Individual colour patches as
multicomponent signals
Gregory F. Grether*, Gita R. Kolluru and Karen Nersissian
Department of Organismic Biology, Ecology and Evolution, University of California, Los Angeles, 621 Charles E. Young Drive South,
Los Angeles, CA 90095-1606, USA
(Received 2 December 2002 ; revised 1 September 2003; accepted 3 September 2003 )
ABSTRACT
Colour patches are complex traits, the components of which may evolve independently through a variety of
mechanisms. Although usually treated as simple, two-dimensional characters and classified as either structural
or pigmentary, in reality colour patches are complicated, three-dimensional structures that often contain multiple
pigment types and structural features. The basic derma l chromatophore unit of fishes, reptiles and amphibians
consists of three contiguous cell layers. Xanthophores and erythrophores in the outermost layer contain
carotenoid and pteridine pigments that absorb short-wave light; iridophores in the middle layer contain crystalline
platelets that reflect light back through the xanthophores; and melanophores in the basal layer contain melanins
that absorb light across the spectrum. Changes in any one component of a chromatophore unit can drastically
alter the reflectance spectrum produced, and for any given adaptive outcome (e.g. an increase in visibility), there
may be multiple biochemical or cellular routes that evolution could take, allowing for divergent responses by
different populations or species to similar selection regimes. All of the mechanisms of signal evolution that
previously have been applied to single ornaments (including whole colour patches) could potentially be applied to
the individual components of colour patches. To reach a complete understanding of colour patch evolu tion,
however, it may be necessary to take an explicitly multi-trait approach. Here, we review multiple trait evolu tion
theory and the basic mechanisms of colour production in fishes, reptiles and amphibians, and use a combi-
nation of computer simulations and empirical examples to show how multiple trait evolution theory can be
applied to the components of single colour patches. This integrative perspective on animal colouration opens up
a host of new que stions and hypotheses. We offer specific, testable functional hypotheses for the most common
pigmentary (carotenoid, pteridine and melanin) and structural components of vertebrate colour patches.
Key words: multiple trait, colour patch, chromatophore, carotenoid, amplifier, mimicry, signal evolution,
handicap, indicator.
CONTENTS
I. Introduction ....... ......... ...... ......... ......... .......... ......... ...... ......... ......... .......... ...... ......... ......... .......... ......... ...... ... 584
II. Multiple trait evolution theory ............................ ...... ......... ......... .......... ...... ......... ......... .......... ...... ......... ... 585
III. Mechanisms of colour production and colour change ....................... ......... ...... ......... .......... ......... ...... ... 586
(1) Chromatophore layers .............................. ......... ...... ......... .......... ......... ...... ......... ......... .......... ...... ......... 586
(2) Dermal chromatophore unit ............. ...... ......... ......... ......... ....... ......... ......... ......... ...... .......... ......... ...... 587
(3) Proximate mechanisms of colour change ................................ ......... ...... ......... ......... .......... ...... ......... 587
IV. Multilayer colour patch model ...................... ...... ......... ......... .......... ...... ......... ......... ......... ....... ......... ......... 589
(1) Xanthophores (and erythrophores) (layer 1) .......................................... ...... ......... .......... ......... ...... ... 589
(2) Iridophores (and leucophores) (layer 2) ....................................... ...... ......... ......... ......... ....... ......... ...... 589
(3) Melanophores (layer 3) ....................... ......... ......... ...... ......... .......... ......... ......... ...... ......... .......... ......... ... 591
(4) Reflective shield of connective tissu e (layer 4) ............................... ...... ......... ......... .......... ...... ......... ... 591
(5) Colour patch reflectance .................... ......... ......... ......... ...... .......... ......... ......... ......... ...... .......... ......... ... 591
* Address for correspondence : E-mail: ggrether@ucla.edu
Biol. Rev. (2004), 79, pp. 583–610. f Cambridge Philosophical Society
583
DOI: 10.1017/S1464793103006390 Printed in the United Kingdom
(6) Colour patch radiance .......................... ......... ......... ...... ......... .......... ......... ...... ......... ......... .......... ...... .... 591
(7) Perceived colour contrasts ......................... ...... ......... ......... .......... ...... ......... ......... ......... ....... ......... ....... 591
(8) Implementation ................................................. ...... ......... ......... .......... ......... ...... ......... .......... ......... ...... . 593
V. Simulation results ................................................. ...... ......... ......... .......... ......... ...... ......... ......... .......... ...... .... 593
(1) Spectral mimicry and fine-tuni ng within the xanthophore layer ......................................... ......... . 593
(2) The amplifying effects of iridophores ......................................... ......... ...... ......... ......... .......... ...... ....... 594
(a) Iridophore reflectivity effects ........................ ......... ......... ....... ......... ......... ......... ...... .......... ......... .... 594
(b) Iridophore ‘blueness’ effects ................... ......... ......... ......... ....... ......... ......... ......... ...... .......... ......... . 596
(3) The amplifying and attenuatin g effects of melanophores ................... ......... ...... ......... .......... ......... . 596
(a) The amplifying effect of melanophores on iridophores .................... ......... ......... ....... ......... ....... 597
(b) The effects of melanophores on xanthophores ...................... ......... ...... ......... ......... .......... ...... .... 598
(4) Summary of simu lation results ........................... ......... ......... ....... ......... ......... ......... ...... .......... ......... .... 598
VI. Individual colour patches as multicomponent traits ....................... ......... ......... ...... ......... .......... ......... .... 599
(1) Xanthophores .................................................... ......... ......... .......... ...... ......... ......... ......... ....... ......... ....... 599
(a) Carotenoids versus pteridines .......... ......... ......... ...... ......... .......... ......... ...... ......... ......... .......... ......... . 599
(b) Multiple carotenoids ................................... ...... ......... ......... .......... ...... ......... ......... .......... ...... ......... . 601
(2) Iridophores ........ ......... ......... ......... ...... .......... ......... ......... ...... .......... ......... ......... ...... ......... .......... ......... .... 602
(a) Detectability ................................................. ......... ......... .......... ...... ......... ......... ......... ....... ......... ....... 602
(b) Indicator value ................................. ......... ......... ...... ......... .......... ......... ...... ......... ......... .......... ...... .... 602
(c) Xanthophore amplification ............... ......... ...... ......... ......... .......... ...... ......... ......... .......... ...... ......... . 602
(d) Xanthophore mimicry and complementarity ............................ ......... ......... ...... .......... ......... ....... 603
(3) Melanophores .............................................. ...... ......... ......... .......... ...... ......... ......... ......... ....... ......... ....... 603
VII. Conclusions ................... ......... ......... ...... .......... ......... ......... ...... ......... .......... ......... ...... ......... .......... ......... ....... 604
VIII. Acknowledgements ...................................... ...... ......... ......... ......... ....... ......... ......... ......... ...... .......... ......... .... 604
IX. References ........................ ......... ......... .......... ...... ......... ......... ......... ....... ......... ......... ......... ......... .......... ...... .... 604
X. Appendix A : the colour polyhedron ............ ......... ......... ......... ...... .......... ......... ......... ...... .......... ......... ....... 608
I. INTRODUCTION
Vertebrate colour patches are complex traits, the
components of which may evolve independently through a
variety of mechanisms (Losos, 1985 ; Thompson & Moore,
1991; Endler, 1992; Macedonia & Stamps, 1994; Berglund,
Bisazza & Pilastro, 1996; Macedonia, Brandt & Clark,
2002). Although usually treated as simple, two-dimensional
characters and classified as either structural or pigmentary
and, if the latter, by pigment type (e.g. melanin, carotenoid
and pteridine), in reality, colour patches are multilayered,
three-dimensional structures that often contain multiple
pigment types and structural features. The standard dermal
chromatophore unit (Bagnara, Taylor & Hadley, 1968;
Bagnara & Hadley, 1973) of poikilothermic vertebrates
consists of three contiguous cell layers (Fig. 1). Xantho-
phores and erythrophores in the outermost layer contain
carotenoid and pteridine pigments that primarily absorb
short-wave light ; iridophores in the middle layer contain
crystalline platelets that reflect light back through the
xanthophores; and melanophores in the basal layer contain
melanins that absorb light across the spectrum. Changes
in any one component of a chromatophore unit can dra-
matically alter the colour produced. For any given adaptive
outcome (e.g. an increase in crypsis), there may be multiple
biochemical or cellular routes that evolution could take,
allowing for divergent responses by different populations or
species to the same selection regimes. Moreover, different
colour patch components may evolve in fundamentally dif-
ferent ways.
Our main thesis herein is that the mechanisms of signal
evolution that previously have been applied to single orna-
ments (including whole colour patches) and multiple or-
naments (including colour patterns) may also be applied to
the individual components of colour patches. This novel
perspective on animal colouration opens up a range of new
questions. For clarity, we present mainly sexual signalling
(mate choice) examples, but the same basic arguments could
be applied, for example, to colour patches used in compe-
tition for territories or other resources.
The organization of this review is as follows. First, we
briefly review multiple trait evolution theory, clarify the
relationships between the various models, and explain how
our hypothesis differs from previous ideas about composite
colour signals (Section II). Next, we review the basic
mechanisms of colour production and colour change in
the skin of fishes, reptiles and amphibians, and provide
examples of colour variants that are understood at a
cellular or biochemical level (Section III). The main pur-
pose of the first two sections is to acquaint readers with
concepts and terminology that will be used subsequently.
In the core of the paper, we derive a general model of
a multilayer colour patch (Section IV) and use computer
simulations to explore the properties of different colour
patch components in a multi-trait context (Section V). Most
of these simulations examine how changes of a specified
nature in one chromatophore layer are likely to affect the
colour perceived and the discriminability of variation in
other chromatophore layers. In the final section, we discuss
the implications of our findings in the context of multiple
584
Gregory F. Grether, Gita R. Kolluru and Karen Nersissian
trait evolution theory and suggest avenues for further re-
search (Section VI).
II. MULTIPLE TRAIT EVOLUTION THEORY
Multicomponent signals may evolve through a variety of
mechanisms operating alone or in combination (Andersson,
1994; Andersson & Iwasa, 1996; Jennions & Petrie, 1997 ;
Brooks & Couldridge, 1999; Rowe & Guilford, 1999 ;
Andersson et al., 2002). These include runaway selection
(Fisher, 1930, 1958), indicator processes (direct and indirect
benefits; Darwin, 1871; Zahavi, 1975; Grafen, 1990),
sensory exploitation (Ryan & Keddy-Hector, 1992), status
signal selection (Pryke , Andersson & Lawes, 2001), chase-
away selection (Holland & Rice, 1998 ; Brooks & Coul-
dridge, 1999; but see Rosenthal & Servedio, 1999), and
species recognition mechanisms (Crapon de Caprona &
Ryan, 1990; Pfennig, 1998; Hankinson & Morris, 2002).
Some traits may also function to enhance, mimic, amplify or
attenuate other traits (Hasson, 1989, 1990). Of these pro-
cesses, indicator mechanisms and signal amplification are
the most relevant to our examination of individual colour
patches as multicomponent trait s, and we focus on these
below.
Amplifiers and attenuators facilitate or hamper receiver
assessment of signalers based on other (indicator) traits
(Hasson, 1989, 1990; Table 1). Amplifiers that remain
unlinked to signaler quality confer a fitness benefit to
high-quality signalers and a cost to low-quality signalers by
enhancing the perception of signaler quality (Hasson, 1991).
For a ‘pure ’ amplifier to persist, the overall benefits to
high-quality signalers must outweigh the costs to low-quality
signalers (Hasson, 1991). Alternatively, amplifiers may
themselves become linked to quality, in which case they
become condition-dependent traits known as amplifying
handicaps (Hasson, 1990, 1997). Handicaps are traits that
are costly to produce, maintain or possess and whose costs
are lower for higher-quality signalers (Hasson, 1997).
Amplifying handicaps are costly to produce and also
amplify the quality-indicating properties of other traits.
For example, white tips and glossy structural colours on
bird feathers may be amplifiers because they make feather
quality (damage, abrasion) more visible, and handicaps
because they increase the probability of feather wear
(Fitzpatrick, 1998a, b). Regardless of the type of amplifier
involved, the amplifier process results in multiple-trait
displays.
Møller and Pomiankowski (1993) formalized several
hypotheses to explain the evolution of multicomponent
sexual displays (see also Zuk et al., 1990; Zuk, Ligon &
Thornhill, 1992; Sullivan, 1994; Johnstone, 1996). They dis-
tinguished among (1) multiple messages, in which different
traits indicate different aspects of signaler quality; (2) redun-
dant signals, which together indicate overall signaler quality;
and (3) unreliable signals, which are arbitrary traits not corre-
lated with signaler quality. The outcomes of these and other
models of multiple trait evolution rely on the costs to re-
ceivers of assessing multiple traits (Iwasa & Pomiankowski,
Xanthophore
layer
Iridophore
layer
Melanophore
layer
Underlying
fascia
Blue-violet
Yellow-green
Red-orange
Fig. 1. The dermal chromatophore unit (Bagnara et al., 1968; Bagnara & Hadley, 1973) of a hypothetical green frog, showing
how the xanthophore, iridophore, melanophore and underlying fascia layers interact to determine the overall colour of the
animal. Wavy lines depict the paths of light of differing wavelengths through the cell layers. Adapted from Bagnara & Hadley (1973).
Colour patch evolution 585
1994; Sullivan, 1994; Johnstone, 1995 ; but see Borgia &
Presgraves, 1998). Some authors have suggested that
evaluating multiple ornaments may entail lower costs to
receivers (Schluter & Price, 1993 ; Sullivan, 1994; Rowe,
1999), for instance, because this enables receivers to evalu-
ate signaler attributes more quickly or accurately, but most
authors have assumed that the costs of assessing multiple
traits are highe r than the costs of assessing single traits. The
added costs of assessing multiple traits are termed the ‘joint
costs of choice’ (Pomiankowski & Iwasa, 1993; Iwasa &
Pomiankowski, 1994). Only if joint-choice costs are lo w can
multiple ornaments be maintained. Joint-choice costs have
been argued to be low for the multiple tail ornaments of
male barn swallows (Hirundo rustica) and the multiple colour
spots of male guppies (Poecilia reticulata), on the grounds that
these traits can be assessed simultaneously by females using
a single sensory modality (Brooks & Couldridge, 1999 ; Kose
& Møller , 1999).
Previous authors have treated vertebrate colour patterns as
composite traits (Fitzpatrick, 1998b ; Brooks & Couldridge,
1999; McGraw et al., 2002). For example, Fitzpatrick (1998b)
suggested that carotenoid, structural and melanin patches
in bird feathers convey multiple messages (general health,
developmental stabilit y and feather quality, respectively).
By contrast, we are proposing that individual colour patches
should be treated as composite traits. We offer this not
as an alternative hypothesis but instead as an added layer
of complexity that has largely been overlooked (but see
Wedekind et al., 1998; Macedonia et al., 2000; Badyaev et al.,
2001; Grether, Hudon & Endler, 2001 a). When viewing
colour patterns, receivers m ay, in effect, be evaluating mul-
tiple ornaments nes ted within multiple ornaments.
III. MECHANISMS OF COLOUR PRODUCTION
AND COLOUR CHANGE
(1) Chromatophore layers
The spectacular array of colours seen in the skin of fishes,
reptiles and amphibians is due to the presence of pigment
cells called chromatophores. These cells include the light-
absorbing melanophores, xanthophores and erythrophores,
and the light-reflecting leucophores and iridophores (Fujii,
1993). Classified by colour, xanthophores and erythro-
phores are yellow to red, leucophores are opaqu e white,
iridophores are iridescent or silvery, and melanophores are
brown to black.
Xanthophores and erythrophores contain carotenoids
and pteridines, which act as coloured filters, preferentially
absorbing short wavelengths (Bagnara & Hadley, 1973).
Carotenoids are hydrophobic molecules that typically
absorb visible light primarily in the 400–500 nm range,
producing yellow to red colours. Two major subclasses
of carotenoids include the pure hydrocarbon carotenes
and the oxygen-containing xanthophylls (Kayser, 1985).
Pteridines are hydrophilic compounds that primarily absorb
light between 340 and 500 nm and may appear yellow
(xanthopterin and sepiapterin), red (drosopterin or erythro-
pterin), or colo urless to the human eye (Kayser, 1985).
Table 1. Terminology associated with multicomponent trait evolution
Term Definition Source
Signaler quality Variable, hidden attributes of signalers that have the potential
to affect the fitness of receivers or their offspring.
Quality indicator A trait that honestly signals some aspect of signaler quality. Some
quality indicators are handicaps, traits that are costly to
produce or possess and whose costs are directly responsible
for signal honesty. Handicaps impose a greater relative cost
on low-quality than on high-quality signalers.
Zahavi (1975); Hasson (1990)
Arbitrary trait A sexually selected trait that is not condition-dependent.
For example, a trait that has evolved via Fisherian runaway
selection or via exploitation of pre-existing receiver sensory
biases.
Fisher (1930, 1958) ; Ryan et al. (1990);
Ryan (1990); Endler (1991 a)
Amplifier (attenuator) A trait not directly preferred by receivers, and not directly
linked to signaler quality, but which increases (decreases)
the ability of receivers to assess signaler quality based
on other traits. Amplifiers are expected to increase the
fitness of high-quality signalers, but decrease the fitness
of low-quality signalers, because of their quality-revealing
properties.
Hasson (1989, 1990, 1991)
Joint costs of mate choice
( joint-choice costs)
Costs associated with choosing based on multiple ornaments,
above and beyond those associated with choosing based on
one ornament. The costs can be equal to or less than the
sum of costs of choosing each trait separately (joint-choice
costs f0), or they can be ‘ super-additive ’, that is greater
than the sum of the costs of choosing based on each
trait singly.
Pomiankowski & Iwasa (1993) ;
Iwasa & Pomiankowski (1994)
586 Gregory F. Grether, Gita R. Kolluru and Karen Nersissian
Iridophores and leucophores produce structural colours
through some combination of specular reflection, construc-
tive interference and scattering (Herring, 1994). Iridophores
contain organized stacks of reflecting crystalline platelets of
guanine, hypoxanthine or uric acid (Bagnara, 1966 ; Fujii,
1993). Light is reflected at the planar surfaces of the platelets
and scattered at the edges, but the colour of the reflected
light is primarily determined by the thickness, spacing
and refractive index of the platelets (Fujii, 1993; Herring,
1994). A given wavelength (l) is reflected most strongly
when the optical thickness (thicknessrrefractive index) of
the platelets and intervening cytoplasmic spaces equals 1/4l
(Huxley, 1968 ; Denton & Land, 1971). Leucophores are
dendritic cells containing globular, membranous vesicles
filled with cytoplasm, purines or colourless pteridines
(Takeuchi, 1976; Fujii, 1993 ; Oliphant & Hudon, 1993).
Leucophores could potentially produce a wide range of
colours through Raylei gh or Mie scattering, depending
on the diameter of the vesicles and refractive indices of the
substances within them (Herring, 1994). However, most
cells identified as leucophores appear white, probably
because the vesicles they contain fall within the size range
in which wavelength-independent scattering occurs (>25l ;
Fujii, 1993).
Melanophores contain melanins – highly polymerized
compounds synthesized from tyrosine (Bagnara & Hadley,
1973) – which absorb light throughout the human visible
spectrum and into the ultraviolet range, and thus cause
darkening of the integument. (Pheomelanins, which are
responsible for some of the yellowish and reddish colours
seen in mammals, have not been reported in fishes, reptiles
or amphibians ; Fujii, 1993.)
(2) Dermal chrom atophore unit
Chromatophores are typically arranged in the dermis in
three or four contiguous cell layers, collectively referred to
as the ‘dermal chromatophore unit ’ (Fig. 1; Bagnara &
Hadley, 1973). Xanthophores (and erythrophores) make
up the outermost layer and are found just beneath the basal
lamina and associated collagen. The second layer is made
up of iridophores and the third layer consists of melano-
phores. A reflective shield of connective tissue occurs
beneath the melanophores in frogs and lizards (Nielsen &
Dyck, 1978 ; Macedonia et al., 2000), and reflective leuco-
phores have been found below the melanophores in some
fishes. Other species lack a reflective shield below the
melanophores (reviewed in Obika, 1988 ; Fujii, 1993).
The dermal chromatophore unit model (Bagnara et al.,
1968; Bagnara & Hadley, 1973) describes how the xantho-
phore, iridophore and melanophore layers interact to de-
termine the overall reflectance spectrum of a colour patch
(see Fig. 1). Carotenoids and pteridines in the xanthophore
layer (if present) differentially absorb shorter wavelengths
(violet to blue). A fraction of the light that passes through
the xanthophores is reflected back out into the environment
by platelets in the iridophores. The remaining light is either
absorbed by the melanophores or reflected back through
the dermal chromatophore unit by the reflective sh ield.
The internal structure of the iridophores affects the colour
as well as the brightness of the colour patch. If the reflecting
platelets are dispersed and oriented randomly within the
cytoplasm, all wavelengths of light will be reflected and
the colour patch will appear silver, yellow, orange or red,
depending on the amounts and types of pigments present in
the xanthophores. If the reflecting platelets are stacked to
produce constructive interference in the short-wavelength
region, the result will be a structural blue in the absence
of xanthophores, or various hues of yellow-green, green or
blue-green if xanthophores are present (see Fig. 1). The
same three functional elements – a filtering layer (caro-
tenoid), a reflecting layer (keratin) and an absorbing layer
(melanin) – are present in a different cellular form in bird
feathers (Bagnara & Hadley, 1973).
(3) Proximate mechanisms of colour change
Most types of colour change in fishes, reptiles and
amphibians, whether within the lifetime of an individual
or over evolutionary time, involve changes in one or more
components of the dermal chromatophore unit. Rapid
(physiological) colour change results from m ovement of
intracellular organelles, while slow (morphological) colour
change involves net gains or losses in pigment content
(Bagnara & Hadley, 1973 ; Fujii, 1993 ; Painter, 2000;
Oshima, 2001). Dispersion of melanosomes causes darken-
ing of the skin, while aggregation of these melanin-contain-
ing organelles cau ses lightening (Bagnara & Hadley, 1973).
Also, migration of melanosomes into the dendritic processes
of the melanophores, which extend over the xanthophores,
can bring about rapid darkening of a colour patch (Taylor
& Hadley, 1970 ; Bagnara & Hadley, 1973; Fujii, 2000).
Dispersion of the pigment-containing organelles in xantho-
phores (xanthosomes) increa ses absorption of short-wave
light by carotenoids and pteridines, thereby intensifying
the colour produced, while aggregation of xanthosomes
has the reverse effect. Changes in the shape or position of
xanthophores relative to iridophores may also contribute
to colour change in some species (e.g. Nielsen, 1978).
Dispersion of iridophore platelets and leucosomes in-
creases the total amount of light reflected, while aggregation
of these reflective organelles has the opposite effect. Also, as
noted above, when iridophore platelets are dispersed in the
cytoplasm, they tend to reflect white light, but when they
are aggregated into stacks, thin-layer interfe rence effects
can drastically alter the spectrum of light reflected (Fujii,
1993). For example, the dark sleeper goby (Odontobutis
obscura) changes from yellow to blue as reflecting platelets in
the iridophores move from dispersed to aggregated positions
(Fujii, Hayashi & Toyohara, 1991). The blue colour appears
to be produced by thin-film interference effects within the
platelet stacks, while the yellow colour appears to be caused
by pigm ents in the xanthophore layer and reflection of
white light by iridophore platelets. Some of the dramatic
colour changes seen in coral reef fishes are caused by co-
ordinated changes in the distance between successive plate-
lets within the aggregated stacks (Fujii, 1993). A similar
mechanism of rapid colour change has been reported for
the lizard Urosaurus ornatus (Morrison, Sherbrooke & Frost-
Mason, 1996).
Colour patch evolution 587
