![[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
3.1 Overview
This study employed a search-and-replace task to determine the effect of VB configuration on spatial knowledge development (Figure 2). It served to directly test the potential for a VB to be used as a point of reference for object sizing and positioning under two different levels of environmental context (low, high) and for three different-sized targets (3", 9", 15"). Two different VB configurations were examined in this effort: a full-body VB (FVB) and no VB (NVB).
Figure 2: Study 1 - Search-and-Replace Study
Search-and-replace tasks have been used in past research on spatial awareness. However, the task was modified for this study to take advantage of the immersive, 3D nature of VR. Subjects were required to replace objects as accurately as possible in three dimensions (height, width, and depth) rather then two dimensions (height and width) commonly associated with conventional computer displays. Also, subjects were required to actually move to the desired replacement location rather than replacing objects from a stationary viewpoint.
It was hypothesized that in the low-context condition, a FVB configuration would result in more accurate target replacements (for all target sizes) then the NVB configuration. This improvement would exist primarily as reduced replacement errors in cube elevation and would be due to the ability of the FVB to act as a scaling reference for different target sizes and as general reference point for target placement. However, this improvement would decline in the high-context condition when more environmental reference cues were made available.
Although not central to this thesis, the following hypothesis was provided for completeness. Targets would be replaced more accurately in high-context conditions than low-context conditions.
3.2 Methodology
3.2.1 Subjects
Eight subjects (7 males, 1 female) from the University of Washington volunteered to participate, ranging in age between 18 and 24 (average age = 20). A height restriction was imposed (due to limitations in varying the VB size) so that all participants had to be between 5'6" and 6'1" tall. Subject eyeheight and bodyheight were measured (height range from 5'6" to 6'1", eyeheight range from 5'1" to 5'9"). All subjects were right-handed. All had normal or corrected-to-normal visual acuity with contacts and were tested for acceptable color vision. One subject had experienced VR several times in the past, four subjects had experienced VR once before, and three subjects had no previous experience with VR.
3.2.2 Apparatus and Stimuli
A SGI ONYX/2 RE2 computer graphics system running Division's dVS software (version 2.0.4i) was used for this study. A Polhemus Fastrak system provided orientation and position tracking of the subject's head (via a HMD), right hand (via a 3D joystick), and body (via a separate Fastrak sensor). A Division dVISOR HMD provided the immersive stereo, full-color, visual display. This HMD had a horizontal FOV of 105 degrees, a vertical FOV of 41 degrees, and a 40 degree overlap. Each eye image consisted of 30,000 full-color, triad pixels (199 horizontal, 150 vertical). The inter-ocular distance was set at 2.55 inches (the default value). A Division 3D joystick was used to track hand movements and to provide for participant movement/interaction in the VE. By depressing the 3D joystick's top-right button, a subject could move forward in the VE, depressing the top-left button allowed the subject to reverse his/her direction. The subject always moved in the direction that he/she was looking (or opposite it if moving in reverse) and was limited in translation movement to the horizontal plane (i.e., the subject could not fly up or down in the VE). By depressing and holding the upper trigger button on the joystick, the subject could pick up and move the virtual targets. The VE was developed using Division's dVISE application software and custom user functions. The subject's body movement was sensed by a single Polhemus tracker attached to a specialized belt worn by the subject and centered on his/her right hip bone at the waist.
An experimenter's station consisted of a computer keyboard and a television monitor. The monitor presented a real-time image to the experimenter that matched the current visual image presented to the subject's right eye and was positioned in front of the experimenter.
The VE consisted of one 20 x 20 x 8 ft room (Figure 3). The room had a black and white tile floor (2'x 2' tiles), surrounded by a 2' wide pale peach border between the tiles and the walls. All four walls were white. One wall had a blue doorway in the middle. An adjacent wall contained two large blue windows, spaced symmetrically. The white ceiling included two parallel brown beams that ran the length of the room. The room contained both direct and ambient lighting which allowed for subtle virtual shading to exist.
Figure 3
The above room configuration was considered the low-context condition. The high-context condition (Figure 4) added the following: red/purple couch, `glass-top' coffee table with a white telephone, brown chair, brown stool, large rosewood table with white telephone, blue/white lamp on a marble pedestal, and dark-brown rectangular wastebasket. There were three high-context layouts (x, y and z) as shown in Figures 5, 6, and 7.
The three different high-context conditions were developed to provide similar contextual information cues while minimizing the potential for asymmetric transfer to occur between high-context and low-context trials. To illustrate this potential for asymmetric transfer, consider the following scenario. If a subject received two of the same high-context layout conditions in a row followed by a low-context condition, the subject could possibly remember and superimpose the high-context layout onto the low-context condition, aiding the memorization/replacement tasks in this condition. This benefit would not exist if the low context condition occurred before the high context conditions, however. Therefore, to avoid this situation from occurring, three different high-context conditions were developed that contained equal information but not equal layouts.
The essential structure of the three high-context layouts is similar and the same objects appear in each layout. The object placements were simply interchanged, so that the general areas of context were always filled with a particular contextual object. Context z was exclusively used for all practice sessions; context x and y were used in all experimental conditions.
Figure 4
Figure 5: Context 'x'
Figure 6: Context 'y'
Figure 7: Context 'z'
A total of six target sets were developed for this study. Each target set consisted of three different-sized, colored, and numbered cubes. The sizes (measured length of each side of the cube) were 3", 9", and 15". Colors (red, green, yellow) were randomly assigned to an object size for each set and were counterbalanced with size across sets. Therefore, each target set had three different colors associated with the three sized cubes. Two of the six target sets (sets #1, #2) were used exclusively for practice trials and the remaining four sets (#3, #4, #5, #6) were used on all experimental trials.
Each target set was developed manually using the following constraints on target placement as guiding principles. Targets could not be placed within 20" of the virtual floor nor within 10" of any virtual wall (measured from target center) to avoid the floor and walls being used as the overriding reference points in all conditions. Also, target centers could not be placed higher than 33" from the floor, to assure that they would only appear in the same areas where a VB image normally appears. No part of any target could touch any context object (i.e., couch, table). Each target had to appear in a separate quadrant of the room, defined by the horizontal plane with the origin being the room center. Each target was orientated in a level position. Lastly, targets could not exist within 4 feet of each other and the goal was to keep the targets dispersed. The six target sets used in this study are shown in Figures 8 through 13, where R, G, and Y stand for the cube's color (red, green, and yellow) and the number above each cube is it's elevation in inches above the floor (represented by the tiled pattern).
Figure 8: Target Set #1 Practice Set
Figure 9: Target Set #2 Practice Set
Figure 10: Target Set #3
Figure 11: Target Set #4
Figure 12: Target Set #5
Figure 13: Target Set #6
Along with counterbalancing color with size across target sets, the following additional constraints were placed on the population of six target sets. Each target size had to appear at least once in each of the four length-width quadrants of the room and no same-sized targets could be positioned within 2 feet of each other across sets. The goal of these constraints was to spread instances of each size throughout the room. Lastly, each same-color grouping of targets had to span at least 3 of the 4 length-width quadrants to minimize a possible color-location grouping. The target diagrams organized by size are shown in Figures 14 through 16.
3" Target and Color Spread
9" Target and Color Spread
15" Target and Color Spread
Two different VB configurations were used (FVB, NVB). The FVB was developed by utilizing a geometry file from an Alias modeling package. The FVB consisted of a torso, two arms, and two legs (Figure 17). All were the same color, a dull light bronze. The head, right hand, and waist were tracked by sensors so that physical movement of these body parts would be transmitted into corresponding movements of the VB in the VE. The right arm could freely bend/move at the shoulder but there was no elbow, finger, wrist, or grasping movements. The left arm was static with respect to the torso. The legs were static, moved in accordance with the torso, and appeared to slide across the ground during translation. There were three differently-sized FVB configurations (small, medium, large) to be fitted for subjects of different heights (5'6" to 5'8.4", 5'8.5" to 5'10.9", 5'11" to 6'1").
Figure 17
The NVB configuration involved no visible body image except a dVISE standard white `arrow', representing the virtual hand, for picking and replacing targets. The arrow moved in accordance with real hand movements. Although the two VB configurations never existed at the same time, Figure 18 shows a image (viewed from the virtual eyepoint) of both the NVB and the FVB arm for comparison.
Figure 18
3.2.3 Experimental Design
A 2 x 2 x 3 within-subjects factorial design was used. A within-subjects design was necessary due to the large individual differences found in previous research on spatial estimation tasks and the relative stability/consistency of these estimations per individual over time. In addition, the effects of a VB on spatial awareness may not be large, requiring the added sensitivity and power of a within-subjects design.
The three independent variables were VB configuration (FVB, NVB), environmental context level (low, high), and target size (3", 9", and 15"). For each VB and context level, subjects experienced all target sizes before moving to the next condition. The dependent variables included overall offset replacement errors (i.e., the total separation between the original cube's centroid and the replacement cube's centroid measured in inches), absolute elevation offset replacement errors (i.e., the centroid separation error in elevation only, also in inches), and time to replace each cube (in seconds).
A short post-study questionnaire provided a few subjective ratings for review (Appendix A). Numerical ratings on a 7-point scale were requested for overall performance (1 = very poor, 7 = excellent) and overall awareness of the specific VB configuration in each trial (1 = never aware, 7 = always aware). Other questions were either forced choice (i.e., which VB was preferred) or free form (i.e. what strategies were used).
Each VB/context condition was digram-counterbalanced across trial order to minimize potential general order effects. Digram-balanced designs not only equate experimental conditions with order but also require that each condition is preceded and followed by each other condition just once (Wagenaar,1969). Each subject received all three different high-context layouts to avoid potential asymmetric transfer from high-context trials to low-context trials (due to memorizing one high-context layout as explained earlier). One of these high-context layouts (context z) was used only during practice trials. The two other high-context layouts (context x, y) were counterbalanced across trial order and VB/context condition to further guard against practice effects. Lastly, two of the six target sets were used exclusively for practice trials (sets #1, #2) and the remaining four sets (#3, #4, #5, #6) were used on all experimental trials. The four experimental target sets were counterbalanced across each VB/context condition rather then trial order. This guaranteed that each target set appeared in each VB/context condition exactly twice, controlling for inter-cell variance due solely to target set differences. The two practice target sets were counterbalanced with practice order. Appendix B shows the experimental ordering for this study.
3.2.4 Task and Procedure
Each subject received written and oral instructions on the task to be accomplished (Appendix C). After reading the instructions and asking any questions about the task, the subject was tested for normal (20/20) visual acuity using the Snellen near-point acuity test on a Keystone Ophthalmic Telebinocular. Eyeheight and body-height were then measured so that the VB height could be best matched to the subject. Next, the subject was briefed on the VR equipment and was fitted with a belt onto which the body-tracking sensor would be attached. The subject was then told to stand on top of a yellow marker on the floor and face a specific direction. This was explained to be the `starting' position that the subject would need to return to at the beginning of each memorization task and each replacement task. The subject was then fitted with the HMD and was told to look down at his/her VB as the body sensor was attached and calibrated to provide an acceptable VB image.
The subject was then handed the 3D joystick and given instructions on how to move in VR along with how to pick and place objects. To move around the virtual room, the subject would press one of two buttons on the joystick while looking in the direction he/she wanted to move. One button allowed the subject to move forward in the VE (in the direction the subject was looking) while the other button allowed the subject to move in reverse. The subject only had to push those two buttons while looking in different directions to completely move in the VE. The subject could also move in the VE by physically moving his/her body in the real world, as his/her body was tracked by three positioning sensors. However, since the subject could only move in a small, constrained area (4 feet) without moving out of the tracking sensor range, most VE movements were accomplished by pressing the two joystick buttons. To pick up a virtual object, the subject pressed the trigger button on the joystick (as described in the Apparatus section).
The subject then spent time maneuvering around the virtual room, picking and placing objects, until satisfied that an acceptable level of proficiency had been reached. It was during this initial phase that the subjects were required to verbally report the colors of each of the targets as a method of verifying acceptable color vision. Following this, each subject participated in two practice trials. Each practice trial utilized a different VB configuration and a different context level. At the end of the first practice trial, the subject had the opportunity to complete an entire second practice trial or spend an abbreviated amount of time getting used to the different VB configuration and room context level. This option was provided to prevent monotony in those who felt fully proficient at the required task along with minimizing overall subject fatigue. No feedback on performance was given because the purpose of practice was strictly to educate the subject on task procedures/protocol, not to train the subject to a certain level of accuracy.
At the conclusion of the practice trials, the first experimental trial began. The subject began the trial standing in the middle of the virtual room (while at the same time standing in the `starting' position in the real, physical room) with a particular VB configuration and a particular context level. One set of three targets was also presented. The subject's task was to move around the virtual room for 2.5 minutes, memorizing the spatial location of the three targets. The subject was warned when only 30 seconds remained. The subject had the ability to shorten this time if he/she so chose (this occurred about 10 percent of the time). Once the time limit was up, the subject was virtually transported instantaneously back to the center of the virtual room and the targets disappeared. The subject was also instructed to physically return his/her real body to the `starting' position at this time (the subject would look down though the bottom of the HMD to align his/her real body in the correct physical starting position).
The subject's task was to replace all three targets, one at a time. A randomly chosen target appeared directly in front of the subject. The subject was instructed to `pick' this target (using his/her virtual hand or arrow) and replace it as accurately as possible. Accuracy was emphasized over speed, however, a 60-second time limit on replacement existed. The subject moved to that target's original location and replaced the target as accurately as possible, not worrying about target orientation. The subject was free to pick and place the target as often as necessary within the 60-second time limit, until satisfied with its location. The subject was warned when only 15 seconds remained. Once the subject reported to the experimenter that he/she was comfortable with the target position (or after the 60-second time limit expired), the original and replacement positions were recorded, offset errors were calculated (absolute errors for the overall offset and each length, depth, and elevation component offset), and the replacement time was recorded. No feedback was provided to the subject. The subject was then transported back to the center of the virtual room (after he/she physically returned to the `starting' position, as before), the replaced target disappeared, and a new target was randomly selected (from the remaining targets) for replacement. After replacing all three targets, the subject received a new VB configuration, a new context condition, and a new target set. The subject had the opportunity to take a short break in between trials. This procedure was repeated for all four VB/context treatments.
3.3 Results
Two dependent variables were analyzed in this study, overall offset replacement error (i.e., radial distance) and absolute (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. In addition, a section details the results of the post-test questionnaire. Time was not analyzed because it could not be temporarily stopped when subjects took breaks during the experiment. Most subjects took the entire time allowed for replacement anyway so the resulting variance was extremely low.
3.3.1 Overall Offset Replacement Error Data
Given the extreme non-normal distribution of the data (as shown in Figure 19), a transformation of the data was required prior to analysis. After removing the one outlier shown in Table 1, the data satisfied the normality assumption by utilizing a log10 transform.
Figure 19
Table 1: `Outlier' Data Point - Overall Offset Error (Study 1)
Subject VB Context Cube Value Grand Mean Deviation from Mean # Level Level Size (inches) (inches) (# of Stand. Dev.) 8 NVB Low 15" 79.4 7.28 > 7 1/2
The mean overall offset error for each condition is shown in Table 2. A three-way, repeated-measures factorial analysis of variance (anova) revealed no significant differences for any variable or interaction. However, the context factor revealed a trend towards significance (F(1,7) = 4.29, p < 0.08), with subjects replacing cubes more accurately in the high-context condition. VB level failed to show an effect.
Table 2
Context Cube Size NVB FVB
3" 5.1 8.4
Low 9" 5.6 3.8
15" 5.0 7.0
3" 3.3 3.5
High 9" 2.8 4.7
15" 4.7 6.8
3.3.2 Abs. Elevation Offset Replacement Data
The extreme non-normal distribution of this data (as shown in Figure 20) also dictated that a transformation occur prior to analysis. After removing the one outlier shown in Table 3, the data satisfied the normality assumption by utilizing a square-root transform.
Figure 20: 
Table 3
Subject VB Context Cube Value Grand Mean Deviation from Mean # Level Level Size (inches) (inches) (# of Stand. Dev.) 1 FVB High 15" 17.5 2.88 > 5.0
The mean elevation offset error for each condition is shown in Table 4. A three-way repeated measures factorial anova revealed a significant context-size interaction (F(2, 14) = 4.62, p < 0.03). However, the Mauchly Test for Sphericity indicated a violation of the assumption of homogeneity (W = 0.554, p < 0.17) for this test. Therefore the Huynh-Feldt epsilon correction was used and the results were still significant (F(1, 11) = 4.62, p < 0.05). This interaction (Figure 21) indicates that only the replacement of small cubes was affected by context level. A post hoc test was conducted on the means of the two context levels within the 3" cube condition. This result was significant using the Scheffe correction (F(1,7) = 15.59, p < 0.01), verifying that replacement of small cubes was significantly aided by the presence of high-context. VB level again failed to show an effect.
Table 4
Context Cube Size NVB FVB
3" 3.8 4.6
Low 9" 2.7 1.5
15" 1.8 2.6
3" 1.0 1.3
High 9" 1.7 2.1
15" 1.5 2.9
Figure 21
3.3.3 Post-Test Questionnaire
A post-test questionnaire was administered after the study to gain input on subject thoughts, strategies, and preferences. Subjects felt confident of their overall performance in replacing cubes (mean rating = 5.4/7.0). They felt slightly less aware of their VB configuration per trial (mean rating = 4.5/7.0). Five of the eight subjects felt that the existence of the FVB aided their performance. Reasons given include "gave sense of position when looking down", "I could place the cubes at a more correct height by comparing them to parts on my body", "helped me know what height and angle I was looking from", and "gave me a better spatial orientation as to what was around me". The remaining three subjects felt that either the body could not be seen enough or that it was a distraction when it was seen (especially the arm). As for preference, three subjects preferred the FVB, one preferred the NVB, and three had no preference. The remaining subject preferred a combination of the FVB body with the NVB arrow for an arm. Justifications of those preferring the FVB included that it "aided in how far away I was from cubes", it provided for more accurate height estimations, and it "seemed more real, like I was in the room". The subject who preferred the arrow felt uncomfortable seeing the body move into and through objects. Justifications for no preference included the task was easy and that the body was not used. The subject preferring the VB combination liked the FVB legs but felt that the arm occasionally blocked his view.
One final note, two of the subjects who felt that the FVB aided their performance stated that they `discovered' the FVB as a reference potential near the end of their study trials.
Seven of the eight subjects felt that the high-context condition aided their performance, the main reason being that it provided more points of reference in all dimensions. Five subjects felt that all cubes were replaced equally well, two felt that the small cubes were replaced most accurately, and one subject felt that the large cubes were most accurately replaced. As for replacement strategies, every subject used the tile floor as their primary memorization strategy. Several subjects also squatted and/or kneeled to view cube height. Context was used when available.
3.4 Discussion
Although there were few significant findings in this study, many lessons were learned as a result of this first attempt at studying the spatial influences of VBs. The significant (and near significant findings) will be discussed first, followed by a discussion of why the focus of this effort, the VB, failed to show any effects.
The context-size interaction that occurred with the abs. elevation offset error is interesting (Figure 21). This same interaction is not significant when overall offset error is considered. In high-context conditions, the smallest cube (3") was replaced significantly more accurately in elevation then the same-sized cube in low-context conditions. In contrast, the medium and large cubes were replaced equally well in both context conditions. What is it about the small cube that made context so important to elevation positioning? The answer is thought to lie in the strategy used by most subjects of using context for assisting in height estimations. All subjects agreed that the elevation offset was hardest to memorize, given that the primary strategy (i.e., use of the tile floor) was useless in this regard. However, many subjects acknowledged in the post-test questionnaire that the availability of context aided in height estimations. Then why wasn't the same improvement shown for the medium and large cubes? A commonly used elevation memorization strategy in the low-context condition was for subjects to squat/bend and use the two virtual windows as reference points. It is hypothesized that larger cubes could be more easily referenced to specific window heights then the smallest cubes. An alternative hypothesis is that the smallest cube was easier to line up with high-context elements
The trend towards significance by the context factor is not surprising. What is surprising is that it failed to reach significance at the alpha = 0.05 level. Given the many studies that have shown context to be effective in spatial tasks, the primary reason for including it in this study was to investigate its influence on any VB effects that may arise. The failure for context to fully achieve significance can most likely be explained though the existence of the tiled floor. As all subjects used the tile grid as their primary memorization strategy, the floor became in essence the major source of context. This had the effect of increasing the amount of context in the low-context condition to a much higher level then planned, the result being a `narrowing of the gap' between these two levels just enough to keep it from reaching significance. One subject relied so heavily on the floor that he wrote, "The grid floor was used for positioning. The extra stuff was just that".
One more point needs to be stated regarding context and the nature of the raw overall offset error distribution. This initial distribution was extremely skewed in the positive direction, presenting multiple data `outliers' that generated the need for a data transformation to take place (Figure 19). It is of interest to note that eight of these ten identified outliers occurred in the low-context condition (including the six highest errors). Based upon the above, it would seem plausible to assert that spatial tasks may be susceptible to a more severe performance degradation in low-context environments then in high-context environments.
Now to the focus of this thesis, the VB. Why did the VB factor fail to show an effect? The easy answer would be to surmise that the VB does not influence spatial awareness. However, studies often generate more questions then they resolve and rarely does the first study of an exploration definitively answer the question at hand. Therefore, this discussion will proceed by identifying discovered deficiencies in this design and by highlighting ideas generated from this first effort in regards to more accurately testing the relationship between VBs and spatial awareness.
The first discovered deficiency of this study was the lack of instructions governing subject movement. Basically the instructions contained no restrictions on movement of any sort, except that the subject had to remain within the range of the Polhemus Fastrak transmitter. As a result, the majority of subjects used this freedom of movement to squat, kneel, and crouch down many times during the course of the experiment. The effect of these motions during trials with the FVB configuration was for the VB image to drop into the floor and out of the subject's view. These movements effectively eliminated the FVB as a factor in the design during the length of time the subjects were in these positions. As a result, any potential effect due to the VB may have been removed in these situations.
Another negative result of this allowance for free movement was the potential jarring or dislodging of the body sensor tracker through sudden severe movements by the subject. Some subjects were very tentative and moved in a controlled fashion at all times. Others would whip around, moving rapidly to obtain different virtual viewpoints. Some of these sudden movements succeeded in jarring the body tracker slightly and on one trial the sensor ripped right off its belt. Due to the nature of the FVB mechanization, slight displacements by the body tracker could result in large displacements by the FVB. The result was a VB image that could potentially become unstable or displaced by these movements. Therefore, any attempts to use the FVB as a point of reference would be susceptible to errors caused by unintended sensor displacements. It should be mentioned that this was not a pervasive problem with all subjects. Only two subjects made movements severe enough to question the consistency of the FVB location. However, it is an issue that needs to be dealt with in future efforts.
A third deficiency of this design may have been that the task was too easy for subjects, making it unnecessary for subjects to reference their VB. Overall separation scores reveal surprisingly accurate performance by the subjects and ratings of their overall performance were high (5.4/7.0) on the post-test questionnaire. This `ceiling' effect may be due in large part to a `floor'. As stated earlier, the tiled floor was used by all subjects as the primary strategy, as it provided a grid by which to easily segment the room. Although pilot studies revealed that the floor could be used as a reference, the magnitude of its potential use was discovered only after experimental trials were well underway. Also, the time to memorize (2.5 minutes) and time to replace (60 seconds) may have contributed to this effect. The implication of this was in effect summarized by a subject when he wrote "I found it easy to do without the body. I actually lost track of which one (VB configuration) I was using".
An interesting finding through the post-test questionnaire was that two subjects `discovered' that their full virtual body could be used as an elevation point of reference towards the end of their trials. Perhaps if they had discovered this fact earlier (i.e., through more extensive practice sessions) it would have altered their strategies. However, the cost associated with longer practices sessions would have been an increased potential for fatigue/discomfort to arise in subjects during later trials. Each subject was in the VE for 35-45 minutes (albeit with breaks as necessary) which is long enough to be concerned about subject fatigue occurring.
A potential factor that was identified in the literature review was the FOV limitations of current VR HMDs. The HMD used for this study (dVISOR) was chosen because it offered one of the largest FOVs on the market (105 degrees horizontal by 41 degrees vertical) . However, this FOV still does not come near the natural FOV of humans (approximately 200 degrees horizontal by 120 degrees vertical) and, therefore, brings to mind the image limitation problem Slater and Usoh (1993c) had in their work with VBs.
Given the identified deficiencies with this first study and the enticing comments by subjects regarding the usefulness of the FVB, it is evident that more studies need to be performed in this area before a definitive answer regarding VBs and spatial awareness is obtained. The study that follows was a revision to this effort in that it eliminated the identified deficiencies with the instruction set.