In general, the results indicate that olfactory cues only can be used to aid navigation since participants made on average 60 percent correct route decisions at both times of testing (64% at t1 and 55% at t2) which is significantly above chance level (one-third). Thereby, wayfinding performance in the olfactory condition did not differ significantly from the visual condition at both times of testing. However, participants who used visual cues performed significantly worse 1 month after learning (t2) compared to their performance immediately after learning (t1), whereas performance in the olfactory condition did not decline over time. Looking at the recognition task, it turns out that visual landmarks were recognized better than olfactory landmarks right after learning and 1 month later.
Odors as landmarks
First, our results suggest that humans can indeed use their sense of smell for successful navigation through their environment. Based on the present results, it cannot be said that odors are of any relevance for human wayfinding, but that they
can be significant because in the standardized laboratory setting, humans were basically able to use olfactory cues for orientation. Appleyard (
1969), Carr and Schissler (
1969), and Siegel and White (
1975), among others, have stated or implied the definition of “a landmark [being] any distinct object or feature that is
noticed and remembered” (Presson and Montello
1988, p.378). According to this definition, the sense of smell would thus be virtually meaningless in human wayfinding because we are often unable to report, label, or consciously perceive odors. However, from our point of view and from the previous findings, landmarks do not necessarily need to be consciously recognized and remembered. Findings suggest that olfactory information needs to be
differentiable, but not
identifiable, for successful recognition (Hamburger and Knauff
2019). We believe that humans use their sense of smell rather implicitly in contrast with other senses such as the visual sense, whose information we process in a more explicit manner. This could also account for other modalities such as the human auditory system. Moreover, the implicit processing of odors could be an explanation for why initial retention is worse at t1 since implicit processing implies that information is processed with less depth. Therefore, it also leads to less interference, and as a result, the acquired information remains stable with minimal loss over time. Since there is a lack of the literature on this topic, we would like to highlight its importance and include it in the future research.
Our finding of humans being able to use olfactory cues for navigation is consistent with the results of Hamburger and Knauff (
2019), who reported 64 percent correct route decisions immediately after the learning phase. One could argue that this results only from spatial sequential learning (e.g., Deroost and Soetens
2006) and not from correct recall of the landmark information as directional cue. However, the previous studies from our research group such as Hamburger and Knauff (
2019), Hamburger and Röser (
2014), or Balaban et al. (
2014) compared a wayfinding task using olfactory landmark information with a control condition in which participants were “beamed” (i.e., teleported) to random intersections of the route, then were presented with the corresponding odor, and had to make the directional decision of this specific intersection. Performance in the wayfinding and control phase was similar, indicating that odors were encoded as landmarks and were used in a map-like mental representation to make adequate route decisions (Hamburger and Knauff
2019).
Although the present study design did not include such a control condition, the experimental setting was mostly similar to that of Hamburger and Knauff (
2019). Since they found no results that could be explained by SSL, we believe that the participants in the present study most likely used cognitive maps including landmarks to find their way through the virtual environment as well. Moreover, we opted not to include a control condition without any landmarks because in such a scenario, participants would only be able to rely on memorizing the route as a sequence of verbal cues such as “left, right, left” and so forth. This cognitive approach would differ significantly from the strategy employed when landmarks are present at decision points (Balaban et al.
2014; Hamburger and Knauff
2019; Hamburger and Röser
2014; for a critical discussion see also Hamburger
2020). Thus, it would not provide us with insights into whether participants perform better with olfactory landmarks. Furthermore, in our experimental design, it is not possible to include an intraindividual comparison between landmark-based wayfinding and non-landmark-based wayfinding, since at least at t2, the subjects would have prior knowledge about the landmarks. Nevertheless, to control whether the differences found were indeed due to different modalities in landmark-based wayfinding, we tested whether SSL was statistically different between the two conditions. Sequential learning occurs at the first and last intersections of the labyrinth according to the recency and primacy effect. For this, we statistically compared the first three and last six route decisions of our experiment between the visual and olfactory conditions. We report here the results of the first three and last six route decisions analogous to existing literature on the primacy and recency effect (Murdock
1962; Baddeley and Hitch
1977), but further statistical comparisons (not reported here in detail) with varying numbers of the first and last intersections showed the same results. Since we did not find any differences, we assume that even if participants used SSL or a combination of SSL and landmark-based wayfinding, the use of these methods was unlikely to differ between the two conditions and, therefore, can be ignored in the interpretation of the present results. This statistical control analysis does not show that subjects learned exclusively in a landmark-based manner, but only that if they learned sequentially, it did not differ between groups, and thus, the group differences found were due to landmark-based learning and not SSL. However, our questionnaire data from subjects as well as anecdotal reports from the previous studies (e.g., Hamburger and Knauff
2019) showed that few subjects reported learning sequentially.
Our findings show that olfactory cues are sufficient to guide navigation in the absence of other cues. In cases of blindness or other visual impairment, olfaction may even be fundamental for orientation. However, more studies are needed to investigate whether olfactory cues are spontaneously relied upon to serve wayfinding in real-world circumstances. In addition, the issue of reliability in a natural environment needs to be further investigated, as the perception of odor marks can be influenced by several factors such as wind direction, interference with other odors, and low consistencies (Koutsoklenis and Papadopoulos
2011).
Multimodal utilization of cognitive maps
Even though it may seem intuitively plausible that the representation of a cognitive map with landmarks would be easier to be created from visual information, we argue against the close association of visual information and propose a rather multimodal utilization of cognitive maps, given the following results of recent research and our own findings of the present study.
Karimpur and Hamburger (
2016), Hamburger and Knauff (
2019), and Arena and Hamburger (
2022) revealed almost equal wayfinding performance for all types of landmarks, regardless of the modality in which they were presented. Participants made on average 64–73% correct route decisions using olfactory landmark information (Hamburger and Knauff
2019; Arena and Hamburger
2022), when using acoustic landmarks, their performance was on average 71–85%, compared to 66–88% correct route decisions when using visual landmarks and 87% correct responses when using written words as cues (Karimpur and Hamburger
2016; Arena and Hamburger
2022). Following these results, our study shows the same pattern, finding no significant difference in the wayfinding task between the visual and olfactory conditions at both times of testing. All of the studies listed also found no significant differences in correct route decisions between the different modalities (Karimpur and Hamburger
2016; Hamburger and Knauff
2019; Arena and Hamburger
2022). This makes a multimodal use of cognitive maps, where our senses work together rather than as separate entities, even more plausible (e.g., Karimpur and Hamburger
2016). Findings in this direction already exist, but could not yet be confirmed by initial studies in our research group (Arena and Hamburger
2023).
Nevertheless, we do not doubt that our visual system provides us with the most important and prominent access to our environment. However, we want to highlight our assumption of humans using all their available senses more or less explicitly in order to optimize their orientation in everyday life. Therefore, future studies should definitely consider a rather multimodal approach on human navigation instead of overinterpreting the experiments on unimodal visual wayfinding and still relying on a visual superiority effect in human navigation (Presson and Montello
1988).
Odor memory as a separate memory system
One month after the leaning phase, wayfinding performance between the visual and olfactory conditions did not differ significantly from each other. That means that participants did not perform better when given olfactory landmarks compared to visual cues. However, visual landmark-based wayfinding performance significantly declined from t1 to t2 whereas wayfinding performance did not decline from t1 to t2 in the olfactory condition. As our second main finding, this suggests that memory of the route was more resistant to decline when participants used odors rather than pictures as landmark information—a result that is consistent with our assumption about odor memory being a separate memory system. In 1996, Herz and Engen already postulated odor memory as being “governed by specific and distinct rules and underlying mechanisms” (p.309).
Moreover, Herz (
1998) found that odors were equivalent in their ability to elicit accurate memory recall compared to verbal, visual, tactile, and musical stimuli. However, the odor-evoked memories were always more emotional. Furthermore, a study using functional magnetic resonance imaging (fMRI) compared activated brain regions during memory recall which was triggered by olfactory versus visual cues (Herz et al.
2004). Neuronal responses demonstrated that when odor cues were used, the amygdala and hippocampal regions showed significantly greater activation than for any other cue (Herz et al.
2004). Additionally, behavioral responses confirmed this since odor-evoked memories were reported as most emotional (Herz et al.
2004). Based on her findings, Herz (
1998) concluded that odors are not superior reminders compared to other sensory stimuli but evoke higher emotional saliency rather than accuracy.
Both studies indeed demonstrate higher emotional saliency for odor-evoked memories. However, in Herz’ study (
1998), the time between encoding and retrieval was only 48 h, and in the fMRI study (Herz et al.
2004), retrieval occurred immediately after encoding. We believe that real differences in memory retrieval between different modalities can only be revealed using a longer time period between encoding and retrieval. Therefore, we set the period between t1 and t2 to 1 month in the present study. The use of a longer retention interval could explain our results, which in comparison with Herz’ (
1998) findings in fact show differences in retrieval accuracy when using visual versus olfactory memory cues.
We believe that the higher emotional saliency of odors leads to a more resistant long-term memory of the route. Therefore, at t1, odors do not result in better wayfinding than pictures since information is (possibly) rather retrieved from working memory. However, at t2 1 month later, the emotional saliency seems quite valuable since wayfinding performance does not decline when using olfactory cues. Hence, in future studies on human wayfinding, attention should be paid on how long information is retained in memory.
The study we conducted is the first research known to the authors that was able to demonstrate the unusual resistance toward decay of odor memory (Herz and Engen
1996) in the form of behavior, i.e., in human wayfinding. Even without neuroimaging techniques, the participants in our experiment exhibited behavior consistent with theories of odor processing, such as the suggested explanation by Engen (
1987) and Lawless (
1978) of odors being represented unitary in memory and therefore resulting in minimal loss over time due to low rates of influence (see
1.3 Odor Memory). In doing so, we were able to reveal an odor superiority effect in terms of long-term memory for the correct route in human wayfinding, which is likely due to the higher emotional saliency of odors compared to pictures.
Recognition does not equal wayfinding
Lastly, we investigated whether visual or olfactory landmarks result in better recognition at both times of testing. Here, results showed that, when it comes to landmark recognition, visual landmarks are better recognized than olfactory cues. This result stands in contrast with our findings in the wayfinding task, where no significant performance differences were found between visual and olfactory landmark information.
One reason for these contrasting findings for the recognition and wayfinding phase could be the design of our study. In the recognition task, participants had to answer as fast and correctly as possible (see
2.3 Procedure). This resulted in participants taking very little time to decode the presented olfactory or visual stimuli as distractors or landmarks. Studies show that response times of the olfactory system range from 600 to 1200 ms (Cain
1976) whereas participants can respond to visual stimuli in only 100 ms (Posner and Cohen
1984). The temporal reaction of pictures is, therefore, 12 times faster than for odors. Given the task design, this could have led to advantages in the correct detection of visual stimuli, as participants did not take enough time to process the olfactory stimulus and thus failed in giving correct responses.
Since the data from the wayfinding task differed across the two conditions (olfactory and visual) from the data of the recognition task, it shows that wayfinding tasks do not equal recognition tasks (i.e., Hamburger and Röser
2014 revealed similar results). Correct recognition of a stimulus alone does not predict correct wayfinding but is often used exclusively to test human navigation skills (i.e., Abu-Obeid
1998; Choi et al.
2016). The underlying cognitive processes in wayfinding and recognition are likely to be different.
However, it should be emphasized that when interpretating the data from the recognition task, special attention must be paid to the experimental design of the present study. The data collected during the recognition phase at t2 cannot be used to draw any conclusions about long-term memory, as the landmarks that are tested in the recognition phase were already partially presented in the previous wayfinding phase. In fact, the repeated presentation of the stimuli 1 month later in the wayfinding phase, just before the recognition phase, primarily relies on short-term memory retrieval in the recognition task. As a result, the present design does not allow for any conclusions regarding long-term memory with respect to the recognition of landmarks. Moreover, it is not possible to draw comparisons between the decline in wayfinding performance in, for example, in the visual condition from t1 to t2, and a potential decline in recognition performance.
However, we conclude that recognition and wayfinding must be interpreted separately and cannot be reconciled with wayfinding. Odor memory appears to coexist with memory for other senses, but our task design was only able to demonstrate a long-term memory odor superiority effect in our wayfinding task in terms of unusual resistance. Moreover, the existing literature suggests that odors are likely to be processed implicitly rather than explicitly, as the latter is the case with other senses such as vision (Degel and Köster
1999). This needs to be further investigated in the future studies.
Limitations and implications for future research
In the following, we will discuss some methodological limitations of the present study and provide implications for future research.
The first potential methodological problem is our chosen retention interval of 1 month between t1 and t2 in order to detect long-term odor memory effects. The previous studies on long-term odor memory revealing retention functions with essentially zero slope used time intervals of 1 month or 1 year (Engen and Ross
1973; Jones et al.
1978; Lawless and Cain
1975). On the other hand, Murphy et al. (
1991) found no significant difference in long-term memory performance between odors and symbols/faces after 6 months. Perkins and Cook (
1990) even reported a significant decline in long-term odor memory after 1 week. As these contradicting findings give little guidance in choosing the optimal retention interval, we aimed to make the period as long as possible in order to make sure to allow for sufficient memory loss. However, due to the time and cost constraints, it was not possible to extend the time interval any further. Our results show that our chosen retention interval was indeed long enough to reveal differences between long-term odor and pictorial memory. However, future research should extend and vary time intervals between encoding and retrieval to investigate those findings more precisely.
Moreover, our results rely on t-tests for dependent samples only. However, the interaction of the ANOVA with repeated measures was not significant. The results of the study must, therefore, be interpreted with caution, and the significance of the results remains limited for now. Therefore, additional studies with more participants are needed in the future in order to support or falsify the present results.
Since the experimental design requires the olfactory cues to be presented by hand by the experimenters, the distance to the nose and the diffusion of the odors in the air are not 100 percent uniform. Therefore, the intensity of the odors may vary from subject to subject, which may affect reliability of the experiment. For more precise results, future research could use an olfactometer (Lorig et al.
1999).
Furthermore, in the wayfinding task, participants did not experience any negative effects of being lost. Consequently, it remains uncertain whether participants would be equally capable to get back on the route by relying on olfactory or visual landmarks. Future research could employ an interactive design for the wayfinding task to explore this question in detail.
Given our results, humans are likely to use all their available senses to orientate themselves. Future studies, therefore, should investigate how olfaction interacts with other senses in human wayfinding. For that, visual, acoustic, haptic, and olfactory stimuli should be used to explore how they complement or compete with each other.
Moreover, a theoretical model is needed to integrate the numerous puzzle pieces of the existing literature—with its (somewhat) ambiguous findings, common theories, and relevant approaches to navigation and olfactory research into a broader and more accurate picture of sensory processing and the use of multimodal landmarks in human spatial orientation.
Furthermore, due to technological progress, virtual environments are increasingly used in everyday life, e.g., in video games (Karimpur and Hamburger
2015), but also in research, such as in the present study. Especially in research, increasing use of virtual environments raises the question of ecological validity (Hamburger and Knauff
2019). Lloyd et al. (
2009) already investigated the equivalence of route-learning performance between real and virtual environments. Their results revealed comparable error rates for wayfinding performance and no differences in strategy usage, indicating that even simple virtual desktop environments provide a useful tool for evaluating and researching navigation. With the presentation of our 3D maze on VR glasses, we increased ecological validity compared to a simple desktop presentation by providing an egocentric, first-person perspective to the participants. Even though, the existing literature supports the methodological approach used in the present study, further studies should transfer our experiment into reality with participants experiencing the visual and olfactory stimuli in an actual real environment. This way, participants are likely to focus more on the olfactory cues, as the presentation of our virtual maze on VR glasses might have caused them to concentrate more on visual information than they would in real life.
Finally, it is important to emphasize that this study is intended to be basic research, not applied research. The goal is to explore the underlying processes of our sensory modalities in wayfinding, not how these modalities are used in everyday navigation. Therefore, it is not necessary for this question to create a virtual environment “as close to reality as possible,” where one could, for example, take a wrong turn and then have to find the right way back. To answer the question whether olfactory stimuli can in principle help human orientation over a longer period of time, our laboratory setting is sufficient. In order to answer application research related questions, a more complex experimental setting with higher external validity would have to be used.