Representation Matters:

Spatial Interface Can Facilitate Target Acquisition

IBM Research - Almaden, 650 Harry Road, San Jose, California 95120, USA

+1 408 927-1912, {selker, imay, zhai}@almaden.ibm.com

ABSTRACT

We present an experiment which demonstrates significant learning and target acquisition advantages of a physical metaphor over a standard icon metaphor in user interface design. Several differences between an icon based interface and a physical based interface contribute. The nonuniformity of layout makes objects easier to identify; the spatial relations and landmarks are more memorable than positions on a surface, and the representations of the physical interface are more closely connected to each other and the semantics of the available task functions.

Keywords: Interface style, Spatial Interface, Graphical Interface, 3D interface.

INTRODUCTION

Which various presentation and representation techniques will have value in specific user interfaces is a matter of design and empirical tests [1, 2]. Evocative proposals for using three dimensions and physical metaphors to "simplify" user experience have existed in many forms and in many contexts for decades. Possibly the most influential interface of this sort was the highly publicized commercial introduction of Magic Cap made by General Magic in which a 3D desk and mall are used as the orienting mechanisms to show people "where" they can find and use tools instead of using the desktop and folder mechanisms. Following this work, Microsoft's Bob and others began using more 3D interfaces in commercial software.

The current study explores the concepts and value of such physical metaphors through a controlled experiment. Psychologists have suggested that contextual information, such as the physical relationships of objects, facilitates information comprehension and memory [4, 5]. The "method of loci" is an example: people can better remember an arbitrary series of objects by mentally associating them with locations in some real or imaginary space [3, 6]. It has also been argued that people unconsciously treat computers like real places [7]. Following these ideas, we examined whether interfaces that incorporate spatial structures like those in the real world would improve users' performance. Conventional graphical user interfaces have transformed many recall tasks in command line interfaces to recognition tasks, thus reducing users' memory load. We view spatial interfaces as the plausible next step.

The Experiment

The interface we studied was chosen from an existing web-based computer system management tool. This interface consisted of regularly-spaced, equally-sized icons, as in any typical conventional GUI interface. Figure 1 shows a mockup that resembles such a style of interface.

Motivated by the hypotheses in last section, we developed a VRML interface for the same system management software. Figure 2 is an mockup that resembles the VRML interface we developed, in which, for example, the representation of network looks like a pipe coming out of a computer and the schedule of maintenance and updates looks like a calendar hanging on the wall. The larger view of the network looks like a globe sitting on the desk. If you click on the "computer," it will open displaying the parts inside as well as status of the parts. Also, selecting the "visual screen" on the "desk" presents the actual screen view of a computer on the network which can then be used by the system adminstrator to interact with that machine.

In this study we specifically examined whether objects in a user interface are more easily found and their locations more easily remembered in a regular iconic presentation (Fig. 1) or a more physically realistic spatial presentation (Fig. 2).

Iconic interface

Fig. 1. The "iconic" interface

Spatial interface

Fig. 2. The "spatial" interface

We conducted a balanced, within subjects experiment. Eleven subjects were asked to find and point to targets on the two interface mockups as quickly as possible without making errors. Each trial of the experiment consisted of a list of six target labels in a random order. The same six labels were used in all trials. Subjects sequentially pointed to the targets corresponding to the labels on the list. Each subject performed four trials with each of the interfaces. The same tests were repeated on a second day.

Experimental results

Fig. 3. Experimental Results

Subjects' trial completion times were recorded. To more reliably assess differences in the time measures, we applied logarithmic transformation on these data for statistical significance tests. Repeated measure ANOVA analysis on the results showed that the subjects' completion times with the spatial interface were significantly shorter than with the iconic interface (F 1, 10 = 9.8, p < .05). As expected, little difference existed between the two interfaces in the first trial, since the subjects had to search for each object at this point in either case. After some practice, subjects presumably developed graphical (iconic) as well as spatial (location) memory associated with each of the objects, helping them to complete the tests faster. To investigate this further, we collapsed the data across the three last trials for each day. We found the interaction between interface (iconic or spatial) and day (1 or 2) to be significant (F 1, 10 = 7.1, p < .05). That is, the advantages of the spatial interface were more evident over time, suggesting that the spatial interface better helped subjects learn and remember the locations of the target objects.

Discussions and Conclusions

Our study shows both performance and learning advantages of 3D / spatial interfaces. The data demonstrating the different learning curves of the two styles of interface is evidence of the demonstrable value of spatial/3D interfaces. Several differences between the icon based interface and the physical based interface might have contributed. The nonuniformity of layout may make objects easier to identify. The spatial relations and landmarks may be more memorable than positions on a flat surface. Also, the representations in the physical interface may be more closely connected semantically to each other and to the available task functions.

We are particular encouraged that the hardware and software (such as VRML) that support 3D interface design and prototyping is increasingly more available on low end computers.

We suggest future research on which characteristics of the spatial interface made the learning and performance improvements, such as non regular layout, textural information and connection between objects and semantic mapping.

Acknowledgment

We thank Dr. D. Christopher Dryer for his insightful comments and suggestions on the study.

REFERENCES

[1] Selker, T. Wolf, C. and Koved, L. A framework for Comparing Systems with Visual Interfaces, Proceedings of Interact 87.

[2] Selker, T. and Appel, A.. Graphics as Visual Language. Handbook of Statistics. Elsevier Science Publishers, Vol 9, 825- 855, 1993.

[3] Ross, J., and Lawarence, K.A. Some observations on menory artifice. Psychometric Science, 13: 107-108, 1968.

[4] Bower, G.H. Analysis of a mnemonic device. American Scientist, 58: 496-510, 1970.

[5] Bransford, J.D., and Johnson, M.K. Contextual prerequisites for understanding: Some investigations of comprehension and recall. Journal of Verbal Learning and Verbal Behavior, 11, 717-726, 1972.

[6] Luria, A. R. The mind of a mnemonist. Basic Books: New York, 1968.

[7] Reeves, B. and Nass, C. The Media Equation: How People Treat Computers, Television, and New Media Like Real People and Places. Cambridge University Press: Cambridge, 1996.