TY - JOUR
AU1 - Garcia, Jair, E
AU2 - Shrestha,, Mani
AU3 - Dyer, Adrian, G
AB - Abstract Color discrimination thresholds proposed by receptor-noise type models are frequently used in animal vision studies to predict a precise limit on the capacity of an animal to discriminate between stimuli. Honeybees and bumblebees are 2 closely related hymenopteran species for which precise data on photoreceptor sensitivities and receptor noise exist, enabling accurate testing on how their vision conforms to model predictions. Color vision has been proved in these species, and they are known to predominantly visit flowers using visual signals to collect nutrition. Surprisingly, however, the natural variability of flower signals has been rarely considered, and recent work also suggests bees may tune color vision through experience. We initially measured the spectral variability of flowers from 2 species: Goodenia ovata and Rosemarinus officinalis where free-flying honeybees were observed constantly foraging from conspecific flowers. We empirically determined honeybee color discrimination thresholds for color stimuli considering either absolute- or differential-conditioning discrimination functions. Secondly, we analyzed greenhouse grown wild-type Antirrhinum majus flower petal spectra as well as spectra from mixta and nivea strains of this species, and empirically determined bumblebee color discrimination considering conditioning experience. In all measured cases, within-flower type spectral variability exceeded a 1.0 Receptor Noise threshold, often by several units. Observed behavioral color discrimination functions considering the respective conditioning procedures closely matched the range of signal variability for both honeybees and bumblebees, showing that color vision in bees cannot be described by a single fixed value, and plasticity is a key component of bee foraging behavior in natural environments. INTRODUCTION Color vision is a key perception that many diurnal animals successfully use in complex, natural environments (Endler 1990; Goldsmith 1990; Kemp et al. 2015). In recent times there has been an increased interest in understanding the diversity of color visual systems in different animals (Siddiqi et al. 2004; Cortesi and Cheney 2010; Igic et al. 2012; Schultz and Fincke 2013; Barry et al. 2015; Fleishman et al. 2016; White et al. 2017), and how it may be possible to interpret what color information a particular species may perceive (Vorobyev and Osorio 1998; Garcia et al. 2017; Olsson et al. 2017; Renoult et al. 2017). Many of the models currently used to explain animal color vision are rooted in colorimetric principles developed to explain human color vision (Wyszecki and Stiles 1982), especially the concept of just-noticeable-difference between stimuli (Schrödinger 1926; MacAdam 1942). The link, however, between how these early models can be applied to explain the different aspects of animal color vision in natural scenarios has rarely been robustly tested. Such a difficulty arises from the inherent complexity to completely understand the relationship between an individual species color perception, and what visual information constitutes important signals for that species. Proof of true color vision requires knowledge of different physiological aspects involved in color vision such as the number and spectral spacing color photoreceptors (Jacobs 2018), the presence of neural processing networks to enable color perception (Kien and Menzel 1977; Yang et al. 2004; Dyer et al. 2011), and behavioral evidence that an animal can use such sensory apparatus to make decisions for biologically relevant stimuli (Kelber et al. 2003). Such complete data exist for very few animal species (Kemp et al. 2015; Olsson et al. 2017); and thus, it has become common practice to “borrow” data from different species to potentially facilitate interpretations of likely perceived color differences for the wide range of species for which color perception is considered (Siddiqi et al. 2004; Martínez-Harms et al. 2012). One potential physiological mechanism that may contribute to the visual perception of stimuli by animals is the limit in discrimination imposed by noise in the photoreceptors (Barlow 1956; Aho et al. 1988; Schnapf et al. 1990; Donner 1992). Vorobyev and Osorio (1998) produced a model allowing for the predictions of how receptor noise (RN) could potentially limit animal color discrimination. This color vision theory assumes that the amplitude of photoreceptor noise is the dominant mechanism that mediates colors choices, with little or no influence of post-receptor neural processing (Vorobyev and Osorio 1998; Vorobyev et al. 2001; Hempel de Ibarra et al. 2014). This RN threshold model was initially calibrated using the honeybee as a key species for threshold perception of color (Vorobyev et al. 2001) since precise measurements of physiological aspects involved in color vision such as photoreceptor sensitivities (Peitsch et al. 1992) and noise data (Howard et al. 1987; Vorobyev et al. 2001) are available; and, empirical wavelength discrimination functions have been measured from free-flying individuals (von Helversen 1972). The physiological assumptions of RN would require that external environmental factors like variation in signal spectral content are minimal, or at least below the perceptual threshold predicted by the RN model (Vorobyev and Osorio 1998). Surprisingly, given the widespread usage of the RN model, no study that we are aware of has robustly validated whether inherent fluctuations in biologically relevant color signals are below the perceptual threshold proposed by the RN model. Flowers, for example, are subject to color variability as there are limitations on how pigments can be produced (Rausher 2008; Hopkins and Rausher 2011; Hopkins and Rausher 2012; Sobel and Streisfeld 2013) and the physical characteristics of the petal epidermis at micro and nano scales (Kay et al. 1981; van der Kooi et al. 2017; van der Kooi et al. 2018). Plant-pollinator interactions are an interesting case for understanding the complexities of color information processing in animal perception as it implies recognizing the role of color as a signal evolved for visual communication. Bees have a phylogenetically ancient trichromatic visual system that precedes the evolution of angiosperms (Briscoe and Chittka 2001), and flowering plants have evolved spectral signals to suit the most effective pollinators (Chittka and Menzel 1992; Dyer et al. 2012; Shrestha et al. 2013, 2014; Shrestha et al. 2016). Indeed, flowers produce distinct color signals for bee pollinators (Shrestha et al. 2013, 2016; van der Kooi et al. 2015; van der Kooi et al. 2016; van der Kooi et al. 2018), enabling discrimination and detection of species (Chittka and Menzel 1992; Bukovac et al. 2016) to best promote flower constancy (Chittka et al. 1999). If color variability were higher than the variability produced by noise at photoreceptor level, RN model predicts that a bee pollinator could very likely misclassify, as different, 2 flowers of the same species. This would reduce flower constancy by the visiting pollinator, and such modeling would lead to very different animal behavior than what is observed (Figure 1, Supplementary Video V1). Interestingly, recent psychophysics and neurophysiological studies on honeybees and bumblebees reveal some evidence of plasticity for color learning, suggesting that neural coding may also be a potentially important factor mediating how animals make color decisions in complex natural environments (Giurfa 2004; Dyer et al. 2011; Sommerlandt et al. 2016; Li et al. 2017). More specifically, color discrimination behavioral experiments with bumblebees (Dyer and Chittka 2004) and honeybees (Giurfa 2004), have demonstrated that the conditioning method using during training in color has a significant effect on performance: bees trained under differential conditioning, where bees learn to associate a stimulus with a reward in the presence of a not rewarding alternative, can discriminate smaller color differences than when trained under absolute conditioning, where bees learn to discriminate a rewarding stimuli in the absence of a distractor. Therefore, individual bees are able to discriminate different color differences depending on context (Avarguès-Weber and Giurfa 2014). Figure 1 View largeDownload slide Flying path of 3 honeybees visiting a flower patch of Rosemarinus officinalis for 14 s. Colored circles indicate the flower visited by each bee, and numbers indicate the respective visit sequence. Complete clip as Supplementary Video V1. Figure 1 View largeDownload slide Flying path of 3 honeybees visiting a flower patch of Rosemarinus officinalis for 14 s. Colored circles indicate the flower visited by each bee, and numbers indicate the respective visit sequence. Complete clip as Supplementary Video V1. In the current study we consider 2 well-established insect models for color vision, Bombus terrestris and Apis mellifera, to formally test if colors of signals that are known to be treated as equivalent by free-flying bees, are indeed correctly classified by the color discrimination threshold proposed by RN theory. We specifically address the question of what is the within-individual flowers color variability (WIV), and the within-species color variability (WSV) for flower color signals from the same species. We considered 2 species that produce flower colors sequentially visited in a flower constant fashion by A. mellifera: Goodenia ovata and Rosemarinus officinalis (Figure 1, Supplementary Video V1). For B. terrestris, we considered 3 morphs of Antirrhinum majus. A. majus represents an important plant model for addressing the potential variability of signals for bumblebees because single mutant lines produce controlled changes in flower characteristics, while not affecting other flower traits that may confound how pollinators interact with flowers (Glover and Martin 1998; Whibley et al. 2006). This enabled us to also test for between-species color variability (BSV) where we considered the color distances between the different morphs of A. majus for bumblebees, and the color distance between G. ovata and R. officinalis for honeybees. Flower spectral data are analyzed and interpreted in relation to the predicted 1.0 JND unit threshold of what color differences should be discriminated based on RN theory (Vorobyev et al. 2001; Schaefer et al. 2007; Hempel de Ibarra et al. 2014; Barry et al. 2015). Under the null hypothesis of the RN model, differences in appearance resulting from natural color variability within individual flowers and within flowers of the same species should be lower than the 1.0 JND color discrimination threshold. However, recent work highlights the importance of validating any prediction from the RN model with behavioral data to better understand what an animal sees (Garcia et al. 2014; Marshall 2018; Ng et al. 2018; Osorio and Vorobyev 2018; Stuart-Fox 2018; Vasas et al. 2018). We thus empirically measure color discrimination with absolute conditioning for free-flying bumblebees and honeybees, to model and understand how behavioral plasticity in the respective bee species could potentially be used to overcome perceptual noise for flower signals that are known to be pollinated by these species. We interpret these new thresholds in relation to previously measured appetitive-aversive differential conditioning thresholds for simultaneous color discrimination for honeybees and bumblebees (Dyer and Neumeyer 2005; Dyer et al. 2008; Garcia et al. 2017). MATERIALS AND METHODS Plant samples and collection We collected flowers of G. ovata and R. officinalis on early austral spring 2017 at our university garden. Fresh flowers from both species were collected on sunny days where free-flying honeybees were observed visiting sequential flowers in a constant fashion (Figure 1, Supplementary Video V1). Flowers were immediately taken to the lab for measurements. The wild type of A. majus has a purple color for a human observer, whereas the MIXTA gene regulates petal cell shape from conical to flat and results in a pink-colored morph, that is, mixta mutant, that can be discriminated by free-flying bumblebees (Noda et al. 1994; Dyer et al. 2007). The nivea mutation consists of a deletion of the single Antirrhinum gene encoding chalcone synthase preventing the synthesis of purple anthocyanins and flavonoid pigments resulting in a white color for a human observer (Wienand et al. 1982; Glover and Martin 1998). Interestingly, nivea flowers are also of an uncolored appearance to bee pollinators due to the reflection of UV radiation (Kevan et al. 1996; Waser and Chittka 1998; Dyer et al. 2007). Thus, within the Antirrhinum flower model there are purple and pink similarly colored morphs, and a “white” dissimilar morph. A. majus flowers for all 3 lines (wild-type, mixta, nivea) were grown from seed at Cambridge University, Department of Plant Sciences. The generation of self-seed from plants genotyped by Southern blotting as homozygous for the mutant or wild-type alleles of the 2 genes was described previously (Glover and Martin 1998). Plants were grown under greenhouse conditions at 23 °C in 4-inch pots in Levingtons (United Kingdom) M3 compost. During growth period plants received supplemental lighting from 400 Osram (Osram, München, Germany) lamps on a 16:8 h light:dark photoperiod. The plant model thus represents genetically homogenous flower types grown under ideal conditions. Spectrophotometry We recorded spectral reflectance from 300 to 650 nm at 0.7 nm intervals from 2 different points on 5 petals of each 7 flowers for Goodenia, and one point on each one of the 6 petals for 12 separate Rosemarinus flowers. A. majus flowers for all 3 lines (wild, mixta and nivea) were measured recording 3-point samples from each sampled flower. Refer to Table 1 for sampling dimensions for each species and lines. Spectral measurements were recorded with an Ocean Optics USB 2000 spectrometer connected to an Ocean Optics PX-2 xenon pulse lamp by means of a 400 µm optical fiber, following published methodologies (Dyer et al. 2011; Shrestha et al. 2013). Raw spectra used for analyses and their representation in Maxwell triangles are available in Supplementary Material S1. Table 1 Numerical dimensions of the tested color samples Species n nf ns Antirrhinum majus (wild type) 3 16 48 A. majus (mixta type) 3 13 39 A. majus (nivea type) 3 13 39 Goodenia ovata 10 7 70 Rosemarinus officinalis 6 12 72 Species n nf ns Antirrhinum majus (wild type) 3 16 48 A. majus (mixta type) 3 13 39 A. majus (nivea type) 3 13 39 Goodenia ovata 10 7 70 Rosemarinus officinalis 6 12 72 Number of point samples (n), total number of flower sampled (nf) and total number of samples available from flowers of a single species (ns) used to calculate the total number of pair combinations available for measuring color variability within individual flowers (WIV), between flowers of the same species/types (WSV) and between flowers of different species/types (BSV). Refer to the Materials and methods sections for formulae used to obtain the total number of combinations for each sampling level. View Large Table 1 Numerical dimensions of the tested color samples Species n nf ns Antirrhinum majus (wild type) 3 16 48 A. majus (mixta type) 3 13 39 A. majus (nivea type) 3 13 39 Goodenia ovata 10 7 70 Rosemarinus officinalis 6 12 72 Species n nf ns Antirrhinum majus (wild type) 3 16 48 A. majus (mixta type) 3 13 39 A. majus (nivea type) 3 13 39 Goodenia ovata 10 7 70 Rosemarinus officinalis 6 12 72 Number of point samples (n), total number of flower sampled (nf) and total number of samples available from flowers of a single species (ns) used to calculate the total number of pair combinations available for measuring color variability within individual flowers (WIV), between flowers of the same species/types (WSV) and between flowers of different species/types (BSV). Refer to the Materials and methods sections for formulae used to obtain the total number of combinations for each sampling level. View Large Color modeling and data analysis Color variability for a given species was calculated as the color difference between samples within a flower, WIV, and color differences between different flowers of the same plant species, WSV; Color differences between flowers of 2 different species/types, BSV were also calculated as a reference for large color differences. Color differences were calculated using the RN model (Vorobyev and Osorio 1998; Vorobyev et al. 2001) for each one of these signal difference categories, modeling a typical open environment daylight illumination of 6500 K (Wyszecki and Stiles 1982) corrected for photon flux, and assuming an Average Green Leaf (AGL) as the adaptation background (Bukovac et al. 2017). To gauge the magnitude of color variability across the within individual, within species and between species levels, we calculated the color differences for all the possible (nk) pair-sample (k = 2) combinations for each category. The total number of pair samples considered for WIF, WIV, and BSB color variability levels are given by: WIF= {(n1k), (n2k),⋯,(nnfk)}, WIV= (nsk), BSV= nsSP1× nsSP2; where n represents the number of point samples recorded for a single flower, nf is the total number of flowers sampled for a given species/type, ns is the total number of samples available from all the flowers sampled for a given species (ns=n × nf) ⁠, and SP1 and SP2 represent flowers from 2 different species. Values of n, nf, and ns are provided in Table 1. RN calculations were performed using photoreceptor spectral sensitivities for A. mellifera and B. terrestris published by (Peitsch et al. 1992). For modeling Apis, we used the noise values reported by Vorobyev and Osorio (1998). Two receptor-noise models were initially constructed for Bombus: a Bombus-β model using noise parameters as reported by Skorupski and Chittka (2010), and a Bombus-α model which assumes the noise parameters of A. mellifera. Reported results in the main manuscript are based on the calculations for the Bombus-α model since the predictions of this model better represent behavioral data for this species (Dyer et al. 2008; Garcia et al. 2017). Results for the Bombus-β model are available as Supplementary Material S2. Behavioral experiments and discrimination function under absolute conditioning Apis mellifera absolute conditioning training Ten marked honeybee workers were individually trained with absolute conditioning (Giurfa 2004; Sommerlandt et al. 2016) to 3 identical target colors (G0: see Dyer and Neumeyer (2005) for details, and Supplementary Material S3 for spectra and color modeling of the stimuli) presented on a horizontally mounted rotating circular board of 50 cm diameter. Experiments were conducted outdoors at the University of Mainz in August 2007 with the same conditions to previous work on appetitive-aversive differential conditioning for honeybees (Dyer and Neumeyer 2005). Each bee was conditioned for 15 landings with a 25% sucrose solution associated with landing on the G0 stimulus. On the 15th landing each bee was satiated so that it returned to the hive to contribute collected sucrose to the production of honey (von Frisch 1967). Each bee thus off loaded collected nutrition and volitionally returned to the testing site after about 2–3 min to continue the experiments (Dyer 2012). All stimuli and apparatus were cleaned with ethanol, and completely fresh stimuli were used for testing. Each bee received 6 non-rewarded trails with stimuli G0 as the target, and stimuli B15, B12, B9, Y15, Y12, or Y9 as the perceptually similar distractor on a given test. In a test, 3 identical targets, and 3 identical distractors, were presented at random positions on the rotating foraging screen, and the frequency of correct landings on the target stimulus during a total of 15 non-rewarded landings were scored as binomial choices. Each bee received all 6 tests, but test order was randomized per bee. Immediately following a test, each bee was allowed to forage again on stimulus G0 for 15 rewarded landings to maintain motivation for subsequent tests. Bombus terrestris absolute conditioning training Eight individually marked bumblebees were individually trained with absolute conditioning (Dyer and Chittka 2004) to 3 identical target colors (G0: see Dyer and Neumeyer (2005) for details, Supplementary Material S3) presented horizontally at random coordinates in a 70 × 50 × 50 (length, width, height) cm3 flight arena illuminated by four 200 Hz Duro-Test 40 W True-lite tubes diffused with a single sheet of ultraviolet transmitting Rosco 216 diffusion screen. Spectral irradiance for this illumination source closely matches that of natural overcast conditions. Intensity was about 1,400 lux and is well above the threshold illumination (approximately 1 lux) that bumblebees require for active foraging (Kapustjanskij et al. 2007); see Dyer et al. (2008) for details. Each bee was conditioned for 15 landings with a 25% sucrose solution associated with landing on the G0 stimulus. On the 15th landing each bee was satiated so that it returned to the nesting box; testing began when the tested bee returned. All stimuli and apparatus were cleaned with ethanol, and completely fresh stimuli were used for testing. Each bee received 6 non-rewarded trails with stimuli G0 as the target, and stimuli B15, B12, B9, Y15, Y12, or Y9 as the perceptually similar distractor on a given test (Supplementary Material S1). In a test, 3 targets and 3 identical distractors were presented at random positions and the number of landings on respective stimuli were scored as the dependent variable. Each bee received all 6 tests, but test order was randomized per bee. Immediately following a test, each bee was allowed to forage again on stimulus G0 for 15 rewarded landings to maintain motivation for subsequent tests. Modeling color discrimination functions Color discrimination functions under absolute conditioning for A. mellifera and B. terrestris were obtained by fitting a 3 and 4 logistic parameters sigmoidal-type function, respectively, to the behavioral data recorded for the 2 species (Garcia et al. 2017). Sigmoidal curves were fitted using the nonlinear generalized least square routine (gnls) available as part of the nlme package (Pinheiro et al. 2013) for the R language for statistical computing v. 3.4.1. (R Development Core Team 2017). Functions describing color discrimination considering appetitive-aversive differential conditioning were taken from Garcia et al. (2017) modeled after behavioral data from Dyer and Neumeyer (2005) for honeybees and Dyer et al. (2008) for bumblebees. RESULTS Color discrimination thresholds under absolute conditioning The discrimination function considering absolute conditioning for A. mellifera is described by a 3-parameter logistic function, whereas the discrimination function for B. terrestris is best described by a 4-parameter logistic function (Figure 2). Coefficients defining each function are provided in Table 2. Figure 2 View largeDownload slide Discrimination functions under absolute conditioning describing the probability of correct choices between 2 different stimuli of increasing chromatic dissimilarity, and corresponding predicted color differences. (a) Bombus terrestris, (b) Apis mellifera, and (c) mean and 95% CI color difference required by Apis (solid green squares) and Bombus (open orange squares) to achieve 50%, 60%, 70%, and 80% of correct choices as predicted by models in (a) and (b). Note that probability values for the 2 species where shifted by ±2 units to ease visual interpretation. In panels a–b circles represent the mean proportion of correct choices for each color difference tested, and bars indicate standard error. The discrimination function for each species is represented by the solid line and the shadowed region indicates the 95% confidence region. Coefficients describing the 2 functions are available in Table 2. Figure 2 View largeDownload slide Discrimination functions under absolute conditioning describing the probability of correct choices between 2 different stimuli of increasing chromatic dissimilarity, and corresponding predicted color differences. (a) Bombus terrestris, (b) Apis mellifera, and (c) mean and 95% CI color difference required by Apis (solid green squares) and Bombus (open orange squares) to achieve 50%, 60%, 70%, and 80% of correct choices as predicted by models in (a) and (b). Note that probability values for the 2 species where shifted by ±2 units to ease visual interpretation. In panels a–b circles represent the mean proportion of correct choices for each color difference tested, and bars indicate standard error. The discrimination function for each species is represented by the solid line and the shadowed region indicates the 95% confidence region. Coefficients describing the 2 functions are available in Table 2. Table 2 Magnitudes of the coefficients defining the sigmoidal functions describing the color discrimination function for Bombus terrestris and Apis mellifera when trained under absolute conditioning Coefficient Value (95% CI) B. terrestris (4-parameter logistic function)  K 0.861 (0.722, 0.915)  r 0.606 (0.071, 0.975)  Mo 0.401 (0.210, 0.526)  xmid 3.08 (2.35, 3.39) A. mellifera (3-parameter logistic function)  K 1.02 (0.845, 2.53)  r 0.863 (0.368, 0.1.49)  Mo 0.164 (0.081, 0.291) Coefficient Value (95% CI) B. terrestris (4-parameter logistic function)  K 0.861 (0.722, 0.915)  r 0.606 (0.071, 0.975)  Mo 0.401 (0.210, 0.526)  xmid 3.08 (2.35, 3.39) A. mellifera (3-parameter logistic function)  K 1.02 (0.845, 2.53)  r 0.863 (0.368, 0.1.49)  Mo 0.164 (0.081, 0.291) Refer to Garcia et al. (2017) for the mathematical formulation of the 2 color discrimination functions defined by the coefficients. View Large Table 2 Magnitudes of the coefficients defining the sigmoidal functions describing the color discrimination function for Bombus terrestris and Apis mellifera when trained under absolute conditioning Coefficient Value (95% CI) B. terrestris (4-parameter logistic function)  K 0.861 (0.722, 0.915)  r 0.606 (0.071, 0.975)  Mo 0.401 (0.210, 0.526)  xmid 3.08 (2.35, 3.39) A. mellifera (3-parameter logistic function)  K 1.02 (0.845, 2.53)  r 0.863 (0.368, 0.1.49)  Mo 0.164 (0.081, 0.291) Coefficient Value (95% CI) B. terrestris (4-parameter logistic function)  K 0.861 (0.722, 0.915)  r 0.606 (0.071, 0.975)  Mo 0.401 (0.210, 0.526)  xmid 3.08 (2.35, 3.39) A. mellifera (3-parameter logistic function)  K 1.02 (0.845, 2.53)  r 0.863 (0.368, 0.1.49)  Mo 0.164 (0.081, 0.291) Refer to Garcia et al. (2017) for the mathematical formulation of the 2 color discrimination functions defined by the coefficients. View Large For B. terrestris, the absolute color discrimination function predicts that color difference of about 3.8 JND are necessary for these bees to correctly discriminate between 2 color stimuli 75% of the time. For A. mellifera, color differences should be about 3.1 JND to achieve the same threshold. Color differences required by these 2 species for correctly discriminating between 2 colored objects 50%, 60%, 70%, and 80% of the time are graphically summarized in Figure 2, panel C. Interestingly, under absolute conditioning, both A. mellifera and B. terrestris perform at chance expectation level when color differences are equal or less than 1.0 JND. Color variability in flowers The range of color variability observed in all sampled flowers and across the 3 variability levels considering: within individual flowers (WIF), within flowers of the same species (WSV), and between species (BSB); is higher than 1.0 JND (Figure 3). The smallest color variability range was observed for the nivea type of A. majus, whereas the largest variability range was observed for the wild type of this species. Figure 3 View largeDownload slide Range of color variability observed for the sampled species across 3 different sampling levels. Panel a: within individual flowers (blue bars), and between flowers of the same species (red bars) color variability. Panel b: between species color variability. Figure 3 View largeDownload slide Range of color variability observed for the sampled species across 3 different sampling levels. Panel a: within individual flowers (blue bars), and between flowers of the same species (red bars) color variability. Panel b: between species color variability. Distributions of color variability within individual flowers (WIV), between flowers of the same species (WSV), and between the different types of A. majus are depicted in Figures 4 and 5. For the wild type, 87.5% of WIV color variability is higher than 1.0 JND and 85.0% of WSV color variability is higher than this threshold (Figure 4a,d). The WIV color variability observed in the mixta type of A. majus was greater than 1.0 JND for 66.7% of the tested pairs, whereas 82.5% of WSV color variability is greater than threshold (Figure 4b,e). Finally, 43.6% of WIV variation and 27.4% of WSV variation for the nivea type was above the 1.0 JND threshold (Figure 4c,f). Figure 4 View largeDownload slide Color variability in Antirrhinum majus. Probability density functions of color variability within individual flowers (WIF) (first row, panels a–c); within flowers of the species (WSV) (second row, panels d–f); and, empirical cumulative probability functions for WSV (third row, panels g–i) of wild type (dark violet bars, first column), mixta type (magenta bars, middle column), and nivea type (gray bars, right column) of A. majus modeled for Bombus terrestris. Dashed red curve in panels a–f represents the probability of discriminating a “blue-like” stimulus (Garcia et al. (2017) for a precise definition of the stimulus) of increasing dissimilarity presented simultaneously to bumblebees when trained under differential conditioned preference (Dyer et al. 2008; Garcia et al. 2017). Vertical red line indicates the color difference required to attain 75% correct choices under differential conditioned preference. Green curve in panels a–f represents the probability of discriminating a stimulus of increasing dissimilarity presented simultaneously to bumblebees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under absolute conditioned preference. Double arrows indicate the difference in color dissimilarity (about 3.0 JNDs) required to make 75% of correct choices on a color discrimination experiment when bumblebees are trained under either differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavioral paradigm. Figure 4 View largeDownload slide Color variability in Antirrhinum majus. Probability density functions of color variability within individual flowers (WIF) (first row, panels a–c); within flowers of the species (WSV) (second row, panels d–f); and, empirical cumulative probability functions for WSV (third row, panels g–i) of wild type (dark violet bars, first column), mixta type (magenta bars, middle column), and nivea type (gray bars, right column) of A. majus modeled for Bombus terrestris. Dashed red curve in panels a–f represents the probability of discriminating a “blue-like” stimulus (Garcia et al. (2017) for a precise definition of the stimulus) of increasing dissimilarity presented simultaneously to bumblebees when trained under differential conditioned preference (Dyer et al. 2008; Garcia et al. 2017). Vertical red line indicates the color difference required to attain 75% correct choices under differential conditioned preference. Green curve in panels a–f represents the probability of discriminating a stimulus of increasing dissimilarity presented simultaneously to bumblebees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under absolute conditioned preference. Double arrows indicate the difference in color dissimilarity (about 3.0 JNDs) required to make 75% of correct choices on a color discrimination experiment when bumblebees are trained under either differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavioral paradigm. Figure 5 View largeDownload slide Probability density (panels a–b, first row) and empirical cumulative functions (panels c–d, second row) of BSB. Between wild and mixta type of Antirrhinum majus (panels a, c first column) and between wild and nivea types for the same species (panels b, d second column). Dashed red curve in panels a–b represents the probability of discriminating a “blue-like” stimulus (see Garcia et al. 2017 for a precise definition of the stimulus) of increasing dissimilarity presented simultaneously to bumblebees when trained under differential conditioned preference (Dyer et al. 2008; Garcia et al. 2017). Vertical red line indicates the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Green curve in panels a–b represents the probability of discriminating a stimulus of increasing dissimilarity presented to bumblebees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Double arrows indicate the difference in color dissimilarity (about 3 JND) required to make 75% of correct choices on a color discrimination experiment when bumblebees were trained under either differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavior paradigm. Figure 5 View largeDownload slide Probability density (panels a–b, first row) and empirical cumulative functions (panels c–d, second row) of BSB. Between wild and mixta type of Antirrhinum majus (panels a, c first column) and between wild and nivea types for the same species (panels b, d second column). Dashed red curve in panels a–b represents the probability of discriminating a “blue-like” stimulus (see Garcia et al. 2017 for a precise definition of the stimulus) of increasing dissimilarity presented simultaneously to bumblebees when trained under differential conditioned preference (Dyer et al. 2008; Garcia et al. 2017). Vertical red line indicates the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Green curve in panels a–b represents the probability of discriminating a stimulus of increasing dissimilarity presented to bumblebees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Double arrows indicate the difference in color dissimilarity (about 3 JND) required to make 75% of correct choices on a color discrimination experiment when bumblebees were trained under either differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavior paradigm. BSV color variability between types of A. majus was high with 90.0% of the samples above 1.0 JND for the wild × mixta comparison and 97.3% of the wild × nivea variability above the 1.0 JND discrimination threshold proposed by the RN model (Figure 5a,b). Distributions of the WIF and WSV color differences for G. ovata and R. officinalis are presented in Figure 6 along with BSV values for the species comparison. Color variability values for G. ovata are above the 1.0 JND threshold 62.2% and 68.5% of the time for WIF and WSV color variability respectively; whereas, 63.5% and 73.3% of WIF and WSV variability are above threshold for R. officinalis. The entirety of the BSV color variability values for the G. ovata × R. officinalis species comparison are above 1.0 JND. Figure 6 View largeDownload slide Observed color variability for G. ovata (yellow bars, panels a, c, e), Rosemarinus officinalis (purple bars, panels b, d, f), and color difference between the 2 species (dashed bars, panels g and h). Distribution of WIV values for Goodenia (a) and Rosemary (b). WSV for Goodenia and Rosemarinus as density histograms (c–d), and as cumulative frequency distributions (e–f). BSV between of Goodenia and Rosemarinus illustrated as a density histogram (g) and as a cumulative density distribution (h). Dashed red curve in panels a–f represents the probability of discriminating either a “yellow-like” (panels a, c) or a “blue-like” stimulus (panels b, d) of increasing dissimilarity presented to honeybees when trained under differential conditioned preference (Dyer and Neumeyer 2005; Garcia et al. 2017). Vertical red line indicates the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Green curve in panels a–d represents the probability of discriminating a stimulus of increasing dissimilarity presented simultaneously to honeybees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Double arrows indicate the difference (about 3 JNDs) in color dissimilarity required to make 75% of correct choices on a color discrimination experiment when honeybees are trained under differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavior paradigm. Figure 6 View largeDownload slide Observed color variability for G. ovata (yellow bars, panels a, c, e), Rosemarinus officinalis (purple bars, panels b, d, f), and color difference between the 2 species (dashed bars, panels g and h). Distribution of WIV values for Goodenia (a) and Rosemary (b). WSV for Goodenia and Rosemarinus as density histograms (c–d), and as cumulative frequency distributions (e–f). BSV between of Goodenia and Rosemarinus illustrated as a density histogram (g) and as a cumulative density distribution (h). Dashed red curve in panels a–f represents the probability of discriminating either a “yellow-like” (panels a, c) or a “blue-like” stimulus (panels b, d) of increasing dissimilarity presented to honeybees when trained under differential conditioned preference (Dyer and Neumeyer 2005; Garcia et al. 2017). Vertical red line indicates the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Green curve in panels a–d represents the probability of discriminating a stimulus of increasing dissimilarity presented simultaneously to honeybees when trained under absolute conditioned preference. Vertical green line represents the color difference required to attain a proportion of correct choices equal to 75% under this training condition. Double arrows indicate the difference (about 3 JNDs) in color dissimilarity required to make 75% of correct choices on a color discrimination experiment when honeybees are trained under differential or absolute conditioned preference. Gray horizontal line represents the chance expectation-choices in the behavior paradigm. DISCUSSION Animal color vision is a topic that has received widespread interest in recent years, although there is often a paucity of data on how the different animal species may use color in complex, ecologically relevant scenarios (Garcia et al. 2017). The RN model is often employed to predict color discrimination in a wide range of animal species by using key noise values of different, sometimes only distantly related, species (Siddiqi et al. 2004; Schaefer et al. 2007; Martínez-Harms et al. 2012; Schultz and Fincke 2013). The threshold calculation of the RN model, based on quality parameterized data for the model honeybee species, predicts that 2 stimuli should be reliably discriminated, around 75% of the times, if separated by a color distance of (ΔS = 1.0) (Vorobyev et al. 2001). Subsequent studies gave units of just noticeable differences to ΔS, so a unit in RN space was defined as the just noticeable difference threshold (JND) (Siddiqi et al. 2004). This definition implies that color differences less than 1 JND are undistinguishable whereas values above 1.0 JND indicate “how much above threshold a color pair is discriminated” (Siddiqi et al. 2004). Surprisingly, however, few studies have attempted to validate with empirical, behavioral data if the RN model does make accurate predictions for classifying naturally occurring stimuli that can be clearly identified as biologically relevant signals. By using flower color stimuli from 3 plant species, A. majus, G. ovata and R. officinalis, we tested if the proposed threshold value of 1.0 for chromatic visual discrimination would reliably classify flowers of the same species using 2 well-established animal models (honeybees, and bumblebees), when considering within individual flowers color variation (WIV) or within species color variation (WSV). Specifically, we consider cases where these respective bees are known pollinators of the flowers, thus validating that the petal color is a biologically relevant signal. Our measurements and analyses allow for an accurate assessment of the probability of the RN model correctly classifying stimuli, if indeed such an experiment was done with no prior knowledge that various stimuli do come from the same signal data set. Data show that within flower signal variability range may extend well above the theoretical 1.0 JND discrimination threshold proposed by the RN model for the 2 bee observers considered and the various flower species analyzed (Figure 4a–f and Figure 6a–d). Statistically, this equates to a chance of a Type I error of α = 0.274 for the nivea type of A. majus (Figure 4g–i), representing a close to ideal case for growing flowers, and up to a Type I α = 0.850 for the wild type of the same species (Figure 4g). This means that for studies of bees visiting flowers the RN model would very frequently make errors about classifying the same biological signal as a different, discriminable color. These results thus indicated that natural flower pigmentation produces signal variability that overwhelms thresholds predictions by the RN model. How might animal vision deal with such real-world natural variability? Several studies have shown that hymenopteran insects including bumblebees (Dyer and Chittka 2004; Dyer et al. 2008), ants (Yilmaz et al. 2017) and honeybees (Giurfa 2004; Avarguès-Weber et al. 2010; Sommerlandt et al. 2016) have a plastic ability to learn small color differences with differential conditioning as was applied to empirically determine the limit of discrimination (Dyer and Neumeyer 2005; Dyer et al. 2008); however, if bees learn a rewarding color stimulus with absolute conditioning of single target types, as is often the case in natural conditions, then color discrimination is more coarse (Giurfa 2004). Indeed, color discrimination data for 2 bee model species suggest that color differences of about 3 JND units are required to reach the 75% correct choices, and even then, discrimination has a non-linear relation with chromatic dissimilarity rather than a binary, discriminate/or not operator (Figure 2). Moreover, a threshold range of near 3 JND is observed when comparing the color difference required to achieve 75% of correct choices when bees are trained under either absolute, or differential conditioned preference (double arrows Figures 4a–f and 6a–d). This is important as under differential conditioning color discrimination can be extremely fine and require differences well below 1.0 RN units to attain the 75% correct choices threshold level (Dyer and Neumeyer 2005; Dyer et al. 2008; Garcia et al. 2017). Interestingly, most of the signal variation within individual flowers and between flowers of the same species considered in this study are below the 75% discrimination threshold for the absolute conditioning (Figures 4a–f and 6a–d). We call this variation “perceptual slop,” which can be defined as the natural tolerance of the visual system to color signal differences, thus suggesting that the color visual system of bees is well-tuned to solve problems arising from signal variability in real-world flowers. The perceptual slop necessary for bees to maintain flower constancy may be exploited by predators, such as crab spiders, which rely on camouflage by background matching for successfully ambushing their prey (Théry and Casas 2002). For example, (Defrize et al. 2010) sampled over a hundred Misunema vatia crab spiders sitting on different flower species and reported color differences of up to 5 JND between the best-camouflaged crab spiders and their host flower. Interestingly, color differences between flowers of the same species equal or larger than this magnitude are not uncommon (Figures 4g–i and 6e–f). Tolerance of bees to inherent color variability present in natural flowers can also explain why honeybees seem to equally prefer empty flowers or flowers occupied with a crab spider matching certain flower colors (Heiling 2005), even if the outcome of such a choice is potentially fatal for a bee. Reliable discrimination of flower color by foraging bee pollinators is fundamental to maintain flower constancy (Chittka et al. 1999). Color discrimination in natural environments frequently occur under complex and perceptually noisy conditions (White and Kemp 2015; Bukovac et al. 2017), so color difference should likely be maximized to facilitate the transmission of the visual signal, a key assumption of the sensory drive theory for the evolution of signals (Basolo and Endler 1998; Boughman 2002). For flower species with very distinct colors, such as G. ovata and R. officinalis (Figure 6g–h), the range of WIV and WSV color variability is smaller than the color difference between species (Figure 3), thus facilitating reliable visual discrimination by bees. In fact, bees trained under absolute conditioning easily and rapidly learn to discriminate between “blue-like” and “yellow-like” targets with high accuracy after just few trials (Giurfa 2007; Avarguès-Weber and Giurfa 2014; Ravi et al. 2016). Considering the signal of A.majus, on the contrary, the range of WIV and WSV of the wild type is equal to, or larger than, color difference between the wild and mixta types (Figures 3 and 5a,c). Interestingly, when bumblebees have to discriminate between these 2 flower types following absolute conditioning, they can do so at a level that is significant from chance expectation, but make mistakes about 30% of the time (Dyer et al. 2007). This means that discriminating between colors whose difference is within the color variability range of one of the stimulus is potentially a difficult perceptual task. In the case of flowers, it is thus possible that in such scenarios, other traits such as scent could be used by bees to overcome the perceptual noise produced by color variability (Leonard et al. 2011; Kantsa et al. 2017). Since the honeybee was a main model used for developing and calibrating the RN model (Vorobyev et al. 2001), our work suggests that when using this theoretical color model for other animal species that the proposed threshold of 1.0 RN unit may be very strict if there is any natural variation in the spectral properties of biologically important signals, like flowers. This means that validation studies of how the RN model makes predictions for known biological signals are essential, and model predictions for unknown stimuli in a study are reported in the context of such essential model error rates; as is custom for classical statistical reporting (Zar 1999). It is also important that when implementing the RN model that precise and consistent data on the noise parameter is used, else the predictions can only ever be relative and should not be considered as comparable in any way to results from separate studies that employed different model parameters (Garcia et al. 2017). Indeed, several recent authors (Garcia et al. 2017; Renoult et al. 2017; Jacobs 2018; Marshall 2018; Ng et al. 2018; Osorio and Vorobyev 2018; Stuart-Fox 2018; Vasas et al. 2018) highlight that behavioral validation is important to confirm results obtained by implementing the RN model. Our results show how natural color variability set thresholds for color discrimination significantly larger than those solely predicted by the effect of photoreceptor noise, and that this variability is more likely to be of importance when animals use color for signaling and camouflage. FUNDING This work was supported by the Australian Research Council (http://dx.doi.org/10.13039/501100000923), Discovery Project (DP160100161) to AGD. We thank Beverley Glover, Heather Whitney, and Matthew Dorling for care of Antirrhinum majus plants and assistance with spectral measurements. We thank 2 anonymous reviewers and the editor for their insightful comments on our manuscript. Conflict of interest: No competing interests declared. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Garcia et al. (2018). REFERENCES Aho AC , Donner K , Hydén C , Larsen LO , Reuter T . 1988 . Low retinal noise in animals with low body temperature allows high visual sensitivity . Nature . 334 : 348 – 350 . Google Scholar Crossref Search ADS PubMed Avarguès-Weber A , de Brito Sanchez MG , Giurfa M , Dyer AG . 2010 . Aversive reinforcement improves visual discrimination learning in free-flying honeybees . PLoS One . 5 : e15370 . Google Scholar Crossref Search ADS PubMed Avarguès-Weber A , Giurfa M . 2014 . Cognitive components of color vision in honey bees: how conditioning variables modulate color learning and discrimination . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 200 : 449 – 461 . Google Scholar Crossref Search ADS PubMed Barlow HB . 1956 . Retinal noise and absolute threshold . J Opt Soc Am . 46 : 634 – 639 . Google Scholar Crossref Search ADS PubMed Barry KL , White TE , Rathnayake DN , Fabricant SA , Herberstein ME . 2015 . Sexual signals for the color-blind: cryptic female mantids signal quality through brightness . Funct Ecol . 29 : 531 – 539 . Google Scholar Crossref Search ADS Basolo AL , Endler JA . 1998 . sexual selection . Trends Ecol Evol . 5347 : 415 – 420 . Boughman JW . 2002 . How sensory drive can promote speciation . Trends Ecol Evol . 17 : 571 – 577 . Google Scholar Crossref Search ADS Briscoe AD , Chittka L . 2001 . The evolution of color vision in insects . Annu Rev Entomol . 46 : 471 – 510 . Google Scholar Crossref Search ADS PubMed Bukovac Z , Dorin A , Finke V , Shrestha M , Garcia J , Avarguès-Weber A , Burd M , Schramme J , Dyer A . 2016 . Assessing the ecological significance of bee visual detection and color discrimination on the evolution of flower colors . Evol Ecol . 31 : 153 – 172 . Google Scholar Crossref Search ADS Bukovac Z , Shrestha M , Garcia JE , Burd M , Dorin A , Dyer AG . 2017 . Why background colour matters to bees and flowers . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 203 : 369 – 380 . Google Scholar Crossref Search ADS PubMed Chittka L , Menzel R . 1992 . The evolutionary adaptation of flower colors and the insect pollinators’ color vision . J Comp Physiol A . 171 : 171 – 181 . Google Scholar Crossref Search ADS Chittka L , Thomson JD , Waser NM . 1999 . Flower constancy, insect psychology, and plant evolution . Naturwissenschaften . 86 : 361 – 377 . Google Scholar Crossref Search ADS Cortesi F , Cheney KL . 2010 . Conspicuousness is correlated with toxicity in marine opisthobranchs . J Evol Biol . 23 : 1509 – 1518 . Google Scholar Crossref Search ADS PubMed Defrize J , Théry M , Casas J . 2010 . Background colour matching by a crab spider in the field: a community sensory ecology perspective . J Exp Biol . 213 : 1425 – 1435 . Google Scholar Crossref Search ADS PubMed Donner K . 1992 . Noise and the absolute thresholds of cone and rod vision . Vision Res . 32 : 853 – 866 . Google Scholar Crossref Search ADS PubMed Dyer AG . 2012 . The mysterious cognitive abilities of bees: why models of visual processing need to consider experience and individual differences in animal performance . J Exp Biol . 215 : 387 – 395 . Google Scholar Crossref Search ADS PubMed Dyer AG , Boyd-Gerny S , McLoughlin S , Rosa MG , Simonov V , Wong BB . 2012 . Parallel evolution of angiosperm colour signals: common evolutionary pressures linked to hymenopteran vision . Proc Biol Sci . 279 : 3606 – 3615 . Google Scholar Crossref Search ADS PubMed Dyer AG , Chittka L . 2004 . Fine colour discrimination requires differential conditioning in bumblebees . Naturwissenschaften . 91 : 224 – 227 . Google Scholar Crossref Search ADS PubMed Dyer AG , Neumeyer C . 2005 . Simultaneous and successive color discrimination in the honeybee (Apis mellifera) . J Comp Physiol A . 191 : 547 – 557 . Google Scholar Crossref Search ADS Dyer AG , Paulk AC , Reser DH . 2011 . Colour processing in complex environments: insights from the visual system of bees . Proc Biol Sci . 278 : 952 – 959 . Google Scholar Crossref Search ADS PubMed Dyer AG , Spaethe J , Prack S . 2008 . Comparative psychophysics of bumblebee and honeybee colour discrimination and object detection . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 194 : 617 – 627 . Google Scholar Crossref Search ADS PubMed Dyer AG , Whitney HM , Arnold SEJ , Glover BJ , Chittka L . 2007 . Mutations perturbing petal cell shape and anthocyanin synthesis influence bumblebee perception of Antirrhinum majus flower color . Arthropod Plant Interact . 1 : 45 – 55 . Google Scholar Crossref Search ADS Endler JA . 1990 . On the measurement and classification of color in studies of animal color patterns . Biol J Linn Soc . 41 : 315 – 352 . Google Scholar Crossref Search ADS Fleishman LJ , Perez CW , Yeo AI , Cummings KJ , Dick S , Almonte E . 2016 . Perceptual distance between colored stimuli in the lizard Anolis sagrei: comparing visual system models to empirical results . Behav Ecol Sociobiol . 70 : 541 – 555 . Google Scholar Crossref Search ADS von Frisch K . 1967 . The dance language and orientation of bees . Cambridge : Harvard University Press . Garcia JE , Greentree AD , Shrestha M , Dorin A , Dyer AG . 2014 . Flower colours through the lens: quantitative measurement with visible and ultraviolet digital photography . PLoS One . 9 : e96646 . Google Scholar Crossref Search ADS PubMed Garcia JE , Shrestha M , Dyer AG . 2018 . Data from: flower signal variability overwhelms receptor-noise and requires plastic color learning in bees . Dryad Digital Repository . https://doi.org/10.5061/dryad.5d8k268. Garcia JE , Spaethe J , Dyer AG . 2017 . The path to colour discrimination is S-shaped: behaviour determines the interpretation of colour models . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 203 : 983 – 997 . Google Scholar Crossref Search ADS PubMed Giurfa M . 2004 . Conditioning procedure and color discrimination in the honeybee Apis mellifera . Naturwissenschaften . 91 : 228 – 231 . Google Scholar Crossref Search ADS PubMed Giurfa M . 2007 . Behavioral and neural analysis of associative learning in the honeybee: a taste from the magic well . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 193 : 801 – 824 . Google Scholar Crossref Search ADS PubMed Glover BJ , Martin C . 1998 . The role of petal cell shape and pigmentation in pollination success in Antirrhinum majus . Heredity (Edinb) . 80 : 778 – 784 . Google Scholar Crossref Search ADS Goldsmith TH . 1990 . Optimization, constraint, and history in the evolution of eyes . Q Rev Biol . 65 : 281 – 322 . Google Scholar Crossref Search ADS PubMed Heiling AM . 2005 . Coloration in crab spiders: substrate choice and prey attraction . J Exp Biol . 208 : 1785 – 1792 . Google Scholar Crossref Search ADS PubMed von Helversen O . 1972 . Zur spektralen Unterschiedsempfindlichkeit der honigbiene . J Comp Physiol . 80 : 439 – 472 . Google Scholar Crossref Search ADS Hempel de Ibarra N , Vorobyev M , Menzel R . 2014 . Mechanisms, functions and ecology of color vision in the honeybee . J Comp Physiol A . 200 : 1 – 23 . Google Scholar Crossref Search ADS Hopkins R , Rausher MD . 2011 . Identification of two genes causing reinforcement in the Texas wildflower Phlox drummondii . Nature . 469 : 411 – 414 . Google Scholar Crossref Search ADS PubMed Hopkins R , Rausher MD . 2012 . Pollinator-mediated selection on flower color allele drives reinforcement . Science . 335 : 1090 – 1092 . Google Scholar Crossref Search ADS PubMed Howard J , Blakeslee B , Laughlin SB . 1987 . The intracellular pupil mechanism and photoreceptor signal: noise ratios in the fly Lucilia cuprina . Proc R Soc Lond B Biol Sci . 231 : 415 – 435 . Google Scholar Crossref Search ADS PubMed Igic B , Cassey P , Grim T , Greenwood DR , Moskát C , Rutila J , Hauber ME . 2012 . A shared chemical basis of avian host-parasite egg colour mimicry . Proc Biol Sci . 279 : 1068 – 1076 . Google Scholar Crossref Search ADS PubMed Jacobs GH . 2018 . Photopigments and the dimensionality of animal color vision . Neurosci Biobehav Rev . 86 : 108 – 130 . Kantsa A , Raguso RA , Dyer AG , Sgardelis SP , Olesen JM , Petanidou T . 2017 . Community-wide integration of floral colour and scent in a Mediterranean scrubland . Nat Ecol Evol . 1 : 1502 – 1510 . Google Scholar Crossref Search ADS PubMed Kapustjanskij A , Streinzer M , Paulus HF , Spaethe J . 2007 . Bigger is better: implications of body size for flight ability under different light conditions and the evolution of alloethism in bumblebees . Funct Ecol . 21 : 1130 – 1136 . Google Scholar Crossref Search ADS Kay QON , Daoud HS , Stirton CH . 1981 . Pigment distribution, light reflection and cell structure in petals . Bot J Linn Soc . 83 : 57 – 83 . Google Scholar Crossref Search ADS Kelber A , Vorobyev M , Osorio D . 2003 . Animal colour vision–behavioural tests and physiological concepts . Biol Rev Camb Philos Soc . 78 : 81 – 118 . Google Scholar Crossref Search ADS PubMed Kemp DJ , Herberstein ME , Fleishman LJ , Endler JA , Bennett AT , Dyer AG , Hart NS , Marshall J , Whiting MJ . 2015 . An integrative framework for the appraisal of coloration in nature . Am Nat . 185 : 705 – 724 . Google Scholar Crossref Search ADS PubMed Kevan P , Giurfa M , Chittka L . 1996 . Why are there so many and so few white flowers ? Trends Plant Sci . 1 : 280 – 284 . Google Scholar Crossref Search ADS Kien J , Menzel R . 1977 . Chromatic properties of interneurons in the optic lobes of the bee. II. Narrow band and color opponent neurons . J Comp Physiol . 113 : 35 – 53 . Google Scholar Crossref Search ADS van der Kooi CJ , Dyer AG , Kevan PG , Lunau K . 2018 . Functional significance of the optical properties of flowers for visual signalling . Ann Bot . doi: https://doi.org/10.1093/aob/mcy119 van der Kooi CJ , Dyer AG , Stavenga DG . 2015 . Is floral iridescence a biologically relevant cue in plant-pollinator signaling ? New Phytol . 205 : 18 – 20 . Google Scholar Crossref Search ADS PubMed van der Kooi CJ , Elzenga JTM , Dijksterhuis J , Stavenga DG . 2017 . Functional optics of glossy buttercup flowers . J R Soc Interface . 14 : 20160933 . Google Scholar Crossref Search ADS PubMed van der Kooi CJ , Elzenga JTM , Staal M , Stavenga DG . 2016 . How to color a flower: on the optical principles of flower coloration . Proc R Soc B . 283 : 20160429 . Google Scholar Crossref Search ADS PubMed 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 – 121 . Google Scholar Crossref Search ADS PubMed Li L , MaBouDi H , Egertová M , Elphick MR , Chittka L , Perry CJ . 2017 . A possible structural correlate of learning performance on a color discrimination task in the brain of the bumblebee . Proc R Soc B . 284 : 20171323 . Google Scholar Crossref Search ADS PubMed MacAdam DL . 1942 . Visual sensitivities to color differences in daylight . J Opt Soc Am . 32 : 247 . Google Scholar Crossref Search ADS Marshall J . 2018 . Do not be distracted by pretty colors: a comment on Olsson et al . Behav Ecol . 29 : 286 – 287 . Google Scholar Crossref Search ADS Martínez-Harms J , Vorobyev M , Schorn J , Shmida A , Keasar T , Homberg U , Schmeling F , Menzel R . 2012 . Evidence of red sensitive photoreceptors in Pygopleurus israelitus (Glaphyridae: Coleoptera) and its implications for beetle pollination in the southeast Mediterranean . J Comp Physiol A Neuroethol Sens Neural Behav Physiol . 198 : 451 – 463 . Google Scholar Crossref Search ADS PubMed Ng L , Garcia JE , Dyer AG . 2018 . Why color is complex: evidence that bees perceive neither brightness nor green contrast in color signal processing . Facets . 3 : 800 – 817 . Google Scholar Crossref Search ADS Noda K , Glover BJ , Linstead P , Martin C . 1994 . Flower colour intensity depends on specialized cell shape controlled by a Myb-related transcription factor . Nature . 369 : 661 – 664 . Google Scholar Crossref Search ADS PubMed Olsson P , Lind O , Kelber A . 2017 . Chromatic and achromatic vision: parameter choice and limitations for reliable model predictions . Behav Ecol . 29 : 273 – 282 . Google Scholar Crossref Search ADS Osorio D , Vorobyev M . 2018 . Principles and application of the receptor noise model of color discrimination: a comment on Olsson et al . Behav Ecol . 29 : 283 – 284 . Google Scholar Crossref Search ADS 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 . 170 : 23 – 40 . Google Scholar Crossref Search ADS PubMed Pinheiro J , Bates D , DebRoy S , Sarkar D . 2013 . nlme: linear and nonlinear mixed effects models . Rausher MD . 2008 . Evolutionary transitions in floral color . Int J Plant Sci . 169 : 7 – 21 . Google Scholar Crossref Search ADS Ravi S , Garcia JE , Wang C , Dyer AG . 2016 . The answer is blowing in the wind: free-flying honeybees can integrate visual and mechano-sensory inputs for making complex foraging decisions . J Exp Biol . 219 : 3465 – 3472 . Google Scholar Crossref Search ADS PubMed R Development Core Team . 2017 . R: a language and environment for statistical computing . Vienna, Austria Renoult JP , Kelber A , Schaefer HM . 2017 . Colour spaces in ecology and evolutionary biology . Biol Rev Camb Philos Soc . 92 : 292 – 315 . Google Scholar Crossref Search ADS PubMed Schaefer M , Schaefer V , Vorobyev M . 2007 . Are fruit colors adapted to consumer vision and birds equally efficient in detecting colorful signals? Am Nat . 169 (Suppl 1) : S159 – S169 . Schnapf JL , Nunn BJ , Meister M , Baylor DA . 1990 . Visual transduction in cones of the monkey Macaca fascicularis . J Physiol . 427 : 681 – 713 . Google Scholar Crossref Search ADS PubMed Schrödinger E . 1926 . Thresholds of color differences . In: MacAdam DL , editor. Sources of color science . Cambridge (MA) : MIT Press . p. 183 – 193 . Schultz TD , Fincke OM . 2013 . Lost in the crowd or hidden in the grass: signal apparency of female polymorphic damselflies in alternative habitats . Anim Behav . 86 : 923 – 931 . Google Scholar Crossref Search ADS Shrestha M , Dyer AG , Bhattarai P , Burd M . 2014 . Flower color and phylogeny along an altitudinal gradient in the Himalayas of Nepal . J Ecol . 102 : 126 – 135 . Google Scholar Crossref Search ADS Shrestha M , Dyer AG , Boyd-Gerny S , Wong BB , Burd M . 2013 . Shades of red: bird-pollinated flowers target the specific colour discrimination abilities of avian vision . New Phytol . 198 : 301 – 310 . Google Scholar Crossref Search ADS PubMed Shrestha M , Lunau K , Dorin A , Schulze B , Bischoff M , Burd M , Dyer AG . 2016 . Floral colours in a world without birds and bees: the plants of Macquarie Island . Plant Biol (Stuttg) . 18 : 842 – 850 . Google Scholar Crossref Search ADS PubMed Siddiqi A , Cronin TW , Loew ER , Vorobyev M , Summers K . 2004 . Interspecific and intraspecific views of color signals in the strawberry poison frog Dendrobates pumilio . J Exp Biol . 207 : 2471 – 2485 . Google Scholar Crossref Search ADS PubMed Skorupski P , Chiitka L . 2010 . Differences in photoreceptor processing speed for chromatic and achromatic vision in the bumblebee, Bombus terrestris . J Neurosci . 30 : 3896 – 3903 . Google Scholar Crossref Search ADS PubMed Sobel JM , Streisfeld MA . 2013 . Flower color as a model system for studies of plant evo-devo . Front Plant Sci . 4 : 321 . Google Scholar Crossref Search ADS PubMed Sommerlandt FM , Spaethe J , Rössler W , Dyer AG . 2016 . Does fine color discrimination learning in free-flying honeybees change mushroom-body calyx neuroarchitecture ? PLoS One . 11 : e0164386 . Google Scholar Crossref Search ADS PubMed Stuart-Fox D . 2018 . Opening the “black box” of modeling animal color vision: a comment on Olsson et al . Behav Ecol . 29 : 284 . Google Scholar Crossref Search ADS Théry M , Casas J . 2002 . Predator and prey views of spider camouflage: both hunter and hunted fail to notice crab-spiders blending with colored petals . Nature . 415 : 133 . Google Scholar Crossref Search ADS PubMed Vasas V , Brebner JS , Chittka L . 2018 . Color discrimination is not just limited by photoreceptor noise: a comment on Olsson et al . Behav Ecol . 29 : 285 – 286 . Google Scholar Crossref Search ADS Vorobyev M , Brandt R , Peitsch D , Laughlin SB , Menzel R . 2001 . Colour thresholds and receptor noise: behaviour and physiology compared . Vision Res . 41 : 639 – 653 . Google Scholar Crossref Search ADS PubMed Vorobyev M , Osorio D . 1998 . Receptor noise as a determinant of colour thresholds . Proc Biol Sci . 265 : 351 – 358 . Google Scholar Crossref Search ADS PubMed Waser N , Chittka L . 1998 . Evolutionary ecology: bedazzled by flowers . Nature . 394 : 835 – 836 . Google Scholar Crossref Search ADS Whibley AC , Langlade NB , Andalo C , Hanna AI , Bangham A , Thébaud C , Coen E . 2006 . Evolutionary paths underlying flower color variation in Antirrhinum . Science . 313 : 963 – 966 . Google Scholar Crossref Search ADS PubMed White TE , Dalrymple RL , Herberstein ME , Kemp DJ . 2017 . The perceptual similarity of orb-spider prey lures and flower colors . Evol Ecol . 31 : 1 – 20 . Google Scholar Crossref Search ADS White TE , Kemp DJ . 2015 . Technicolor deceit: a sensory basis for the study of color-based lures . Anim Behav . 105 : 231 – 243 . Google Scholar Crossref Search ADS Wienand U , Sommer H , Schwarz Z , Shepherd N , Saedler H , Kreuzaler F , Ragg H , Fautz E , Hahlbrock K , Harrison B , et al. 1982 . A general method to identify plant structural genes among genomic DNA clones using transposable element induced mutations . MGG Mol Gen Genet . 187 : 195 – 201 . Google Scholar Crossref Search ADS Wyszecki G , Stiles WS . 1982 . Color science concepts and methods, quantitative data and formulae . 2 nd ed. New York : John Wiley & Sons, Inc . Yang EC , Lin HC , Hung YS . 2004 . Patterns of chromatic information processing in the lobula of the honeybee, Apis mellifera L . J Insect Physiol . 50 : 913 – 925 . Google Scholar Crossref Search ADS PubMed Yilmaz A , Dyer AG , Rössler W , Spaethe J . 2017 . Innate colour preference, individual learning and memory retention in the ant Camponotus blandus . J Exp Biol . 220 : 3315 – 3326 . Google Scholar Crossref Search ADS PubMed Zar JH . 1999 . Biostatistical analysis . 4 th ed. Upper Saddle River (NJ) : Prentice Hall . © The Author(s) 2018. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Flower signal variability overwhelms receptor-noise and requires plastic color learning in bees
JF - Behavioral Ecology
DO - 10.1093/beheco/ary127
DA - 2018-11-27
UR - https://www.deepdyve.com/lp/oxford-university-press/flower-signal-variability-overwhelms-receptor-noise-and-requires-fC0MjYF6LH
SP - 1286
VL - 29
IS - 6
DP - DeepDyve
ER -