Polymorphic mimicry in Papilio dardanus: mosaic

EVOLUTION & DEVELOPMENT
5:6, 579–592 (2003)
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, has
a female-limited polymorphic mimicry. This polymorphism is
controlled by allelic variation at a single locus with at least 11
alleles. Many of the alternative morphs are accurate mimics of
different species of distasteful butterflies. Geneticists have
long been interested in the mechanism by which a single gene
can have such diverse and profound effects on the phenotype
and in the process by which these complex phenotypic effects
could have evolved. Here we present the results of a morphometric analysis of the pleiotropic effects of the mimicry gene
on the array of elements that makes up the overall pattern.
We show that the patterns controlled by mimicking alleles
are more variable and less internally correlated than those
controlled by nonmimicking alleles, suggesting the two are
subject to different degrees of selection and mutational
variance. Analysis of the pleiotropic dominance of the alleles
reveals a consistent pattern of dominance within a coevolved
genetic background and a mosaic pattern of dominant and
recessive effects (including overdominance) in a heterologous
genetic background. The alleles of the mimicry gene have big
effects on some pattern elements and small effects on others.
When the array of big phenotypic effects of the mimicry gene
is applied to the presumptive ancestral color pattern, it
produces a reasonable resemblance to distasteful models
and suggests the initial steps that may have produced the
mimicry as well as the polymorphism.
INTRODUCTION
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
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
& BLACKWELL PUBLISHING, INC.
579
580
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
Fig. 1. The homozygous female forms of Papilio dardanus. The specimens are arranged in three groups based on their pattern. The black
portions are considered pattern, and the colored portions are background. The names of the forms and their allelic designation are given
below 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 the
planemoides group. A male is shown underneath group 2. Males are monomorphic and identical for all genotypes.
the species, but three of these alleles (H c, H h, and H T)
produce significantly different phenotypes in different populations (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 hippocoonides; 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 (H T) 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-
Nijhout
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 maintained 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 malelike 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.
Polymorphic mimicry in P. dardanus
581
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 variation 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 photographed 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
582
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
Fig. 2. Diagrammatic representation of the forewing pattern. (A) Names of the wing veins and the names of the pattern elements as used in
Figs. 3 and 4, and Table 5. (B) Coordinates recorded for digitization of pattern landmarks (circles) and the distance measures (lines) used in
Tables 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 the
distance measurements used in Table 3. Dotted lines indicate distance measures used to calculate the general size of the wing.
variation. Here we restrict our analysis to the patterns on the
forewings, because these have undergone the greatest radiation in form during the evolution of mimicry.
Correlated variation of pattern elements
in 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 proximal 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 genetically 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 distance. 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, geographic 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
5d 0.85
B
4d
3d
2d
6d
5d 0.92
5d
4d
3d
2d
4d 0.61 0.79
4d 0.82 0.91
3d −0.05 0.25 0.32
3d 0.73 0.81 0.93
2d 0.27 0.53 0.42 0.81
2d 0.79 0.75 0.84 0.87
1d 0.34 0.53 0.27 0.50 0.71
1d 0.68 0.56 0.55 0.56 0.82
Axes indicate the pattern element (see Fig. 2A). Bold numbers indicate
correlations that are significantly different from zero.
Nijhout
Polymorphic mimicry in P. dardanus
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)
B
A
6d
5d
4d
3d
6d
2d
5d
4d
3d
583
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.
2d
5d
0.81
5d
0.74
4d
0.73 0.88
4d
0.47 0.69
3d
0.66 0.74 0.91
3d
0.36 0.41 0.82
2d
0.64 0.76 0.85 0.89
2d
0.29 0.31 0.73 0.87
1d
0.60 0.67 0.77 0.77 0.93
1d
0.21 0.21 0.54 0.71 0.89
Axes indicate the wing cells (see Fig. 2A). Bold numbers indicate
correlations that are significantly different from zero.
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 nonmimetic patterns.
It is possible that the reduction of internal correlations
enabled the evolution of accurate mimicry. In the Nymphalidae 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.
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 venation 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
584
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
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
A
1d(1)
2d(2)
3d(3)
4d(4)
5d(5)
6d(6)
1p(7)
2p(8)
3p(9)
4p(10)
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
3p(9)
5p(11)
5p(11)
0.09
B
1d(1)
2d(2)
3d(3)
5d(5)
6(6)
1p(7)
2p(8)
2d(2)
0.68
3d(3)
0.29
0.69
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
0.20
0.25
0.02
−0.21
0.23
Pattern elements are coded as shown in Fig. 2A. Parenthetical numbers next to pattern element refer to distance
measures shown in Fig. 2B. Bold numbers indicate correlations that are significantly different from zero.
scoring system, 1 5 10%, 2 5 20%, 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 within
geographic races
When crosses are made between different genotypes within
a geographic race, the various alleles typically have a
Nijhout
Polymorphic mimicry in P. dardanus
585
Table 4. Coefficients of variation of the color pattern elements in Papilio dardanus
Venation
meriones (female)
meriones (male)
tibullus male
polytrophus male
polytrophus cenea
polytrophus hippocoonides
dardanus hippocoonides
antinorii hippocoonides1
C&S broods 4687147812
C&S broods 4695147012
C&S brood 46832
All Pattern Elements
4.9
4.4
11.9
7.3
5.9
5.8
8.9
9.4
7.8
8.5
8.1
Proximal Patterns
Distal Patterns
F
F
F
F
33.8
36.3
36.3
33.9
44.6
41.8
46.2
7.4
8.3
29.2
11.1
16.3
16.6
14.9
17.0
21.3
24.1
21.9
7.4
8.3
9.0
11.1
26.9
21.2
21.0
24.2
32.1
32.2
34.1
Names of geographic races are italicized; names of female forms are in roman letters.
1
This form is usually called niavioides and is an imperfect mimic.
2
Data from female offspring of interracial crosses by Clarke and Sheppard (1963) and preserved in the British Museum, Natural History.
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 heterozygous 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 (H ni)
but also as a heterozygote between the planemoides and
trophonius alleles (H pl/H T), yielding the so-called synthetic
niobe (Clarke and Sheppard 1960a). The heterozygote
between the leighi and trophonius alleles (H l/H T) gives a
unique phenotype called salaami, and the heterozygote
between the planemoides and poultoni alleles (H pl/H p)
gives a unique phenotype called dorippoides (Clarke and
Sheppard 1960a).
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
cenea
dionysos
dorippoides
hippocoonides
hippocoon
lamborni
leighi
meriones
natalica
niobe
ochracea
planemoides
poultoni
trophonius
salaami
dorippoides
synth niobe
P. phorcas yellow
P. phorcas green
1
4
1
1
2
2
2
2
1
2
1
3
1
2
1
1
1
1
4
4
2
4
2
2
4
4
3
4
2
4
2
4
2
4
2
2
2
2
4
4
2
4
2
2
5
5
3
6
2
6
2
4
2
5
4
3
2
3
5
5
3
4
4
2
4
4
4
4
3
4
4
4
2
4
4
4
2
4
5
5
3
4
1
1
1
1
1
1
4
1
1
4
2
1
1
1
1
1
5
5
3
4
1
1
1
1
1
1
4
1
1
5
1
1
1
1
1
1
5
5
3
6
2
2
2
1
3
2
5
3
2
6
3
2
2
1
2
2
5
5
4
5
4
3
4
4
4
4
5
4
4
4
4
3
3
1
3
4
5
5
5
6
8
8
8
8
8
8
8
8
8
6
8
8
8
8
8
8
8
8
0
4
0
0
0
0
0
4
0
2
0
3
3
1
0
0
0
0
3
1
0
0
0
0
0
0
0
1
0
1
0
1
1
0
0
0
0
0
2
1
0
0
0
0
1
1
0
1
0
1
0
0
1
0
1
0
0
0
2
0
0
0
0
0
4
4
4
4
0
4
4
4
2
4
3
4
0
4
1
0
0
2
2
0
3
3
2
3
0
3
3
2
2
3
3
2
0
2
0
0
0
1
1
0
1
1
1
1
0
1
1
1
1
2
1
1
0
1
0
0
0
2
0
0
0
0
1
0
0
1
0
2
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
2
1
3
1
3
3
3
3
3
1
3
3
3
3
3
3
3
3
3
3
3
2
3
1
3
3
3
3
3
2
3
3
3
3
3
3
3
3
3
3
3
0
3
2
0
3
3
3
3
0
3
3
3
1
3
3
3
0
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 described
in 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 the
posterior (p) and anterior (a) halves (e.g. 2pd is the posterior portion of the pattern in cell 2d).
586
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
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 combinations 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 (H pl) 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 (H h) and natalica
(H na) 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 between
geographic 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 are
grouped 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 unique
heterozygous patterns (see text) are indicated. (Right) Morphic effects of various alleles on different elements of the color pattern. Morphic
effects refer to alleles in the left-most column: black, hypermorphic (large pattern dominant to small pattern); red, hypomorphic (large
recessive to small); blue; codominant (intermediate between homozygotes); green, overdominant. In all cases, the overdominant pattern is
smaller than that of both allelic homozygotes.
Nijhout
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 independent 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.
Polymorphic mimicry in P. dardanus
587
Big effects may give insight into the origin
of mimicry
Each 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
588
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
Fig. 4. Matrix of big effects of the various alleles on different elements of the color pattern. Gray, difference score 5 3; black, difference
score 5 4. Diagram of forewing illustrates the locations and polarity of the big effects of the mimicry gene on the elements of the color
pattern.
(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 either
black and green (A) or yellow and brown (B).
Nijhout
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 pigmentation to achieve a white field is sufficient. In both the prehippocoon and the pre-cenea patterns, relatively minor additional modifications need to be applied to achieve a remarkably 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,
Polymorphic mimicry in P. dardanus
589
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 close
resemblance to a hippocoon pattern is obtained in a single step. A single-step transformation to a cenea-like pattern requires pattern changes
that 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.
590
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
Fig. 7. Derivation of the planemoides from Papilio phorcas is accomplished by simple diminution of the proximal and distal black areas (and
a change in background color).
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 correlations 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
Nijhout
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 (H h and H na) and recessive to others
suggests that the hippocoon-like phenotype may have evolved
independently more than once. This interpretation is supported 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 independently from the ancestral pattern. It is possible, then, that the
cenea allele (H c) evolved from the group that led to the
hippocoon (H h) and natalica (H na) 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 evolutionary step (Vane-Wright 1976; Clarke et al. 1985) produced
male-like females and was the origin of the female color
dimorphism.
Acknowledgments
This work was supported by a grant from the Human Frontiers
Science Program. I thank Alfried Vogler, Ali Cieslak, Toshio
Polymorphic mimicry in P. dardanus
591
Sekimura, Dick Vane-Wright, Susan Paulsen, and, in particular,
Steve Collins for many helpful and insightful discussions and Trey
Greer for developing DIGIT, the digitization software used in this
study.
REFERENCES
Bernardi, G., Pierre, J., and Nguyen, T. H. 1985. Le polymorphisme et le
mimétisme 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 of
Batesian mimicry. I. Single-locus models. J. Theoret. Biol. 55: 283–303.
Charlesworth, D., and Charlesworth, B. 1975b. Theoretical genetics of
Batesian mimicry. II. Evolution of supergenes. J. Theoret. Biol. 55: 305–
324.
Charlesworth, D., and Charlesworth, B. 1975c. Theoretical genetics of
Batesian 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, Papilio
dardanus, 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 dardanus
Brown. I. Race cenea from South Africa. Genetics 44: 1347–1358.
Clarke, C. A., and Sheppard, P. M. 1960a. The genetics of Papilio dardanus
Brown. II. Races dardanus, polytrophus, meseres, and tibullus. Genetics
45: 439–457.
Clarke, C. A., and Sheppard, P. M. 1960b. The genetics of Papilio dardanus
Brown. III. Race antinorii from Abyssinia and race meriones from
Madagascar. Genetics 45: 683–698.
Clarke, C. A., and Sheppard, P. M. 1960c. The evolution of mimicry in the
butterfly 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 dominance
under disruptive selection. Heredity 14: 23–87.
Clarke, C. A., and Sheppard, P. M. 1962. The genetics of Papilio dardanus
Brown. IV. Data on race ochracea, race flavicornis, and further
information on races polytrophus and dardanus. Genetics 47: 909–920.
Clarke, C. A., and Sheppard, P. M. 1963. Interactions between major genes
and polygenes in the determination of the mimetic patterns of Papilio
dardanus. Evolution 17: 404–413.
Clarke, C. A., and Sheppard, P. M. 1971. Further studies on the genetics of
the mimetic butterfly Papilio memnon. L. Philos. Trans. R. Soc. Lond. B
263: 35–70.
Clarke, C. A., Sheppard, P. M., and Thornton, I. W. B. 1968. The genetics
of 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 morphology. I. Developmental morphological integration and differential dominance. Mol. Dev. Evol. 296B: 58–79.
Fisher, R. A. 1930. The Genetical Theory of Natural Selection. Oxford
University 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 developmental
processes as sources of dominance. Genetics 159: 423–432.
592
EVOLUTION & DEVELOPMENT
Vol. 5, No. 6, November^December 2003
Haldane, J. B. S. 1924. A mathematical theory of natural and artificial
selection. Part I. Trans. Cambr. Phil. Soc. 23: 19–41.
Koch, P. B., and Nijhout, H. F. 2002. The role of wing veins in colour
pattern 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: developmental
genes that contribute to speciation. Evol. Dev. 5: 269–280.
Nijhout, H. F. 1985. Independent development of homologous pattern
elements 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 the
phenotypic 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 Wing
Patterns. Smithsonian Institution Press, Washington, DC.
Nijhout, H. F. 1994. Developmental perspectives on the evolution of
butterfly 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 the
buckeye butterfly (Precis coenia and Precis evarete): evidence against the
constancy of G. Evolution 50: 1585–1597.
Paulsen, S. M., and Nijhout, H. F. 1993. Phenotypic correlation structure
among 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 the
world. J. East Afr. Uganda Nat. Hist. Soc. 209: 4–22.
Sheppard, P. M. 1959. The evolution of mimicry; a problem in
ecology 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 an
adaptation. Evol. Biol. 10: 163–206.
Vane-Wright, R. I. 1976. An alternative hypothesis on the evolution of
Papilio 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 of
three African swallowtail butterflies, Papilio dardanus, P. phorcas and P.
constantinus: cladistic analysis (Lepidoptera: Papilionidae). Syst. Entomol. 16: 275–291.
Vane-Wright, R. I., and Smith, C. R. 1992. Occurrence and significance of
natural hybrids between Papilio dardanus and P. phorcas (Lepidoptera:
Papilionidae). Syst. Entomol. 17: 269–272.