TY - JOUR
AU - Dunlap, Aimee, S
AB - Abstract Animals reduce uncertainty in their lifetime by using information to guide decision making. Information available can be inherited from the past or gathered from the present. Therefore, animals must balance inherited biases with new information that may be in conflict with those potential biases. In our study, we set up color pairings such that an arbitrarily chosen focal color, human-orange, would result in an inherent bias in comparison to 3 other colors tested resulting in equal, medium, and strong preference differences. We chose color pairings through a series of preferences tests across 8 colonies of bumble bees. We subsequently used these pairings with rewards that varied in quality (good or bad states) and consistency (steady and fluctuating) in order to investigate how inherited biases affect the foraging choices of bumble bees when new information is gathered. We found that the preexisting color biases within our bees were only maintained when the reward associated with those colors was steady, even if paired with mediocre sugar concentrations. When maintained, we observed that other aspects of bee choice also reflected this bias, including increased sampling for the preferred color and an increased likelihood of choosing that color in a subsequent choice. Thus, environmental change and reward differences interact with the level of preexisting bias to determine whether inherited information is more heavily weighted than newly gathered information, and even a strong preexisting bias can be quickly erased with experience under some conditions. INTRODUCTION Food sources in nature are ephemeral, with food quality, abundance, and appearance varying temporally and spatially (Shettleworth et al. 1988; Kawaguchi et al. 2006; Goulson et al. 2007). Therefore, nectar and pollen foragers are faced with the dilemma of keeping track of changing environments (i.e., being able to discriminate between rewarding and unrewarding flowers) and make decisions as to when, where and what they should forage to gain sufficient energy for growth, survival, and reproduction (Pyke et al. 1977; Kawaguchi et al. 2006; Dunlap and Stephens 2012). In highly changing environments foragers learn to associate cues such as floral color with reward thereby allowing the forager to track changes in rewards (Weiss 1997; Gumbert 2000). And it is easy to imagine how more accurate and rewarding choices can lead to increased foraging efficiency and more total food resources acquired (Raine and Chittka 2008), which will in turn increase their fitness (Kawaguchi et al. 2006). In the absence of correlated cues (i.e., when a forager is faced with incomplete information), the forager can take actions to reduce uncertainty by gathering information typically at the cost of time and energy (Dall et al. 2005; Stephens 2008). Foragers can track environmental changes by sampling a potential resource—analogous to exploration of environment without necessarily receiving immediate reward (Stephens and Krebs 1986; Shettleworth et al. 1988), which allows them to detect changes in fluctuating food patches (Shettleworth et al. 1988). Moreover, animals gain experience via sampling which can result in an adaptive modification of their behavior (learning). Therefore, sampling facilitates an increase in their overall fitness (Raine and Chittka 2008) by allowing foragers to maximize their rate of food intake by choosing a fluctuating patch when it is in a good state and otherwise feeding at a stable/steady mediocre patch (Shettleworth 2010). Previous models of information sampling and tracking assume a world in which there are no preexisting preferences of the stimuli indicating the possible resources, for instance the colors associated with steady or fluctuating rewards. Intuition tells us that animals should more quickly sample a substrate they already prefer and may never sample a substrate when it is indicated by a stimulus that is not preferred. As such, we see in nature that selection favors signals, for example, specific floral signals, that exploit the preferences and perceptual abilities of animal pollinators (Weiss 1997; Schiestl and Johnson 2013). The cognitive basis of pollinator attraction to floral signals is dependent on both innate behavioral responses and on learning as well as on the interaction between these factors (Weiss 1997; Gumbert 2000; Schiestl and Johnson 2013). Innate sensory preferences can be the outcome of unilateral adaptation of pollinators to flowers, reciprocal adaptation leading to coevolution or even non-floral visitation behaviors such as finding mates (Lunau and Maier 1995; Ellis and Johnson 2012; Schiestl and Johnson 2013). Such innate sensory preferences or receiver biases can be sensory in nature; thus, signals can be detected efficiently. However, it can also be based on neuronal or higher cognitive processes in the perceiver’s brain (Endler and Basolo 1998). In contrast to receiver biases, innate sensory preferences in pollinators can also reflect the outcome of selection for efficient usage of the most rewarding flowers in a specific habitat (Raine and Chittka 2007; Schiestl and Johnson 2013). Therefore, preferences can evolve due to long-term patterns of reward for a population and subsequently passed genetically to new members of the population, or preferences can develop within a lifetime based upon favorable experiences, thus preference may also be learned (Schiestl and Johnson 2013; Jones et al. 2015). Oligoletic bees that rely on specific host plants for pollen often show sensory preferences for host plants that display signals such as floral color or scent. These preferences are thought to have a genetic basis and can be so strong that it will lead to bees rejecting suitable pollen sources even if their plant of preference is absent from the habitat (Schiestl and Johnson 2013). However, bumble bees, generalist foragers that extensively use associative learning, also rely on innate preferences when choosing flowers as observed by Müller (1881) who noted bees’ fondness for blue colored flowers in nature (as cited in Gumbert 2000). Therefore, even generalist insects seem to demonstrate an unlearned preference for certain colors and although such preferences can be modified by learning, their preexisting sensory biases can be very resilient (Menzel 1985; Leadbeater and Chittka 2007a; Leadbeater and Chittka 2009; Jones et al. 2015). From our examples, of both specialist and generalist foragers, we see that information is available to animals from a variety of sources (Menzel 1985; Dall et al. 2005; Jones et al. 2015). This information may be categorized as “personal information” which is generated through direct locomotory and sensory interactions with the environment (Menzel 1985), “socially acquired information” that is obtained from other foragers (Dall et al. 2005; Kawaguchi et al. 2006) and phylogenetic experience that is genetically encoded and can account for preexisting sensory biases (Menzel 1985; Jones et al. 2015). Thus, in ever-changing environments foragers must continuously attempt to reduce uncertainty through information gathering and use of preexisting knowledge. It is also equally important that they learn to discard information that is unreliable or inaccurate in order to make optimal decisions that will ultimately affect survival and reproduction (Koops 2004; Dall et al. 2005). Our study investigates implications of foraging theory in the Common Eastern Bumblebee, Bombus impatiens from a unique perspective of animals balancing the value of inherited information (sensory biases) with individual experience (i.e., information gathered while sampling), as optimization of foraging is likely dependent on the respective costs and benefits of these 2 information sources (Mery and Burns 2010). We use a framework of bees choosing between 2 resources: one which is steady in value, and one which fluctuates between a good and a bad state. By manipulating the relative strength of the innate bias between the colors indicating each resource, and whether the preferred color indicates a steady or fluctuating reward, we can look at the interactions between innate preference, sampling, and tracking change. We specifically focus on answering 4 questions; 1) How do inherited biases affect when and how frequently new information is sampled? 2) How does inherited bias affect the speed with which new information is learned? 3) How do inherited bias and acquired information interact in enabling bees to track changes in reward? And because few resources in nature are completely predictable and stable, we also ask 4) How does inherited bias affect choice of a resource with a variable reward in foraging? METHODS Husbandry Four colonies of B. impatiens (Koppert Biological Systems, Michigan), were used in the main experiment, with one colony actively participating in the experiment at any one time. Bees from an additional 6 colonies were used in preference testing prior to the main experiment. After arrival, we relocated the colony into a wooden nest box with plexiglas lid (43 cm × 23 cm × 10 cm) and placed a sucrose source in an adjacent feeder box of same size connected by clear plastic tubing (3 cm diameter). We kept colonies under a 12 h light dark cycle and fed 20% sucrose ad libitum with ground honeybee-collected pollen given every other day upon completion of trials. Although most colonies were fed with 20% sucrose, sucrose concentration given varied if colony needed motivation to complete trials, for example, some colonies were fed 50% sucrose for short periods to elicit foraging behavior. Foragers were tagged with numbered discs glued between the wings on the dorsal region of the thorax. Maze design To assess bee decisions, we used a serial Y-shaped maze which allowed for a choice between 2 colors (as illustrated and described in Figure 2, Dunlap et al. 2017). The maze was constructed from a matte gray foam board base, Plexiglas Y-shaped segments, and clear Plexiglass lids. The maze consisted of 10 individually connected Y-segments. The 2 arms of a maze were marked with rectangles of colored craft foam placed centrally on each side. To ensure the choices were well marked with the colors, we also attached square pieces of matching craft foam to the wall at the end of each arm. We placed a glass microscope slide (100 mm × 30 mm) on each rectangular stimulus, onto which the sucrose solution reward could be pipetted (directly over a target marked with a black line on the craft foam). After making a single choice, the bee moved into the next Y-maze segment for the next choice. Following 10 choices, the bee was allowed to return to the colony before another foraging trip through the maze. Before every trial, the floor of each maze segment was covered in a clean removable clear plastic liner to allow for the removal of scent “footprints” that might be used as cues to foragers, and each glass microscope slide was also replaced. Wooden rectangular doors restricted individuals once a color selection was made, so that only one arm of the maze could be experienced by the bee in a given trial and so that bees moved forward rather than backwards through the maze. The maze was attached to the colony box with 3-cm plastic tubing, where bees could be diverted from visiting the feeder box towards the maze instead. To ensure that only the bee being tested could enter the maze; we used a series of doors placed into the tubing. Upon completion of the final maze segment, the bee returned to the nest box through more tubing. Training We chose motivated foragers for the experiment while simultaneously training bees to walk through the maze and forage from the target stimuli. We connected the colony box to a training segment using clear plastic tubing similar to that used to connect the feeder box. The training segment consisted of a single Y-maze, identical to those used in the serial trials. For training, we used a gray and white craft foam stimulus under the microscope slides, selected to prevent color bias in the experimental maze segments during trials. This training segment was used to familiarize foragers with the experimental setup, as individuals would travel through the training segment to reach the feeder box. 5 µL of 22.5% sucrose was pipetted onto the microscope slide along the marked target in the training segment and was replenished upon consumption. All tagged foragers successfully foraging sucrose from training segments and returning to the nest box were recorded. “Successful” foragers, which were subsequently used in the experiment, were those consistently active, tagged foragers that sampled from the training segments and returned to the nest box on multiple occasions. Color preference trials We conducted a series of simple preference trials to choose the colors used in the main experiment. We arbitrarily chose a single color, human-orange, as a focal color to which each additional color’s preference difference would be recorded, with the goal of finding a strong, moderate, and equal preference difference. Color preference trial treatments consisted of 80 choices, where every color stimulus was equally rewarding with a 50% sucrose solution. We calculated preferences for each bee as a proportional choice, and then averages were calculated across all bees tested, with each bee only being tested once. This series of preference tests was carried out across 8 colonies of bumble bees to minimize any potential colony-specific preferences. From these preliminary trials using 10 total colors paired with orange we then chose 3 focal colors to use in the experiment: high preference—orange strongly preferred over purple; medium preference—orange moderately preferred over blue; and equal preference—orange equally preferred to yellow. The preference data of these colors are in Figure 1. Figure 2b shows the spectral reflectance measurements of these colors, measured using an Ocean Optics USB-200+UV-VIS spectrometer with a PX-2 pulsed xenon light source. Figure 2a shows the locations of these colors in hexagonal color space for Bombus (Chittka 1992), calculated using the spectral sensitivities of B. impatiens (Skorupski and Chittka 2010). The hexagon is rooted against the gray background that comprised the floor and walls of the full maze apparatus. Figure 1 View largeDownload slide Measured differences in preference for the stimulus colors chosen for the experiment. The total percent chosen for each color averaged across all bees is shown. The x axis represents the color preference in relation to the focal color orange, with orange and purple representing a high preference difference, blue and orange representing a medium preference difference, and yellow and orange as the equal preference treatments. Figure 1 View largeDownload slide Measured differences in preference for the stimulus colors chosen for the experiment. The total percent chosen for each color averaged across all bees is shown. The x axis represents the color preference in relation to the focal color orange, with orange and purple representing a high preference difference, blue and orange representing a medium preference difference, and yellow and orange as the equal preference treatments. Figure 2 View largeDownload slide (a) Color hexagon calculated for the stimuli used in the experiment, centered on the background; following Chittka 1992. (b) Spectral reflectance measures for the colored stimuli used in the experiment, in comparison to the background color of the apparatus. The human colors are identified as 1) orange, 2) yellow, 3) dark blue, and 4) purple. Figure 2 View largeDownload slide (a) Color hexagon calculated for the stimuli used in the experiment, centered on the background; following Chittka 1992. (b) Spectral reflectance measures for the colored stimuli used in the experiment, in comparison to the background color of the apparatus. The human colors are identified as 1) orange, 2) yellow, 3) dark blue, and 4) purple. To further describe the colors of these stimuli we calculated 3 additional values for each color, following Spaethe et al. (2001). The distance of each color to the background, calculated in hexagon units is 0.73 for orange, 0.51 for yellow, 0.10 for blue, and 0.23 for purple. The brightness contrast, calculated as the sum of the receptor type excitation values, is 1.17 for orange, and 1.51, 2.12, and 1.99 for yellow, blue, and purple, respectively. And finally, as green contrast is relevant for visual processing of various kinds (reviewed in van der Kooi et al. 2018), we calculated the distance of each stimulus color for the green photoreceptor from the background. Here, orange is 0.37, whereas yellow, blue, and purple are 0.31, 0.14, and 0.02, respectively. Experimental design and procedure We designed the experiment as a 3 × 2 factorial, with the 3 levels of preference strength (equal, medium, and high), and 2 levels of change, where the reward of the focal color either remained steady throughout the trials or fluctuated between a bad and then good state. We manipulated the state of the resource through changes in concentration of the sucrose reward, where a steady reward was 22.5%, and the fluctuating resource alternated between a bad (5%) and good (50%) state. This allows us to look at how change interacts with preference to determine choice (Figure 3). We randomized the order of the treatments into complete experimental blocks of 6 treatments each, for a total of 8 blocks, or 48 subjects. The randomized blocked order allowed for a relatively equal allocation of treatments across time and colonies. For each choice trial, we randomized the placement of the focal color (right or left). Figure 3 View largeDownload slide Design of trials. The y axis is the level of reward, whereas the x axis is choice number. One resource provides a steady, mediocre reward whereas the second resource fluctuated between a poor reward and a good reward every 20 trials. The orange focal resource could be either fluctuating or stable in reward, and could be paired either yellow, blue, or purple to provide an equal, medium, or high level of difference. Figure 3 View largeDownload slide Design of trials. The y axis is the level of reward, whereas the x axis is choice number. One resource provides a steady, mediocre reward whereas the second resource fluctuated between a poor reward and a good reward every 20 trials. The orange focal resource could be either fluctuating or stable in reward, and could be paired either yellow, blue, or purple to provide an equal, medium, or high level of difference. After the subject entered the first maze segment, wooden doors were used to confine the bee into the segment until a color choice was made, that is, proboscis extension or antennation towards sucrose. We recorded color choice and the wooden door was removed allowing the individual to proceed to the next segment. This procedure was repeated for the length of the maze. Upon completion of a block of trials, that is, 10 individual choices, we allowed the bee to return to the nest to empty her crop and fill honey pots. We then cleaned the maze and set up for the next series of choices from the same individual until 80 total choices were made (8 runs through the maze). We removed an individual from the experiment if no evidence of foraging or acknowledgment of sucrose occurred after 15 minutes and these bees were subsequently frozen. Dependent measures We calculated a series of dependent measures based upon the bees’ choices to assess their frequency of sampling the fluctuating resource after experiencing it in a poor state, their tendency to choose the fluctuating resource, the types of errors they made, and the effect of experience upon their choices. Choice of the focal resource For each bee, our focal resource was marked with human-orange coloration. To assess the role of experience in influencing this choice, we split the 80 choices of each bee into blocks of 10 consecutive trials each (for a total of 8 blocks) and calculated a proportion for each block. Sampling window Following the classic sampling literature, we define a sampling event as a choice of the fluctuating resource after it was experienced in the bad state (e.g., Stephens 1987; Shettleworth 1989). After determining sampling events for each bee, we calculated the sampling window: the number of trials elapsed between each sampling event; analogous to the Stephens 1987 model. Choice of fluctuating resource We calculated the simple proportion of choices on the fluctuating resource. Errors of win-stay and lose-shift rules A simple learning rule is win-stay, lose shift: stay on the resource if it is good and leave the resource if it is bad. Sampling models assume nothing more sophisticated than this simple rule, and find that the optimal sampling window increases in more fixed environments and decreases as fixity decreases (Stephens 1987). While much evidence in animals, including bees (e.g., Keasar et al. 2002), shows that they use decision rules much more complicated than this in the real world, we can use violations of win-stay, lose-shift to understand their choices in a little more detail. Specifically, by looking at the types of errors bees make of these rules, we can gain some insight into whether they are tending to stick to resources that are rewarding and leaving them quickly when they are not. For instance, one might imagine a bee with a strong initial preference for resource A might be less likely to choose a different resource even after experiencing A in a poor state, and once switching to B, may be more likely to leave B even if it is rewarding. We counted win-stay errors, as a bee gaining a good reward on the fluctuating resource and then choosing the steady resource on the following trial. We counted lose-shift errors as a bee experiencing the fluctuating resource in a poor state and then choosing the same resource in the following trial. Because different rates of sampling can influence how frequently an individual bee might have the opportunity to make each type of error, we then scaled the totals of each type of error for each bee by the number of times that error was possible. This then allows for a more independent measure of the tendency to make a specific type of error from the effects of sampling. RESULTS Choice of focal resource We first look at bee choices in terms of choosing the focal color (Figure 4). To assess the role of experience in choice, we used a repeated measures Anova, with a main effect of whether the focal color (human-orange) was fluctuating or steady in reward, a main effect of the level of preference difference (low, medium, or high), and a repeated measure of this choice across blocks of trials. We find significant effects of fluctuating/steady reward (F1,42 = 31.231, P = 0.000002), of preference level (F2,42 = 7.061, P = 0.00227), and of the interaction between these 2 factors (F2,42 = 7.145, P = 0.00224). Block is only significant within a 3-way interaction between reward type and preference level (F14,294 = 31.231, P = 0.0142). When the human-orange resource fluctuated in reward, regardless of the original difference in preferences strength, bees chose that human-orange resource relatively equally across the blocks. We do show a trend of switching away from the focal color when the preference difference was high to start with: this preference significantly declines across the blocks (contrast analysis comparing first 2 blocks to final 2 blocks: F1,42 = 6.059, P = 0.0180). However, when human-orange remains in a steady state, original preference differences are maintained for the high preference treatment and increased from baseline in the medium preference treatments (the proportion of bees choosing human-orange is above 0.8 across all trial blocks in these treatments). When there is no initial preference, we can see a trend in which bees are picking the color that contains the better resource following the reward shifts every 20 trials (every 2 blocks on the graph). Figure 4 View largeDownload slide Proportional choices of the bees on the focal color orange. The x axis reflects averages across blocks of 10 choices. The fluctuating reward changes every 20 choices starting in a bad state in trial block 1. Figure 4 View largeDownload slide Proportional choices of the bees on the focal color orange. The x axis reflects averages across blocks of 10 choices. The fluctuating reward changes every 20 choices starting in a bad state in trial block 1. Sampling window We next look at sampling window (Figure 5). Because sampling window is a measure across trials for each bee, we have a simple factorial Anova, with a main effect of fluctuating or steady reward, and of strength preference difference. We find that both main effects are statistically significant. Sampling windows are longer when the focal resource gives a steady reward (F1,42 = 7.352, P = 0.00967), and sampling windows tend to be longer the more preferred the resource is (F2,42 = 3.590, P = 0.03635). These effects approach a significant interaction (F2,42 = 3.038, P = 0.0585), with any difference in sampling windows emerging only when the focal resource gives a steady reward. Figure 5 View largeDownload slide Statistically significant interaction of preference level and reward type upon the sampling window (number of trials between sampling events). Figure 5 View largeDownload slide Statistically significant interaction of preference level and reward type upon the sampling window (number of trials between sampling events). Choice of fluctuating resource How preference and resource type affect bees’ choice of whichever resource is fluctuating? We present these data in Figures 6 and 7. We analyze these data with a repeated measures Anova, looking at choices across blocks of trials, with main effects of reward as fluctuating or steady, and the level of preference difference. We find significant effects of fluctuating/steady reward (F1,42 = 11.505, P = 0.00152), of preference level (F2,42 = 7.145 P = 0.00213), and of the interaction between these 2 factors (F2,42 = 7.0612, P = 0.00227; Figure 5). Block is only statistically significant within its interaction with preference level (F12,294 = 2.0562, P = 0.01415; Figure 7 shows this interaction). Bees with an equal preference between the colors of resources choose the fluctuating resource equally, regardless of whether the focal color is fluctuating or steady. This is not the case for bees choosing between colors for which there was a baseline preference difference. Here bees choose resources equally when their focal color is fluctuating, with neither treatment differing from 0.50 choice. However, bees show a very strong propensity to choose their preferred color when it is providing a steady reward. This is not completely reflecting a maintenance of that original color preference, because bees show a decreased propensity to choose the color associated with fluctuating reward across the blocks of trials in these medium and high preference treatments (Figure 7). Figure 6 View largeDownload slide Significant interaction of preference level and reward type on the proportion of choices of the fluctuating resource. Figure 6 View largeDownload slide Significant interaction of preference level and reward type on the proportion of choices of the fluctuating resource. Figure 7 View largeDownload slide Full interaction of preference level, reward type and experience (blocks of trials). The interaction of block with preference type is statistically significant. Figure 7 View largeDownload slide Full interaction of preference level, reward type and experience (blocks of trials). The interaction of block with preference type is statistically significant. Errors The types of errors the bees make give a window in how they are valuing the resources. We analyzed the win-stay and lose-shift errors separately, looking at the total proportion of errors made of each type across all trials for each bee. We analyzed these values using an Anova with main effects of preference level (equal, medium, or high) and resource reward type (fluctuating or steady). For win-stay errors we find statistically significant effects of preference level (F2,36 = 3.999, P = 0.02701), reward type (F1,36 = 10.97, P = 0.00212), and of the interaction between the 2 (F2,36 = 5.6607, P = 0.00728; Figure 8a). Bees make more win-stay errors when the focal color fluctuates in reward and when there is no difference in color preference. Win-stay errors become rarer for favored colors that are not changing in value. We find different results with the pattern of lose-shift errors, in that here, preference level is not statistically significant (F2,41 = 0.585, P = 0.5617), although it does significantly interact with reward pattern (F2,41 = 5.0672, P = 0.0108; Figure 8b). Reward pattern alone is statistically significant (F1,41 = 7.3405, P = 0.00980). We do not see an effect of the pattern of reward when bees are faced with a choice between more equal preference colors. Differences emerge as preference levels increase, with bees failing to shift from a loss more frequently when the focal color gives a reward that is steady in value. Figure 8 View largeDownload slide Errors of learning. (a) In a fluctuating patch, the proportion of bee choices to stay with the high rewarding color once sampled (win-stay), as opposed to choosing to the color providing the steady (win-shift), but mediocre resource. The x axis represents the state of the focal color. (b) In a fluctuating patch, the proportion of bee choices to abandon a high rewarding color once sampled and instead return to the color providing the steady, mediocre resource. The x axis represents the state of the focal color orange. Figure 8 View largeDownload slide Errors of learning. (a) In a fluctuating patch, the proportion of bee choices to stay with the high rewarding color once sampled (win-stay), as opposed to choosing to the color providing the steady (win-shift), but mediocre resource. The x axis represents the state of the focal color. (b) In a fluctuating patch, the proportion of bee choices to abandon a high rewarding color once sampled and instead return to the color providing the steady, mediocre resource. The x axis represents the state of the focal color orange. DISCUSSION Our results show that bees tested exhibited a sensory bias towards the human color-orange when compared to blue (medium preference) or purple (high preference) and, equal preference when exposed to yellow. This color preference was further demonstrated by a small sampling window when the focal color was presented as the fluctuating resource. However, sampling windows became longer when the focal color was the steady resource. The exception to both of these changes were when there was no preference; here bees sampled more and at shorter intervals when the resource of their preferred color fluctuated, less when it offered steady rewards and, randomly in the absence of an unlearned (equal) preference. Although we noted that bees demonstrated this bias towards human-orange, consistency of reward offered compounded decision making. This resulted in bees choosing other colors when the focal (preferred) color offered rewards that changed over time. We also found that bees were more likely to stay when they encounter a steady reward paired with their preferred color (human-orange) even if the reward offered is better on the less-preferred option. This was mirrored by the win-stay errors where the bees were more likely to abandon a good state and return to a mediocre state when that good state was not paired with their preferred color. In the absence of an unlearned (equal) preference, sampling was random regardless of if the focal color was fluctuating or steady. Overall, we noted that bees avoided the fluctuating resource when their preferred color offered steady rewards, but this preference disappeared once the originally preferred color offered changing rewards (Figures 6 and 7). Thus, initial preferences were only maintained when the accompanying reward of that stimulus was predictable, and not when it was changing in value even though the average reward was the same. Flowers send out a variety of signals which may indicate varying rewards (Raine and Chittka 2012). Through sensory biases that elicit innate behavioral responses and learning from information gathered through time, animals can regulate the wide variety of plant signals to assess their relative importance (Raine and Chittka 2007) and even reduce uncertainty (Koops 2004; Dall et al. 2005). The benefit of either source of information is based on environmental heterogeneity with innate behavioral responses being optimal during unchanging environmental states where reward cues are stable over short-term. While learning, that is, personal information gathered by sampling, is best in a moderately changing environment, as this information can be used to update the current state of the environment and inform on not too distant future decisions (Mery and Burns 2010). As an animal learns new information it must suppress old and sometimes conflicting information while a new association is formed (reversal learning paradigm) (Raine and Chittka 2012). We found that across the bees we tested in the initial preference trials, individuals showed a preference for human-orange in comparison to others presented, blue (medium preference) or purple (high preference) (Figure 1). Given that inexperienced bees were tested, this reflects their innate sensory bias. This sensory bias may have resulted from reproductive processes that facilitate the transfer of forging information from one generation to the next (Jones et al. 2015). We would predict that this information reflects color or brightness traits of the most profitable flowers experienced by generations past (Raine and Chittka 2007), although given that the bees we tested have been bred in captivity for many generations, it is difficult to make a direct link to ecology in this case. Regardless, during the lifespan of an individual information inherited is weighted against and new information gathered from the environment (Koops 2004; Wagner and Danchin 2010; Danchin 2013) and this will influence choice. As such a worker bee’s choice of focal plant species reflects both unlearned preferences and learned information about current reward levels gained through individual sampling (Kawaguchi et al. 2006; Leadbeater and Chittka 2007b; Leadbeater and Chittka 2009). Moreover, we noted that although many studies have shown bees preference for blue (Gumbert 2000; Briscoe and Chittka 2001; Raine and Chittka 2007) preference data across a number of our pilot studies have shown us that the preference for blue is greatly affected by the brightness of each color involved in the choice, so a darker blue can be less preferred than a bright yellow or orange. For instance, a light and bright blue was much more highly preferred than the human-orange. In our study, we carefully chose our 3 color pairings so as to reduce the risk of “special pairings” as seen in Worden and Papaj (2005) where the bees innate bias for orange affects learning on green. We also chose color pairings did not result in a large amount of individual variability in initial preference. In the real-world information gained by sampling incurs costs such as energy expenditure and exposure to predators. Therefore, it is important that animals balance these costs with benefits incurred (Bell 1990). Thus, foragers are constantly faced with the decision of how often to sample in an attempt to avoid sampling too much or too little (Dunlap and Stephens 2012). In a perfect world, bees will always choose the highest reward after sampling regardless of their biases. As Fortin (2003) noted with sheep, foragers will become more selective when they consumed high-quality food as information collected while foraging can influence subsequent decisions (Fortin 2003). As bees sampled and individuals gathered new information which informed subsequent choices we found that they abandoned their color bias when other colors reliably offered greater rewards, that is, bees will choose a color that presents high rewards in a steady state over a color it prefers that offers alternating food sources. In our study, we specifically noted that bumble bees were more likely to continue foraging on the same color, when that color was preferred even if the reward offered is better on the un-preferred option. However, this decision was affected by the consistency of that reward. Therefore, we found that preference levels and effects of fluctuating/steady rewards were statistically significant as inherited biases influenced frequency of sampling, specifically, bumble bees would make more sampling choices at shorter intervals if their preferred color offered fluctuating rewards (Figures 4 and 5). The observation of pollinators selecting flowers that reduce uncertainty of rewards was also noted by Caraco (1980) and Real (1980). In nature, this behavior would be of adaptive value as it maximizes possible rewards and minimizes the likelihood that the forager will select poor rewards (Real et al. 1982). Our results also corroborate the findings of reversal learning in bumble bees, as shown by researchers such as Chittka (1998), Raine and Chittka (2012) and Strang and Sherry (2014) and as noted in nature (Sherry and Strang 2015). Similarly, evidence of reversal learning was also found in honeybees by Von Helversen (1974) (as cited in Strang and Sherry 2014) and Dyer et al. (2014), with Dyer et al. (2014) aptly classifying bees as “fickle-circumspect,” “deliberative-decisive” or “stays.” The first would incrementally adjust their preferences when new evidence is gathered from their feeding environment, the second would change its preference, if that preference becomes unrewarding or there is a more rewarding option and “stays” remained loyal to their preference (Dyer et al. 2014). Although, we observed a range of similar behaviors in our study animals, overall, regardless of preference we noted behavioral flexibility as is seen in the wild despite the bumblebee’s florally constant nature (Sherry and Strang 2015). In nature, foragers keep track of change nectar sources via environmental cues that indicate changes in reward levels (Weiss 1991), or in the absence of such cues, the forager will sample various floral sources in its habitat (Shettleworth et al. 1988). Naïve bees use innate rules to find food, as such sensory biases towards particular colors are useful when first exploring the world as these biases are often associated with the most rewarding flowering (Raine and Chittka 2007). And although learning is advantageous for bumble bees in nature (Raine and Chittka 2012) they must overcome their biases to make new associations. We also know that bumble bees are greatly affected by patterns of reward; for instance a long history of studies on risk sensitivity in bees shows that if the reward is high enough (e.g., Harder and Real 1987), or food stores are low enough (Cartar and Dill 1990), bees will choose resources with variable rewards. Otherwise, they display risk-averse behaviors and avoid a resource with a variable reward. In our experiment, we find that how bees react to a resource with a changing reward interacts with their initial bias for the color cue of the resource. Bumble bees in our study exhibit changes in their propensity to choose the fluctuating resource in the presence of innate preferences (Figures 6 and 7). However, when preference in equal, we also find no strong effect of which color varies in reward and which is changing. While such a result may always be due to a lack of visual discrimination between the options, in this case we do know this species can discriminate the orange and yellow color pairing from a previous work using these identical color stimuli (Dunlap et al. 2016). It is also possible that in a close color discrimination task, learning may be required for bees to better discriminate the colors (e.g., Dyer and Chittka 2004); in this study, bees may not have been able to quickly learn such a discrimination due to some aspect of the fluctuating rewards for one resource. When there is a color bias, bees will continue to forage on the color preferred when it offers constant rewards. However, if the color of preference offers changing rewards the bee will continue to forage on this color even though rewards are variable. This was corroborated by our win-stay and lose-shift scenarios (Figure 8a,b) where we found that in the absence of an initial preference, a bee is more likely to stay on the fluctuating color after experiencing a good reward whether or not it is a preferred color. However, when bees are offered a color to which they show a bias in comparison to less-preferred colors, bees are only more likely to stay when that color is preferred. Therefore, in the high and medium preferences the likelihood for a bee to stay with the high rewarding color is higher when it is the preferred color. We do acknowledge the merit of using more complex decision rules to investigate bumblebee foraging choices and the drawbacks of using the win-stay and lose-shift approach only as color shift probabilities vary with differences in rewards offered and experience (Keasar et al. 2002). However, by using the win-stay and lose-shift approach we are able to look at sampling when the best reward varies, as it does in this study and to look at the pattern of choices and when bees are switching or not switching to the other resource. It also specifically allows us to focus on situations where bees remain on a resource despite experiencing a lower value or leave a resource despite a good reward. Overall, bumble bees sampled more when preferred options offered unstable rewards. However, they were less likely to resample their newly learned less-preferred option even if it offered more rewards. Therefore, we found that inherited biases greatly affect decision making in bumble bees as they generally only choose options that must be learned if their innately preferred resource is paired with an unreliable reward, that is, they allocate more weight to innate biases than individual experience when those biases offer steady rewards. In our study, we focused on broad categories of equal, medium and high preference levels paired with fluctuating and steady rewards. Although we were able to investigate how preference interacted with rewards we were unable to determine the critical threshold points at which at which bumble bees will abandon their innate preferences. These thresholds are likely to be present both in terms of levels of preference between equal and medium, and in terms of different levels of potential rewards. In the future, we would like to further look at this interaction in more detail to determine the critical preference threshold points. As this will elucidate at what juncture do animals abandon past information, inherited biases, for information gathered by personal experience in the present. FUNDING We thank Ben Abts, Darshan Patel, and Stephanie Wooten for assistance with bumble bee husbandry and data collection and Avery Russell for help with color calculations. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Maharaj et al. (2018). REFERENCES Bell WJ . 1990 . Searching behavior patterns in insects . Annu Rev Entomol . 35 : 447 – 467 . 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 Caraco T . 1980 . On foraging time allocation in a stochastic environment . Ecol . 61 : 119 – 128 . Google Scholar Crossref Search ADS Cartar RV , Dill LM . 1990 . Why are bumble bees risk-sensitive foragers ? Behav. Ecol. Sociobiol . 26 : 121 – 127 . Google Scholar Crossref Search ADS Chittka L . 1992 . The color hexagon: a chromaticity diagram based on photoreceptor excitations as a generalized representation of color opponency . J Comp Physiol A . 170 : 533 – 543 . Chittka L . 1998 . Sensorimotor learning in bumblebees: long-term retention and reversal training . J. Exp. Biol . 201 : 515 – 24 . Dall SR , Giraldeau LA , Olsson O , McNamara JM , Stephens DW . 2005 . Information and its use by animals in evolutionary ecology . Trends Ecol Evol . 20 : 187 – 193 . Google Scholar Crossref Search ADS PubMed Danchin É . 2013 . Avatars of information: towards an inclusive evolutionary synthesis . Trends Ecol Evol . 28 : 351 – 358 . Google Scholar Crossref Search ADS PubMed Dunlap AS , Nielsen ME , Dornhaus A , Papaj DR . 2016 . Foraging bumble bees weigh the reliability of personal and social information . Curr Biol . 26 : 1195 – 1199 . Google Scholar Crossref Search ADS PubMed Dunlap AS , Papaj DR , Dornhaus A . 2017 . Sampling and tracking a changing environment: persistence and reward in the foraging decisions of bumblebees . Interface Focus . 7 : 20160149 . Google Scholar Crossref Search ADS PubMed Dunlap AS , Stephens DW . 2012 . Tracking a changing environment: optimal sampling, adaptive memory and overnight effects . Behav Processes . 89 : 86 – 94 . 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 , Dorin A , Reinhardt V , Garcia JE , Rosa MG . 2014 . Bee reverse-learning behavior and intra-colony differences: simulations based on behavioral experiments reveal benefits of diversity . Ecol Model . 10 : 119 – 131 . Google Scholar Crossref Search ADS Ellis AG , Johnson B . 2012 . Pollinator mediated floral divergence in the absence of pollinator shifts . In: Patiny S , editor. Evolution of plant-pollinator relationships . New York (NY) : Cambridge University Press . p. 237 – 262 . Endler JA , Basolo AL . 1998 . Sensory ecology, receiver biases and sexual selection . Trends Ecol Evol . 13 : 415 – 420 . Google Scholar Crossref Search ADS PubMed Fortin D . 2003 . Searching behavior and use of sampling information by free-ranging bison (Bos bison) . Behav Ecol Sociobiol . 54 : 194 – 203 . Goulson D , Cruise JL , Sparrow KR , Harris AJ , Park KJ , Tinsley MC , Gilburn AS . 2007 . Choosing rewarding flowers; perceptual limitations and innate preferences influence decision making in bumblebees and honeybees . Behav Ecol Sociobiol . 61 : 1523 – 1529 . Google Scholar Crossref Search ADS Gumbert A . 2000 . Color choices by bumble bees (Bombus terrestris): innate preferences and generalization after learning . Behav Ecol Sociobiol . 48 : 36 – 43 . Google Scholar Crossref Search ADS Harder LD , Real LA . 1987 . Why are bumble bees risk averse ? Ecol . 68 : 1104 – 1108 . Google Scholar Crossref Search ADS Jones PL , Ryan MJ , Chittka L . 2015 . The influence of past experience with flower reward quality on social learning in bumblebees . Anim Behav . 101 : 11 – 18 . Google Scholar Crossref Search ADS Kawaguchi LG , Ohashi K , Toquenaga Y . 2006 . Do bumble bees save time when choosing novel flowers by following conspecifics ? Funct Ecol . 20 : 239 – 244 . Google Scholar Crossref Search ADS Keasar T , Rashkovich E , Cohen D , Shmida A . 2002 . Bees in two-armed bandit situations: foraging choices and possible decision mechanisms . Behav Ecol . 13 : 757 – 765 . 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 . 1 – 14 . doi: [PubMed] Koops MA . 2004 . Reliability and the value of information . Anim Behav . 67 : 103 – 111 . Google Scholar Crossref Search ADS Leadbeater E , Chittka L . 2007a . Social learning in insects–from miniature brains to consensus building . Curr Biol . 17 : R703 – R713 . Google Scholar Crossref Search ADS Leadbeater E , Chittka L . 2007b . The dynamics of social learning in an insect model, the bumblebee (Bombus terrestris) . Behav Ecol Sociobiol . 61 : 1789 – 1796 . Google Scholar Crossref Search ADS Leadbeater E , Chittka L . 2009 . Bumble-bees learn the value of social cues through experience . Biol Lett . 5 : 310 – 312 . Google Scholar Crossref Search ADS PubMed Lunau K , Maier EJ . 1995 . Innate color preferences of flower visitors . J Comp Physiol a-Sensory Neural Behav Physiol . 177 : 1 – 19 . Maharaj G , Horack P , Yoder M , Dunlap AS . 2018 . Data from: influence of preexisting preference for color on sampling and tracking behavior in bumble bees . Dryad Digital Repository. http://dx.doi.org/ https://doi.org/10.5061/dryad.21qb8m0 Menzel R . 1985 . Learning in honey bees in an ecological and behavioral context . In: Hölldoboler B , Lindauer M , editors. Experimental behavioral ecology and sociobiology . Sunderland (MA) : Sinauer Associates . p. 55 – 77 . Mery F , Burns JG . 2010 . Behavioural plasticity: an interaction between evolution and experience . Evol Ecol . 24 : 571 – 583 . Google Scholar Crossref Search ADS Müller H . 1881 . Alpenblumen . Leipzig (Germany) : Engelmann . Pyke GH , Pulliam HR , Charnov EL . 1977 . Optimal foraging: a selective review of theory and tests . Q Rev Biol . 52 : 137 . Google Scholar Crossref Search ADS Raine NE , Chittka L . 2007 . The adaptive significance of sensory bias in a foraging context: floral colour preferences in the bumblebee Bombus terrestris . PLoS One . 2 : e556 . Google Scholar Crossref Search ADS PubMed Raine NE , Chittka L . 2008 . The correlation of learning speed and natural foraging success in bumble-bees . Proc Biol Sci . 275 : 803 – 808 . Google Scholar Crossref Search ADS PubMed Raine NE , Chittka L . 2012 . No trade-off between learning speed and associative flexibility in bumblebees: a reversal learning test with multiple colonies . PLoS One . 7 : e45096 . Google Scholar Crossref Search ADS PubMed Real L . 1980 . Fitness, uncertainty and the role of diversification in evolution and behavior . Am Nat . 115 : 623 – 638 . Google Scholar Crossref Search ADS Real L , Ott J , Silverfine E . 1982 . On the tradeoff between the mean and the variance in foraging: effect of spatial distribution and color preference . Ecol . 63 : 1617 – 1623 . Google Scholar Crossref Search ADS Schiestl FP , Johnson SD . 2013 . Pollinator-mediated evolution of floral signals . Trends Ecol Evol . 28 : 307 – 315 . Google Scholar Crossref Search ADS PubMed Sherry DF , Strang CG . 2015 . Contrasting styles in cognition and behaviour in bumblebees and honeybees . Behav Processes . 117 : 59 – 69 . Google Scholar Crossref Search ADS PubMed Shettleworth SJ . 1989 . Animals foraging in the lab: problems and promises . J Exp Psychol Anim Behav Process . 15 : 81 – 87 . Google Scholar Crossref Search ADS Shettleworth SJ . 2010 . Cognition, evolution, and behavior . UK : Oxford University Press . Shettleworth SJ , Krebs JR , Stephens DW , Gibbon J . 1988 . Tracking a fluctuating environment: a study of sampling . Anim Behav . 36 : 87 – 105 . Google Scholar Crossref Search ADS Skorupski P , Chittka L . 2010 . Photoreceptor spectral sensitivity in the bumblebee, Bombus impatiens (Hymenoptera: Apidae) . PLoS One . 5 : e12049 . Google Scholar Crossref Search ADS PubMed Spaethe J , Tautz J , Chittka L . 2001 . Visual constraints in foraging bumblebees: flower size and color affect search time and flight behavior . Proc Natl Acad Sci U S A . 98 : 3898 – 3903 . Google Scholar Crossref Search ADS PubMed Stephens DW . 1987 . On economically tracking a variable environment . Theor Popul Biol . 32 : 15 – 25 . Google Scholar Crossref Search ADS Stephens DW . 2008 . Decision ecology: foraging and the ecology of animal decision making . Cogn Affect Behav Neurosci . 8 : 475 – 484 . Google Scholar Crossref Search ADS PubMed Stephens DW , Krebs JR . 1986 . Foraging theory . USA : Princeton University Press . Strang CG , Sherry DF . 2014 . Serial reversal learning in bumblebees (Bombus impatiens) . Anim Cogn . 17 : 723 – 734 . Google Scholar Crossref Search ADS PubMed Von Helversen O . 1974 . Das experiment: dressurversuche mit bienen . Biol unserer Zeit . 4 : 119 – 124 . Google Scholar Crossref Search ADS Wagner RH , Danchin É . 2010 . A taxonomy of biological information . Oikos . 119 : 203 – 209 . Google Scholar Crossref Search ADS Weiss MR . 1991 . Floral colour changes as cues for pollinators . Nature . 354 : 227 – 229 . Google Scholar Crossref Search ADS Weiss MR . 1997 . Innate colour preferences and flexible colour learning in the pipevine swallowtail . Anim Behav . 53 : 1043 – 1052 . Google Scholar Crossref Search ADS Worden BD , Papaj DR . 2005 . Flower choice copying in bumblebees . Biol Lett . 1 : 504 – 507 . Google Scholar Crossref Search ADS PubMed © 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 - Influence of preexisting preference for color on sampling and tracking behavior in bumble bees
JO - Behavioral Ecology
DO - 10.1093/beheco/ary140
DA - 2019-03-04
UR - https://www.deepdyve.com/lp/oxford-university-press/influence-of-preexisting-preference-for-color-on-sampling-and-tracking-1jYEhSsvjE
SP - 150
VL - 30
IS - 1
DP - DeepDyve
ER -