenlandica lacks suchspots but has a purple area in the lowerbasal part of the beak (Fig. 1). Thesepurple markings are not nectar guides;the flowers are nectarless. It seemslikely that they are species-specific rec-ognition marks for the bumblebees, andthey may have been developed by se-lection for ethological isolation. Thesame suggestion can be made for thedifferences in floral scent between thetwo species.Phlox glaberrima and Phlox pilosa

have overlapping ranges in eastern NorthAmerica. Their usually pink flowers arepollinated by butterflies. A white-flowered morph occurs in P. pilosa. Thismorph is rare in most parts of the area ofP. pilosa but is predominant in the zoneof sympatric overlap with pink-floweredP. glaberrima. Here it would of coursefavor species discrimination by butter-flies. Analysis of pollen grains on floralstigmas shows that butterflies do transfersignificantly less pollen between pink P.glaberrima and white P. pilosa than be-tween P. glaberrima and pink P. pilosa.Production of hybrid seeds ofP. pilosa 9x P. glaberrima is also reduced in thecombination of white P. pilosa x P.glaberrima as compared with pink P.pilosa x P. glaberrima. The flower colorin the white morph of P. pilosa thus hasa selective advantage and enhances etho-logical isolation under conditions of sym-patry between P. pilosa and P. glaber-rima (22, 23).

Origin of Mechanical Isolation as aBy-Product of Character Displacement

Two species with overlapping distribu-tion areas are often more highly differen-tiated in their sympatric zone than in theirallopatric areas. The morphological andphysiological characters involved usuallyrelate to secular or nonreproductive ecol-ogy. The differentiation in the sympatriczone (character displacement) is gener-ally explained as a result of interspecificcompetition and selection for divergencewith respect to secular factors in theenvironment. However, the same pro-

Review: Grant

Proc. Natl. Acad. Sci. USA 91 (1994)

cess of selection for ecological diver-gence can also come into play in the stageof pollination and bring about reproduc-tive character displacement.A case of reproductive character dis-

placement is found in the species Sola-num lumholtzianum and Solanum grayi,which occur in Arizona and Mexico withan overlap in their ranges. In their allo-patric areas both species have largeflowers and are pollinated by large bees(Bombus, Xylocopa, etc.); in the sym-patric area S. lumholtzianum continuesto have large flowers, whereas the flow-ers of S. grayi are small and are polli-nated by small bees (Nomia, etc.). Me-chanical isolation occurs in the sympat-ric zone. It appears to be a product ofselection for ecological divergence inrelation to pollinators (27). A parallelcase of reproductive character displace-ment occurs in the species pair Fuchsiaencliandra and Fuchsia parviflora (19).

Selection for ecological divergenceand selection for reproductive isolationas discussed in the preceding section aredifferent selection modes. However,they both take place in a sympatric field,and both can produce floral isolation.The floral isolation develops in the areaof sympatric overlap of two species. Itmay be difficult to say which mode isoperating in any given case. Indeed,both modes could well be operatingsimultaneously in the same speciespair.

Effects of Hybridization

A pair of florally isolated species is oftenreproductively isolated in one area ofsympatric contact but hybridizes in an-other area. This pattern has been found inAquilegia, Epimedium, Phlox, Ipomop-sis, Diplacus, Penstemon, Rhinanthus,Salvia, Heliconia, Ophrys, and Platan-thera (Table 1 and references cited there-in). What are the later-generation effectsof the hybridization? There are severalpossibilities.Let us restrict this discussion to floral

isolation composed of the Salvia type ofmechanical isolation combined with theAquilegia type of ethological isolation.The contrasting plant species have dif-ferent kinds of pollinating animals. Me-chanical isolation between them de-pends on structural differences andethological isolation depends on signal-ing differences in their flowers. Hybrid-ization generally produces floral char-acters in the intermediate range. Indi-vidual F1 hybrids with intermediatefloral characters bridge the gap betweenthe parental species for the pollinators,and the floral isolating barrier breaksdown. The extent and direction of theevolutionary changes in the hybrid pop-ulation in later generations are deter-

mined among other factors by the pol-linators themselves.The pollinators exert selective pres-

sures on the hybrid population. If thenormal pollinators of parental species Aare abundant, active, and effective inthe area of the hybrid population, andthe normal pollinators of species B areless so, the later-generation progeny ofthe hybrids can be expected to reverttoward the characters of species A. Or ifthe two different sets of normal pollina-tors are approximately the same in num-ber of flower visits and pollination ef-fectiveness, their combined selectivepressures should produce later-genera-tion derivatives with intermediate floralcharacters suited for both types of pol-linators (74). Species groups in Aquile-gia, Ipomopsis, and Diplacus have pop-ulations and population systems of hy-brid derivation that exemplify thesealternative courses of development (74).

Hybridization between florally iso-lated but interfertile species can also beexpected to engender some recombina-tion types for floral characters in F2 andlater generations. Most recombinationtypes will probably have a floral mecha-nism that is not well ada,pted for anyavailable pollinator, but this is not theonly possible fate of a new type, as wewill see.

Origin of Floral Isolation by HybridSpeciation

One of the known modes of speciation inplants is hybrid speciation with the seg-regation of external barriers. Most exam-ples involve secular ecological isolatingbarriers (72). We are interested here in apossible parallel process involving floralisolating barriers.Straw (75) suggested that some par-

ticular recombination types for floralcharacters in a hybrid population mightbe preadapted for a flower-visiting ani-mal that occurs in the area of the hybridpopulation and is different from the nor-mal pollinators of the parental species.The new flower type and new pollinatorcould then develop into a separate iso-lated species with a distinctive pollina-tion system of its own.Straw (75) described a case in a dip-

loid species group in Penstemon whichseemed to conform to this model. Thehummingbird-pollinated Penstemoncentranthifolius and carpenter-bee-pollinated Penstemon grinnellii hybrid-ize sporadically. Some hybrid progenyresemble a third species, Penstemonspectabilis, which is pollinated bypseudomasarid wasps. Furthermore, P.spectabilis is ecologically and morpho-logically intermediate between P. cen-tranthifolius and P. grinnellii. Straw(75) suggested, therefore, that P. spec-tabilis originated as a diploid hybrid

segregate from P. centranthifolius x P.grinnellii which captured a new anddifferent pollinator and developed into anew florally isolated species.

This hypothesis has been widely cited.I accepted it years ago. However, thepremise of a hybrid origin of Penstemonspectabilis is not supported by strongevidence and is inconsistent with someold and some recent evidence. First, P.spectabilis is a member of a group of fivespecies (Penstemon pseudospectabilis,etc.) with a widespread distribution frominterior southern California to Arizona.This group represents a branch of Pens-temon commensurate with the P. cen-tranthifolius and P. grinnellii groups. Analternative hypothesis which should beconsidered is that P. spectabilis origi-nated in the common way, by allopatricspeciation on the diploid level, and thatits resemblance to some hybrid progenyof P. centranthifolius x P. grinnellii iscoincidental.

Second, a recent allozyme study ofthisgroup of penstemons by Wolfe and Elis-ens (76) has identified enzyme allelescharacteristic of P. centranthifolius andof P. grinnellii. These alleles are notpresent in P. spectabilis in the frequen-cies expected ifP. spectabilis is a productof hybrid speciation of P. centranthifo-lius x P. grinnellii. According to Wolfeand Elisens (76) the allozyme evidence isconsistent with allopatric speciation andsubsequent introgression in the group butnot with any recent events of hybridspeciation.The formation of a new floral isolating

barrier by hybridization and recombina-tion requires a rare coincidence in spaceand time between a novel flower type anda new available pollinator. This is possi-ble but must be very rare in nature, andits occurrence remains to be demon-strated.

Sympatric Origin by Flower-ConstantPollination

Flower-constant behavior in pollinatinganimals brings about positive assortativemating in a plant population. Assortativemating can theoretically lead to sympat-ric speciation. Models of sympatric spe-ciation resulting from flower-constantpollination have been considered (4, 58).The flower characters used as cues byflower-constant pollinators could start aspolymorphic variants in the populationand go on to develop into species-distinguishing characters within a sym-patric field.

This is a theoretical possibility which isbeset with great theoretical difficulties.The model requires complete assortativemating, which in turn requires completeflower constancy over a sequence offlowering seasons. But flower constancyis subject to lapses, especially when one

8 Review: Grant

Proc. Natl. Acad. Sci. USA 91 (1994) 9

type of flower becomes an inadequatesource of food. The lapses can occur inany flowering season or plant generationand prevent the sympatric speciationprocess from going through to comple-tion.

Evidence for the occurrence of thismode of sympatric speciation in plants islacking. However, the possibility war-rants further study, as Wells et al. (58)suggest.

Conclusions

The Salvia type of mechanical isolationis often combined with the Aquilegiatype of ethological isolation. The evi-dence indicates that this common pat-tern of floral isolation is a product ofallopatric speciation in Aquilegia, Im-patiens, Ipomopsis, Mimulus, Diplacus(with certain reservations), Penstemon,Salvia, Monarda, and Lobelia.The unique type of ethological and

mechanical isolation found in the genusOphrys may also develop by allopatricspeciation, although this question re-quires further study.

Selection for reproductive isolationper se will account for many cases ofethological isolation of the flower-constancy type. The evidence that itdoes so-is largely circumstantial. Visualand olfactory features that distinguishrelated species, and have no other dis-cernable function, are widespread inbee-pollinated and lepidopteran-polli-nated angiosperms. Detailed evidencesupporting a selective origin of etholog-ical isolation, however, is available inonly a few cases (mainly Phlox), andmore such evidence is needed.Where plant species compete for a

limited supply ofpollinators in a commonsympatric area, selection for ecologicaldivergence may go into action and pro-mote a partitioning of the pollinators bysize and type. Mechanical isolation is abyproduct ofthis process in Solanum andFuchsia.Two sympatric modes of origin of flo-

ral isolation have been proposed and dis-cussed in the literature. One is hybridspeciation followed by the segregation ofa new flower form that captures a newtype of pollinator. The other is the sep-aration of a polymorphic variant for floralcharacters from other members of itspopulation by flower-constant behaviorand assortative mating. These modes aretheoretically possible but also theoreti-cally difficult, improbable, and unknownin any actual plant group.Many questions remain. It is not clear

how the Pedicularis type of mechanicalisolation originates. Also unclear is themode of origin of ethological isolation inorchids and other plant groups pollinatedby perfume-collecting euglossine bees.The contribution of monotropy/oligo-

tropy to floral isolation is poorly under-stood. There is good evidence for phylo-genetic coevolution between Ficus andfig wasps, but we know little about theorigin ofwasp-flower relationships at thespecies level in Ficus.

I thank Billie L. Turner, Ethan J. Temeles,and Karen A. Grant for reading the manuscriptand making helpful suggestions.

1. Dobzhansky, T. (1937) Genetics and theOrigin ofSpecies (Columbia Univ. Press,New York), 1st Ed.

2. Dobzhansky, T. (1941) Genetics and theOrigin ofSpecies (Columbia Univ. Press,New York), 2nd Ed.

3. Stebbins, G. L. (1950) Variation and Ev-olution in Plants (Columbia Univ. Press,New York).

4. Grant, V. (1949) Evolution 3, 82-97.5. Dobzhansky, T. (1951) Genetics and the

Origin ofSpecies (Columbia Univ. Press,New York), 3rd Ed.

6. Werth, E. (1955) Ber. Dtsch. Bot. Ges.68, 163-166.

7. Grant, V. (1952) Aliso 2, 341-360.8. Straw, R. M. (1956) Am. Nat. 90, 47-53.9. Sprague, E. F. (1962) Aliso 5, 181-209.

10. Grant, K. A. & Grant, V. (1964) Evolu-tion 18, 196-212.

11. Grant, V. (1976) Evolution 30, 625-628.12. Grant, V. (1992) Pros. Natl. Acad. Sci.

USA 89, 11828-11831.13. Suzuki, K. (1984) Bot. Mag. (Tokyo) 97,

381-396.14. McNaughton, I. H. & Harper, J. L.

(1960) New Phytol. 59, 15-26.15. Delgado, A. 0. & Sousa, M. (1977) Bol.

Soc. Bot. Mex. 37, 5-52.16. Jones, E. E. (1978) Evolution 32, 189-

198.17. McGuire, A. D. (1993) Am. J. Bot. 80,

147-152.18. Brantjes, N. B. M. (1982) Plant Syst.

Evol. 141, 41-52.19. Breedlove, D. E. (1969) Univ. Calif.

Publ. Bot. 53, 1-69.20. Armbruster, W. S., Herzig, A. L. &

Clausen, T. P. (1992) Am. J. Bot. 79,1374-1381.

21. Wood, C. E. (1975) J. Arnold Arbor.Harv. Univ. 56, 413-426.

22. Levin, D. A. & Schaal, B. A. (1970) Am.Nat. 104, 273-283.

23. Levin, D. A. (1978) Evol. Biol. 11, 185-317.

24. Levin, D. A. & Kerster, H. W. (1967)Evolution 21, 679-687.

25. Levin, D. A. (1969) Evolution 23, 444-455.

26. Grant, V. & Wilken, D. H. (1988) Bot.Gaz. (Chicago) 149, 213-221.

27. Whalen, M. D. (1978) Syst. Bot. 3, 77-86.

28. Hiesey, W. M., Nobs, M. A. & Bjork-man, 0. (1971) Experimental Studieson the Nature ofSpecies (Carnegie Inst.of Washington, Washington, DC), Vol.5.

29. Beeks, R. M. (1962) Aliso 5, 83-122.30. Grant, V. (1993) Proc. NatI. Acad. Sci.

USA 90, 7729-7730.31. Macior, L. W. (1977) Bull. Torrey Bot.

Club 104, 148-154.

32. Grant, V. (1994) Biol. Zentralbl. 113, inpress.

33. Kwak, M. M. (1978) Taxon 27, 145-158.34. Whitten, W. M. (1981) Am. J. Bot. 68,

435-442.35. Rosatti, T. J. (1986) J. Arnold Arbor.

Harv. Univ. 67, 1-64.36. Johnston, M. 0. (1991) Evolution 45,

1468-1479.37. Hopper, S. D. & Burbidge, A. H. (1978)

Aust. J. Bot. 26, 335-350.38. Stiles, F. G. (1975) Ecology 56, 285-301.39. Stiles, F. G. (1979) Brenesia 15 (Suppl.),

151-180.40. Heslop-Harrison, J. (1953) New Con-

cepts in Flowering-Plant Taxonomy(Heidemann, London).

41. Kullenberg, B. (1961) Zool. Bidr. Upp-sala 34, 1-340.

42. Stebbins, G. L. & Ferlan, L. (1956) Ev-olution 10, 32-46.

43. Nilsson, L. A. (1983) Bot. J. LinneanSoc. 87, 325-350.

44. Dodson, E. H. & Frymire, G. P. (1961)Ann. Mo. Bot. Gard. 48, 137-172.

45. Van der Pijl, L. & Dodson, C. H. (1966)Orchid Flowers: Their Pollination andEvolution (Univ. Miami Press, Coral Ga-bles, FL).

46. Nilsson, L. A., Jonsson, L., Ralison, L.& Randrianjohany (1987) Biotropica 19,310-318.

47. Holm, R. W. (1950) Ann. Mo. Bot. Gard.37, 477-560.

48. Woodson, R. E. (1954) Ann. Mo. Bot.Gard. 41, 1-211.

49. Kephart, S. R. & Heiser, C. B. (1980)Evolution 34, 738-746.

50. Grant, V. & Temeles, E. J. (1992) Proc.Natl. Acad. Sci. USA 89, 9400-9404.

51. Grant, V. (1950) Bot. Rev. 16, 379-398.52. Kugler, H. (1970) Blutenokologie (Fisch-

er, Stuttgart, Germany), 2nd Ed.53. Free, J. B. (1970) Insect Pollination of

Crops (Academic, London).54. Heinrich, B. (1979) Bumblebee Econom-

ics (Harvard Univ. Press, Cambridge,MA).

55. Mather, K. (1947) Heredity 1, 175-186.56. Lewis, H. & Lewis, M. (1955) Univ.

Calif. Berkeley Publ. Bot. 20, 241-392.57. Dronamraju, K. R. (1960) Nature (Lon-

don) 186, 178.58. Wells, H., Wells, P. H. & Smith, D. M.

(1983) J. Apic. Res. 22, 33-44.59. Vogel, S. (1966) Osterreich. Bot.

Zeitschr. 113, 302-361.60. Dressler, R. L. (1981) The Orchids: Nat-

ural History and Classification (HarvardUniv. Press, Cambridge, MA).

61. Dressler, R. L. (1982) Ann. Rev. Ecol.Syst. 13, 373-394.

62. Linsley, E. G. (1958) Hilgardia 27, 543-599.

63. Linsley, E. G. & MacSwain, J. W.(1958) Evolution 12, 219-223.

64. Linsley, E. G. & MacSwain, J. W.(1959) Univ. Calif. Berkeley Publ. Ento-mol. 16, 1-46.

65. Linsley, E. G., MacSwain, J. W. &Raven, P. H. (1963) Univ. Calif. Berke-ley Publ. Entomol. 33, 1-58.

66. Linsley, E. G., MacSwain, J. W. &Raven, P. H. (1964) Univ. Calif. Berke-ley Publ. Entomol. 33, 59-98.

67. Wiebes, J. T. (1963) Tijdschr. Entomol.106, 1-112.

Review: Grant

10 Review: Grant Proc. Natl. Acad. Sci. USA 91 (1994)

68. Wiebes, J. T. (1979) Annu. Rev. Ecol.Syst. 10, 1-12.

69. Ramirez, W. (1970) Evolution 24, 680-691.

70. Dobzhansky, T. (1970) Genetics of theEvolutionary Process (Columbia Univ.Press, New York).

71. Grant, V. (1991) The Evolutionary Pro-cess (Columbia Univ. Press, New York),2nd Ed.

72. Grant, V. (1981) Plant Speciation (Co-lumbia Univ. Press, New York), 2ndEd.

73. Grant, V. & Grant, K. A. (1965) Flower

Pollination in the Phlox Family (Colum-bia Univ. Press, New York).

74. Grant, V. (1993) Proc. Natl. Acad. Sci.USA 90, 990-993.

75. Straw, R. M. (1955) Evolution 9, 441-444.76. Wolfe, A. D. & Elisens, W. J. (1993)

Am. J. Bot. 80, 1082-1094.