![[Previous Chapter]](/icons/button-back.gif){kind=icon}
![[Table of Contents]](/icons/button-contents.gif){kind=icon}
![[Next Chapter]](/icons/button-next.gif){kind=icon}
by Mark Draper
4.1 Overview
Given that squatting, kneeling, and sudden severe movements by subjects may have contributed to the failure of the VB variable to achieve significance, a slightly revised version of Study 1 was conducted (Figure 22). Essentially, the only difference in this revised study was the inclusion of a restriction preventing any squatting/ knee bending along with instructions to only make slow, controlled movements. Subjects were still allowed to bend at the waist as necessary.
Figure 22
Given the low power of this study due to the small sample size, it was hypothesized that the result would be more likely a trend towards significance for the VB-context interaction. Specifically, the FVB would aid replacement accuracy in low-context and not in high-context conditions, particularly in regards to elevation. The context factor was again hypothesized to have an overall effect. It would also be interesting to see if the context-size interaction would again achieve significance under these new conditions and smaller subject size. No hypothesis was generated on the size factor.
4.2 Methodology
4.2.1 Subjects
Four male subjects from the University of Washington volunteered to participate, ranging in age between 25 and 37 (average age = 29). Subject height ranged from 5'7" to 5'11". Two subjects had previously experienced VR several times (but had not participated in Study 1) and two subjects had no previous experience with VR. All subject restrictions from the original study also applied to this effort.
4.2.2 Apparatus and Stimuli
The same as the original study.
4.2.3 Experimental Design
The same as the original study.
4.2.4 Task and Procedure
The same as the original study, except that subjects were not allowed to squat, bend at the knee, or perform any sudden, severe movements.
4.3 Results
As in the previous study, two dependent variables were analyzed, overall offset replacement error and abs. elevation offset replacement error. Overall replacement error provided an integrated measure of replacement accuracy while abs. elevation offset error provided for a more specific testing of the hypothesis that the primary benefit of a FVB would be in elevation accuracy. Finally, a section is included to summarize the results of the post-test questionnaire.
It should be noted that one subject's data set were eliminated entirely and replaced with a randomly selected additional subject, because of two extreme responses on one trial. The subject in question interchanged the locations of the two larger cubes during replacement causing extreme offset errors to occur in both cases (Table 5). Given that the required solution to keep this subject's data (substitution of these two cases with the grand mean) would have essentially made a small sample size smaller, the subject's entire data set was eliminated and replaced by a fifth subject's data.
Table 5
VB Context Cube Value Grand Mean Standard Level Level Size (inches) (inches) Deviation Full High 9" 128.8 8.6 22.8 Full High 15" 127.7 8.6 22.8
4.3.1 Overall Offset Replacement Error Data
The non-normal distribution data (as shown in Figure 23) required a transformation prior to analysis. After removing the one outlier shown in Table 6, the data satisfied the normality assumption by utilizing a square-root transform.
Figure 23
Table 6
Subject VB Context Cube Value Grand Mean Deviation from Mean
# Level Level Size (inches (inches) (# of Stand. Dev.)
)
2 FVB Low 3" 25.6 4.6 > 5.5
The mean overall offset error for each condition is shown in Table 7. A three-way, repeated-measures factorial anova revealed no significant differences for any variable or interaction. However, the context factor revealed a trend towards significance (F(1,3) = 7.49, p < 0.072), with subjects replacing cubes more accurately in the high-context condition. VB level failed to show an effect.
Table 7
Context Cube Size NVB FVB
3" 3.8 4.4
Low 9" 3.8 5.5
15" 5.1 5.5
3" 3.0 3.6
High 9" 3.0 3.0
15" 2.7 4.2
4.3.2 Abs. Elevation Offset Replacement Data
A normal distribution of this data was achieved simply by removing (and replacing with the grand mean) the one outlier shown in Table 8.
Table 8
Subject VB Context Cube Value Grand Mean Deviation from
# Level Level Size (inches) (inches) Mean (# of Stand.
Dev.)
1 FVB Low 9" 10.5 2.65 > 3.5
The mean elevation offset error for each condition is shown in Table 9. A three-way, repeated-measures factorial anova revealed no significant main effects or interactions. Interestingly, the trend towards significance of the context factor found earlier failed to occur with these data. VB level again failed to show an effect.
Table 9
Context Cube Size NVB FVB
3" 3.0 2.8
Low 9" 1.0 2.9
15" 4.4 2.8
3" 2.0 2.6
High 9" 2.0 2.7
15" 0.8 2.8
4.3.3 Post-Test Questionnaire
The same post-test questionnaire used in Study 1 was administered after this study to gain input on subject thoughts, strategies, and preferences. Subjects felt fairly confident of their overall performance in replacing cubes (mean rating = 4.8/7.0). They felt slightly more aware of their VB configuration per trial (mean rating = 5.5/7.0). Two of the four subjects felt that the existence of the FVB aided their performance. Reasons given include "slight help in elevation accuracy" and "the arrow was more difficult to judge depth with". One of the two subjects that felt that the FVB did not aid performance implied that it had in fact been utilized when he wrote that the body was "used only for reference points". The other subject saw no benefit to having a body in this task. As for preference, two subjects preferred the FVB, one preferred the NVB, and one had no preference. The justification of those preferring the FVB was that it was slightly easier to gage distances with the FVB than with the arrow. The subject who preferred the arrow felt it less distracting and believed its color (white) made it easier to identify when it intersected a cube. The justification for no preference was that both configurations provided equal points of reference.
All four subjects felt that the high-context condition aided their performance, the main reason being that it provided more points of reference in all dimensions (especially height). It also allowed subjects to get better oriented in the room. Two subjects felt that small cubes were replaced most accurately, one believed the medium cube, and one subject felt that all cubes were replaced equally well. As for replacement strategies, every subject used the tile floor as their primary memorization strategy. Context was used when available.
4.4 Discussion
The purpose of this study was to primarily look at the effect of revised instructions on replacement accuracy and to continue the development of a VB knowledge base. Given that its small sample size resulted in low power, it is easily understood why no factor achieved significance. There was hope for at least a trend towards significance of the VB-context interaction, however, which failed to occur. Nonetheless, interesting questions and ideas continued to emerge from this study as well.
First to the one trend towards significance of the context factor. This trend actually appears equal to (or slightly stronger then) the same trend found in Study 1, which is surprising given the sample-size differential. It appears that the restriction on subject movements slightly increased the need to use available context for reference. This is understandable when considering that the primary reason for squatting and kneeling in the low-context condition (Study 1) was to gage target height by using the windows as reference points. This strategy was eliminated in this study. However, it is perplexing that this trend towards significance of the context factor is lost when only elevation offsets are included. It is hypothesized that this was due to the small sample-size and the new subject strategy of `measuring' the amount of arm/arrow insertion into the top of a cube as a major clue to elevation.
It was interesting that the context-size interaction failed to achieve even a trend towards significance given that it was found to be significant in the first study. The previous explanation given that this effect was related to the subject strategy of using squats/kneels to reference cube height to the windows has been reinforced by the results of this study. By eliminating the squatting, the context-size interaction disappeared.
Why was there no trend towards significance of any VB related factor? Subject movement can be effectively eliminated as a reason as a result of this study. Subjects always had a FVB image in which to view on the trials when it was present and subjects did not perform movements that would have allowed for body-sensor displacements to occur. The other previously identified possibilities remain valid, however. The potential `ceiling effect' (i.e., the task being too easy) remains a strong possibility, as the results of this study are similar to those found in Study 1. Subjects again performed very well replacing cubes in all conditions. The floor was used by all as a grid for length/width location and timing appeared to be more then adequate in all instances. FOV remains a potential factor also. Although subjects were very aware of their VB on all trials (rating = 5.5/7.0), this awareness appears to be more one of arm/arrow vs. FVB/NVB. A larger FOV would definitely be a major aid in the testing of potential FVB effects.
Perhaps a reason for the failure of the VB factor to show an effect may be due to the nature of the task (as opposed to the ease of the task discussed earlier). Given that alternative measures of spatial awareness exist in ecological psychology and given assertions by ecological psychologists (Gibson, 1979) that one's body image is an important information source in the ambient optical array, perhaps a more ecologically-based task is better suited to test for VBs effects. The following study tested the above possibility.