This research was prompted by the real-world challenges faced by international nuclear safeguards inspectors. These inspectors must develop accurate spatial knowledge about complex nuclear facilities, yet they have little control over the route that they take through the facility, they typically cannot use digital devices that could help them to track their location, and they may not have access to facility maps other than complex blueprints. In this study, we investigated three different types of paper-based facility maps, which reflect the types of maps that would typically be available to safeguards inspectors. We led participants on a route through a former nuclear facility and manipulated whether they studied a map before entering the facility or carried a map with them while learning the route. We tested various aspects of their spatial knowledge and their memory for items in the building. Our results suggest that simple maps are better for supporting spatial knowledge development in this context, but that the effects of map use vary considerably due to individual differences in sense of direction. We recommend that International Atomic Energy Agency (IAEA) inspectors self-assess their sense of direction, and that people with a low sense of direction might consider avoiding map exposure, as it does not seem to improve their spatial learning. In addition, since carrying a map during route-learning did not benefit spatial learning and had detrimental effects on participants’ attention to their environment, we recommend that inspectors avoid referring to maps during their inspection unless it is critical for their inspection duties.
International nuclear safeguards inspectors conducting facility inspections for the IAEA face interesting cognitive challenges when performing inspections in the field. They are led through a complex, often unfamiliar, nuclear facility by a host and have little to no control over their path through the facility due to safety, security, or operational constraints defined by their hosts. The main goals of their inspections are to detect the diversion of nuclear materials, the misuse of safeguarded facilities, and the development of undeclared nuclear facilities. Although inspectors typically have a pre-determined list of tasks that they need to accomplish during the inspection, such as collecting material measurements or verifying the design of the building, they are also expected to pay careful attention to their surroundings in order to identify potential anomalies that might warrant further verification activities under their international safeguards agreements or the Additional Protocol (International Atomic Energy Agency, 1997). As part of this overall awareness of their environment, inspectors are expected to monitor their route through the facility to ensure whether they have visited all of the rooms that they were supposed to visit, and detect whether certain areas of the building were avoided or hidden.
Types of spatial knowledge
The main experimental question that this study sought to address was how to present building layout information to inspectors in a way that supports the development of their spatial knowledge of the environment but does not impede their ability to attend to other aspects of the environment. Inspectors must pay close attention to their environment as they move through a facility, both to ensure their safety as they move through a hazardous industrial environment, and to be on the lookout for any subtle cues that could indicate the diversion or misuse of nuclear materials or facilities. At the same time, they must build and maintain their spatial knowledge of the facility and their location within it. If the inspectors have more complete spatial knowledge of the facility, they will be better able to detect anomalies in the processes and operations of the facility, the layout of the building, or instances in which their guide leads them on circuitous routes, avoiding certain areas of the facility.
The question of what constitutes spatial knowledge is rather complex. It has been shown that there are at least three different levels of spatial knowledge: landmark, route, and survey knowledge (Siegel & White, 1975). Landmark knowledge refers to one’s memory for objects encountered in the environment (detached from the object’s location) and has recently been shown to develop in the absence of overt attention (van Asselen, Fritschy, & Postma, 2006). Route knowledge is defined as knowledge of an environment that is anchored by a series of actions taken at specific decision points. Route knowledge is more abstract that landmark knowledge and can contain the path between different landmarks, but is not thought to encompass aspects of the environment that were not encountered as part of a learned route nor to include metric knowledge about the distance between landmarks. Finally, survey knowledge is the most detailed representation of the relationships between all places in the environment. Survey knowledge represents the space from an allocentric point of view (i.e., relationships between places represented in terms of cardinal directions, angular degrees, or some other objective measurement) as opposed to an egocentric point of view as in route knowledge (i.e., turning left or right based on the individual’s location in the environment). Survey knowledge is marked by the ability to estimate straight-line directions and/or distances between landmarks, especially those that were never traversed between, and is believed to require effortful attention to generate.
Although Siegel and White (1975) originally suggested that these three levels of spatial knowledge develop sequentially, Montello (1998) proposed a competing framework suggesting simultaneous development of survey knowledge in parallel with the other levels of spatial knowledge. Indeed, Ishikawa and Montello (2006) demonstrated that large individual differences exist in the type of spatial knowledge that individuals learn about a novel environment after repeated exposures: some people developed survey knowledge almost immediately, some people never did, and still others showed a continuous progression in their knowledge. Similarly, Hölscher, Meilinger, Vrachliotis, Brösamle, and Knauff (2004) found that even people who were highly familiar with a particular building showed poor survey knowledge, even if they had excellent route knowledge. The evidence from these prior studies suggests that one model of spatial knowledge development may not fit all learners in all situations.
Influence of learning conditions on spatial knowledge development
Some work exists in the cognitive science literature regarding what types of learning conditions best support the development of spatial knowledge, although none of it specifically addresses the constraints faced by IAEA inspectors. For example, it has been shown that being passively led along a route leads to worse spatial knowledge than if the individual has active control over their navigation (for review, see Chrastil & Warren, 2012). As such, IAEA inspectors already start at a disadvantage because they have no choice but to be passively led through the facility. Moreover, much of the cognitive science literature considers spatial learning and testing that takes place exclusively in a virtual environment (e.g., Carassa, Geminiani, Morganti, & Varotto, 2002; Chrastil & Warren, 2013, 2015), such as watching videos and being tested in a virtual environment (e.g., Meilinger, Knauff, & Bülthoff, 2008), learning maps of an environment without ever seeing it in person (e.g., Coluccia, 2008; Garden, Cornoldi, & Logie, 2002), or learning a route through a combination of virtual environments and still photographs (e.g., Gaunet, Vidal, Kemeny, & Berthoz, 2001). Since it is rarely possible to create virtual, video, or other digital representations of nuclear facilities, many of these findings from the cognitive science literature are not applicable to the IAEA. Similarly, there are tight restrictions on bringing electronic devices into nuclear facilities. While there have been several studies of digitally aided navigation and wayfinding (cf. Münzer, Zimmer, & Baus, 2012; Richter, Dara-Abrams, & Raubal, 2010; Schmid, Richter, & Peters, 2010; Schwering, Krukar, Li, Anacta, & Fuest, 2017), we cannot employ digital aids in this problem space. However, we can draw upon what other researchers have learned from comparisons of different methods of route learning.
There have been several studies that have compared the efficacy of learning a pre-determined route from a map versus other sources. In studies that have compared learning a route from a map versus learning through direct experience (typically by following an experimenter along a guided route), two studies have found approximately similar levels of spatial learning under both conditions (Ishikawa, Fujiwara, Imai, & Okabe, 2008; Richardson, Montello, & Hegarty, 1999). However, Ishikawa et al. (2008) found that the map group walked more slowly and made more errors than the direct experience group. This is especially relevant to IAEA safeguards inspectors because they have only a limited amount of time to complete their verification activities due to their global verification obligations as well as pressure from the nuclear facility operators to complete their activities quickly, since these activities usually cause a disruption to operational activities and, therefore, lost revenue.
Additionally, Richardson et al. (1999) noted alignment effects in the map condition (Levine, Marchon, & Hanley, 1984), in which pointing errors were higher when the participant was misaligned in space with the original orientation of their map, suggesting that people build an orientation-specific representation when learning from a map. Thorndyke and Hayes-Roth (1982) found better survey knowledge in map learners, but better procedural knowledge (i.e., routes between locations) in their direct experience group. However, their direct experience group worked in the building where the testing took place, so their effects could be due to prolonged exposure to the environment prior to the experiment.
Other studies have compared learning a route from a map versus learning from a GPS device. While the use of GPS is not an option for IAEA safeguards inspectors due to limited functionality of these devices indoors and restrictions on the use of electronic devices in nuclear facilities, GPS-aided spatial learning is similar to the guided navigation that safeguards inspectors experience while being led through a facility. Research in this area has shown that GPS tends to produce worse spatial learning outcomes, particularly on complex parts of the route (Ishikawa et al., 2008; Willis, Hölscher, Wilbertz, & Li, 2009). Willis et al. (2009) posited that the piecemeal way in which the map was displayed to participants in the GPS condition may have contributed to a more fragmented knowledge of the configural layout of the environment, relative to the map condition in which the entire environment was visible at once.
Li, Brown, Pinchin, and Blakey (2015) compared learning a route through a complex, indoor environment (a hospital) from a map versus a verbal description and found that participants learned the route equally well from both sources, although the map group was able to walk the route in reverse faster than the verbal description group.
Other studies have assessed the impact of different types of GPS-style wayfinding aids on spatial learning. For example, Löwen, Krukar, and Schwering (2019) compared map schematizations that emphasized different types of features along a route. The maps could emphasize local features (such as local landmarks), global features (structural features, such as city or area boundaries), or both. They found that accentuating local features improved participants’ route knowledge, but not their survey knowledge, while accentuating global features improved participants’ survey knowledge, but not their route knowledge. Similarly, Münzer et al. (2012) found tradeoffs between learning routes and learning configural information when participants used mobile navigation assistance systems that presented different types of information. When the presentation mode provided configural information, participants had better configural knowledge, but poorer wayfinding performance. When the presentation mode emphasized the route, participants had better wayfinding performance, but poorer configural knowledge. The researchers also found that individual differences in the participants’ sense of direction had a substantial impact on both types of learning.
Several studies have compared learning a route from a map versus pictures and/or videos of the route. In the same study just described, Li et al. (2015) found that people were better able to learn a route through a hospital from a video of the route than a map alone, as measured by their speed in traversing the route both forward and backward, although there were no differences in the number of errors made. Münzer and Stahl (2011) compared learning an indoor route from a map, egocentric photos of the route, or an egocentric video of the route. In terms of errors made, the map did not differ from the combination of the two egocentric conditions; however, the map condition did induce more hesitations than the video. Across these two studies, it seems that the ability to reproduce a specific route in an indoor environment is roughly equally supported by both maps and pictures/videos, with the caveat that the maps may induce higher cognitive load, as evidenced by taking longer to complete the route or showing more hesitations along the way. However, neither of these studies probed other aspects of spatial learning, like survey knowledge of the environment, so we cannot say how these two learning conditions extend beyond reproducing a specific route.
At least two studies have compared learning an indoor route from a map versus a virtual environment. Richardson et al. (1999) found the worst learning outcomes from their virtual environment compared to their map and direct experience conditions. On the other hand, Bliss, Tidwell, and Guest (1997) measured the ability of firefighters to reproduce a route through an unfamiliar building after learning from a map or from a virtual walkthrough of the route. They found equally good performance between the two conditions in terms of both time to complete the route and number of errors made. However, one caveat to note is that these studies are both at least 20 years old, and so it could be the case that newer, more immersive virtual reality environments may show different effects if compared directly to learning from maps.
To our knowledge, only a handful of previous studies directly compare the efficacy of different types of paper or non-interactive maps on wayfinding or spatial knowledge development. In one such study, Dillemuth (2005) tested whether participants were better able to learn an outdoor route from viewing a detailed aerial photograph versus a “generalized” map of a college campus (in which the buildings, grass, paths, etc., were filled in with a solid color). The maps were statically displayed on a handheld electronic device that did not include GPS to update position or orientation. They found that subjects performed better on their wayfinding measures when using the simple generalized map, particularly on time to route completion and number of stops made on the route. They also found that users did more zooming in and out when using the more detailed map, suggesting that more attention was required to use the detailed map. Interestingly, participants with higher sense of direction (as measured by the Santa Barbara Sense of Direction Scale, or SBSOD) performed much better with the simple than detailed map, contrary to what may be predicted.
In another study, Liben, Myers, and Kastens (2008) asked whether different types of map shapes (i.e., circular or square) and angle (i.e., drawn from directly overhead or at an oblique angle to display the buildings in slight relief) enabled college students to more accurately locate themselves during a route-learning task. They found no differences in accuracy based on map type, but they did observe response-time differences in which the square oblique map was the fastest and square flat maps were slowest. This was interpreted as showing that participants generated the strongest spatial knowledge when using the square oblique map. Additionally, they found that people were less likely to rotate the square map to match their orientation, which could have contributed to alignment effects in their results. Unfortunately, neither of these studies measured other types of spatial knowledge (i.e., landmark or survey knowledge), so we cannot say for sure whether either of these map types supported the development of spatial knowledge more generally, or whether they were simply better suited to the tasks assessed.
In a study of map schematization, Meilinger, Hölscher, Büchner, and Brösamle (2007) tested participants’ ability to localize themselves and complete wayfinding tasks inside of a complex building, using either a standard floor plan or highly schematized maps that provided only route information. For the self-localization tasks, the participants performed equally well with either type of map. For the wayfinding tasks, participants performed better when using a schematized map. While the highly schematized maps were useful for wayfinding, they explicitly excluded survey information. For the IAEA safeguards inspectors, developing survey knowledge is extremely important, since they must be able to understand how the rooms in a facility relate to one another and to the route on which they are guided by the facility operators. Therefore, although schematized maps may be useful for indoor navigation in other contexts, they would not support the cognitive needs of the safeguards inspectors.
Current study
The current study is unique in that we are testing participants in a former nuclear hot-cell facility, which constitutes an extremely complex, indoor environment that was designed around safety and operational requirements rather than easy human navigation. In this way, we ensure that many of the physical and visual characteristics of the environment are well-matched to those that will be faced by safeguards inspectors in the field. This is also true of the experimental conditions that we tested. We tested three types of maps that are representative of the types of maps that safeguards that inspectors have access to when visiting a facility. We obtained actual facility maps from the group that manages the building in which our experiment took place. One map was a simple facility map, intended for space management purposes. The second map was a blueprint of the facility, which was highly complex and included a great deal of extraneous detail. The third map was a 3D representation of the simple facility map, created using the widely available software tool, (SketchUp [Computer Software], 2018) (www.sketchup.com). All three of the maps showed the overall layout of the building accurately, but all three also had minor inconsistencies between the map and the building, due either to changes that had taken place in the building over time (such as the removal of a door) or a lack of detail in the map (such as a temporary rolling partition not appearing on the map). These minor inconsistencies between a map and a facility are highly common in industrial facilities, where the building’s usage changes over time, but older maps are still in use. All maps were provided to participants on standard-sized paper because IAEA inspectors are typically unable to bring electronics into safeguarded facilities. As such, interactive or digital maps were not included as test conditions.
Our experimental conditions allowed us to ask several questions that are unique relative to the existing literature. First, we asked whether studying a map of the building prior to completing a guided route-learning task improves spatial learning. Next, we asked whether the ability to carry and refer to a map when learning a guided route provides an additional benefit to spatial knowledge, beyond the benefit of studying the map beforehand. Finally, we asked whether the level of detail in the map impacted spatial learning. We predicted that studying a map before the route-learning task would improve spatial knowledge, and that carrying the map during the route-learning task would provide an additional benefit. We also predicted that participants would receive a greater benefit from the simpler maps than from the complex blueprint, which was difficult to read and contained extraneous, distracting information.
We measured learning at every level of spatial knowledge (i.e., landmark, route, and survey), in addition to testing the participants’ awareness of details in their environment via a non-spatial memory test. We chose tasks that have been shown to reflect landmark knowledge (i.e., a landmark recognition memory task; Wenczel, Hepperle, & von Stülpnagel, 2017), route knowledge (i.e., drawing the guided route on an outline of the building; Labate, Pazzaglia, & Hegarty, 2014), and survey knowledge (i.e., a verbal pointing task, indicating angular direction between landmarks; Rand, Creem-Regehr, & Thompson, 2015). We also chose tasks that simultaneously tapped into multiple levels of spatial knowledge, including filling in a building outline with the name and location of learned landmarks and navigating a novel shortcut between two landmarks (Labate et al., 2014). The non-spatial memory task was designed to assess participants’ ability to maintain awareness of their environment by testing their recognition of incidental landmarks (i.e., van Asselen et al., 2006; Wenczel et al., 2017).
We also included a measure of individual differences in sense of direction, given the large differences in spatial knowledge acquisition found between people with good and poor senses of direction (Hegarty, Richardson, Montello, Lovelace, & Subbiah, 2002; Münzer et al., 2012; Wolbers & Hegarty, 2010). We administered the SBSOD (Hegarty et al., 2002), which is a 15-question scale that has been shown to be highly correlated with measures of spatial knowledge acquired from direct experiences in an environment. This allowed us to test whether the map study conditions had differential impacts on the development of spatial knowledge in individuals depending on their sense of direction. We predicted that participants with a poorer sense of direction would have lower levels of spatial knowledge, even with the aid of the maps.