香京julia种子在线播放

All measurements were conducted in the Hohe Tauern National Park located in the main chain of the Eastern Alps in Austria. All sampling sites were located in the subalpine to alpine range between 1,400 and 2,600 m a.s.l. The sites were scattered roughly along the Großglockner panoramic road in the National Park. Experiments were carried out in the years 1994–2020 (nectar measurements: 1994–2020; spectral reflectance measurements: 2006–2020). Research and sampling permits for sites within the National Park core region were issued by the “Land Salzburg” (permit no. 21 301-RI/547/161-2010, to JN).

Plants were either identified in the field or taken to the lab for species-level identification using adequate identification keys. For consistency, all identifications followed the nomenclature by Fischer et al. (2008). Some species could only be identified to the level of a species-group of closely related species, which is indicated by the suffix “agg.” (for aggregate; e.g., Saxifraga oppositifolia agg., Thymus praecox agg.). All sampled species are listed in Supplementary Table 1.

To include only those plant species that are actually used by bumble bees as a nectar source, we used a database containing museum specimens, observations and literature records pertaining to the majority of field observations made on bumble bees in Austria involving over 42,000 flower visitation records collected between 1848 and 2021. A minimum criterion of five databased observations of bumble bee visits for a certain plant species was set, to include the species in the analysis. For high altitude alpine plants for which the observation density is generally lower (e.g., Phyteuma globulariifolium, Saxifraga spp., alpine Salix spp.), we accepted a threshold of three databased records (in N = 5 plant species). All plant species that failed to meet these criteria were removed from the analysis.

Nectar measurements were performed as a series of five measurements during consecutive 2-h intervals, covering the entire day between 7 am and 5 pm CET (7–9 am, 9–11 am, 11 am–1 pm, 1–3 pm, and 3–5 pm). Measurements were only performed on days without precipitation, and with a mean relative humidity below 80% during the day (11 am–4 pm). Since weather conditions are fairly unpredictable in the alpine area, some measurement series were aborted and missing time intervals were completed on the next appropriate day.

For each series, the nectar volume and concentration were measured in at least 10 flowers (5 in rare species), with a maximum of three different flowers measured per individual plant and time interval. Flowers were randomly selected from the plants/inflorescences. In each measurement series, individuals were investigated from the same population. We then calculated the mean reward quantity for each species and time interval. Two different values were used in the analysis. First, we selected the maximum value across the time series. This value relates to the maximum nectar standing crop of flowers that were shielded from visitors for up to 24 h (in the bagged condition), and thus represents a value that is comparable to previous studies (Shrestha et al., 2020). Secondly, we selected the mean value across the time series, which serves as a more realistic standing crop that would, on average, be available for visitors throughout the day.

Open flowers are usually depleted by nectar-seeking visitors. To estimate both the intrinsic nectar production of a flower and the nectar standing crop that is actually available for visitors, we recorded series of measurements from flowers, which were covered to exclude visitors, and from unprotected (“open”) flowers. For the intrinsic nectar production, a white plastic mesh was placed around the flowers and inflorescences in the early morning, before the first flower visitors were active (referred to as “bagged” in the text). A mesh width of 0.8–1.5 mm allowed sufficient airflow to minimize the microclimatic influence of solar radiation, extreme temperatures and humidity, which are known to affect nectar amount and concentration, but excluded all potential visitors (Corbet et al., 1979; Kearns and Inouye, 1993). Nectar measurement series on bagged flowers were carried out in parallel to series on open flowers at the same site and day.

Nectar was extracted from flowers following the protocol of Corbet (2003), with microcapillaries of 5.0 μl (Assistant, Germany), 1.0 μl, 0.5 μl, or 0.2 μl (Drummond, Great Britain) volume. For flowers with several separated nectaries, the nectar from all nectaries was extracted. The length of the liquid column was measured to the nearest 0.5 mm and converted to μl volume. The contained liquid was then expelled onto a hand refractometer suitable for small volumes, and calibrated for concentrations between 0 and 50% sucrose equivalent (Bellingham and Stanley, Great Britain). When a higher nectar concentration was anticipated, the nectar was diluted by dipping the microcapillary into distilled water (Aqua dest.) before the measurement (Corbet, 2003). The nectar volume was then determined before and after dilution, and the ratio was used to calculate the original sugar concentration. For flowers containing very small nectar volumes (<0.05 μl), the nectar was generally diluted with Aqua dest. to allow measurement with the refractometer. Conversion of the volumes was performed as described above. In the first years of the study (1994–2000), measurements on flowers with very small volumes (<0.1 μl) were conducted by pooling the nectar from several flowers. The total volume was then divided by the number of flowers used to reach the threshold volume for measurement with the refractometer. For extremely small nectar volumes (<0.02 μl) concentration measurements failed in a few cases. We then used the average concentration from the other flowers in the same 2 h interval to determine the sugar content of the flower. In open flowers, nectar can concentrate during hot and dry weather conditions. When concentrations exceeded 80%, measurements were excluded from the analysis, since bees do not visit flowers with highly concentrated nectar (own unpublished observations; see also Harder, 1986).

Since both nectar volume and nectar concentration are strongly influenced by ambient conditions, we calculated the amount of sugar in the nectar (Bertsch, 1983). To calculate the total amount of reward per flower, our measurements were converted to mg sucrose equivalent/flower using standard procedures (Cruden and Hermann, 1983; Kearns and Inouye, 1993; Corbet, 2003). In brief, the measured nectar volume was multiplied by the measured sugar concentration. The sugar concentration measured with the refractometer (calibrated for weight:weight concentration) was corrected to the appropriate unit (weight:volume concentration) using a correction factor according to Cruden and Hermann (1983).

Since flowers often form units, e.g., inflorescences, flower baskets or synflorescences consisting of individual inflorescences/baskets (Fischer et al., 2008), we also aimed to quantify the nectar reward of the entire flowering units. We first defined a “flowering unit” from a visual perspective. Each entity that constitutes a separate visual cue during approach was treated as a single unit. Parts of inflorescences which stand far apart, form no continuous unit and which force visitors to fly between, were treated as separate units (e.g., Aconitum degenii and Adenostyles alliariae). For each plant species in our dataset, we counted the number of individual flowers per unit in at least 10 plant individuals and calculated the mean.

Flowers and inflorescences were collected and brought to the lab for spectral measurements. Spectral measurements were performed by measuring a small (c. 5 mm²) area of a given plant organ that was mounted on black insulation tape. All measurements were performed with either a USB2000 spectrometer equipped with a DH2000 BAL light source, or a JAZ spectrometer unit equipped with a pulsed Xenon light source (Ocean Optics, Dunedin, FL, United States). The spectrometers were calibrated against a white standard (WS-1-SL, Ocean Optics). Measurements were performed with a bifurcated fiber optics probe, with the incident and measuring angle set at 45° with respect to the surface normal, following standard protocols (Chittka and Kevan, 2005). A single measurement was performed on each plant individual. For single flowers or unicolored compound inflorescences (e.g., in Knautia, Scabiosa, and Valeriana) we measured the region of the most prominent flower organ, facing the viewing direction of the visitor (usually the upper surface of a petal). For plants with multicolored flowers or inflorescences, e.g., many Asteraceae, we only measured the part of the inflorescence that occupied the majority (>50%) of the surface (usually the upper surface of the petal lips). Since the flowers of dioecious species (Salix sp.) differ in their reward and appearance and are found on different plant individuals, they were treated as independent data points in our analysis. In Trifolium pratense, two distinct color morphs with a general form appearing pink for a human observer, and an alpine form appearing white, were measured and treated as independent data points. In both cases, nectar and color measurements were also performed separately. Several specimens (between 1 and 127, median 3) were measured per species, depending on availability of the flowers.

To estimate how bee visitors perceive the flower color, we used the color hexagon (Chittka, 1992), a bee-specific color space that is widely employed in pollinator studies and has been repeatedly tested in laboratory settings (Chittka et al., 1992; Giurfa et al., 1995; Raine and Chittka, 2005; Théry et al., 2005; Dyer et al., 2008, 2012; Leonard et al., 2011). Color loci were calculated according to standard procedures (Chittka and Kevan, 2005) using standard illumination (D65; Wyszecki and Stiles, 1982) and photoreceptor spectral sensitivity functions specific for Bombus terrestris (Skorupski et al., 2007). Hymenopteran photoreceptor sensitivities are phylogenetically conserved and similar among bee species (Briscoe and Chittka, 2001) and were confirmed to be similar across bumble bee species in particular (Peitsch et al., 1992; Briscoe and Chittka, 2001; Skorupski et al., 2007; Skorupski and Chittka, 2010). We used an average reflection spectrum of green foliage as adaptation background (Chittka and Kevan, 2005). For each measurement, we determined the (absolute) green receptor contrast, the position of the locus in the color space, and the color contrast as the Euclidean distance between the hexagon center and the color locus (Spaethe et al., 2001; Chittka and Kevan, 2005). Brightness, considered as the summed response of all three photoreceptors, is used by bees only during phototactic response (Menzel and Greggers, 1985) and is not regarded as an important spectral feature during foraging (Ng et al., 2018).

Color hue refers to the direction of a locus in the color space and was calculated as the angle between the lines connecting the hexagon center with the blue corner (set as 0°) and the color locus, respectively. Color locus angles are reported as positive values in the clockwise direction with respect to the reference line (see Figure 1A). However, it must be noted that although several studies used angles as a measure for hue, there are no universally accepted standards on how to report it. Different studies used different reference lines and rotation directions (Chittka et al., 1994; Dyer et al., 2012; Shrestha et al., 2014 vs. Tai et al., 2020). For further analysis, the hues (angles) were binned to six categories, which correspond to distinct classes of reflectance functions (Chittka et al., 1994). The categories are referred to as “blue” (B; 330°–30° in the hexagon space), “blue-green” (BG; 30°–90°), “green” (G; 90°–150°), “green-UV” (GU; 150°–210°), “ultraviolet” (UV; 210°–270°), and “UV-blue” (UB; 270°–330°). In addition, we also plotted the hues to a finer scale of 10° bins to make the data comparable to other studies, which used this bin size (e.g., Chittka et al., 1994; Dyer et al., 2012; Shrestha et al., 2014). Flower species with color contrasts < 0.1 hexagon units were assigned as “achromatic” and excluded from further analysis (N = 3 species, see below).

To test for phylogenetic signal in the color and nectar reward data, we constructed a species-level phylogenetic tree, including all of our studied taxa. We initially used the phylogenetic tree (“ALLOTB” tree) published by Smith and Brown (2018). Tree manipulation was performed in R (version 4.1.1; R Development Core Team, 2021) using the packages “phytools” for R (Version 0.7-80; Revell, 2012), “ape” for R (Version 5.5; Paradis and Schliep, 2019), and “picante” for R (Version 1.8.2; Kembel et al., 2010). The tree was pruned to include only those taxa contained in our dataset. For two taxa which include distinct color morphs (Trifolium pratense) or different sexes (Salix waldsteiniana) of a single species, we introduced a dichotomy with branch length zero. Multitomies in the tree were resolved using the function “multi2di” in the “ape” package. The final tree used in our analysis can be found in the Supplementary Material.

To test whether color traits, number of flowers/inflorescence or nectar reward of the study species show phylogenetic signal, we calculated Pagel’s λ (Pagel, 1999). For continuous traits, λ calculation and significance tests were performed in the “phytools” package. Since the distribution of nectar reward quantities (i.e., sucrose equivalents) was significantly different from a normal distribution (p < 0.05; Shapiro-Wilks test), all values were log10-transformed for the analysis. For color category, λ was calculated using the “fitDiscrete” function in the “geiger” package. Significant difference between the fitted model and the null model (λ = 0.00; no phylogenetic signal) was tested using a log-likelihood ratio test.

To test whether plant species are uniformly distributed among the color categories, we used a Chi-square test, followed by an analysis of the standardized residuals (Sharpe, 2015). To test whether nectar reward quantity differed significantly between color categories, we performed phylogenetic ANOVA using the package “geiger” for R (Version 2.0.7; Pennell et al., 2014), using log10-transformed nectar values as dependent and hexagon color category as independent variable. Our dataset contained a single species in the “UV”-category (Crepis aurea), which was removed prior to the ANOVA. Independent analyses were performed for the maximum and mean nectar reward and for the “open” and “bagged” treatments. Significant results in the omnibus test were followed by a post-hoc test, comparing all possible combinations and adjusting the p-level using the Bonferroni method.

Whether nectar reward quantity differed significantly between “open” and “bagged” flowers was tested using a paired t-test. To test the relationship between nectar reward quantity and visual and other traits we used phylogenetic generalized linear mixed models (PGLMM) with a Gaussian distribution. The model included log10-transformed nectar reward data as response variable, species (both as phylogenetic and non-phylogenetic covariate) as random factor, and color contrast, green contrast, brightness and the number of flowers per inflorescence as continuous covariates. All continuous variables were scaled prior to model preparation to facilitate interpretation of the effect sizes (Schielzeth, 2010). Separate models were calculated for each combination of nectar data (maximum & mean) and treatment (open and bagged). Model calculation was performed using the “pglmm” function in the “phyr” package for R (Version 1.1.0.; Li et al., 2020). All analyses were performed using the base version of R (Version 4.1.1; R Development Core Team, 2021) and the cited packages.

We obtained full nectar (both open and bagged flowers) data and spectral measurements from 108 samples. Three species with color contrasts < 0.1 hexagon units were excluded from further analysis (Pedicularis recutita, Salix hastata female, and Vaccinium myrtillus). The remaining 105 samples constituted 103 unique species with two additional samples from a second color morph (Trifolium pratense) and a second sex of a dioecious species (Salix waldsteiniana).

Of the 112 flowering plant species for which bumble bee visits have been recorded in the study region (local communities of Rauris, Fusch and Heiligenblut above 1,400 m a.s.l.), the analyzed sample comprises 103 species, which accounted for the vast majority (98 %) of all recorded bumble bee visits (N = 4,070) in that region.

The mean color loci of the samples were not distributed uniformly in the color space (Figure 1A), and the distribution of plant colors among the color categories differed significantly from a uniform distribution (N = 105, Chi² = 125.46, p < 0.05). The majority of samples were found in the “blue-green” category (N = 54), followed by the “blue” category (N = 30; Figure 1B). Analysis of the standardized residuals indicated that the observed frequencies in these two categories are significantly higher than expected by chance, while the categories “UV”, “UV-blue,” and “green-UV” had significantly less observations than expected. Color frequency distribution, when analyzed at a finer scale, showed a pronounced peak at 60°, which corresponds to the central part of the “blue-green” sector in the hexagon (Figure 1). Phylogenetic signal for flower color, calculated as Pagel’s λ was estimated to be λ = 0.87, a value significantly different from λ = 0.00 (p < 0.05; Table 1 and Figure 2).

Nectar reward quantity (expressed in mg sucrose per flower) varied among species and treatment (Supplementary Table 1). Bagging had a significant effect on the measured reward quantity; flowers shielded from visitors had a significantly higher reward quantity than flowers which could be depleted by visitors (maximum nectar reward: t(104) = 5.57, p < 0.05; mean nectar reward: t(104) = 6.88, p < 0.05). Pagel’s λ for the reward per flowers showed significant phylogenetic signal both in the mean and maximum value and in the “bagged” and “open” treatments (Table 1).

Reward quantity of bagged flowers differed significantly between the color categories for the maximum nectar reward values [Phylogenetic ANOVA: F(4,99) = 3.51, p < 0.05; Figure 3A], but just failed significance for the mean nectar reward values [F(4,99) = 2.94, p = 0.07; Supplementary Figure 1A]. While the individual flowers in the “blue” category had, on average, higher reward quantities than in the other categories, pair-wise post hoc comparison did not identify significant differences between any of the combinations after Bonferroni correction.

Reward quantities were generally smaller in open flowers. Their distribution across color categories did not differ significantly among categories for the maximum nectar reward values [F(4,99) = 2.18, p = 0.20; Figure 3B] and for the mean nectar reward values [F(4,99) = 2.26, p = 0.18; Supplementary Figure 1B]. When we extrapolated the nectar reward quantity to the entire functional unit (inflorescence), quantities did not differ significantly for both maximum [bagged: F(4,99) = 3.37, p = 0.32; open: F(4,99) = 1.84, p = 0.56; Figure 4] and mean [bagged: F(4,99) = 2.81, p = 0.49; open: F(4,99) = 1.97, p = 0.59; Supplementary Figure 2] quantity measures.

Phylogenetic generalized linear mixed models for maximum (Tables 2, 3) and mean (Supplementary Tables 2, 3) nectar reward quantities identified a strong influence of species, followed by a smaller effect of the phylogeny-corrected species term. In the fixed effects, we identified equally strong negative effects of color contrast and the number of flowers per inflorescence, as well as small, non-significant effects of all other tested variables (Tables 2, 3 and Supplementary Tables 2, 3). In other words, nectar reward quantity correlated negatively with color contrast (i.e., flowers with higher color contrast contained less nectar) and flower number (i.e., inflorescences with fewer flowers had more nectar per flower).

Our analysis showed that the color loci of bumble bee visited flowers were scattered throughout the color hexagon, resulting in a variety of hues (angles) and chromatic contrasts to the background (distance to hexagon center). Interestingly, only three species appeared achromatic to bees. Achromatic cues are difficult to detect under natural conditions, and bees may not utilize them for flower detection and identification (Ng et al., 2018). The number of flowers found in each of the major bee-color categories (sensu Chittka et al., 1994) differed significantly from a uniform distribution, with the majority falling into the “blue-green” sector. This pattern, as well as that obtained when analyzed at a finer resolution (Figure 1C), showed remarkable similarity with data from other habitats and locations, e.g., Germany (Giurfa et al., 1995), Australia (Dyer et al., 2012), Nepal (Shrestha et al., 2014), New Zealand (Bischoff et al., 2013), and Taiwan (Tai et al., 2020). Interpretation and comparison of the distribution is problematic as it may depend on sampling strategy, habitat type, pollinator species and the choice of visual system selected for the color modeling (for a discussion, see Shrestha et al., 2019). For instance, most of the above-mentioned studies either combined samples from large regions rather than local communities and/or did not consider the pollinator composition.

Both abiotic factors and biotic factors are assumed to influence the flower color distribution. Previous studies demonstrated influences of e.g., day length and precipitation (Arista et al., 2013), soil composition (Horovitz, 1976), and vacuole pH (Grotewold, 2006). For the alpine environment, selection pressures for adaptations to cope with temperature extremes and high irradiance are likely to influence the observed color frequencies (van der Kooi et al., 2019; Dalrymple et al., 2020).

For biotic selection pressures, color frequency differences have been hypothesized as resulting from selection by different pollinator assemblages, but (experimental) proof for this hypothesis is rare. Recently, data from regions that lack bees (Maquarie Island; Shrestha et al., 2016) or social bees (New Zealand; Ishii et al., 2019), show different plant color distributions and thus provide support for this hypothesis. For regions with highly overlapping visitor spectra, like the Alps, comparing color frequencies as function of pollinator group is more complex. Plant-visitor networks showed considerable overlap of visitor groups for most investigated plant species in this ecosystem (Lefebvre et al., 2018). While it can be assumed that many of the flower visitors also serve as pollinators, experimental proof of the actual pollinator identity and its share in the overall pollination of most generalist plant species is largely lacking. These bits of information are, however, crucial in understanding pollinator-mediated selection on traits like e.g., color, since the strength of selection can be assumed to critically depend on the pollination efficiency of the different pollinators of a plant species (Trunschke et al., 2021). To better understand the origin of the color frequency distribution that we observed in our study, we need further detailed information about the base line in the entire community (i.e., spectral reflectance data from all of the c. 400 flowering plant species that occur in the region) and quality information about the predominant pollinator(s) for each of them, which will be a challenge for future generations of pollination ecologists.

Previous studies, which attempted to link flower color with reward quantity, either used literature data only (Giurfa et al., 1995), extrapolated nectar production rates from short measurement sequences to the entire day (Chittka et al., 2004; Raine and Chittka, 2007a), or measured sugar content of the nectar standing crop of bagged flowers only (Shrestha et al., 2020). In our study, we measured the nectar production capacity of the species (bagged flowers) as well as the nectar standing crop of open flowers, as a more direct measure of what is actually available during a typical day. As expected, the nectar standing crop was lower than the production capacity, due to depletion by visitors. Nectar reward quantities were more similar between color categories in open flowers, suggesting that those flowers that produce more nectar are preferentially depleted in the field under normal conditions. It is unclear whether these higher visitation rates originate initially from random or targeted visits of flower visitors, which have learned that certain flowers are more rewarding than others (Goulson et al., 2007).

Visual signals are used by flowering plants to convey information about their species identity and allow for easier detection in the usually cluttered visual environment. They can hold information about the reward or promote learning of the association between floral traits and the reward. Flower spectral reflectance is a complex mixture of different qualities that can be employed by the bee visual system separately or in combination. These qualities involve color contrast (contrast between the background and the flower color), achromatic contrast (modulation of the green receptor channel), brightness (the sum of the three photoreceptor excitations) and color hue. Aside from brightness, which is sometimes used in bee vision studies but has not been shown to be of importance for bees (Spaethe et al., 2001; Ng et al., 2018) all other signals and cues have been found to be relevant in bee foraging. Color contrast correlates with detection speed (Spaethe et al., 2001; Streinzer et al., 2009) and bees are known to prefer flowers of higher contrast when given a choice (Rohde et al., 2013). Achromatic contrast is used in object detection (Giurfa et al., 1996; Dyer et al., 2008), but is probably not or only rarely used as a sole cue in flower detection (Martínez-Harms et al., 2010; Lunau et al., 2011; Ng et al., 2018). Finally, hue is employed by bees to identify different reflectance spectra independent of lighting conditions and other visual traits (Reser et al., 2012). Bumble bees can learn to discriminate very small differences in hue when trained appropriately (Dyer and Chittka, 2004), but in the real world, such fine discrimination ability is probably of little value, given the existing variation of flower color within species, which sometimes overlaps with color of other species (Jersáková et al., 2016; Garcia et al., 2020).

Flower visiting insect have been shown to have (species-specific) innate preferences for certain colors (Lunau and Maier, 1995) which have been interpreted to help them find rewarding flowers more quickly during their first foraging flights. In a field study with Bombus terrestris, Raine and Chittka (2007b) first determined the strength of the innate preference for “UV-blue” and then let them forage in the surrounding, where UV-blue flowers were also the most rewarding ones. They found a significantly higher colony-level success of colonies that showed a strong innate color preference, indicating that these preferences may be adaptive if color correlates with reward quantity. Due to the large variation of rewards in our study, color cannot be considered as a reliable signal for nectar reward quantity. After sampling the vast majority of plant species visited by bumble bees in our study region, we, however, found that flowers with colors from the preferred color categories do have, on average, higher reward production at the single flower level (Figure 3A). Smaller differences have also been found for open flowers (Figure 3B) and when comparing entire inflorescences (Figure 4), though the differences were statistically not significant. We thus conclude that color may be a weak (but honest) indicator for reward quantity, and that this overall (small) advantage may indeed allow the bees to increase foraging success, compared with an entirely random search. Similar correlations between the innately preferred color categories of bees and the reward quantity were also found in Central Europe (Giurfa et al., 1995; Raine and Chittka, 2005), but not e.g., in Australia (Shrestha et al., 2020). While in the European region, social bees are assumed to have a large share in the overall pollination of entomophilous plants, in the Australian communities, (social) bees are not the major pollinator guild. Furthermore, pollinator/visitor identity was not investigated in that study, which limits the comparison with our results.

For a complete understanding of how strongly innate color preferences affect flower color in the Alps, we must know the relative contribution of all flower visitors of a plant species to its pollination success, and to analyze in detail whether different nectar traits (like volume, concentration, sugar content, and sugar composition) differ between major visitor groups and color categories. Our study surprisingly showed a negative correlation between the color contrast and nectar reward, which seems to stand in contrast to the observation that bumble bees prefer flowers of high color contrast (Rohde et al., 2013). However, some previous studies found contrasting results regarding the relationship between color contrast and reward quantity (Kantsa et al., 2017; Shrestha et al., 2020). Color contrast is a highly variable trait in flower communities (Garcia et al., 2021) and it is currently not known how bees use this visual feature while foraging in natural environments.

Interestingly, we found no statistical difference of nectar rewards among color categories when we calculated the nectar reward of the entire inflorescence (Figure 4 and Supplementary Figure 2). Flower and floral display size have been shown to correlate with nectar reward quantity and may constitute an honest signal of reward quantity for potential visitors (Ortiz et al., 2021). In our study, we found a significant negative relationship between nectar reward quantity of individual flowers and flower number of an inflorescence. From a plant’s perspective, grouping several flowers with smaller reward quantity to larger units would constitute a strategy to attract more potential pollinators due to a larger display size (Spaethe et al., 2001; Wertlen et al., 2008) and thus promote learning through higher reward quantities that can be gathered during a single visit. While larger inflorescences may provide a larger total amount of reward, the energy and time needed to collect these rewards must also be considered in foraging economics (Harder et al., 2001). In future studies, one will need to investigate in more detail how the interplay between nectar reward of individual flowers, variation of inflorescence size (e.g., number of flowers) and spatial distribution of plant individuals within a population affect the foraging economics and color preferences of bumble bees (Geslin et al., 2014).