by Glenna A. Satalich
Because of time limitations, equipment failure, subject no shows and four subjects reporting simulator sickness; only data from 57 subjects were analyzed in this study. At least four subjects participated in each cell. All statistical data analysis was conducted with SPSS for Windows. The means reported in this thesis are weighted means. This approach produced results by having the means dependent on the number of subjects in each cell, not just the mean of the cell itself.
All of the results reported can be assumed to have met the assumptions for analysis of variance, unless otherwise stated. These assumptions are homogeneity, normality and independence. Homogeneity was measured by using either the Bartlett-Box or Cochran method. Normality was observed by reading histograms and Boxplots of the data.
There were reliable male-female differences; F(1,56) = 4.39, p <.05. The means for gender were; male (n=38), M=22.04 s(SD=8.08); and female (n=19), M=17.03 s(SD=7.6). The Guilford Zimmerman reports descriptive statistics for college students, the mean for college aged males is 20, s(SD=10.32) and for college aged females the mean is 12 s(SD=8.67).
Table 2 Means for The Guilford Zimmerman Test
Grand Mean = 21 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 16 19 15 23 25 20 no Map During 22 16 14 20 28 20 21
An interaction of the treatments; exploration and map during, also was significant F(2,50)=3.14, p <.05. The means for this interaction can be found in Table 4.
The means show that for those in the Active Guided condition of exploration, not having a map had detrimental effects. A possible explanation for this result, is that subjects in the active guided group were often observed looking down at the path and not observing their surroundings. If they did not have the map during, while looking down, then they would have little information about where they were in the building. A graph of the interaction can be seen in Figure 15. A total of eight subjects had averages over 90 degrees revealing complete disorientation. Seven of these were in the experimental conditions and one in the control condition.
Table 3 Means for Orientation (average degrees away from targets)
Grand Mean = 61 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 69 45 53 56 82 71 no Map During 56 64 46 58 52 87 68
Table 4 Orientation means for Explore by Map During Interaction
Self Exploration Active Guidance Passive Guidance Map During 62.9 64.2 62.3 No Map During 54.8 75.4 57.6
Figure 15 Orientation Interaction of Exploration and Map During
Grand Mean = 21 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 10 23 19 19 26 26 no Map During 13 27 22 13 30 20 19
An interesting interaction was also revealed for the map before and map during; F(1,39)=3.89, p < .05. The means are given in Table 7 and shown graphically in Figure 16. The means show that people do their worst when they never have a map for route distances. Contrasts for the route estimations between two landmarks proved significant. The first contrast looked at those who studied a map and explored the environment versus those who only studied the map but had no map in exploration; t10.7=3.54, p<.05; map before with no map during M=39.38, control M=17.1. The second contrast compared those who explored the environment with maps neither before nor during, and the control subjects who studied the map but had no exploration; t10.8=6.45, p<.05, no maps M=66.13, control M=17.1. This second contrast again shows that secondary survey knowledge produced more accurate estimations than those with procedural knowledge. The control contrasts clearly showed that exploration of the virtual environment caused a deterioration in estimation performance.
Table 6 Route Estimation Means (between 2 objects)
Grand Mean = 46 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 57 39 37 57 46 44 no Map During 50 29 38 17 66 77 54
Table 7 Route means for Interaction of Map Before by Map During
Map Before No Map Before Map During 44.8 41.8 No Map During 47.3 66.3
Figure 16 Route Est. Interaction for Map Before by Map During
The Euclidean distance estimation tasks were used to measure survey knowledge. In the first task subjects used themselves as a reference point to give straight line distances to the landmarks they had previously pointed at and given route estimations to. The second task was giving straight line distance estimations between two objects. The hypothesis was that those who had a map to study beforehand would be better at the distance estimations than those who did not. It was also hypothesized that the control group would do better than those who did not have a map at all because of the benefits of having secondary survey knowledge.
The ANOVA results for the first task did not reveal any significant differences between the experimental conditions or contrasts with the control condition (means are given in Table 8.) The ANOVA for the second survey estimation task did show a benefit from studying the map before going into the virtual environment; F1,39=7.30, p < .01, map before M=16.86 s(SD=11.8) and the no map before M=28.90 s(SD=10.0). Means for all conditions are given in Table 9. The first contrast comparing the control group to those who explored the environment and did have the map to study before was not significant. The second contrast indicated that the control group was superior in judging straight line distances than those who did not have a map to study did show significance (t8.0=2.531, p<.05). The mean for the control group was M=11.93 s(SD=9.16) and for contrasting groups M=24.80 s(SD=10.1).
Table 8 Straight Line Distance Estimations (Self as Reference: absolute distance from true distance in feet)
Grand Mean = 18 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 11 17 18 20 26 21 no Map During 16 19 36 14 19 15 19
Table 9 Straight Line Distance Estimations (Between 2 Objects: absolute distance from true distance in feet)
Grand Mean = 19 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 15 22 15 26 17 25 no Map During 16 20 10 12 29 21 23
These results show that having a map before entering the virtual environment was beneficial and that exploring the virtual environment did not add to performance beyond the information gained from the map.
Wayfinding is possibly the most valid way to assess navigational knowledge. The first wayfinding task was to find the classroom where subjects had seen two whiteboards on the wall saying "I will not talk in class" (figure 17). The guided tours subjects saw this classroom last. Figure 13 shows the routes. For those subjects who studied the map beforehand, the classrooms were not differentiated on the map. When the subjects were asked if they remembered seeing such a room, all 52 experimental subjects answered affirmatively. Since the control subjects did not explore the environment, they did not before hand the location of the target classroom.
Figure 17 Wayfinding 1 Target Room
The wayfinding task was scored using the paths shown in Figure 13. A maximum of four points was possible. The first point was given if the subject turned left immediately. The second point was given if the subject then took the next left without detouring. Two points were given if the subject then entered the door on the left (target room) without detour. One point was given if the subject detoured and went into another classroom but eventually came back to the target door from the hallway. One point was also awarded if the subject entered the target room from the adjoining small room. The routes were monitored and timed while in progress. When the subject recognized that they were in the correct room the time was noted. The subject was then asked to return to the entrance door. Up to four points were also awarded for the return trip; two for exiting the through the hallway door, one for turning right and one for stopping at the entrance/exit door. Time and route taken were also monitored.
All subjects participated in both wayfinding tasks. In the first wayfinding task 40 of the 52 subjects turned in the proper direction (left). Only 19 subjects found the target room within the 5 minute time limit. Of those 19 subjects 16 returned to the entrance door within the 5 additional minutes. Only one subject performed flawlessly.
The mean performance for the first wayfinding task are shown in Table 10. An ANOVA showed that those subjects who had a map before they entered the environment for the exploration period performed better, F1,39 = 8.43, p < .05; M(map before group) 2.0 and the no map before group M=1.3. Thirteen of the nineteen subjects who found the room had seen the map before, the other six had not. A marginal trend was found for those in the map during condition; F1,39 = 3.71, p < .07; Mean for map during M=1.3, Mean for no map during M=2.0. This suggests that the map during condition may have interfered with wayfinding. Twelve of the thirteen subjects in the map before condition returned to the entrance/exit door, and four of the six in the no map before condition returned. The scores of subjects who had a map before were significantly higher on the return trip than those who did not; F1,39 =4.26, p < .05. The mean for map before was M=1.5 and the mean for no map before was M=.84. Four of the 6 subjects in the no map before condition who found the room, did not have a map during exploration. The same 4 subjects found their way back to the starting point. The control group performed well, as three out of the five subjects found the target room, even though it was not specifically designated on the map. Two of the three control subjects also successfully returned to the entrance/exit door. All of the subjects took the initial left turn. Again the control subjects performed as well as or better than each of the other cells in the experimental design. There were no differences in the contrasts between the controls and those in the virtual environment experimental conditions.
The average time to find the target room was 164 seconds and the average time it took the subject to return to the entrance/exit door was 47 seconds. This time difference is significant t16 = 7.47, p<.001(paired t test). Clearly those who found the target room knew how to get back to the starting point. There were no significant time differences between the exploration conditions and the control group.
Table 10 Means for finding the target room in Wayfinding Task 1 ( four points possible)
Grand Mean = 1.67 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 2 2.4 1.2 .6 .75 1.2 no Map During 2 2.5 2 1.8 2.5 1.2 1.5
The ANOVA completed for the points scores in the first wayfinding task showed that the covariate was significant for both finding the target room, F1,39=4.13, p < .05 and the return phase also; F1,39=4.20, p < .05. The effects size for (2) for finding the target room was .096 and for returning was .097. These results are in line with Infield's research showing that the Guilford-Zimmerman can be used as a predictor of navigation performance on land and now can be used in Virtual Environments.
The second wayfinding task presented special problems to the subject because the two main routes to the target room were blocked with fire walls (Figure 18). The second target room was a laboratory that had one table (Figure 19). The two laboratories in the building were specified on the map as having one table or two. Subjects that were in the experimental conditions were reminded that there was a large double helix (DNA molecule) on top of the table. All experimental subjects and control subjects said they were familiar with the room or at least the specifications of the room.
Figure 18 A Fire wall in Wayfinding 2
Figure 19 Wayfinding 2 Target Room
The second wayfinding task was scored with a maximum score of nine points (Figure 14). The first point was given for starting in the correct direction (right). The second point was given for moving towards the first hallway to the right. This hallway was the shortest route to the laboratory but was blocked by fire. The fire could not be seen till the hallway was entered. The third point was scored if the next move was to go along the area near the bathrooms towards the other laboratory. If the subject then turned right and went down to the next hallway another point was earned. Many subjects went into the laboratory with the two tables, and points were not given for this maneuver. The fourth point was awarded if the subject turned right to go down the second hallway. Many subjects were observed being apprehensive about this move, as they could see another fire wall down the hall, which would block their way, if they were to try to continue all the way down the hall (figure 18). The fifth point was given if they continued walking down the hall towards the fire wall. When they neared the fire wall a decision had to be made to either turn right, left or turn around and go back. Two points were given if the subject went left into the executive office. An additional point was given when the subject exited either the secretary's or the far executive office door. If the subject then turned left and entered the lab door straight a head of them and into the target room the additional two points were given.
A half point was given if they went right down the bisecting hallway instead of turning left into the first executive office. If the subject then turned left and walked towards the lab with detouring into the offices on the right side of hallway another half point was given. If the subject then turned right and then left into the lab door they were given one point. The total points that could be awarded for taking this tour were 6 points. One subject in the experimental condition completed the nine point tour flawlessly, as did one subject in the control condition. Four experimental subjects and one control subject completed the return trip using the shortest route.
Five subjects initially turned left at the entrance door instead of right. Two of the five had turned incorrectly on the first wayfinding task. The results that more subjects turned in the correct direction for the second task are in line with having previous experience in the building during the first wayfinding task. During the first wayfinding task subjects thoroughly explored the left wing of the building and possibly remembered that the lab was not in that wing. Twenty nine of the 57 subjects (including control subjects) entered the laboratory where there were the two tables. Many of the subjects commented that they knew the task would not be that easy but they checked it out anyway. Twenty-two of the 52 experimental subjects found the target room within the five minute time limit. (See mean scores for Wayfinding 2 Task in Table 11) Of the 22 successful subjects, 15 found there way back to the entrance/exit door. There were no significant differences across the experimental conditions. The contrast between the control and those who did not have access to any maps was significant, t6.3 = 2.68, p <.05. The average time it took a subject to find the target room if successful was 200 seconds. The average time it took the subject to return was 129 seconds. Using the paired t-test the time difference was shown to be significant: t14=8.13, p<.001. There were no significant time differences between the exploration groups and the control group.
Table 11 Means for finding the target room in Wayfinding Task 2 ( nine points possible)
Grand Mean = 3.88 Self Exploration Active Guided Passive Guided Control (Map/noMap Before) MB nMB MB nMB MB nMB Map During 4.25 3 3.8 4.2 4.25 3 no Map During 4.25 2.75 3 4.8 6.2 2.5 3.75
Many of the subjects made insightful comments that highlight the four steps involved in wayfinding. The first step in wayfinding is orientation: Knowing where one is in respect to objects nearby and one's relative position in the world. Comments made by the subjects included the following: "I miss the lines for orienting myself. I don't think I was careful enough to pick out a landmark before I turned around", "Where's my map? You took away the lines too!", "It's hard without the lines."
The second step in wayfinding was deciding which route would lead to their destination. The task specified that the shortest route be used. Comments related to this step were most prevalent in the second wayfinding task when subjects attempted to turn right down the first hallway and were blocked by fire. Specific comments were, "Oh sure! block the easiest route," and "You took away the direct route."
The third step of route monitoring generated the most comments, both positive and negative. Positive comments included "Well that's a classroom, so I must be in the classroom neighborhood," and in the second wayfinding task as one subject turned left into the executive room and then quickly moved through the three offices stated "Oh here we go to the lab." The negative comments were made when some subjects ventured over to the right wing during the first wayfinding task and recognized that they were not in the correct area for the classrooms: "What am I doing over here", "I am going the wrong way", "What is going on? I'm in the total wrong part of the building?" Many just stated "I'm in the wrong wing." The most insightful comments for the second wayfinding task were "Where did all these rooms come from?", and "Oh, the lab has two doors." There were no specific comments about reaching the destination, but subjects groaned when asked to return to the entrance/exit door.
The comments and behaviors of these subjects clearly show that Regian's (1991) random walk is not an appropriate control. Only five subjects in the first wayfinding task reentered a room in which they had previously been. Six subjects entered the same room twice, unintentionally, in the second wayfinding task. Four of these six subjects kept going back into the lab with the two tables. Many subjects traversed hallways more than necessary while trying to find a target room, but few subjects traversed the same hallway more than twice than needed. Subjects, even when lost, would not backtrack without purpose.
Each of the three exploration conditions elicited unique subject behaviors. Self-exploration was the most interesting to watch. Some subjects were curious and wanted to check every room and every object. They physically squatted to look under tables and desks. Other subjects would wander the hallways, look at the pictures, go to the open-faced rooms, and then finally start entering the rooms through the doorways. A few subjects who studied the map before entering the building seemed to have a destination in mind and what they wanted to see first. Subjects could be grouped into those who were curious versus those who were hesitant or unsure of where to go.
Subjects in the active guided group seemed to have the most difficulty. They were told that they should stay near the path but that they did not have to stay directly on it. Many subjects looked straight down instead of watching what was around them so as to stay on the path. Similarly, when they came to an area where the line was straight ahead for some distance, subjects would use the turbo button to zoom down the line without noticing what was around them.
The passive guided group performed very well, considering there were hardware and software difficulties that required them to stay near the head-tracking receiver when going through doorways. These subjects often leaned backwards to see what was in some of the rooms that they did not enter. A few subjects verbally reported that they wished the tour would stop so they could better view things around them.
It was difficult for the experimenter to discern if subjects were attending to the map unless they specifically reported something about the map. Most subjects commented that they would prefer an arrow (rather than just a crosshair) to indicate both position and orientation.