Introduction

Also there are many 3D information visualization tools, there is little data to support their added value. In a recent "browse-off" at CHI'97, and in more systematic empirical studies [7], 2D visualizations like tree-maps or hierarchical file browsers did surprisingly well in comparison with a dynamic 3D hyperbolic browser. This lack of 3D advantage is consistent with a more extensive empirical literature comparing static 2D and 3D data visualization graphs (e.g., [1,2]).

There are a number of information processing approaches that provide partial accounts for these effects (e.g. [6,9]). Our approach is to examine the perceptual characteristics that influence visualization design. In this paper, we briefly discuss one aspect of that analysis---the way in which dimensional interaction can help to explain undesirable artifacts in 3 D visualization.

Appropriate interaction among dimensions produces "perceptual redundancy" or "emergent properties" that enhance perceptual displays [4]. In some cases, two different dimensions (say height and color saturation) both represent the common property (say frequency of a document access); such perceptually redundant dimensions should facilitate information extraction (e.g., finding the most frequently accessed documents). In other cases, two dimensions are perceptually "integral;" they function together as a new whole to provide emergent visual properties. For example, lengths (number of hits per day) and width (number of days examined) dimensions will result in a figure with area (total number of hits).

Although such redundant or emergent properties can improve the display of information, their inappropriate use can explain some of the unexpected results in design of 3 D visualization.

Misapplied redundancy

Most spreadsheets include a tool to make 3D graphics from the same data that a represented with 2D display is. This does not improve processing through redundancy, since a volume is not just a 3D amplification of a 2D planar surface; differences in volumes are perceived as substantially greater than differences in heights, resulting in what Tufte has called the Lie Factor [10]. A simple solution in this case is to match the number of information-bearing perceptual dimensions to the number of relevant data dimensions.

Spurious integrality

Although the mapping of perceptual dimensions to information is arbitrary, the interaction among the chosen dimensions can be significant. For example, in conducting a large bibliographic search, they resultant matches are presented in an ordered list that is hard to scan or conceptualize. One solution is to break up the search task into lower dimensional entities that can be isolated and reduced [8]. Another option is to represent all of the data in the more especially compact for, such as a spiral, with distance from the center along the spiral as symmetric for rank in the ordered list [3,5]. These representations contain new perceptual groupings that have no well-defined relationship to the source data. In informal assessment of such displays, for example, we have noted that there might be an alignment of data in one quadrant of the spiral. However, that perceptual grouping is an artifact of the display that is meaningless to the viewer. Changing the scale of the mapping or the tightness of the spiral eliminates that group but can easily produce others that are completely devoid of meaning.

Conclusions

Descriptions of dimensional interaction make it clear that 3D displays need to assign meaning not only to individual dimensions but to the perceptual relations among those dimensions. These brief examples illustrate how perceptual principles like dimensional interaction can be used as guides in avoiding distracting artifacts in information visualization and can thus serve as an important part of the design process.

References

[1] Barfield, W., & Robless, R. (1989). The effects of two- or three-dimensional graphics on the problem solving performance of experienced and novice decision makers. Behaviour and Information Technology, 8, 369-385.

[2] Carswell, M.C., Frankenberger, S., & Bernhard, D. (1991). Graphing in depth: Perspectives on the use of three-dimensional graphs to represent lower-dimensional data. Behaviour and Information Technology, 10, 459-474.

[3] Cugini, J., Piatko, Christine, Laskowski, Sharon (1996). Interactive 3D Visualization for Document Retrieval. Workshop on New Paradigms in Information Visualization Manipulation, ACM Conference on Information and Knowledge Management (CIKM'96).

[4] Garner, W. R. (1978). Aspects of a stimulus: Features, dimensions, and configurations. In E. Rosch & B. Lloyd (Eds.), Cognition and categorization. Hillsdale, NJ: Lawrence Erlbaum Associates.

[5] Gershon, Nahum. (1997). Information visualization: 4. Case studies. CHI '97 Tutorial Notes on Informatino Visualization. ACM.

[6] Kosslyn, S. (1989). Understanding charts and graphs. Applied Cognitive Psychology. 3, 185-226.

[7] Lamping, J., Rao, R., & Pirolli, P. (1995). A Focus+Context Technique Based on Hyperbolic Geometry for Visualizing Large Hierarchies. Proceedings of CHI'95. ACM.

[8] Plaisant, C., Bruns, T., Shneiderman, B., & Doan, K. (1997). Query previews in networked information systems: The Case of EOSDIS. Proceedings of CHI'97. ACM.

[9] Simkin, D. & Hastie, R. (1987). An information-processing analysis of graph perception. Journal of the American Statistical Association, 82, 454-465.

[10] Tufte, E. (1989). The visual display of quantitative information. Cheshire, CT: Graphics Press.

Posters Presentation

A Visualization Architecture for Enterprise Information by Lester Lee

Using Theories of Perceptual Dimensional Interaction to Improve Information Visualization Design by Marc M. Sebrechts

Dosimetry with GAF Chromic: Experimental verification of software accuracy in stereotactic radiosurgery by R. Foroni, et al.

Table of Contents