1 Introduction

1.1 Bumble bee foraging plays a crucial role in the stability of natural and agricultural ecosystems

As generalist buzz-pollinators, bumble bee foragers are vital in both natural and agricultural systems (Corbet et al. 1981; Heemert et al. 2015; Hegland and Totland 2008; Klein et al. 2008; Mänd et al. 2002; Motten 1986). The observed population decline of many bumble bee species thus begets concern for the long-term stability of these ecosystems (Cameron and Sadd 2019; Goulson et al. 2008; Grixti et al. 2009). Bumble bee declines have been linked to numerous factors, including but not limited to habitat loss, disease, and pesticide exposure (Goulson et al. 2015). In addition to direct pressures on bumble bee populations, there are indirect effects via modulation of sensory signals used by bumble bees while foraging. For example, air pollution can alter the odors bumble bees will encounter on foraging bouts (Fuentes et al. 2013, 2016; Leonard et al. 2018; Lusebrink et al. 2015; McFrederick et al. 2008). The ecological and conservation impact of anthropogenic changes in sensory signals depends on how reliant bumble bees are on those signals during foraging. Given that bumble bee foraging is critical to both pollination services and colony fitness (Whitehorn et al. 2012), a thorough understanding of bumble bee foraging behavior is a critical component of creating a sustainable anthropocene.

1.2 Bumble bee foraging is a complex, multi-phase behavior

Consider encountering a bumble bee on a flower. Is this its maiden foraging voyage, or is it a seasoned worker? Is it learning how to extract nectar from a new type of flower, or has it returned to that very flower every day for the past 5 days? When it flies off, will it continue to a nearby flower, move to the next patch, search for a new patch of flowers, or return to its colony? The idea that bumble bee foraging is not monolithic, but rather an assemblage of sub-behaviors that rely on their own sensory information and experiential context, is not novel. Authors already seek to contextualize their studies in terms of which foraging situations are relevant, despite the lack of a formal vocabulary in the literature (Burns 2005). To synthesize, integrate, and unify the extensive literature on bumble bee foraging, we propose a framework that considers behavioral state and state transitions in the context of spatial scale and experience. Specific behaviors will emerge from convergence of these three dimensions, informed by and determining relevant sensory information (Figure 1). For example, consider an experienced forager with full pollen baskets that departs a flower and begins to navigate back to their colony. If that individual encounters an innately attractive floral odor plume, they may not be receptive to the signal — ignoring it and continuing on their return journey. However, a naive forager that has just left the hive and is hunting for resources is likely to be very receptive to that same odor signal. There is a staggering amount of additional work needed to fully deconstruct the relationship between neural processing and behavior — however, as larger bodies of knowledge are acquired and synthesized, we can locate support for the idea of a neural state that underlies emergent behavior. This shows up in physiological work looking at the effect of neuromodulatory signals on learning and sensory sensitization (Dacks et al. 2012; Farooqui et al. 2003; Gage et al. 2013; Kloppenburg and Mercer 2008; Scheiner et al. 2006); and in more deeply conceptual discussions of the relationship between appetitive motivation and behavioral performance (Baracchi et al. 2017).

Figure 1.
figure 1

A naïve bee completing a single foraging trip will execute multiple discrete foraging phases.

Full size image

1.3 Beyond bumble bees: commonalities across central place foragers

Given the broad similarities in ecology and life history across central place foragers, as well as the common phylogenetic origin of hymenoptera, this framework could reasonably apply to other species, such as honey bees and ant spp. However, we should not assume that states and state transitions will manifest identically across species. While ants, bumble bees, and honey bees are eusocial with foragers provisioning a central colony to support a reproductive queen/s, there are real differences in their life history and behavior that will show up as variability in strategy or relevant sensory information in this framework. For example, scent-trails in ants facilitate a type of navigation, route-following, that is not accessible to their flying relatives (Wolf 2011). The principal review project here is focused on bumblebees, but applicability to honey bees and other central place foragers is acknowledged and discussed.

1.4 Deconstructing foraging behaviors into states and related state-transitions

This framework for foraging behavior provides both a review of existing bumblebee literature in the context of its underlying complexity, and proposes commensurate terminology. In addition, a clearer understanding of foraging phases will help us better understand bumble bee foraging as a whole; and, in particular, how it is challenged by our anthropogenic world.

We propose three major behavioral states: search (S), acquisition (A), and navigation (N); and their commensurate state-transitions: initiation of search or navigation (Ø → S, Ø → N), transition from search or navigation to acquisition (S → A, N → A), transition from acquisition to search or navigation (A → S, A → N), and conclusion of navigation or search (N → Ø, S → Ø). A specific foraging-phase is emergent from a behavioral state or state-transition in the context of scale and experience (Figure 1, Table I). The structure of the framework’s notation emphasizes this, with the behavioral state/state-transition as the “function” acting on scale and experience as the “inputs.” For example, an experienced forager that is actively collecting pollen from a flower would be notated A(l,e) where “A” indicates acquisition, “l” indicates a local spatial scale, and “e” indicates an experienced forager. This acknowledges the real similarities in relevant sensory information utilized by foragers in the same state/transition, while allowing room for understanding the differences that variable contexts might provide. This classification foregrounds the neuroethological differences between an experienced “forager” returning to the colony with a full crop and a first-time “forager” leaving the nest with a preference for a floral-odor that has recently been deposited in the colony’s honeypots (Molet et al. 2008a). These two behaviors, currently carrying the same descriptor, rely on different sensory information and are driven by different motivations. Not all forager-phase distinctions will be so clear as this example; and not every forager will proceed through these behaviors identically, nor will every forager meet our “definitions” of a phase. However, the idiosyncrasy of individuals’ behaviors does not diminish the utility of scaffolding how foraging operates on a population level.

Table I This table lists out individual foraging phases based on their values in each of the three framework dimensions (behavioral states and transitions, spatial scale, and background). The last column lists vocabulary currently used to reference those foraging behaviors. Many current terms encompass multiple discrete phases. The second-to-last column indicates what type of sensory information is likely to drive or motivate a behavior (before the /) as well as what type of sensory information an individual in that phase might be sensitive to (after the /). Sensory information with a “1” is based on experimental evidence from citations discussed in the text for at least one model organism, while sensory information marked “2” is hypothesized or inferred from discussed citations but lacking direct experimental evidence. Innate attraction is typically marked with.“2” because experimental evidence is used to answer what floral characteristics are innately attractive, and their role in motivating behavior is inferred. The phase marked with an * would only be applicable to honey bees, due its reliance on waggle dance communication
Full size table

1.5 Why aren’t states and state-transitions sufficient on their own? The role of forager experience

The impact of forager experience on behavior is a well-studied phenomenon. Foraging experience impacts efficiency of resource acquisition (Klein et al. 2019; Peat and Goulson 2005; Saleh and Chittka 2006), and may also influence what sensory information an individual uses in decision-making (Burns 2005; Molet et al. 2008a). A naive forager has not had the opportunity to learn rewarding floral cues or floral handling skills, and is thus likely to rely on innate preferences. In contrast, an experienced forager is more likely to select floral cues that have previously been rewarding (Cartar 2004). In addition, they typically demonstrate higher efficiency, likely due to better proficiency at resource extraction from learned flowers/floral patches (Peat and Goulson 2005). Foraging experience is also likely to impact how a bumble bee navigates to and extracts resources from floral patches given their learning abilities and demonstrated tendency for floral constancy (Lihoreau et al. 2010; Makino 2013; Makino et al. 2007; Makino and Ohashi 2017; Makino and Sakai 2005, 2007; Ogilvie and Thomson 2015; Thomson et al. 2019). When exhibiting constancy for learned flower types, experienced foragers will benefit from increased speed of floral handling (Chittka and Thomson 1997; Ogilvie and Forrest 2017; Ogilvie and Thomson 2015) — although this phenomenon is not universal across central place foragers (Chittka and Muller 2009; Dornhaus 2008). In addition, bumble bees and honey bees have been found to demonstrate traplining in multiple contexts (Buatois and Lihoreau 2016; Lihoreau et al. 2010, 2012; Ohashi et al. 2007; Saleh and Chittka 2006; Woodgate et al. 2017), representing experiential modification of foraging behavior. However, it is important to recognize that while individual experience can influence foraging behavior, it is unlikely to dictate foraging behavior. Indeed, Saleh and Chittka noted inconsistencies within individual trapliners, despite behavioral trends that were significant on a population level (Saleh and Chittka 2006). Likewise, lifelong radar tracking of bumble bees showed that some bees remained strongly exploratory over their foraging career, rather than settling strongly into learned routes (Woodgate et al. 2016).

In addition to an individual’s lived experience, recent social experience is likely to be a strong modulator of foraging behavior in some states/transitions. The presence or absence of social cues (Pearce et al. 2017; Stout and Goulson 2001; Witjes and Eltz 2007), such as foraging pheromone (Molet et al. 2008a; Strube-Bloss et al. 2015), has been shown to impact individual bumble bees’ behavior. Likewise, social recruitment and communication between honey bees via the waggle dance (and accompanying odor cues) are a robustly explored phenomenon (Biesmeijer and Seeley 2005; Esch et al. 2001; Johnson and Wenner 2015; Naug and Gilley 2014; Samuelson et al. 2021; Thom et al. 2007).

Given that their behavioral will dramatically shift what sensory stimuli are relevant to an individual forager, as well as how stimuli are responded to, this framework incorporates a “Background” dimension with three conditions: naive (n), primed (p), and experienced (p). A “primed” bee represents an individual-forager who has received and responded to species-relevant communication about potential resources: waggle dances in the honey bee (Biesmeijer and Seeley 2005) and foraging pheromone in the bumble bee (Granero et al. 2005).

1.6 Spatial scale influences what sensory information is available to inform states and state-transitions

While some authors directly address the impact of experience on foraging decisions (Orbán and Plowright 2014, Avarguès-Weber et al. 2015), spatial scale is often missing from the conversation. For example, in a comprehensive review that starts with the sentence, “How do bees first find flowers?,” many of the studies reviewed are executed on a small spatial scale that is more relevant to local within-patch foraging than floral search (Orbán and Plowright 2014). The methodological difficulties of performing experiments on spatial scales that inform search behavior are directly addressed by the authors; but without having more nuanced vocabulary available, studies’ conclusions can be misinterpreted. In this framework, we consider spatial scale from the context of a forager’s ability to perceive a floral patch, leading to three broad categorizations: local (l), intermediate (i), and distant (d). “Local” is loosely construed as the physical distance over which a forager could reasonably encounter either a visual or an integrated visual-olfactory signal from a flower/flowering plant within the confines of its working memory, and is synonymous with a “patch” in this framework. This loose definition is bolstered by findings that the local floral density, on a scale of 100–900 m2, influences bumble bee visitation rates (Thomson 1981). The difference between two flowers within a single patch or two flowers in adjacent patches is not always obvious; this distinction varies depending upon the local floral landscape. The precise plant-density necessary to be considered a patch will be variable depending on the visual characteristics of relevant floral species as well as the forager’s visual resolution and flight speed. For a bumble bee with an angular resolution of 1.39° (Macuda et al. 2001) and a flight speed of 3 m/s (Heinrich 1976), a resolvable floral signal would need to be encountered within 6 m (Raine and Chittka 2007a). If this individual is foraging on a large flowered echinacea plant with a high number of blooms (Sprayberry 2018), inter-plant distances of 34.2 m would constitute a patch. However, a bumblebee with an angular resolution of 7° (Spaethe and Chittka 2003) feeding on a much smaller penstemon plant (Sprayberry 2018) would decrease that interflower distance to 6.1 m. For honey bees with an estimated working memory of 5 s (Zhang et al. 2005), a visual resolution of 5° (Ibarra et al. 2014), and a flight speed of 0.8 m/s (Baird et al. 2006), those same flowers would result in inter-plant distances of 11.8 and 4.2 m respectively. “Intermediate” would be a distance at which olfactory information alone is available. In contrast to the limited range of visual information from flowers, insect antennae have been measured responding to odor plume contact at distances of 20 m from the source (Murlis et al. 2000). Given the limited spatial resolution of hymenopteran visual systems (Horridge 2005) and the ability of odor information to travel a substantial distance from a flower, olfactory information will often be encountered in isolation (Sprayberry 2018). Floral scents are known to be attractive to honey bees (as reviewed by Srinivasan and Reinhard (2009)); and bumble bees are capable of locating food resources using odor information alone (Spaethe et al. 2007; Sprayberry et al. 2013). This implies that isolated odor information is behaviorally relevant to central place foragers. “Distant” is the distance from a floral resource at which no sensory information originating from the flower is likely to be available to foragers. At this range, we expect animals to be actively searching the environment for floral sensory information, or using additional cues to navigate to a known location (see “Search” and “Navigation” sections below).

Given that defining spatial context in this framework heavily depends on defining a patch, it is necessary to acknowledge that the spatial differentiation between a “flower” and a “patch” is not a stark divide. Indeed, in many situations, the difference between the two can only be determined by examining an individual forager’s history. For example, the first flower a bee visits in a patch would represent both patch and floral selection. When the bee departs, is it departing from the flower or the patch? This in turn begets the question: is the difference between a flower and a patch so esoteric that it need not be considered? In natural landscapes, floral resources can be spatially disparate, resulting in foragers flying long distances. For a bumble bee engaged in active search, you could imagine those distances being the consequence of a prolonged search period without floral encounter. In this scenario, the forager itself would not be changing its behavior relative to the spatial scale of the environment, rendering consideration of spatial scale irrelevant. However, search behavior is characterized by indirect or turning paths (Bartumeus et al. 2005; Spaethe et al. 2006; Viswanathan et al. 1999; Woodgate et al. 2017). Radar studies have shown that experienced foragers engage in straight flights to resources (Lihoreau et al. 2012; Osborne et al. 2013; Woodgate et al. 2017, 2016), a flight pattern that indicates intentional flight to a patch, rather than continuous searching behavior. This evidence of transit behavior is implicitly evidence for the behavioral relevance of patches, as bumble bee behavior acknowledges the distinction.

2 Foraging initiation (Ø → S, Ø → N)

Foraging initiation represents the transition into either a search state (Ø → S) or a navigation state (Ø → N). In terms of foraging, the impetus to leave the colony is likely motivated by both intrinsic factors such as the colony’s need for nutrients, as well as extrinsic factors such as social communication from returning foragers (Biesmeijer and Seeley 2005; Esch et al. 2001; Molet et al. 2008a, b). While the literature on conspecific communication in honey bees is far more prolific (see reviews by Couvillon (2012) and Slessor et al. (2005)), social communication, particularly between workers, is also germane to bumble bee behavior (Avarguès-Weber et al. 2018; Crall et al. 2018; Dornhaus and Chittka 2004; Dunlap et al. 2016; Granero et al. 2005; Leadbeater and Chittka 2011; Loukola et al. 2017; Molet et al. 2008a; Pearce et al. 2017; Rogers et al. 2013; Saleh et al. 2007; Stout and Goulson 2002; Strube-Bloss et al. 2015; Wilms and Eltz 2007). Returning foragers can disperse a recruitment pheromone (containing eucalyptol, ocimene, and farnesol) into the nest airspace, which increases the number of workers that subsequently depart the nest to forage (likely Ø → S(d,e/ p/ n)) (Granero et al. 2005; Witjes and Eltz 2007). Bumble bees also indirectly communicate information about rewarding food sources through floral odor cues: floral scent in the colony activates dormant-workers and biases post-departure resource selection, indicating they are leaving with an olfactory search image (likely Ø → S(d,p)) (McAulay et al. 2015; Molet et al. 2008a).

Nutritional status of a colony additionally appears to be an important factor in individuals’ likelihood of leaving to forage, scaling in response to nutritional needs (Pelletier and McNeil 2004; Weinberg and Plowright 2015). Even the response to recruitment pheromones can be influenced by colony nutritional-status (Molet et al. 2008b) implying that the response to neural identification of foraging pheromones is state-dependent. It is relevant to note that these studies were all performed on workers — there is no literature on how queens or males respond to scent cues encountered prior to colony departure.

3 Search (S)

Search behavior as a movement phenomenon is well studied across species and can be reasonably modeled with probabilistic movement distributions through a landscape (Bartumeus et al. 2005; Bartumeus and Catalan 2009; Edwards et al. 2007; Langrock et al. 2012; Reynolds et al. 2007). Data from multiple bumblebee species show that foragers travel from 500 m to 1.75 km — spatial scales that are much larger than a single meadow or patch (Hellwig and Frankl 2000; Lihoreau et al. 2012; Osborne et al. 1999, 2013). Honey bees have been shown to have even larger foraging ranges, reaching over 6 km (Beekman and Ratnieks 2000). Therefore, colonies are unlikely to restrict themselves to a single foraging-location over a lifetime. Bees are likely using sensory-guided search and navigation to discover spatially disparate resource patches. Given that the spatial resolution of bumble bees’ visual systems is unlikely to result in perception of flowering plants at distances of more than 20 m (Sprayberry 2018), visual components of floral cues are unlikely to be the primary or sole cue utilized in patch search and discovery. Multiple species of bumble bees have been shown in laboratory studies to be capable of utilizing floral odorants to locate food resources (Spaethe et al. 2007; Sprayberry et al. 2013) and social odor cues to locate feeders in darkness (Chittka et al. 1999b). In addition, modeling work indicates that increasing the strength of a floral odor cue increases the distance at which a bumble bee is likely to encounter sensory information from flowers (Sprayberry 2018). Collectively this implies that odor information is valuable for bees engaged in transit to a novel patch, which would require searching behavior (S(d)) that transitions to sensory-based navigation (N(i/l)) if a resolvable cue is encountered and acted upon (Figure 1).

A naive forager departing a colony for the first time is likely to engage in a learning-search at a small spatial scale to locate local landmarks and other sensory cues that indicate nest location (Collett et al. 2013b; Doussot et al. 2021; Ibarra et al. 2009; Lobecke et al. 2017; Philippides et al. 2013; Riabinina et al. 2014; Robert et al. 2018, 2016), reviewed by Collett and Zeil (2018), in order to prepare for larger scale search. At conclusion of this learning bout, if a forager does not return to their nest, they will likely transition to a sensory-based search as discussed above (S(d,n)). We would expect a forager in this phase to be responsive to innately attractive floral cues (Orbán and Plowright 2014), while odor-primed foragers (S(d,p)) might be predisposed to a particular floral scent (Molet et al. 2008a).

Upon patch localization, the spatial scale of search behavior shifts (S(l,n)) and the number of resolvable visual cues from flowers should increase substantially (Figure 2). As experience grows, the likelihood of a forager engaging in search behavior may also shift. As addressed in “navigation,” the ability of foragers to learn routes to resources can mean that experience reduces the need to search. So, when might an experienced forager search? The senescence of a learned resource could trigger a transition to searching at a large spatial scale to locate novel patches (S(d,e)) or at a smaller spatial scale for novel resources within a known patch (S(l,e)) (Figure 2). Ogilvie and Thomson demonstrated that bumble bees are more likely to stay faithful to a learned site when the active resource is removed and switch to foraging from a different floral species, rather than move to a new patch containing the original floral species (Ogilvie and Thomson 2016). A patch with enough resources over time — both throughout the day and a forager’s lifespan — could preclude the need for a forager to search for distant patches and increase the prevalence of S(l,e) phase foragers in that patch.

Figure 2.
figure 2

A Experience is likely to modify foraging behavior. The light grey circles around the focal flowers indicate the distance at which an odor plume is likely to be encountered. The dashed blue line indicates a naïve bee that searched and located a profitable resource, executing a behavior comprised of the phases S(d,n)» N(i,n)» N → A(l,n)» A(l,n)» S(l,n)» {S → A(l,n)» A(l,n)}*repeated» A→N(d,e)» N → Ø (d,e)» S → Ø (l,e). As that individual gains experience, it establishes more efficient route to exploit the patch, as represented by the solid purple line, and completes the behavior with a phase composition of N(d,e)» {N → A(l,e)» A(l,e)}*repeated» A→N(d,e)» N → Ø (d,e)» S → Ø (l,e). B Senescence of learned resources will result in experienced foragers entering back into search-base foraging phases, either locally (dotted purple line) or at a larger scale (dash-dot dark purple line).

Full size image

Interestingly, the vast majority of floral search studies in bumblebees have been performed at spatial scales less than or equal to 3.6 m (Dyer et al. 2008; Hudon and Plowright 2010; Ings et al. 2009; Kulahci et al. 2008; Leadbeater and Chittka 2005; Lunau 1992; Molet et al. 2008a; Renner and Nieh 2008; Sprayberry et al. 2013; Suchet et al. 2010; Wilms and Eltz 2007; Witjes and Eltz 2007). At these scales, a searching bumble bee is likely to rapidly encounter a resolvable floral signal (Sprayberry 2018)). Therefore, our current understanding of search behavior is primarily at a local spatial scale, indicating a need for ecologically relevant studies designed to target intermediate and distant scales.

4 Navigation (N)

Search has a random or semi-random component to animal movement, whereas navigation uses sensory information to guide the animal along a predictable trajectory. In terms of this framework that could include odor-guided navigation, waggle-dance informed navigation, and return to a learned destination. While lab and modeling studies imply that odor-guided navigation is likely in bees (Spaethe et al. 2007; Sprayberry 2018; Sprayberry et al. 2013), there is a paucity of studies directly testing their odor navigation strategies, a phenomenon that is much better characterized in other insects (see reviews by Carde and Willis (2008) and Murlis et al. (1992)). While we would expect navigation-phases to be the province of experienced bumblebees only, social communication in honey bees facilitates priming of both naive and experienced foragers to navigate through a novel environment based on the waggle-dance information passed along by nestmates (Esch et al. 2001). However, we can reasonably expect foragers to learn routes to known resources with experience, as numerous studies demonstrate traplining, revisitation of resources, and changes in flight trajectories over time (Lihoreau et al. 2010; Ohashi et al. 2007; Osborne et al. 2013; Woodgate et al. 2016). How insects in general, and bees in particular, navigate to distant locations is a richly studied field (see reviews by Collett and Zeil (2018), Srinivasan (2020), and Wolf (2011)). Seminal work by Srinivasan’s lab implicated the use of optic flow odometry in honey bees (Srinivasan et al. 1997), which contributes to path integration via the central complex (Collett et al. 2013a; Heinze et al. 2018; Moël et al. 2019). The demonstrated importance of optic flow in bumble bee flight control (Frasnelli et al. 2018) implies that the neural and neuromuscular substrates for a similar strategy are likely present. However, recent work on habitual route following found that neither optic flow or path integration adequately explained B. terrestris navigation behavior, implicating an integrated guidance system that employs both visual and movement memories (Bertrand et al. 2021).

5 Transitioning to acquisition: resource selection (S → A/N → A)

Selection represents a transition from either a search state or a navigation state to acquisition. The relevant sensory information is likely to be dependent on the precise nature of the transition (Table I). Transition from search (S → A) is relevant at multiple spatial scales and life history conditions. At distant spatial scales, this transition represents patch selection, while at local spatial scales, this represents floral selection — although it could be argued that the first flower selected in a patch holds both titles.

Naive bumble bees (n), naive honey bees (n), and primed bumble bees (p) should be locating resources via exploration of their environment. Therefore, we would expect those animals to transition into acquisition from search (S → A (d)) or olfactory-navigation (N(i)). Primed honey bees (p) and experienced bees (e) are more likely to transition from navigation at large spatial scales (N → A(d)).

Bees need floral resources (i.e., pollen and sugars) for both their own and colony survival (Pereboom 2000; Russell et al. 2017; Vaudo et al. 2016). Efficient foraging requires the selection of profitable flowers. For all foragers, encounter with and assessment of relevant sensory information is likely to trigger this moment of transition. There are many sensory cues that bees utilize during selection, but there is not one specific set of features that all pollinators rely on. Rather, it is a constellation of characteristics including but not limited to flower color and shape, floral odor, social cues, and memory (Orbán and Plowright 2014). There are substantial data indicating the importance of visual cues in regard to selection (Burns 2005; Dyer et al. 2008; Dyer 1998; Essenberg et al. 2015; Hudon and Plowright 2010; Keasar 2000; Kulahci et al. 2008; Marden 1984; Møller 1995). Bumble bee species have demonstrated innate color preferences, often in the blue/violet range of the spectrum (Chittka and Raine 2006; Ings et al. 2009). However, innate preferences can be overridden by experience, as has been shown for color, shape, and size (Arnold and Chittka 2019; Essenberg et al. 2015; Gumbert 2000; Hudon and Plowright 2010; Lunau et al. 1996; Lunau 1991; Møller 1995; Rodríguez et al. 2004), indicating that sensory preferences between (n) and (e) foragers could be quite different.

In addition to visual cues, flowers provide mechanosensory (Goyret 2006), gustatory (Ruedenauer et al. 2015), electrostatic (Clarke et al. 2013), and olfactory information (Lawson et al. 2018; Leonard et al. 2011; Odell et al. 1999; Reinhard et al. 2010; Schaeffer et al. 2019; Schiestl 2015; Srinivasan and Reinhard 2009). As with color, there is evidence that bumble bees have innate odor preferences (Renner and Nieh 2008; Suchet et al. 2010). In the selection phase, visual and odor cues will be encountered together; thus, multimodal studies are most relevant in understanding selection behavior. The presence of floral odor in addition to visual signals provides a more information-rich signal, enhancing floral discrimination and attraction (Kunze and Gumbert 2001; Lawson et al. 2017, 2018; Leonard et al. 2011). Redundancy in floral cues across sensory modalities might also facilitate recognition in a broader range of environmental conditions (Lawson et al. 2018).

A substantial body of work suggests that foraging experiences will, via associative learning, impact the selection behavior of bees. Indeed, numerous studies have found floral constancy in bees (Chittka et al. 1999a; Grüter et al. 2011; Lihoreau et al. 2010; Makino and Ohashi 2017; Makino and Sakai 2004, 2005; Miyake and Sakai 2005; Ogilvie and Thomson 2016; Thomson et al. 2019), a phenomenon that relies on associative learning. This in turn implies that naive foragers (n) operate with a different set of “rules” within a floral patch as they build up experience (e). The impact of experience on selection behavior is also evidenced by foragers regularly returning to a given patch. Multiple studies have shown that bumble bees are capable of traplining behavior, where an efficient path is set through a known patch and repeated on return visits (Burns 2005; Lihoreau et al. 2010, 2012; Ohashi et al. 2007; Saleh and Chittka 2006; Woodgate et al. 2017). In these situations (St.NtA), foragers appear to be selecting flowers from memory, rather than employing local sensory cues. In addition to learned routes and preferences, resource specialization likely impacts flower selection, with pollen specialists being more likely to select flowers for pollen harvesting and nectar specialists being more likely to focus on nectar acquisition (Hagbery and Nieh 2012; Russell et al. 2017).

In addition to the sensory information provided by flowers, social cues may play a strong role in floral selection. The most extensively studied social communication cue in bumble bees is scent marks. Scent marks on flowers seem to indicate recent exploitation of that resource, resulting in reduced selection by bumble bees (Goulson et al. 1998, 2000; Witjes and Eltz 2007). These scent marks appear to be passive footprint odors that are left behind when bumble bees forage (Saleh et al. 2007; Wilms and Eltz 2007) and are variable between individuals and colonies (Pearce et al. 2017). Electrostatic cues may also signal recent visitation by another individual via detection of a weakened floral-electric field (Clarke et al. 2013; Eskov 2018; Kaplan 2013; Koh et al. 2019; Koh and Robert 2020; Montgomery et al. 2019; Sutton et al. 2016). Avoidance of recently visited flowers should increase foraging efficiency. However, this avoidance is not universal: both experience and context impact how bumble bees respond to scent marks. Naive foragers do not show preference or avoidance when given the choice between marked and unmarked flowers, suggesting that experience is relevant in the interpretation of scent marks (Leadbeater and Chittka 2011; Saleh et al. 2007, 2006). Moreover scent marks can be interpreted differently depending on floral handling time (Saleh et al. 2007), with the repulsive effects of scent marks lasting longer on more complex flowers requiring longer handling times. Social transmission of information is not limited to flower rejection in bumble bees; positive information about what resources to seek and how to extract them is also utilized by foragers. Indeed, the presence of a conspecific can increase the attractiveness of a resource (Avarguès-Weber et al. 2018), and multiple studies have indicated that bumble bees can learn by visual observation of conspecifics (Alem et al. 2016; Loukola et al. 2017).

6 Acquisition (A)

For this framework, the behavioral state of acquisition encompasses the physical extraction of nutrients (nectar, pollen, floral oils, etc.). This is necessarily enacted after floral selection, and exclusively operates at small spatial scales (Table I). At some level, acquisition ability for a particular flower type is determined by forager anatomy: for example, long-tongued bumble bees can forage in flowers with deeper corollas than short-tongued bumble bees (Heinrich 1976; Lundberg and Ranta 1980). In addition to honest access via the floral limb, bumble bee species are known to “rob” nectar either by forming a hole at the base of the flower, or by accessing holes made by previous foragers (Irwin et al. 2010).

Foragers utilize a variety of sensory cues during nutrient acquisition. Visual information in the form of floral patterns encourages “legitimate access” of nutrients by bumble bees (Leonard and Masek 2014) and increases foraging efficiency by facilitating faster nectary localization (Leonard and Papaj 2011). Chemo-tactile and gustatory cues are used to evaluate the quality of resources, impacting acquisition and future choices (Brito-Sanchez et al. 2008; Gegear et al. 2007; Ruedenauer et al. 2015, 2016). The burgeoning field of nectar microbiology is exploring how bacteria and yeast within nectar impact chemosensory information passed along to bumblebees and subsequent nectar-feeding responses (Rering et al. 2017; Schaeffer et al. 2019). Unsurprisingly, given their capacity to learn, foraging experience has positive impacts on acquisition efficiency in bumble and honey bees — increasing pollen and nectar extraction rates (Heinrich 1979a; Klein et al. 2019; Raine and Chittka 2007b). Thus, we might expect the time spent in an acquisition state at a single flower to be shorter for an experienced bee (A(l,e)) as compared to a naive one (A(l,n)).

7 Transitioning to movement states: resource departure (A → S/ A → N)

Departure is the transition away from acquisition to either search or navigation (A → S or A → N). The spatial scale of forager behavior will differentiate between “floral departure” (local scale: l) or “patch departure” (distant scale: d). As addressed above, the definition of “flower” versus “patch” is also dependent on spatial scale: a flower in isolation is a patch unto itself, while a flower in close proximity to others of its kind would be considered a patch-component. In the second scenario, if a bee left a flower to travel to a nearby flower, we would consider that floral departure. However, if a bee left that flower and then moved past all other local resources for sights unknown, we would consider that a patch departure. The motivations for leaving either are likely rather similar. This begets the question under what conditions do bees choose to leave? Potential motivations for departure may include insufficient nectar and/ or pollen availability (Cibula and Zimmerman 1987), interactions with other bees (Rogers et al. 2013), or a need to deposit gathered resources in the hive (Cibula and Zimmerman 1987). The majority of foraging experiments in lab contexts are necessarily asking questions about floral departure, given that their spatial scale is small enough that a bumble bee can traverse the entire arena in a few seconds (Cibula and Zimmerman 1987; Dukas and Real 1993; Heinrich 1979b; Rogers et al. 2013). Foraging experiments in a field context are better able to ask specific questions about patch versus flower departure, although that requires the ability to measure post-departure flight paths. The methodological complexities involved in experimentally examining departure flights are comprehensively reviewed by Mola and Williams (Mola and Williams 2019). Much of the literature looking at the flower/patch departure is asking whether or not bumble bees are optimizing their foraging efficiency, as they forage in patches which often dynamically differ in quality. Foragers making decisions on when to leave a flower or patch based on local and recent information will likely have higher resource acquisition rates. Lab-based foraging studies find that recent experience (in terms of nectar acquisition rates) influences departure decisions in bumble bees, resulting in longer stays in rewarding areas (Cibula and Zimmerman 1987; Dukas and Real 1993). Likewise, queen bumble bees have been found to roughly adhere to a “win-stay, lose-leave” strategy while foraging (Hodges 1985). Existing data on bumble bees foraging in the field do not contradict these findings; overall they indicate that rewarding experiences result in longer stays at flowers and within a patch (Cartar 2004; Kato 1988). Although the use of “floral” versus “patch” departure in the literature is variable, motivations for both floral and patch departure are likely very similar — indicating that conclusions from departure studies may be broadly applicable.

8 Foraging conclusion (N → Ø, S→ Ø)

The transition out of foraging would be construed as “foraging conclusion.” This will typically manifest as a return to the hive, although individual bumblebees have been known to stay away from their colony overnight and resume foraging the next day (Woodgate et al. 2016). However, foraging conclusion will primarily operate as a transition out of the navigation state (N → Ø). At a large spatial scale (N → Ø(d,)), this will manifest as foragers traveling to the vicinity of their nest. The strategies for this are similar to those covered in the navigation section (reviewed by Wolf (2011)). At a local spatial scale (S → Ø(l,)), this would entail location and recognition of their nest. Colony location is learned via “learning flights” by departing bees (Collett and Zeil 2018; Doussot et al. 2020; Ibarra et al. 2009; Lobecke et al. 2017; Philippides et al. 2013; Riabinina et al. 2014; Robert et al. 2018), as discussed in the search section. What cues are learned on these early flights? Visual information such as landscape structure, landmarks, compass cues, and optic flow cues are known to be important (Brünnert et al. 1994; Collett et al. 2013b; Collett and Zeil 2018; Doussot et al. 2020; Ibarra et al. 2009; Kheradmand and Nieh 2019; Lobecke et al. 2017; Philippides et al. 2013; Riabinina et al. 2014; Zeil 2012). In addition to visual information, insects are also known to utilize olfactory cues in homing (Steck et al. 2009). Given that bumble bee nests contain olfactory markers (Rottler et al. 2013), it is reasonable to hypothesize that this could provide additional homing information.

9 Applications and implications

This framework codifies foraging phases based on the three dimensions of behavior states/transitions, spatial scale, and background, and subsequently identifies relevant sensory modalities for both motivating and modulating emergent behavior (Table I). However, this framework should not be read as a rigid structure capable of predicting the precise sequence of behaviors an individual forager will execute upon leaving its colony. Rather, the framework seeks to identify, on a population level, what neuroethological differences we might expect between different phases of foraging behavior. Patterns that are measurable but not universal at the population-level imply noise at the individual-level. This type of individual variability is demonstrated in the bumble bee literature. Heinrich’s canonical paper on bumble bee foraging found flight distance probabilities for individuals ranging from 50 to 80%, values that are meaningful in terms of a population but not predictive of an individual’s behavior (1979). Cartar’s tests of two subtly different foraging theories offer additional support for noisy adherence to foraging “rules”: their analysis showed statistical support for both theories, again implying that individual variability precludes universal rules of behavior (Cartar 2004). This begets the question, is this kind of framework a useful tool? The actions of an individual bee have low impact on colony fitness. Rather, the ecological impacts of behavior emerge from the actions of the collective population; therefore, understanding the emergent patterns and outcomes of this collective is extraordinarily relevant to the behavior, ecology, and conservation of bumble bees and other central place foragers. This type of systematic review allows a clearer picture of which foraging phases are best understood, which require a substantial increase in experimental focus, and which are most susceptible to anthropogenic disruption.

This approach reveals drastically understudied foraging phases. Many lab-based bumble bee studies are performed at scales of less than 1.2 m (Dukas and Real 1993; Kulahci et al. 2008; Rodríguez et al. 2004; Saleh and Chittka 2006). At this size, bumble bees will likely encounter the visual signal of a flower immediately upon entering an experimental arena, indicating a “local” scale. These means that the vast majority of published resource selection and search studies are relevant to floral selection (S → A(l,)/ N → A(l,)) and within-patch search (S(l,)) rather than patch selection (S → A(d/i,)/ N → A(d/i,)). Outliers in controlled foraging experiments reach spatial scales of up 8 m (Burns 2005) which ensures that bumble bees entering an arena are less likely to see flowers instantly, but would still be considered a “local” spatial scale given that an individual could cross the arena in under 3 s (Heinrich 2004). Given that habitat fragmentation disrupts pollinator resources (Goulson et al. 2010), forager search ability is likely crucial to colony provisioning and therefore fitness. Experimental disambiguation of search (S()) versus selection (S → A()/ N → A()) foraging phases is also relevant to floral ecology and reproductive success (Dohzono et al. 2008; Dukas 2005; Hegland and Totland 2008). Existing studies imply that olfactory information is likely pivotal for locating novel resources during search behavior (Spaethe et al. 2007; Sprayberry et al. 2013), but this needs to be confirmed with targeted experimental studies — particularly in a field context. This need is especially relevant given the growing body of work showing anthropogenic disruption of floral odor information (Fuentes et al. 2013, 2016; Girling et al. 2013; Lusebrink et al. 2015; McFrederick et al. 2008; Reitmayer et al. 2019; Ryalls et al. 2022). From the context of this framework, we might expect odor pollution to be most problematic for foragers in long distance searches (S(d,)) who may no longer recognize contaminated floral odorants, or are less likely to encounter a resolvable signal to trigger a transition into olfactory navigation (N(i,)). Until we have a better understanding of how odor cues operate in search behavior at field scales, and how relevant search success is to colony fitness, we will not adequately understand the impacts of anthropogenic odor signal modification on bumble bee conservation.