The Ecology ofButterflies in Britain

Edited byRoger L H. Dennis

Figures prepared byDerek A. A. Whiteley

Oxford New York Tokyo

OXFORD U N I V E R S I T Y PRESS1992

Contents

List of contributors *'"

1 Islands, regions, ranges, and gradients

Roger L. H. Dennis

1.1 Butterflies on British islands1.2 Butterfly distributions on the British mainland1.3 Focusing on regional and local issues

Adult behaviourTim G. Sh reeve

2.1 The significance of behaviour patterns2.2 Regulating body temperature2.3 Finding nutrients2.4 Mate-locating behaviour2.5 Communication and courtship

. c2.6 Behaviour and butterfly biology

Eggs and egg-layingKeith Porter

3.1 Egg production3.2 Finding and recognizing larval hostplants3.3 Distribution of egg load3.4 Patterns in egg-laying

4 Butterfly populationsMartin S. Warren

4.1 What is a population?4.2 Measuring and monitoring butterf ly populations4.3 The structure of butterfly populations4.4 Natural population fluctuations4.5 Mortality factors affecting population size

Contents

5 Avoidance, concealment, and defence 93Paul M. Brakefield and Tim G. Shreeve with Jeremy A. Thomas

5.1 Adult defences 935.2 The defences of eggs 1035.3 The defences of larvae 1065.4 Adaptations to living near ants 1095.5 Pupal defences 1155.6 Butterflies and multiple defence mechanisms 118

Monitoring butterfly movements 120

Tim G. Shreeve

6.1 The components of movement , 1206.2 Variability in butterfly movement 1226.3 Local movements of butterfly adults 1246.4 Migration and dispersal 1336.5 Unresolved issues 138

Butterflies and communities 139Keith Porter with Caroline A. Steel and Jeremy A. Thomas

7.1 Biological communities 1397.2 Resource partitioning and the niche 1407.3 Interactions between butterflies and other animals 1477.4 Interactions between butterflies and plants 1557.5 Diversity, ecological succession, and butterfly communities 171

Diversity within populations 178Paul M. Brakefield and Tim G. Shreeve

5.1 Genetic variation, natural selection, and evolution 1788.2 Polymorphism and wing pattern forms 1798.3 The development of wing patterns 1868.4 Enzyme polymorphism 1888.5 Seasonal polyphenism 1888.6 Life history variation 191

Case studies in evolution 197

Paul M. Brakefield and Tim G. Shreeve

9.1 The meadow brown: continuous variation and adaptation 1979.2 The large heath: evolution of races 2099.3 The speckled wood: geographic variation in Europe 213

Con f en f s

10 An evolutionary history of British butterflies 217Roger L. H. Dennis

10.1 Evolution before glaciers 21710.2 Evolution with glaciation 22110.3 The pattern of butterfly arrivals 22710.4 Butterfly adaptations to Britain's Post Glacial environments 23110.5 The evolution of subspecies, races, and character gradients 236

11 The conservation of British butterflies 246Martin S. Warren

11.1 Changing butterfly populations 24611.2 Causes of decline of British butterflies 25011.3 Early attempts at conservation 25711.4 The ecological approach to conservation 26211.5 Strategies for conservation 26511.6 Future prospects 274

AppendicesAppendix 1 A check list of British butterflies and their hostplants 275Appendix 2 Traditional classification of butterfly breeding biotopes in Britain 280Appendix 3 (a), (b) Summaries of the Joint Committee for the Conservation

of British Insects codes on collecting and insect introductions 284Appendix 4 Useful addresses of societies, journals, specialist books;

equipment, livestock, and scientific institutions 287

Glossary 289

Bibliography 300

Index 335

Case studies in evolutionPaul M. Brakefield and Tim G. Shreeve

Genetic variation is all important to evolutionarychange. It forms the basis of adaptation to specific ornovel environments and also of divergence betweenpopulations and speciation. This chapter uses workon three European satyrine butterflies to i l lustratethe processes involved in such divergence. Firstly,studies of local differences, and of progressive clinalchanges between populations are described. Thesemay reflect changes in adaptation to the environ-ment. Secondly, examples of more extensive pheno-typic and genetic divergence characteristic ofdisjunct races or subspecies are discussed. Suchdivergence may be the prelude to complete repro-ductive isolation. Once this stage has been reachedand genes cannot be exchanged by hybridizationbetween two groups of populations, they havebecome distinct species. The same processesinvolved in adaptation to novel environments or inlocal divergence between populations are also theultimate basis of speciation events.

Reproductive isolation may involve differences atmany genes resulting in the inviability of hybrids. It

may also be associated with more specific changes inthe genes underlying the mating systems which leadto successful sexual contacts within a population.Broadly speaking such effects are thought of asinvolving postulating and premating isolatingmechanisms, respectively. In practice most specia-tion events are likely to involve some combination ofboth. Unfortunately research on British butterflieshas yet to contribute significantly to the understand-ing of premating isolating mechanisms. Some recentinterpretations of the fossil record have described apattern of comparatively short periods of rapidspeciation and adaptive radiation of new lineagesinterspersed with periods of apparent stasis orrelative stability (Eldridge 1987). This pattern ofevolutionary change is known as 'punctuated equi-librium'. It is not, however, necessary to invoke anynew processes to account for such a pattern. Thework on the three species of satyrines has largelybeen concerned with phenotypic variation in wingpattern.

9.1 The meadow brown: continuous variation and adaptation

The meadow brown Maniola jurtina has long beenrecognized as a particularly variable butterfly com-prising a wide range of different forms, races, andsubspecies (Thomson 1973). It is a common species,forming more or less discrete populations in a widevariety of grassland habitats throughout Britain andmost of continental Europe (Pollard 1981; Brake t ie ld19820,fr; Heath et al. 1984). Adult emergence inBritain ranges from about 100 to 2000 per hectarewith an expectation of l i fe of some five to twelvedays. The butterfly is rather sedentary in areas offavourable habitat with individuals, on average,

ranging over about one hectare during their lifetime.A wide range of nectar sources are used by adultsand larvae feed on a variety of grasses

In early work on the Isles ot Scilly, Oowdeswell c tal. (1949) noted that the number of small black spotsnear the margins of the ventral surface of eachhindwing of Mainola tiirtiita varies trom O to ï rarelysix (Fig. 9.1). They chose to use this variat ion as anindex of the fine adjustment and adaptation otpopulations. Although can- is needed in scoring thepresence of these small spots (Brakefield andDowdeswell 1983), hindwing spot-number provides

198 Paul M. Brakcfield and Tim G. Shreevc

Fig. 9.1 Diagram of variation in the spot pattern on thewing ventral surface of the meadow brown Mamolajurtina. Top row shows variation in the forewingeyespot (unshaded area is of brighter fulvous colora-tion): A, small black eyespot with single white pupil(characteristic of males); B, large spot with single pupil(characteristic of females); C, a more extreme femalephenotype showing a very large eyespot with twopupils (f. bioculata) and with two additional spots (f.addenda). D-L illustrate nine of the thirteen commonlyoccurring hindwing spot phenotypes: D, nought spot,(0); E, costal 1 (Cl); F, anal 1 (Al); G, costal 2 (C2); H,splay 2 (S2); I, costal 3 (C3); J, median 3 (M3); K, splay 4(S4); L, all 5. Not shown: A2, A3, C4 and A4. The refer-ence numbers of the spots are indicated. Different sizedhindwing spots present an idea of changes in relative(not absolute) spot size. (After Brakefield 1990a.)

a simple means of scoring their total area. The size ofeach indivdiual spot is actually a threshold characterin that when it is present it varies in a truly con-tinuous manner; butterflies with no spots representall individuals below the threshold at which spotsare expressed phenotypically. Such quantitativecharacters are usually jointly determined by theinteracting effects of a number of genes or poly-genes, each of small effect. The same phenotype canbe determined by different combinations of thesepolygenes. Differences between populations, andthe results of processes of natural selection influenc-ing quantitative characters, are usually recorded interms of the population mean (or spot average) andvariance. They cannot be represented in terms ofgene frequency changes as is possible for Mendeliancharacters such as polymorphisms controlled by oneor a small number of major genes (see chapter 8).Systems of polygenes which determine continuousvariation in phenotypic characters are critical inevolution since they provide the basis for smoothadaptive change.

9.1.1 Field surveys

A large number of populations of Mamola jurtina inEurope have now been surveyed to compare thefrequency distributions for hindwing spot-numberwhich are referred to as spot frequencies (reviews byFord 1975; Dowdeswell 1981; Brakefield 1984). Someexamples of spot frequencies are shown in Fig. 9.2.Males tend to be more highly spotted than femalesso that the sexes must be analysed separately(Fig. 9.2a). The following discussion is based onsome patterns of variation in the most intensivelysampled regions of the species range. The signi-ficance of these patterns is discussed in relation toprocesses which are involved in generating andmaintaining them.

The earliest survey work by Ford and his co-workers (Ford 1975) involved populations on theIsles of Scilly archipelago situated off the south-westcoast of England (Fig. 9.2). Their study rose toprominence in the 1960s because of the controversyamong evolutionary biologists over the relativecontribution to geographical and population differ-entiation of natural selection and adaptation toparticular environments on the one hand, and ofrandom genetic drift on the other (discussed fur ther

Ca] Tean, 1946

r,o

40

20

m.0 1 2 3 4 5 0 1 2 3 4 $

Spot-number

CbJSt. Martin's

40

.•n

•O 0 1 2 3 4 5C

l» Tresco

<D 40O)<5 20

0 1 2 3 4 5

St. Mary's

40

ZO

The Isles of ScillySt Helen's Whi te Island

v~st' M a r t m s

km.0 1 2 3 4 5

St. Helen's

20

0 1 2 3 4 5

Tean, part

aen

0 1 2 3 4 5

White Island

0 1 2 3 4 5

Fig. 9.2 Variation in the number of spots on the ventral surface of the hindwings of the meadow brown Maniolajurtina in the Isles of Scilly. (a) A summary of the frequency of males and females with different spot-numberscollected by Ford and his co-workers in their original study on Tean. Note the difference between the sexes. Theirmark-release-recapture experiments showed that the population inhabiting areas 1, 3, and 5 of the island werelargely isolated from one another by unfavourable habitat in areas 2 and 4. (Adapted from Ford 1975.) (b) Map of theIsles of Scilly showing the spot-frequency distributions lor l i-malcs Uuind on three large islands and on three smalli s lands (Adapted Irom lord 1975.) (c) Diagram ot the hypothesis put torward by Ford to account tor thedissimilarity among small island populations and the uniformity on large islands. The proposed occurrence ofspecialized environments on small islands and of mixed habitats on large ones is illustrated by different shadings.( A l t e r mater ia l in Ford 1975; Berry 1977; Dowdeswell 1981 and Shorrocks 1978.)

200 Paul M. Rrakcficld and Tim G. Shrcnv

below). Islands are ideal study sites because forsedentary species these processes can be examinedunder conditions of very low individual movementsand gene flow between islands.

Most female populations on the three large islands(>275ha) of the Isles of Scilly showed a 'flat-topped' spot frequency distribution with similarnumbers of 0-, 1-, and 2-spot butterflies (Fig. 9.2b).In contrast, females on the small islands (< 16 ha)exhibited a variety of spot frequencies which tendedto fall into three groups: unimodal at 0-spots,bimodal at 0- and 2-spots and unimodal at 2-spots(Dowdeswell et al. 1960; Creed et al. 1964).

The other major series of studies by Ford's grouphas been concerned with the so-called 'boundaryphenomenon' which occurs along the south-westpeninsula of England. Females in populations inCornwall to the west and extending some distanceinto Devon are more highly spotted than in popula-tions fur ther east and extending throughoutsouthern England (Fig. 9.3). In some years thetransition between the western populations tendingto be bimodal, with peaks at 0- and 2-spots, to thoseunimodal at 0-spots in the east has been an abrupt ora sharp one. For example, in 1956 when firstdiscovered, the boundary along the east-west study

Southern England(& Continental Europe)

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

West T

O O

'c 'tC Sra _"

- S5

i0Cra— i

COraLJ

— r—

0)

"rä

raEn

— r—cow

U

East

Fig. 9.3 The 'boundary phenomenon' for the hindwing spot-number in females of the meadow brown Maniolajurtma across the south-west peninsula of England. The forms of spot-frequency distributions found by Ford and hisco-workers to be characteristic of populations in southern England and in Cornwall art' i l lustrated. The zone of thetransition between these forms occurred, in some years, along the line indicated. The results of surveying spot-number variation in populations in 1956 along a transect bisecting the transition zone are also shown (inset). (Aftermaterial in Ford 1975; Berry 1977; and Parkin 1979.)

Case studie» in evolution 201

transect was associated with two adjoining fieldsseparated by a hedge (Creed cf al. 19S9; see Fig. 9.3).In other years the change-over is much less abruptwith a more or less steep clinal change over sometens of kilometres between regions characterized bythe western and eastern types of spot frequencies.Male spot-numbers, other spot characters (seebelow), and the allele frequencies for several poly-morphic enzymes (Handford 1973; Brakefield andMacnair, unpublished data) are also associated withfluctuations or clines across the boundary region.These observations have again led to some con-troversy involving alternative explanations based oneither divergence during a period of past isolation(allopatric differentiation) or on present-day spatialchanges in natural selection (sympatric evolution).

A study of museum material was also used toexamine variability in spotting throughout the rangeof the species. Dowdeswell and McWhirter (1967)interpreted their data as demonstrating a large, moreor less central region extending into southernEngland characterized by a rather uniform spottingwith other, more peripheral regions, including theIsles of Scilly and Cornwall, showing different formsof spot frequency. Such regions were described asstabilization areas and transitions between themwere considered to be sharp as in the boundaryregion in southwest England. However, when afurther study of this type of Frazer and Wilcox (1975)and the field surveys within particular regions byether workers are considered, it is unclear howprecise the distinctions between stabilization areasare. Certainly enclaves of populations with differingspot frequencies, such as the higher spotted popula-tions of the Grampian Mountains in Scotland, canoccur within these areas (see Brakefield 1984).

9.1.2 The inheritance of wing-spotting variation

The evolutionary significance of such survey resultscan only be assessed with any confidence when thebasis of the phenotypic variation in terms of geneticand environmental effects has been rigorouslyexamined. To take an extreme situation, if differ-ences in spot-number depend solely on the environ-ment during development, as for example iftemperature alone accounts for the amount of spotpigment synthesized, then the field data will merelyreflect the developmental conditions within popula-

tions. They will then be of no evolutionary signi-ficance. In practice, the control of continuousvariation will usually be intermediate between thisextreme and variation arising solely from geneticinfluences. Geneticists assume that variation in aquantitative character results from a combination ofgenetic and environmental differences (see Falconer1981). Their initial aim is to divide the total or pheno-typic variation (Vr) into its components, the geneticvariance resulting from additive effects of the poly-genes influencing the character ( V4) and the environ-mental variance resulting from external effects (Vt).The heritability (/r) of a character is then the propor-tion of the total variance which is additive:

h2- VA/Vr.

The heritability of a character can be estimated fromthe degree of resemblance between relatives. Forexample, if a graph of the mean values for a char-acter of the offspring of different families is plottedagainst the corresponding values for the parents (i.e.mid-parent values), then the line fitted through thepoints represents the degree of resemblance betweenoffspring and parents (see Fig. 9.4). The slope of this,so-called, parent-offspring regression line estimatesthe heritability of the character and can varybetween 0 and 1. A value close to 1 indicates thatoffspring are closely similar to their parents and thatsubstantial genetic variation exists with little environ-mental influence. The measurement of the resemb-lance between relatives gives heritability one of itsmost important properties, namely as a predictor ofthe response to directional selection. For example,say that the character of wing-size has a highheritability in a particular species, then if only thelargest butterflies in a population provide theparents of the next generation, this will have ahigher mean value than the original unselectedpopulation and one more similar to that of theselected parents. Thus a rapid change will beapparent in the population mean for the character.This type of response is also evident when artificialselection is applied on highly heritable variation inplant breeding or animal husbandry programmes. Aheritability of zero, as occurs in the absence ofadditive genetic variance, precludes a response toany form ot selection.

Reliable estimates for heritability in Maniolajtirtina were obtained by breeding 16 families,

202 Paul M. Brakeficld and Tim G. Shrccvc

(a) Males

'S 3

k|c5 2

c~.

/T=0.66± 0.11

Cb) Females

h''=Q 89+ 0.1 1

! L

3 4 "W 2 3

Mean spot-number of parents [mid-parent value]

Fig. 9.4 The genetics of variationin the number of spots on theventral surface of the hindwingsof the meadow brown Maiuolajurtina. The graphs illustrate thefamily resemblance betweenoffspring of each sex and theirparents. The estimates ofheritability and their standarderrors as given by the indicatedregression lines are shown. (AfterBrakefield 1984; courtesy ofAcademic Press.)

comprising a total of 1340 butterflies, from a Dutchstock (Brakefield and van Noordwijk 1985). Thefamilies and their parents were all reared on growinggrasses at similar temperatures and humidities in anunheated laboratory. Parent-offspring regressionanalyses yielded estimates of heritability (Fig. 9.4).These indicate that there is a fairly high heritabilityand substantial additive genetic variance for spot-number. However, some caution is necessary ininterpreting these estimates since they can only beconsidered precise for the stock used and theparticular conditions in which the butterflies werereared. Indeed it is recognized that such estimatestend to overestimate these parameters in naturalpopulations.

Some additional experiments were performedwith one of the families of Mamola jurtina to examinefurther the possibility of direct effects of temperatureon spotting during the period of pattern determina-tion in the early pupal stage (see section 8.3). Thebackground to these experiments includes muchearly work which demonstrated that wing patternscould be altered by subjecting the pupae to extremesof temperature (see Kühn 1926; Köhler and Feldetto1935). Every pattern element was found to have itsparticular sensitive period. Lorkovic (1938, 1943) andH0egh-Guldberg (1971«, 1974) found that prolongedcooling of the pupae of two lycaenids led to effects

on the underside spot pattern. Furthermore, H0egh-Guldberg and Hansen (1977) found that a lowerspot-number was sometimes produced in thenorthern brown argus Aricia artaxerxcs by subjectinginsects, just before or after pupation, to one of moreperiods of cooling at between 2 and 5 °C for 9-12 hours. Their experimental pupae also yieldedsome rare forms which may provide an explanationfor such aberrations in nature (see chapter 8). Thesimilar experiments to those of H0egh-Guldberg andHansen on the fami ly of M. jurtina failed to detect anenvironmental effect on hindwing spotting (Brake-field and van Noordwijk 1985). However, this doesnot preclude the possibility of some more indirect, sofar undetected, effect occurring at some earlier stageof development.

The genetical and breeding experiments withManiola jurtina suggest that differences in spot fre-quency distributions between populations are likelyto reflect genetic differentiation, whether arisingfrom random effects or the influences of naturalselection. The same is likely to apply to changeswhich have been detected within populations over anumber of generations. Some of these changes(Dowdeswell and Ford 1955; Dowdeswell ctal. 1957)have been associated with an altered habitat as forexample on Tean in the Isles of Scilly where removalof cattle led to more luxuriant vegetation cover and

Casestudies in evolution

increased spotting. Other examples have beenassociated with exceptional droughts or warm anddry summers on islands (Dowdeswell et al. 196(1;Bengtson 1978) or with climatic changes rn Englandand Italy (Creed et al. 1959, 1962; Scali 1972). Atlower altitudes in some Mediterranean regionsM. iiirtnin shows a modified life cycle with an earlieradult emergence followed by a mid-summer aestiva-tion of females prior to oviposition in late summer(Scali 1971; Masetti and Scali 1972; Scali and Mast- t t i1975). Scali and his colleagues found that the femalespot frequency changed from a 'flat-topped' form toone unimodal at 0 spots over the period of aestiva-tion, indicating a 65-70 per cent selection against 2-5 spotted specimens.

Detailed measurements of variation in the com-plete spot pattern of Mar.iola jurtina on the ventralsurface of both wings revealed the existence ofnumerous phenotypic correlations. Butterflies ofeach sex which have more hindwing spots also tendto have larger apical forewing eyespots. Theserelationships also extend to populations (Brakefield1984). The forewing eyespot is the most striking andwell developed element of the spot pattern, with awhite pupil and a large black area which contrastmarkedly with the background wing colour It isabout twice as large in females as in males and oftenbipupilled though females usually exhibit fewer andsmaller hindwing spots. The eyespot is alsoextremely variable in size and often in shape withinand between populations of each sex. Hindwingspots in females tend to be more often positioned ormost heavily expressed towards the costal area ofthe wing which is closest to the forewing eyespot.These types of phenotypic correlations are probablycharacteristic of variable species of satyrines andother butterflies (see Frazer and Willcox 1975;Ehrlich and Mason 1966; Mason et ni. 1967, 1968;Dennis et al. 1984) and in M. jurtina hindwing spot-number can be conveniently taken as a 'trend'character for the expression of the complete spotpattern. It will be interesting to determine bv moredetailed genetic analyses or selection experimentswhether or not such phenotypic correlations arebased on genetic correlations involving commoneffects of some genes on different spot characters.Studies on the positioning of the hindwing spotshave, however, shown that this spot-placing varia-tion is largely independent of spot-number

(McWhirter and Creed 1971). This is emphasized inScotland where a steep dine ot increasing costalityin the position of hindwing spots with altitude is notassociated with any corresponding change in spotfrequency (Brakefield 1984).

Analysis of reared material demonstrates a sub-s tan t ia l genetic component in determination ot tor t 'wing eyespot-size (/r — 0.59-0.80; Brakefield andvan Noordwijk 1985). The families also revealedevidence for a component of polygenic control forbipupillation of the forewing eyespot (f. biocitlata),for the shape of the eyespot in males, for thepresence of additional spots on the forewing(f. addenda) and for the presence of white pupils inthe hindwing spots (f. infra-fwpillata) (Fig. 9.1).Similarly, additive genetic variance exists for theposition of the hindwing spots (h2 — 0.35-0.57). Onefeature of both hindwing spot-number and spot-placing is that those estimates of Ir for offspring ontheir same-sex parent are higher than those on thatof the opposite sex (Braketield and van Noordwijk1985). In other words, males apparently resembletheir male parent more closely than their femaleparent, and females are more like their femaleparent. This seems to be due to sexual differences inthe inheritance and expression of the characters.Certain of the individual hindwing spots can becharacterized as typically female while others aretypically male. This will cause a greater resemblanceto the same-sex parent in respect of both spotnumber and spot position

9.1.3 Wing-spotting and natural selection

This pattern ot partial sex-dependence for expres-sion of spotting is of particular interest because it isconsistent with a model to account for variability inspol pat terns in populations (see chapter 5, Fig. 5.3).This model provides an a l ternat ive to the assump-tion that the spots themselves are ' t r i v i a l ' or un-important to the indiv idual It predicts that visualselection by insectivorous birds and l izards operatesdif ferent ly for males and females of Maiiwln jitrtina.These differences in selection are associated withdifferences in the resource requirements ot the sexesand in adult behaviour (Brakefield 1982a). Themarked sexual dimorphism in spot pattern is animportant component of the model. The large eve-spot is frequently hidden when females are at rest,

204 Paul M. Brakcfii'ld and Tim G. Shrenv

but as in the grayling Hipparchia semele (Tinbergen1958; chapter 5) it is sometimes exposed whendisturbed. Overall, the model predicts that an evenwing-spotting will be most advantageous in deflect-ing attacks by predators to the more active butter-flies. Male M. jurtina are generally more active thanfemales and have a more evenly expressed spotting.

Mark-release-recapture experiments performedin populations near Liverpool also suggested that thedistances moved by butterflies of each sex within ahabitat-area increased with spot-number (Brakefield1984). A particularly clear example of this tendencyis illustrated in Fig. 9.5. If such greater dispersaloccurs more generally within populations of thespecies and if it reflects a general increase in activityand frequency of changing behaviour and shiftingposition, which tends to attract searching predators,then the relationship with spot expression is con-sistent with the model. Further capture-recaptureexperiments on other populations and in largerhabitat areas would be most useful. Interestingly,there is some evidence for differences in behaviourbetween two hindwing spotting phenotypes in the

speckled wood Pararge aegeria (Shreeve 1987). Theselection against spotting, particularly on the hind-wing, in inactive butterflies depends on their restingbackground and the effectiveness of their crypsis(see chapter 5). Circular spots may tend to be moreconspicuous in grasslands dominated by linearshapes and thus attract the attention of searchingpredators (a comparable example is shown inFig. 5.2).

Many of the components and details of the modelrequire testing, although Bengtson (1978, 1981)working in Sweden found differences in the fre-quency of wing damage (see Fig. 5.1) betweenspotted and unspotted females consistent with dif-ferences' in their activity and consequently in theirexposure or 'apparency' to predators. However, themodel is of value in the absence of alternativehypotheses about how precisely natural selectioninfluences the phenotypic variation.

An important finding from the breeding pro-gramme was that in some larger families there weredifferences in the timing of adult emergence betweenthe different hindwing spot phenotypes represented

Nought spot

Splay 2

Fig. 9.5 The movements ofmarked females of two hindwingspot-types of the meadow brownManutla jurtina at Hightown nearLiverpool in 1976. Open cirdrsshow capture-release sites andclosed circles show recaptures atthe same site as release. Arrowsindicate the direction of moves.Each map (a) and (b) shows 25movements selected at randomfrom the complete sample for aspot type. Divisions on the edgeof each map are at intervals of15 m. The illustrated patterns ofmovement are significantlydifferent. (After Brakefield 1979.)

Case studies in evolution

by particular combinations of the hindwing spots.Overall such differences led to declines in spottingduring the emergence period. There may be differ-ences in development rates between the phenotypeswhich result from other (pleiotropic) effects of thespotting genes than those on wing pattern. Suchnon-visual effects may be critical in the overallselection influencing spotting. These observationsare consistent with numerous findings of intra-seasonal changes in spot-number within populations(examples in Creed et al. 1959; Dowdeswell 1962;Scali and Masetti 1975; see Fig. 9.6b). When theyoccur such changes are always in the direction of adeclining spot average with time. The species ischaracterized by extreme geographical variation inthe timing and length of emergence. There are con-sistent differences between populations with a trendtowards a more synchronized emergence—begin-ning later and ending earlier—northwards throughEngland and Wales (Brakefield 1987d). There is alsoa clear trend towards an earlier emergence in yearswhen the month of June is comparatively warm.Figure 9.6a shows that differences in the timing ofemergence may also occur in the field between spottypes which differ in the positioning of the spots. InScotland the cline of increasing anality of spottingwith altitude may be related to such effects (Brake-field 1984). It is tempting to suggest that spotting ispartly related to spatial and habitat-related patternsof variation in timing and length of emergence.

Earlier research on the larvae of Maniola jurtina byDowdeswell (1961, 1962) which involved a com-parison of the spotting in adults reared from wild-collected, late-instar larvae with that of flying adultsfrom populations in Hampshire suggested that intra-seasonal changes in spotting could result fromdifferential parasitism by the ichneumonid Apantelestetricus. However, the causal nature of the observedrelationship between level of parasitism and spottinghas not been demonstrated (Brakefield 1984).

9.1.4 Interpreting geographical variation

With this background information in mind theresults of the survey work on the Isles of Scilly andsouth-west England can be discussed with regard toalternative explanations for the patterns of popula-tion differentiation. The spotting of Maniola jiirtiiiuon the three large islands of the Isles of Scilly is

20 r

.

100

III

10/7 15 /7 20/7 25/7

Dale. July-August 1978

Cb)

30/7

20/6 1 /7 1/8

Date, June-Sept 1984

1/9

Fig. 9.6 Emergence patterns and hindwing spottingvariation in the meadow brown Maniola jurtina. (a)Cumulative daily totals of males of two-spot types infresh condition and captured for the first time in twopopulations near Liverpool, (b) The intra-seasonalchange in female spot average for four populations inthe boundary region. Each point is based on about 30females. (After Brakefield 1979, 1984; Brakefield andMacnair, unpublished data.)

similar whereas the small island populations aremore variable tending towards one of three differentspot frequencies (Fig. 9.2b). Ford and his co-workersbelieve that those populations on the large, anddiverse islands result from natural selection produc-ing a gene complex simultaneously adapted to awide range of environments. In contrast, they

206 Paul M. BrakefiL'ld and Tim G. Slirccvc

suggest that the small islands are each characterizedby one of a range of different environments and thatconsequently selection has favoured a more specificgene complex closely adapted to specialized condi-tions (Fig. 9.2c). McWhirter (1957) suggested thatfemales unimodal at 0-spots are associated withareas of open, often exposed grassland whilst thosein luxuriant habitats with patches of scrub tend toshow high spot averages. The discontinuity inspotting found between populations on the two endsof White Island is also a clear example of thisassociation (see Ford 1975). This difference is, there-fore, consistent with the predictions of the model forvisual selection described above and with selectionfor low dispersal rates in more exposed, open, andwindswept habitat and higher ones in a more mixed,diverse habitat. The high spotted and low densitypopulations of the Grampian Mountains in Scotlandwhich occupy a mixed habitat of bracken scrub andpoor grassland also fit this pattern (Brakefield 1982«;1984). However, in many other regions populationsin widely differing habitats show similar spottingDennis (unpublished data) found spotting patternsto be similar in 10 widely separated populations inNorth Wales on very different rock types and associ-ated with different habitat types.

An alternative to an explanation based on differ-ences in natural selection related to habitat is one inwhich random effects are paramount. Waddington(1957) considered that the differences between smallislands resulted from periods of intermittent randomgenetic drift associated with periods of very smallpopulation size known as bottlenecks. A similarreasoning was developed by Dobzhansky andPavlovsky (1957), who suggested that the smallisland populations were derived from small foundergroups with differing gene frequencies from whichrelatively stable but different gene pools developed.Such effects do seem to be reflected in frequencies ofmelanic forms of the spittlebug Philacnus spumariuton the Isles of Sally (Brakefield 1987c, 1990b). How-ever, Ford and his colleagues have countered suchhypotheses for Maniola jurttna with their observa-tions of a population passing through an extremebottleneck in size, with no subsequent disruption ofspot frequency (Creed et ai 1964), and an example ofa change in habitat due to the removal of cattle onTean being associated with one in spotting. Un-fortunately recent fieldwork on the Isles of Scilly

found that a switch in land use on smaller, un-inhabited islands away from grazing by cattle has ledto the spread of bracken scrub through grasslandand the extinction or decline of many colonies ofM. jurtina (Brakefield 1987c).

The studies of the 'boundary phenomenon' insouth-west England (Fig. 9.3) also led to alternativehypotheses to account for the population differentia-tion. Ford and his co-workers consider that theirdata demonstrate the existence of powerful naturalselection which differs on each side of the boundary.Some laboratory experiments with Drosophila fruitflies have shown how such disruptive patterns ofselection practised on artificial populations maysometimes lead to divergence and effective isolation(see Sheppard 1969). Ford (1975) discusses theboundary phenomenon in relation to sympatricevolution in which distinct races or local forms canarise without isolation past or present. Handford(1973) elaborates on this interpretation, suggestingthat there is a switch-over between two co-adaptedgenetic systems at a critical point in an environ-mental gradient. Genes are said to be co-adapted ifhigh fitness depends upon specific interactionsbetween them. Oliver's (I972a,b; 1979) hybridizationstudies provide some evidence of such geneticsystems in several species of butterfly. He foundvarious forms of disruption in development and inthe relative timing of emergence of males andfemales in hybrids between geographically well-separated populations of, for example, wall brownLasiommata incgera, and small pearl-bordered f r i t i l -lary Bolana selene.

Dennis (1977) discusses evidence, particularlyfrom palaeobotanical studies of pollen, which showsthat from about 9500 to 5000 years ago a disjunct orallopatric distribution of Maniola jurtina may haveoccurred in southern Britain. During the warm dryclimate of the Boreal period (see Table 10.2), forestcover would have led to populations being restrictedto the granite or sandstone upland areas and to thecoastal fringe of Cornwall and Devon in the west andto the wide expanse of interconnected calcareousuplands in southern and south-east England. Dennisuses this reasoning to develop the alternativeexplanation for the boundary phenomenon based ondivergence in allopatry originally suggested byClarke (1970). The boundary is then considered torepresent the zone to which two groups of popula-

,

Case studies in evolution

lions of M. Jurtina have expanded their range after aperiod of isolation and divergence. The discretenessof the main population groups may be maintained bysome form of selection against hybrids. Alter-natively, the boundary zone may be in the process ofdecay involving a progressive change to a shallowcline, and eventual uniformity and mixing of genepools. Such a breakdown may be slowed by somepresent-day changes in selection from east to westand the comparatively low dispersal rate of thespecies. In practice it is extremely diff icult to distin-guish between differentiation evolving with or with-out allopatry, especially without a detailedknowledge of the geological and biological history ofthe region across which a present day hybrid /one orsteep cline occurs (Endler 1977).

The most difficult feature of the boundary pheno-menon to account for is the frequent shift in the geo-graphical position of the boundary itself. This maybe up to 60 km east or west between generations(Creed et al. 1970). Clarke (1970) in suggesting thatthe boundary region may be a zone of hybridizationalso postulated that the shifts could result from indi-viduals within this zone being particularly prone todevelopmental instability. Ford (1975) argues thatsuch instability would lead to a mosaic of popula-tions with differing spot frequencies, which is notfound. Recent work by P. Brakefield and M. Macnair(unpublished data) has suggested that progressivedeclines in spotting during adult emergence (seeFig. 9.6b) in combination with variation betweenyears in the timing of the f l ight period and/orsampling dates could contribute to such (apparent)shif ts in the position of the boundary.

Brakefield and Macnair's research was based onanalysis of morphometric data from a grid of overseventy populations covering the whole boundaryregion. This reveals a more complex pattern of geo-graphical differentiation than the results of Ford andhis co-workers obtained by using one or twotransects (Fig. 9.7). The morphometric data for eachspot-character in males and females showed a seriesof more or less coincident clines of varying steepnessfrom east to west along the peninsula between 1982and 1984. The width of the clines was of the order ota few tens of kilometres rather than in single figures.No simple boundary was detected running fromnorth to south across the peninsula. Various types ofmultivariate analysis of the morphometric data for

. :I

aco

Fig. 9.7 Results of a multivariate analysis of spottingvariables in the meadow brown Manwla jurtnia for agrid of populations covering the boundary region in thesouth-west peninsula of Fngland in 1982 and 1484. (Seealso Fig. 9.3.) Pseudo three-dimensional plots for thefirst principal component over the boundary region.The component is a linear combination of variables andmeasures the overall extent of development ot wing-spotting in females. Note: plot ignores sea areas. Majortowns: F, Exeter; T, Torquay; P, Plymouth; O, Oke-hampton; B, Bodmin. ( A l t e r Brakefield and Macnair,unpublished data.)

all spot-characters failed to separate the populationsinto two discrete clusters corresponding to a westernand an eastern grouping. Rather, they indicated asingle cluster with western populations tending toone side and eastern to the other. Overall theseresults are more consistent with a pattern ot fairly

208 Paul M. Brakcficld and Tim G. Shrew

steep clines resulting from some form of present dayspatial change in selection in response to an environ-mental gradient rather than with a hybrid zonefollowing extensive genetic differentiation in allo-patry. Such an environmental gradient might, forexample, involve a change in climate and the effectsof spot genes on timing of emergence. However,such environmental gradients have been in existencethroughout the Holocene (Dennis 1977; Table 10.2)and at this stage one can only speculate about thenature of selective influences across the boundaryregion.

9.1.5 Variation in genitalia

Thomson (1973, 1975, 1976) has made some inter-esting studies of another aspect of the phenotypicvariability of Maniola jurtma. He investigated thedistribution of different morphological forms of thegenitalia (Fig. 9.8). This geographical variability doesnot appear to parallel that in spot pattern. It is mostmarked in the form of the valve of males but is also

evident in the shape of the paired gnathos. Althoughsuch characters can be difficult to measure, two maintypes—the 'eastern' and the 'western'—are distin-guished. These differ particularly in the shape of thedorsal process. The change-over between thesetypes apparently occurs along a more or less broad/.one running southwards from Sweden, TheNetherlands, Belgium and north-east France tosouth-east France, and the Mediterranean. Thisregion is characterized by a transitional type of valveand is much narrower in the south (Fig. 9.8). Unfor-tunately the inheritance of the variation in genitaliahas not been examined. However, Thomson (1987)has found that changes in allele frequencies at 10polymorphic enzyme loci (see chapter 8) are associ-ated with the change-over in types of genitalia. Thisdistribution of genitalia types and enzyme variationmay reflect some form of hybrid zone or secondarycontact between two races or similar entities whichdiverged in allopatry. Such isolation was probablyassociated with changing spatial patterns of 'réfugia'of favourable habitat occurring in the period of gross

(b)

Fig. 9.8 Geographical variation inmale meadow brown Maniola jurtinagenitalia forms over Europe. (AfterThomson 1973, 1975, 1987; courtesyof G. Thomson.) (a) The two prin-cipal forms of the valve of the malegenitalia: left, 'western' form; right,'eastern' form. 1, dorsal process; 2,distal process; 3, valve. The simplecoefficient b-a gives positive valueswhen populations are 'western' andnegative when values are 'eastern'.(b) The distribution of » , 'eastern',• , 'western', and D , transitionalin.i lc genitalia forms.

Case studies in evolution 209

climatic fluctuations of the glacial advances andretreats of the Pleistocene (see section 10.5). Thom-son considers that both the eastern and westerntypes of valve originated from ancestral populationsin western Asia (Iran). These populations are repre-sented today by a so-called 'primitive' valve struc-ture. The two main types of valve may, however,have originated independently in the eastern and

western regions of the Mediterranean (see Dennis1977; Dennis et al. 1991). Within Britain there is adiversity of valve forms, transitional types tending tobe restricted to dry calcareous landscapes (Thomson1973, 1975). This diversity occurs at individual sitesand there may also be an element of seasonality invalve shape (Shreeve 1989).

9.2 The large heath: evolution of races

Apart from the variation in genital morphology, thestudy of Maniola jurtina has examined processesinvolved in microevolution and the adaptation oflocal populations to particular environmental condi-tions. The differences examined between popula-tions have been rather limited in extent, oftenconcerning gradual or clinal changes in the relativefrequency of different spotting types with any pair ofpopulations differing in mean number, size, or posi-tion of spots. Research on another satyrine butterfly,the large heath Cocnouymplm tulha. has begun toexamine more substantial differences in wing pat-tern which are probably associated with the evolu-tion of races in different geographical regions.

Coenonympha tulha is a northern or alpine speciesand together with the small mountain ringlet Ercbiacpifihwii probably represents the most ancient of ourbutterfly fauna (see section 10.3). Colonies of C. tullia are confined to peat mosses, lowland raised bogs,damp acid moorland, and upland blanket bog fromsea-level up to at least 800m. The species has adisjunct distribution in Britain, although this is likelyto be partly due to major losses of lowland raisedbogs. Several major races or subspecies are recog-nized by taxonomists. The use of these terms issomewhat imprecise but they implicate moredistinctive and far-reaching differences betweenpopulations inhabiting different regions than thoseinvolving hindwing spot pattern in Maniola jurtina.The races of Coenonympha tullia are distinguishedprincipally on the basis of development of the sub-marginal rings of eyespots especially on the ventralsurface of both wings (Fig. 9.9). These are moststriking in populations inhabiting some ot the peatmosses of England from lowland Cumbria to Shrop-shire. This phenotype, which is also characteri/ed bya comparatively dark ground colour, is known as

ifflz 'us (an important synonym is philoxenus). Popula-tions of the race polydama with intermediate expres-sion of these spots are found in the central lowlandsof Scotland southwards to Cumbria and NorthWales in the west and Yorkshire and Lincolnshire inthe east. The eyespots are greatly reduced both innumber and size in populations of the race scot tea inthe Scottish Highlands and Orkney. The butterfliesare also comparatively pale in ground colour.

Some indication of more extensive genetic differ-entiation than that associated with the phenomenastudied in Maniola jiirtina is given by some limitedcrossing experiments made by Ford (1949). Hereared some 'intersex' specimens from a matingbetween material from Merioneth in Wales andCaithness in northern Scotland, that is polydama Xscotica (none were observed for a cross of Merionethand Carlisle in northern England). This form ofdisruption in the more distant cross indicates sub-stantial disruption of the mechanism of sex deter-mination. Oliver's (1972fl,/ '; 1979) experiments,which included crosses of Lasiommata megera fromEngland and France, quantified in a more completemanner the disruption of features such as the sexratio and the emergence pattern ot males andfemales commonly found in hybrids betweendistantly spaced populations. Such effects indicatesubstantial disruption of development and hybridinviability, probably as a result of the breakdown ofcoadapted gene complexes.

A detailed morphometric analysis of material fromBritish collections of Cocnoinfiiifha titllia has beendone by Dennis and Porter over the last decade or so.Their latest results (Dennis et al. 1986) describe amultivariate analysis of records of spot presence andmeasurements of spot size and wing area in samplestrom thirty-one localities in the British Isles

210 Paul M. Brakcfield and Tim G. Shreeve

davus

July sunshine (h day )

Fig. 9.9 Variation in wing-spottingamong British populations of the largoheath Cocnoni/inplia tullia. (Redrawnfrom sources in Dennis et al. 1986;courtesy of Entomologist's Gazette.) (a)Examples of the ventral spotting toillustrate the degree of variation fromnorth to south in Britain, (b) Sites fromwhich population samples wereobtained for morphometric analysis, (c)A non-metric scaling plot of the malepopulations for some f i f t y spotvariables; most of the geographicalvariation can be accounted for in asingle dimension. Lines join nearestneighbour populations from singlelinkage cluster analysis of the samedata. Clusters specify traditionalsubspecies, (d) Relationship between themean spot-number (index of spotdevelopment) for the samples of malesand the mean dai ly hours of sunshine inJuly at the sites. The fitted regressionline is shown, (e) A male of the racedavus collected at Whixhall Moss,Shropshire, shows symmetrical damageto the hindwings likely to have resultedfrom the partial evasion of an attack bya bird or lizard. Forty per cent of themales sampled in 15 of the populationsgiven in (b) had symmetrical wingdamage to the hindwings.

(Fig. 9.9b). A non-metric two dimensional scalingplot (see p. 6) produces three dusters correspond-ing to populations of the three phenotypes associ-ated with the described subspecies (Fig. 9.9c). Theclusters to the left and right of Fig. 9.9c includesamples from the north of Scotland and from Shrop-

shire, Cheshire, and Cumbria, respectively. Thelarger, central envelope comprises the geographic-ally more heterogeneous group covering the regionfrom the central lowlands of Scotland to northernEngland with North Wales. The two Irish collectionsappear to be somewhat intermediate between scotica

Case studies ni evolution 211

and polydama. It is interesting that the positionwithin the polydama envelope tends to reflect thegeographical order either southwards from Scotlandor westward into North Wales. This relationship canbe seen more clearly within the array for the ScottishHighlands and lowland English mosses.

These findings suggest several possible evolution-ary explanations. For example, the local disjunctdistribution of Cocnonympha tullia, throughout muchof its range, may tend to mask what is essentially apattern of equivalent clines in spotting extendingboth northwards and westwards from the lowlandEnglish mosses in response to an environmentalgradient. At the other extreme, the three phenotypesmay correspond to three (or four) groups of popula-tions which have diverged genetically from eachother during a past period of allopatry and isolation.If so, why is there evidence of clinal change withineach race? This could represent a common responseto similar environmental gradients within eachregion. The occurrence of intersexes when distantpopulations are crossed is consistent with this typeof explanation. One can then ask whether thedivergence in spotting patterns reflects:

(1) a by-product of more general genetic differentia-tion unrelated to adaptation processes;

(2) pleiotropic effects of selected differences in genescontrolling spotting unrelated to the spottingphenotypes per se; or

(3) an integral component of adaptation to differ-ences in the environment inhabited by each race.

A further possible complication is that the hetero-geneous grouping of populations with intermediatephenotypes could represent a zone of mixing orintrogression between the races with extremephenotypes produced by a spread and mixing ofpopulations resulting from recent human activities.

Unfortunately, as Endler (1977) has emphasized, itis very difficult to distinguish between hypothesesinvolving past allopatry and differentiation withsecondary contact and those concerning responses topresent-day selection regimes and environmentalgradients (see preceding section). This applies toCocnonympha tullia since it is difficult to speculateabout the likely historical changes in distribution ofthe population groupings in relation to both eco-logical and climatic constraints. If future workdemonstrates more profound ecological differences,

and confirms the major genetic differentiationbetween populations of the lowland mosses andthose of the Scottish Highlands, then the hypothesisinvolving secondary introgression becomes moreattractive. The often patchy distribution of coloniesof C. tullia even in regions where it is relatively welldistributed makes it rather difficult to distinguishreal discontinuities in distribution and gene flowbetween populations. The northern and southernraces of the brown argus, which are now usuallygiven the taxonomie status of species (Anaa agestisand A. artaxerxcs), provide a particularly interestingexample of well separated population groupingsassociated with both genetically controlled differ-ences in wing pattern and other morphologicalcharacters and with ecological differences especiallyin larval hostplants (section 103).

The model described in the previous section relat-ing variation in spotting between meadow brownbutterflies to differences in adult activity patternsand exposure to vertebrate predators can be appliedto Coemwympna tullia (Brakefield 1979, 1984; Denniset al. I984. 1986). The basic components suggest thatthe combination of a habitat with mixed vegetationand high sunshine experienced by the lowlandShropshire populations will favour the evolution ofeyespots for deflection of predator attacks.Figure 9.9e illustrates a specimen from WhixhallMoss in Shropshire with symmetrical damage to thehindwing consistent with deflection by the eyespotsof an attack by a bird or lizard away from the body otthe insect (see also Fig. 5.1). At the other extreme itcan be argued that in northern Scotland a comparat-ively homogeneous moorland habitat and a coolerclimate which reduces adult activity will favoursmaller eyespots which enhance crypsis. Analogousto the scenario for Maniola j u r t i i i a . large eyespots inthese conditions might attract predators to restinginsects which were for the most part incapable ofescape. It is also noteworthy that when other speciesof heaths on the continent are considered, forexample, Coenoin/mpha arcaiua, C. dons C. gardctta,and 0. niuehcn, populations at lower altitudes showa more pronounced spotting than those of moremountainous regions.

Dennis et al. (1986) found a close positive associa-tion between spotting in Coenonympha tullia and theduration of bright sunshine in the adult f l ight period(Fig 9 9d) Although there are statistical associations

212 Paul M. Brakcficld and Tim G. Slirccvc

with other climatic variables, that with sunshine isthe highest and most consistent with geographicalchanges in levels of adult activity. An increase inactivity with more frequent changes in behaviourand position could in turn shift the balance of selec-tion on wing pattern towards an emphasis on morehighly developed spotting functioning primarily inthe deflection of attacks by predators. Interestingly,in the scotica phenotype the pupils of small spots aresilver in colour which may tend to increase their con-spicuousness in sunshine by reflective effects. Signi-ficant positive correlations between spotting andsunshine have also been found for Mamola jurtina inBritain (Brakefield 1979) suggesting that this may bea more general relationship. Dennis et al. (1986)summarize the incidence of wing damage in theirmaterial of C. tullia which was consistent with failedattacks by predators (see chapter 5). There was evid-ence of a high level of prédation by birds, with 43 percent of males and 58 per cent of females bearing atleast one damage mark. They had noted earlier(Dennis et al. 1984) that the pattern in the position-ing of such damage was consistent with the hypo-thesis that larger wing spots are the best decoys.

Dennis et al. point out that certain populations ofCoenonympha tullia in Britain, including some withwell expressed spots, are associated with grasslandhabitats at the margins of woodlands or where sub-stantial invasion by birch scrub has occurred Theseare habitats where the abundance of foraging birdsand heterogeneity of resting backgrounds are greaterthan in uniform grassland. Interestingly an addi-tional spot frequently occurs at the anal border of thehindwing in populations near Witherslack, Cumbriawhich is the most wooded locality. This is consistentwith the influence of the strongest directional selec-tion for spotting. Shapiro and Cardé (1970) describedan increase of spotting in some woodland satyrinesof the genus Lcthc in an area of New York StateUSA. Brakefield (1979, 1984) suggests that there maybe a general tendency for woodland satyrines to bemore heavily spotted, possibly because their restingbackgrounds tend to be more heterogeneous and lessdominated by the linear shapes of grasslands. How-ever, this does not seem to apply to the dry seasonforms of tropical species which are inactive and relyon a resemblance to dead leaves for survival (seeFig. 5.2). Dennis et al. (1986) suggest that more

numerous and larger eyespots are favoured in condi-tions of changing light and shade because butterfliesmay be less able to monitor, and therefore avoid,approaches by predators and that the low light levelsnecessitate more contrasting spots to divert predatorattention

Although some interesting explanations of geo-graphical variability in wing pattern have been givento some butterflies such as large heaths, it must beemphasized that tests for the entire chain of causalmechanisms are still required. Evidence that birdscan select for small differences in wing pattern ordetails such as eyespots is discussed in chapter 5. Incommon with most research on variation in wingpattern's where visual selection by predators isimplicated, we need to know much more about theprecise incidence and mechanics of feeding bypredators on living Lepidoptera in the wild. InCoenonyrnpha tullia we also need to know much moreabout its ecology before we can be more confidentabout our adaptive explanation for variation in wingpattern There are no quantitative data describingthe postulated differences in activity levels betweenpopulations of different regions, although they couldbe readily obtained.

The other major area for which empirical datamust be obtained before we can be more definiteabout the role of natural selection in the observedspatial differentiation of wing pattern is the mode ofcontrol of the phenotypic variation. There is noinformation from breeding work about the inherit-ance of spotting variation in Coenonympha tullia.Indeed Turner (1963) suggested that the distributionof the races of C. tiillni is to some extent correlatedwith temperature: for example, on Islay in theWestern Isles, Scotland, which has a mean annualtemperature comparable to that of Shropshire, asmall proportion of individuals approach the davit»phenotype, with many more of intermediatespotting development. Dennis et al. (1986) confirmedthe close statistical correlations between spottingand temperature variables for July. Although sucheffects are quite consistent with the model of selec-tion developed above, there is a need to establishthat the development of spotting is insensitive tosuch direct environmental effects (see previous sec-tion and chapter 8).

Case studies in evolution 213

9.3 The speckled wood: geographic variation in Europe

The studies of Maniola jurtina illustrate clinal varia-tion and microevolution, and those of Coenonymphatullia, evolution and differentiation in isolation.More substantial variation, in wing pattern andseasonal polyphenism (see section 8.5) is exhibitedby the speckled wood Pararge ae$ena in Europe. Thisis discussed below in relation to the evolution andmaintenance of distinct forms within a continuousdistribution.

Within its whole range (Fig. 9.10) Pararee aey^cna,occupies a variety of habitats, including forest,grassland, and wasteland. Some geographic regions

are characterized by a gradual change of the wingphonotype, but there are also more abrupt changesdespite an apparently continuous distribution. Thereare also differences in the l i fe histories of popula-tions from different areas (Thompson 1952; Robert-son 1980; Wiklund et al. 1983; Shreeve 1985; Nylin etal. 1989; E. Lees, personal communication) whichmay not coincide with changes in the adult pheno-type. This whole pattern of variation is not easilyexplained by either clinal variation in response topresent-day environments or by adaptation to pastisolation but rather by some combination of the two.

I III. n,, In, Illl In, Ilii IlllFig. 9.10 Geographic variation in the phenotype of the speckled wood Parargc acgcria in Furope. The three mainforms, P. aegcria aegeria in Iberia and North Africa. P. acgeria firds over most of Europe, and P. acgeria italienrestricted to the Italian peninsula are clinally variable Most Mediterranean island forms resemble adjacentmainland forms, but those from Sicily, Crete, and Cyprus are probably unique. Size, wing shape, pattern, and colourof the various forms are represented on an arbitrary scale from 1 to 10 by histograms. A. Size: 1, smallest (Africa), 10,largest (Scotland); B. wing shape (forewing angularity): 1, most indented (Africa), 10, least indented (northernEurope); C. pattern (dorsal pale patches): 1. most extensive (Africa) , 10, least extensive (Scotland); D dorsal patchcolour: 1, most orange (Africa) . 10. pole cream colour (Scotland). Subspecies tireis is the most seasonably variableDistribution in eastern Furope is not well documented but may extend to the Ural mountains and Pripyat marshes

214 Paul M. Brakefield and Tim G. Shreeve

9.3.1 Species, subspecies, and geographic variation

The three species within the genus Pararee (seeHiggins and Riley 1983) are distinguished by theirwing shape and colour. P. aegeria, which occursthroughout mainland Europe, western Asia, NorthAfrica, and Madeira is the smallest and most variablespecies. P. xiphia, is the largest and darkest speciesand is restricted to Madeira. P. xiphioides, whichoccurs on the Canary Islands is intermediate in sizeand colour between P. xiphia and P. aegeria.Numerous races of P. aegeria have been described byvarious authors as varieties or subspecies (e.g. Verity1916; Gaede 1931), but two subspecies are uni-versally recognized, P. aegeria aegeria from southernEurope and P. aegeria tircis from northern Europe.They are distinguished by the colour of the paleareas on the upper wing surfaces; in the former theseare reddish orange and in the latter creamy yellow.The southern European form also has more scal-loped margins on the forewing.

Within the range of the southern aegeria formthere is obvious clinal variation, specifically anincrease in wing expanse and a decrease in theintensity and size of the reddish orange patches onthe upper wing surface northwards and withaltitude. The wing shape appears constant and thereis l i t t l e indication of seasonal polyphenism. Insouthern Spain and in Africa the species is con-tinuously brooded but is least abundant in the drysummer months. Within this area individuals fromthe higher parts of the Pyrenees are strikinglydifferent, being much larger, with more hair-scalesand darker wing coloration. Presumably it is aresponse to thermoregulatory needs in the coolerenvironment (see section 2.2).

The northern tircis form is more uniform in wingshape and pattern throughout its range. It is usuallyseasonally polyphenic, spring adults having largerand more extensive pale wing areas than those flyingin summer. This variation is more extreme in thenorth. However, populations in eastern Finland,central Sweden, and probably north-western Russiahave only one generation a year (Wiklund et al.1983), and may therefore be less variable (see alsosection 8.5). Although coloration is one of thefeatures which distinguishes tircis from the southernEuropean aegeria, this coloration is variable. At lowaltitude in southern parts of its range it is similar in

colour, but not pattern, to the southern form. Atnorthern latitudes and at high altitude in southernareas the phenotype of tircis is very similar to thatfound in southern Britain.

Individuals from peninsular Italy are di f ferentfrom those of all other mainland European areas,both in wing shape and pattern, though this pheno-type is also located on certain Aegean islands, inSyria, and Lebanon. This similarity of Italian andMiddle-Eastern races was first commented on byVerity (1916) but Larsen (1974) disputes this view.He points out that current geographic distributionpatterns cannot support any inferred similaritybetween Italian and Middle-Eastern races. Thephenotype does not extend into the Alps but doesoccur along the northern Dalmatian Coast.

On each of the Mediterranean islands the pheno-type is unique but on most it has a close aff ini ty tothe southern aegeria form. Interestingly, thoselocated on Corsica and Sardinia, which are only25 km apart, are very different, resembling (;ms andaegeria, respectively. Neither resemble those fromIta ly , the nearest mainland area.

9.3.2 Adaptations and evolutionary arguments

The complex pattern of geographic variation doesnot always follow geographic boundaries and is noteasily explained, particularly the abrupt boundariesin southern France, and in northern I ta ly (see Dennisetal. 1991).

However, clinal variation in the wing colorationwithin each of the northern and southern Europeansubspecies follows a consistent pattern which,together with seasonal polyphenism, can beexplained by adaptations to present-day environ-mental gradients In each, the pale wing areasbecome more orange southwards and with loweraltitude, a change which can be related to back-ground matching. In a woodland habitat in Britainmore than one half of individuals are damaged in amanner which is consistent with having beenattacked by birds (Shreeve 1985). An examination ofpopulations in 25 different sites in France reveals asimilar incidence (Shreeve, unpublished data).Therefore, an appropriate form of crypsis may sub-stantially enhance survival (see chapter 5). Whensettled in a potentially active state, the butterflybasks with open wings and when the general wing

Case studies in evolution 215

pattern mak hes that of the background the butterflyis difficult to detect. Although the effectiveness ofthe various colour forms on different backgroundsneeds to be tested, orange coloration probablyaffords better crypsis in drier southerly, low-lying,habitats. A reduction in the contrast between darkand light wing areas of negerin may decrease con-spicuousness in more open habitats where the con-trast between light and shade is less extreme than indensely wooded areas. This may also account for thedevelopment of orange coloration on the uppersurfaces of individuals from the Isles of Scilly,described by Howarth (1971) as subspecies insula.Observations of the species in [-"ranee (Shreeve,unpublished data) demonstrate that in southernareas individuals spend more time resting on bareearth and dead leaf litter than on vegetation, but inmore northern areas the main resting background isvegetation, usually in woodland. Reduction of thewing area occupied by the pale patches with increas-ing latitude and altitude can be directly related tothermoregulatory constraints (see chapter 2), themost melanized individuals occurring in the coolestand cloudiest regions, such as the higher Pyrenees.

The geographic pattern of seasonal polvphenismcan be related to patterns of habitat use and seasonalchanges in these habitats. In southern regions, suchas the southern part of the Rhône valley, the mostfrequently used habitats are open evergreen forestand ditches, and water-courses in very open agri-cultural areas. Unlike most northern habitats, whichare chiefly dense deciduous and coniferous wood-land, these are probably less seasonally variable inboth colour and light intensity. Therefore, seasonalpolyphenism, which may enhance backgroundmatching is likely to be generally favoured in thenorth but not in the south.

Individuals from three Mediterranean regions-Spain, Italy, and Greece—differ, though thev are insimilar climatic zones. Three explanations may beoffered to account for these differences. Firstly, theymay all originate from a single source in the southernrange of the species, variation between these geo-graphic areas being related to local adaptation todifferences in climate, predators, and habitat.Secondly, the present distribution of forms may bethe result of colonization by distinct forms from twoor more areas since the List major ice retreat some15 000 years ago, each a d j u s t i n g to current condi-

tions. Thirdly, variation may be the consequence ofphenotypic plasticity with different environmentaland/or climatic conditions producing differentphenotypes (see section 8.5).

At present it is impossible to provide a definitiveexplanation for the pattern of variation throughoutEurope or to provide an accurate post-Pleistoceneframework of range expansion (see Dennis et al.1991; section 10.5). However, it is possible that thespecies has spread from more than one geographic-area. Examination of museum material and fieldwork in July 1987 and 1988 reveals an apparentabsence of phenotypes intermediate between Parareeacgena aegena and P. aegena tircis in their main con-tact zone in southern France. In the Rhône valley theformer subspecies is located in agricultural areas atlow elevation and the latter at higher elevation inboth agricultural and woodland habitats, the twoforms occurring some 1-2 km apart. Similar pat ternsof distribution are recorded for other species andsubspecies in southern Europe (see Dennis ct at.1991). This lack of intermediates may be explainedby abrupt habitat differences but a comparisonbetween chromosome numbers of P. ae^ena ae^ena(N — 27) and P. aegeria tiras (N — 28) (Robinson1971) points to genetic dissimilar i ty between the twosubspecies, perhaps indicative of different geo-graphic origins The lack of intermediates may thenindicate hybrid inviability or the existence of pro-mating barriers. However, the degree of geneticdif ferent ia t ion of these Uvo forms is not quantified. Itis possible that the phenomenon may be a moreextreme example of the genital valve boundaryphenomenon described tor Maniola i i t r t i na . Thedistance between aegcria and tims is not great andmay be traversed by flying individuals. Furthermore,these observations were made at the driest time ofyear when population density was low and dampareas for egg-laying scarce. At other times, parti-cularly in spring, population density may be higherand with more abundant and widespread suitablelarval resources, the distance between aegena andtin !•• mav be less

I he distr ibut ion ot the \ , i r ions lorms ot Put a i ynegerin is currently changing. In southern Framethere is some evidence tha t the southern Furopeanlorni is expanding its range northwards (Robertson,personal communication) perhaps m response tochanges ot landuse In Sweden there are two

216 Paul M. Brakefield and Tim G. Shrcci'c

geographically separate races of different origin, aunivol t ine race from Finland and a bivoltine racefrom Denmark, the latter being recent colonists(Nordstrom 1955).

Parargc aegena aegeria is a recent colonizer ofMadeira, being first found on the island in 1967 byHoegh-Goldberg (Oehmig 1980). The impact of thiscolonizer on the endemic P. xiphia is of considerableinterest. Lace and Jones (1984) suggest that it mayreplace the endemic species, citing the disappear-ance of P. xiphia from low-lying areas in the south ofthe island as tentative evidence of the 'taxon cycle'(Wilson 1961; Ricklefs and Cox 1972). In this cycle arecent colonist is supposed to be competitivelysuperior to a resident species because the invadingorganism has escaped from its natural enemieswhereas the resident has a suite of enemies. Thiseffect is presumably thought to outweigh the closeradaptation of the resident to the environment.

More recent work (Owen et ai 1986; Jones et al.1987; Shreeve and Smith, in press; Lace and Jones, inpress) provides evidence of limited habitat partition-ing and overlap and of potential competitive inter-actions between the two species. Because Parargcxiphia is more active in cool, and P. aegena in warmtemperatures (Shreeve and Smith, in press), thedistribution of the two species changes with theseason. P. xiphia tends to be most abundant in cool,high altitude laurel forest in summer but more dis-persed into more open, lower pine forest and agri-cultural areas in winter. P. aegena tends to be morecommon in low altitude warm habitats. Althoughthere may be interactions between individuals of thetwo species in relation to mate-location (Lace andJones, in press), competitive displacement of adultsis unlikely as the larger endemic species tends tooccupy canopy layers and the colonizer groundlayers while engaging in patrolling and perching (seechapter 2). There is a possibility that interactionsduring development, perhaps via shared naturalenemies have an effect on the dynamics of any inter-action between the two species: larvae of bothspecies can occur on the same foodplant (Brachy-podmm sylvaticum) when the two occur togetherWhilst the exact nature of the interaction betweenthese two species remains in doubt, changes in agri-

cultural practices may well have caused the loss ofP. xiphia from low-lying areas before colonization ofthe island by P. aegena aegeria.

The time of origin and establishment of Parargexiphia on Madeira is not known. It is probable thatthis species, and P. xiphioides, are derived fromP. aegena or P. aegeria-tike ancestors possibly fromNorth Africa or southern Europe, the isolatedpopulations then evolving into distinct, reproduct-ively isolated species. The marked dissimilarity ofthese species from P. aegena imply a relatively longperiod of isolation. However, the strength of pastand present selection on the adult phenotype is notknown; intense selection can cause rapid change andphenetic differentiation in isolation can be rapid(Dennis 1977). If change has been rapid it is possiblethat divergence of wing pattern of the differentisland species has occurred over a few thousandgenerations, though differences of other features ofboth island species, such as genitalia (Higgins 1975),larval morphology (Shreeve, personal observation),behaviour, and thermorégulation (Shreeve andSmith, in press; A. Smith, T. Shreeve, and M. Z.Baez, unpublished data) suggest a longer period ofisolation (see section 10.5). Some of the Mediter-ranean island races are probably of more recentorigin than the Atlantic island species. They may bethe result of either recent colonization events, asperhaps in the case of the Balearics, or originate fromearlier Pleistocene invasions.

Pararge is not unique in its complex pattern ofvariation and probable diverse origin of geographicraces. Other equally complex patterns occur, forexample in species of Melanargia (Higgins 1969;Descimon and Renon 1975; Wagener 1982; Mazel1986; Tilley 1986), Erebia (Lorkovic and de Lesse1954), Lysandra (de Lesse 1960,1969,1970), and Hip-parchia (de Lesse 1951). In Erebia there are identifi-able hybrid zones.

In this chapter studies of three species have beenexamined in detail to illustrate the processes ofdivergence, differentiation, and the formation ofreproductively isolated forms of species. Theseissues are put into the wider historical perspective inchapter 10.