polymorphic mimicry in papilio dardanus: mosaic dominance, big...

Polymorphic mimicry in Papilio dardanus: mosaic dominance,

big effects, and origins

H. Frederik Nijhout

Department of Biology, Duke University, Durham, NC 27708, USA

Author for correspondence (e-mail: [email protected])

SUMMARY The mocker swallowtail, Papilio dardanus, hasa female-limited polymorphic mimicry. This polymorphism iscontrolled by allelic variation at a single locus with at least 11alleles. Many of the alternative morphs are accurate mimics ofdifferent species of distasteful butterflies. Geneticists havelong been interested in the mechanism by which a single genecan have such diverse and profound effects on the phenotypeand in the process by which these complex phenotypic effectscould have evolved. Here we present the results of a morpho-metric analysis of the pleiotropic effects of the mimicry geneon the array of elements that makes up the overall pattern.We show that the patterns controlled by mimicking allelesare more variable and less internally correlated than those

controlled by nonmimicking alleles, suggesting the two aresubject to different degrees of selection and mutationalvariance. Analysis of the pleiotropic dominance of the allelesreveals a consistent pattern of dominance within a coevolvedgenetic background and a mosaic pattern of dominant andrecessive effects (including overdominance) in a heterologousgenetic background. The alleles of the mimicry gene have bigeffects on some pattern elements and small effects on others.When the array of big phenotypic effects of the mimicry geneis applied to the presumptive ancestral color pattern, itproduces a reasonable resemblance to distasteful modelsand suggests the initial steps that may have produced themimicry as well as the polymorphism.

INTRODUCTION

The female-limited polymorphic mimicry of the African

Mocker Swallowtail, Papilio dardanus, represents one of the

most spectacular and puzzling cases of evolution in the animal

world. Papilio dardanus is broadly distributed throughout

sub-Saharan Africa. The males of P. dardanus look identical

across this entire range, but the females occur as at least

14 distinct morphs (Fig. 1). The majority of these morphs are

Batesian mimics of several distasteful species of Danaidae and

Acraeidae (Poulton 1924; Ford 1936; Bernardi et al. 1985).

The remaining morphs, although highly distinctive, do not

resemble any known species of butterfly and are not believed

to be mimics. As many as six of the female morphs may occur

together in a given population. Perhaps the most remarkable

feature of the female color morphs of P. dardanus is that their

diversity is controlled by allelic variation at a single locus, the

mimicry gene,H. There are at least 11 alleles at this locus, and

the various diploid combinations of these alleles account for

the diversity of highly distinctive female color patterns.

The mimicry gene

There are at present two competing hypotheses about the

nature of the mimicry gene: the supergene and the regulatory

gene hypotheses. Clarke and Sheppard (1960d) suggested that

the diverse phenotypic effects of this locus could be explained

if it was actually a tightly linked cluster of genes, each with an

independent effect on the phenotype. Such a supergene could

arise by recombination events that move genes with related

functions for pattern specification to a common region of a

chromosome; recombination within such a gene cluster could

be inhibited by an inversion. This way, each allele of the

supergene would lock together a unique combination of

alleles of the constitutive genes to achieve its particular

phenotypic effect. The alternative hypothesis is that the

mimicry gene is a regulatory gene that controls the expression

of a number of unlinked genes that affect various aspects

of the color pattern (Nijhout 1991). Each allele of this

regulatory gene would control a different temporal or spatial

pattern of expression in the genes it regulates. The regulated

pattern genes could be monomorphic in a given geographic

race, and differentiation of these genes between races could

account for the observed breakdown of mimicry in interracial

hybrids.

The exact number of alleles of this gene is not known,

because all possible crosses have not yet been done, but it

appears at present that there are at least 11 alleles with major

effects on the pattern. Most alleles have the same phenotypic

effect wherever they are found across the geographic range of

EVOLUTION & DEVELOPMENT 5:6, 579–592 (2003)

& BLACKWELL PUBLISHING, INC. 579

the species, but three of these alleles (H c, Hh, and HT)

produce significantly different phenotypes in different popula-

tions (Clarke and Sheppard 1960a,b,c, 1962). The form

ochracea appears to be controlled by the same allele (H c) as

cenea. The form cenea is widespread and mimics the danaids

Amauris echeria and A. albimaculata, whereas ochracea

occurs in a small area in Northern Kenya where it mimics

the darker local form, Amauris echeria septentrionis. The form

hippocoon is probably controlled by the same allele (h) as

hippocoonides. Hippocoon is more widespread than hippo-

coonides; their distributions do not overlap because the two

forms mimic different geographic races of Amauris niavius. In

addition, in the race polytrophus, there is a yellowish form of

hippocoonides, called trimeni, which appears to be produced

by modifier genes. Finally, the form lamborni from central

Kenya may be controlled by the same allele (HT) as the more

widespread form trophonius. The differences in the color

patterns associated with these three alleles are presumably due

to differences in the genetic background in which the alleles

are expressed.

The evolution of mimicry

Batesian mimicry is believed to originate by means of an

initial mutation that has a sufficiently big effect on the

phenotype to give a passable resemblance to a protected

model, followed by the accumulation and selection of

mutations in modifier genes that progressively refine the

mimicry (Fisher 1930; Carpenter and Ford 1933; Sheppard

1959; Clarke and Sheppard 1960c; Charlesworth and Charles-

Fig. 1. The homozygous female forms of Papilio dardanus. The specimens are arranged in three groups based on their pattern. The blackportions are considered pattern, and the colored portions are background. The names of the forms and their allelic designation are givenbelow each specimen. The text refers to group 1 as the hippocoon, group 2 as the cenea group, and group 3 as the male-like or theplanemoides group. A male is shown underneath group 2. Males are monomorphic and identical for all genotypes.

580 EVOLUTION & DEVELOPMENT Vol. 5, No. 6, November^December 2003

worth 1975a; Turner 1977; Charlesworth 1994). Charlesworth

and Charlesworth (1975b,c) calculated the conditions under

which mimicry will evolve, and their calculations suggest that

modifying mutations that refine the mimicry will be main-

tained if they are tightly liked to the gene that conferred the

initial advantage, thus providing a plausible explanation for

the evolution of a supergene.

In P. dardanus, this simple scenario is complicated by the

fact that mimicry is polymorphic and that the various morphs

are not equally distributed among geographic races. Among

the unresolved questions are (a) whether each of the 14

morphs evolved independently from a common ancestor or

(b) whether some morphs evolved from others and (c)

whether the geographic distribution of morphs due to

independent origins in different regions or (d) to differential

dispersal and extinction. Given the complexity and the wide

geographic distribution of the polymorphism, it seems

parsimonious to assume that each of the alleles arose only

once and is identical across its entire geographic range and

that the evolution of geographic races occurred after the

evolution of the polymorphism. The alternative possibility,

that the same polymorphic mimicry arose several times

independently in different geographic races of a monomorphic

ancestor, seems to be less probable. An intermediate between

these two extremes is that a limited polymorphism of perhaps

three or four alleles arose first, followed by a geographic

dispersal and a subsequent evolution of additional alleles

unique to particular geographic races. The latter is almost

certainly true for the form ochracea (derived from cenea),

hippocoonides (derived from hippocoon), lamborni (derived

from trophonius), and of the nonmimetic male-like female

morph from Ethiopia (a loss of mimicry). Whether the male-

like females from Madagascar represent a primitive condition

or are also due to a secondary loss of mimicry remains an

open question (Vane-Wright and Smith 1991).

The diversity of P. dardanus female forewing patterns can

be classified into three groups (Fig. 1). The first one, which we

will call the hippocoon group, is the largest and has a pattern

of large bold areas of black and color. The black portions of

the wing pattern constitute ‘‘pattern,’’ whereas the white or

colored portions of the wing pattern constitute ‘‘background’’

(Nijhout 1994). Pattern diversity in this group consists

primarily of diversity in the background color and to a lesser

degree in differences in the width of the black areas of the

pattern. The second is the cenea group, which consist of

two forms, cenea and ochracea. Here the black portions of the

pattern are greatly expanded and the background is reduced

to an array of spots. The third group is the planemoides

group. It has relatively simple patterns of dark areas in the

proximal and distal portions of the wing, separated by a

single band of background color. In this group we also

include the male patterns and the male-like female patterns,

which have the distal black pattern but not the proximal one.

Although we know little about the nature and molecular

mechanism of action of the mimicry gene, we can nevertheless

learn a great deal about how this gene controls various

aspects of the color pattern by systematic analysis of how

different aspects of the color pattern vary under different

breeding conditions. The beginnings of such an analysis, using

morphometrics and a comparative analysis of pattern variat-

ion in color forms and interracial hybrids, is the subject of this

article. This analysis suggests a hypothesis about the first steps

in the evolution of mimicry and shows that the subsequent

diversification of forms and refinement of their mimicry

involved significant changes in the patterns of dominance of

the many pleiotropic effects of the mimicry gene.

MATERIALS AND METHODS

Specimens for this study were obtained from collections and

experimental crosses preserved in the British Museum, Natural

History, the National Museum of Kenya, and the African Butterfly

Research Institute (Nairobi, Kenya). Because populations of

P. dardanus can be quite variable, summaries on every group of

specimens reported here were either from collections from single

locales or the progeny of single crosses. Specimens were photo-

graphed with a 3 megapixel digital camera, and morphometrics

were done using digitization software developed for the purposes of

this study. The landmarks measured are shown in Fig. 2. These

landmarks were recorded as (x,y) coordinates, and distances

between various pars of coordinates were calculated using a

spreadsheet program (Microsoft Excel). In all cases, the effects of

body size were removed by regressing the measures of the size of a

pattern element on the mean of four distance measures of the

length and separation between wing veins (Fig. 2B, dotted lines)

and taking the residuals of the regression (using JMP, SAS

Institute, Cary, NC, USA). The data reported here refer to these

residuals. Repeatability of measurement was established by

digitizing 12 landmarks on 10 specimens, four different times, at

intervals of a week to a month. The mean coefficient of variation

(standard deviation� 100/mean) of repeatability on 18 distances

calculated from these measurements was 1.03, with a standard error

of 0.37. Correlations reported here are Pearson product-moment

correlations; 95% confidence intervals were estimated using

Fisher’s z-transformation (Sokal and Rohlf 1981; Paulsen and

Nijhout 1993). If the 95% confidence interval contained zero, the

correlation was assumed to not be significantly different from zero.

RESULTS

The definition of the female forms has been based almost

entirely on subjective qualitative descriptions of the color

patterns shown in Fig. 1. To make this comparison more

objective and to develop a quantitative database to study

morphological evolution, we have undertaken an extensive

morphometric analysis of pattern diversity and pattern

Polymorphic mimicry in P. dardanus 581Nijhout

variation. Here we restrict our analysis to the patterns on the

forewings, because these have undergone the greatest radia-

tion in form during the evolution of mimicry.

Correlated variation of pattern elementsin males and female morphs

The color patterns of P. dardanus are considerably simpler

than those of the well-studied nymphalid butterflies. The

elements of the nymphalid groundplan (Nijhout 1991) are

absent, and the entire pattern consists of two regions of black

pattern elements. One extends inward from the distal margin of

the wing and is present in both males and females. The other,

restricted to females, occupies the leading edge and proxi-

mal portion of the wing. As in other butterflies, color pattern

development is compartmentalized by the wing veins (Nijhout

1994; Koch and Nijhout 2002). As a consequence, the

portions of the pattern in each intervein region can become

developmentally uncoupled to various degrees (Nijhout 1985).

Elements of the color pattern that are developmentally

uncoupled will exhibit a certain degree of independent

variation from individual to individual. A high degree of

correlated variation between two elements indicates that they

share many developmental determinants. A low degree of

correlated variation indicates that variation of each portion of

the pattern is caused by different factors. Here we measure

only phenotypic correlations. In other butterflies, phenotypic

correlations are good predictors of genetic correlations

(Paulsen 1994, 1996), suggesting that phenotypic variation is

due to individual genetic variation rather than to individually

different environmental effects on each portion of the pattern.

The genetic correlation between different portions of the

pattern indicates the degree to which those portions are gene-

tically coupled; high genetic coupling is a constraint on the

ability of those parts to respond independently to selection.

The correlated variation in the width of the black margin

in adjoining intervein regions in males and females of

the monomorphic Madagascan race meriones are shown in

Table 1. In males the correlated variation declines with dis-

tance. In females the patterns in the anterior part of the wing

are correlated with each other, as are those in the posterior

portion of the wing, but there is no correlated variation

between those two regions of the wing.

We were able to obtain sufficiently large series (430

specimens from a single locale) of males of two additional

geographic races, polytrophus and tibbulus, for a comparative

correlation analysis. These results are shown in Table 2. In

tibbulus the correlations decline strongly with distance,

whereas in polytrophus the patterns remain highly correlated

despite increasing distance between them. Evidently, geo-

graphic races can differ considerably in the correlations

among various portions of their color pattern.

In all cases summarized in Tables 1 and 2, where the

decline in correlation with distance is strong, there appears to

Table 1. Correlations among the widths of the marginal

pattern elements (taken along the midline of each

wing cell) in females (A) and males (B) of the

monomorphic nonmimicking meriones race

A

6d 5d 2d5d 0.85

4d 0.61

3d −0.05 0.25

2d 0.27 0.42 0.81

1d 0.34 0.27 0.50

B

6d 5d

5d 0.92

4d 0.82 0.91

3d 0.73 0.81

2d 0.79 0.75

1d 0.68 0.560.71

0.79

0.32

0.53

0.53

3d4d 2d3d4d

0.93

0.870.84

0.820.560.55

Axes indicate the pattern element (see Fig. 2A). Bold numbers indicatecorrelations that are significantly different from zero.

Fig. 2.Diagrammatic representation of the forewing pattern. (A) Names of the wing veins and the names of the pattern elements as used inFigs. 3 and 4, and Table 5. (B) Coordinates recorded for digitization of pattern landmarks (circles) and the distance measures (lines) used inTables 1–4. The landmarks used for pattern element are the distances between a vein and the edge of a black portion of the pattern,measured along the midline of an intervenous area. These distances are indicated by solid lines, and the numbers on each line refer to thedistance measurements used in Table 3. Dotted lines indicate distance measures used to calculate the general size of the wing.


be a greater step at wing vein M3 or M2, so that the patterns

on each side of this boundary are significantly more highly

correlated with each other than they are with patterns on the

other side of this boundary. In many species of butterflies

there is an abrupt change in the shape or color of pattern

elements on either side of wing vein M3 (Nijhout 1991),

suggesting that some of the determinants of the color pattern

are different in the anterior and posterior portions of the

forewing.

We were also able to obtain a sufficiently large sample of

females from two mimicking forms, cenea and hippocoonides,

of the race polytrophus for correlation analysis. In these forms

the color pattern is more complicated in that in each wing

cells there are two black patterns: a distal pattern that extends

in from the wing margin and a proximal pattern that extends

out from the discal cell. The sizes of the patterns at the wing

cell midlines were measured, as illustrated in Fig. 2, and the

correlations among these are shown in Table 3. In the form

cenea only 4 of 66 correlations were significantly different

from zero, and in hippocoonides only 2 of 45 correlations

were significantly different from zero. By chance alone one

would expect three and two of the correlations to be

significant (at the 5% level) in cenea and hippocoonides,

respectively, so it appears that there are probably no

significant correlations among any portions of the pattern in

these two forms. The absence of correlated variation among

pattern elements in mimicking forms stands in contrast to the

neighbor and regional correlations observed in the nonmi-

metic patterns.

It is possible that the reduction of internal correlations

enabled the evolution of accurate mimicry. In the Nympha-

lidae there are generally low correlations among the various

elements of the color pattern (Nijhout 1991; Paulsen and

Nijhout 1993; Paulsen 1996), but it is not yet clear whether

this is also the case for the simpler patterns of the Papilionida.

The findings of Koch and Nijhout (2002) on the role of wing

veins in compartmentalization of the color pattern of Papilio

xuthus suggest that high correlated variation among the

patterns in different wing compartments is likely to be the

primitive condition in the Papilionidae.

Variability of the pattern

Our studies on the correlated variation of pattern elements

revealed a substantial amount of phenotypic variability in the

various forms of P. dardanus. Assuming a similar mutation

load, patterns that are subject to strong selection should

exhibit less genetic and phenotypic variability than patterns

that are under weaker selection. Accordingly, we measured

the variability of the color pattern in mimetic and nonmimetic

forms. We also measured the variability of patterns of three of

the interracial crosses done by Clarke and Sheppard (1960b,

1963), in which they observed a breakdown of mimicry. The

results are presented in Table 4 in the form of coefficients

of variation. To provide a basis of comparison, we also

measured the variability of several physical landmarks in the

venation pattern of the wing whose absolute dimensions were

in the same range as those of the various features of the color

pattern that were measured (Fig. 2).

The color patterns of the nonmimetic forms have

approximately the same degree of variability as the structural

features of the wing. Interestingly, the patterns of mimetic

females are substantially more variable than those of the

nonmimetic forms. This increased variability is not due to a

greater variability of the substrate on which the pattern

develops, because the coefficients of variation of their venat-

ion pattern are about the same as those of the nonmimetic

forms. The variability of the pattern of naturally occurring

imperfect mimics (niaviodes) and of the progeny of interracial

crosses is greater than that of the pure female forms. Table 4

partitions the variability of the non-male-like forms into that

attributable to variation in the marginal patterns (which they

have in common with the male-like forms) and the basal

patterns (which are unique to the female mimicking forms). It

is clear that the basal patterns are more variable than the

marginal patterns, although both sets of pattern elements are

more variable in the mimetic than in the nonmimetic forms.

Effects of the alleles of the mimicry locus

If the diversity of female forms evolved by the sequential

addition of mutations that affect different portions of the

pattern, then it would be interesting to know whether these

independent effects can be detected within today’s diversity

and variation of the pattern. For the studies that follow, we

established a character matrix for all the homozygous female

forms (Table 5). In Table 5, the size of each pattern element is

scored as the percent of the wing cell covered by the element,

based on the average of a range of 10–20 specimens. In the

Table 2. Correlations among the widths of the marginal

pattern elements (taken along the midline of each

wing cell) of males of the races polytrophus (A)and tibbulus (B)

A

6d 5d 4d 3d 2d

5d 0.81

4d 0.73

3d 0.66 0.91

2d 0.64 0.85

1d 0.60 0.93

0.88

0.74

0.76 0.89

0.67 0.77 0.77

B

6d 5d 4d 3d 2d

5d 0.74

4d 0.47

3d 0.36 0.82

2d 0.29 0.73

1d 0.21 0.89

0.69

0.41

0.31 0.87

0.21 0.54 0.71

Axes indicate the wing cells (see Fig. 2A). Bold numbers indicatecorrelations that are significantly different from zero.


scoring system, 1510%, 2520%, and so on. In cases where

the entire wing cell is black, the pattern is made up by fusion

of the proximal and distal pattern elements in that wing cell.

In such cases the score of each pattern element is given as the

largest value recorded for that element in all genotypes. In

addition, we measured the sizes of pattern elements for all the

types of heterozygotes of whose identity we could be certain.

These data are used to study the patterns of dominance

relationships and the relative dimensions of the effects of the

alleles of the mimicry gene.

Dominance relationships withingeographic races

When crosses are made between different genotypes within

a geographic race, the various alleles typically have a

Table 3. Correlations among the widths of the pattern elements in females of the mimetic

forms cenea (A) and hippocoon (B) from the race polytrophus

2d(2) 0.68

3d(3) 0.29

5d(5) −0.03 −0.14 0.01

6d(6) −0.01 −0.18 −0.22 0.53

1p(7) −0.08 0.18 0.45 0.01 0.04

2p(8) 0.06 0.03 −0.11 −0.06 0.089 −0.14

3p(9) −0.09 −0.09 0.00 −0.05 0.27 0.42 −0.35

5p(11) 0.08 0.16 0.28 −0.06 0.022 0.24 −0.29 0.04

6p(12) 0.07 0.08 0.12 0.08 −0.21 0.20 0.25 0.02 0.23

0.69

1d(1) 2d(2) 3d(3) 4d(4) 5d(5) 6d(6) 1p(7) 2p(8) 3p(9) 4p(10) 5p(11)

2d(2) 0.65

3d(3) 0.53 −0.16

3d(4) 0.11 0.21 0.57

5d(5) 0.40 0.56 0.31 0.19

6d(6) 0.14 0.20 0.31 0.20 0.66

1p(7) −0.05 0.14 −0.03 0.33 0.03 −0.12

2p(8) 0.03 −0.24 −0.12 0.30 −0.17 −0.44 0.65

3p(9) −0.37 −0.16 −0.50 −0.30 −0.28 −0.01 0.23 0.13

4p(10) −0.11 −0.07 −0.41 −0.40 −0.13 −0.02 0.095 0.30 0.72

5p(11) −0.12 0.01 0.11 0.09 −0.17 0.11 −0.12 −0.35 −0.08 −0.23

6p(12) −0.34 0.27 −0.43 −0.07 −0.29 −0.09 0.10 0.30 0.23 0.22 0.09

1d(1) 2d(2) 3d(3) 5d(5) 6(6) 1p(7) 2p(8) 3p(9) 5p(11)

A

B

Pattern elements are coded as shown in Fig. 2A. Parenthetical numbers next to pattern element refer to distancemeasures shown in Fig. 2B. Bold numbers indicate correlations that are significantly different from zero.


well-defined dominance relation to each other, and the

various female morphs segregate as clear Mendelian traits

(Clarke and Sheppard 1959, 1960a,b,e, 1962). Most hetero-

zygous genotypes exhibit full dominance of one of the alleles,

with only three notable exceptions. The niobe phenotype can

be obtained with the niobe allele of the mimicry locus (Hni)

but also as a heterozygote between the planemoides and

trophonius alleles (Hpl/HT), yielding the so-called synthetic

niobe (Clarke and Sheppard 1960a). The heterozygote

between the leighi and trophonius alleles (H l/HT) gives a

unique phenotype called salaami, and the heterozygote

between the planemoides and poultoni alleles (Hpl/Hp)

gives a unique phenotype called dorippoides (Clarke and

Sheppard 1960a).

Table 4. Coefficients of variation of the color pattern elements in Papilio dardanus

Venation All Pattern Elements Proximal Patterns Distal Patterns

meriones (female) 4.9 7.4 F 7.4

meriones (male) 4.4 8.3 F 8.3

tibullus male 11.9 9.0 F 29.2

polytrophus male 7.3 11.1 F 11.1

polytrophus cenea 5.9 26.9 33.8 16.3

polytrophus hippocoonides 5.8 21.2 36.3 16.6

dardanus hippocoonides 8.9 21.0 36.3 14.9

antinorii hippocoonides1 9.4 24.2 33.9 17.0

C&S broods 4687147812 7.8 32.1 44.6 21.3

C&S broods 4695147012 8.5 32.2 41.8 24.1

C&S brood 46832 8.1 34.1 46.2 21.9

Names of geographic races are italicized; names of female forms are in roman letters.1This form is usually called niavioides and is an imperfect mimic.2Data from female offspring of interracial crosses by Clarke and Sheppard (1963) and preserved in the British Museum, Natural History.

Table 5. Character matrix of pattern elements of the female forms of Papilio dardanus, and the two color forms of

Papilio phorcas

Wing Cell

Form 1d 2pd 2ad 3d 4pd 4ad 5d 6d 7d 1p 2pp 2ap 3p 4pp 4ap 5p 6p Dp Dm Dd

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

cenea 4 4 4 4 4 4 6 5 6 4 0 0 0 2 1 2 0 3 3 3

dionysos 1 2 2 4 1 1 2 4 8 0 0 0 0 2 1 0 0 1 1 2

dorippoides 1 2 2 2 1 1 2 3 8 0 0 0 0 0 0 0 0 3 3 0

hippocoonides 2 4 5 4 1 1 2 4 8 0 0 1 4 3 1 0 0 3 3 3

hippocoon 2 4 5 4 1 1 1 4 8 0 0 1 4 3 1 0 0 3 3 3

lamborni 2 3 3 4 1 1 3 4 8 0 0 0 4 2 1 1 0 3 3 3

leighi 2 4 6 4 1 1 2 4 8 4 1 1 4 3 1 0 0 3 3 3

meriones 1 2 2 3 4 4 5 5 8 0 0 0 0 0 0 0 0 1 2 0

natalica 2 4 6 4 1 1 3 4 8 2 1 1 4 3 1 1 0 3 3 3

niobe 1 2 2 4 1 1 2 4 8 0 0 0 4 3 1 0 0 3 3 3

ochracea 3 4 4 4 4 5 6 4 6 3 1 0 4 2 1 2 1 3 3 3

planemoides 1 2 2 2 2 1 3 4 8 3 1 1 2 2 1 0 0 3 3 1

poultoni 2 4 5 4 1 1 2 3 8 1 0 0 4 3 2 0 0 3 3 3

trophonius 1 2 4 4 1 1 2 3 8 0 0 1 3 3 1 0 0 3 3 3

salaami 1 2 3 4 1 1 1 1 8 0 0 0 4 2 1 0 0 3 3 3

dorippoides 1 2 2 2 1 1 2 3 8 0 0 0 0 0 0 0 0 3 3 0

synth niobe 1 2 3 4 1 1 2 4 8 0 0 0 4 2 1 0 0 3 3 3

P. phorcas yellow 4 4 5 5 5 5 5 5 8 3 2 2 1 0 0 0 1 3 3 3

P. phorcas green 4 4 5 5 5 5 5 5 8 1 1 0 0 0 0 1 2 3 3 3

The dimension of each pattern element is scored on a scale from 0 to 10, representing the fraction of the wing cell covered by black pattern, as describedin text. Wing cells are coded as in Fig. 2A. Wing cells 2 and 4 often have an asymmetrical pattern and independent measurements were taken of theposterior (p) and anterior (a) halves (e.g. 2pd is the posterior portion of the pattern in cell 2d).


The dominance relationships of the effects of the mimicry

gene on different portions of the pattern are illustrated in

left-hand panel of Fig. 3, for all the heterozygous combina-

tions for which we have been able to obtain clear

morphological data from the literature and from unpublished

breeding experiments. In most heterozygotes, one allele

typically has a dominant effect on all elements of the color

pattern, though there are several major exceptions to this

general rule. The first is the planemoides (Hpl) allele, which

appears to have a mix of dominance effects on different

portions of the pattern. The second is the case of the

heterozygotes with unique phenotypes (the forms salaami,

dorippoides, and synthetic niobe). In these forms, several

pattern elements exhibit overdominance, meaning that the

phenotypic value of the heterozygote is greater than that of

either homozygote. Although overdominance for fitness is

widespread and well understood (as heterozygote superiority),

overdominance for morphological traits is exceptionally

uncommon, and the developmental mechanisms by which it

could arise are not obvious.

The cenea (H c) allele exhibits two patterns of dominance

with respect to members of the hippocoon group. It is

dominant with respect to the hippocoon (Hh) and natalica

(Hna) alleles and recessive to all the rest. This suggests that

members of the hippocoon group belong to two genetic

classes, perhaps reflecting different episodes in the evolution

of mimicry.

Dominance relationships betweengeographic races

The relatively clean patterns of dominance obtained from

crosses within a geographic race stand in contrast to the

results obtained when crosses are made between geographic

races. In such cases, one obtains highly variable heterozygous

phenotypes that differ from the within-race crosses in that no

two individuals are quite identical in their pattern and that the

outlines of the patterns are often irregular and poorly defined.

Clarke and Sheppard (1960b,c,e, 1963) suggested that this

variable dominance was due to the fact that within a

Fig. 3. Matrices of the pleiotropic effects of various alleles of the mimicry gene on different elements of the color pattern. Genotypes aregrouped by allele in the left-most column (alleles are designated by the superscripts shown in Fig. 1). Many of the genotypes appear twice,grouped by each of their alleles. (Left) Dominance relationships among the alleles refer to the alleles in the left-most column: black,dominant; red, recessive; blue, codominant; green, overdominant; gray, no difference in pattern. Interracial hybrids and uniqueheterozygous patterns (see text) are indicated. (Right) Morphic effects of various alleles on different elements of the color pattern. Morphiceffects refer to alleles in the left-most column: black, hypermorphic (large pattern dominant to small pattern); red, hypomorphic (largerecessive to small); blue; codominant (intermediate between homozygotes); green, overdominant. In all cases, the overdominant pattern issmaller than that of both allelic homozygotes.


population the mimicry alleles occur in a coadapted

genetic background. When this coadapted gene complex is

broken up by hybridization between geographic races, the

mimicry gene no longer produces a well-defined and stable

phenotype.

Figure 3 shows that the clean dominance relationships of

pattern elements break down in heterozygotes from interracial

crosses. Here we illustrate crosses using the ‘‘yellow’’ Hya

allele (of the antinorii race from Ethiopia) and Hym allele (of

the meriones race from Madagascar). In interracial hybrids,

these alleles have different dominance relationships for

different elements of the pattern: some elements are dominant,

whereas others are recessive in the same heterozygote. The

pleiotropic patterns of dominance are quite different for the

two alleles, suggesting that these two male-like forms are

genetically quite dissimilar and may therefore have indepen-

dent evolutionary origins.

This mosaic dominance of the pleiotropic effect of the

mimicry gene may be related to the differential dominance of

pleiotropic QTL effects described in mouse mandibles by

Ehrlich et al. (2003). In P. dardanus, however, the mosaic

dominance appears only in the aberrant genetic environment

of interracial hybrids that show a breakdown of mimicry. The

evolution of a coadapted gene complex that supports accurate

mimicry within a geographic race appears to be associated

with the loss of mosaic dominance. It is not obvious at present

what is gained by having the same sign of dominance of all

the pleiotropic effects of a gene, but it is a very clear

evolutionary signal, so it must have had some consequences

for fitness.

Hyper- and hypomorphic effects

In addition to the dominance effects on different portions of

the pattern, each allele also has a unique effect on the

dimensions of those pattern elements. An allele can have

either a hypermorphic effect, if it makes a pattern element

larger than the allele with which it is compared, or a

hypomorphic effect, if it makes the pattern element smaller. It

is most common for dominant alleles to have hypermorphic

effects on the trait they control. The morphic effects of the

various alleles of the mimicry gene on different portions of

the pattern are illustrated in the right-hand panel of Fig. 3.

The alleles clearly have a mixture of morphic effects on the

different portions of the color pattern. Dominant alleles do

not make all elements of the pattern larger than the recessive

alleles with which they are juxtaposed. Indeed, most effects of

dominant alleles are hypomorphic. The mixture of morphic

effects indicates that the mimicry allele does not affect each

pattern element through the same physiological mechanism.

In many (but not all) cases, the morphic effects are similar in

adjoining blocks of pattern elements, indicating, perhaps, that

these blocks share a common genetic control.

Big effects may give insight into the originof mimicryEach allele of the mimicry gene either enlarges or reduces

different portions of the color pattern, but the magnitude of

this effect is not the same for all parts of the pattern. Insofar

as the initial step in the evolution of mimicry is likely to have

been due to a genetic effect of large magnitude, we

investigated whether the relative dimensions of the phenotypic

effects of modern mimicry alleles could be used to detect the

footprint of the original mutation of big effect. We determined

the size of the effect of an allele on a given pattern element by

the difference of the score of that element in the homozygotes

and all available heterozygotes for each allele. Difference

scores of three or greater are given in Fig. 4. The big effects

were found to be restricted to a relatively small set of

locations on the wing (Fig. 4). The overall effect is to reduce

the size of the marginal elements in wing cells 1, 2, 4, and 5

and the proximal element in wing cell 1 and to increase the

size of the proximal element in wing cell 3.

To determine whether these big effects could have

produced a passable mimicry when superimposed on the

ancestral pattern of P. dardanus, we need to understand what

that ancestral pattern must have looked like. Most authors

have assumed that the male pattern of P. dardanus represents

the ancestral form of the pattern from which the female forms

must have been derived (Ford 1936; Clarke and Sheppard

1960c; Turner 1963). This assumption is presumably based on

the homogeneity of male forms across the species’ range and

the fact that several geographic races possess male-like female

forms. The pattern of male P. dardanus is, however, very

different from that of other papilionids, and it seems more

reasonable to seek the ancestral pattern among the species

that are most closely related to P. dardanus.

There is general agreement that P. dardanus is most closely

related to P. phorcas and P. constantinus, although authors

differ on which of these species is the sister species to

P. dardanus (Vane-Wright and Smith 1991, 1992; Clarke et al.

1991; Caterino and Sperling 1999; Vane-Wright et al. 1999).

Of the two, P. phorcas is polymorphic, with boldly patterned

black and green males and dimorphic females (Fig. 5). One of

the female morphs is male-like, and the other has a yellow and

brown pattern that has a general similarity to that of related

papilionids. Papilio constantinus is monomorphic, and both

males and females have a pattern that resembles that of the

brown female morph of P. phorcas.

It seems possible, therefore, that the polymorphism of

P. dardanus is not uniquely derived but is related to the

pattern dimorphism of its sister species, P. phorcas. The

ecological function of the polymorphic pattern of P. phorcas

is unclear, because neither of the morphs appears to be a

mimic. The black and green male morph is clearly the derived

pattern, so it may be maintained by sexual selection, as is

believed to be the case for the male pattern of P. dardanus


(Ford 1953; Sheppard 1959). The polymorphic female form of

P. phorcas is believed to have originated as a male-mimicking

‘‘transvestitism’’ from a primitively sexually dimorphic color

pattern (Vane-Wright 1976; Clarke et al. 1985).

If P. phorcas and P. dardanus share a recent common

ancestor, then the evolution of a female polymorphic pattern

in P. dardanus may have been facilitated by a preexisting

polymorphism resembling that of P. phorcas. If this is the

case, then the male pattern of P. dardanus may have been

derived from the male pattern of P. phorcas and the female

polymorphism of P. dardanus from the female dimorphism of

P. phorcas. If we take the patterns of P. phorcas to represent

the primitive extreme, then we may have a more reasonable

basis from which to deduce the evolution of the P. dardanus

patterns than if we assume the male P. dardanus pattern to be

primitive.

This possibility was explored by examining the big

effects of the various alleles of the mimicry gene. Figure 6

illustrates the pattern modifications obtained when the

pattern changes of big effect (from Fig. 4) are applied to the

yellow female pattern of P. phorcas. The altered pattern bears

a passable resemblance to a hippocoon-like target phenotype.

So it seems possible that this small suite of changes could have

given rise to the hippocoon class of patterns.

Fig. 5. Patterns of Papilio phorcas. All males are of the black and green phenotype (A), whereas females are dimorphic and can be eitherblack and green (A) or yellow and brown (B).

Fig. 4. Matrix of big effects of the various alleles on different elements of the color pattern. Gray, difference score53; black, differencescore54. Diagram of forewing illustrates the locations and polarity of the big effects of the mimicry gene on the elements of the colorpattern.


The transition from a phorcas-like pattern to a cenea-like

pattern is more difficult. Using only the pattern elements that

express the big effect, it would require an enlargement of 3p,

1p, 1d, and 4d and no change in 2d and 4d. This suite of

changes is inconsistent with the observed correlation of

changes within this set of pattern elements (3p becoming

larger and all others becoming smaller, or vice versa with the

alternative allele). A cenea-like pattern can, however, be

achieved from a hippocoon-like pattern by increasing 1p, 1d,

2d, 4, and 5d and keeping 3p unchanged (Fig. 6). This change

involves a simple partial reversal of the changes that led from

phorcas-like to hippocoon-like, requiring only that the

magnitude of the reversal is greater in cells 1 and 5 so that

these cells become completely filled with the black pattern.

To achieve the resemblance to a model, the altered black

portion of the pattern must be accompanied by a change in

background pigmentation. A suppression of background pig-

mentation to achieve a white field is sufficient. In both the pre-

hippocoon and the pre-cenea patterns, relatively minor addi-

tional modifications need to be applied to achieve a remark-

ably close resemblance to a target model phenotype (Fig. 6).

Whether these transformations resemble the actual effect

of early mutations and whether they provide sufficient

resemblance to distasteful species to improve the fitness of

its bearer is, obviously, a matter of conjecture. It seems,

nevertheless, that it is possible to evolve at least two of the

major mimicry patterns in relatively few steps.

The third main mimicry pattern is that of the planemoides

form (Hpl). This could be derived most simply from the P.

phorcas pattern by narrowing the distal black portion of the

pattern along the entire length of the wing (Fig. 7) and by

altering the background pigmentation to a reddish brown.

The changes required to produce the planemoides pattern are

relatively simple and are not related to the previous two. The

scenario outlined above thus suggests that a hippocoon-like

form arose first and that this form gave rise to the cenea-like

form and that the planemoides-like form arose independently.

So the first step in the evolution of mimicry could involve

only a single locus. The hippocoon-like pattern is likely to

have evolved twice (see Discussion), and the planemoides and

cenea-like patterns may have originated once each. The

further refinement of any one of the female forms could be

accomplished by a relatively small number of additional

genetic changes.

DISCUSSION

The mimetic wing patterns of P. dardanus have different

patterns of variation and internal correlation than nonmi-

Fig. 6. Imposing the big effects of the mimicry gene (see Fig. 4) on the presumptive ancestral Papilio phorcas-like pattern. A reasonably closeresemblance to a hippocoon pattern is obtained in a single step. A single-step transformation to a cenea-like pattern requires pattern changesthat are not supported by the big effects of the mimicry gene. Transformation of a hippocoon-like pattern to a cenea-like pattern, however,requires only reversals of some of the big effects.


metic ones. Mimetic patterns exhibit a higher degree of

individual variability than nonmimetic patterns. If the degree

of variability is a function of the mutation-selection balance,

then these findings suggest that nonmimetic patterns either

are under stronger selection or experience less mutational

variance than mimetic patterns. It is possible that the greater

individual variability of mimetic pattern is due to the fact that

these genotypes are more protected against predators and

thus less subject to selection and that there is, for some reason,

stronger normalizing selection on nonmimetic and male

patterns than on mimetic patterns. Alternatively, it is possible

that mimetic patterns require the activity of many more genes

than do nonmimetic ones and would therefore be more

sensitive to standing genetic variation and newly introduced

mutational variance.

Mimetic wing patterns exhibit a lower degree of correlated

variation among the various elements of the pattern than do

nonmimetic wing patterns. If the pattern of genetic correla-

tions resembles the pattern of phenotypic correlations, as is

the case in other butterflies (Paulsen 1994, 1996), then these

findings suggest few if any genetic correlations among the

elements of the color pattern of the mimicking forms. The

absence of correlation in the mimicking patterns may be due

to the accumulation of the many genes with small effects that

have been hypothesized to be responsible for refining the

details of the mimicry (the second stage in the evolution of

mimicry). If this is the correct interpretation, then it implies

that these modifier genes affect the morphology of the color

pattern on a spatial scale that is the same size or smaller than

the scale at which these measurements were made.

The various alleles of the mimicry gene affect the

dimension of the pattern elements to different degrees. Here

the effect is rather complex in that a given allele may increase

the size of some pattern elements while decreasing the size of

others. This is unlike the case in Heliconius, where dominant

alleles typically increase the size of a pattern element (Nijhout

et al. 1990), although a recent report shows substantial

evolution of dominance in Heliconius (Naisbit et al. 2003). In

P. dardanus the effect on the size of the pattern is independent

of the dominance effect of the allele. Thus, a dominant effect

can either increase or decrease the size of an element, with the

latter being somewhat more common than the former. This

mix of hyper- and hypomorphic pleiotropic effects of a given

allele is difficult to explain on the basis of a common

mechanism of action. This effect is consistent with both the

supergene and the regulatory gene hypotheses. In either case,

it can be best explained if variation in each pattern element is

affected by at least some unique genes or by unique

combinations of genes. Whether these genes are linked in a

supergene, as in the case of Papilio memnon (Clarke et al.

1968; Clarke and Sheppard 1971), or whether they are

dispersed across the genome and controlled by a polymorphic

regulatory gene remains an open question.

The alleles of the mimicry locus have a complex set of

pleiotropic and dominance effects on various features of the

forewing pattern. Each allele affects a different combination

of pattern elements, and the size of each of the pattern

elements affected by a given allele is altered to a different

degree. Within a geographic race, the different alleles show a

clear and unambiguous pattern of dominances, in which the

degree of dominance is the same for all elements affected by a

given allele. This stands in contrast to the mosaic pattern of

dominance observed in interracial crosses, where a given allele

may have a dominant effect on some elements and a recessive

effect on others. The similarity of dominance effects within a

race may be a manifestation of the so-called coadaptation of

genes within a population. The mosaic of dominances in

interracial crosses would then be due to a breakdown of this

coadaptation when an allele is placed in a genetic background

that is different from the one in which it evolved. If this is the

correct interpretation, it suggests that the similarity of

dominance effects within a population is an evolved property.

It is possible that when a mimicry allele first originated, it

had a mosaic of dominant and recessive pleiotropic effects on

Fig. 7.Derivation of the planemoides from Papilio phorcas is accomplished by simple diminution of the proximal and distal black areas (anda change in background color).


different portions of the pattern, just as modern alleles have a

mosaic of morphic effects on different parts of the pattern.

Perhaps initially all dominant effects were associated with

increases in the size of pattern elements and all recessive

effects with decreases in size. A favorable combination of

hyper- and hypomorphic pleiotropic effects could then be

stabilized by the evolution of a genetic background that made

all those effects dominant (e.g., Gilchrist and Nijhout 2001).

This process would be repeated with the origin of each new

mimetic locus, resulting in the order of dominance observed.

This scenario for the evolution of dominance relations is

different from the traditional Haldane’s sieve, which suggests

that a new favorable allele can be established most easily if it

is dominant than if it is recessive so that new phenotypes tend

to be dominant over preexisting phenotypes (e.g., Clarke et al.

1985). Under both scenarios, the order of dominance is likely

to reflect the order of origin of the phenotypes, but under

Haldane’s sieve this would be a purely statistical effect,

whereas under the alternative scenario suggested above,

dominance is an evolved adaptation.

The hippocoon pattern group is the largest and most

diverse of the three pattern groups in P. dardanus (Fig. 1). The

fact that the cenea allele (H c) is dominant to some alleles in

the hippocoon group (Hh and Hna) and recessive to others

suggests that the hippocoon-like phenotype may have evolved

independently more than once. This interpretation is sup-

ported by the finding on the dimensions of the phenotypic

effects of the various alleles on the pattern. Application of the

big phenotypic effects to the presumptive primitive pattern of

the P. dardanus ancestor suggests that a hippocoon-like mimic

can be achieved in one step but that a cenea-like mimic most

likely requires at least two evolutionary steps and is probably

derived from a hippocoon-like pattern rather than indepen-

dently from the ancestral pattern. It is possible, then, that the

cenea allele (H c) evolved from the group that led to the

hippocoon (Hh) and natalica (Hna) alleles, to which it is

dominant. The other alleles in the hippocoon group may have

been derived independently and at a later time, and this may

explain why they are dominant to the cenea allele. This

scenario is in accordance with the suggestion of Turner (1963)

that the hippocoon form probably arose before any of the

other mimetic forms. Clarke et al. (1985, 1991) argued that in

Papilio phorcas, the male-like pattern of the female is the

derived form. This suggests that the species may initially have

been sexually dimorphic (with brown/yellow females and

black/green males) and that a so-called transvestite evolu-

tionary step (Vane-Wright 1976; Clarke et al. 1985) produced

male-like females and was the origin of the female color

dimorphism.

AcknowledgmentsThis work was supported by a grant from the Human FrontiersScience Program. I thank Alfried Vogler, Ali Cieslak, Toshio

Sekimura, Dick Vane-Wright, Susan Paulsen, and, in particular,Steve Collins for many helpful and insightful discussions and TreyGreer for developing DIGIT, the digitization software used in thisstudy.

REFERENCES

Bernardi, G., Pierre, J., and Nguyen, T. H. 1985. Le polymorphisme et lemimetisme de Papilio dardanus Brown. Bul. Soc. Entomo. France 90:106–155.

Carpenter, G. D. H., and Ford, E. B. 1933. Mimicry. Methuen, London.Caterino, M. S., and Sperling, F. A. H. 1999. Papilio phylogeny based on

mitochondrial cytochrome oxidase I and II genes. Mol. Phylogene. Evol.11: 122–137.

Charlesworth, B. 1994. The genetics of adaptation: lessons from mimicry.Am. Nat. 144: 839–847.

Charlesworth, D., and Charlesworth, B. 1975a. Theoretical genetics ofBatesian mimicry. I. Single-locus models. J. Theoret. Biol. 55: 283–303.

Charlesworth, D., and Charlesworth, B. 1975b. Theoretical genetics ofBatesian mimicry. II. Evolution of supergenes. J. Theoret. Biol. 55: 305–324.

Charlesworth, D., and Charlesworth, B. 1975c. Theoretical genetics ofBatesian mimicry. III. Evolution of dominance. J. Theoret. Biol. 55: 325–337.

Clarke, C., Clarke, F. M.M., Collins, S. C., Gill, A. C. L., and Turner, J. R.G. 1985. Male-like females, mimicry and transvestitism in butterflies(Lepidoptera: Papilionidae). Syst. Entomol. 10: 257–283.

Clarke, C. A., Gordon, I. J., Smith, C. R., and Vane-Wright, R. I. 1991.Phylogenetic relationships of three African swallowtail butterflies, Papiliodardanus, P. phorcas and P. constantinus: new data from hybrids(Lepidoptera: Papilionidae). Syst. Entomol. 16: 257–273.

Clarke, C. A., and Sheppard, P. M. 1959. The genetics of Papilio dardanusBrown. I. Race cenea from South Africa. Genetics 44: 1347–1358.

Clarke, C. A., and Sheppard, P. M. 1960a. The genetics of Papilio dardanusBrown. II. Races dardanus, polytrophus, meseres, and tibullus. Genetics45: 439–457.

Clarke, C. A., and Sheppard, P. M. 1960b. The genetics of Papilio dardanusBrown. III. Race antinorii from Abyssinia and race meriones fromMadagascar. Genetics 45: 683–698.

Clarke, C. A., and Sheppard, P. M. 1960c. The evolution of mimicry in thebutterfly Papilio dardanus. Heredity 14: 163–173.

Clarke, C. A., and Sheppard, P. M. 1960d. Super-genes and mimicry.Heredity 14: 175–185.

Clarke, C. A., and Sheppard, P. M. 1960e. The evolution of dominanceunder disruptive selection. Heredity 14: 23–87.

Clarke, C. A., and Sheppard, P. M. 1962. The genetics of Papilio dardanusBrown. IV. Data on race ochracea, race flavicornis, and furtherinformation on races polytrophus and dardanus. Genetics 47: 909–920.

Clarke, C. A., and Sheppard, P. M. 1963. Interactions between major genesand polygenes in the determination of the mimetic patterns of Papiliodardanus. Evolution 17: 404–413.

Clarke, C. A., and Sheppard, P. M. 1971. Further studies on the genetics ofthe mimetic butterfly Papilio memnon. L. Philos. Trans. R. Soc. Lond. B263: 35–70.

Clarke, C. A., Sheppard, P. M., and Thornton, I. W. B. 1968. The geneticsof the mimetic butterfly Papilio memnon. L. Philos. Trans. R. Soc. Lond.B 254: 37–89.

Ehrlich, T. H., Vaughn, T. T., Koreishi, S. F., Linsey, R. B., Pletscher, L. S.,and Cheverud, J. M. 2003. Pleiotropic effects on mandibular morphol-ogy. I. Developmental morphological integration and differential domi-nance. Mol. Dev. Evol. 296B: 58–79.

Fisher, R. A. 1930. The Genetical Theory of Natural Selection. OxfordUniversity Press, Oxford.

Ford, E. B. 1936. The genetics of Papilio dardanus Brown (Lep.). Trans.Entomol. Soc. Lond. 85: 435–466.

Ford, E. B. 1953. The genetics of polymorphism in the Lepidoptera. Adv.Genet. 5: 43–87.

Gilchrist, M. A., and Nijhout, H. F. 2001. Nonlinear developmentalprocesses as sources of dominance. Genetics 159: 423–432.


Haldane, J. B. S. 1924. A mathematical theory of natural and artificialselection. Part I. Trans. Cambr. Phil. Soc. 23: 19–41.

Koch, P. B., and Nijhout, H. F. 2002. The role of wing veins in colourpattern development in the butterfly Papilio xuthus (Lepidoptera:Papilionidae). Eur. J. Entomol. 99: 67–72.

Naisbit, R. E., Jiggins, C. D., and Mallett, J. 2003. Mimicry: developmentalgenes that contribute to speciation. Evol. Dev. 5: 269–280.

Nijhout, H. F. 1985. Independent development of homologous patternelements in the wing patterns of butterflies. Dev. Biol. 108: 146–151.

Nijhout, H. F., Wray, G. A., and Gilbert, L. E. 1990. An analysis of thephenotypic effects of certain color pattern genes in Heliconius(Lepidoptera: Nymphalidae). J. Linn. Soc. 40: 357–372.

Nijhout, H. F. 1991. The Development and Evolution of Butterfly WingPatterns. Smithsonian Institution Press, Washington, DC.

Nijhout, H. F. 1994. Developmental perspectives on the evolution ofbutterfly mimicry. BioScience 44: 148–157.

Paulsen, S. M. 1994. Quantitative genetics of butterfly wing patterns. Dev.Genet. 15: 79–91.

Paulsen, S. M. 1996. Quantitative genetics of the wing color pattern in thebuckeye butterfly (Precis coenia and Precis evarete): evidence against theconstancy of G. Evolution 50: 1585–1597.

Paulsen, S. M., and Nijhout, H. F. 1993. Phenotypic correlation structureamong elements of the color pattern in Precis coenia (Lepidoptera:Nymphalidae). Evolution 47: 593–618.

Poulton, E. B. 1924. Papilio dardanus, the most wonderful butterfly in theworld. J. East Afr. Uganda Nat. Hist. Soc. 209: 4–22.

Sheppard, P. M. 1959. The evolution of mimicry; a problem inecology and genetics. Cold Spring Harbor Symp. Quant. Biol. 24:131–140.

Sokal, R. R., and Rohlf, F. J. 1981. Biometry. Freeman, San Francisco, CA.Turner, J. R. G. 1963. Geographical variation and evolution in the males of

the butterfly Papilio dardanus Brown (Lepidoptera: Papilionidae). Trans.R. Entomol. Soc. Lond. 115: 239–259.

Turner, J. R. G. 1977. Butterfly mimicry: the genetical evolution of anadaptation. Evol. Biol. 10: 163–206.

Vane-Wright, R. I. 1976. An alternative hypothesis on the evolution ofPapilio dardanus Brown. Proc. R. Entomol. Soc. Lond. 41: 1–7.

Vane-Wright, R. I., Raheem, D. C., Cieslak, A., and Vogler, A. P. 1999.Evolution of the mimetic African swallowtail butterfly Papilio dardanus:molecular data confirm relationships with P. phorcas and P. constantinus.Biol. J. Linn. Soc. 66: 215–229.

Vane-Wright, R. I., and Smith, C. R. 1991. Phylogenetic relationships ofthree African swallowtail butterflies, Papilio dardanus, P. phorcas and P.constantinus: cladistic analysis (Lepidoptera: Papilionidae). Syst. En-tomol. 16: 275–291.

Vane-Wright, R. I., and Smith, C. R. 1992. Occurrence and significance ofnatural hybrids between Papilio dardanus and P. phorcas (Lepidoptera:Papilionidae). Syst. Entomol. 17: 269–272.


polymorphic mimicry in papilio dardanus: mosaic dominance, big...

Documents