Navigation and Wayfinding in Virtual Reality:
Finding Proper Tools and Cues to Enhance Navigation Awareness

by Glenna A. Satalich

CHAPTER 2

Literature Review

This review of the literature will report the effects of two different ways of presenting environmental information; map study and exploration.

Navigation

There was no difference in performance between the map-study groups. This finding suggests that once a person has studied a map to an error free criteria their navigational knowledge does not progress beyond that point. The map-learning subjects made larger errors in route estimations than Euclidean distance estimations. The navigation subjects showed the opposite effect; larger errors for the Euclidean distance than for route estimation distance. The group that had only one to two months exposure had greater Euclidean distance judgment errors than those who had more experience. Those in the navigation condition who had at least 6 months experience performed on an equal basis as those in the map-learning condition for giving Euclidean distances. This group also did better on the route estimations than the map-learning subjects. The orientation tasks provided evidence that Euclidean representations are better learned from direct experience then secondary survey knowledge. For both orientation tasks the navigation subjects were more accurate than the map-learning subjects. For object location, the most experienced navigation group performed equally well as the map-learning group. The navigation group with the most experience overall performed better than any of the other groups. In summary, the most experienced navigation group attained complete navigational knowledge, and primary survey knowledge was superior to secondary survey knowledge.

Goldin and Thorndyke (1982) examined the different types of knowledge developed from a direct navigation experience and from a simulated experience. The direct navigation experience involved a bus tour of an area. The simulated experience was a filmed auto trip of the same area. In both tours the subjects were passive observers. An additional factor was the other type of information subjects received, a verbal description during the tour, prior map study or no additional information. The map study consisted of looking at a map with landmarks and routes for 10 minutes prior to taking the tour. The narrative provided during the tours gave names of the streets on the route, landmarks, the distance between intersections and the current compass direction.

Participants performed 6 tasks after the tour. The tasks were; location recognition, location sequencing, landmark location, route and Euclidean distance estimations and taking a basic spatial ability test. The spatial ability test was the Kit of Factor-Referenced Cognitive Tests (French, Ekstrom, and Price, 1963). These are tests of visual memory, spatial visualization and perceptual independence abilities.

The subjects in the film group identified tour locations better than those who were on the actual tour. Within the film group the control group performed better than the two groups with added information, narration and map study. This was not the case for the subjects on the real tour, where there was no difference. For the location-sequencing task film groups again performed more accurately than the those on the real tour. The narrative condition had a negative effect in performance across both conditions. The orientation test provided opposite results. Those on the real tour were about 10 degrees more accurate than those in the film group. While both groups contained subjects who were more than 90 degrees off the mark (complete disorientation, Ciccone et. al., 1978), there were significantly more subjects in the film group who were disoriented than in the real tour group. Both the film and the real tour groups who were allowed to study a map beforehand, did worse on the orientation task than those in the control and narrative groups. The route distance estimation tasks revealed no differences between the groups.

Landmark placement and Euclidean estimations tasks also showed no differences between the film and tour groups. In post-hoc tests though, differences did appear for those in the film group having access to additional information. Those who had the map to study beforehand showed a higher performance than all of the other five groups, and those who had the narrative had a lower performance than the other five groups.

The results of the study indicate that people can learn about environments from a simulated medium such as film. In both conditions though, the subjects were passive. Thorndyke suggests that interaction, either by a person driving the tour or interacting with the environment in a simulation, might show different results. The results of this study also show that within a condition, such as the film condition, adding an additional navigational aid has consequences and that the consequences are task dependent. In one test we see that having a map or narration hinders performance, while in another task the map improves performance and narration lowers it. These results are important when deciding what aid to use to introduce someone to a new environment.

Wayfinding

In 1985 Streeter, Vitello and Wonsiewicz performed a wayfinding study comparing navigational aids for people driving in a car. There were 7 routes chosen for the subjects to drive. They ranged from 3 to 20 miles. The routes were divided into three categories. The first category, limited access, was defined as one that used the major highway for more that 50% of the route, but contained seven turns within the route. The second category, moderately difficult were local routes where each had 6 turns. The final category, were complicated local road routes where each had 13 turns. There were 2 routes for each category, and a subject was asked to drive 1 of the 2 routes from each category. All subjects also drove a trial route that was four miles long and contained only three turns. The comparison was between those who had a tape recorded narrative of where to go, those who had a customized route map of the area, those who had both, and those who had none. In the last condition, the drivers were instructed to use any method they wanted to find their way using a standard road map. The measures recorded were time (seconds) to target, mileage, number of turns, and errors. Errors included turning in the wrong direction, turning on the wrong street, and passing the destination. For the drivers who had the tape recorder, the number of rewinds was also recorded. The experimenter sat silently in the back seat during the test.

The control group, those without any navigational aids given to them by the researcher, drove significantly more miles, took more turns than necessary, and took more time completing the task, than the other three groups. Within the experimental conditions those who had the narrative drove fewer miles than those with the customized map. The narrative group also took marginally less time than the customized map group. The customized map group made significantly more errors than both the narrative group and the customized tape - narrative combination group. The combination group made marginally more errors than the narrative only group.

This study showed that while wayfinding it is more difficult to interpret a map than it is to receive directions. The finding that the combination of tools did not produce better performance than the narrative alone is what is important, since more tools may not mean better performance.

Regian and Shebilske (1990) conducted studies of the use of VR as a training medium for visual-spatial tasks. One of their experiments involved wayfinding. The environment used in the wayfinding study was a virtual maze. It consisted of three stories, with four rooms in each story. Each room was the same size and each contained a unique color-coded object, either a star, cube sphere, or pyramid. The color could be red, green or blue. Every room was connected to at least one adjoining room by a hallway or passageway leading to a room above. The walls were all colored gray, the floors red and the hallways yellow. Subjects were given verbal directions on where to move through the virtual environment. Three different tours were taken. Following the tours the subjects were free to self-explore the environment for one hour. The testing phase of this experiment consisted of 3 wayfinding tasks. The subject started in one room of the complex and then tried to take the shortest path to a room having a target object. The first route could be completed by visiting a minimum of eight rooms, the second route in four rooms and the third route in six rooms. A perfect result would mean that all of the rooms would be visited at least once for a total of 18 rooms. The dependent measure was the number of rooms traversed for each wayfinding task.

The authors performed a comparison between these data and data from a random walk algorithm and found significant differences. The results of this study showed that the mean number of rooms for each tour was 10.1, 4.8 and 6.7. The authors claim that this shows that subjects can learn navigation spatial-navigational skill in a virtual environment. However, their design and methodology does not support this claim. The major problem was the Monte Carlo method used to show that learning occurred. By using this method a person could enter a total of 14 rooms for the first tour (8 rooms was the minimum) and still show a significance learning effect. The performance in the last tour (6 rooms minimum) could not be ruled out by chance. Previous research has already shown that "people do not act like randomly moving automata that make unbiased decisions at each point where a decision has to be made" (Peponis, et. al., 1990). Two rules that Peponis observed in wayfinding behavior, is people avoid unnecessary backtracking and also that people tend to find the area that gives them the best visual access to other areas. Regian, et. al., should have run a control group for the same wayfinding task. The control group would have had no experience in the building, and therefore would have had to search for the unique object, using strategies that would probably be much more efficient than a random search.

Butler, Acquino, Hissong and Scott (1993) conducted 4 experimental studies examining newcomers wayfinding in a complex building. Only the first three studies will discussed. Their first study examined whether signs or the use of You-Are-Here (YAH) maps would benefit people who were visiting a building for the fist time. These two experimental conditions were contrasted to a control group who had no added information. The dependent measure was the amount of time it took to reach two different rooms. The results showed that for the first room those with the maps took significantly longer than those who had no aids, and both of these groups took significantly longer than those who used the signs. For the second room, again those with the maps took longer than the other two groups. Those who used the signs did not differ from the control.

The second study by Butler et. al., (1993), was initiated because there were multiple routes to most locations in the building, and the best signed route had to be determined. This study examined possible routes to see if people preferred, and would use, the simplest (those with the least decision points) or the most efficient (the shortest distance). Subjects examined 7 routes leading to one location, by viewing a video tape of each route. While the video tape was playing, appropriate floor plans would be displayed, and a narrative was heard pointing out turns, decision points, etc. Subjects were then asked to rate each route on a scale of 0 to 10 with 0 being a route they would never take and 10 being the best route. After rating the routes for their personal preference, subjects were given comments about each of the routes by people who were confined to a wheelchair or visually impaired. The subjects were then asked to rate the routes as if they were confined to the wheelchair and then as if they were visually impaired. A final rating was then requested for the best overall route keeping all of the factors they had learned about in mind. The results showed that the subjects preferred routes that conserved energy (the shorter routes), regardless of how complex they were. Significant rating changes were made depending on the physical impairment being asked about. The final ratings showed that the shortest routes were picked but weighted more for those in a wheelchair than those visually impaired.

The third study generalized the results of the second. The second study only looked at routes that were going to one location, while the third study examined a single starting point with two routes going to 2 different locations. Subjects were told that the end locations were either bathrooms or vending machines. The routes were not real routes but were derived by crossing complexity ratings (Leeuwenberg, 1967) and distance. Subjects reviewed 56 pairs of routes and again distance was found to be the significant factor that determined which path they chose. Shorter distances were chosen despite their complexity. Both the second and third studies show that efficiency based on distance is more important than complexity. However the authors do raise the point that complexity may be more important for wayfinding in a building when verbal directions are given.

The wayfinding literature indicates that receiving directions verbally or through signs is advantageous when trying to find a goal. In contrast, the use of maps can be disadvantageous. What these studies did not do is have the subjects repeat their performance without any navigational aids in order to test their navigational awareness. This factor may not be important in structures that are rarely visited, but ideally for people who work in a building this factor could be important, such as during an emergency in a hospital, or if a building was on fire. A person may not have the time or opportunity to read signs in a building or ask directions. A series of studies conducted by Moeser (1988) tested student nurses in a hospital and the results showed that they did not have survey knowledge of the building even after 3 years. Moeser mentions anecdotally that the nurses verbally reported to her that they use the signs in the building to locate areas within the hospital. Which may be evidence that having signs relieves the cognitive load to the point where survey knowledge is inhibited. The question that this thesis will address is, what navigational tools under different exploration conditions would benefit wayfinding at a later period where the tools would not be available.

Darken and Sibert, (1993) reported on an informal study looking at toolsets for wayfinding in virtual environments. The tools available to the participants were flying (the ability to rise above the virtual environment), spatial audio markers, visual markers (breadcrumbs), coordinate feedback, grid navigation and two map-views of the world. The map-views available were track-up and North-up. Track-up maps change dynamically so that the user is always represented in the middle of the map, and the map revolves so that the users's forward view is always presented at the top of the map. A North-up map has a similar central representation of the user, but the representation rotates not the map. North is always representated at the top of the map. Only one type of map could be chosen and seen at one time. While the mechanics of how to use the tools were explained, the benefits and what information could be retrieved from them were not. Only one type of tool was available to the subject during each scenario for each condition. The environment was a large landscape with grid markings on the floor. The landscape was sparsely populated with objects such as ships and rectangles. The focus of this study was to examine how subjects would use a specific tool under three different types of conditions. The first condition was just to explore the environment. The second condition was a naive search, where the subject knew what the object looked like but did not know its location. The final condition was an informed search, where the subject knew both the description and location of the object. The subjects were instructed to search for the target. When the subject was close to the target a bell would sound and they were told to return to the starting position as efficiently as possible. The subjects were encouraged to talk aloud as they moved through the environment and as they used the navigation tool to help them.

The coordinate tools were mainly utilized in two ways. The first was at the start position, where the subject would move along one axis of the grid lined floor and note the feedback. During the return portion of the task they commented that they had remembered their initial position. The breadcrumbs were used more as landmarks than as trail making objects, which they were originally intended for. The first breadcrumb was usually dropped at the starting position. It was also observed that groups of breadcrumbs were dropped in a shape to show directional information.

Informal observations indicated that people used the different tools in a variety of ways. One of the most useful tools was a synthetic sun, which improved performance in both the search and return phases. In fact before the sun was added all subjects moved in the incorrect direction in the homing phase. The conclusion of this report was that subjects showed different behaviors when they used different tools in wayfinding.

Summary

The literature on navigation and wayfinding brings to light that at present we are unsure of how to introduce a person to new environments so that they will gain navigational awareness efficiently. The Thorndyke and Hayes-Roth (1982) and Goldin and Thorndyke (1982) studies show that exposure time and navigational tools can affect the process. Their studies revealed that map study before entering an environment can be beneficial but that when used alone it does not give someone complete navigational awareness. The research also points out that depending on the type of introduction to an environment, an additional navigational aid can have a positive or negative effect. The results also show that learning of an environment can occur without actually being there but the results do not reveal whether it is better to be an active participant or a passive observer when being exposed to a new environment.

The wayfinding literature shows that people are better at finding a target location when using signs or narrative directions than with using maps, and that adding navigational tools can have a positive effect or a negative effect depending on the task and the original tool or cue used. The research in wayfinding also showed that given a choice people prefer shorter routes to longer routes to a destination despite the complexity of a shorter route.

Both sets of literature point to a need to examine in further detail how to introduce someone to a new environment, what navigational aids are beneficial during initial exposure and their after-effects when the aid is may no longer available.