VISFILES

Wayfinding2: The Lost World

T.Todd Elvins
San Diego Supercomputer Center

November 97 Columns



T.Todd Elvins

Introduction

In my last column I defined some terminology and posed two problems with regard to wayfinding in virtual environments:

• Acquisition: How can travelers best find their way to locations of interest and find their way back to locations previously visited?
• Facilitation: How should worlds and world browsers be designed to facilitate skilled wayfinding behavior?

I then reviewed a portion of the literature on real-world wayfinding to set the stage for this column on virtual environment research. This area of research raises two more problems that we will want to discuss:

• Transportation: In virtual environments, what modes of travel foster efficient cognitive map construction and maintenance?
• Familiarization: Do affordances such as 2D maps assist in wayfinding if studied before or during a visit to a virtual environment ?

The answer to the acquisition question will be dependent on solutions to the other three, and a set of conclusive solutions to all four could someday evolve into a set of principles, and a methodology, for designing legible worlds and usable world browsers.

Before starting, it is important to note the distinction between wayfinding and navigating. Navigating is most often defined as the process of following a course by making directed movements, while wayfinding is the process of determining the strategy, direction, and course needed to reach a desired destination. Without wayfinding, a navigator won't know in which direction to steer and without navigating, a wayfinder will not have the means to move toward their destination. We will see that different types of wayfinding tasks require different types of strategies.

One set of real-world travelers provides some interesting lessons in wayfinding. For centuries the native people of the Puluwat (Poo-loo-watt) Atoll in the South Pacific have found their way between islands across hundreds of miles of featureless ocean without the use of instrumentation [6]. Relying on a dead reckoning mode of travel (where one's position is determined by the distance and direction traveled since leaving the last known location), and cues in their environment such as bird flight paths, underwater reefs, stars, relative orientation of landmarks, and many others, the Puluwatans sail their tiny canoes with amazing accuracy. The Puluwatans have developed a set of three rules by which they can maximize their likelihood of success: (1) set out in a direction so as to arrive in the vicinity of the destination, (2) hold a steady course and a running estimate of current position, and (3) ensure that when near, there is some technique for locating the destination island and heading toward it.

Figure 1. A Puluwatan star compass after Goodenough (1953) [7].

Figure 1 shows a type of compass labeled with directions to 28 stars by which Puluwatan navigators, such as the famous Hipour (hi-power) the Navigator, can set a course (by our definition Hipour is actually a wayfinder). One navigational strategy used in conjunction with the star compass, and not well understood by anthropologists, is based on imaginary landmarks, which the Puluwatans refer to as sealife. Sealife is an ordered and named inventory of landmarks (or seamarks) extending out in 28 individual star compass directions from every island. For example, if one were to sail toward Vega from Sau Island, one would expect to pass two white dolphin, then a red-tailed tropical bird, then a single large barracuda, and so on. The sealife paths and landmarks play a major role in village folklore but the names of each are known only to the specially trained navigators.

Let's review how Puluwatans solved the four wayfinding problems listed above in a real world situation. They solved the acquisition problem by developing a set of three rules that maximize their chances of successfully finding their way. To follow these rules our other three problems must be solved. The facilitation problem was solved by redesigning a sparsely featured ocean to include invisible landmarks. Whether the sealife is seen or not (they are rarely seen), the navigators sail with confidence knowing that the sealife is there. Mentally rehearsing canoe mode travel fosters formation of a cognitive map thus solving the transportation problem. Motivation to mentally practice a route and to utilize all available cues is strong because their very lives depend it. Puluwatans solve the familiarization problem through navigational training, personal experience, and storytelling. During training, the navigators learn to use a compass affordance, and to find their way in a sparsely structured area using actual, relative, and invisible landmarks. Can the problems of wayfinding in virtual environments be solved in similarly clever ways ? Let's take a look.

More Invisible Wayfinding Cues

In 1988 Fred Brooks observed that people form precise world-models of familiar environments such as bedrooms and offices [2]. People can find their way, "in the dark and reach for objects without looking at them." Personal visits to a region enable people to effortlessly form rich 3D cognitive maps (often without even knowing they are doing it) and to find their way in the dark. This mode of travel is not so different from the way Puluwatans use landmarks that are only seen in their mind. Brooks goes on to state that, "forming similarly accurate mental models of virtual worlds requires hours of exposure to these worlds, plus every feasible cue." In developing affordances to enable navigation of virtual worlds, Brooks repeatedly found that a traveler, "needs to see both a view of the virtual world and a map showing where he is and where he is looking." Brooks observed that travelers began by studying the map and then proceeded to world navigation as their cognitive map matured. These conclusions favor filling a virtual environment with wayfinding cues to solve the facilitation problem, and using 2D maps to solve the familiarization problem; at least for some tasks.

Virtual Buildings

Satalich's research supports Brook's observations [11]. Satalich studied the general question of how best to introduce travelers to a new virtual environment and what navigational aids are beneficial during initial exposure. Conducting a series of empirical experiments, Satalich familiarized subjects with a virtual building of 39 rooms (shown in Figure 2) in three different ways:

• Self exploration: subjects traveled through the building in any manner they wanted
• Active guided: subjects followed a pre-determined path through the building
• Passive guided: subjects moved through the building at a constant speed under the control of the software. This is not unlike the invisible trolley to be discussed momentarily.

Half of the subjects also saw a north-up 2D map of the building during the familiarization tour. The control group had no tour but instead studied a 2D map before the search task. Half of the non-control subjects also studied the 2D map before the search task.

Figure 2. A plan view of Satalich's virtual building. Subjects started at the entrance and visited all the rooms during the familiarization. There is at least one door to each room (not shown).

The results showed that subjects who studied the map performed better than subjects who did not study the map. Surprisingly the control subjects, who saw only the 2D map, performed as well as the subjects who received the familiarization tours. Perhaps the subjects performed better with the map because they had been using maps for a lifetime while they had never previously visited a virtual environment. Perhaps the virtual environment was not sufficiently interactive or detailed, or perhaps the technology was just too immature.

We would have expected -- based on real world wayfinding literature -- that the solution to our transportation problem would involve a self exploration mode of travel. Satalich's results suggest that, under the given conditions, the task of relocating a particular room is best accomplished using 2D map study familiarization. Certainly further research is needed to explain this unexpected result.

Virtual Coastlines

Darken was the first to apply lessons learned from real-world wayfinding studies (primarily [9]) to virtual environments [4]. His dissertation argues that: "real-world wayfinding and environmental design principles are effective in designing virtual worlds which support skilled wayfinding behavior." In his timed within-subject study, subjects started from an virtual oil tanker platform and tried to find their way to other ships in a virtual ocean/island environment. Subjects were asked to complete two tasks

• Conduct an organized exhaustive search for five ships
• Return to the oil tanker in four virtual worlds, each of which embodied a different treatment:
  1. The control treatment: no wayfinding assistance provided
  2. The grid treatment: a radial grid was overlaid on the 3D world
  3. The map treatment: a 2D "you-are-here" map was visible
  4. The map/grid treatment: both 2 and 3

To construct the grid treatment environment, Darken added a radial grid similar to a wireframe dartboard to the virtual environment as shown in Figure 3. The map treatment added a small 2D you-are-here map to the display. The map/grid treatment added both. Subjects performed better in treatment #2 than in treatment #1, and better yet in treatment #3. Surprisingly, using both the map and grid (treatment #4) resulted in performance inferior to using the map by itself. Darken's conclusions include the following:

• When not given a source of directional information, disorientation will inhibit both wayfinding performance and spatial knowledge acquisition.
• A structure must be imposed on the world if an organized exhaustive search is to be attempted.
• Path following is a natural spatial behavior. Subjects frequently used features such as coastlines or grid lines as if they were paths.
• A map allows for optimizations to be made to search strategies. This is because the map can be considered a supplement to survey knowledge.
• Dead reckoning ... appears to be more easily understood and implemented in virtual spaces [than in the real world].

Figure 3. Darken's radial grid from a bird's eye view. Tall colored posts are positioned at the North, South, East, and West boundaries. Circles are color coded Red (R), Yellow (Y), and White (W).

Familiarization solutions advocated by this study include the following:

• Show organizational elements (paths, landmarks, districts, etc.) and particularly the organizational principle.
• Always show the observer's position.
• Orient the map with respect to the observer such that the forward-up equivalence principle is accommodated.

Darken was successful in showing that Lynch's basic theory (a structured environment facilitates wayfinding [9]) holds in virtual environments. In regard to our facilitation problem, Darken's first three conclusions suggest that: (1) virtual environments should be structured to enable a traveler to form a cognitive map and find their way using that map, and (2) virtual environment browsers should be designed to aid travelers in perceiving cues and structure in the environment.

A Virtual City

One final example study does not use a 2D you-are-here map. The study by Gillner and Mallot offered subjects no global information at all [5]. Instead subjects were placed at a home position in a virtual hexagonal maze resembling an affluent neighborhood with streets, hedges, and houses. Subjects were then shown a picture of a landmark as seen from a target viewpoint. While searching for the viewpoint, they traveled along street segments on a constant speed invisible trolley, and re-oriented themselves in increments of +/- 60 degrees with mouse clicks.

Figure 4. A small part of the hexagonal city from a bird's eye view. Small black-filled circles represent buildings, larger hollow circles represent trees and hedges, and lines represent road segments.

The objective was to find the shortest route between the starting point and the target viewpoint. If the subject reached the target viewpoint by a non-optimal route, or moved more than one segment off of the optimal route on the way, they were advised that they would have to start again. Analyzing the results of the experiment, the investigators found that the subjects readily learned optimal routes without a map using only local landmarks and route decisions. Additionally, subject performance improved during the course of twelve trials/targets in the hexagonal city. Further analysis showed that the improvement was due to the subjects piecing together -- into cognitive maps -- the landmarks, intersections, and routes seen while looking for previous target viewpoints.

This is an interesting result. The personal exploration required to complete the task led to high-quality primary survey knowledge which is the basis for skilled wayfinding. A number of observations and issues come to mind:

• The environment provided sufficient cues, structure, and landmarks for a cognitive map to be built. The hexagonal city gives us a baseline for what objects must be included in a solution to our facilitation problem.
• Although the experiment started off with subjects performing a naive search for target viewpoints, by the time they refined eleven optimal routes, the twelfth task had taken on many characteristics of a primed search. For example, the twelfth task might require locating a fish taco stand seen during the seventh task.
• A number of the objects in the virtual city had been previously specified as the subject of target viewpoints and so had taken on personal meaning and hence become landmarks.
• Since the subjects performed all twelve tasks, usually several times each, in the same virtual environment, the subjects saw landmarks from multiple vantage points. This is another means of strengthening a mental representation of a 3D landmark.
• Survey knowledge was constructed without the use of a 2D map, suggesting that not all virtual environment tasks require a map for familiarization.
• The invisible trolley allowed the subjects to look around and see landmarks while they were being shuttled to the next intersection. The trolley design and the nature of the task ensured that the subjects perceived landmarks along the way.

Transportation by trolley is an excellent example of how software affordances can be designed to foster the creation and maintenance of a cognitive map. In this case the software-controlled translation through the scene helped the traveler to perceive landmarks in the traveler's vicinity. On an unrelated but important topic, other studies have shown that software-controlled movement can cause cyber motion sickness in some people. Gillner and Mallot do not mention motion sickness in their paper, but it may be the case that, since subjects were in control of their view angle while riding the trolley, motion sickness was mitigated.

Other Studies

Several studies have found that virtual environment task familiarization has positive transfer training effects to real world tasks [1, 8, 13]. In particular, Witmer found that virtual environments were effective in familiarizing subjects with real world routes [14].

Real world wayfinding research suggests that studying a map before entering an unfamiliar region is helpful for creating secondary survey knowledge. Secondary knowledge is, however, inferior compared with primary survey knowledge which is built through a personal exploration type of transportation [10]. It will be interesting in the future to see if this theory holds in virtual environments. Butler showed that having a map during a real world familiarization period may or may not be helpful during a subsequent search task depending on the search task to be performed [3]. Butler also found that posted signs are generally advantageous, and Streeter found that taped verbal directions are extremely advantageous [12].

Discussion

This column's primary goal was to try to identify wayfinding design principles for: virtual worlds, virtual world browsers, travel modes, and familiarization affordances. From the Puluwatans and Fred Brooks we learned the importance of a cognitive map; cognitive maps are so powerful that people can skillfully find their way in invisible environments and based upon invisible landmarks. Satalich and Witmer agree that, although maps work, better methods for efficiently familiarizing first time visitors with routes, landmarks, and other spatial information may exist. Experiments by Brooks, Satalich, and Darken showed that previewing and/or carrying a 2D map of a 3D world is an effective familiarization solution and results in more expedient wayfinding for some tasks. Gillner and Mallot showed that humans readily build cognitive maps of virtual environments given: a sufficiently detailed environment, a task requiring landmark searching, and a clever transportation solution. For traveling in virtual environments (and perhaps in the real world), the penultimate spatial information affordance of the future will: (1) be as fun, fast, and easy to use as a map of Disneyland, (2) be adaptable to different types of tasks and different modes of travel, and (3) help travelers perceive cues in their environment.

T.Todd Elvins is a Staff Scientist at San Diego Supercomputer Center, and a Computer Engineering Ph.D. candidate at the University of California, San Diego. His research interests include human factor-based user interface design, data and information visualization, image synthesis, and web-based imaging. He can be contacted at:

T.Todd Elvins
San Diego Supercomputer Center
University of California
San Diego
MC 0505
La Jolla, CA
92093-0505, USA

Website

The copyright of articles and images printed remains with the author unless otherwise indicated.

References

1. Bliss, J.P., Tidwell, P.D., Loftin, R.B., Johnston, B.E., Lyde, C.L., Weathington, B.L., "An experimental evaluation of virtual reality for training teamed navigation skills," Presence: Teleoperators and Virtual Reality, To appear.
2. Brooks, F.P. Jr., "Grasping reality through illusion -- interactive graphics serving science," Proceedings of the ACM CHI Conference on Human Factors in Computing, Washington, D.C., 1988, 1-11.
3. Butler, D.L., Acquino, A.L., Hissong, A.A., and Scott, P.A., "Wayfinding by newcomers in a complex building," Human Factors, 35(1), 1993, 159-173.
4. Darken, R.P., Wayfinding in Large-scale Virtual Worlds, Ph.D. dissertation. School of Engineering and Applied Science, The George Washington University, January, 1996.
5. Gillner, S., and Mallot, H.A., "Navigation and acquisition of spatial knowledge in a virtual maze", Technical report #45, Max Planck-Institut für biologische Kybernetik, http://www.mpik-tueb.mpg.de/bu.html
6. Gladwin, T., East is a Big Bird - Navigation and Logic on Puluwat Atoll, Harvard University Press, 1970.
7. Goodenough, W.H., Native Astronomy in the Central Carolines, Museum Monographs, Philadelphia: The University Museum, University of Pennsylvania, 1953.
8. Loftin, R.B., and Kenney, P.J., "Training the Hubble space telescope flight team," IEEE Computer Graphics and Applications, 15(5), September, 1995, 31-37.
9. Lynch, K., The image of the city, M.I.T. Press, 1960.
10. Presson, C.C., and Hazelrigg, M.D., "Building spatial representations through primary and secondary learning," Journal of Experimental Psychology: Learning, Memory and Cognition, 10, 1984, 716-222.
11. Satalich, G. A., Navigation and Wayfinding in Virtual Reality: Finding the Proper Tools and Cues to Enhance Navigational Awareness, Masters Thesis, Department of Computer Science, University of Washington, 1995. http://www.hitl.washington.edu/publications/
12. Streeter, L.A., Vitello, D., and Wonsiewicz, S.A., "How to tell people where to go: comparing navigational aids," Int. J. Man-machine Studies, 22, 1985, 549-562.
13. Tate, D.L., and Sibert, L., "Virtual environments for shipboard firefighting training," Proceedings of the 1997 ACM Virtual Reality Annual International Symposium, Albuquerque, New Mexico, 1997, 61-68.
14. Witmer, G.G., Bailey, J.H., Knerr, B.W., Parsons, K.C., "Virtual spaces and real world places: transfer of route knowledge," Int. J. Human-computer Studies, 45, 1996, 413-428.