![[Previous Chapter]](/icons/button-back.gif){kind=icon}
![[Table of Contents]](/icons/button-contents.gif){kind=icon}
CHAPTER 8. CONCLUSIONS.
Perception of Distances - Conclusions.
A significant size task finding in this study is the extent to which distances are perceived to be smaller in the simulated environments. The horizontal and vertical dimensions of spaces in the simulation conditions were all perceived to be significantly smaller than in the real spaces. It would appear that this misperception is due to the well documented size-constancy phenomenon, whereby sizes and distances appear to be smaller when seen through a truncated field of view (Dolezal 1982, Alfano 1990).
The results also indicate that perceived distances are underestimated by a larger amount in the virtual environment than in both the monoscopic and stereoscopic walkthrough types of representations. These findings would tend to suggest that head-position tracking somehow affects the process by which people visual search the display in such a way as to diminish their perception of distances. Perhaps, when people are head-tracked, they tend to gaze more often at the edges of eyephones, in which case they would be more affected by the distortion in the eyephones than if they focused their gaze in the center of the optics.
Implications of the Distance Estimate Conclusions.
The underestimating of distances in virtual environments is large enough to raise concern about the uses of the tool in its present configuration for doing distance evaluations. Virtual environments, if used with a display configuration similar to the one in this study, could lead users to make considerable errors in judgments of volume size.
There are however several short term solutions to this problem. Since the underestimation of distance is consistently around 20%, users could use this information as a reminder to correct their initial perceptions of volume sizes. Another safeguard would be to include many familiar elements of scale in the model. Many participants indicated that the task would have been easier had there been more elements of scale (Appendix B: Data Figures).
A more durable solution for decreasing the effect of the size-constancy illusion would be to increase the display's field of view. At present unfortunately, it is very difficult and expensive to increase substantially the field of view in a eyephone display. It is a technology which to this day can only be afforded by the military.
Retro-projection is however another technique for presenting virtual environments which does not involve using the eyephones and yet which offers a large field of view. In this configuration, the participant stands in front of very large screens. The images are projected from behind the screen. As the number of screens increase, so does the field of view and the sense of immersion. One of the drawbacks of this configuration is that it is much more cumbersome than when using Head Mounted Displays.
An example of this display was demonstrated in the Virtual Building Walkthrough project at the University of North Carolina. In this set-up, participants looked at a very big screen projected by a Barcodata system (Brooks 1986). A more recent example is "The Cave" which was presented at the SIGGRAPH '92 exposition. It is a virtual environment in which the big screens completely surround the participant. In this scenario, the virtual model is displayed to the participant's entire visual field of view, just as it is in the eyephones.
At least some of the underestimating of distances must have been related to head-position tracking because estimates in this condition were slightly worse than in the non-tracked stereoscopic condition. Explanations for this effect could be the result of a relatively complex process and there is, unfortunately, no short term solution. It is clearly an area which will require further research.
Perception of Orientation - Conclusions
While none of the results for the orientation task are statistically significant, they do indicate that participants viewing the simulations were fairly less precise in their angle estimates than those in the real museum. If however, the relative simplicity of the task is taken into consideration, then even a small difference in the degree of variation between the Real and the simulated conditions would confirm the difficulty people have in building an accurate cognitive map of simulated environments.
There are several reasons to believe that this is the case. First of all, it would agree with the numerous observations at the H.I.T.Lab in which people were lost and confused about their orientation when in virtual environments. It would also coincide well with the many projects [1] at the lab which where instigated out of a concern for this problem.
There is evidence that the limited field of view in the displays could disrupt people's ability to build a good cognitive map and therefore reduce their accuracy in the pointing task. As previously quoted in Chapter 7, "the overlap of peripheral and fovea information is necessary for veridical perception to occur, ... and that restricting the field of view will interfere with both perception and visuomotor performance " (Alfano 1990). If that is the case, it could explain why it is so difficult to know where one is in a virtual environment. The exploration of the effect of varying the field of view on cognitive mapping tasks is yet another instance where further research is required.
One more reason to believe that people had difficulties building a cognitive map of the spaces is the fact that they had no kinesthetic feedback to concord with the visual illusion of movement. In this respect, these participants were missing what might be an essential interaction for "understanding" the space; that of pacing the space to get a physical sense for its size along with the visual spatial information. It would be interesting to compare the H.M.D configuration to the configuration at U.N.C in which people move their viewpoint by physically walking. It would then be possible to determine the extent to which the physical feedback of walking helps in building a cognitive map of an environment.
Unfortunately, the study was conducted in such a way that makes it difficult to establish whether the imprecision of the estimates in the simulation conditions are due to people's inability to build an accurate cognitive map of the spaces, or whether it was impossible to do the task accurately even with an accurate cognitive map. The limited field of view of the displays implied that people had very little spatial information with which to base their orientation, even when they were guided to the corners in the rooms.
Implications for the Perception of Orientation Conclusions.
In any event, regardless of the many valid concerns about orientation in virtual environments, the pointing task in the study did not reveal any conclusive differences either between the simulations and the real museum, nor among the simulations themselves. However, this does not imply that orientation is not a problem in virtual environments. Results from the informal sketch task vary in accuracy across the four display conditions, suggesting that there were differences to be measured in the pointing task.
One implications from these conclusions is the need to develop better methods for measuring people's sense of orientation in virtual environments. The model used for this study was too simple and the path of the visit was too informative. It was also ill-adapted for the display conditions in which participants had a limited field of view. Perhaps a completely different task is required, one which would require participants to actually find their way to designated target destinations.
Until further research can be conducted, architects wishing to use this tool could limit the problem of spatial orientation in virtual environments by providing the participant with additional layout information, such as a plan view of the space. This would enable users to correct and update their cognitive map of the spaces.
There have also been attempts to include additional information within the environments. One such project was for the design of a fighter cockpit virtual display. Along with a virtual representation of the landscape environment, the pilots were to be provided with a miniature view of their plane as seen from above. The miniature plan view is just one of many solutions which could help users to always know where they are.
Perception of the Qualities Spaces - Conclusion.
Overall, the results indicate that all the simulation conditions were successful at representing the general feel of the spaces. However, under closer inspection, results from the Tracked condition indicate that it was slightly more effective at representing the "feel" of the spaces than the other displays conditions. While not statistically significant, it does suggest the head-tracking device, combined with the stereo eyephones, enables participants to make accurate evaluations about the way a place "feels".
This is probably due to the fact that people feel as if they are really inside the model. Every motion of their head reenforces the illusion of presence in the virtual space. Feeling one is present in a space seems essential to being able to properly qualify it. This would explain why, on average, participants' qualification of the space in the Tracked condition was closer to the Real condition than were the other two simulation conditions.
Although virtual environments were expected to be more effective than walkthroughs for communicating the qualities of spaces, their actual difference in the study was not statistically significant. One possible explanation for the lack of greater distinction between the two conditions is that the test space was relatively simple. As a museum space, it was not designed to have a lot of character. The results were also potentially diluted during the process of selecting adjectives because I included only those which could appropriately define the space, thereby narrowing the variability in the responses.
Implications for the Perception of the Qualities of Spaces.
Results from the description task begin to demonstrate one of the specific advantages of using virtual interfaces to represent architectural spaces; that of being able to accurately assess how a place would feel, and conversely, how one would feel in a place. The tool, as it exists in this study, is at least as effective if not more so than the walkthroughs for conveying how a place would feel.
Overall Conclusion.
As measured with the three perceptual tasks, it becomes apparent that virtual environments do not fully satisfy the requirements for completely replacing those forms of architectural representations which are meant to convey the basic spatial characteristics of proposed spaces. The virtual interface used in this study is not quite good enough for making quantitative judgments of spaces. It is difficult to orient oneself in virtual spaces and distances are underestimated. However, the interface is adequate for making qualitative evaluations of architectural spaces. Using this interface, people's perception for the way the modeled spaces feel would rather accurately predict their perception of the feel of the real space.
In part , the difficulty of making accurate quantitative judgments is due to the narrowness of the field of view. These difficulties can be reduced if the horizontal field of view of the eyephones is increased or if a display technology using a wider field of view is used. The source of other problems are more complex; the effect head-position tracking has on the estimates is one example, and the lack of kinesthetic feedback for movement is another. Both of these areas would require further research.
While the type of virtual interfaces in this study are not adequate for quantitative tasks, they could be used to supplement other existing forms of architectural representations which are perhaps more effective at conveying the quantitative attributes of spaces but less effective at conveying how the spaces would feel. Typically, designers could continue to use drawings to convey the size and location of spaces while supplementing these with a virtual environment for conveying how the spaces would feel.
Another reason for supplementing existing techniques of spatial representation with virtual environments is that they are very easy to use. Interpreting spatial information using virtual interfaces is perhaps as simple and intuitive as it is to interpret real spaces. To interact with the virtual environment, participants used their body and head to look around as they would have in real spaces. This was certainly much easier than trying to interact with the walkthrough representation. Participants complained extensively about the movement metaphor in the walkthrough condition (see Appendix B: Data Figures - Participant Evaluation of Interface).
Similarly, virtual interfaces seemed to be more enjoyable to use than the walkthroughs. Many participants in the Tracked condition wanted to continue viewing the museum even after the tour ended. This could not be said of participants in the walkthrough type conditions. Perhaps this is also why the virtual interface was judged to be more professionally acceptable in many hypothetical instances. In any event, the appeal of the interface and its ease of use are its own very important distinctions. They confirm the significance of this technology and endorse continued research.
Future Research.
This study yielded conclusive results indicating that the limited field of view in the simulation conditions affected people's ability to judge distances accurately. Preliminary results also indicated that the restricted field of view and the lack of kinesthetic feedback might explain why it is so difficult for people to understand their orientation in the virtual environments. All of these difficulties were also found to be compounded by head-position tracking. To clarify these findings, the following areas of research would be required.
Studying effect of limiting the field of view in the periphery on:
* people's ability to build cognitive maps
* the perception of sizes and distances
Studying the effect of the head-position tracking on:
* the process of visually searching the display
* people's ability to build cognitive maps
Studying the effect of kinesthetic feedback for walking on:
* people's ability to build cognitive maps
* the perception of sizes and distances
There were also several instances where the results from the study were inconclusive. Some of the methods of measurement were either inappropriate or they were improperly conducted. The following areas could benefit from a re-investigation:
* using an alternative method for measuring people's understanding of the spaces
* using the complete adjective checklist in the qualitative study
Another area of research is comparing the use of an interaction device which is "anchored" to the physical space (as the Spaceball was the case in this study) to the use of the same device when held in the participant's hand, and the effect this has on:
* the process of visually searching the display
* people's ability to build cognitive maps
Extended Areas of Future Research
There are many other distinctive aspects of virtual environments which could affect the way people perceive spaces. Some of these characteristics could be exploited if they were found to improve users' spatial perceptions. Some of the sample topics are the following:
Exploring the effect of penetrable/impenetrable surfaces on:
* people's ability to build cognitive maps
* people's subsequent interaction with the real space
Exploring the effect of non-constrained vertical movement on:
* people's ability to build cognitive maps
* people's subsequent interaction with the real space
Exploring the effect of being able to render surfaces transparent on:
* people's ability to build cognitive maps.