Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators

Abstract

Alpine flowers face multiple challenges in terms of abiotic and biotic factors, some of which may result in selection for
certain colours at increasing altitude, in particular the changing pollinator species composition, which tends to move from
bee-dominated at lower elevations to fly-dominated in high-alpine regions. To evaluate whether growing at altitude—and the
associated change in the dominant pollinator groups present—has an effect on the colour of flowers, we analysed data collected
from the Dovrefjell National Park in Norway. Unlike previous studies, however, we considered the flower colours according
to ecologically relevant models of bee and fly colour vision and also their physical spectral properties independently of
any colour vision system, rather than merely looking at human colour categories. The shift from bee to fly pollination with
elevation might, according to the pollination syndrome hypothesis, lead to the prediction that flower colours should shift
from more bee-blue and UV-blue flowers (blue/violet to humans, i.e. colours traditionally associated with large bee pollinators)
at low elevations to more bee-blue-green and green (yellow and white to humans—colours often linked to fly pollination) flowers
at higher altitude. However, although there was a slight increase in bee-blue-green flowers and a decrease in bee-blue flowers
with increasing elevation, there were no statistically significant effects of altitude on flower colour as seen either by
bees or by flies. Although flower colour is known to be constrained by evolutionary history, in this sample we also did not
find evidence that phylogeny and elevation interact to determine flower colours in alpine areas.

Full text

ORIGINAL PAPER
Flower colours along an alpine altitude gradient, seen
through the eyes of fly and bee pollinators
Sarah E. J. Arnold Æ Vincent Savolainen Æ
Lars Chittka
Received: 23 June 2008 / Accepted: 30 January 2009 / Published online: 17 February 2009
Ó Springer Science+Business Media B.V. 2009
Abstract Alpine flowers face multiple challenges in
terms of abiotic and biotic factors, some of which may
result in selection for certain colours at increasing altitude,
in particular the changing pollinator species composition,
which tends to move from bee-dominated at lower eleva-
tions to fly-dominated in high-alpine regions. To evaluate
whether growing at altitude—and the associated change in
the dominant pollinator groups present—has an effect on
the colour of flowers, we analysed data collected from
the Dovrefjell National Park in Norway. Unlike previous
studies, however, we considered the flower colours
according to ecologically relevant models of bee and fly
colour vision and also their physical spectral properties
independently of any colour vision system, rather than
merely looking at human colour categories. The shift from
bee to fly pollination with elevation might, according to the
pollination syndrome hypothesis, lead to the prediction that
flower colours should shift from more bee-blue and UV-
blue flowers (blue/violet to humans, i.e. colours tradition-
ally associated with large bee pollinators) at low elevations
to more bee-blue-green and green (yellow and white to
humans—colours often linked to fly pollination) flowers at
higher altitude. However, although there was a slight
increase in bee-blue-green flowers and a decrease in bee-
blue flowers with increasing elevation, there were no sta-
tistically significant effects of altitude on flower colour as
seen either by bees or by flies. Although flower colour is
known to be constrained by evolutionary history, in this
sample we also did not find evidence that phylogeny and
elevation interact to determine flower colours in alpine
areas.
Keywords Flower colour  Pollinator diversity 
Insect vision  Alpine flowers  Pollination
Introduction
Plants growing in mountainous regions are faced with a
range of challenges. As well as having to contend, poten-
tially, with high winds, desiccation and extremes of cold,
they also face increased ultraviolet exposure and pollinator
limitation when the temperatures and winds grow too
extreme for pollinating insects to fly (Totland et al. 2000).
Many strategies employed for dealing with such habitats
have already been investigated in depth (Totland et al.
2000), but what still warrants further investigation is how
flowers at high altitude might exhibit specific adaptations
in terms of their pigmentation.
Why might some colours in high-alpine habitats be more
beneficial than others? Flower colour is under selection by
pollinators (Kevan and Baker 1983; Rodriguez-Girones
and Santamaria 2004; Tastard et al. 2008; Waser 1983;
Whibley et al. 2006). There are indeed several studies
associating shifts in flower colours with shifts in pollinator
type (Altshuler 2003; Bradshaw and Schemske 2003). In
alpine areas the numbers of pollinators present will
Handling editor: Neal Williams
S. E. J. Arnold (&)  L. Chittka
Research Centre for Psychology, School of Biological and
Chemical Sciences, Queen Mary University of London, Mile
End Road, London E1 4NS, UK
e-mail: s.e.j.arnold@qmul.ac.uk
V. Savolainen
Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, UK
V. Savolainen
Imperial College London, Silwood Park Campus,
Ascot, Berks SL5 7PY, UK
123
Arthropod-Plant Interactions (2009) 3:27–43
DOI 10.1007/s11829-009-9056-9

decrease overall with increasing altitude, and will change
in composition; some insect groups are less able to function
at very high elevations than others (Kearns 1992; Totland
1992). Therefore, different pollinator guilds dominate at
different elevations and therefore the selective forces on
flower traits might be expected to differ. The pollination
syndrome hypothesis, which has been used as the basis for
studies of pollination systems for many years (Faegri and
van der Pijl 1978), postulates a strong association between
different pollinator guilds and particular suites of floral
characteristics, in particular aspects of morphology and
colour (e.g. the zygomorphic, closed and blue/purple ‘‘bee
flowers’’; large, white ‘‘moth flowers’’ with long corolla
tubes). Based on this framework, a changing pollinator
composition at different elevations may be expected to lead
to different colours prevailing at different altitudes
depending on the dominant pollinator types and the colours
that appeal to them.
For example, the ability of flies to forage at higher
elevations than bees (Kearns 1992;La
´
zaro et al. 2008;
Totland 1993) might perhaps lead us to expect that the
flower colours traditionally thought to be associated with
fly pollination (appearing white and yellow to humans)
would be more abundant at high altitudes. Flowers
appearing white (and also pink) to humans are mostly
blue–green for bees and other trichromatic insects (Kevan
et al. 1996), whereas yellow flowers can be either green or
UV-green, depending on their UV reflectance, to such
insects (Chittka et al. 1994).
Bees appear to be especially dominant at low to medium
elevations (i.e. below the treeline, in sub-alpine habitats)
(La
´
zaro et al. 2008), their large body size allowing foraging
in the relative cold, but the high energetic requirement
demanded by maintaining their flight muscles at a high
enough temperature perhaps restricting their activities at
very high elevations (Arroyo et al. 1982). Bees of many
species have an innate preference for UV-blue and blue
flowers (Giurfa et al. 1995; Raine et al. 2006), leading
perhaps to an expectation based on the pollination syn-
drome hypothesis that where bees dominate as pollinators
(e.g. below the treeline in mountainous regions), bee-blue
and UV-blue flower species should be more numerous.
However, this preference is modifiable by learning, as are
the innate preferences in many other pollinating insects
such as hoverflies and butterflies (Lunau and Maier 1995).
There have been previous attempts to document the
effects of altitude on flower colours present (see Totland
et al. (2000) for a summary)—Weevers (1952) observed
that there were more blue flower species in upland areas
than in lowland areas (both in Switzerland above 1,100 m
and Java above 1,500 m), and Kevan (1972) and Savile
(1972) observed that flowers in alpine areas and arctic
regions (which are climatically similar to alpine areas)
tended to consist of a higher proportion of white and yel-
low species. McCall and Primack (1992) observed that
purple and yellow flowers were the most visited colours in
lowland woodland, whilst yellow and white were the two
most visited colours in alpine tundra, with blue–purple
flowers being much less frequently visited. Some of these
earlier studies, however, contain primarily observational
recordings that are not well supported by statistical power.
More importantly, perhaps with the exception of Kevan
(1972), some of these studies have considered flower col-
our principally from the human perspective, without fully
taking into account the more recent understanding of pol-
linator visual systems and how these differ from human
eyes (Chittka and Kevan 2005; Chittka and Menzel 1992
;
Menzel and Shmida 1993).
These differences are fundamental: all insects so far
extensively tested have UV receptors with a maximum
sensitivity between around 330 and 375 nm (i.e. in the UV
range where human eyes have no sensitivity)—this
includes bees and other hymenopterans, lepidopterans,
coleopterans, hemipterans, dipterans, etc. (Briscoe and
Chittka 2001). Bees, the most important pollinators in
Norway at all but the high-alpine elevations (La
´
zaro et al.
2008), also have blue and green receptors, but typically
lack red receptors (Peitsch et al. 1992) (see Fig. 1a). Other
insects, including many butterflies and flies, have rather
different colour vision systems, in some cases more com-
plex than those of bees or humans (Briscoe and Chittka
2001; Morante and Desplan 2008). Figure 1b shows the
spectral sensitivities of the four photoreceptors contribut-
ing to colour vision in the blowfly, Lucilia sp.
We investigated whether the flower community growing
at high altitude has a different pollinator-relevant colour
composition to that of lower altitude areas, by using a data
set collected along a transect in the Norwegian Dovrefjell
mountains from 700 m to 1,600 m elevation. This is of
especial interest in light of the recent study by La
´
zaro et al.
(2008), in which plant communities at different elevations
in southern Norway were surveyed for floral colour and
morphology, and this was combined with visitation data.
The study found evidence of association between traits
(including colour) and pollinator, showing that flowers in
alpine areas generally seem to be visited by pollinators that
could be predicted according to the pollination syndrome
hypothesis. Thus, it seems that the predominance of pol-
linator types (and subsequently the main foraging strategies
in evidence) varies with elevation and could potentially
have strong effects on which flower species are most
abundant.
In our study we consider flower colours as seen by their
pollinators, firstly using the well-studied model of bee
colour vision, secondly using a model of fly colour vision,
and also using the raw reflectance spectra of the flowers,
28 S. E. J. Arnold et al.
123

thereby considering their colours without bias towards any
vision system. As bees and flies are the most important
pollinators in most Norwegian alpine habitats (La
´
zaro et al.
2008), we considered how the colours might appear to
these pollinators (using the bee colour hexagon model
(Chittka 1992) and a model of the blowfly (Lucilia sp.)as
described in Troje (1993)). Additionally, we analysed the
raw spectral properties of the flowers. This encompasses
wavelengths invisible to humans (\400 nm), but does not
impose a particular visual system on to the results. To
investigate whether the colours of flowers were constrained
by their phylogeny, we tested whether phylogenetic dis-
tance correlates with differences in colours, and whether
there is an interaction between elevation and phylogeny
that affects flower colour. This is an important consider-
ation because of the following possibility: if closely related
flowers tend to have similar colours and also occur in
similar elevation ranges, this could result in detection of an
association between colour and elevation, but the cause
would be phylogenetic constraints rather than selection for
particular colours at particular altitudes. It is also possible
that there may be no statistical association between ele-
vation and colour, and equally no statistically detectable
association between phylogeny and flower colour, but
when the effects of phylogeny and elevation are combined
there could be a tendency for certain related groups of
flowers, when growing at particular elevations, to exhibit
particular colours more often.
Materials and methods
Study sites and data collection
The study site was located in the Dovrefjell–Sunndalsfjella
National Park (formerly Dovrefjell National Park) in
Norway, near to Oppdal. Data were collected in June 1992
in the altitude range 700–1,600 m a.s.l. (sub-alpine to high-
alpine), using Kongsvoll Biological Station (62° 18
0
N, 9°
36
0
E, 900 m above sea level) as a basis. The entire altitude
range from the valley floor to the Knutshø (also variously
spelt Knutshøa or Knudshø) peaks (south peak altitude:
1,680 m; 62° 18
0
N; 9° 41
0
0 E; north peak altitude:
1,684 m; 62° 18
0
N; 9° 40
0
E) east of the station was sur-
veyed for 7 days, mostly using footpaths as trunk routes
and thoroughly exploring the territory around. All flower-
ing species found in this survey were noted, along with the
elevation at which they were recorded. It is possible that
some comparatively rare species might have escaped
attention, but the vast majority of common species have
been included (c.f. West and West (1910)), and also
confirmed by local expert Simen Bretten (personal com-
munication). Spectrophotometer readings from 300 nm to
700 nm (i.e. including the ultraviolet range) were taken of
the flowers of all species present, using the methods
described in Chittka and Kevan (2005) and Dyer and
Chittka (2004). (All spectral reflectance curves are avail-
able from the Floral Reflectance Database http://www.
reflectance.co.uk (Arnold et al. 2008).) A total of 74 spe-
cies were sampled from this location and are listed in
Appendix 1.
Effect of elevation on bee colour composition
of the community
We divided the surveyed territory into three elevation
ranges: lower altitudes (700–1,000 m), intermediate alti-
tudes (1,000–1,300 m) and high altitudes (1,300–1,600 m),
and recorded which species were found in each, and which
spanned more than one range. At this location, the low
altitude group corresponds to the vegetation of mountain
meadows, stream beds, and some forests (mainly birch);
the intermediate group covers the first zone above the tree
line although scattered dwarf birch (Betula nana) and Salix
trees still occur here (West and West 1910); 1,300 m is the
lower boundary of permafrost in the Dovrefjell (Sollid
et al. 2003); the high altitude vegetation is dominated by
lichens and comprises flowering plants growing on rocky,
unstable soils (West and West 1910). The range sampled
Fig. 1 Spectral sensitivities of the (a) bee photoreceptors and (b)fly
photoreceptors that contribute to colour vision. Bee photoreceptors
shown are UV-, blue- and green-sensitive for Bombus terrestris
(Skorupski et al. 2007); fly receptors are for Lucilia sp. (Hardie and
Kirschfeld 1983)
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 29
123

still extends into regions that can at times be too cold for
many pollinator species to fly and thus the dominant types
of pollinators will change significantly within the range
sampled (Totland 1993; Totland et al. 2000), with an
increase in muscoid fly species, a decrease in bee and
beetle species and possibly an increase in butterfly species.
For bee pollinators, we categorised the flower species by
colour, according to their loci in the bee colour hexagon
(Chittka 1992; Gumbert et al. 1999). The colour hexagon is a
graphical representation of the discriminability of different
colours to a bee, based on the relative excitations of the three
types of bee photoreceptor (UV, blue and green) elicited by
the colours, and two unspecified colour opponency mecha-
nisms. Previous studies have indicated that the division of
the colour hexagon into six particular categories corre-
sponds well to the actual distributions of flower colours
present in nature (Chittka et al. 1994). Thus, we classified
flowers as either bee-blue, blue–green, green, UV-green, UV
or UV-blue (see Appendix 1), as shown in Fig. 2a.
Effect of elevation on fly colour composition
of the community
We also looked at patterns in flower colour as seen by flies.
Many dipteran species have five photoreceptor types, of
which four are used for colour vision (Morante and Des-
plan 2008; Troje 1993). These are typically referred to as
R7p (most sensitive in the UV), R7y (highest sensitivity to
violet light), R8p (peak sensitivity to blue) and R8y (peak
sensitivity to green). Two general types of ommatidia are
present in fly eyes, containing either the two p-type
(‘‘pale’’) receptors or the y-type (‘‘yellow’’) receptors—
named according to how they appear in transmitted light
(Troje 1993).
The model we used is that of Troje (1993), based on the
blowfly Lucilia sp., in which spectral stimuli across quite
wide ranges are not discriminated amongst, but are dis-
criminated from stimuli in other spectral ranges, with
category boundaries at 400 and 515 nm. The opponent
system takes the difference in relative excitations between
the two p-type receptors and the two y-type receptors and
the receptor of each pair stimulated most strongly deter-
mines the colour the fly perceives. This results in four
colour categories (p? y?,p- y?,p- y- and p? y-)
which could be regarded as fly-UV, -blue, -yellow and fly-
purple (purple referring in human vision to a colour where
the shortest and longest wavelength receptors are stimu-
lated most strongly in combination), with all stimuli within
one category being chromatically indistinguishable to the
fly. The fly colour loci are plotted in Fig. 2b, with the four
quadrants labelled.
Fig. 2 a Bee colour hexagon of the flower species measured. The six
segments correspond to the six bee colour categories used in this
analysis (b = blue, bg = blue–green, g = green, ug = UV-green,
u = UV and ub = UV-blue). Loci are calculated according to the
relative stimulation of the three receptor types (UV, blue, green)
elicited by the stimulus. ‘‘Low’’, ‘‘medium’’ and ‘‘high’’ refers to the
highest elevation category in which the species was recorded. b
Colours of flowers from the study site, according to how they would
be perceived by the blowfly. Flower species are categorised according
to the highest elevation range in which they were recorded. The
model is used is that from Troje (1993) for Lucilia sp.; colours in the
same quadrant of the graph are not discriminated by the fly, meaning
that all flowers appear (clockwise, from top-left) fly-blue, UV, fly-
purple or fly-green
30 S. E. J. Arnold et al.
123

We used Microsoft Excel with the Bootstrap add-in
(available from http://www3.wabash.edu/econometrics/)to
investigate whether there is an association between flower
colour and elevation. We compared all the species that
occurred in the same altitude group or combination of
groups (e.g. low and medium) pairwise, counting the total
number of times flower species occurring across the same
ranges also shared the same colour, using either bee colours
or fly colours in separate analyses. This yielded a measure,
N
\
, of the association between flower colour and altitude
ranges of the flowers, derived from summing the counts of
incidences in which species found across the same altitude
groups share the same colours. For bee colours, N
\
= 251
in our actual data, and for fly colours, N
\
= 343.
We reassigned the flower colours across the sample
10,000 times, whilst keeping the altitude range over which
each species is found constant, and recalculated N
\
with
each trial, tracking how it varied and producing frequency
distributions shown in Fig. 3. If particular colours are
strongly dominant at some altitudes, N
\
will be
disproportionately high, whilst if particular colours are
found at unusually low frequency in a particular altitude
range, N
\
will be low. Using the frequency distributions of
N
\
obtained from our randomisations, we were able to
ascertain whether the value corresponding to the original
data differed significantly from expected, and thus, whether
there was an association between elevation and colour. We
will subsequently refer to this randomisation test as ‘‘Ele-
vation versus Colour analysis’’.
Distributions of spectral characteristics of flowers
by elevation group
Since an alternative possibility to pollinator-mediated
selection is that flowers’ pigments are selected by abiotic
factors, we also analysed the spectra independently of the
consideration of any visual system. We simplified the
spectra to the values obtained at 50 nm intervals over the
range originally measured. This is justified because flowers
typically have smooth reflectance functions, with only two
or three strong changes in reflectance over a range from
300 nm to 700 nm (Chittka and Menzel 1992; Chittka et al.
1994). We performed a principal components analysis
(PCA) on these data using SPSS for Windows. To test
whether the coordinates fell into distinct clusters according
to altitude range, we performed a MANOVA on the PCA
scores, testing whether the groups of species from different
elevations yield significantly different scores. This pro-
vides information on whether the reflectance spectra of
flowers at the different elevations differ in terms of their
physical properties, regardless of the visual system that
perceives the flowers.
Effect of phylogeny on flower colour
It is possible that phylogeny is a stronger predictor or
constraint of flower colour than any selective action of
pollinators or abiotic factors within a habitat, as evidenced
by the findings in Chittka (1997) that some plant families
have flowers in only two or three bee colour groups, with
fewer flowers of other colours. It may also be the case that
plant clades are themselves distributed according to
elevation.
We constructed a phylogenetic tree of the species studied,
using published DNA sequence information of ribulose-1,5-
bisphosphate carboxylase (rbcL) gene (Appendix 2) and
ribosomal ITS1 (internal transcribed spacer 1) (Appendix 3)
to resolve multiple species within genera. The rbcL gene has
already been extensively used in phylogenetic studies as it is
well conserved throughout the angiosperms (Chase et al.
1993). We used the GenBank database (http://www.
ncbi.nlm.nih.gov/Genbank/) to search for rbcL sequences
for the species present in the habitat. When a complete or
Fig. 3 Frequency distributions obtained for the value of N
\
when the
flower species are categorised by (a) bee colour and (b) fly colour. N
\
is a measure of the number of times flower species that are present
across the same altitude range share the same colour. If, at any
elevation, one flower colour becomes disproportionately dominant or
rare relative to chance, species occurring at that elevation will be
either much more or much less likely to share the same colour, giving
rise to an unusually high or low N
\
value. The thin dotted lines
indicate the boundaries for the upper and lower 2.5% most extreme
N
\
values
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 31
123

near-complete rbcL sequence was not available for a par-
ticular species we recorded, we substituted a sequence from a
species of the same genus.
In some cases, no sequence was available for any spe-
cies in the genus; in these cases we used a close relative
from the same family. Thus, we used a Dimorphotheca
sinuata sequence in the place of Antennaria dioica and a
Platanthera ciliaris sequence in the place of the two
Dactylorhiza species. This is justified by the studies of Kim
and Jansen (1995) and Aceto et al. (1999), which place
Dimorphotheca and Antennaria, and Platanthera and
Dactylorhiza close together on phylogenetic trees. For
three species (Viscaria alpina, Tanacetum vulgare and
Hieracium sp.), we were unable to find any appropriate
substitute sequences (sequences for other species from the
same family were already included in the analysis, but we
were unable to find species that were more closely related
to the three above species than to the others in their fam-
ilies), and so these species remain unresolved in this study
and were excluded from the subsequent statistical analysis.
Appendix 2 lists the species originally recorded in the
habitat, and also the data relating to the sequences used to
resolve the relationships, including the species from which
the rbcL sequences were obtained, the accession details of
the samples and the relevant references.
We aligned the sequences using PAUP* (Swofford
2002), then constructed a tree using maximum parsimony.
We used a heuristic search with the Tree Bisection
Reconnection (TBR) swapping algorithm. We performed
1,000 replicates for stepwise addition, saving only the five
best trees from each replicate. The best trees produced
were used to create a strict consensus tree, with the two
monocot genera (Tofieldia and Dactylorhiza/Platanthera)
being used to root the tree following the Angiosperm
Phylogeny Group (2003). Two genera (Saxifraga and Si-
lene) contained three species that were all present at the
study site. In order to resolve the relationships between the
species within these genera, we used the ITS1 sequence
information from the species in these genera.
Using MacClade (Maddison and Maddison 1992), we
then manually substituted the original species names from
the habitat in place of the species providing the rbcL
sequence information. Because of the relatively small
number of taxa in our sample, a few genera (specifically
those genera in the Brassicaceae and Saxifragaceae) were
misplaced according to the current phylogeny published by
the Angiosperm Phylogeny Group. In these cases, we cor-
rected the tree in MacClade according to the most recent
information of the APG using http://www.mobot.org/
MOBOT/APGroup/, moving taxa to their correct branches
as given in this resource.
All major lineages contained at least two different bee
colours, though the Ericaceae in this sample consisted only
of bee-blue and bee-blue-green species. We used Mac-
Clade (Maddison and Maddison 1992) to test whether the
distribution of colours with respect to the known phylogeny
deviated significantly from random, and also whether
species on the tree showed a pattern in their maximum
elevations relative to their phylogeny. We tested for ran-
dom versus non-random distribution of traits by shuffling
the characters (colour or maximum elevation) 1,000 times
and testing whether the tree lengths obtained differed sig-
nificantly from the tree length obtained from the actual
data. If the characters are clustered in the actual tree, this
tree length will be significantly shorter than for the trees
subsequently created with random reassignments of char-
acters. This test will be referred to as ‘‘Elevation versus
Phylogeny analysis’’.
To investigate whether the two variables, phylogeny and
altitude, have a combined effect on the colours of flowers,
we constructed a distance matrix based on phylogenetic
distance between species on the tree. As the sequence data
did not perfectly correlate with established trees, we did
not use rbcL genetic distance in our measure as this would
produce anomalous distances; instead we measured dis-
tance in terms of the number of nodes in the tree between
pairs of species. We created two other similar distance
matrices in SPSS, based firstly upon elevation ranges of the
species, and secondly on colour distances derived from raw
spectral reflectance data for the flower species.
If evolutionary history constrains flower colour, one
would anticipate that the colour distance matrix would
correlate significantly with the phylogenetic distance
matrix. Furthermore, if there was a combined effect of
phylogeny and elevation on flower colours, an aggregate
distance matrix containing combined information on phy-
logenetic distance and dissimilarity of elevation range
would be expected to correlate with a matrix of colour
distances—i.e. the closer species are in evolutionary his-
tory and elevation range, the more similar their colours. We
used the ade4 package in the R statistical package (R
Development Core Team 2004) with 1,000 repeats to test
whether this was the case.
Results
The colour composition of the flower populations in the
different elevation groups is shown in Fig. 4; graphs show
the bee colours of the flowers, as used in the analyses, and
also the species as classified by human colours, for
reference.
32 S. E. J. Arnold et al.
123

Effect of elevation on bee colour composition
of the community
The commonest bee colour at all altitudes was blue–green
(52% of flowers overall), and the proportion of blue–green
flower species increased from low to high altitudes, from
45% in the low altitude group to 67% in the high altitude
group. Whilst not significant (see below), this trend is in
line with predictions based purely on the concept of pol-
lination syndromes—at high altitudes where flies are the
dominant pollinator type, flowers should, according to this
hypothesis, be more likely to be human white, i.e. bee-
blue-green, than at lower elevations. By contrast, the pro-
portion of bee-blue flowers (usually blue or purple to
humans) declined with increasing altitude; only 5% of the
flower species recorded above 1,300 m were blue to a bee’s
eyes compared with 21% of species occurring at the lowest
elevations below 1,000 m.
Figure 3a shows the frequency distribution of N
\
values
obtained from the randomisation when species are classified
by bee colour, and the actual value of N
\
obtained from the
dataset. The analysis showed no significant tendency for
species in the same altitude group to share the same colour
more often than chance (Elevation versus Colour analysis,
N
\
= 251, P = 0.144), and this holds when species found
at each elevation are considered individually (low eleva-
tion, N
\
= 221, P = 0.076; medium, N
\
= 33, P = 0.608;
high, N
\
= 23, P = 0.457). This indicates overall that no
flower colour is more dominant than expected at any par-
ticular elevation; were a colour category to predominate
more at any particular elevation or, equally, to decrease in
importance, it would skew the value of N
\
and cause an
unexpectedly high or low value to be obtained. The value of
N
\
obtained for the actual dataset is somewhat lower than
many of the random values obtained; although this is by no
means significant, it suggests that, perhaps surprisingly, the
aggregation of colours by elevation is somewhat less than
chance would predict, i.e. the dominance of common col-
ours in some altitude groups is (non-significantly) less than
expected by chance.
0%
20%
40%
60%
80%
100%
(a)
(b)
700-1000 1000-1300 1300-1600
Altitude (m asl)
Proportion of colours
blue
blue-green
green
UV-green
UV
UV-blue
0%
20%
40%
60%
80%
100%
700-1000 1000-1300 1300-1600
Altitude (m asl)
Proportions of colours
blue
green
pink/purple
red
white
yellow
Fig. 4 Relative proportions of
different flower colours present
at the survey site, with
increasing elevation. Flowers
are classified on their
appearance according to a) the
bee visual system and b) the
human visual system. (Number
of species recorded at each
elevation range: 700–1,000 m,
58; 1,000–1,300 m, 27;
1,300–1,600 m, 18.)
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 33
123

Effect of elevation on fly colour composition
of the community
The commonest fly colour categories were ‘‘fly-yellow’’
(50% of species) and ‘‘fly-blue’’ (38% of species); ‘‘fly-
UV’’ and ‘‘fly-purple’’ categories contained only three and
six species, respectively. Increasing elevation was associ-
ated with an increase in the proportion of fly-yellow flower
species (51–73%) and a decrease in fly-blue and fly-purple
flowers (from 36% to 20%, and 8 to 0% from the low
elevation group to the high elevation group, respectively).
However, these changes are not statistically significant: the
randomisation analysis revealed no trend for flowers
growing in the same altitude ranges to share the same fly
colour (Elevation versus Colour analysis, N
\
= 343,
P = 0.594). This indicates that no colour of flower is more
or less dominant than expected by chance at any elevation.
The frequency distribution of N
\
values obtained in the
randomisation is shown in Fig. 3b, including the value
corresponding to the actual data.
Distributions of spectral characteristics of flowers
by elevation group
The results of the principal components analysis are shown in
Fig. 5. The distribution of colours from all three altitude
groups appear to overlap heavily, and indeed the MANOVA
reveals no difference between the groups of points (Wilks’
lambda, F = 0.247, P = 0.911, hdf = 4.0, edf = 140),
indicating that the distributions of reflectance spectra present
in each elevation group are statistically indistinguishable, at
least with the sample sizes available to us.
Effect of phylogeny on flower colour
The phylogenetic tree of the flower species present at the
study site is shown in Fig. 6, with the colours included for
reference purposes. The tree length when maximum ele-
vation was mapped on to the tree is significantly shorter
than chance (Elevation versus Phylogeny analysis, n = 71,
actual tree length = 24, P = 0.029), indicating that
growing at high elevations is a trait that occurs non-ran-
domly with respect to phylogeny. By contrast, in this
analysis, the tree length for colour did not differ signifi-
cantly from random (Elevation versus Phylogeny analysis,
n = 71, actual tree length = 27, P = 0.250), giving no
evidence in this particular analysis of a pattern of colour
relative to phylogeny, perhaps because our sample included
mostly distantly related species, where the phylogenetic
signal is weak.
When we compared a matrix of phylogenetic distances
with a matrix of colour distances based on raw flower
spectra, there was also no significant correlation between
the matrices (Mantel test, r = 0.0169, n = 71, P = 0.174).
Although at certain scales there are inevitably constraints
of evolutionary history on flower colour, the effect in this
sample is not significant.
We also found no significant correlation between the
matrix of colour distances and the aggregated matrix of
phylogenetic distance and dissimilarity in altitude range
when the colour distances were derived from raw spectra
(Mantel test, r = 0.0215, n = 71, P = 0.123). This indi-
cates that phylogenetic distance across the data set does not
interact with altitude to affect flower colour.
Discussion
Previous authors have made observations about the colours
of flowers in alpine areas, stating that the colour compo-
sition of high altitude and arctic communities differs from
those at lower elevations (Kevan 1972; Savile 1972;
Weevers 1952). In this study of flower colours along a
single transect in the Norwegian alpine flora, we sought to
test whether any of these observations can be supported
statistically. Unlike the studies by Savile (1972) and
Weevers (1952), we considered the flower colours as they
would be seen by insect pollinators, as this better reflects
the selective pressures on those flowers. We also analysed
how spectral properties of flowers were distributed over a
range of altitudes, making no a priori assumptions about
the visual systems viewing any of the flower species.
PC1
PC2
Low
Medium
High
Fig. 5 Principal components analysis of spectral reflectance data.
Here, flowers are classified as low, medium or high elevation based
on the maximum elevation at which they were recorded (\1,000 m,
\1,300 m and \1,600 m, respectively). Variation accounted for by
principal component 1: 45.177%; principal component 2: 25.871%
34 S. E. J. Arnold et al.
123

We have focused on the visual systems of bees and flies
because these are the two dominant pollinator types in
Norwegian alpine habitats. However, other pollinators are,
of course, present in plant communities. The recent study
by La
´
zaro et al. (2008) noted that butterflies in the Nor-
wegian mountains are of relatively minor importance when
compared to the above groups (constituting just 2% of the
flower-visiting insects at the lowest altitude study site,
increasing to 7.9% at the highest altitudes). Butterflies’
innately preferred colours can vary vastly depending on
species and individual so it is difficult to generalise (Lunau
and Maier 1995; Neumayer and Spaethe 2007); diurnal
lepidopterans also have very variable numbers of photo-
receptor types (Briscoe 2000; Briscoe and Bernard 2005;
Sison-Mangus et al. 2006). For these reasons, we have
chosen not to include a quantitative analysis of flower
colours as seen by butterflies here.
The diversity of butterfly visual systems, species-to-
species differences in colour preferences and the compar-
atively small contribution they make to the total pollinator
visitation relative to that of bees and flies means that
attempting to model the flowers according to their visual
system was not feasible. Equally, there is insufficient
information about beetle colour vision to be able to make
predictions about how they view flowers, and this group
makes a very small contribution to pollination in Norwe-
gian alpine habitats (La
´
zaro et al. 2008).
The study by La
´
zaro et al. (2008) found an association
between pollinator type (e.g. bees, flies, butterflies, etc.)
and flower colour and other aspects of morphology in
Norwegian habitats at various altitudes. Given this, and the
fact that the pollinator community changes in composition
at different elevations, a possible prediction arising from
this is that as different pollinator types change in
Fig. 6 Phylogenetic tree of the
species recorded at the study
site. We used published rbcL
(the large subunit of the
ribulose-1,5-bisphosphate
carboxylase) gene sequences
from the GenBank database to
resolve relationships to the
genus level, and ITS1
(ribosomal internal transcribed
spacer 1) sequences to resolve
species within genera where
necessary
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 35
123

importance at different elevations, and each group is
associated with particular colours of flower, then the col-
ours of flower species present should also vary in
accordance with these preferences. Although visit fre-
quency taken alone does not perfectly assess an insect’s
contribution to pollination of a particular plant, visit fre-
quency is one measure of total interaction (Va
´
zquez et al.
2005) and therefore a colour that is associated with more
visits from a pollinator is probably also receiving more
benefit from that pollinator than a flower of another colour
visited less frequently. This could apply regardless of
whether the colour association is based on innate prefer-
ences (Lunau and Maier 1995) or the result of pollinators
learning which flowers are most suitable for them (Raine
et al. 2006), given flower morphology and rewards.
However, our analysis provides no evidence for altitude
variation in the distribution of flower colours, either as
perceived by bees or by flies. Indeed, even when consi-
dering the flower colours without any model of insect
perception, no differences between the altitude groups
emerged; the PCA indicates that the distributions of spec-
tral reflectance properties in the three altitude categories
are statistically indistinguishable. The lack of association
between elevation and colour is unlikely to be a result of
insufficient data: all species found along our transect dur-
ing a full week’s careful survey were recorded, and the
length of the transect spanned sufficient distance that there
were substantial changes in the habitat type, from wood-
land and stream beds to unstable, rocky high-alpine
substrates. Although the transect began at around 700 m,
and did not extend to such low elevations as in La
´
zaro et al.
(2008), the change in habitat types suggests a significant
change in pollinator composition, such that a change in
flower colour composition of the communities could be
anticipated.
It is known from other studies that evolutionary history
constrains flowers’ colours (Kalisz and Kramer 2008;
Menzel and Shmida 1993), since not all families have the
biochemical pathways to produce particular pigments.
Some plant lineages only contain particular floral pigments,
and therefore flowers in those groups can only assume a
limited range of colours. However, most plant families are
ultimately capable of producing a variety of colours
(Chittka 1997), and in our study there is no evidence of a
combined effect of phylogeny and elevation predicting
flower colour similarities according to relatedness and
shared elevation range.
There are a number of reasons why such a lack of
change in the proportions of flower colours with changing
elevation, even when the effect of phylogeny is factored
into the analysis, may be observed. The first is the phe-
nomenal learning ability of insect pollinators. Even though
certain pollinator types have specific innate preferences for
colours of flowers (Giurfa et al. 1995; Lunau et al. 1996;
Raine et al. 2006), many are able to learn to overcome
these preferences easily if a flower is sufficiently rewarding
(Menzel 1985). Therefore, simply because a flower’s col-
our matches the innate preference of a dominant pollinator,
this may not necessarily constitute a fundamental selective
advantage for the flowers, since the vast majority of pol-
linator visits will be by experienced individuals (Raine and
Chittka 2007). The pollination market hypothesis, in fact,
advocates that a range of distinct and discriminable colours
in a habitat would be most advantageous to plants (Fried-
man and Shmida 1995; Gumbert et al. 1999). Even if the
diversity of pollinators decreases with elevation, rather
than appealing to the innate preferences of those remaining
pollinator types, a plant species may benefit more from
displaying floral signals that are distinctive and recognis-
able (Chittka et al. 1999).
There can also be selection for certain pigments for
reasons other than pollinator preference, particularly
because of their protective ability on the plant. Examples
include protecting against desiccation, cold, drought, her-
bivory or UV damage (Ben-Tal and King 1997; Chalker-
Scott 1999; Chittka et al. 2001; Fineblum and Rausher
1997
; Mori et al. 2005; Warren and Mackenzie 2001), and
to other challenges such as herbivory (Johnson et al. 2008),
all factors which are likely to differ in importance at dif-
ferent elevations. Similarly, the increase in ultraviolet at
high elevations can be damaging to some plant cells, and it
has been found that floral pigments such as anthocyanins
(usually conferring a blue or red colouration) may also
confer protection against UV damage (Mori et al. 2005). It
is therefore conceivable that in some cases the pigments
favoured by physical factors conflict with those that poll-
inators may favour, resulting in an overall observation that
colour frequencies at different altitudes do not differ, in
spite of various selection pressures favouring particular
colours in particular circumstances. Based on current
knowledge of the protective effects of anthocyanin pig-
ments, one might expect that the flowers of plants
subjected more often to extreme environmental conditions
such as high altitude would bear such pigments in
increased quantities. However, at high altitudes, selection
according to the concept of fly pollination syndrome might
dictate that flowers would be principally pollinated by flies
and therefore ‘‘should’’ be white or yellow (La
´
zaro et al.
2008). This could result in a trade-off situation in which
flowers must compromise between the colours that appeal
to pollinators’ innate preferences, and those which serve
other protective functions. Analogous trade-offs in which
traits or behaviours are beneficial in some contexts and
disadvantageous in others are relatively abundant in nature,
for example when zooplankton face trade-offs between
risks of UV damage and risk of predation when migrating
36 S. E. J. Arnold et al.
123

between depths in lakes or developing protective pigments
(Boeing et al. 2004; Hansson 2000).
Overall, our study demonstrates that there is no simple
pattern to the colours of flowers in mountainous areas as
elevation increases, whether flowers are considered
according to either of two insect colour vision models or
based only on their reflectance spectra, and that the polli-
nator types present cannot account for the lack of
differences if considered purely within the context of the
pollination syndrome concept.
Acknowledgements The raw data for this study were col-
lected when LC worked under the auspices of the Institute of
Neurobiology, Free University of Berlin. Financial support came
from a Leibniz Award from the German Research Foundation
(DFG) to R. Menzel. We wish to thank Simen Bretten for help with
identification of the plants, and Neal Williams and three anonymous
reviewers for their helpful comments on the manuscript. SEJA was
supported by a Biotechnology and Biological Sciences Research
Council/Co-operative Award in Science and Engineering (CASE)
studentship in association with Kew Enterprises (BBS/S/L/2005/
12155A).
Appendix
See Tables 1, 2 and 3.
Table 1 List of plant species analysed, including colour information (as seen by bees, and also humans, after categorising the flower colour into
one of six human colours) and elevation of collection (in m a.s.l.)
Family Species name Bee colour Human colour Fly colour Elevation
Campanulaceae Campanula rotundifolia Blue Blue p- y? (‘‘blue’’) 700–800
Caryophyllaceae Silene dioica Blue Pink/purple p- y? (‘‘blue’’) 700–1,200
Caryophyllaceae Viscaria alpina Blue Pink/purple p- y? (‘‘blue’’) 1,050–1,180
Fabaceae Astragalus alpinus Blue Pink/purple p- y? (‘‘blue’’) 700–1,500
Fabaceae Oxytropis lapponica Blue Pink/purple p- y? (‘‘blue’’) 700–900
Fabaceae Trifolium pratense Blue Pink/purple p- y? (‘‘blue’’) 700–900
Fabaceae Vicia cracca Blue Pink/purple p- y? (‘‘blue’’) 700–800
Lentibulariaceae Pinguicula vulgaris Blue Pink/purple p- y? (‘‘blue’’) 700–1,300
Onagraceae Chamerion angustifolium Blue Pink/purple p- y? (‘‘blue’’) 700–900
Plantaginaceae Veronica alpina Blue Blue p- y? (‘‘blue’’) 1,100
Polemoniaceae Polemonium caerulum Blue Pink/purple p- y? (‘‘blue’’) 700–1,000
Primulaceae Primula stricta Blue Pink/purple p- y? (‘‘blue’’) 700–900
Primulaceae Primula scandinavica Blue Pink/purple p- y? (‘‘blue’’) 1,050–1,150
Ranunculaceae Aconitum lycoctonum subsp. septentrionale Blue Pink/purple p- y?
(‘‘blue’’) 700–900
Violaceae Viola canina Blue Blue p- y? (‘‘blue’’) 700–900
Apiaceae Anthriscus sylvestris Blue–green White p- y- (‘‘yellow’’) 700–920
Asteraceae Achillea nobilis Blue–green White p- y- (‘‘yellow’’) 700–900
Asteraceae Antennaria dioica Blue–green Pink/purple p- y- (‘‘yellow’’) 920–1,200
Asteraceae Erigeron borealis Blue–green White p- y- (‘‘yellow’’) 1,400–1,500
Boraginaceae Myosotis decumbens Blue–green Blue p- y? (‘‘blue’’) 800–1,000
Brassicaceae Draba incana Blue–green White p- y- (‘‘yellow’’) 700–1,500
Brassicaceae Draba oxycarpa Blue–green Yellow p- y- (‘‘yellow’’) 1,600
Caryophyllaceae Cerastium alpinum Blue–green White p- y? (‘‘blue’’) 700–1,100
Caryophyllaceae Silene acaulis Blue–green Pink/purple p- y? (‘‘blue’’) 1,000–1,600
Caryophyllaceae Silene vulgaris Blue–green White p- y- (‘‘yellow’’) 700–900
Caryophyllaceae Stellaria nemorum Blue–green White p- y- (‘‘yellow’’) 700–900
Crassulaceae Sedum annuum Blue–green Red p- y- (‘‘yellow’’) 800–900
Diapensiaceae Diapensia lapponica Blue–green White p- y- (‘‘yellow’’) 1,040–1,100
Dipsacaceae Knautia arvensis
Blue–green Pink/purple p- y? (‘‘blue’’) 700–900
Ericaceae Andromeda polifolia Blue–green White p- y- (‘‘yellow’’) 1,040–1,260
Ericaceae Harrimanella hypnoides Blue–green White p- y- (‘‘yellow’’) 1,040–1,160
Ericaceae Kalmia procumbens Blue–green Pink/purple p- y? (‘‘blue’’) 1,040–1,160
Ericaceae Phyllodoce caerulea Blue–green Pink/purple p- y? (‘‘blue’’) 700–1,500
Ericaceae Vaccinium vitis-idaea Blue–green Pink/purple p- y- (‘‘yellow’’) 700–900
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 37
123

Table 1 continued
Family Species name Bee colour Human colour Fly colour Elevation
Ericaceae Vaccinium myrtillus Blue–green Red p- y? (‘‘blue’’) 1,000–1,100
Fabaceae Trifolium repens Blue–green White p- y- (‘‘yellow’’) 700–900
Orchidaceae Dactylorhiza maculata Blue–green Pink/purple p- y? (‘‘blue’’) 700
Orchidaceae Dactylorhiza fuchsii Blue–green Pink/purple p- y? (‘‘blue’’) 700
Orobanchaceae Pedicularis lapponica Blue–green Yellow p- y- (‘‘yellow’’) 700–800
Orobanchaceae Pedicularis oederi Blue–green Yellow p- y- (‘‘yellow’’) 920–1,600
Papaveraceae Papaver radicatum Blue–green Yellow p- y- (‘‘yellow’’) 900
Polygonaceae Bistorta vivipara Blue–green White p- y- (‘‘yellow’’) 960–1,100
Polygonaceae Rumex acetosa Blue–green Red p- y- (‘‘yellow’’) 700–800
Polygonaceae Rumex acetosella Blue–green Red p- y? (‘‘blue’’) 700–900
Primulaceae Trientalis europaea Blue–green White p- y- (‘‘yellow’’) 900–1,200
Ranunculaceae Pulsatilla vernalis Blue–green Pink/purple p- y? (‘‘blue’’) 1,000–1,200
Ranunculaceae Ranunculus glacialis Blue–green Pink/purple p- y- (‘‘yellow’’) 1,600
Rosaceae Dryas octopetala Blue–green White p- y- (‘‘yellow’’) 900–1,500
Rosaceae Prunus padus Blue–green White p- y- (‘‘yellow’’) 700–800
Rubiaceae Galium boreale Blue–green White p- y- (‘‘yellow’’) 700–900
Saxifragaceae Saxifraga stellaris Blue–green Pink/purple p- y- (‘‘yellow’’) 1,400–1,500
Saxifragaceae Saxifraga oppositifolia Blue–green Pink/purple p- y? (‘‘blue’’) 1,400–1,500
Saxifragaceae Saxifraga cespitosa Blue–green White p- y- (‘‘yellow’’) 1,100–1,600
Tofieldiaceae Tofieldia pusilla Blue–green White p- y- (‘‘yellow’’) 1,400
Asteraceae Tanacetum vulgare Green Yellow p- y- (‘‘yellow’’) 800
Brassicaceae Erysinum strictum Green Yellow p- y- (‘‘yellow’’) 700–900
Crassulaceae Rhodiola rosea Green Green p- y- (‘‘yellow’’) 1,160–1,600
Fabaceae Astragalus frigidus Green Yellow p- y- (‘‘yellow’’) 700–1,000
Fabaceae Lathyrus pratensis Green Yellow p- y- (‘‘yellow’’) 700–800
Orobanchaceae Melampyrum sylvaticum Green Yellow p- y- (‘‘yellow’’) 700–960
Orobanchaceae Melampyrum pratense Green Yellow p- y- (‘‘yellow’’) 700–800
Ranunculaceae Trollius europaenus Green Yellow p- y- (‘‘yellow’’) 900
Rosaceae Alchemilla glabra Green Green p
- y- (‘‘yellow’’) 700–1,000
Rosaceae Potentilla crantzii Green Yellow p- y- (‘‘yellow’’) 900–1,600
Rosaceae Geum rivale UV Pink/purple p? y? (‘‘UV’’) 700–1,000
Geraniaceae Geranium sylvaticum UV-blue Pink/purple p? y? (‘‘UV’’) 700–1,000
Orobanchaceae Bartsia alpina UV-blue Pink/purple p? y? (‘‘UV’’) 700–1,500
Plantaginaceae Veronica fruticans UV-blue Blue p- y? (‘‘blue’’) 700–900
Asteraceae Hieracium spec. UV-green Yellow p? y- (‘‘purple’’) 900
Asteraceae Taraxacum officinale UV-green Yellow p? y- (‘‘purple’’) 700–1,000
Ranunculaceae Caltha palustris UV-green Yellow p? y- (‘‘purple’’) 900–1,000
Ranunculaceae Ranunculus acris UV-green Yellow p? y- (‘‘purple’’) 700–1,500
Rosaceae Potentilla erecta UV-green Yellow p? y- (‘‘purple’’) 700–900
Violaceae Viola biflora UV-green Yellow p? y- (‘‘purple’’) 800–1,500
Species names are as in Norsk Flora (Lid and Lid 2005)
38 S. E. J. Arnold et al.
123

Table 2 Details of the rbcL sequences used to build the phylogenetic tree, including accession and citation details for the species providing rbcL
sequences
Species measured Species sequence used Family Accession Citation
Achillea nobilis Achillea millefolium Asteraceae EU384938 Panero and Funk (2008)
Aconitum septentrionale Aconitum racemulosum Ranunculaceae AY954488 Wang et al. (2005)
Alchemilla glabra Alchemilla mollis Rosaceae AMU06792 Soltis et al. (1993)
Andromeda polifolia Andromeda polifolia Ericaceae AF124572 Kron et al. (1999)
Antennaria dioica Dimorphotheca sinuata Asteraceae EU384966 Panero and Funk (2008)
Anthriscus sylvestris Anthriscus aemula Apiaceae D44554 Kondo et al. (1996)
Astragalus alpinus/frigidus Astragalus membranaceus Fabaceae EF685978 Guo et al. (2007)
Bartsia alpina Bartsia alpina Orobanchaceae AF190903 Olmstead et al. (1992)
Bistorta vivipara
(syn. Polygonum viviparum)
Polygonum cuspidatum Polygonaceae AB019031 Inamura et al. (1998)
Caltha palustris Caltha palustris Ranunculaceae AY395532 Silvertown et al. (2006)
Campanula rotundifolia Campanula trachelium Campanulaceae DQ356118 Antonelli (2008)
Harimanella hypnoides
(syn. Cassiope hypnoides)
Cassiope mertensiana Ericaceae L12603 Kron and Chase (1993)
Cerastium alpinum Cerastium glomeratum Caryophyllaceae M83542 Manhart et al. (1991)
Chamerion angustifolium
(syn. Epilobium angustifolium)
Epilobium rigidum Onagraceae AF495763 Levin et al. (2003)
Dactylorhiza maculata/fuschii Platanthera ciliaris Orchidaceae AF074215 Cameron et al. (1999)
Diapensia lapponica Diapensia lapponica Diapensiaceae L12612 Kron and Chase (1993)
Draba incarna/oxycarpa Draba nemorosa Brassicaceae NC_009272 Hosouchi et al. (2007)
Dryas octopetala Dryas drummondii Rosaceae U59818 Swensen (1996)
Erigeron borealis Erigeron tenuis Asteraceae EU384973 Panero and Funk (2008)
Erysimum hieracifolium Erysimum capitatum
Brassicaceae AY167980 Cummings et al. (2003)
Galium boreale Galium mollugo Rubiaceae AY395538 Silvertown et al. (2006)
Geranium sylvaticum Geranium albanum Geraniaceae DQ452884 Fiz et al. (2008)
Geum rivale Geum macrophyllum Rosaceae U06806 Soltis et al. (1993)
Hieracium spec. N/A
Knautia arvensis Knautia intermedia Dipsacaceae Y10698 Backlund and Bremer (1997)
Lathyrus pratensis Lathyrus pratensis Fabaceae AY395544 Silvertown et al. (2006)
Kalmia procumbens Kalmia procumbens Ericaceae U49288 Kron and King (1996)
Melampyrum sylvaticum/pratensis Melampyrum sylvaticum Orobanchaceae AM503854 Li et al. (2008)
Myosotis decumbens Myosotis discolor Boraginaceae AY395552 Silvertown et al. (2006)
Oxytropis lapponica Oxytropis anertii Fabaceae EF685981 Guo et al. (2007)
Papaver radicatum Papaver sp. Goldblatt 12541 Papaveraceae AM235045 Forest et al. (2007)
Pedicularis oederi/lapponica Pedicularis coronata Orobanchaceae AF206803 Soltis et al. (1999)
Phyllodoce caerulea Phyllodoce caerulea Ericaceae AF419829 Kron (2001)
Pinguicula vulgaris Pinguicula caerulea Lentibulariaceae L01942 Albert et al. (1992)
Polemonium caerulum Polemonium reptans Polemoniaceae L11687 Olmstead et al. (1992)
Potentilla erecta/crantzii Potentilla fruticosa Rosaceae U06818 Soltis et al. (1993)
Primula stricta/scandinavica Primula stricta Primulaceae AF394975 Trift et al. (2002)
Prunus padus Prunus padus Rosaceae AF411485 Jung et al. (2001)
Pulsatilla vernalis Pulsatilla cernua Ranunculaceae AY954492 Wang et al. (2005)
Ranunculus acris/glacialis Ranunculus acris Ranunculaceae AY395557 Silvertown et al. (2006)
Rumex acetosa and acetosella Rumex acetosella Polygonaceae D86290 Yasui and Ohnishi (1996)
Saxifraga stellaris/
oppositifolia/
cespitosa
Saxifraga stellaris Saxifragaceae AF374732 Soltis et al. (2001)
Sedum anuum and Rhodiola rosea
(syn. Sedum rosea)
Sedum rubrotinctum Crassulaceae L01956 Albert et al. (1992)
Silene acaulis/dioica/vulgaris Silene dioica Caryophyllaceae EF646928 Muir and Filatov (2007)
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 39
123

References
Aceto S, Caputo P, Cozzolino S, Gaudio L, Moretti A (1999)
Phylogeny and evolution of Orchis and allied genera based on
ITS DNA variation: morphological gaps and molecular conti-
nuity. Mol Phylogenet Evol 13:67–76. doi:10.1006/mpev.
1999.0628
Albert VA, Williams SE, Chase MW (1992) Carnivorous plants:
phylogeny and structural evolution. Science 257:1491–1495.
doi:10.1126/science.1523408
Altshuler DL (2003) Flower color, hummingbird pollination, and
habitat irradiance in four Neotropical forests. Biotropica 35:344–
355
Anderberg AA, Stahl B, Ka
¨
llersjo
¨
M (1998) Phylogenetic interrela-
tionships in the Primulales inferred from cpDNA rbcL sequence
data. Plant Syst Evol 211:93–102. doi:10.1007/BF00984914
Antonelli A (2008) Higher level phylogeny and evolutionary trends in
Campanulaceae subfam. Lobelioideae: molecular signal over-
shadows morphology. Mol Phylogenet Evol 46:1–18. doi:
10.1016/j.ympev.2007.06.015
Arnold SEJ, Savolainen V, Chittka L (2008) FReD: the floral
reflectance spectra database. Nat Preced: doi: 10.1038/npre.
2008.1846.1031
Arroyo MTK, Primack R, Armesto J (1982) Community studies in
pollination ecology in the high temperate Andes of central Chile.
I. Pollination mechanisms and altitudinal variation. Am J Bot
69:82–97. doi:10.2307/2442833
Backlund A, Bremer B (1997) Phylogeny of the Asteridae s.str. based
on rbcL sequences with particular reference to the Dipsacales.
Plant Syst Evol 207:225–254. doi:10.1007/BF00984390
Ben-Tal Y, King RW (1997) Environmental factors involved in
colouration of flowers of Kangaroo Paw. Sci Hortic (Amster-
dam) 72:35–48. doi:10.1016/S0304-4238(97)00071-X
Boeing WJ, Leech DM, Williamson CE, Cooke S, Torres L (2004)
Damaging UV radiation and invertebrate predation: conflicting
selective pressures for zooplankton vertical distribution in the
water column of low DOC lakes. Oecologia 138:603–612. doi:
10.1007/s00442-003-1468-0
Bradshaw HD Jr, Schemske DW (2003) Allele substitution at a flower
colour locus produces a pollinator shift in monkey flowers.
Nature 426:176–178. doi:10.1038/nature02106
Bremer K (2000) Early Cretaceous lineages of monocot flowering
plants. Proc Natl Acad Sci USA 97:4707–4711. doi:10.1073/
pnas.080421597
Briscoe AD (2000) Six opsins from the butterfly Papilio glaucus:
molecular phylogenetic evidence for paralogous origins of red-
sensitive visual pigments in insects. J Mol Evol 51:110–121
Briscoe AD, Bernard GD (2005) Eyeshine and spectral tuning of long
wavelength-sensitive rhodopsins: no evidence for red-sensitive
photoreceptors among five Nymphaline butterfly species. J Exp
Biol 208:687–696. doi:10.1242/jeb.01453
Briscoe AD, Chittka L (2001) The evolution of colour vision in
insects. Ann Rev Ent 46:471–510. doi:10.1146/annurev.ento.
46.1.471
Cameron KM et al (1999) A phylogenetic analysis of the Orchida-
ceae: evidence from rbcL nucleotide sequences. Am J Bot
86:208–224. doi:10.2307/2656938
Chalker-Scott L (1999) Environmental significance of anthocyanins in
plant stress responses. Photochem Photobiol 70:1–9. doi:10.1111/
j.1751-1097.1999.tb01944.x
Chase MW et al (1993) Phylogenetics of seed plants: An analysis of
nucleotide sequences from the plastid gene rbcL. Ann Mo Bot
Gard 80:528–548?550–580
Chittka L (1992) The color hexagon: a chromaticity diagram based on
photoreceptor excitations as a generalized representation of
colour opponency. J Comp Physiol A Neuroethol Sens Neural
Behav Physiol 170:533–543
Chittka L (1997) Bee color vision is optimal for coding flower color,
but flower colors are not optimal for being coded—why? Isr J
Plant Sci 45:115–127
Table 3 Details of the ITS sequences used to discriminate species
within genera
Species Accession Citation
Saxifraga cespitosa AF087604 Conti et al. (1999)
Saxifraga oppositifolia AF087592 Conti et al. (1999)
Saxifraga stellaris AF374827 Soltis et al. (2001)
Silene acaulis U30949 Desfeux et al. (1996)
Silene dioica U32568 Desfeux et al. (1996)
Silene vulgaris U30969 Desfeux et al. (1996)
Table 2 continued
Species measured Species sequence used Family Accession Citation
Stellaria nemorum Stellaria media Caryophyllaceae AF206823 Soltis et al. (1999)
Tanacetum vulgare N/A
Taraxacum officinale Taraxacum officinale Asteraceae AY395562 Silvertown et al. (2006)
Tofieldia pusilla Tofieldia pusilla Tofieldiaceae AJ286562 Bremer (2000)
Trientalis europaea Trientalis europaea Primulaceae U96655 Anderberg et al. (1998)
Trifolium pratense/repens Trifolium pratense Fabaceae AY395564 Silvertown et al. (2006)
Trollius europaenus Trollius laxus Ranunculaceae AY954486 Wang et al. (2005)
Vaccinium vitis-idaea/myrtillus Vaccinium vitis-idaea Ericaceae AF419837 Kron (2001)
Veronica alpinus/fruticans Veronica anagallis-aquatica Plantaginaceae AY034021 Wagstaff et al. (2002)
Vicia cracca Vicia cracca Fabaceae AY395566 Silvertown et al. (2006)
Viola biflora/canina Viola philippica Violaceae AB354436 Tokuoka (2008)
Viscaria alpina N/A
‘‘N/A’’ indicates that a suitable sequence from a species in the same genus or a closely related genus was not available
40 S. E. J. Arnold et al.
123

Chittka L, Kevan PG (2005) Flower colour as advertisement. In:
Dafni A, Kevan PG, Husband BC (eds) Practical pollination
biology. Enviroquest Ltd., Cambridge, ON, Canada, pp 157–196
Chittka L, Menzel R (1992) The evolutionary adaptation of flower
colors and the insect pollinators’ color vision systems. J Comp
Physiol A Neuroethol Sens Neural Behav Physiol 171:171–181
Chittka L, Shmida A, Troje N, Menzel R (1994) Ultraviolet as a
component of flower reflections, and the colour perception of
Hymenoptera. Vision Res 34:1489–1508. doi:10.1016/0042-6989
(94)90151-1
Chittka L, Thomson JD, Waser NM (1999) Flower constancy, insect
psychology, and plant evolution. Naturwiss 86:361–377. doi:
10.1007/s001140050636
Chittka L, Spaethe J, Schmidt A, Hickelsberger A (2001) Adaptation,
constraint, and chance in the evolution of flower color and
pollinator color vision. In: Chittka L, Thomson JD (eds)
Cognitive ecology of pollination. Cambridge University Press,
Cambridge, pp 106–126
Conti E, Soltis DE, Hardig TM, Schneider J (1999) Phylogenetic
relationships of the silver saxifrages (Saxifraga, sect. Ligulatae
Haworth): implications for the evolution of substrate specificity,
life histories, and biogeography. Mol Phylogenet Evol 13:536–
555. doi:10.1006/mpev.1999.0673
Cummings MP, Nugent JM, Olmstead RG, Palmer JD (2003)
Phylogenetic analysis reveals five independent transfers of the
chloroplast gene rbcL to the mitochondrial genome in angio-
sperms. Curr Genet 43:131–138
Desfeux C, Maurice S, Henry JP, Lejeune B, Gouyon PH (1996)
Evolution of reproductive systems in the genus Silene. Proc R
Soc Lond B Biol Sci 263:409–414. doi:10.1098/rspb.1996.0062
Dyer AG, Chittka L (2004) Biological significance of discriminating
between similar colours in spectrally variable illumination:
bumblebees as a study case. J Comp Physiol A Neuroethol Sens
Neural Behav Physiol 190:105–114. doi:10.1007/s00359-003-
0475-2
Faegri K, van der Pijl L (1978) The principles of pollination ecology.
Pergamon Press, Oxford
Fineblum WL, Rausher MD (1997) Do floral pigmentation genes also
influence resistance to enemies? The W locus in Ipomoea
purpurea. Ecology 78:1646–1654
Fiz O, Vargas P, Alarcon M, Aedo C, Garcia JL, Aldasoro JJ (2008)
Phylogeny and historical biogeography of Geraniaceae in
relation to multiple major increases and decreases in mitochon-
drial climate changes and pollination ecology. Syst Bot 33:326–
342. doi:10.1600/036364408784571482
Forest F et al (2007) Preserving the evolutionary potential of floras in
biodiversity hotspots. Nature 445:757–760. doi:10.1038/nature
05587
Friedman JW, Shmida A (1995) Pollination, gathering nectar and the
distribution of flower species. J Theor Biol 175:127–138. doi:
10.1006/jtbi.1995.0125
Giurfa M, Nu
´
n
˜
ez J, Chittka L, Menzel R (1995) Colour preferences of
flower-naive honeybees. J Comp Physiol A Neuroethol Sens
Neural Behav Physiol 177:247–259
Gumbert A, Kunze J, Chittka L (1999) Floral colour diversity in plant
communities, bee colour space and a null model. Proc R Soc
Lond B Biol Sci 266:1711–1716. doi:10.1098/rspb.1999.0836
Guo H, et al. (2007) Identification of Radix Astragali by DNA sequence
of its ITS, rbcL, matk, cox1,andNAD1-intron2. Direct submission:
School of Traditional Chinese Materia Medica, Shenyang Pharma-
ceutical University, Shenyang, Liaoning 110016, China
Hansson LA (2000) Induced pigmentation in zooplankton: a trade-off
between threats from predation and ultraviolet radiation. Proc R
Soc Lond B Biol Sci 267:2327–2331. doi:10.1098/rspb.2000.
1287
Hardie RC, Kirschfeld K (1983) Ultraviolet sensitivity of fly
photoreceptors R7 and R8: evidence for a sensitising function.
Biophys Struct Mech 9:171–180. doi:10.1007/BF00537814
Hosouchi T, Tsuruoka H, Kotani H (2007) Sequencing analysis of
Draba nemorosa chloroplast DNA. Direct submission: NCBI
Genome Project, National Centre for Biotechnology Informa-
tion, NIH, Bethesda, MD 20894, USA
Inamura A, Ohashi Y, Sato E, Yoda Y, Masuzawa T, Yoshinaga K
(1998) Intraspecific sequence variation of chloroplast DNA and a
molecular phytogeographic study of Polygonum cuspidatum.
Direct submission: Shizuoka University, Faculty of Science, Oya
836, Shizuoka 422–8529, Japan
Johnson ET, Berhow MA, Dowd PF (2008) Colored and white sectors
from star-patterned Petunia flowers display differential resis-
tance to corn earworm and cabbage looper. J Chem Ecol 34:757–
765. doi:10.1007/s10886-008-9444-0
Jung YH, Han SH, Oh YS, Oh MY (2001). Direct submission:
Department of Biology, College of Natural Sciences, Cheju
National University, 1 Ara-Dong, Jeju 690–756, Korea
Kalisz S, Kramer EM (2008) Variation and constraint in plant
evolution and development. Heredity 100:171–177. doi:10.1038/
sj.hdy.6800939
Kearns CA (1992) Anthophilous fly distribution across an elevation
gradient. Am Midl Nat 127:172–182. doi:10.2307/2426332
Kevan PG (1972) Floral colors in the high arctic with reference to
insect-flower relations and pollination. Can J Bot 28:2289–2316.
doi:10.1139/b72-298
Kevan PG, Baker HG (1983) Insects as flower visitors and pollinators.
Ann Rev Ent 28:407–453. doi:10.1146/annurev.en.28.010183.
002203
Kevan PG, Giurfa M, Chittka L (1996) Why are there so many and so
few white flowers? Trends Plant Sci 1:280–284. doi:10.1016/
1360-1385(96)20008-1
Kim KJ, Jansen RK (1995) ndhF sequence evolution and the major
clades in the sunflower family. Proc Natl Acad Sci USA
92:10379–10383. doi:10.1073/pnas.92.22.10379
Kondo K, Terabayashi S, Okada M, Yuan C, He S (1996)
Phylogenetic relationship of medicinally important Cnidium
offcinale and Japanese Apiaceae besed on rbcL sequences.
J Plant Res 109:21–27. doi:10.1007/BF02344283
Kron KA (2001). Direct submission: Wake Forest University,
Winston-Salem, NC 27109–7325, USA
Kron KA, Chase MW (1993) Systematics of the Ericaceae, Empetr-
aceae, Epacridaceae and related taxa based upon rbcL sequence
data. Ann Mo Bot Gard 80:735–741. doi:10.2307/2399857
Kron KA, King JM (1996) Cladistic relationships of Kalmia,
Leiophyllum and Loiseleuria (Phyllodoceae, Ericaceae) based
on rbcL and nrITS data. Syst Bot 21:17–29. doi:10.2307/
2419560
Kron KA, Judd WS, Crayn DM (1999) Phylogenetic analyses of
Andromedeae (Ericaceae subfam. Vaccinioideae). Am J Bot
86:1290–1300. doi:10.2307/2656777
La
´
zaro A, Hegland SJ, Totland Ø (2008) The relationships between
floral traits and specificity of pollination systems in three
Scandinavian plant communities. Oecologia 157:249–257. doi:
10.1007/s00442-008-1066-2
Levin RA et al (2003) Family-level relationships of Onagraceae based
on chloroplast rbcL and ndhF data. Am J Bot 90:107–115. doi:
10.3732/ajb.90.1.107
Li M et al (2008) Development of COS genes as universally
amplifiable markers for phylogenetic reconstructions of closely
related plant species. Cladistics 24:727–745. doi:10.1111/j.1096-
0031.2008.00207.x
Lid J, Lid DT (2005) Norsk Flora, 7th edn. Det Norske Samlaget,
Oslo
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 41
123

Lunau K, Maier EJ (1995) Innate colour preferences of flower
visitors. J Comp Physiol A Neuroethol Sens Neural Behav
Physiol 177:1–19
Lunau K, Wacht S, Chittka L (1996) Colour choices of naive bumble
bees and their implications for colour perception. J Comp
Physiol A Neuroethol Sens Neural Behav Physiol 178:477–489
Maddison WP, Maddison DR (1992) MacClade: analysis of phylogeny
and character evolution. Sinauer Associates, Inc., Sunderland,
MA
Manhart JR, Hugh JH, Wilson D (1991) Phylogeny of the
Caryophyllales. Direct submission, GenBank
McCall C, Primack R (1992) Influence of flower characteristics,
weather, time of day, and season on insect visitation rates in
three plant communities. Am J Bot 79:434–442. doi:10.2307/
2445156
Menzel R (1985) Learning in honey bees in an ecological and
behavioral context. In: Ho
¨
lldobler B, Lindauer M (eds) Exper-
imental behavioral ecology, vol 31. Gustav Fischer Verlag,
Stuttgart, pp 55–74
Menzel R, Shmida A (1993) The ecology of flower colours and the
natural colour vision of insect pollinators: the Israeli flora as a
study case. Biol Rev Camb Philos Soc 68:81–120. doi:10.1111/
j.1469-185X.1993.tb00732.x
Morante J, Desplan C (2008) The color-vision circuit in the medulla
of Drosophila. Curr Biol 18:553–565. doi:10.1016/j.cub.2008.
02.075
Mori M, Yoshida Y, Matsunaga T, Nikaido O, Kameda K, Kondo T
(2005) UV-B protective effect of a polyacylated anthocyanin,
HBA, in flower petals of the blue morning glory, Ipomoea
tricolor cv. Heavenly Blue. Bioorg Med Chem 13:2015–2020.
doi:10.1016/j.bmc.2005.01.011
Muir G, Filatov D (2007) A selective sweep in the chloroplast DNA
of dioecious Silene (section Elisanthe). Genetics 177:1239–1247.
doi:10.1534/genetics.107.071969
Neumayer J, Spaethe J (2007) Flower color, nectar standing crop, and
flower visitation of butterflies in an alpine habitat in central
Europe. Entomol Gen 29:269–284
Olmstead RG, Michaels HJ, Scott KM, Palmer JD (1992) Monophyly
of the Asteridae and identification of their major lineages
inferred from DNA sequences of rbcL. Ann Mo Bot Gard
79:249–265. doi:10.2307/2399768
Panero JL, Funk VA (2008) The value of sampling anomalous taxa in
phylogenetic studies: Major clades of the Asteraceae revealed.
Mol Phylogenet Evol 47:757–782. doi:10.1016/j.ympev.2008.
02.011
Peitsch D, Fietz A, Hertel H, de Souza J, Ventura DF, Menzel R
(1992) The spectral input systems of hymenopteran insects and
their receptor-based colour vision. J Comp Physiol A Neuroethol
Sens Neural Behav Physiol 170:23–40
R Development Core Team (2004) R: a language and environment for
statistical computing. R Foundation for Statistical Computing,
Vienna, Austria
Raine NE, Chittka L (2007) The adaptive significance of sensory bias
in a foraging context: floral colour preferences in the bumblebee
Bombus terrestris. PLoS ONE 2:e556–e557. doi:10.1371/journal.
pone.0000556
Raine NE, Ings TC, Dornhaus A, Saleh N, Chittka L (2006)
Adaptation, genetic drift, pleiotropy, and history in the evolution
of bee foraging behavior. Adv Stud Behav 36:305–354. doi:
10.1016/S0065-3454(06)36007-X
Rodriguez-Girones MA, Santamaria L (2004) Why are so many bird
flowers red? PLoS Biol 2:1515–1519. doi:10.1371/journal.pbio.
0020350
Savile DBO (1972) Arctic adaptations in plants. In: Canada Depart-
ment of Agriculture, Ottawa
Silvertown J et al (2006) Absence of phylogenetic signal in the niche
structure of meadow plant communities. Proc R Soc Lond B Biol
Sci 273:39–44. doi:10.1098/rspb.2005.3288
Sison-Mangus MP, Bernard GD, Lampel J, Briscoe AD (2006)
Beauty in the eye of the beholder: the two blue opsins of
lycaenid butterflies and the opsin gene-driven evolution of
sexually dimorphic eyes. J Exp Biol 209:3079–3090. doi:
10.1242/jeb.02360
Skorupski P, Do
¨
ring TF, Chittka L (2007) Photoreceptor spectral
sensitivity in island and mainland populations of the bumblebee,
Bombus terrestris. J Comp Physiol A Neuroethol Sens Neural
Behav Physiol 193:485–494
Sollid JL, Isaksen K, Eiken T, Ødega
˚
rd RS (2003) The transition zone
of mountain permafrost on Dovrefjell, southern Norway. In: 8th
International Conference on Permafrost, vol. 1–2, Zu
¨
rich,
Switzerland, pp 1085–1089
Soltis DE, Morgan DR, Grable A, Soltis PS, Kuzoff R (1993)
Molecular systematics of Saxifragaceae sensu stricto. Am J Bot
80:1056–1081. doi:10.2307/2445753
Soltis PS, Soltis DE, Chase MW (1999) Angiosperm phylogeny
inferred from multiple genes as a tool for comparative biology.
Nature 402:402–404. doi:10.1038/46528
Soltis DE et al (2001) Elucidating deep-level phylogenetic relation-
ships in Saxifragaceae using sequences for six chloroplastic and
nuclear DNA regions. Ann Mo Bot Gard 88:669–693. doi:
10.2307/3298639
Swensen SM (1996) The evolution of actinorhizal symbioses:
evidence for multiple origins of the symbiotic association. Am
J Bot 83:1503–1512. doi:10.2307/2446104
Swofford DL (2002) PAUP*: phylogenetic analysis using parsimony
(*and other methods). Sinauer Associates, Inc., Sunderland, MA
Tastard E, Andalo C, Giurfa M, Burrus M, The
´
baud C (2008) Flower
colour variation across a hybrid zone in Antirrhinum as
perceived by bumblebee pollinators. Arthropod-Plant Interact
2:237–246. doi:10.1007/s11829-008-9046-3
Tokuoka T (2008) Molecular phylogenetic analysis of Violaceae
(Malpighiales) based on plastid and nuclear DNA sequences.
J Plant Res 121:253–260. doi:10.1007/s10265-008-0153-0
Totland Ø (1992) Pollination ecology in alpine plant communities in
southern Norway: effect of abiotic and biotic factors on insect
visitation and interspecific interactions. University of Bergen,
Norway
Totland Ø (1993) Pollination in alpine Norway: flowering phenology,
insect visitors, and visitation rates in two plant communities. Can
J Bot 71:1072–1079
Totland Ø, Eide W, Grytnes JA (2000) Is there a typical alpine
flower? In: Totland Ø (ed) The Scandinavian association for
pollination ecology honours Knut Fægri, vol 1. Det Norske
Videnskaps-Akademi, Oslo, pp 139–148
Trift I, Ka
¨
llersjo
¨
M, Anderberg AA (2002) The monophyly of
Primula (Primulaceae) evaluated by analysis of sequences from
the chloroplast gene rbcL. Syst Bot 27:396–407
Troje N (1993) Spectral categories in the learning behaviour of
blowflies. Z Naturforsch 48c:96–104
Va
´
zquez DP, Morris WF, Jordano P (2005) Interaction frequency as a
surrogate for the total effect of animal mutualists on plants. Ecol
Lett 8:1088–1094. doi:10.1111/j.1461-0248.2005.00810.x
Wagstaff SJ, Bayly MJ, Garnock-Jones PJ, Albach DC (2002)
Classification, origin, and diversification of the New Zealand
hebes (Scrophulariaceae). Ann Mo Bot Gard 89:38–63. doi:
10.2307/3298656
Wang W, Li R-Q, Chen Z-D (2005) Systematic position of
Asteropyrum (Ranunculaceae) inferred from chloroplast and
nuclear sequences. Plant Syst Evol 255:41–54. doi:10.1007/
s00606-005-0339-z
42 S. E. J. Arnold et al.
123

Warren J, Mackenzie S (2001) Why are all colour combinations not
equally represented as flower-colour polymorphisms? New
Phytol 151:237–241. doi:10.1046/j.1469-8137.2001.00159.x
Waser NM (1983) The adaptive nature of floral traits: ideas and
evidence. In: Real LA (ed) Pollination biology. Academic Press,
New York, pp 241–285
Weevers T (1952) Flower colours and their frequency. Acta Bot Neerl
1:81–92
West W, West GS (1910) Sketches of vegetation at home and abroad.
V. The ecology of the Upper Driva Valley in the Dovrefjeld.
New Phytol 9:353–374. doi:10.1111/j.1469-8137.1910.tb
05557.x
Whibley AC et al (2006) Evolutionary paths underlying flower color
variation in Antirrhinum. Science 313:963–966. doi:10.1126/
science.1129161
Yasui Y, Ohnishi O (1996) Comparative study of rbcL gene
sequences in Fagopyrum and related taxa. Genes Genet Syst 71:
219–224. doi:10.1266/ggs.71.219
Flower colours along an alpine altitude gradient, seen through the eyes of fly and bee pollinators 43
123