Journal of
Plant Ecology
VOLUME 10, NUMBER 2,
PAGES 397–405
April 2017
doi:10.1093/jpe/rtw029
Advance Access publication
18 April 2016
available online at
academic.oup.com/jpe
Weak correlation of flower color
and nectar-tube depth in temperate
grasslands
Julia Binkenstein1,*, Martina Stang2, Julien P. Renoult3 and
H. Martin Schaefer4
1
Chair of Nature Conservation and Landscape Ecology, Faculty of Environment and Natural Resources, University of
Freiburg, Tennenbacher Straße 4, 79106 Freiburg, Germany
2
Section Plant Ecology and Phytochemistry, Institute of Biology Leiden, Leiden University, Sylviusweg 72, 2300 RA Leiden,
The Netherlands
3
Institute of Arts Creation Theory and Aesthetics, UMR 8218-CNRS, 47 r. des bergers, 75015 Paris, France
4
Fundacion Jocotoco, Lizardo García, E9-104 y Andrés Xaura, P.O. Box 17-16-337, Quito, Ecuador
*Correspondence address. Chair of Nature Conservation and Landscape Ecology, Faculty of Environment
and Natural Resources, University of Freiburg, Tennenbacher Straße 4, 79106 Freiburg, Germany. E-mail:
[email protected]
Abstract
Aims
Nectar is one of the most common floral rewards offered to pollinators by plants. Depending on the plant species, nectar is offered
openly or in tubes of various lengths restricting accessibility of
this resource for flower visitors with short mouthparts. If attracting pollinators that match floral morphology increases pollination
efficiency, flowers could profit from signaling nectar-tube depth to
pollinators. Since flower colors are important signals in plant–pollinator communication, we investigated whether and which different chromatic or achromatic aspects of flower color might indicate
nectar-tube depth or whether flower colors facilitate the differentiation between flowers with long nectar tubes by means of high
chromatic uniqueness.
Methods
To this end, we collected flower reflectance spectra of 135 grassland
plant species. We analyzed flower colors as raw reflectance spectra
in principal component analysis (PCA) and in the color space of
honeybees.
Important Findings
The correlation between flower colors and tube depths was weak.
From the bee’s point of view, blue flowers had on average deeper
tubes than green, blue-green and UV-green flowers potentially allowing insects to predict tube depths based on blue color. Spectral purity
did not correlate with nectar-tube depth, nor did the chromatic
uniqueness of flower colors in the honeybee color space. Dominant
wavelength showed a significant but very weak correlation with tube
depth. The achromatic green contrast decreased with increasing tube
depth as did brightness; thus deep tubes were less conspicuous than
shallow tubes. Chromatic components resulting from PCA did not
or only slightly correlate with tube depth. Our results illustrate that
flower colors may have a certain potential to indicate tube depth, i.e.
nectar accessibility, from a bee’s perspective.
Keywords: vision, pollination efficiency, plant–animal interaction,
nectar, tube depth
Received: 27 March 2015, Revised: 14 December 2015, Accepted:
12 March 2016
INTRODUCTION
Up to 90% of all flowering plant species depend on animal
pollinators for their reproduction. To attract pollinators, recent
angiosperms commonly offer floral nectar as reward (Simpson
and Neff 1983). Floral nectar can be offered in an accessible
manner like in open Apiaceae flowers or can be hidden in
nectar tubes of various depths. Nectar-tube depth is a floral
trait restricting nectar accessibility to a certain range of flower
visitors with adequate morphological traits (Ibanez 2012). The
degree of morphological match between flowers and pollinators can strongly influence aspects of pollination efficiency,
e.g. rates of pollen removal and deposition, which can directly
affect plant reproductive success (Hiei and Suzuki 2001; Ibanez
2012; Nilsson 1988; Nagano et al. 2014; Pauw et al. 2009). Deep
nectar tubes allow plants to protect their nectar against less efficient pollinators and limit their pollinator spectrum to a small
range of pollinators that match their floral morphology.
© The Author 2016. Published by Oxford University Press on behalf of the Institute of Botany, Chinese Academy of Sciences and the Botanical Society of China.
All rights reserved. For permissions, please email: [email protected]
398
Pollinators have to make economic decisions when foraging
(Goulson 1999). A high frequency of wrong landing decisions
can sum up to considerable costs, i.e. landing on flowers of plant
species that give no access to their nectar may consume foraging
time and energy, whereas missing flowers with accessible nectar
means missing potential energy. Especially, when the proportion of rewarding flowers in a bee’s foraging area is low, accurate
flower choice is important (Burns and Dyer 2008). For a pollinator with a given morphology, i.e. proboscis length, nectar accessibility and the efficiency of nectar extraction can strongly vary
between plant species as a function of nectar-tube depth (Inouye
1980). Therefore, pollinators would strongly profit from being
able to learn floral traits associated with tube depth to assess
which flowers provide most easily exploitable rewards for them
in a given plant community. Several studies have demonstrated
that foraging decisions in pollinators are strongly linked to learning and memory (Riveros and Gronenberg 2012). Honeybees
and other pollinators are able to learn associations like those
between flower color and nectar rewards (Giurfa 2007; Riveros
and Gronenberg 2012). Morphological flower traits, particularly
nectar-tube depth, have been shown to be closely related to pollinator foraging choices and thereby filter interactions between
plants and animals in pollination networks (Junker et al. 2013;
Stang et al. 2007). If pollinators were able to pay attention to floral traits associated with tube depth, they could easily distinguish
the length of floral nectar tubes from a distance before landing and thereby assess accessibility of rewards without wasting
energy to land on a flower and try to access its nectar.
Flower color is one of the most important floral traits used for
detection and recognition of flowers by flower visitors. Flower
color facilitates the initial recognition of flowers for naive pollinators, and can be a predictor of the reward type, e.g. pollen or nectar, and the amount of floral rewards (Giurfa et al.
1995; Menzel and Shmida 1993). Furthermore flower color
plays an important role in flower constancy behavior of bees,
i.e. the tendency to restrict flower visits to certain plant species
or morphs within one foraging trip (Chittka et al. 1999). Flower
constancy increases pollination efficiency, since it reduces
negative effects of heterospecific pollen deposition (Feinsinger
1987 and references therein). Flower color has been shown to
correlate with various plant traits, e.g. secondary compounds
(Irwin et al. 2003), inter- and intraspecific differences in nectar
rewards (Giurfa et al. 1995; Raine and Chittka 2007, Willmer
et al. 2009), or the general suitability of a flower type for different pollinator groups like insects and birds (Shrestha et al.
2013). Corolla color and nectar accessibility are commonly
used for the classification of flowers as in pollination syndromes (Faegri and van der Pijl 1979). In their study, Junker
et al. (2013) illustrated that floral reflectance, i.e. flower color,
was next to phenology, flower height and tube depth one of
the four most important floral traits influencing the specialization degree of flower visitors. Therefore, the importance of
flower colors to explain network structure may be proximally
explained as it acts as an informative cue to flower visitors
about nectar-tube depth, i.e. the accessibility of nectar.
Journal of Plant Ecology
In order to understand the relationship between flower
color and nectar-tube depth, we test two hypotheses in our
study. The first hypothesis is: Flowers indicate nectar accessibility, i.e. tube depth, by petal coloration to attract pollinators that match their own morphology, because this increases
pollination efficiency and thereby plant reproductive success.
Particularly, plants with long-tubed flowers could profit from
an association between petal color and tube depth, because
short-tongued visitors might avoid these owing to associative learning as they cannot reach the nectar in those longtubed flowers. Thereby interference with legitimate, i.e.
long-tongued, pollinators might be reduced. A short-tongued
insect on a long-tubed flower may represent an interference
because it could alter the electric potential of the flower at
least for some seconds (Clarke et al. 2013). Other pollinators, specifically long-tongued bumblebees which are legitimate pollinators for long-tubed flowers, could misinterpret
an altered electric potential as a cue for diminished floral
resources as they are able to detect floral electric fields (Clarke
et al. 2013). Furthermore, at a high density of flower visitors,
short-tongued insects probing the accessibility of long-tubed
flowers may simply inhibit legitimate pollinators to reach the
flower by their physical presence. Pollinators would profit
from an indication of tube depth by petal color as learning
that association would enable economic foraging decisions. By
increasing foraging efficiency of pollinators, plants would promote the pollinator community in their habitat, which they
would profit from in the following year. The second hypothesis is based on the idea that plants may use petal coloration to
facilitate the differentiation between plant species with deep
nectar tubes by pollinators. Generalized plant species with
short or no nectar tubes may have similar but bright colors
like white facilitating the detection of food sources for opportunistic pollinators like flies. In contrast, deep-tubed flowers
depend on long-tongued pollinators for efficient pollination.
Distinct flower colors of plants with long-tubed flowers would
promote flower constancy of long-tongued bees. increasing
pollination efficiency (Chittka et al. 1999). Accordingly, the
classical pollination syndromes state that bee-pollinated plants
appear to be considerably more diverse in flower colors than
plants pollinated by short-tongued insects like flies (Willmer
2011). We regard the distinctness of flower colors of certain
plant species compared to other species in the color space of
honeybees as chromatic uniqueness. We analyze this uniqueness using a measure of density in color space. Specifically, we
hypothesize: The deeper the nectar tube of flowers, the more
they differ in color from other species.
METHODS
Study area
We collected data on flower colors from three German regions
in the framework of the Biodiversity Exploratories project in
2009 (Fischer et al. 2010): Hainich-Dün (HD), SchorfheideChorin (SC) and Schwäbische Alb (SA) are located in Central,
Binkenstein et al. | Weak correlation of flower color and nectar-tube depth 399
Northeast, and Southwest of Germany, respectively. We
collected flower colors in early and in late summer in each
region.
Color measurements
We measured the reflectance spectra of 20 flowers from different individuals of 125 plant species found in the six plant
communities (for full species list see supplementary Table S1).
Color measurements were performed with an AvaSpec-2048
Fiber Optic Spectrometer and an AvaLight-XE as a standardized light source (both Avantes, RB Eerbeek, The Netherlands).
Reflectance was measured as the proportion of a standard
white reference tile (WS-2) at an angle of 90° (Binkenstein
et al. 2013). For each color, we calculated a mean spectrum
between 300 and 700 nm. In addition, we retrieved 10 floral
reflectance spectra (Allium vineale, Anthemis arvensis, Arenaria
serpyllifolia, Astragalus glycyphyllos, Conyza canadensis, Euphorbia
helioscopa, Galium aparine, Matricaria recutita, Pimpinella major
and Vicia hirsuta) from The Floral Reflectance Database FReD
(Arnold et al. 2010). We excluded one species, Silene latifolia,
from this study because its deep white flowers are predominantly pollinated by nocturnal sphingid and noctuid moths
(Young 2002), which is unique in our data set. Response to
color parameters likely differs between nocturnal and diurnal pollinators. To model background reflectance, we used a
mean spectrum of green leaves measured at the mean height
of flowers. We used one mean background spectrum for all
regions, because reflectance spectra of green foliage were very
similar in all regions.
the color hexagon (Chittka 1992), since it has been shown
to predict behavioral and ecological data very well (Chittka
1999; Spaethe et al. 2001) and is therefore well-known and
widely used in recent ecological studies. Chittka et al. (1994)
proposed a system to categorize reflectance spectra according to their maximum reflection: 300–400 nm = UV (u), 400–
500 nm = blue (b), 500–600 nm = green (g), 600–700 nm = red
(r; visual sensitivity for red colors is very low in bees, see
Peitsch et al. 1992). We use these terms to describe colors in the
following. The color hexagon can be divided into six equally
sized parts representing UV, blue, green and mixed color categories, i.e. blue-green, UV-green and UV-blue (Fig. 1). It is
not known whether bees make color categorizations according to hexagon color categories. Several studies have shown
that honeybees are able to categorize objects based on different visual features (see Benard et al. 2006). Likewise, there
is evidence from several studies that honeybees may be able
to categorize colors (Benard and Giurfa 2008). At least, bees
are able to generalize across flower colors, which is a basic
requirement for categorization (Gumbert 2000, reviewed in
Avarguès-Weber et al. 2012).
The chromatic distance between color loci in the color hexagon is calculated as Euclidean distance, which indicates their
discriminability. Bumblebees do not reliably discriminate
color distances smaller than 0.04 units in the color hexagon,
distances between 0.04 and 0.11 units can be discriminated
if bees receive differential conditioning, and distances greater
than 0.11 hexagon units are generally discriminated reliably
Spectra processing
To analyze the correlation between colors and nectar-tube
depth from a pollinator’s perspective, we chose the visual system of the honeybee (Apis mellifera) as a model. In temperate
grasslands, honeybees are abundant and efficient pollinators
for the majority of flowering plant species (Faegri and van der
Pijl 1979). Their color vision system is well known (for review
see e.g. Avarguès-Weber et al. 2012) and very similar to that of
many other bee species (Peitsch et al. 1992). Data from several
studies suggest that the most frequent flower colors in the
Northern Hemisphere as well as in Australia and even flower
colors measured over a considerable elevational range in
Nepal seem to have evolved to be maximally discriminable by
hymenopteran insects, illustrating the importance of hymenopteran color vision for flower color evolution (Chittka and
Menzel 1992; Chittka et al. 1994; Dyer et al. 2012; Shrestha
et al. 2014). Honeybees possess three types of photoreceptors
with sensitivity maxima in the short (344 nm, UV), medium
(436 nm, blue) and long wavelength range (544 nm, green;
Peitsch et al. 1992). Their inputs are evaluated by two opponent color coding mechanisms, which is universal in hymenoptera (Chittka et al. 1992). Flower colors can be plotted in a
two dimensional chromaticity diagram based on the sensitivities of the photoreceptor types. Different models have been
proposed to draw chromaticity diagrams. We decided to use
Figure 1: flower color loci of 135 plant species from three German
regions (Hainich-Dün, Schorfheide-Chorin, Schwäbische Alb) in the
color hexagon. The centre of the color hexagon represents the leafgreen background. The angular position of a color locus in the hexagon is determined by the relative excitation (E) of the photoreceptor
types (Blue, Green, UV). Filled circles connected by a line indicate the
spectral loci of monochromatic lights.
400
(Dyer 2006). These thresholds have likewise been applied in
studies of other bee species (Dyer et al. 2012).
Different components of flower colors affect the behavior
of honeybees and bumblebees, i.e. chromatic components like
dominant wavelength and purity as well as achromatic components like the green contrast (Rohde et al. 2012; Papiorek
et al. 2013). The green contrast of a flower is calculated as
the receptor-specific contrast between target and background
perceived by the long-wavelength receptor (i.e. the green
receptor). Values of 1 represent a lack of achromatic contrast
between target and background, whereas the achromatic contrast increases with deviation from 1 towards higher or lower
values (Niggebrügge and Hempel de Ibarra 2003). To facilitate
the interpretation of green contrast values, we applied the
following transformation on green contrasts (gc; Binkenstein
and Schaefer 2015): │(gc) − 1│. Transformed excitations of
0 indicate no achromatic contrast, increasing values indicate increasing achromatic contrast. Because no other receptor types are involved, this signal is achromatic (Land and
Chittka 2013). The green contrast cannot be inferred from the
color hexagon. The dominant wavelength of a flower color
is the locus of the corresponding monochromatic light in the
hexagon (spectral locus). We calculated purity as the chromatic distance between target and background divided by the
chromatic distance between dominant wavelength and background (Lunau et al. 1996). Dominant wavelength, purity and
green contrast are related to the human perception of hue,
saturation and brightness. At small visual angles (5°–15°), i.e.
in case of very small or faraway objects, the green contrast
mediates detection of flowers (Giurfa et al. 1996; Giurfa and
Vorobyev 1998). At large visual angles (> 15°), honeybees use
only chromatic cues to detect and recognize flowers (Hempel
de Ibarra et al. 2002).
We estimated the chromatic uniqueness of flower colors
from the density distribution of all flower colors across the
color space. We first estimated a multivariate kernel density function to characterize the spatial distribution of flower
colors in the color hexagon (R-function npudens in package
np; Li and Racine 2003). We set the bandwidth, which specifies how smooth the density function is, to 0.04 for both x
and y as this value corresponds to the lowest discrimination
thresholds in bumblebees (Dyer 2006). This value is very close
to the optimal bandwidth (0.06 for x and 0.02 for y) estimated
for the density function by a likelihood cross-validation procedure (R-function npudensbw in package np). We then used
the density function to evaluate the focal density of flower
colors at each locus occupied by a flower color. Low focal density corresponds to high uniqueness (see supplementary Fig.
S1). We transformed focal density to uniqueness as follows:
uniqueness = 1 − focal density + (max (focal density)).
In a previous study, we found a significant phylogenetic signal of bee-subjective chromatic components of flower colors
(Binkenstein et al. 2013). Therefore, any correlations between
tube depth and chromatic components of flower colors from
the bee’s perspective could reflect an unbalanced sampling
Journal of Plant Ecology
across plant families instead of an ecological relationship. In
addition to the analysis of flower colors from a bee’s perspective, we analyzed reflectance spectra without considering a
visual model for pollinators. This allows an unbiased analysis
of the correlation between colors and morphology of flowers.
Reflectance data are multi-dimensional, i.e. each wavelength
for which reflectance has been recorded represents one dimension (i.e. one variable in analyses). To reduce this high dimensionality of reflectance spectra, we applied principal component
analysis (PCA). In case of reflectance spectra, PC1 is usually
highly correlated with brightness, while other PCs are correlated with chromatic cues (Cuthill et al. 1999). We decided to
keep as many PCs as necessary to explain at least 95% of the
variation in original spectra, which were four PCs in our case.
Tube depths
We measured the tube depth of 95 species using a dissecting microscope connected to a monitor screen (see Junker
et al. 2013 for details). We estimated the nectar tube-depths
of another 40 species from Jager (2009), wherein excellent
drawings with scales are given. Tubes were formed by hairs,
the receptacle, the calyx, the corolla, filaments or a combination of organs. We use the term nectar tube for all of these
structures. To compensate intraspecific variations and inaccuracies resulting from different methods to estimate tube
depths (measurement and literature survey), we approximated tube depths into classes of 2 mm, e.g. 0.1–2, 2.1–4,
4.1–6 and used mean class values for further calculations. In
case a nectar tube was absent and nectar glands were openly
accessible tube depth was scored as 0mm.
Statistical analyses
We used Kruskal–Wallis test to test for differences in tube
depths between hexagon colors, since data were distributed
non-normally (Shapiro–Wilk test) and variances were heterogeneous (Levene test). In case Kruskal–Wallis test revealed
significant differences, we applied Mann–Whitney U-tests
(with Holm correction) for multiple comparisons. To test
whether tube depth is correlated to chromatic uniqueness or
to chromatic and achromatic color components from the bees’
point of view (purity, dominant wavelength and green contrast) as well as without any vision model (PCs) we conducted
multiple linear regression analyses (global models) and simple linear regression analyses. Prior to regression analyses we
performed Goldfeld–Quandt test to test for homogeneity of
variances (function gqtest in R package lmtest; Zeileis and
Hothorn 2002). Tube depth was log-transformed to improve
normality. Regression residuals did not excessively deviate
from normality in any case. We assessed collinearity between
the four bee-subjective color parameters. These parameters
were in no case correlated with Spearman’s r > 0.7 suggesting an acceptable degree of multicollinearity (Dormann et al.
2013). We used Spearman’s r because parameters were not
normally distributed. Statistical analyses were conducted in R
2.15.3 (R Development Core Team 2013).
Binkenstein et al. | Weak correlation of flower color and nectar-tube depth 401
RESULTS
We identified 135 flowering plant species across all communities (see supplementary Table S1 for full species list).
Blue flowers had significantly deeper tubes than blue-green,
green and UV-green flowers (bee subjective colors; Figs 2 and
3). However, in community-specific analyses (one for each
region and study period) this difference in tube depth was
not consistent, which may be attributed to the lower sample
size (plant species number, i.e. 47–69) per community compared to the global analysis (supplementary Fig. S5). Tube
depth did not significantly correlate with purity, but with
dominant wavelength in both simple and multiple regression
analyses (Tables 1 and 3). However, dominant wavelength
explained only 3.3% of the variation in tube depth. Both variables are chromatic color components assessed from the bees’
perspective.
Phylogenetic constraints are not expected to strongly affect our
results by the unbalanced sampling across plant families. We found
species from 27 different plant families (supplementary Table S1;
supplementary Fig. S7). The most common plant families were
Asteraceae and Fabaceae being represented by 24 and 19 species,
respectively. The majority of both plant families was represented by
three different nearly evenly distributed hexagon colors (supplementary Fig. S7). Another two plant families represented by more
than 10 species were Apiaceae and Lamiaceae. Dominating colors
in these two groups were blue-green in Apiaceae and blue in
Lamiaceae. These colors significantly differ in tube depths in our
analysis. However, both families comprise only 11 species each.
All other families had less than 10 species and therefore are not
expected to influence our results in the sense of phylogenetic
constraints.
The achromatic component of flower colors as evaluated by bees, i.e. the green contrast, significantly decreased
with increasing tube depth (Table 1). Likewise, green contrast predicted tube depth when considered separately (in
simple regression analysis; Table 3). However, green contrast accounted for only 7 % of the variation in tube depth
(Table 3).
The first four principal components received from PCA
explained 96.35% of the variation among reflectance
spectra (PC1: 52.87%, PC2: 23.40%, PC3: 17.03%, PC4:
3.05%). PC1 was mainly correlated with intensity of
reflectance (brightness), not with specific wavelengths as
other PCs (supplementary Figs S2 and S3). PC1 scores of
flowers significantly predicted tube depth in our global
model (multiple regression analysis), i.e. tube depth significantly decreased with increasing brightness of flower
colors (Table 2). However, effect strength was weak (3.3%,
Table 3). Whereas PC1 was positively correlated with all
wavelengths, PC2 and PC3 varied more strongly with
wavelengths: PC2 correlated positively with blue wavelengths and negatively with UV and green wavelengths
(bee colors, supplementary Fig. S2). PC3 positively correlated with UV and blue wavelengths and negatively
correlated with green wavelengths (for examples see supplementary Fig. S3). Tube depth was not significantly
associated to PC2 and PC3. PC4 was negatively correlated
with UV, blue and green wavelengths and positively correlated with red wavelengths. In the red wavelength part
of the spectrum (>600 nm) the bees’ chromatic sensitivity
is very low. Therefore, for bees PC4 may not be very important, but it could be for other insect species. PC4 predicted
20.07% of the variation in tube depth (Tables 2 and 3).
Chromatic uniqueness of flower colors in the color hexagon did not significantly correlate with tube depth in any
analysis (Tables 1 and 3).
Results for individual plant communities (i.e. region and
season) only differ from the full species pool results in case of
dominant wavelength, which does not significantly correlate
with tube depth in individual plant communities (see supplementary Figs S4–S6; supplementary Tables S2–S5).
Figure 2: nectar tube depths of flowers mapped in the honeybee’s color hexagon according to their respective flower color loci (see Fig. 1).
402
Journal of Plant Ecology
DISCUSSION
In our study, we aimed to understand whether flower colors
are associated with nectar-tube depths in a way that is perceivable for pollinators. Bee-blue flowers, which appear blue,
violet or pink to human observers, had on average the longest
tubes. We found that bee-blue flowers had significantly longer
tubes than bee-green, bee-blue-green and bee-UV-green flowers. Therefore, at least bee-blue flowers may be able to signal
the depth of their tubes to flower visitors and thereby attract
flower visitors that more often match their morphology than
differently colored flowers. Bee-green and UV-green flowers
are yellow to human observers. Bee-blue-green flowers are
white or pink to humans. For example, most Apiaceae fall into
the group of blue-green flowering species. They appear white
to human observers and have no nectar tube. Insects with
short mouthparts are restricted to flowers with short tubes
when searching for accessible nectar. In contrast, blue and
violet flowers, which are often long-tubed are preferred by
bumblebees and honeybees, because these flowers are often
the most profitable ones for these insects (Raine and Chittka
2007). These differences in tube depths between color categories are largely in accordance with a study of Menzel and
Shmida (1993) who found similar patterns in the Israeli flora.
In their study about 70% of all flowers with long tubes fell into
the blue and UV-blue categories, whereas only about 10% of
long-tubed flowers were blue-green and none were UV-green
(Menzel and Shmida 1993). These authors furthermore found
that long tubed flowers were often UV-blue, which we cannot
confirm for the German grassland flora in our study. However,
the UV-blue category had the lowest species number (n=13)
of all color categories in our study, reducing statistical power.
Despite significant differences between tube depths of some
color categories the correlation between flower colors and tube
depth was not consistent across communities (see supplementary Fig. S5) and was not very strong for hymenopteran eyes:
the variation in tube depths was large in all color categories
(mean relative standard deviation: 61.45%), which implies a
lack of tight functional correlations between tube depth and
flower color. Thus, the combination of flower color and tube
depth commonly used for the classification of flowers, e.g. in
Table 2: predictive effect of four principal components of floral
reflectance spectra on tube depth (log-transformed) in 135 plant
species based on multiple regression analyses
Parameter
Intercept
Figure 3: mean tube depths (log-transformed; ± SD) within hexagon color categories of 135 plant species (b = blue, bg = blue-green,
g = green, ub = UV-blue, ug = UV-green; see Fig. 1). Numbers in bars
indicate species numbers. Different letters above error bars indicate
significant differences (P < 0.05) in tube depth between hexagon categories (Mann–Whitney U-test with Holm correction).
Table 1: predictive effect of four bee-subjective color parameters
on tube depth (log-transformed) in 135 plant species based on
multiple regression analyses
Parameter
Estimate
Intercept
2.391
SE
0.677
t
3.531
Estimate
SE
t
P
<0.001
1.216
0.060
20.182
PC1
−0.168
0.061
−2.738
0.007
PC2
−0.080
0.060
−1.324
0.188
PC3
0.075
0.060
1.254
0.212
PC4
0.374
0.060
6.208
<0.001
Significant P-values are indicated in bold.
Table 3: predictive effect of eight color parameters on tube depth
(log-transformed) in 135 plant species based on simple linear
regression analyses
Paramater
Slope
Intercept
r2
T
P
Green contrast
−0.255
1.470
0.071
−3.358
0.001
Dominant wavelength
−0.003
2.798
0.033
−2.372
0.019
P
Uniqueness
0.019
0.988
0.014
1.719
0.088
<0.001
Purity
0.364
1.061
0.004
0.724
0.47
Green contrast
−0.186
0.086
−2.169
0.032
PC1
−0.164
1.217
0.033
−2.346
0.021
Dominant wavelength
−0.003
0.001
−2.185
0.031
PC2
0.087
1.221
0.004
1.252
0.213
Uniqueness
0.019
0.012
1.569
0.119
PC3
0.075
1.220
0.001
1.078
0.283
Purity
0.785
0.487
1.611
0.109
PC4
0.372
1.218
0.207
5.997
<0.001
Significant P-values are indicated in bold.
Significant P-values are indicated in bold.
Binkenstein et al. | Weak correlation of flower color and nectar-tube depth 403
pollination syndromes, is not very informative at least for most
entomophilous flowers in Germany.
In our study, we also asked whether color purity or dominant wavelength indicate tube depth. Purity and dominant
wavelength are color components that affect the behavior of
hymenopteran flower visitors. For example, Rohde et al. (2012)
and Papiorek et al. (2013) reported that honeybees and bumblebees prefer pure colors over less pure ones and exhibit a strong
fidelity for the dominant wavelength of a trained color. In contrast, Gumbert (2000) found a general preference for a dominant wavelength of 410 nm in bumblebees. However, although
purity and dominant wavelength have the potential to affect
the foraging behavior of pollinators, purity is not and dominant
wavelength is only very weakly correlated to tube depth.
In contrast to chromatic components of flower colors, tube
depth significantly decreased with the achromatic component
evaluated by hymenopterans, i.e. the green contrast. Green
contrast is an important cue used by bees for the long-range
detection of flowers. But can the green contrast indicate nectar
accessibility? Blue-green flowers have relatively short tubes
and often very bright colors. In contrast to blue-green flowers,
long-tubed blue flowers show a low overall reflection across
the wavelength range visible for bees, i.e. they are darker.
Thus, the correlation between green contrast and tube depth
may mirror the difference in tube depth between blue and
blue-green flowers. Furthermore, green contrast accounted
for only 7% of the variation in tube depths. Thus, although
green contrast could be part of a pollination syndrome including long nectar tubes, bees cannot be expected to reliably distinguish tube depths only based on green contrast information.
The analysis of PCA data in relation to tube depth shows that
the weak correlation between colors and tube depths that we
found in the honeybee color space is probably not limited to
honeybees, i.e. it does not result from the modalities of this particular visual system. As green contrast, PC1 was strongly correlated with brightness and therefore likewise revealed a negative
correlation of brightness with tube depth indicating that the
perception of this correlation is probably not restricted to the
visual system of Hymenoptera. Whereas PC2 and PC3 did not
correlate with tube depth, PC4 was a relatively good predictor
for tube depth. On the one hand the overall variation of reflectance spectra was poorly correlated to PC4 (3.05%). Therefore,
the predictive value of PC4 may be very limited in a natural
context. On the other hand, the very small particular region of
the spectrum PC4 is correlated with (red, i.e. 600–700 nm) may
be an important one for particular pollinator taxa. The study of
Papiorek et al. (2013) indicates that even subtle spectral changes
may influence honeybee behavior. However, the correlation
between tube depth and PC4 mainly resulted from the characteristics of only three taxa: Dianthus carthusianorum and Silene
dioica have very long flower tubes (>10 cm) and their flowers
strongly reflect (ca.80%) in the red wavelength range PC4 is
correlated with (600–700 nm), whereas all Apiaceae have no
nectar tubes and reflect only 40–60% of the light above 600nm.
Summarizing, the results based on PCA-scores indicate that the
lack of a tight correlation between petal colors and tube depths
is a general phenomenon. If flowers signal tube depth, floral
signals like scent may be at least as or more important to allow
flower visitors to predict nectar accessibility or function in addition to color for that purpose (Leonard et al. 2011).
We reject our second hypothesis that deep-tubed flowers have more distinct colors than short-tubed flowers. The
chromatic uniqueness of flowers was not correlated with tube
depth, i.e. long-tubed flowers were not more distinct in flower
color from the rest of the species pool than short-tubed flowers.
Thus, flower colors are not very likely to induce flower constancy more strongly in long-tubed flowers than in short-tubed
flowers. One reason for this finding may be that short-tubed
flowers may likewise strongly profit from flower constancy.
Flower constancy behavior has been demonstrated not only in
long-tongued flower visitors, but likewise in short-tongued species like hoverflies that cannot exploit nectar from deep tubes
(Goulson and Wright 1998). Flowers presenting their nectar
openly may even profit more strongly from flower constancy
than long-tubed flowers, since long-tubed flowers possess more
additional mechanisms to promote the deposition of conspecific
pollen. Plants with rather complex floral morphologies can, for
example, place their pollen grains on specific parts of the pollinators body that precisely touch the stigma of conspecific flowers to avoid mixture with foreign pollen grains (Sargent 2004).
Thus, the common notion that bee-pollinated plants appear to
be more diverse in flower colors than plants pollinated by shorttongued insects like flies (Willmer 2011) does not translate into
a higher chromatic uniqueness of deep-tubed flowers, since
even opportunistic pollinators like short-tongued flies may benefit their food plants by flower constancy behavior.
In conclusion, we found no correlation between chromatic
uniqueness and floral tube depth indicating that long-tubed
flowers do not use flower colors to promote flower constancy
to a larger extent than short-tubed flowers to increase pollination efficiency. Furthermore, plants do not seem to use chromatic or achromatic components of flower colors to specifically
attract pollinators that match their morphology to increase pollination efficiency. On the one hand, our hypotheses are based
on a co-evolutionary scenario between plants and pollinators.
The regions where we collected our data have been considerably modified by diverse land-use practices within the past centuries (Fischer et al. 2010). This probably changed the species
diversity of plants and animals and affected the species composition in these regions. The lack of tight functional relationships
between tube depths and flower colors in our data might to a
certain degree refer to this modification of ecosystems. On the
other hand, floral displays provide various combinations of different floral signals, i.e. they are complex signals (Leonard et al.
2011). Therefore, if plants profit from signaling nectar accessibility, other signal components (like scent) alone or in combination with color may indicate nectar accessibility, which needs
to be examined in future studies. Particularly, the achromatic
contrast may in combination with other signal components
constitute a pollination syndrome including deep nectar tubes.
404
SUPPLEMENTARY MATERIAL
Supplementary material is available at Journal of Plant Ecology
online.
FUNDING
The work has been funded by the DFG Priority Program
1374 ‘Infrastructure-Biodiversity-Exploratories’ (Scha1008/51) and by the FAZIT-Stiftung. Field work permits were issued
by the responsible state environmental offices of BadenWürttemberg, Thüringen and Brandenburg (according to §
72 BbgNatSchG). Data collection complies with the current
laws of Germany.
ACKNOWLEDGEMENTS
We thank A. Bogenrieder and S. Boch for helping to identify plant
species. We thank B. Kreuzinger-Janik, K. Kohlberg, J. Kramer and
D. Behringer for field assistance. We thank the managers of the three
Exploratories, Swen Renner, Sonja Gockel, Kerstin Wiesner and Martin
Gorke for their work in maintaining the plot and project infrastructure; Simone Pfeiffer and Christiane Fischer giving support through
the central office, Michael Owonibi for managing the central data base
and Markus Fischer, Eduard Linsenmair, Dominik Hessenmöller, Jens
Nieschulze, Daniel Prati, Ingo Schöning, François Buscot, Ernst-Detlef
Schulze, Wolfgang W. Weisser and the late Elisabeth Kalko for their
role in setting up the Biodiversity Exploratories project.
Conflict of interest statement. None declared.
Journal of Plant Ecology
Chittka L, Beier W, Hertel H, et al. (1992) Opponent colour coding
is a universal strategy to evaluate the photoreceptor inputs in
Hymenoptera. J Comp Physiol A 170:545–63.
Chittka L, Shmida A, Troje N, et al. (1994) Ultraviolet as a component
of flower reflections, and the colour perception of Hymenoptera.
Vis Res 34:1489–508.
Chittka L, Thomson JD, Waser NM (1999) Flower constancy, insect
psychology, and plant evolution. Naturwissenschaften 86:361–77.
Clarke D, Whitney H, Sutton G, et al. (2013) Detection and learning of
floral electric fields by bumblebees. Science 340:66–9.
Cuthill IC, Bennett ATD, Partridge JC, et al. (1999) Plumage reflectance and the objective assessment of avian sexual dichromatism.
Am Nat 153:183–200.
Dormann CF, Elith J, Bacher S, et al. (2013) Collinearity: a review of
methods to deal with it and a simulation study evaluating their
performance. Ecography 36:27–46.
Dyer AG (2006) Discrimination of flower colours in natural settings
by the bumblebee species Bombus terrestris (Hymenoptera: Apidae).
Entomol Gener 28:257–68.
Dyer AG, Boyd-Gerny S, McLoughlin S, et al. (2012) Parallel evolution of angiosperm colour signals: common evolutionary
pressures linked to hymenopteran vision. Proc Roy Soc B Biol Sci
279:3606–615.
Faegri K, van der Pijl L (1979) The Principles of Pollination Ecology, 3rd
ed. Oxford, UK: Pergamon Press.
Feinsinger P (1987) Effects of plant species on each other’s pollination: is community structure influenced? Trends Ecol Evol
2:123–26.
REFERENCES
Fischer M, Bossdorf O, Gockel S, et al. (2010) Implementing large-scale
and long-term functional biodiversity research: The Biodiversity
Exploratories. Basic Appl Ecol 11:473–85.
Arnold SEJ, Faruq S, Savolainen V, et al. (2010) FReD: the floral
reflectance database—a web portal for analyses of flower colour.
PLoS One 5:e14287.
Giurfa M (2007) Behavioral and neural analysis of associative learning in the honeybee: a taste from the magic well. J Comp Physiol A
193:801–24.
Avarguès-Weber A, Mota T, et al. (2012) New vistas on honey bee
vision. Apidologie 43:244–68.
Giurfa M, Vorobyev M (1998) The angular range of achromatic target
detection by honeybees. J Comp Physiol A 183:101–10.
Benard J, Giurfa M (2008) The cognitive implications of asymmetric
color generalization in honeybees. Anim Cogn 11:283–93.
Giurfa M, Nunez J, Chittka L, et al. (1995) Colour preferences of
flower-naive honeybees. J Comp Physiol A 177:247–59.
Benard J, Stach S, Giurfa M (2006) Categorization of visual stimuli in
the honeybee Apis mellifera. Anim Cogn 9:257–70.
Giurfa M, Vorobyev M, Kevan P, et al. (1996) Detection of coloured
stimuli by honeybees: minimum visual angles and receptor specific
contrasts. J Comp Physiol A 178:699–709.
Binkenstein J, Renoult JP, Schaefer HM (2013) Increasing land-use
intensity decreases floral colour diversity of plant communities in
temperate grasslands. Oecologia 173:461–71.
Binkenstein J, Schaefer HM (2015) Flower colours in temperate forest
and grassland habitats: a comparative study. Arthropod Plant Interact
9:289–99.
Burns JG, Dyer AG (2008) Diversity of speed-accuracy strategies benefits social insects. Curr Biol 18:R953–54.
Chittka L (1992) The colour hexagon: a chromaticity diagram based
on photoreceptor excitations as a generalized representation of
colour opponency. J Comp Physiol A 170:533–43.
Goulson D (1999) Foraging strategies of insects for gathering nectar and pollen, and implications for plant ecology and evolution.
Perspect Plant Ecol Evol Syst 2:185–209.
Goulson D, Wright NP (1998) Flower constancy in the hoverflies
Episyrphus balteatus (Degeer) and Syrphus ribesii (L.) (Syrphidae).
Behav Ecol 9:213–19.
Gumbert A (2000) Color choices by bumble bees (Bombus terrestris):
innate preferences and generalization after learning. Behav Ecol
Sociobiol 48:36–43.
Chittka L (1999) Bees, white flowers, and the color hexagon – a reassessment? No, not yet. Naturwissenschaften 86:595–97.
Hempel de Ibarra N, Giurfa M, et al. (2002) Discrimination of coloured
patterns by honeybees through chromatic and achromatic cues. J
Comp Physiol A Sens Neural Behav Physiol 188:503–12.
Chittka L, Menzel R (1992) The evolutionary adaptation of flower
colours and the insect pollinators’ colour vision. J Comp Physiol A
171:171–81.
Hiei K, Suzuki K (2001) Visitation frequency of Melampyrum roseum
var. japonicum (Scrophulariaceae) by three bumblebee species and
its relation to pollination efficiency. Can J Bot 79:1167–74.
Binkenstein et al. | Weak correlation of flower color and nectar-tube depth 405
Ibanez S (2012) Optimizing size thresholds in a plant-pollinator interaction web: towards a mechanistic understanding of ecological
networks. Oecologia 170:233–42.
Peitsch D, Fietz A, Hertel H, et al. (1992) The spectral input systems
of hymenopteran insects and their receptor-based colour vision. J
Comp Physiol A 170:23–40.
Inouye DW (1980) The effects of proboscis and corolla tube lengths
on patterns and rates of flower visitation by bumblebees. Oecologia
45:197–201.
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.
Irwin RE, Strauss SY, Storz S, et al. (2003) The role of herbivores in
the maintenance of a flower color polymorphism in wild radish.
Ecology 84:1733–43.
R Development Core Team (2013) R: A Language and Environment for
Statistical Computing. Vienna, Austria: R Foundation for Statistical
Computing.
Jager E (2009) Rothmaler - Exkursionsflora von Deutschland, Bd. 3:
Gefäßpflanzen. Atlasband, 11th ed. Heidelberg, Germany: Spektrum
Akademischer Verlag.
Riveros AJ, Gronenberg W (2012) Decision-making and associative
color learning in harnessed bumblebees (Bombus impatiens). Anim
Cogn 15:1183–93.
Junker RR, Blüthgen N, Brehm T, et al. (2013) Specialization on traits
as basis for the niche-breadth of flower visitors and as structuring
mechanism of ecological networks. Funct Ecol 27:329–41.
Rohde K, Papiorek S, Lunau K (2012) Bumblebees (Bombus terrestris) and honeybees (Apis mellifera) prefer similar colours
of higher spectral purity over trained colours. J Comp Physiol A
199:197–210.
Land M, Chittka L (2013) Vision. In Simpson SJ, Douglas AE (eds).
The Insects: Structure and Function. Cambridge: Cambridge University
Press, 708–37.
Sargent RD (2004) Floral symmetry affects speciation rates in angiosperms. Proc Roy Soc B Biol Sci 271:603–8.
Leonard AS, Dornhaus A, Papaj DR (2011) Flowers help bees cope
with uncertainty: signal detection and the function of floral complexity. J Exp Biol 214:113–21.
Shrestha M, Dyer AG, Boyd-Gerny S, et al. (2013) Shades of red: birdpollinated flowers target the specific colour discrimination abilities
of avian vision. New Phytol 198:301–10.
Li Q, Racine JS (2003) Nonparametric estimation of distributions with
categorical and continuous data. J Multivariate Anal 86:266–92.
Shrestha M, Dyer AG, Bhattarai P, Burd M (2014) Flower colour and
phylogeny along an altitudinal gradient in the Himalayas of Nepal
(R Shefferson, Ed). J Ecol 102:126–35.
Lunau K, Wacht S, Chittka L (1996) Colour choices of naive bumble
bees and their implications for colour perception. J Comp Physiol A
178:477–89.
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 68:81–120.
Nagano Y, Abe K, Kitazawa T, et al. (2014) Changes in pollinator fauna
affect altitudinal variation of floral size in a bumblebee-pollinated
herb. Ecol Evol 4:3395–407.
Niggebrügge C, Hempel de Ibarra N (2003) Colour-dependent target detection by bees. J Comp Physiol A Sensory Neural Behav Physiol
189:915–8.
Nilsson L (1988) The evolution of flowers with deep corolla tubes.
Nature 334:147–9.
Papiorek S, Rohde K, Lunau K (2013) Bees’ subtle colour preferences: how bees respond to small changes in pigment concentration. Naturwissenschaften 100:633–43.
Pauw A, Stofberg J, Waterman RJ (2009) Flies and flowers in
Darwin’s race. Evolution 63:268–79.
Simpson B, Neff J (1983) Evolution and diversity of floral rewards. In
Jones C, Little R (eds). Handbook of Experimental Pollination Ecology.
New York, NY: Van Nostrand Reinhold, 277–93.
Spaethe J, Tautz J, Chittka L (2001) Visual constraints in foraging
bumblebees: flower size and color affect search time and flight
behavior. Proc Natl Acad Sci USA 98:3898–903.
Stang M, Klinkhamer PGL, Meijden E (2007) Asymmetric specialization and extinction risk in plant–flower visitor webs: a matter of
morphology or abundance? Oecologia 151:442–53.
Willmer P (2011) Pollination and Floral Ecology. Princeton, NJ:
Princeton University Press.
Willmer P, Stanley DA, Steijven K, et al. (2009) Bidirectional flower
color and shape changes allow a second opportunity for pollination. Curr Biol 19:919–23.
Young HJ (2002) Diurnal and nocturnal pollination of Silene alba
(Caryphyllaceae). Am J Bot 89:433–40.
Zeileis A, Hothorn T (2002) Diagnostic checking in regression relationships. R News 2:7–10.