The Gender Factor in Virtual Reality Navigation and Wayfinding Joaquin Vila, Ph.D. Applied Computer Science Illinois State University javila@.ilstu.edu Barbara Beccue, Ph.D. Applied Computer Science Illinois State University bbeccue@.ilstu.edu Sachin Anandikar Applied Computer Science Illinois State University Abstract An area of interest that remains to be more thoroughly investigated is that of human navigation and wayfinding in VR. This study attempted to examine the effect of gender on VR navigation and wayfinding. A tool named Virtual Reality Navigation Tool (VRNT) was designed to conduct experiments for this research. Four experiments were conducted to track navigational patterns employed by the subjects while interacting with the D maze environment. As each subject was performing a prescribed task, the VRNT was tracking navigational behavior, specifically, the number of turns taken, time spent and the number and order of rooms traversed. The analysis of the tracking revealed some common navigational behaviors. The study results suggest that the tendency to take left and right turns is influenced by gender; whereas, the time traveled and numbers of rooms visited are not.. Introduction Virtual Reality (VR) is an artificial environment that is experienced through sensory stimuli as sights and sounds provided by a computer. In such environments, the user s actions partially determine what happens in the environment. With modern advances in technology, virtual reality is being increasingly utilized in a wide range of applications such as avionics, visualization of architectural modeling and games. Users navigate these spaces through immersion interfaces; however, not much research has been done on the usability and adaptability of these interfaces. An area of interest that remains to be more thoroughly investigated is that of human navigation and wayfinding in VR. Having a better understanding of navigational patterns employed by humans when interacting with these environments will enable software developers to design more intuitive and effective VR interfaces. The goal of this research project was to determine prominent patterns that subjects employ while navigating through a VR environment. In order to study human factors that affect navigation in virtual reality, differences in patterns of navigation and wayfinding based on demographic characteristics were studied. Specifically this study looks at the effect that gender has on navigation and wayfinding in VR... Significance of Research Project Navigation is one factor that has been researched in both real world situations and the two-dimensional world of user interface design (UID). Researchers have gained some insight into the navigational patterns of different groups of users based on user characteristics [,, 0] As researchers and interface designers work with threedimensional worlds, they need to know whether the findings related to two-dimensional interfaces or to real world behaviors still apply. The research in wayfinding and navigation in the real world and traditional UID has shown that signs, landmarks and maps enhance navigational knowledge, but there has been little research done with respect to the effects of demographic differences in navigation of VR environments. This study attempted to find differences in the navigation patterns employed by subjects of different gender. The outcome of this research could aid VR developers in designing more effective applications given users demographic profiles... Definition of Navigational Terms To better describe the process and research background for this project, it is important to define some key words and terms that are frequently used in VR research. Naive Search is a task that involves searching for a target when the subject has no prior knowledge of the whereabouts of the target. This type of search implies the need for an exhaustive search [6]. Primed Search is a task that involves searching for a target when the subject knows the location of the target. This type of search implies the use of a non-exhaustive search [6]. Exploration is any wayfinding task in which there is no target [6].
Navigation Awareness is defined as having complete navigational knowledge of an environment []. Wayfinding is the process used by a person to orient oneself and navigate in an environment []. A small world is one that can be seen from a single viewpoint. A large world, as defined by Kuipers and Levitt [8], is the space whose structure is at a significantly larger scale than the observations available at an instance. Density of a VR can be categorized into three groups: sparse, dense and cluttered. A sparse world is a world that is open and has few navigational aids. An open sea or large empty room modeled in VR with doors on all sides is an example of a sparse world. Dense worlds can be characterized by the presence of relatively large objects and clues for navigation. Examples of these types of worlds can be urban areas with very closely placed buildings. A cluttered world is one that has many clues. A library is a cluttered world []. Activity of a VR environment depends on the mobility of objects over time [6]. In a static world, properties of objects and their locations do not change over time; on the other hand, in a dynamic world, objects are moved about thus increasing the complexity of navigational tasks. A static world provides a more controlled environment. Procedural knowledge is ego referenced and usually gained by personal exploration of a new area []. A navigator is said to have procedural knowledge when the navigator can successfully go from one landmark to another on a single known route but does not recognize alternative routes. Survey knowledge recognizes alternate routes between landmarks. It is attained through multiple explorations of an environment using multiple routes. With survey knowledge the mental representation of the world is seen from a bird s eye point of view []... Navigation & Wayfinding The techniques for navigation in virtual worlds are derived from navigational techniques in the real world. Extensive studies have been conducted on human behavior in navigation and wayfinding in the real world by [, 7]. These studies in real world navigation suggest that exploration and the study of maps are two ways of familiarizing oneself with the environment. Satalich [] conducted research in the area of navigation and wayfinding. In this study a comparison was made between an experimental and a control group in a VR environment. The results indicated that having a map before entering the virtual environment improved performance but exploring the virtual environment did not. Satalich observed that the navigational decisions of the subjects were not random. She noted that subjects, even when lost, would not backtrack without purpose. In addition, Satalich s [] wayfinding research indicates that people prefer a shorter route to a longer route despite the complexity of the shorter route. Satalich suggested that wayfinding performance is better when navigational tools like marks and signs are used. Darken and Sibert [] considered cognitive map formation and map design by cartographers and planners. The study concluded that subjects take advantage of environmental clues in predictable ways to facilitate an exhaustive search. They use the clues when they are highly predictable in motion and visible from the entire environment. Audio and visual clues can be combined to make a target easier to find. Darken and Sibert [6] also suggested that the tools people use strongly influence their behavior. Further, Darken and Sibert [5] suggests that lack of directional clues and spatial organization leads to ineffective search strategies and frequent disorientation. Some conclusions drawn from their work are:. Wayfinding is disoriented if proper and adequate directional clues are not used.. In a large VR environment the exhaustive search is difficult, if not impossible.. A structure must be imposed on VR to make it easier for the subject to chunk the environment.. Maps can be used to optimize the search strategy since they supplement the survey knowledge. 5. Dead reckoning, which is the ability to infer position from past locations over time, is natural and intuitive in VR. To achieve complete navigational knowledge in a new large-space environment, Siegel & White [] have described a process called the "Sequential and Hierarchical" model as described below. The main steps in the Sequential and Hierarchical model include:. Landmark Recognition: A landmark is an object that can be identified by distinct shape, size and color. Objects can give directional instructions thus becoming salient landmarks. Some objects have personal meaning and may become landmarks.. Routes or Links: Routes or links are formed after traveling between two landmarks. Images and landmarks are formed while traveling, then
recalled while again traveling the same route or link.. Primary Survey Knowledge: This is achieved after traveling significantly so that straight-line distances and alternate routes between landmarks can be determined.. Secondary Survey Knowledge: This type is added by Satalich [] because a subject can attain partial survey knowledge solely through the use of maps to learn about an environment. 5. Chunking of the Environment: It is necessary to chunk a large environment into smaller regions. Landmarks located in different regions may be susceptible to slight distance distortions. This can be visualized as clustering of small towns to cities, then to states, and then to countries. This helps users zoom in and out of the representation. Because of chunking, the virtual environment is divided into smaller parts. The first two steps in the Sequential and Hierarchical model are necessary for procedural knowledge, and the third and fifth are necessary for complete survey knowledge. The fourth step of secondary survey knowledge is an intermediate between the procedural and survey knowledge. This study will deal with procedural knowledge and survey knowledge issues in virtual worlds. Satalich [] classified the concept of wayfinding in the following four steps:. Orientation: Determining where one is in respect to nearby objects and the target location.. Route Decision: Choosing a route that will get one to a destination.. Route Monitoring: Monitoring the route one has taken to confirm that one is on the correct route and is going in the right direction.. Destination Recognition: Recognition that one has reached the correct destination, or at least a point nearby. Regian and Shebilske [9] conducted a wayfinding study in a virtual reality environment that was a virtual maze consisting of three stories with four rooms on each floor. Each room contained a unique color-coded geometric object. After being taken on three different verbally directed tours, the subjects were given one hour of free time to explore the environment. Following this learning time, the subjects were given wayfinding tasks in which they had to find a specified object. They were told to take the shortest path to the specified object. The researchers found significant differences between the number of rooms visited by the subjects during each task and the number of rooms visited by a random walk algorithm for each task. This research addressed the ability of subjects to learn navigational skills in a virtual environment; however, it did not indicate how they learn i.e. by self-exploring, by verbally guided tours or by a combination. The study did not investigate navigational patterns that were followed by the subjects. In reviewing literature related to navigation patterns in Human Computer Interaction (HCI), it was found that the research involving navigation in three-dimensional environments focuses on navigational awareness, spatial ability or wayfinding []. The two-dimensional research on navigation patterns more closely aligns with wayfinding than with the other two areas. Furthermore, it seems as if navigational awareness is the result of exploring an environment to gain procedural and then survey knowledge; whereas, wayfinding is a dynamic process that uses navigational knowledge to reach a destination. However, the literature on navigation and wayfinding indicates that VR designers are still not able to conclusively decide on the way in which one should be introduced to a new VR environment. Since research in VR is lacking information on how gender affects navigational patterns, more research is required in this area. Further inquiry into initial exposure to virtual worlds and the after-effects when the navigational aids are not available could provide information to help designers to be more aware of the user needs. More knowledge and better understanding is this area could aid in the design and creation of more natural and intuitive ways to navigate in VR. In an attempt to address issues of navigation and wayfinding, research carried out by Vila, Beccue and Furness [] gives insight into many important issues to be considered while designing the tools that could facilitate research in VR navigation. They designed a flexible system named Virtual Reality Navigation Research Tool (VRNRT) that relieves the researcher of the time-consuming process of designing and creating their own research tool. VRNRT allows a researcher to track navigation patterns while subjects navigate a customized VR environment.. VRNT Use for Design To facilitate experimentation on VR navigation, a Virtual Reality Navigation Tool (VRNT) was designed and implemented. VRNT allowed the user to explore a three-dimensional, polygon graphics environment. VRNT bases its maze design on the work of Darken and Sibert [5], Satalich [] and Vila et al []. The maze was chosen as the VR environment as it offers a simple world with sparse density [8]. The researcher can specify certain factors that may have an effect on the user s navigational decisions. Thus the study of these factors
and user s reaction to them will be enabled through the use of VRNT to customize a VR environment and to track the user s actions. VRNT incorporates the following components: a demographic survey, D maze builder, tracker log and navigation pattern visualizer. The demographic survey enables the researcher to gather and record information about the subjects such as age, gender, level of education and computer literacy. The D maze builder permits the researcher to specify the size and the color schema of the maze. While the subject is navigating the environment and completing the tasks, his/her movements are tracked and logged at predetermined time intervals by the tracker log. More specifically, the user s coordinates are recorded every t seconds. The log of the subject s movements throughout the virtual reality environment contains timestamps, coordinates (for straight line navigation) and angle (for turning navigation). The pattern visualizer allows researchers to replicate navigational paths employed by subjects who have used VRNT. The visualizer provides a top view animation of a navigational path by color-coding its progression over time. VRNT has facilitated research efforts into navigational preferences since it has relieved the researcher of the time consuming and tedious effort of constructing an appropriate VR environment in which to conduct experiments. In addition, the fact that it is Webbased increases its potential subject base. With the proliferation of the Internet, subjects can be selected from all over the globe to sample their navigational preferences and to gain an understanding of cultural differences if any. This tool was designed to facilitate research in several areas related to the navigation of three dimensional computer environments such as navigational knowledge, navigational preferences and wayfinding. VRNT allows customization of D mazes. Through maze designs, researchers can evoke a subject s natural route decision making, if any. For example, a maze design could depict a building with multiple rooms, each having four doors. If the subject were free to explore the environment, what would the navigation path be? Are there any demographics such as gender, age, and righthandedness that correlate to navigational patterns? Are there any environmental factors, such as color of the walls, size of the rooms, which affect navigation? Do some maze designs that affect the sense of orientation more than others? In a maze, how many turns can a subject recall before complete disorientation occurs?. Methodology In order to answer the questions posed in this research a set of experiments were prepared using VRNT. All of the experiments were conducted using mazes organized as 9 x 9 matrices of rooms. The top view of the tailored VR maze is shown in Figure. Each of the rooms is identified with a number. Each room has four walls and only one of the walls in the maze has a different color. The color of the walls for all the rooms is gray with the exception of the one wall that was green. Four experiments were designed and conducted with a convenience sample. Each experiment consisted of tracking navigational patterns employed by subjects when traversing the VR maze. The task given to the subjects in each of the four experiments was to find the greencolored wall. The bold numbers in Figure show the actual location of the colored walls for each of the experiments. For each experiment the colored wall was placed in a different quadrant given that the Cartesian origin was room. In all of the experiments, the user started in Room and was facing north. Figure shows the VR maze as it appeared to the subjects. Figure. VR Maze Design
mean varied from 8.8 to.9. The mean was consistently higher for females than males. Table Mean Number of Left Turns Taken in Each of the s by Gender Mean Number of Left Turns Taken Gender Male 8. 9.0 5.5 7.97 Female 0.7.90 0. 8.8 Total 8.80 0.98 6.50 7.78 Figure. Subject s view of the VR maze. Sixty-eight subjects ( females and 7 males) participated in the study. Prior to the actual experiment, participants completed a demographic survey. All participants completed the same set of experiments. The navigational data collected from the 68 subjects was refined and converted into four measurements for each experiment, namely number of left turns, number of right turns, number of rooms visited and total time spent. Descriptive Statistics and t tests were employed respectively to understand the sample set and determine significant differences between navigational patterns based on gender.. Results The descriptive statistics for the dependent variables are shown in Tables through. Table shows the mean time spent in all four experiments by gender. The mean time spent ranged from approximately two minutes (8. seconds) to over four minutes (76. seconds). The minimum and maximum time spent ranged from less than one minute to more than ten minutes. The mean time was less for males than for females in each of the experiments. Table Mean Time Spent in Each of the s by Gender Gender Mean Time Spent in Seconds Male 5.0 88.7 8. 59.0 Female 76. 5.8 0.6 0.9 Table shows the mean of the total number of left turns taken in each experiment by gender. The mean for males varied from 5.5 to 9.0; whereas, for females the Table shows the mean of the total number of right turns taken in each experiment by gender. The mean for males varied from 5.7 to.7; whereas, for females the mean varied from 8.85 to.. The mean of the total number of right turns taken does not have any consistent pattern among the four experiments. Table Mean Number of Right Turns Taken in Each of the s by Gender Gender Mean Number of Right Turns Taken Male.70 0. 5.7.56 Female 9.. 8.85 0.9 Total.68.65 6.78 0.9 Table shows the mean of the total number of rooms visited in each experiment by gender. The mean for males varied from 9.09 to 50.9; whereas, for females the mean varied from.85 to 6.. The mean was higher for females than males in three of the experiments. Gender Table Mean Number of Rooms Visited in Each of the s by Gender Mean Number of Rooms Visited Male 7..86 9.09 50.9 Female 5.8 6..85. Total 8.97.5.8 8.0 To analyze the results and answer the research questions, t-tests were used to determine if there were any statistically significant differences between the navigational performance of males and females in the experiments. Tables 5, 6 and 7 show the results of the t tests.
In order to answer the first research question (Ho ) that gender has no effect on the way subjects navigate in VR, in terms of number of left and right turns taken while traversing a VR environment, the two dependent variables, number of right turns and number of left turns, were combined into a ratio: left turns/right turns. This ratio helped in reducing the error in statistical calculation, as the ratio focused on the tendency of turning (either to the right or to the left) as opposed to just the number of right turns and the number of left turns taken by each subject. Table 5 shows the t-test results for the ratio of left to right turns by gender. The resulting t value (t=-.8) was significant at the p =.05 level. Thus, in the first exposure to the VR environment (experiment ), there is sufficient evidence to reject the null hypothesis (Ho ) that there is no significant difference between the performance of male and female subjects in the tendency of turning left or right. In examining the t-test results of the experiments reported in Table 5, the researchers noted that there was a decreasing effect of gender on turning tendency as exposure to the VR environment increased. Table 5 t-test for Ratio of Left to Right Turns by Gender Ratio of Left To Right t-test Ratio P Turns -.8 0.0* -.8 0. -.096 0.77-0. 0.87 * p <.05 Table 6 summarizes the t-test results used to answer the second research question (Ho ), that gender has no effect on the way subjects navigate in terms of number of rooms traversed in a VR environment. The resulting t values for each of the experiments were not significant at the p =.05 level. Thus, the null hypothesis (Ho ) was accepted. Table 6 t-test for Number of Rooms Traveled by Gender Number of Rooms t-test Ratio P Traveled -0.880 0.8-0.80 0.780 0.86 0.69 -.0 0.97 Table 7 summarizes the t-test results used to answer the third research question (Ho ), that gender has no effect on the way subjects navigate in terms of time spent in a VR environment. The resulting t values for each of the experiments were not significant at the p =.05 level. Thus, the null hypothesis (Ho ) was accepted. Table 7 t-test for Time Spent by Gender Time Spent t-test Ratio P -0.8 0.656 -.86 0.7-0.068 0.96 -.86 0.07 5. Summary This research project investigated the effect of gender on VR navigation and wayfinding. Four experiments were conducted to track navigational patterns employed by the subjects while interacting with a D maze environment. As each subject was performing a prescribed task, VRNT tracked navigational behavior, specifically, the number of turns taken, time spent and the number and order of rooms traversed. The analysis of the tracking revealed some common navigational behaviors. The study results suggest that the tendency to take left and right turns is influenced by gender during the initial exposure to a VR environment. The researchers noted that there was a decreasing effect of gender on turning tendency as exposure to the VR environment increased. Furthermore, in this study, the time traveled and numbers of rooms visited in the VR environment was not influenced by gender. There is a need for more research to determine additional factors that affect navigational behavior in VR environments. This research is needed to gain a good understanding of navigational patterns employed by different types of users with the intent to formulate VR design guidelines that could aid VR designers in creating more effective VR applications for specific groups. 6. References [] Beasley, R. E. and Vila, J. (99). The Identification of Navigation Patterns in a Multimedia Environment: A Case Study, Journal of Educational Multimedia and Hypermedia. Vol., No.. pp. 09-. [] Beccue, B. and Vila, J. (99). AIT: An Analytic Tool for Identifying User-Preferences in a Multimedia Environment, International Conference Proceedings of PRIISM-9. Maui, Hawaii, January, 99. pp. -.
[] Butler, D. L., Acquino, A. L., Hissong, A. A., and Scott, P.A. (99). Wayfinding by Newcomers in a Complex Building. Human Factors, 5() 59-7. [] Darken, R., & Sibert, J. (99). A Toolset for navigation in virtual environments. In the Proceedings of ACM User Interface Software & Technology, 57-65. [5] Darken, R., & Sibert, J. (996a). Navigating Large Virtual Spaces. International Journal of Human-Computer Interaction, 8(), 9-7 [6] Darken, R., & Sibert, J. (996b). Wayfinding strategies and behaviors in large virtual worlds. In the proceeding of ACM CHI 96: Human Factors in Computing Systems, -9. [7] Goldin, S.E., and Thorndyke, P.W. (98). Simulating Navigation for Spatial Knowledge Acquisition. Human Factors, (), 57-7. [8] Kuipers, B. J., & Lewitt, T.S. (988). Navigation and Mapping in Large-Scale Space. AI Magazine, 9(), 5-. [0] Rendon, G. (99). The Identification of User Preference Patterns in a Multimedia Learning Environment, Unpublished Master Thesis, Applied Computer Science Department, Illinois State University. [] Satalich, G. (995). Navigation and wayfinding in virtual reality: Finding proper tool and cues to enhance navigation awareness. Unpublished master s thesis, University of Washington, Seattle, WA. http://www.hitl.washington.edu/ publications/satalich [] Siegel, A.W., and White, S.H., (975). The Development of Spatial Representation of Large-Scale Environments. In H.W. Reese (Ed.), Advances in Child Development and Behavior. New York: Academic Press [] Vila, J., Beccue, B., & Furness, G. (998). User Interface Design for Virtual Reality: A Research Tool for Tracking Navigation. Proceedings of the st Hawaii International Conference on System Sciences (HICSS'98). [9] Regian, J.W. and Shebiske, W. L. (990). Virtual Reality: An Instructional Meduim for Visual-Spatial Tasks, Journal of Communications. Vol., No.. pp. 6-9.