45
estimating distances, often underestimating them, in VEs compared to walking in the real world.
Many VEs use embodied interaction and gestures as an interaction method.
We also wanted to address this issue by integrating this interaction method into some of our applications (publications II, III, and IV). This type of interaction is not equivalent to real-world wayfinding, but it still incorporates the body-based sensory modalities used in these scenarios.
Ultimately, choosing varying environments for evaluating spatial ability comes down to the setup and premise of the experiment: computer-generated VEs provide researchers an “ecologically relevant environment”
in which to examine human behavior with “high control over the environment’s properties” (Waller, 2005). For this reason, VEs are an exceptionally useful tool, even with their limitations, for assessing individual distinctions in spatial cognition.
Richardson et al. (1999) reported that users commonly experience greater difficulty in forming spatial knowledge about VEs than the real world. This may lead to poorer performance in wayfinding tasks in VEs. Some factors may improve this performance, for example, increasing the FOV (McCreary and Williges, 1998), embodied interaction (Zanbaka, 2004; publications II and IV), and heightening the visual information available to the wayfinder (Gillner and Mallot, 1998).
4.2 LANDMARKS IN VIRTUAL ENVIRONMENTS
Landmarks provide information that helps individuals to identify their location and orientation, and often serve as the main component in route planning through virtual and real-world large-scale environments. Vinson (1999) suggested various guidelines for landmark design in VEs. Some of these guidelines are introduced and discussed briefly in the following chapters:
VE should contain several landmarks.
Once the traveler gains experience with a particular route, he or she increases the representational precision of distances and positions of landmarks (Evans, 1981), which in turn might change the spatial representation from route knowledge to survey knowledge. This allows the traveler to adopt the most suitable perspective of the environment for a wayfinding task (Thorndyke and Hayes-Roth, 1982). Vinson (1999) suggested that the types of landmarks to be included in VEs should follow Lynch’s (1960) categorization. These include paths, edges, districts, nodes, and landmarks. Each of these has a specific function, but each individual object can also have more functions than just one (Table 4). For example,
landmarks can also be used as focal points and reference points while traveling.
Type Examples Function
Paths Street, canal, transit line
Channel for traveler movement
Edges Fence, river Indicate district limits Districts Neighborhood Reference point Nodes Town square, public
building Focal point for travel
Landmarks Statue Reference point into which one does not enter
Table 4. Landmark types and functions (Vinson, 1999)
All five types (paths, edges, districts, nodes, and landmarks) should be included in a VE.
Because most individuals are used to navigating large-scale environments in the real world by wielding these types as reference points, it is important to include all of them in VEs. The designer of a VE usually has the benefit of choosing the type of landmark and its location in the environment. As stated before, the transfer of knowledge between VEs and the real world is relatively good if the level of fidelity and immersion is sufficient (Witmer et al., 1996). The relationship between landmark features and recalling them has been examined by Evans et al. (1984). Consequently, they presented a set of features that make landmarks and buildings more memorable or easier to locate. Some features can also contribute to both memorability (marked with m in the list below) and location recall (marked with l in the list below). These features (from Evans et al., 1984) are as follows:
Significant height m
Complex shape m
Bright exterior l
Large, visible signs m
Expensive building materials and good maintenance l
Freestanding (visible) lm
Surrounded by landscaping m
Unique exterior color, texture l
m Improves memorability
I Improves location recall
……………
47
In addition, Vinson (1999) suggested the use of landmarks from natural environments, including fabricated items such as roads, sheds, and fences;
land contours such as hills, slopes, and cliff faces; and water features such as lakes, streams, and rivers. From these elements, we can provide the third guideline:
Make the landmarks distinctive by using the features presented by Evans et al. (1984) for urban environments and Vinson (1999) for natural environments.
As a fourth guideline, Vinson (1999) suggests that designers:
Use concrete, non-abstract objects as landmarks.
Ruddle, Payne, and Jones (1997) concluded that memorable landmarks increase effective wayfinding. In their study, they used familiar three-dimensional objects such as cars. In the same study, they suggested that abstract objects such as complex paintings did not help the traveler in his or her task. In natural environments, any manufactured constructs stand out from the rest of the environment (Whitaker, 1996). Expert orienteers most often rely on land contours and water features in addition to synthetic constructs when traveling in natural environments (Whitaker and Cuqlock-Knopp, 1992).
A landmark must be distinguishable from its environment.
Landmarks presented in VEs should be distinctive compared to other nearby landmarks, as those objects that contrast with their surroundings stand out from their environment (Evans et al., 1984). Confusing one landmark with another is a very common mistake in wayfinding tasks, and in natural environments, this error has been named the parallel error (Darken and Banker, 1998). In addition, landmarks should have distinct sides that have enough differences so travelers can tell from which direction they are looking at it.
The saliency of a landmark can be increased by placing other objects nearby.
In some situations, landmarks can complement each other. For example, placing a colorful landmark among many monochromatic ones makes it prominent in a landscape, rendering it a good landmark for wayfinding purposes. For example, consider a landmark that is symmetrical on every side. It is very challenging for a traveler to discern the direction he or she is viewing it from. Inserting another, distinct object next to it makes the orientation easier.
Place landmarks along travel nodes and at decision points.
The memorability of a landmark is also affected by its location in the environment (Evans et al., 1984), especially if it is located on a major travel node or at an intersection. The most convenient place for a landmark is at the decision points, that is, those locations where travelers need to reorient themselves. Placing landmarks relatively close to each other also supports those travelers who rely on piloting (traveling from landmark to landmark).
For example, the following positions for a building contribute to their memorability (marked with m in the list below) and location recall (marked with l in the list below):
Located on a major path m
Visible from a major road lm
Direct access from a street lm
Located at an important choice point m
m Improves memorability
I Improves location recall
These general guidelines can be useful in designing landmarks and navigable content for VEs, but do not really consider realistic representations of urban or natural environments. For example, when using photographic images or videos for VE content, one can only choose the locations (e.g., a plaza or a street corner) that are presented. Each application presented in this dissertation provides this kind of VE content.
To design these environments, one should consider locations with prominent landmarks that are in high contrast with their surroundings.
Another model for landmark presentation was presented in Publication I.
This model highlights landmarks based on their saliency, which is calculated with three distinct properties: semantic, cultural, and structural.
This model was then evaluated in a collaborative VE (see Publication II).
Because many wayfinding experiments were conducted a long time ago, the technology for implementing more sophisticated three-dimensional VEs did not exist. Wayfinding evaluations are often performed in three-dimensional mazes where the participant navigates from the start point to the goal. In these mazes, landmarks are often presented with either three-dimensional models (see Figure 18a) or with two-three-dimensional icons (see Figure 18b). These models are not very good representations of the real world because of their simplifications. Interactive omnidirectional videos of real-world environments would be a better representation for wayfinding studies because of their realism. These representations were introduced in publications VI and VII.
……………
49 Figure 18. a) A three-dimensional maze where landmarks are presented as icons (Sharma et al., 2017). b) A three-dimensional maze where landmarks are presented as three-dimensional objects (Astur et al., 2016).
Gender Differences in Wayfinding in Virtual Environments
Most research on individual differences in wayfinding in VEs has focused on gender (Walkowiak, Foulsham and Eardley, 2015). Results from these studies generally suggest that males outperform females in spatial tasks, and that these differences are even larger in VEs than in real-world scenarios (Astur, Ortiz and Sutherland, 1998). Castelli, Corazzini, and Geminiani (2008) found that males perform more efficiently (i.e., complete a task faster) and make fewer errors than females while completing a wayfinding task where they must utilize a survey strategy. Moffat et al.
(1998) and Waller (2000) reported similar results in their studies. Lin et al.
(2012) have provided possible explanations for these distinctions. They suggested that males are more explorative in their wayfinding. Males also traveled large distances even when they were still not familiar with the environment. This was not evident in females, who adopted more conservative strategies during the wayfinding task. This exploratory nature of wayfinding among males was also reported by Coluccia et al. (2007).
Wayfinding experiments are commonly conducted with either the virtual Morris Water Task (vMWT) (Morris, 1984) or the multiple T-maze (Tolman, 1948). In the vMWT, participants try to find their way through a virtual water maze by using various navigational cues. The avatar in this maze is usually controlled by a mouse or a keyboard. Time and distance to locate the goal across trials are then used to describe the user’s wayfinding performance. Virtual corridor mazes contain a start location, interconnecting corridors, and a goal. Task completion time, errors made, and number of trials to the criterion are usually used for measuring performance (e.g., Moffat, Hampson and Hatzipantelis, 1998). The same variables apply to virtual wayfinding measures as to real-world wayfinding tasks, and more traditional tasks such as a mental rotation task and paper map tasks are used to complement the results from these wayfinding tasks performed in VEs (e.g., Astur et al., 2016).
Walkowiak, Foulsham, and Eardley (2015) suggested that one variable affecting wayfinding in VEs is computer experience. In their study, females
who reported more computer and video game experience completed the wayfinding maze task faster and made fewer errors during its completion.
Similar results were reported by Lin et al. (2012) and Head and Isom (2010).
Because less computer experience is detrimental to the user’s feeling of computer self-efficacy (Compeau and Higgins, 1995), some of these effects may be due to a lack of familiarity with this type of task.
In general, wayfinding experiments in VEs follow the same procedures as in real life. For this reason, the same differences between genders have been reported. For example, males recall non-vivid descriptions, which are generally harder to memorize, more effectively than females (Tom and Tversky, 2012). This difference diminished once the descriptions were made more vivid, thus benefiting both genders. Males have also consistently outperformed females in locating information, but there are also studies that did not find gender variations in information location (see, e.g., Tom and Tversky, 2012; Halpern, 2000; Wolbers and Hegarty, 2010). The problem with these evaluations is that they are simplified versions of real-life scenarios, and they often do not utilize the individual’s motor, vestibular, and proprioceptive systems. Taube, Valerio, and Yoder (2013) suggested that ruling out these factors needs to be considered when reporting any wayfinding measurements. More sophisticated VEs could provide these functions in addition to more traditional spatial tasks, thus complementing the results gained from these experiments.