An Exploration of Techniques to Improve Relative Distance
Judgments within an Exocentric Display
From the first experiment it is clear that image rotation improved relative distance judgments made within an exocentric view. However, it is also clear that rotation increases the length of time taken to make such a judgment. Since the use of rotation results in greater accuracy, one might argue that such an expenditure of time is worth the increase in accuracy. However, there may be situations in which distance judgments are time-critical. There may also be situations in which the computer platform being used does not have the computational power available to adequately support run-time rotation of a complex scene.
Thus, it is worth exploring different viewing methods to determine if the information conveyed by manual rotation can be presented to the user in a more rapid and less computationally-intensive manner. For example, it may be that the only benefit of rotation is that it provides a number of discrete views along a range of axes, as suggested by Sollenberger and Milgram (1993). If this is rotation's primary benefit, then it may not matter how that information is provided; i.e., people will be able to make accurate judgments as long as they have access to the different views. There may be a way of providing such information more rapidly than using the continuous manual rotation technique examined in the first experiment.
On the other hand, in addition to viewing along different axes, people may be consciously using selected rotation strategies as described in the first experiment. This position suggests that people truly need to control the manipulation of the 3D environment, through their own rotation strategies, in order to make accurate distance judgments. If this is true, we would expect lower accuracy in relative distance judgments when computer-controlled techniques are used.
If we compare different rotation techniques, where some allow the subject to completely control the use of rotation and others were more controlled by the computer, we could test the following hypotheses: When making distance judgments within a 3D spatial display, do people need only missing axis information regardless of control over rotation? We assume, of course, that the computer-controlled techniques offer views along all major axes. Additionally, if this hypothesis is true, there may be a way of providing such axis information more rapidly than the time it takes to manually rotate the image of the 3D space. If, on the other hand, people truly need to implement their own rotation strategy when making relative distance judgments, then we would expect higher accuracy for the subject-controlled rotation techniques as opposed to the computer-controlled techniques.
Four rotation techniques were used in this experiment. The first rotation technique was "Manual Rotation". This technique was identical to the rotation technique used in the first experiment. This first technique was considered to be a subject-controlled method, as subjects used the number key pad to rotate the world to obtain any desired view (see Figure 3.2).
The second technique was known as "Discrete Views". In this technique, subjects saw the world from one of four pre-determined views by pressing a single keyboard key (the "j" key was used). Subjects were allowed to freely cycle through 1) the initial view of the world, 2) a top down view, 3) a head-on view and 4) a view which was rotated -90 degrees from the initial view (Figures 4.1 - 4.4).
Figure 4.1: Discrete Views - View 1 Figure 4.2: Discrete Views - View 2
Figure 4.3: Discrete Views - View 3 Figure 4.4: Discrete Views - View 4
From these four discrete views, a subject obtained all axis information of the 3D space. The Discrete Views method was considered a computer-controlled technique.
The third technique combined the Manual Rotation technique and the Discrete View technique. It was considered a subject-controlled method since it included manual rotation. The purpose of this technique was to provide the ease of getting to specific discrete views with the flexibility of altering those views if desired. It was known as the known as the "Discrete Views + Manual Rotation" technique.
The fourth technique used was called the "Animated Views" technique. This was a computer-controlled technique which smoothly rotated the world to each of four the predetermined positions. As with Discrete Views, subjects cycled through the four animated views as many times as they liked. For each animation, the world changed along two axes before reaching the predetermined position. This feature allowed the user to see at least two axes change within one rotation. Based on the written comments in the first experiment, the ability to observe the relative movements of objects within the world was perceived to be an important cue for making relative distance judgments for some users. Thus, the overall purpose of this technique was to allow the user to see the relative movement of objects within the world as the world moved from one view to another.
As in the first experiment, subjects were again shown four differently colored cubes (yellow, blue, green and lavender) hovering over a terrain, and were asked to indicate, via the keyboard, "which colored cube is closest to the white cube" within the scene. Subjects were asked to make this judgment as quickly and accurately as possible. The apparati, in terms of processor and display, were the same as in the first experiment. There was, however, no use of stereo glasses or head tracking devices. Division Inc.'s dVISE software was again the base software used for creating the stimuli. Several additional software routines, written in C, were added to the routines for the first experiment in order to provide the different rotation techniques. The same 18 stimulus configurations used per condition in the first experiment were again used for the 18 judgments for each condition in experiment two. The geometric field of view and initial eyepoint elevation angle were also the same as in the first experiment.
4.3 Experimental Design and Procedure
This second experiment was 4 X 1 repeated measures design. The four treatments were:
1) Manual Rotation
2) Discrete Views
3) Discrete Views + Manual Rotation
4) Animated Views
As described above, rotation methods one and three were considered subject-controlled techniques, while methods two and four were considered computer-controlled rotation techniques.
12 naive subjects were run, 8 males and 4 females, ranging in age from 19 to 49. All subjects were different from those used in experiment one so that no subject would have had experience with any particular rotation technique. The order of the 4 treatments was counterbalanced across subjects, using a digram-balanced Latin Square design (Wagenaar, 1969) to control for any asymmetrical transfer of practice from one condition to another. The task was to determine which colored cube was closest to the white cube. Each treatment used the same 18 object position configurations. The order of the 18 trials was randomized within each treatment so that subjects would not memorize response orders. Also, the colors assigned to the stimulus cubes were rotated in order to again discourage recall of previously seen configurations and to guard against responses biased by chromostereopsis.
With regard to the experimental procedure, subjects were given a written overview (Appendix C) of the experiment. After reading the overview, subjects signed a Subject Consent Form and then had their near-point visual acuity measured using a Keystone Ophthalmic Telebinocular device. All subjects had 20/30 or higher visual acuity. Subjects were seated at the computer display 57 cm away from the monitor. The four different rotation techniques were explained to the subjects, and they were allowed to practice making relative distance judgments using each of the different techniques. The practice session judgments ranged in difficulty across all rotation techniques in order to encourage the use and exploration of each rotation technique. During the practice trials, subjects were given feedback on their judgments.
At the beginning of each treatment, subjects were told which rotation technique would be available to them. At the end of each treatment, subjects were asked to rate their confidence in the accuracy of their responses for the latest condition on a scale of 1 to 10, where 1 indicated no confidence in their judgments and 10 indicated extreme confidence in their judgments. After completing all treatments, subjects were asked to fill out a written post-experiment questionnaire (Appendix D).
For each treatment, the following data were calculated for each subject:
1) the subject's "Accuracy" as the percentage of correct judgments made within the treatment.
2) the subject's "Time" as the average time (in seconds) that it took the subject to make each judgment within the treatment condition.
3) "Confidence" as the rating on a scale of 1 to 10 that the subjects gave at the end of each treatment indicating how confident they were in their answers.
4.4.1 Dependent Variables of Accuracy, Time and Confidence
Table 4.1 presents the means and standard deviations for Accuracy associated with the different rotation techniques in this experiment.
Table 4.1: Means and Standard Deviations for Accuracy in Experiment 2
Mean Std. Dev.
Manual Rotation 75.3 6.9
Discrete Views 70.9 6.4
Discrete + Rotation 73.6 7.5
Animated Views 68.8 5.8
An ANOVA was performed across rotation techniques and showed a tendency toward significance among the means (F(3,33) = 2.63, p < .06). The accuracy means for the two subject-controlled techniques of Manual Rotation and Discrete Views + Manual Rotation were slightly greater than the computer-controlled techniques of Discrete and Animated Views (Figure 4.5).
Figure 4.5: Accuracy as a Function of Rotation Technique
A T-test comparing subject-controlled rotation techniques with computer-controlled rotation techniques revealed a significant difference on subjects' accuracy (T = 2.38, df = 46, p < .02). Subjects performed better using the subject-controlled techniques with a mean accuracy of 74.5, versus a mean accuracy for the computer-controlled techniques of 69.8.
Table 4.2 presents the means and standard deviations for Time associated with the different rotation techniques in this experiment.
Table 4.2: Means and Standard Deviations for Time (seconds) in Experiment 2
Mean Std. Dev.
Manual Rotation 34.3 14.3
Discrete Views 24.2 4.8
Discrete + Rotation 33.3 12.5
Animated Views 31.9 6.9
An ANOVA performed across the different rotation techniques revealed a significant difference among the means (F(3,33) = 4.75, p < .007) for Time. It took significantly less time to make a relative distance judgment using the rotation technique of Discrete Views than with the other techniques (Figure 4.6). A planned comparison showed a significant difference between Discrete Views and the other three rotation techniques (F=2.59, df=44, p < .013).
Figure 4.6: Time (in seconds) as a Function of Rotation Technique
A correlation analysis was performed across the independent and dependent variables of the experiment. There were no significant correlations noted among the variables, including subject Confidence ratings. A t-test on Gender and Time showed a significant difference between the length of time that males and females took to make their judgments (t = -2.95, df = 46, p < .008). The mean Time taken by males was 27.3 seconds per judgment compared to 38.1 seconds for females. A t-test on Gender and Accuracy also showed a significant difference between male and female subjects' scores (t = -2.07, df = 46, p < .04). The mean Accuracy for males was 73.5 and for females was 69.3. These results suggest that males, on average, took less time to make their judgments and were also more accurate.
4.4.2 Post-Study Questionnaire
On the post-study questionnaire, subjects were asked to describe the usability of the Manual Rotation method on a scale from 1 to 7, with 1 signifying "very easy to use" and 7 indicating that rotation was "very difficult to use". Subjects responded with a mean score of 2.8 indicating that rotation was a little easier to use than for those subjects in the first experiment.
Subjects were also asked to evaluate how useful they thought the other three rotation techniques were in helping them make relative distance judgments during the experiment. On a scale of 1 to 7, where one represented "not useful at all" and 7 represented "extremely useful", subjects rated the Discrete Views + Manual Rotation method as the technique that was most useful with a mean score of 4.6, followed by the Animated View method, with a mean score of 4.3. The mean score for Discrete Views alone was 3.6. Over 60% of the subjects believed they had developed a particular strategy for using all of the techniques except for the Animated Views method. When asked if they felt they had developed a technique for the Animated Views method, no subjects responded with a "Yes".
As a final question, subjects were asked to rank each of the four rotation methods in terms of their preference for using it. They were told to assume that all the methods used allowed them to perform the task equally. They were asked to place a 1, 2, 3, and 4 next to each item, with a 1 signifying their most preferred method and 4 signifying their least preferred method. Table 4.3 shows the total weighted score for each of the rotation techniques in ascending order. More preferred rotation methods will have a lower total weighted score.
Table 4.3: Subject Preference Rankings of Rotation Techniques
Methods Control Weighted ScoreThe results show a strong preference for the subject-controlled rotation techniques with subjects indicating that they prefer these methods as a first or second rotation choice. The Discrete Views + Manual Rotation technique was strongly preferred by subjects as their first choice rotation method, and Manual Rotation was preferred for the second rank position. Discrete and Animated Views, the two computer-controlled techniques, were ranked as subjects' third and fourth preferred method choice.
Discrete Views + Rotation Subject 16
Manual Rotation Subject 24
Animated Views Computer 34
Discrete Views Computer 36
The quantitative results of this experiment support the hypothesis that there is some performance benefit in allowing subjects to control how they manipulate the world view for this relative distance judgment task. It is also very clear from the data that subjects preferred the two subject-controlled techniques over the two computer-controlled techniques as seen by the relative rankings of the rotation methods. Their highest preference, however, was for the one rotation technique which combined both subject and computer-control. These rankings are consistent with comments made by subjects in which many explicitly stated that they preferred the control that they had when manual rotation was available, yet they also saw benefit in the Discrete Views technique. Other subjects mentioned being frustrated, at times, by the Discrete (when available without Manual Rotation) and Animated Views techniques, if there was a view they wanted to see but could not get access to.
The most negative comments occurred in describing the Animated Views technique. One common frustration noted was that the image would be rotating in a direction that subjects had no interest in seeing. Other subjects said that they were frustrated when the animation would stop. This suggests that these subjects were perhaps using some motion cues but then became frustrated when they could not control the use of those cues.
With regard to the Discrete Views technique, several people noted that the "top down" and "head-on" views were the most helpful for them. Several subjects specifically noted that they would "jump" to a particular view that they liked to either confirm a decision or to get a quick understanding of the world as a whole. When the Discrete Views technique was provided with Rotation, subjects noted that they would again "jump" to their favorite view and then rotate the world to refine their perspective on the 3D scene.
It is also worth noting that subjects took significantly less time to make a decision using the Discrete Views (without rotation) method. Subjects took approximately a third less time than the other methods and still maintained relatively similar accuracy (see figures 4.5 and 4.6). This finding supports the assertion that having access to specific, alternate views is a key component in making accurate relative distance judgments. As noted above, subjects mentioned that they liked the fast access to different views, which they thought were very helpful; however, most still preferred this method in conjunction with Manual Rotation.
Similar to experiment 1, the differences in performance again reported in this experiment across male and female subjects, is in accordance with reported higher spatial ability of males in a range of spatial tests (Masters and Sanders, 1993). Different from experiment 1, however, is that in this experiment the males took significantly less time than females when making judgment decisions. Male's ability to make more rapid spatial judgments than females, especially when the task involves mental rotation, has been reported (Gallagher and Johnson, 1992). It may be that males made more rapid judgments than females because the techniques in this experiment involved more rotation (on average) than in experiment 1. Such conclusions need to be tempered, however, with the fact that in this experiment there was a highly uneven number of male and female subjects (eight and four respectively).
It is also worth noting that subjects maintained very similar performance both in accuracy and time taken to make a judgment across experiments one and two when Manual Rotation was available (Figure 4.7).
Figure 4.7: Manual Rotation Accuracy and Time Across Experiments 1 and 2.