Navigational Techniques to Improve the Display of Large Two-Dimensional Spaces

By David Beard and John Walker

Summary and Analysis by Kamika Lawrence

Summary

This paper addresses the issue of navigating through information spaces that are larger than can be displayed on the screen. Beard and Walker perform a controlled experiment to analyze two techniques, map windows and zooming, to aid in finding information. Map windows are miniature windows that display the entire information space and have some indication of the user’s current position in the window. Map windows allow for easy roaming throughout the information space. The experiment compared map windows and zooming with scroll bars.

The previous works section of the paper discusses some the current techniques used to navigate through two-dimensional information spaces and gives some of the issues with them.

Hypertext is a common technique that uses a two dimensional information space. Hypertext is a collection of labeled nodes connected by labeled lines. The method for finding nodes in hypertext is human visual search. However, this can become difficult. Users may become lost or make unsuccessful search decisions.

They mention some other techniques and systems that try to solve the problem of viewing data that is larger than the screen, but they don’t go into detail about the systems. Some of the systems mentioned include the spatial information management system and a building walkthrough system.

Navigation consists of the viewing and manipulation of he computer screen to show another portion of the information space.

Two elements of navigation are cognitive and mechanical. The cognitive element of navigation is the basic information about the current location of focus and where others areas are located that users carry in memory. The mechanical element of navigation is the use of motor skills to move the user’s focus of attention.

Three methods used to display information when it is too large to be displayed at once are truncating, shrinking, and distortion. Truncating crops the information space so only a portion is displayed. This is a common technique in text editors. The shrinking technique does not display items below a certain resolution. The distortion technique makes items in the center of focus appear larger than other items farther away from the center.

The techniques used to manipulate the display vary in the ease of which users can allow diagonal and curvilinear movement, and the ease with which users can rapidly access remote regions of the space.

Some problems that occur in navigating over two-dimensional information spaces are as follows:

• In hypertext systems, there is a question of whether or not the tree or graph should be updated every time the user manipulates the display.
• Users may require simultaneous, detailed viewing of several portions of the information space.

• Information spaces that are large often use a recursive information space concept. The recursive information space concept, where any detailed view could be a mapped view, increases the richness of the information space, but it possibly increases the complexity of the mental model for the user.

The Study

The objective of the navigational techniques is to allow easy access to various locations in a two-dimensional information space. The three techniques used are a map window, zooming, and roaming. The map window has a trade off of usage of screen space vs. increased notion of the current location in the information space. Zooming allowed drawing a wired framed box in a map window and this information is then displayed in the main window. Zooming also allows for magnification of the map and main windows. The roam technique allowed the wire framed box in the map window to change locations, however, the box was a fixed size and there it did not allow the changing of magnification of either of the windows.

The experimenters came up with the following hypotheses:

1. Roam and roam-and-zoom each will be faster than the scroll bars.
2. Each of the techniques with the map window would be faster than without.

The roam, zoom and scroll techniques were each done with and without the map window. Since the map window was small enough that the users could not locate nodes from the map window.

The study consisted of two different within-subject, repeat measure, three-by-two experiments. The first experiment gave the users the task of finding a word in a balanced binary tree. The second task was to move from the apex node of a binary tree to displaying the right-most leaf node and back five times. Throughout the experiment, the analysis of variance technique was used to measure the significance of the results.

For the first task, each of the six subjects ran ten trials. The results were significantly better (with 90 % confidence) when using the map windows than without. The use of roam and roam-and-zoom were both significantly better (with 90 % confidence) than scroll bars.

The second task each of the six subjects ran six trials. With 99.5 % confidence, the use of the map window was significantly better than without. Both roam and roam-and-zoom were significantly better than scroll bars with 99.5 % confidence.

There seemed to be little difference between roam-and-zoom and roam techniques.

Analysis

I don’t really know that much about controlled experiments but it seemed to me that the authors did a good job of eliminating external influences in the results. For example the subjects were computer science graduate students so mouse manipulation would not affect the navigational technique. The experimenters used both a complex and a simple problem for the subjects to use when comparing the techniques.

Questions

1. Were the same examples given to all the users in training so that none of the users may have had better training than any of the others?