Using your nose to find your way: ethological comparisons between human and non-human species

Abstract

Olfaction is arguably the least valued among our sensory systems, and its significance for human behavior is often neglected. Spatial navigation represents no exception to the rule: humans are often characterized as purely visual navigators, a view that undermines the contribution of olfactory cues. Accordingly, research investigating whether and how humans use olfaction to navigate space is rare. In comparison, research on olfactory navigation in non-human species is abundant, and identifies behavioral strategies along with neural mechanisms characterizing the use of olfactory cues during spatial tasks. Using an ethological approach, our review draws from studies on olfactory navigation across species to describe the adaptation of strategies under the influence of selective pressure. Mammals interact with spatial environments by abstracting multisensory information into cognitive maps. We thus argue that olfactory cues, alongside inputs from other sensory modalities, play a crucial role in spatial navigation for mammalian species, including humans; that is, odors constitute one of the many building blocks in the formation of cognitive maps.

1. Introduction

In Western cultures, olfaction is often considered to be the least relevant of the senses, being the one sense that most people agree to sacrifice when posed with the question in the classroom or over cocktail party banter. Put differently, olfactory information is thought to be of low significance for goal attainment, especially when compared to the predominant role of the visual sense. Navigation is no exception to this rule. In the behavioral domain of spatial navigation, humans are generally characterized as purely visual navigators, a view that greatly undermines the potential impact of olfactory cues on human wayfinding. In fact, the inferiority of olfaction as a source of meaningful sensory information has been assumed and reinforced for centuries. Philosophers and scientists of the 19^th and 20^th century deemed vision to be the sense of reason and civilization, while the sense of smell was denounced to be of considerably lower order, and believed to be used only by the blind, the “savage,” or “primitive” peoples (Dunglison, 1841; Edinger, 1893; Gould & Pyle, 1900; Longet, 1860; Titchener, 1915). The following excerpt (published the same year as Edgar Allan Poe’s The Murders in the Rue Morgue, featuring a protagonist “savage” at the heart of the story (Poe, 1841)) illustrates the view characteristic of this era:

[…] As an intellectual sense, the smell is not entitled to a higher rank than the taste. Its mediate functions are very limited. It enables the chemist, the mineralogist, and the perfumer to discriminate bodies from each other We can, likewise, form a slight – but only slight – idea by it, regarding the distance and direction of bodies, owing to the greater intensity of odours near an odorous body, than at a distance from it. Under ordinary circumstances, the information of this kind, which we derive by olfaction, is inconsiderable; but in the blind, and in the savage, who are accustomed to exercise all their external senses more than the civilized individual, the sphere of utility and accuracy of this sense is largely augmented […] (Dunglison, 1841, p. 134).

Interestingly, the author suggests that direction and distance information can be derived from olfactory cues; however, such information was thought to be significant only to those who had considerable training using their sense of smell, either because they lost their sight or because their “uncivilized” lifestyle required it (Dunglison, 1841). Examples of the latter case often referenced “[…] savage tribes [who] with their large, open nostrils not only recognize their enemies but also track game the same as hounds […]” (Gould & Pyle, 1900, p. 398). The direct comparison of so-called “savages” to animal species, dogs in this example, illustrates the historically formative division between the animalistic olfactory sense and the “civilized” visual sense. Although centuries have passed since this formal division emerged, its main ideas persist even today, and bias scientific research in human subjects towards a focus on the visual, rather than the olfactory sense.

Spatial navigation is a problem that, in itself, does not favor a specific sensory domain. In fact, an organism may use any of the available sensory systems to estimate its physical location relative to other objects, and use this information to approach or avoid specific target objects in the environment. However, experimental studies performed on human subjects almost exclusively use visual cues to study the behavioral aspects, as well as the neural mechanisms, of spatial cognition and spatial memory (for an overview, see Ekstrom et al., 2018; Epstein et al., 2017; Schiller et al., 2015). Despite the narrow focus on visually-guided navigation in the existing scientific literature, it can be assumed that complex behaviors, including spatial navigation, exploit multisensory input (Pasqualotto & Proulx, 2012). And thus, even though humans may prioritize visual input during navigation, sensory information from other domains also shapes the process of wayfinding.

Among the sensory systems, olfaction is particularly well-suited to guide navigation behaviors. This is because the relevant stimulus, the odor, has specific features that provide important spatial information. First, odor intensity typically decreases with increasing distance from its source (Jacobs, 2012). This feature embodies the prerequisite of efficient navigation, as it allows for predictions of what is to be encountered next in a given environment. Second, olfactory cues can overcome physical and diurnal boundaries (e.g., walls, nighttime) in ways that visual cues cannot. Finally, the sense of smell is omnipresent, and olfactory functioning is preserved in critical periods of animal development. Even in humans, infants locate their mothers’ nipples by scent, indicating that olfactory cues can be used for spatial orientation at a very young age when visual cues are not yet available (Varendi et al., 1994).

Despite these arguments emphasizing the potential value of odorous cues in navigational tasks, their use in humans is not widely acknowledged within the scientific community. Studies investigating the behavioral aspects of human odor navigation are rare (Hamburger & Knauff, 2019; Jacobs et al., 2015; Porter et al., 2007), and only one experiment thus far has tested the neural mechanisms of the phenomenon (Bao et al., 2019). In contrast, literature on olfactory navigation in non-human species is abundant, and contains copious information on behavioral strategies as well as neural mechanisms that characterize the use of olfactory cues during spatial tasks. Using an ethological approach, the current review draws from scientific studies on olfactory navigation describing how different species have adapted to accomplish the navigational goal of source localization. There is strong evidence, dating back more than 70 years with Tolman’s seminal work (Tolman, 1948), that mammalian species can interact with the environment by integrating multisensory information into cognitive maps. Such maps, reliably capturing the spatial relationships between locations, can be used to navigate landscapes more flexibly (Wang et al., 2020). Thus, we argue that olfactory cues constitute one of the many building blocks in the formation of mental maps. Here, we highlight how the evolutionary development of sophisticated behavioral strategies has paralleled the development of complex neurophysiological processes and mechanisms, in the service of pathfinding and behavior. While much of this article focuses on navigational strategies in non-human species, in the final section of this review, we will provide an outlook on ongoing and future research aimed at revealing the behavioral strategies and neural mechanisms that support olfactory navigation in otherwise “civilized” humans, which in many ways resemble those in other mammals.

2. Odor-guided behavior in single cell organisms

Odor tracking has been relevant for survival since the earliest beginnings of life, roughly two to three billion years ago, and has retained a fundamental status throughout evolution. Organisms as simple as bacterial prokaryotes use chemotaxis (Table 1), that is, directional movement in response to chemical cues or gradients, to approach nutrients and avoid toxins (Adler, 1966; Adler, 1975; Gottfried & Wilson, 2011). Bacteria use a specific kind of serial sampling strategy (Table 1), the mechanisms of which are well understood (Berg, 2000; Bi & Sourjik, 2018), perhaps due to the simplicity of the organism itself. Motile bacteria perform a random walk, consisting of alternating states in which the organism either tumbles or runs. When attractants bind to a chemoreceptor on the membrane, runs are extended, thus biasing the random walk towards locations of higher concentration (Berg, 2000; Bi & Sourjik, 2018). Interestingly, the system mediating this behavior also comprises a negative feedback mechanism; this allows the bacterium to maintain its sensitivity to increasing concentrations of the attractant along the concentration gradient by regulating the sensitivity of the chemoreceptor (Falke et al., 1997). As this feedback mechanism works on a slightly slower time scale than the signaling itself, the bacterium is effectively endowed with a rudimentary form of short-term memory that is used to compare chemical concentrations from adjacent time frames several seconds apart (Bi & Sourjik, 2018). This temporal integration may not correspond to the mechanisms that other species have adopted to encode and recall odor intensity at past locations, but it powerfully illustrates the early evolutionary development of olfactory strategies that bring organisms closer to attractive stimuli and further away from repellent stimuli (Gottfried & Wilson, 2011).

Table 1.

Overview of navigation strategies.

Strategy		Mechanism
Stimulus response strategy Strategy by which the animal reaches a goal location through a specific behavioral response to a (changing) external stimulus	Chemotaxis Used to move to areas of high (low) odor concentration Used across a wide range of animals, from single cell organisms, such as bacteria (Berg, 2000; Bi & Sourjik, 2018), to humans (Porter et al., 2007), irrespective of their habitat	Sampling of odorants at different locations consecutively (“serial sampling strategy”
		Sampling of odorants at two places simultaneously via bilateral chemosensors (“stereo sampling strategy”)
	Odor-gated anemotaxis Used to move upwind (crosswind) when encountering (losing contact with) the odor plume Used by a number terrestrial animals, including moths and flies (Gaudry et al., 2013), as well as rodents (Liu et al., 2020);	Visual/mechanoreceptive assessment of wind direction and speed
	Beaconing Used when navigating at a limited spatial scale as distant sensory cues need to be within the perceptual range Used, e.g., by salmon (Dittman & Quinn, 1996), fruit flies (Saxena et al., 2018), ants (Graham & Cheng, 2009), bats (Geva-Sagiv et al., 2015), rodents (Clark, & Taube, 2009; Jain et al., 2017), and humans (Lehnung et al., 1998)	Navigation toward distant sensory, often visual, cues
	Route following Used when navigating in highly regular environments Used, e.g., by rodents (Gire et al., 2016) as well as humans (Hamburger & Knauff, 2019)	Navigation along a specific route where the direction changes at specific decision points (i.e., habitual behavior)
Cognitive map strategy Strategy by which the animal reaches a goal location through the integration of multiple cues into a comprehensive mental map	Path integration Used when navigating at a limited spatial scale, as error accumulates over time in the absence of external sensory information Used by ants (Buehlmann et al., 2014; Collett & Carde, 2014; Wolf, 2008), rodents (Rowland et al., 2016), and humans (Etienne & Jeffery, 2004)	Navigation based on an estimation of the present location from the past trajectory
	Landmark-based navigation Used to orient the cognitive map relative to the present location and relative to other buildings, objects, and places in the environment Used by rodents (Rowland et al., 2016; Zhang & Manahan-Vaughan, 2015) as well as humans (Hamburger & Knauff, 2019))	Navigation based on the individual's memory of how different landmarks relate to one another in a familiar environment
	Mental navigation/Abstract cognition Used to simulate the consequences of certain actions and thereby help decision-making Used by rodents (e.g., in the context of simulating different paths at decision points within an environment by hippocampal “sweeps”; for an overview, see Lisman & Jensen, 2013) as well as humans (e.g., Bao et al., 2019; Behrens et al., 2018; Bellmund et al., 2016; Bellmund et al., 2018; Constantinescu et al., 2016; Doeller et al., 2010; Horner et al., 2016; Tavares et al., 2015)	Mental simulation or imagination of (future) trajectories through a physical or conceptual (abstract) space based on the known relationship between different landmarks, stimuli, or abstract concepts

The simple chemotactic strategy presented here is characterized by a minimal level of complexity, and yet allows for efficient behavior. While chemotaxis may be sufficient to satisfy the needs of single-cell organisms, more developed organisms need to solve more complex problems than following olfactory concentration gradients to find a food source or avoid a toxic substance. As such, with increasing levels of task complexity, organisms need to implement increasingly sophisticated search algorithms that allow them to catch prey, navigate in turbulent odor plumes, migrate to mating and/or homing sites, find specific goal locations in the environment to maximize reward, and solve complex spatial problems. The gradual refinement of olfactory navigation strategies and the importance of multisensory integration for navigation behaviors across evolution will be discussed in the following sections.

3. Behavioral strategies of olfactory navigation

3.1. Strategies of olfactory navigation in non-mammalian species

Many studies have shown that olfactory cues are of high relevance in non-mammalian navigation (for an overview, see Baker et al., 2018). A thorough analysis of the corresponding literature is beyond the scope of this review. Instead, we present a selection of striking scientific findings that illustrate examples of successful navigational strategies. The focus will be directed towards both similarities and differences in olfactory navigation across species, and the relevance of alternative, often supplementary, navigational strategies will be discussed.

3.1.1. Strategies of olfactory navigation in pelagic animals

Highly sensitive chemoreceptive abilities have been described for pelagic animals, that is, animals that inhabit the open sea. In the midst of the ocean, only limited numbers of sensory inputs (geomagnetic, hydrodynamic, chemical) are available (Nosal et al., 2016), thus providing few reliable cues to guide behavior. In this environment, olfaction may provide exclusive signals to find nutrients. Under these conditions, copepods, a group of small planktonic crustaceans (Lombard et al., 2013), as well as shrimp (Hamner & Hamner, 1977), follow the scent trails of sinking nutrient particles. Importantly, neither visual nor hydrodynamic cues seem to account for this behavior, thereby ruling out possible alternatives to a purely chemosensory mechanism. However, it is unclear whether the observed behavior truly represents scent tracking in the sense of following a chemical gradient. For example, shrimp consistently descended after the detection of a scented trail, regardless of the concentration gradient (Hamner & Hamner, 1977). Likewise, copepods are not sensitive to the direction of a chemical gradient (Doall et al., 1998), suggesting that these animals have developed an adaptive behavioral strategy to catch falling food. This behavioral response may make sense from an evolutionary standpoint, but does not require sophisticated mechanisms to track rapid changes in the odor stimulus across time and space. The food-falling catch strategy may simply offer the best combination of energy expenditure and energy intake to ensure homeostasis in these species.

In contrast, larger and evolutionarily more advanced pelagic animals have developed more complex mechanisms. Sharks, for instance, not only use olfactory cues to facilitate underwater navigation (Nosal et al., 2016; Gardiner et al., 2015), but also exploit bilateral odor arrival-time differences to localize the source of an odor from a distance (Gardiner & Atema, 2010). This strategy is particularly useful for animals navigating in turbulent odor plumes where odor concentration changes drastically from one location to the next, and where the simple chemotactic strategies discussed above are not sufficient for successful navigation. In their experiment, Gardiner and Atema (2010) stimulated the two nares of a small shark species with odor pulses of varying odor concentration and onset delay. They found that sharks consistently veered to the side corresponding to the earliest arriving odor stimulus, while odor concentration per se was irrelevant for steering behavior. This result highlights the importance of bilateral odor arrival time differences, as well as the advantages of greater spacing between the nares. A key implication is that the greater the spatial distance between the two nares, the greater the spatial resolution at a given swimming speed. At its evolutionary extreme, the hammerhead shark has truly achieved the apogee of binaral sensing (Figure 1).

Figure 1. Evolutionary advantage of increased inter-naris spacing.

(A) Differences of odor arrival time as a function of angle of attack (i.e., the angle at which the animal approaches the odor plume). At a given swimming speed, the time delay between odor arrivals at the two nares increases with increasing angle of attack. (B) Differences of odor arrival time as a function of inter-naris spacing. At a given swimming speed and a given angle of attack, greater anatomical spacing between the nares will increase the relative time delay of odor arrival at the two nares. In turn, animals with a greater distance between the two nares can increase their swimming speed while maintaining a high spatial resolution. Figure adapted with permission from Gardiner & Atema (2010).

Fast swim speed in combination with high spatial resolution of odor signals is necessary for predators hunting live, moving prey. In contrast, for more sedentary predators, simpler mechanisms probably suffice. The difference in hunting styles may explain the difference in olfactory strategies between planktonic animals (such as shrimp or copepods) and sharks; whereas the former detect their prey and rely on the predictability of gravity and its downward pull, the latter hunt for prey whose behavior is a lot less predictable. For sharks, this may force them to develop more refined olfactory navigation strategies, such as the binaral comparison of odor arrival time and/or odor concentration, to track a moving target object across time and space. Interestingly, this stereo sampling strategy (Table 1) is used not only by sharks, but also by a variety of other species (Catania, 2013; Gaudry et al., 2013; Jones & Urban, 2018; Liu et al., 2020; Porter et al., 2007), as will be discussed below (see sections 3.1.2, 3.2.1, and 3.2.2).

Importantly, the ability to use olfactory cues for successful spatial navigation may depend not only on the animal’s lifestyle, but also on its habitat. For example, inhabiting the deep, open sea where visual cues are sparse requires a higher level of olfactory spatial skills compared to living in the reef where visual cues are readily available. In this context, one study compared the size of the olfactory bulb, the brain region that receives direct input from the olfactory sensory neurons, across 60 different cartilaginous fishes (Yopak et al., 2015). The largest olfactory bulbs (relative to the size of the remaining brain) were found in pelagic sharks that rely heavily on olfactory cues to navigate their environment. This finding is in line with the olfactory spatial hypothesis (Jacobs, 2012), which assumes that olfaction serves an essential function in spatial navigation and predicts that olfactory structure size should scale with navigational demand. Connecting this idea with the above evidence, larger olfactory bulbs in pelagic sharks may help track the presence of dynamic odorous stimuli or chemical gradients over large distances, link salient locations in olfactory space, and even form cognitive maps of the subterranean space within their hunting range.

3.1.2. Strategies of olfactory navigation in insects

Moving forward in the timeline of evolution, animal species started to populate the land. Along with the transition from water to land, animals had to adapt to the transmission of odor molecules through the air. Airborne odors are greatly influenced not only by the mechanisms of molecular diffusion, but also by the presence of wind. The resulting turbulent diffusion dominates the development of odor plumes, in which odor molecules are distributed inconsistently, thus exposing the navigator to a highly intermittent chemical signal (Murlis et al., 1992). In such an environment odor gradients are noisy, rather than smooth, and simple spatial or temporal comparisons of odor concentration are insufficient for successful navigation. Of note, several research groups have analyzed odor plume dynamics (Celani et al., 2014; Connor et al., 2018; Vergassola et al., 2007), highlighting the important point that odor plumes provide relevant information which in turn can be exploited by navigating animals. For example, Vergassola and colleagues (2007) suggested that the probability of contact with an airborne odor depends on the distance from the source, and that the set of odor encounters occurring along the search trajectory may carry information about the source location. In a similar manner, researchers have elegantly characterized the probability density functions of stimulus concentration within the plume at various locations from the source (Celani et al., 2014; Connor et al., 2018; Murlis et al., 1992). Whereas both high and low odor concentrations could be perceived at locations close to and far from the source, the probability density function showed a higher probability of encountering a high concentration odor closer to the source.

A potentially useful strategy that many terrestrial animals have adopted is the integration of chemical cues with visual and mechanoreceptive inputs carrying information about wind direction and speed. For example, in a turbulent environment, the presence of an odorant will prompt upwind movement, whereas the absence of odor promotes cross-wind, or casting, behavior. This strategy is termed odor-gated anemotaxis (Table 1), as the detection of an olfactory stimulus controls the response to a cue originating from a different modality, in this case, somatosensory (Vickers, 2000). Importantly, odor-gated anemotaxis and chemotaxis are complementary rather than mutually exclusive strategies and can be used simultaneously while animals are navigating in odor plumes to localize odor sources (Gaudry et al., 2012).

Likewise, terrestrial animals typically have access to an abundance of discrete and often visually defined objects, allowing for beaconing (Table 1). This strategy is characterized by the navigation toward distant sensory cues. In and of itself, this strategy is limited in scale because discrete visual stimuli located at a large distance or hidden by other objects are outside of the perceptual range. However, visual information can be combined with wind direction (Currier & Nagel, 2018) and olfactory cues (Saxena et al., 2018) to guide orientation behavior and may, in this multisensory context, enhance behavioral performance (Saxena et al., 2018).

The use of multisensory cues (including olfactory cues) has been demonstrated in numerous insect species, including walking species such as desert ants (Buehlmann et al., 2014; Collett & Carde, 2014; Wolf, 2008). When foraging for food, desert ants engage in extensive cross-wind terrestrial search to increase the chances of encountering an odor plume, and subsequently follow it in an upwind direction (Buehlmann et al., 2014). Once the food source is found, desert ants return to it using path integration (Table 1). While this strategy may bring the ant nearer to the feeding site, search paths may be long and tortuous without the guidance of an odor plume. Interestingly, after the initial encounter with an odor plume emanating from a food source, ants seem to adjust their future navigation behavior by approaching the food source from downwind locations (Wolf, 2008). These results suggest that the animals use the wind direction as an important factor to improve foraging efficiency.

The assessment of wind parameters also supports odor source localization in flying insect species (Álvarez-Salvado et al., 2018; Gaudry et al., 2012). Optic flow as well as mechanosensory cues during flight provide information about wind speed and wind direction, ultimately affecting the animal’s search behavior. For example, moths and flies will decrease their speed and steer upwind upon encountering an odor plume; in contrast, when the odor is lost, animals will initiate cross-wind flight, or casting behavior, to increase the chance of reencountering the odor plume (Budick, & Dickinson, 2006; Mafra-Neto & Cardé, 1994; Saxena et al., 2018). Critically, the degree of upwind bias depends on the frequency, rather than the duration, of odor contact, suggesting that odor timing plays an important role in successful source localization (Demir et al., 2020). Importantly, strategies relying on the combination of odor- and wind-sensing yield successful navigation in a turbulent environment. However, under still-air conditions, comparisons of odor concentration between antennae (the equivalent to inter-nostril comparisons in mammals; see Catania, 2013; Porter et al., 2007) may prove more useful (Gaudry et al., 2013).

3.1.3. Strategies of olfactory navigation in birds

Many of the challenges faced by flying insects are also experienced by birds. However, commonly studied insect species (e.g., ants, fruit flies, locusts, and moths) navigate at a relatively small spatial scale, whereas homing pigeons or migratory birds can (and need to) navigate over large distances, up to tens of thousands of kilometers. This long-distance navigation represents a highly complex behavior and clearly requires more advanced strategies than those observed in fruit flies or moths.

Olfactory cues seem to play a role in avian navigation across large distances (for review, see Wallraff, 2004). In classical studies conducted in homing pigeons (Papi et al., 1971; Papi et al., 1972; Wallraff, 1980) as well as wild bird species (Fiaschi et al., 1974; Wallraff et al., 1995), it was demonstrated that birds heavily rely on olfactory cues during spatial navigation. In these studies, birds were typically deprived of olfactory information, either via transection of the olfactory nerve, or via naris occlusion. Experimental birds lacking their olfactory sense typically displayed high levels of disorientation and were not able to navigate successfully (Fiaschi et al., 1974; Papi et al., 1971; Papi et al., 1972; Wallraff, 1980; Wallraff et al., 1995). Importantly, visual landmarks did not seem to play a significant role during navigation behaviors unless birds were released within a short distance from the homing site (≤30km; Wallraff, 2004), suggesting that learned topographical features of the visual landscape are useful only at a relatively small spatial scale.

Empirical evidence of impaired navigation performance in birds lacking olfactory input has led researchers to investigate which olfactory cues these species use along their routes. Wallraff (2004) suggested that birds use large-scale chemical gradients in the atmosphere for the purpose of navigation. That is, they may exploit the relative proportions of specific odorant compounds in the air to assess their position and direction of flight. Although some researchers have been hesitant to accept the suitability of atmospheric odor gradients as navigational cues (for review, see Gagliardo, 2013; Walcott et al., 2018), these gradients were found to be remarkably stable, despite changing wind directions. In fact, the presence of wind may generate useful directional information and allows the bird to show homing behavior even when released in areas in which the animal has never been before. An explanation of how this could be achieved is shown in a simplified version in Figure 2. Briefly, birds may use odor gradients in combination with information regarding the direction and speed of the wind, as well as the position of the sun to form map-like representations of the environment, ultimately allowing them to return to their home base. More recent analyses have confirmed that wild bird species indeed exploit wind speed-dependent odorant distributions during navigation (Abolaffio et al., 2018). Taken together, these findings not only illustrate another example of an efficient navigational strategy resulting from the integration of olfactory, visual, and mechanoreceptive inputs, but also provide an instance in which multisensory integration fundamentally contributes to the creation of mental maps.

Figure 2. Olfactory navigation in birds.

(A) Learning phase. In this phase, the bird is at its home site (H). Wind from the North-East direction blows odor molecules (blue dots) down its gradient, increasing the amount (concentration) of odor that the bird encounters. The bird forms associations between the windborne odor and the wind direction. Wind direction is inferred from the time of day (via circadian rhythms) and the position of the sun. The red circle indicates the current position of the bird. (B) Operant phase. The bird will be able to use formed associations to navigate homeward (in a South-West direction) when released at the site where the odor was at high concentration. Figure adapted with permission from Gagliardo, 2013.

Mental, or cognitive maps (Table 1) are an important concept in the spatial navigation literature (Tolman, 1948). Cognitive maps describe world-centered representations of space, reliably capturing the distance and direction between various locations. This information can be used to flexibly navigate by planning routes that have never been taken before (Wang et al., 2020). In contrast, stimulus-response (S-R) strategies are much more limited in scope. For example, animals engaged in S-R behaviors may show habitual responses to specific external cues (i.e., route following; Table 1), but cannot deviate from learned routes when these routes are obstructed. In turn, by employing cognitive maps, that is, having an understanding of how different locations and objects relate to one another, animals can implement more flexible behavioral strategies even when the usual route is no longer available. Cognitive maps thus provide an important evolutionary advantage that is exploited not only by birds, but also by mammals.

3.2. Strategies of olfactory navigation in mammalian species

The transition from non-mammalian to mammalian species is accompanied by a switch between vastly different research traditions. Ethological studies in non-mammalian species present important behavioral data on olfactory navigation strategies, as they often focus on navigation in the animals’ natural habitats, thereby providing ecologically valid conditions for olfactory navigation. However, they do not generally allow for the tight control of experimental conditions. To illustrate this point, note that many of the studies cited in earlier sections of this review were implemented in the field (Abolaffio et al., 2018; Buehlmann et al., 2014; Fiaschi et al., 1974; Gardiner et al., 2015; Nosal et al., 2016; Papi et al., 1971; Papi et al., 1972; Wallraff, 1980; Wallraff et al., 1995). In contrast, neuroscientific and psychological research in mammals focuses on the examination of spatial navigation under highly-controlled laboratory conditions to identify the factors supporting (olfactory) navigation behaviors (Geva-Sagiv et al., 2015). The corresponding literature highlights both similarities and differences between non-mammalian and mammalian species.

3.2.1. Strategies of olfactory navigation in non-human mammals

Mammals have a broad behavioral repertoire supporting olfactory navigation. This includes the demonstration of chemotactic and anemotactic strategies, the use of beaconing and route following, and the formation of cognitive maps (Table 1). The breadth of behavioral strategies allows for manifold interactions with complex, highly variable environments and enables efficient navigation under a multitude of external conditions, with each strategy having relative advantages and disadvantages in specific situations or environments.

Two important chemotactic strategies mentioned earlier are (1) the consecutive sampling of odor concentration at various locations in space to localize an odor source (serial sampling strategy); and (2) the simultaneous sampling of odor concentration at the two nares (stereo sampling strategy). These behaviors were observed in rodents during odor source localization using both scented trails (Jones & Urban, 2018) and odor gradients (Findley et al., 2020; Liu et al., 2020). For example, when following scented trails, mice used the concentration differences between consecutive sniffs, rather than absolute odor concentration, to guide tracking behavior; that is, they initiated corrective turns upon perceiving a decrease in concentration, and proceeded in the same direction when registering an increase in concentration along the trail (Jones & Urban, 2018). Interestingly, the same mice showed significant and systematic lateral bias when nares were occluded unilaterally. This shift was generally small, approximately 1mm in magnitude, suggesting that bilateral comparisons of odor concentration contribute to odor localization at a fine spatial scale. In two recent studies (Findley et al., 2020; Liu et al., 2020), mice navigated through odor plumes transmitted in the air, rather than tracking an odor trail on the ground. Despite the fundamentally different nature of the task, temporal differences in sniff-to-sniff concentration were declared as the main factor contributing to odor-guided navigation in both studies. In contrast, stereo sampling of odor concentration was found to contribute to task performance only to a minor degree.

Despite the rather subtle differences in performance of mice using bilateral odor cues compared to those that do not (Findley et al., 2020; Jones & Urban, 2018; Liu et al., 2020), Catania (2013) suggested that stereo sampling of odorants can have important behavioral consequences for Eastern American moles, which depend on olfactory cues to localize their prey. In this study, the airflow to the two nares of the animals was crossed, thus reversing the odor information arriving at the two nostrils, but without affecting serial odor sampling. Whereas the moles showed perfect performance (100% accuracy) under normal airflow conditions, they could no longer locate the food (0% accuracy) when delivery of the bilateral odor cues was scrambled. Instead, animals searched to the left or to the right of the food, illustrating the relevance of bilateral odor cues at close proximity to the odor source. Catania (2013) thus suggested that serial and stereo sampling strategies are used at different spatial scales, a proposal that is in line with the findings cited earlier (Findley et al., 2020; Jones & Urban, 2018; Liu et al., 2020) and that intuitively makes sense. To illustrate this point, consider the example of an odor plume. The concentration gradient is relatively shallow at a large distance from the source; thus, large movements and serial sampling of odor concentration at different locations in space can provide crucial directional information. In contrast, at closer proximity to the source the gradient is steep, and smaller movements and bilateral comparisons of odor intensity give more fine-grained information about the location of the odor source.

In addition to purely chemotactic strategies, it is likely that rodents, as well as other mammals, use wind parameters to localize odor sources. The role of anemotactic strategies in odor source localization in rodents is poorly understood, and not many studies have devoted their attention to this topic (Gumaste et al., 2020). However, some authors reported a positive correlation between whisking and sniffing behavior (Kleinfeld et al., 2014; Kurnikova et al., 2017; Moore et al., 2013), suggesting an intrinsic relationship between the two. Furthermore, Liu et al. (2020) reported that mice consistently approached an odor source from the North-East direction, opposite to the wind direction. This example not only demonstrates the difficulty of establishing stable and smooth gradients even in the laboratory (Gershow et al., 2012), but also offers a direct parallel to the behavior observed in ants (Wolf, 2008), which also consistently approached the odor source from a downwind location. These findings suggest that rodents (and possibly other mammals), just like insects or birds, use wind information to render their search more efficient.

A special feature of successful odor navigation is the capability to switch between strategies, depending on the animals’ needs and abilities, as well as the present external stimulus conditions. This may refer to the intensive use of serial sampling strategies at a distance from the odor source, and the reliance on bilateral olfactory cues near it (Catania, 2013). Likewise, animals may rely on wind parameters when wind is present, but exploit other strategies during periods in which wind is absent. Whereas these changes in behavior typically occur within a single trial, shifts in strategy may also occur as a consequence of experience with a particular task. For instance, when mice were presented with odor plumes originating at fixed reward locations in the environment, over the course of the experiment, they switched from a purely odor-based approach behavior to a habitual response strategy (Gire et al., 2016). That is, after having learned that the reward was presented at a limited number of locations, they simply explored the potential reward locations while disregarding the odor cues. Using this habit-based strategy, the animals were actually faster in finding the reward locations, despite taking more of a trial-and-error approach. Indeed, habitual responses work very effectively in regular, highly consistent environments, but begin to break down as the search space becomes more complex. In comparison, a study on odor source localization in mice (Gumaste et al., 2020) was designed to minimize the use of habit-based strategies by ending the trial as soon as the first reward location was explored. Intriguingly, the mice showed robust odor source localization even when the investigators increased the complexity of the odor plumes by removing the inlet flow straightener to create increased air turbulence, illustrating the advantage of strategies supporting flexible behavior over those relying on stereotyped S-R relationships.

Behavioral flexibility is also a key aspect of the cognitive map theory of spatial navigation (Tolman, 1948) presented earlier. While much of the relevant research focuses on the neural processes related to the formation of cognitive maps (see section 4.2), there is some behavioral evidence that mammals, and rodents in particular, develop mental maps of their (olfactory) environment. For example, a recent study employed an odor-cued spatial navigation task in which rats were trained to retrieve a food reward in one of the four arms of an elevated plus maze (Poo et al., 2020). Within a given trial, the reward location was indicated by a specific odor presented at the beginning of a trial in one of the four arms. For example, smelling grass-like odor in any of the four arms at the time of trial initiation required the animal to retrieve a reward in the West arm; in contrast, the perception of cheese-like odor in any of the four arms at the beginning of the trial required the animal to retrieve a reward in the North arm. Importantly, this behavioral task cannot be successfully performed by relying on S-R strategies, as a given odor identity is associated with a specific arm during reward retrieval, but not during the trial initiation period. Instead, a more comprehensive understanding of space is required. The authors reported high levels of performance, suggesting that rodents indeed developed a cognitive map of the external environment.

The examples above show that mammalian species make use of a variety of olfactory navigation strategies, readily switch between strategies depending on the environmental conditions, and rely on multisensory integration to form mental maps of the environment. Thus, it appears that throughout evolution, olfaction has persistently played an important role in spatial navigation, and has remained crucial for survival across (non-)mammalian species. The question that remains unresolved is whether and how humans use olfactory information to guide navigation behavior.

3.2.2. Strategies of olfactory navigation in humans

Humans are commonly described as “visual animals”, undermining the potential relevance of olfactory cues in human wayfinding. Early advocates of this premise often pointed to the anatomical fact that the olfactory lobe decreases in relative size in higher-developed animals, and is smallest in relative terms in the human species (Edinger, 1893). This evolutionary adaptation was later believed to have been founded in the achievement of a bipedal position that prevented the nose from being close to the ground, a presumed prerequisite for an acute sense of smell (Titchener, 1915; Sarafoleanu et al., 2009). This hypothesis is not a wholly satisfying explanation, since odors can be transmitted through air, and are, for instance, used for navigational purposes in flying insects and birds who clearly do not have their primary olfactory organ close to the ground. Thus, despite anatomical differences in olfactory areas between species, a more advanced analysis of human olfaction, and particularly its function, is necessary to test its potential involvement in spatial navigation.

Although theories have long suggested an intrinsic association between navigation and olfactory prowess (Jacobs, 2012), empirical evidence for such a link in humans was missing until recently. In a set of studies, Dahmani and colleagues (2018; 2020) tested a population of healthy individuals on spatial memory and olfactory identification tasks, and found a positive correlation between spatial navigation performance and olfactory abilities. Interestingly, spatial memory and olfactory identification in human subjects could be predicted by cortical thickness measures in the medial orbitofrontal cortex, an area closely involved in higher-order olfactory processing (Dahmani et al., 2018). Although these neuroimaging results are purely anatomical, and the spatial navigation task itself did not involve odor cues, the authors suggested that spatial memory and olfactory abilities not only may correlate behaviorally, but may also share neural substrates.

Pioneering research studies on olfactory navigation in humans have been performed, exploring the use of various odor-based navigation strategies. In an early study, Porter and colleagues (2007) asked human subjects to follow a chocolate-scented trail in an open grassy field, while wearing blindfolds, earplugs, and gloves to prevent the use of visual, auditory or tactile cues. Two-thirds of the participants successfully tracked the scented path. Interestingly, improvement of scent-tracking performance over the time course of a 4-day training, as measured by an increase in tracking velocity, was positively correlated with sniffing frequency, suggesting that higher tracking speed requires a faster sensory acquisition. The same researchers further investigated the role of stereo sampling strategies in human scent-tracking, finding that by shunting left and right sniff samples into a common input to the nose, subjects had more difficulty tracking the odor path, suggesting the relevance of stereo olfaction in humans.

While there has been considerable debate over the question whether binaral directional information is derived from trigeminal, rather than pure olfactory, cues (Croy et al., 2014; Frasnelli et al., 2009; Kleemann et al., 2009; Kobal et al., 1989, Porter et al., 2005), a recent study showed that binaral odor concentration disparities can subconsciously affect behavior relevant for navigation without activating the trigeminal system (Wu et al., 2020). In this experiment, the researchers administered different concentrations of non-trigeminal odors to the two nostrils during a self-motion judgment task. The authors reported that a 4:1 ratio of odor delivered to the two nostrils yielded a consistent bias in self motion judgment toward the nostril at which the higher concentration was delivered. Critically, results were not explained by differences in airflow between the two nostrils, thus ruling out an important confound. In addition, concentration differences across nostrils were not consciously perceived by participants, reconciling earlier negative results (Croy et al., 2014; Frasnelli et al., 2009; Kleemann et al., 2009; Kobal et al., 1989) with present findings. These data suggest that humans use both serial and stereo sampling strategies during scent tracking, even if unconsciously so, highlighting the similarity of odor-based S-R behaviors across species.

Another study showed that odors can serve as salient landmarks during human spatial navigation (Hamburger & Knauff, 2019). In their experiment, the authors instructed participants to remember a route through a virtual environment in which specific odors presented at specific intersections required a specific response (for example, turn left or turn right). Once participants had passively experienced the correct route once, the wayfinding phase was initiated. During that phase, participants navigated along the same route, but this time, when reaching an intersection and being presented with the corresponding odor, participants made an active decision about which direction to proceed in. Participants, on average, made 64% correct navigational decisions (chance level: 33%). In a separate control condition, in which participants were “beamed” to four randomly selected intersections and asked to make navigational decisions based on the odor presented to them, a comparable percentage of correct decisions (63%) was reported. Although the authors interpret their findings as evidence for odor-based cognitive maps, it should be noted that the control condition only controls for sequential learning, but not the use of S-R strategies. That is, participants could have simply memorized the association of a particular odor with a specific directional decision (see route following, Table 1), rather than integrating the odor landmarks into a coherent mental map (see landmark-based navigation, Table 1).

There is nonetheless evidence that humans can use olfactory information for the formation of cognitive maps. For instance, one study has shown that participants can use odor gradients to map an arbitrary location in space (Jacobs et al., 2015). In this experiment, participants deprived of any sensory input other than smell were led to a random location within a room diffused with two odors emanating from two different sources in the room. After brief sampling and spatial disorientation, subjects then had to try to return to their original location in the room. Strikingly, even though participants only had access to olfactory information, 70% of all participants still performed better than chance in “homing” to the previously sampled location, suggesting that humans can use odor gradients to create mental maps of the environment to guide successful navigation behavior.

Overall, results in human subjects suggest that the disregard of human olfaction in navigation may not accurately reflect ultimate abilities. In everyday life, humans may prioritize visual (over olfactory) information when navigating an environment, but this does not preclude the possibility that olfactory information may contribute to human wayfinding. Of note, all experimental studies introduced here focus exclusively on olfactory navigation strategies. To model more naturalistic search spaces, it will be important to design environments containing both olfactory and non-olfactory sensory cues, and which may help elucidate the relative contributions of visual and olfactory streams in human spatial navigation.

4. Neural mechanisms of olfactory navigation

Above, we have explored the development and refinement of olfactory navigation strategies across the timeline of evolution. Behavioral strategies supporting odor-based navigation range from simple chemotactic strategies to the formation of cognitive maps. An outstanding question is how the different behavioral strategies are implemented at the neurobiological level, and how different brain mechanisms may support differently complex behavioral strategies.

The development of neuroimaging methods and analyses in both animal and human models have allowed researchers to map navigation behaviors onto their corresponding neural circuits (Geva-Sagiv et al., 2015). Establishing such links can provide a starting point to understand how animals overcome the challenges of navigation, including olfactory navigation. Such challenges may include the need to accurately respond to a wide range of different odor concentrations, encode rapid fluctuations in odor intensity, or integrate discrete odors as well as odor gradients into cognitive maps. Analyses of such problems require the assessment of neural responses in the olfactory system during behavior. Recent studies, mostly performed in insect and rodent species, have offered potential solutions by examining the behavioral strategies used, and linking them to their neural correlates.

4.1. Neural mechanisms of olfactory navigation in insects

Odor plumes, and the ways in which these plumes shape behavioral responses, have been studied extensively in insects. An important advantage of studies using insect species is that they optimally allow for the identification of behavioral strategies as well as the underlying neural mechanisms, an undertaking that is much more difficult to implement in pelagic sharks, desert ants, or wild bird species. Importantly, the brain structures identified in the insect olfactory system typically have an analogue in the mammalian brain (Su et al., 2009). That is, although anatomically different, there is a striking functional similarity in the early stages of olfactory processing between species that can be exploited when translating findings obtained in insects to mammalian species.

In search for an odor source, animals need to maintain sensitivity over a very wide range of odor concentrations. That is, depending on the relative position with respect to the plume at any given time, an insect navigator might encounter an infinitesimal trace of odor, or might encounter an odor of high concentration near the heart of the plume. To this end, several mechanisms are in place to ensure sensitivity at low odor concentrations and to prevent saturation at high odor concentrations. For example, flies have adapted odorant receptors that show a high affinity to fruit odors (Hallem & Carlson, 2006), or to the smell of conspecifics (Van der Goes van Naters & Carlson, 2007; Wilson, 2013). In addition, olfactory receptor neurons (ORNs) expressing the same odorant receptor converge onto the same area (glomerulus) within the antennal lobe (analogue of the mammalian olfactory bulb), potentially helping to strengthen the signal in downstream projection neurons (PNs) (Syzszka, & Galizia, 2015) and thus increase odor sensitivity. On the other hand, in the case of prolonged exposure to high-intensity odor, ORN firing rates will slowly decrease over time (De Bruyne, et al., 1999; De Bruyne et al., 2001), helping prevent saturation. In addition, Kazama & Wilson (2008) suggested that short-term depression at ORN-PN synapses in response to strong odor stimulation may further contribute to the maintenance of high sensitivity at high stimulus concentration. A recent investigation has further shown that ORNs dynamically adjust their responses to naturalistic odor stimuli by decreasing gain not only with increases of average stimulus intensity, but also as a function of average stimulus variance (Gorur-Shandilya et al., 2017), a mechanism that may be critical for navigation in odor plumes where signal variance is high. Collectively, empirical evidence suggests that the olfactory system is equipped with sophisticated gain-control mechanisms to dynamically encode the widely varying range of sensory input within an odor plume.

Another major challenge that flying insects face is the highly turbulent odor environment. Odor plumes are assembled from a highly intermittent odor signal that fluctuates in both time and space. One important question is how a small flying creature is able to track such a dynamic odor stimulus, and how its brain adjusts movement trajectories within odor plumes. A study cited earlier (see section 3.1.2) suggested that the frequency of odor contact may be an important factor in defining the cast-and-surge behavior observed in fruit flies (Demir et al., 2020), suggesting that the temporal dynamics of the odor stimulus is key. While research has indeed shown that odor-evoked activity in ORNs (Nagel & Wilson, 2011) and PNs (Geffen et al., 2009; Jacob et al., 2017; Kim et al., 2015; Vickers et al., 2001) may encode rapidly fluctuating odors whose time course mimics odor encounters in turbulent odor plumes, it is an ongoing debate which parameters are critical for successful navigation. Reports suggest that the insect olfactory system has access to absolute odor concentration (Vickers et al., 2001), change in odor concentration (Kim et al., 2015), fine-scale temporal dynamics (Ache et al., 2016; Geffen et al., 2009; Vickers et al., 2001), and low-frequency events in the odor plume (Jacob et al., 2017) in order to guide navigation behavior. In this context, information-theoretical analyses of real-world odor plumes (Boie et al., 2018; Victor et al., 2019) may help to determine which information is most efficient for odor source localization. At the same time, complementary contributions from other sensory modalities – such as mechanoreceptive inputs signaling wind direction (Suver et al., 2019) – are likely to be integrated with olfactory information to optimize navigation strategies.

While turbulent odor plumes represent one potential cause of the fluctuating odor signal, another reason for intermittency may be the employment of active sampling behaviors. Insects may actively sample their environment via wing fanning (Loudon & Koehl, 2000), the coordinated movement of their antennae (Suzuki, 1975), or their head (Gomez-Marin et al., 2011; Gomez-Marin & Louis 2012), critically influencing the temporal patterning of the odor signal (Tripathy et al., 2010). For example, Huston and colleagues (2015) found that locusts repeatedly sweep their antennae through odor plumes. Interestingly, the animals did not increase the time of odor contact but instead increased the frequency of odor contact. Brief contact of the antennae with the odor stream resulted in transient oscillatory waves at 20Hz in the local field potential (LFP) recorded from the mushroom bodies (analogue of the mammalian piriform cortex), and the magnitude (power) of these odor-evoked oscillations closely estimated the location of the odor stream. The authors further reported that odor-induced changes in antennal movement enhanced neural information about the odor location and reduced the error in estimating the odor edge location. These results suggest not only that stimulus fluctuations can be encoded in the olfactory system, but also that animals may actively increase intermittency to facilitate source localization.

4.2. Neural mechanisms of olfactory navigation in mammals

The neural processes supporting odor navigation in insects provide an important starting point for investigating the same phenomenon in mammals. Similar to insect species, mammals employ active odor sampling mechanisms (Catania, 2013; Findley et al., 2020; Jones & Urban, 2018; Porter et al., 2007), which in turn may influence neural activity in the olfactory system (Parker et al., 2020). In mammals, “sniffing” essentially aligns odor-evoked activity with the breathing cycle, which presumably allows for optimal detection and perception of the odor stimulus (Wachowiak, 2011). In addition, many anatomical structures identified in insects (antenna, antennal lobe, mushroom bodies) have corresponding functional analogues in mammalian species (nose, olfactory bulb, piriform cortex) (Su et al., 2009). However, a crucial distinction consists in the amount of computational resources devoted to spatial memory. Although the mushroom bodies (Barnstedt et al., 2016; Barron & Klein, 2016) and the central complex (Varga et al., 2017) of the insect brain are considered to be involved in (place) learning and navigation behaviors, insects do not have large, sophisticated brain systems dedicated to spatial memory. In contrast, the hippocampal formation found in vertebrate species is a network that is critically involved in the formation of cognitive maps and thus plays a crucial role in mammalian odor navigation. Prior to giving an overview of the neurobiological literature on olfactory cognitive maps, we will briefly outline the neural circuits supporting S-R strategies.

Similar to the olfactory system in insect species, the mammalian olfactory system is equipped with the necessary mechanisms to encode sensory inputs across a wide range of signal intensities, which is important for navigation in smooth gradients as well as turbulent odor plumes. For example, ORNs expressing the same odorant receptor converge onto the same glomerulus within the olfactory bulb (Bozza & Kauer, 1998; Buck, 2004; Mombaerts et al., 1996), thereby amplifying weak signals. In a similar way, odor concentration may be normalized in the olfactory bulb (Banerjee et al., 2015; Economo et al., 2015; Roland et al., 2015; Sirotin et al., 2015); that is, low odor concentrations are amplified and high odor concentrations are diminished. This normalization effectively represents a gain control mechanism, ensuring that the entire spectrum of odor input can be captured. While concentration-dependent odor coding is very prominent at the level of the olfactory bulb, in piriform cortex, odors typically retain their perceptual identities across odor intensities (Krone et al., 2001; Laing et al., 2003), suggesting that cortical odor representations are concentration invariant. However, recent work has shown piriform cortex may employ distinct neural codes to represent odor identity vs. odor intensity (Bolding & Franks, 2017; Stern et al., 2018). Whereas the former may be encoded by the specific population of neurons participating in the ensemble response, the latter may be encoded by the temporal dynamics of the response, with higher-intensity stimuli exhibiting reduced onset latencies (Stern et al., 2018). This suggests that information about odor identity and odor intensity is reliably mapped in piriform cortex, possibly enabling animals to coordinate chemotactic and anemotactic behavior during plume navigation.

The neural mechanisms for chemotactic strategies, including both serial and stereo sampling strategies, in mammalian species have been studied, albeit under somewhat artificial conditions. For instance, Parabucki and colleagues (2019) recently showed that a population of neurons in the rodent olfactory bulb specifically coded for odor concentration changes across sniffs, providing a potential neural mechanism for serial sampling strategies. Respective neurons typically increased their firing rate in the case of increasing stimulus concentration, or decreased their firing rate when the concentration decreased across consecutive sniffs, thus enhancing the contrast between neural responses for different concentrations. While the step stimuli used in the experiment were within the range of concentration differences typically experienced in odor plumes, and discriminable for behaving animals, olfactory stimuli encountered under natural conditions are far less regular. Furthermore, the animals were not engaged in a behavioral task, and only passively smelled the odors presented to them. Therefore, the current available evidence for neural mechanisms of inter-sniff concentration comparisons may be limited to conditions of olfactory stimulation that are rarely encountered in real-life situations.

In another study, Kikuta et al. (2010) found correlational evidence for a neural process enabling bilateral concentration comparisons. Neurons in the rodent anterior olfactory nucleus pars externa showed an excitatory response to odor stimulation of the ipsilateral nostril, and an inhibitory response to odor stimulation of the contralateral nostril. Neural responses were phase-locked to the respiratory cycle, potentially creating temporal windows for a direct comparison between inputs to the individual nares. Importantly, the degree to which excitatory responses were suppressed during bilateral odor stimulation was proportional to the concentration used to stimulate the contralateral nostril. That is, the greater the odor concentration delivered to one nostril, the greater the suppression in AON neurons on the opposite side. This suppression effect was linear, suggesting a simple subtraction mechanism. Even though the described results hint at a neural mechanism capable of exploiting simultaneously sampled bilateral olfactory cues, the animals were anesthetized during electrophysiological recordings. Such circumstances greatly reduce the ecological validity of the reported findings. Future research will need to exploit more causal approaches to assess the relevance of the described neural circuits for behavior. For instance, Esquivelzeta Rabell and colleagues (2017) demonstrated that a unilateral lesion of the AON disrupts orienting behavior to a unilaterally presented odor, and that unilateral optogenetic activation of the AON systematically biases orientation toward the side of stimulation, suggesting a causal role of the AON in olfactory navigation.

Chemotactic strategies and the corresponding neural substrates are clearly important for mammalian navigation. However, these simple algorithms are not sufficient to explain the sophisticated navigational strategies observed in mammalian species. An alternative theory of spatial navigation, involving the formation of cognitive maps, has long been substantiated with plausible neural processes, including place cells, grid cells, and other spatially sensitive neurons in the medial temporal lobe (Marozzi & Jeffery, 2012; McNaughton et al., 2006; Moser et al., 2017; O’Keefe & Dostrovsky, 1971; O’Keefe & Nadel, 1978; Yartsev & Ulanovsky, 2013). Briefly, place cells are neurons found in the hippocampus that fire action potentials when the animal is in a particular location of the environment. Their receptive fields are highly confined to a small region in space, and are therefore referred to as “place fields”. In contrast, grid cells are neurons found in the entorhinal cortex (ERC), which exhibit a hexagonally symmetric grid-like structure of receptive fields, providing the animal with distance and direction information that can be used to self-orient in space (Hafting et al., 2005; Stensola et al., 2012). As such, grid cells may provide a metric, or a coordinate system, for navigating landscapes in which positional information of individual landmarks can be coded. The integration of positional, directional, and distance information provides the substrate for generating coherent cognitive maps. Together, place cells and grid cells help establish cognitive maps of a spatial environment, enabling animals to self-orient and navigate effectively.

Most of the above work has been based on research using environments defined by visual cues. However, this evidence has now begun to expand into other sensory domains, including the olfactory domain (Baker et al., 2018; Marin et al., 2021), with concurrent development of new technologies offering precise control of chemical stimuli in both space and time (Connor et al., 2018; Radvansky & Dombeck, 2018; Tariq et al., 2019). A purely computational study first indicated the potential role of olfactory cues in place field generation and stabilization (Kulvicius et al., 2008), an idea that later received support by various empirical investigations using live animals navigating in real (Zhang & Manahan-Vaughan, 2015) as well as virtual (Fischler et al., 2019; Radvansky & Dombeck, 2018) environments. Of note, although visual cues, if available and stable, may dominate the formation and maintenance of cognitive maps (Aikath et al., 2014; Knierim et al., 1995; Save et al., 2000), olfactory cues can significantly influence existing place fields (Jeffery & Anderson, 2003; Zhang & Manahan-Vaughan, 2015; see also Figure 3). Olfactory input could be incorporated into the cognitive map through direct connections between the lateral ERC (Knierim et al., 2014), a direct target of projections from piriform cortex (Burwell & Amaral, 1998; Chapuis et al., 2013; Kerr et al., 2007; Luskin &Price, 1983), and the hippocampus (Leitner et al., 2016; Li et al., 2017), which have previously been implicated in olfactory-spatial association learning (Igarashi et al., 2014).

Figure 3. Place cells may be anchored to olfactory cues depending on the environmental conditions.

(A) Place cells may be insensitive to olfactory cues during the day, when visual cues are reliable. Left: the environment contains four discrete odors emanating from four different locations (colored odor clouds) and one external visual landmark (top, a Japanese gate). A place field forms as a consequence of exploring the environment during the day. Right: if the olfactory cues are rotated by 90°, but the external visual landmark remains in its original position, then the place cell will retain its place field relative to the stable external visual landmark. (B) Place cells may be sensitive to olfactory cues during the night, when visual cues are unreliable. Left: same as in A, but now the place field forms as a consequence of exploring the environment during the night. Right: if the odor cues are rotated by 90°, then the place field will rotate with reference to the fixed landmark. Portions of figures adapted with permission from Zhang & Manahan-Vaughan (2015), and Pilly & Grossberg (2013).

Notably, a recent study (Poo et al., 2020) reported the occurrence of place cells in the piriform cortex, the primary olfactory cortical brain area, during an odor-cued spatial navigation task (see also section 3.2.1). In this task, rodents learned to associate a particular odor identity with a specific reward location. Interestingly, separate neuronal populations in the piriform cortex coded for odor identity, location, or both. This is the first empirical evidence for place cell-like representations in a primary olfactory area. Whether spatially selective neurons in the piriform cortex support olfactory navigation exclusively, or play a more general role in spatial navigation tasks, remains to be shown. Likewise, the influence of reward value – as well as top-down input from cortical and subcortical areas – on representations of olfactory space remained largely unexplored in the study, providing fuel for future research.

The exploration of odor-induced place fields (Fischler et al., 2019; Poo et al., 2020; Radvansky & Dombeck, 2018; Zhang & Manahan-Vaughan, 2015) raises important questions about the validity of results obtained using visually defined environments. For example, a recent article (Lebedev et al., 2018) claimed that the use of scent marks may provide an alternative explanation for the apparent conservation of map-like representations in the ERC in the absence of visual cues (Hafting et al., 2005). Lebedev and colleagues (2018) suggested that the maintenance of precise map-like representations when navigating up to 30 minutes in the dark is unlikely to be explained by path integration mechanisms, as one would expect a substantial drift. In contrast, the authors proposed that animals in Hafting et al. (2005) used scent marks established during the preceding light phase to guide navigation behaviors in the absence of visual cues. That is, sensory cues in addition to proprioception may be required to maintain map-like representations in the dark. To test this hypothesis, future research studies should tightly control corresponding behavioral variables by measuring scent traces, sniffing behavior, whisking, and general locomotion patterns.

While most of the relevant literature focuses on odor navigation in rodent species, a recent study from our lab set out to test whether humans can use intensity gradients to form map-like representations of olfactory space (Bao et al., 2019). For this purpose, a two-dimensional, perceptual space was created out of binary odor mixtures, in which each of the two odor components (banana and pine) independently varied in perceived intensity, thus comprising the two coordinate axes of the space. At the beginning of each trial, participants were presented with an initial “start” odor mixture of a given banana and pine intensity, and then they viewed slider bars on a computer screen, informing them how much to expect the relative proportions of the two odors to change. Subjects were then instructed to imagine what the “end” odor would smell like, after which they were presented with a second odor mixture, and had to decide whether or not this stimulus matched their prediction (Figure 4). Prediction accuracy was higher than chance, suggesting that humans can mentally “navigate” from one odor mixture to another (see Mental navigation/Abstract cognition in Table 1). Of note, this behavior was supported by grid-like representations in the ERC, the ventromedial prefrontal cortex (vmPFC), and the anterior piriform cortex (APC).

Figure 4. Human odor navigation experiment.

(A) Experimental design. A two-dimensional space was defined by two odorants, whose intensity could vary independently from each other. Each x-y-coordinate represented a unique odor mixture. Participants were instructed to navigate from a start odor to an imagined end odor and then decide if the actual end odor matched their prediction (blue arrow; on trajectory) or not (red arrow; off trajectory). (B-D) Based on the observation that grid cells in a given subject all share the same angular orientation (φ) within a local area of entorhinal cortex (ERC) (Doeller et al. 2010), the prediction was that greater fMRI activity should be observed when trajectories in olfactory space are aligned to the grid angle, with hexagonal (six-fold) periodicity (orange bins, B and C), compared to when trajectories were misaligned (blue bins, B and C). Interestingly, Bao et al. (2019) found grid-like representations in the entorhinal cortex (ERC), ventromedial prefrontal cortex (vmPFC), and anterior piriform cortex (APC shown in panel D). Figure adapted with permission from Bao et al. (2019).

These findings suggest that 2-D arrays of odor intensities, which themselves are not topographically represented in the olfactory epithelium, map onto grid-like representations that can support spatial orientation and navigation within an olfactory space. Interestingly, the involvement of APC implies that the sensory modality used for navigation, in this case, olfactory, has an impact on the composition of the grid-like network (Bao et al., 2019). Finally, from an ecological perspective, to the extent that (1) odor concentration decreases with distance from its source, and (2) perceived odor intensity monotonically scales with concentration (Conover, 2007; Gire et al., 2016; Jacobs, 2012, Vickers et al., 2001), the odor intensity array devised here plausibly models the type of spatial environment that an olfactory navigator could successfully use to orient and find their way. Ultimately, it remains to be shown whether navigation in a physical space defined by olfactory cues is supported by the same grid-like neural mechanisms.

5. Conclusions

Throughout evolution, olfactory cues have represented a highly valuable source of spatial information. Olfactory cues are often integrated with sensory input from other modalities to allow the organism to flexibly switch between strategies (Metaxakis, et al., 2018). This flexibility increases the probability of success under various environmental conditions, as one can select the strategy that is most beneficial towards reaching the goal in a given situation at a given time. For example, rodents performing odor source localization in a windy environment can sense the wind direction using their vibrissae and approach the odor source from a downwind location (Liu et al., 2020). In contrast, under still-air conditions animals need to rely on bilateral and serial comparisons of odor concentration (Catania, 2013). Once familiarized with the environment, the animal may use or habitual response strategies (Gire et al., 2016). However, these strategies are only successful in highly regular environments and do not result in successful localization of the odor source when the complexity of the environment increases (Gumaste et al., 2020). Cognitive maps of the olfactory environment may overcome these limitations, by providing a flexible representation of space and incorporating visual landmarks, odorant distributions, and cues from other sensory modalities (Fischler et al., 2019; Poo et al., 2020; Radvansky & Dombeck, 2018; Zhang & Manahan-Vaughan, 2015).

In synthesizing the literature, we contend that the ability to exploit environmental cues using all sensory systems, including olfaction, is key to successful navigation across different species. Notably, there is no reason to refute the idea that this description also applies to humans. Rather, there is simply not enough scientific evidence to evaluate its applicability, given that the history of human spatial navigation research has relied almost exclusively on visually defined environments. This bias may partly be due to general hesitation to consider or accept the relevance of olfaction for human behavior. Scientific and philosophical views of the 19^th and 20^th century have critically influenced our perceptual and cognitive regard toward the sense of smell, or the lack thereof (McGann, 2017). Long-held assumptions, partly propagated by the media, continue to fuel the idea that olfaction is a “minor” sense. For instance, sharks are illustrated in movies as malicious predators that can smell a single drop of blood over many kilometers – this depiction is not only incorrect, but also reinforces the association between olfaction and primitive basic instincts.

To conclude, we argue that olfaction, along with other sensory systems, provides crucial spatial information that allows many species, including humans, to navigate the environment and interact with external stimuli. Although few efforts have been made to understand the potential contributions of olfaction to human spatial navigation, it is important to investigate the phenomenon to assess its opportunities and limitations, as well as its relevance in everyday life and in the lives of those with an impaired sense of smell. Without more rigorous consideration, the questions surrounding human olfactory navigation will remain unsubstantiated. Despite the fact that olfaction has been considered a human non-sense for centuries, it is time to change that impression.

Highlights.

• Spatial navigation is a core function of olfaction in many species.

• Olfactory search strategies have become more sophisticated across evolution.

• This change is accompanied by the development of more complex neural mechanisms.

• Even in mammalian species, odor cues contribute significantly to wayfinding.

• We discuss the relevance of these insights for human olfactory navigation.