Do Metaphors Make Web Browsers Easier to Use?  

Elissa D. Smilowitz
Claris Corporation
5201 Patrick Henry Drive, MS: C-62
Santa Clara, CA 95052
(408) 987-7508
elissa_smilowitz@claris.com      

ABSTRACT

Since Visicalc’s metaphorical ledger and the Xerox Star’s desktop metaphor, interface designers have been incorporating metaphors into user interfaces. User interface guidelines for most of the popular operating systems encourage the use of metaphors in interface design. They suggest that applications should build on the user’s real-world experience by exploiting concrete metaphors thereby making applications easier to use. Surprisingly little research supports the popular belief that metaphors in user interfaces facilitate performance.

This research explores the use of metaphors in interface design. Are user interface (UI) metaphors effective in facilitating performance, and if so, how can they be designed to be most effective? The World Wide Web serves as the application domain for this research.

A series of experiments show that UI metaphors can facilitate users’ interactions. However, various metaphors are not equally effective, some are no better than non- metaphoric interfaces. In Experiment 1, a metaphoric interface was better than a non-metaphoric interface, and icons did not contribute to the effectiveness of the metaphor. In Experiment 2, a good integral metaphor was superior to a composite metaphor. These findings contribute to user interface design by establishing some of the characteristics of effective metaphoric interfaces. Furthermore, it provides some practical directions for a UI metaphor for the World Wide Web.

KEYWORDS: Metaphors, user interface design, empirical evaluation, Internet, World Wide Web, Mosaic

INTRODUCTION

Metaphors are ubiquitous in the user interfaces of today’s computers. Software designers are incorporating metaphors into a variety of software from operating systems to information retrieval applications. Technological advances have made more and more realistic depiction of these metaphors possible. However, technological feasibility does not insure psychological utility.  

The motivations for using metaphor in the design of user interfaces are similar to the reasons metaphors have long been popular in education. Many educators have observed that giving students comparisons can help them learn. For example, an analogy commonly used in teaching about electricity is “Electricity is like water”. Imagine electricity flowing as water does. You can then imagine the wires as pipes carrying water (electrons). It follows that your wall plug can be thought of as a high-pressure source which can be tapped by inserting a plug [6]. These types of comparisons are also used in teaching in the domain of human-computer interaction. For example, a physical metaphor for electronic storage is to think of “storage locations as buckets.” Experimental studies of the effectiveness of metaphor in teaching programming concepts have been conducted. Mayer showed that many programming constructs in BASIC (i.e., memory locations) could be learned more easily when they were presented in the context of a concrete metaphor [10]. Thus, educators in many domains believe that students can import conceptual relations and operations from one domain to another.  

Research has demonstrated that there may be great value to teaching about existing computer systems through the use of metaphor (i.e., the text editor is like a typewriter) [10, 11]. Yet, little research has been conducted as to how to design a user interface around a metaphor. Does metaphor provide the same advantage in the domain of user interface design?

Current interface metaphors exploit a variety of prior knowledge of familiar objects and activities. For example, in desktop operating systems (e.g., the Macintosh system), routine file management tasks become familiar operations on familiar objects such as getting a folder from a file cabinet [2]. These systems frequently carry the metaphor even further by displaying documents as icons and showing simultaneous activities in separate places like stacks of papers on a real desktop [9].  

To a large extent, the effects of metaphor on users’ performance with a system are unknown. Although many guidelines encourage the use of interface metaphors and many interface metaphors are currently employed [1, 8], there is little if any empirical research demonstrating the claimed benefits of metaphor. Beyond that there is no research exploring how to design the most effective metaphor. Little is known about what characteristics enhance the power and utility of interface metaphors.  

The research summarized here is an initial attempt to understand how metaphors can facilitate users’ performance, and to identify the characteristics of metaphors that make them enhance performance. This series of experiments demonstrates a performance advantage provided by UI metaphors compared to non- metaphoric interfaces. The findings also identify some of the characteristics that contribute to the effectiveness of metaphor.  

The World Wide Web was chosen as the application domain in which to study UI metaphors. Several alternative user interfaces were designed based on the task of searching the World Wide Web for information. Thus, the following studies also provide a practical direction for the design of an effective metaphoric interface for exploring the Web.

EXPERIMENT 1: DO METAPHORS AID PERFORMANCE?

The first step is to establish if metaphors do in fact facilitate users’ performance. An important issue is how to convey a UI metaphor. There is a substantial body of work in linguistics, philosophy and education about metaphor and metaphor comprehension. Yet how this applies to UI metaphor is not completely understood. In a graphical user interface, it is commonly assumed that UI metaphors are carried both by the terminology (i.e., function names) and graphics (i.e., icons). In fact, many people incorrectly assume that a UI metaphor cannot exist without the presence of graphics. While it may be that the metaphor is much more compelling and apparent if incorporated into the interface graphics, it is certainly not a defining characteristic of a metaphor. The UNIX pipe command is an example of a metaphoric command. Therefore, this experiment attempts to establish what aspects of the interface contribute to the metaphor.

Subjects

Forty undergraduate students at New Mexico State University participated in partial fulfillment of an experimental credit requirement in introductory psychology. All subjects had used a computer mouse before but none had prior experience with Mosaic or any World Wide Web browser. The forty subjects were randomly divided between each of four conditions.

Materials

The experimental paradigm was based on the Mosaic software application used to search for information on the World Wide Web. The terminology for the two no- metaphor conditions was based completely on Mosaic’s function names. The terminology for the two metaphor conditions was based on a library metaphor. Specifically, each of the key functions in Mosaic was mapped to a concept that is found or could be used in a library (table 1). The library domain was chosen because it’s primary function of searching for information is analogous to the primary task performed in Mosaic; that of searching for information on the World Wide Web. All subjects performed tasks on an HP Vectra 486/66 XM personal computer with a VGA resolution color monitor.

                        

Mosaic Terminology (No-metaphor)

Library Terminology (Metaphor)

World Wide Web

Library

Open URL

Search bookshelves

Window History

Stack of viewed books

Hotlist

Bookmark

Annotate

Post-it notes

Document

Book

Home

First

Forward

Next

Back

Previous

Clone Window

Copier

Save as

Translator

Find in current

Find text

Internet Resources Metalist

Reference section

Table 1. Function labels used in No-metaphor and Metaphor conditions.

Design

There were four conditions which differed in the interface terminology (function labels) subjects received -- either based on a library metaphor or not -- and in the use of icons -- either library icons or no icons. Thus, in one condition, subjects saw library metaphor terms along with icons representing a library, whereas subjects in the second condition saw the metaphor terms but with no icons. Likewise, subjects in the third and fourth conditions received the Mosaic terms (no-metaphor terms) either with icons or without. Thus, there were two between-subjects factors, terminology (library metaphor or no- metaphor/mosaic terms) and icon (library icons or no icons).

All subjects performed 3 blocks of 10 tasks per block. Task structure was identical across all three blocks, however the specific tasks were slightly different. Task order within block and block order were randomized. Three dependent measures were collected for each subject: task time, number of errors, and task completion. Task time was measured in seconds from the start of the task until the completion of the task. Subjects controlled the start of each task. The dependent measure, number of errors, was calculated by subtracting the minimum possible number of mouse clicks to successfully complete the task from the actual number of times the subject clicked the mouse button on an icon or label. Therefore, the more clicks per task the greater the number of errors. The task completion dependent variable was based on successful completion of the task. If a subject exceeded 25 clicks on a particular task, the task would be terminated and the subject would be allowed to move on to the next task. An experimental session lasted between 30 minutes and 1 hour.

Procedure

Subjects were asked to fill out a questionnaire which asked about their computer background. Subjects were then given written instructions about the computer tasks they would be performing. In the instructions, all subjects were told that they would be performing 30 tasks on the computer, and that it was to their advantage to try to learn as they performed the tasks. All subjects were shown a picture of what the computer interface would look like. Specific instructions about the tasks varied according to condition. Subjects in the two metaphor conditions were told that the tasks were based on a metaphor of a library, and that thinking about a library would help them to perform the tasks. Subjects in the two no-metaphor conditions were told that the basic task was to search for information.

Figure 1. Example task in condition with metaphor and icons.

Then subjects performed a series of thirty tasks on the computer. The tasks were presented in three blocks of ten tasks. Subjects read the task in an area on the computer screen and then tried to figure out how to perform the task. They were given auditory feedback for incorrect clicks. When subjects correctly performed the task they would be given auditory and visual feedback, and allowed to continue to the next task. If a subject exceeded 25 clicks on a given task, they were shown a message to that effect and allowed to continue to the next task.

For example, in the metaphor/icon condition, a subject might be given the task ‘See if there is a book about Virtual Reality.’ The subject sees the library metaphor terminology and icons on the computer screen (figure 1). To successfully complete the task the subject would click on the icon labeled ‘Reference Section’.

In contrast, in the no-metaphor/icon condition a subject might be given the task, ‘See if there is a document about Virtual Reality.’ The subject sees the mosaic terminology (no-metaphor) and icons on the computer screen (figure 2).


Figure 2. Example task in condition with no-metaphor and icons.

To complete the task, the subject would click on the icon labeled ‘Internet Resources Metalist’. The metaphor/no icon and no-metaphor/no icon conditions were similar to the metaphor/icon and no-metaphor/icon conditions respectively except the icons were replaced by gray squares.

Upon completion of the thirty tasks, subjects were asked to fill out a questionnaire concerning their preference for the interface and terminology. Subjects in the two metaphor conditions were also asked to rate how much the metaphor helped them perform the task.

RESULTS AND DISCUSSION

Performance Measures

In total, each subject completed 3 blocks of 10 tasks in one of four conditions -- metaphor/ icon; no-metaphor/ icon; metaphor/ no icon; no-metaphor/ no icon. A 10 (tasks) x 3 (blocks) x 2 (metaphor/ no-metaphor) x 2 (icon/ no icon) ANOVA for the number of errors in each task was performed. There was a significant main effect for metaphor F(1, 36) = 20.3. On average, subjects made fewer errors with the metaphor interface than with the no- metaphor interface (figure 3). Also, there was no significant main effect of icon.


Figure 3. The effects of metaphor and icon on number of errors.

There were significant main effects of block F(2, 72) = 60.7 and task F(9, 324) = 21.2. The maximum probability of Type I error for all results reported as significant in this experiment is 0.05 unless otherwise stated. There was also a significant metaphor by task interaction F(9, 324) = 4.1. The effect of metaphor was larger for some tasks than for others, but the metaphor average was always better than the no-metaphor. All subjects showed improvement over the blocks, showing strong practice effects (Block 1 mean - 6.6; Block 2 - 4.0; Block 3 - 2.6).

An analysis of variance on the task time dependent measure showed a similar pattern of results. Overall, subjects performed tasks faster in the metaphor conditions than in the no-metaphor conditions (an average of 34 s and 59 s per task respectively). Subjects again showed improvement over blocks as measured by faster task times (Block 1 mean - 80 s; Block 2 - 37 s; Block 3 - 23 s).

Task completion was measured as a binary value, 1 = success, and 0 = failure. Chi-square analyses were performed on each of the factors. The metaphor X2(1) = 43.2, task X2(9) = 184.5 and blocks X2(2) = 15.4 factors were significant with p < 0.01. The effects of metaphor on task completion also depended on the particular task.

There was no significant effect of icon on task performance in any of the dependent measures. This suggests that the inclusion of icons does not appear to improve performance in the metaphor conditions or worsen performance in the no-metaphor conditions.

All three dependent measures showed a similar pattern of results, suggesting that the use of a metaphor provides a large advantage in initial performance. Based on this data, the use of metaphor appears to facilitate performance. Furthermore, it appears that the effect of the metaphor is provided through the terminology (i.e., function labels) and not at all through the graphics (i.e., icons). It is surprising that the icons did not appear to have any effect at all as shown by the icon/no icon manipulation. Future research needs to be performed to better understand this lack of icon effect. Perhaps, the icons were not compelling enough or meaningful enough to aid in task performance.

The significant interaction between metaphor and tasks suggests that for some tasks the metaphor provides greater advantage than for other tasks. For example, in task 4 ‘Make it easy to get back to this place in the book/document later.’, there was a significant advantage in the metaphor conditions over the no-metaphor conditions for all three dependent measures. Subjects took significantly less time to complete the task, made fewer errors on this task, and there was a greater number of subjects who successfully completed the task. This particular task required the subject to click on the icon which was labeled ‘bookmark’ in the metaphor conditions (‘hotlist’ in the no-metaphor conditions).

Preference Measures

Subjects completed a post-session questionnaire in which they answered questions about their preference for the condition in which they participated. Subjects were asked to rate on a scale from 1 to 5, the ease of use of the interface, how much they liked the interface, and the helpfulness of the terminology. Subjects in the two metaphor conditions were also asked to rate on a scale of 1 to 6, the helpfulness of the metaphor. Analyses of variance were conducted on the answers to each question. On average, subjects perceived the metaphoric interface as easier to use than subjects using the non-metaphoric interface, (2.03 vs. 3.03 respectively), F(1, 36) = 9.99, p< 0.01. None of the other questions resulted in significant differences between the conditions.

Subjects were shown a picture of each of the four experimental conditions and were asked to rank order the four conditions from the condition they believed would be the easiest to participate in to the hardest. A Chi-square analysis of the seven unique rank orders subjects specified revealed significant differences with the most frequently specified order being 1(easiest)- metaphor, icon; 2- metaphor, no-icon; 3-no metaphor, icon; 4 (hardest)- no- metaphor, no-icon.

The preference measures demonstrate a pattern of results consistent with the performance data. Subjects perceived the metaphoric interface as significantly easier to use than the non-metaphoric interface, which mirrors the performance data. They did not perceive a difference due to the use of icons, again consistent with the performance data. When subjects were asked to rank order the conditions from the condition they believed would be the easiest to perform to the hardest, they most frequently ranked the metaphor, icon condition as easiest and the no- metaphor, no-icon condition as hardest. While these results are based on subject’s preference (perceptions) rather than their performance, it is encouraging to see a consistent picture developing. Subjects clearly perceive metaphoric interfaces as easier than non-metaphoric interfaces.

EXPERIMENT 2: COMPOSITE METAPHORS

Experiment 2 investigates the characteristics that contribute to the effectiveness of a UI metaphor -- specifically the integrality of the metaphor. Composite metaphors are common in user interfaces and frequently advocated by user interface designers. A composite metaphor is a combination of metaphors that are not necessarily related to each other but together represent the structure of the system. For example, a keyboard, processor, and diskette connected to a monitor is not easily compared to any one thing in a new user’s experience. It shares common elements with a blackboard, a typewriter, a television screen, a copying machine, and a tape recorder. Given that no one thing is like this computer device, a composite metaphor can be used to represent this device to a new user [3]. Sometimes, mismatched or incomplete correspondence between the source and the target domains in a metaphor comparison is addressed by composite metaphors [2].  

While composite metaphor may help to minimize the amount of mismatches between the base (metaphor) and target domain, perhaps we need to take a step back and question whether mismatches are so bad. Perhaps, it is better to preserve the overall structure of the target domain by designing the interface around a single metaphor, rather than a composite. Granted there may be more mismatches between the two domains, but the unified structure provided by an integral metaphor might outweigh the cost of mismatches.  

There appears to be some controversy amongst researchers and user interface designers as to whether it is better to use an integral metaphor or a composite metaphor. Carroll, Mack and Kellogg [2] suggest composite metaphors can be useful beyond increasing the coverage of a target domain, by helping the learner generate more and different kinds of inferences about the nature of the target domain. Gick and Holyoak [7] argue that analogical mappings from multiple source domains can help people more efficiently create a single description in which the metaphor mappings are integrated and abstracted into a more direct representation of the target domain. Rummelhart and Norman [11] also claim it is better to use many different conceptual models, each one simple, each making a different point than one overall model which does not fit perfectly. In contrast, Gentner might argue that it is better to design the UI metaphor around a single integral structure. In her structure mapping theory Gentner [4] proposes the systematicity principle which states that, a predicate that belongs to a mappable system of mutually interconnecting relationships is more likely to be imported into the target than is an isolated predicate [5]. This would favor the view that an integral metaphor is better than a composite metaphor.  

This experiment attempts to establish whether an integral metaphor is superior to a composite metaphor in terms of user’s performance. The controversy as to which type of metaphor is more effective and the fact that these theories are drawn from areas such as education, raises the question of which type of metaphor is most effective in user interface design. In the domain of user interface design, there is little known about the characteristics of an effective metaphor. This experiment could provide a clear direction for user interface design beyond the current generalization that metaphors are beneficial.

Pilot Study

In order to develop the necessary materials (i.e., the composite metaphor) for this experiment, a pilot study was conducted to enable development of a composite metaphor based on empirical data. An interface based on a travel metaphor was created as an alternative to the library metaphor. Each function in Mosaic was mapped to a concept within the domain of travel. For example one could conceptualize navigating through the Web as traveling from place to place in search of certain information. Thus, common travel accessories were mapped to the current Mosaic terminology. Performance measures on 2 integral metaphors (library and travel) were used to select the tasks to be combined in the composite metaphor.  

While the primary purpose of the pilot study was to develop a composite metaphor, the results of the study were interesting in their own right. The pilot study served to replicate the performance advantage shown by the library metaphor compared to the non-metaphoric interface. It also revealed that not all metaphors are beneficial, in that the travel metaphor did not show a performance advantage compared to it’s non-metaphoric control.

Subjects

Thirty New Mexico State University students participated and were compensated either by receiving $5.00 payment or partial fulfillment of an experimental credit requirement in introductory psychology. The same characteristics as in first experiment were used as criteria for participation.

Materials and Procedure

The same experimental paradigm as was used in Experiment 1 was used in this experiment. Table 2 shows the function labels used in each of the three conditions.  

Subjects were randomly assigned to one of the three conditions -- integral library metaphor, integral travel metaphor, composite metaphor (mix of library and travel).                            

Library

Travel

Composite

Library  

World

Library and World

Search bookshelves

Travel to a place

Search bookshelves

Stack of viewed books

Travel Log

Stack of viewed books

Bookmark

Shortcuts

Bookmark

Post-It notes

Message Pad

Message Pad

Current Book

Current Place

Current Screen

First

Go to Start

First

Next

Go Forward

Counter-balanced

Previous

Backtrack

Counter-balanced

Copier

Photograph

Photograph

Translator

Photo Album

Photo Album

Find text

Find in Current Place

Find in Current Place

Reference Section

TourBook

TourBook


Table 2 Function labels for each experimental condition.

Subjects in the integral library metaphor condition were told that the interface was based on a library metaphor. Subjects in the integral travel metaphor condition were told that the interface was based on a travel metaphor. Subjects in the composite metaphor condition were told that some parts of the interface were based on a library metaphor and that other parts were based on a travel metaphor. In the integral library metaphor condition, a subject might be given the task ‘See if there is some information about Virtual Reality.’ The subject sees the library metaphor function labels and icons on the computer screen (figure 1). To successfully complete the task the subject would click on the icon labeled ‘Reference Section’. In the integral travel metaphor condition a subject given the same task sees the travel metaphor function labels and travel icons on the computer screen (figure 4). To complete the task, the subject would click on the icon labeled ‘TourBook’.

In the composite metaphor condition the correct response for the same task would be identical to the integral condition on which it was based. In this case, the interface element was drawn from the travel metaphor (figure 5), and therefore the correct response would be ‘TourBook’, the same as in the integral travel condition.

Half the interface elements in the composite metaphor condition were identical to the interface elements in the integral library metaphor condition and the other half were identical to the elements in the integral travel metaphor condition.

RESULTS AND DISCUSSION

In total, each subject completed 3 blocks of 10 tasks in one of three conditions -- integral library metaphor (LM); integral travel metaphor (TM); composite of library and travel (CM).


Figure 4. Example task in integral travel metaphor condition.


Figure 5. Example task in composite metaphor condition.

Analyses of variance on the average time to complete a task (task time), average number of errors, and average rate of successful task completion were performed. The analyses on each measure were broken into two sub- analyses, one focusing on the 5 tasks that used the library interface elements in the composite condition as compared to the corresponding tasks in the integral library condition. The other analysis compared the 5 tasks that used the travel interface elements of the composite condition to their corresponding integral travel tasks. The reason for this approach to the analysis is to enable identical tasks to be compared to each other under two conditions, either in the context of an integral metaphor or in the context of a composite metaphor (combination of multiple metaphors).

Task Time Dependent Measure

The left side of figure 6 shows the mean task times in the three conditions for the five tasks using the library elements in the composite condition. An analysis of variance of mean task time for these tasks revealed a significant main effect of domain, F(2, 33) = 8.128 and blocks, F(2, 66) = 164.311. T-tests showed that the average task time was significantly longer for the composite library tasks (69.9 s) than the same tasks performed in the integral library condition (39.9 s). T-tests also showed the significant advantage of the integral library task times as compared to the analogous tasks in integral travel condition (85.2 s).


Figure 6. The effect of metaphor on mean task time for library tasks and travel tasks.

These findings show that the overall context of the UI metaphor matters. Specifically, metaphor elements in the context of an integral metaphor are easier to use than the same elements in the context of a composite metaphor. It is important to remember that exactly the same task is given in both the integral library condition and composite condition and should be performed in exactly the same way. Yet in the composite condition performance was worse than the identical task in the integral condition.

The right side of figure 6 shows the mean task times in the three conditions for the five tasks using the travel elements in the composite condition. An analysis of variance of mean task time for these tasks revealed a significant main effect of domain F(2, 33) = 6.664, blocks F(2, 66) = 83.777, and tasks F(4, 132) = 12.918. T-tests comparing the travel metaphor (57 s) to the composite metaphor (49.2 s) did not reveal a significant difference. Again there was a significant difference between the tasks in the integral travel condition and the analogous tasks in the integral library condition (20.4 s). The fact that performance in the composite metaphor did not show a difference from the travel metaphor suggests that it is not simply that integral metaphors are better than composite metaphors. It may be that the composite metaphor serves to dilute the beneficial effect of a good integral metaphor, but if the metaphor is weak to begin with as was evident from the results of the pilot study, it does not matter whether it is mixed with other metaphors (composite).

These results suggest that integral metaphors are better than composite metaphors when the metaphor is good, but does not matter when the metaphor is bad. Further research needs to be conducted to understand precisely why the integral metaphor is better than the composite metaphor for a good metaphor, but not so for a poor metaphor. A possible explanation for this finding could be that the performance on the integral travel metaphor had reached the boundary (floor effect) and could get no worse, resulting in no performance difference with the composite travel tasks. An alternative explanation is the mixing of the metaphors in the composite condition, resulted in the dilution of the good metaphor, thus making the composite significantly worse than the library, but the dilution of the poor metaphor (travel metaphor) resulted in minimizing the poor performance of the travel metaphor.

Task Completion and Error Dependent Measure

An analysis of variance for the average number of errors in each task, and chi-square analyses on the number of task completions revealed a similar pattern of results to the task time measure. The chi-square analysis showed a significantly lower rate of task completion of the library tasks in the composite metaphor condition compared to the same tasks in the library metaphor condition. This further supports the benefit of integral metaphors as compared to composite metaphors.

Preference Measures

Subjects also completed a similar post-session questionnaire as was administered in the previous experiment. Analyses of variance were conducted on the answers to each question. The analysis of the ease of use question and helpfulness of the metaphor question revealed significant metaphor main effects. On average, subjects rated the library metaphor as easiest to use and most helpful.

These results are consistent with the findings of the performance data, showing that people not only perform better with a good integral metaphor (i.e., library metaphor) but they also prefer them over a composite metaphor and a poor metaphor (i.e., travel metaphor).

In summary, both subject’s performance and preference showed that for a good metaphor, it is better to use an integral metaphor than a composite metaphor. However, for a poor metaphor it doesn’t matter. These findings appear to support Gentner’s theory that it is important to map the overall structure even at the expense of mismatches. Clearly, there may be cases when it is not possible to develop an interface around a single metaphor, and in this case it may be necessary to use a composite metaphor. However, in choosing a composite metaphor it may be beneficial to select various metaphors that are related in some way.

CONCLUSION

The results of these experiments begin to provide the much needed empirical data to validate the popular belief that UI metaphors can facilitate the use of software. However, it also cautions us that not all metaphors are good. Good metaphors facilitate performance, and poor metaphors are no more effective than non-metaphoric interfaces. Interestingly, the metaphor advantage appears to be carried in the language or terminology (i.e., function labels) in an icon-based system, not through the graphics. This is a somewhat surprising finding in that with the advent of the graphical user interface many people have put much emphasis on the graphics. It is not known if these results generalize to the latest variation on graphical user interfaces, the contextual (non-symbolic) graphical interface (i.e., General Magic’s Desktop or Microsoft’s Bob).  

Clearly the next question is what are the characteristics of a good metaphor. Experiment 2 has begun to uncover the qualities of a good metaphor by showing us that integral metaphors are superior to composite metaphors, but again we are left asking what makes the metaphor good. Perhaps, an essential characteristic of a good metaphor is it’s mapping to the target domain, or it’s similarity. Further research needs to address these issues and pin- down more precisely the fundamental qualities or characteristics of a good metaphor.  

Since the goal of this research was to provide some clear directions for the design of UI metaphors, it is appropriate to close with some recommendations. The following guidelines were developed based on this research and further the current understanding of how to design UI metaphors.  

  1. Designing an interface around a metaphor can provide substantial benefits in the user’s use of the software. Be wary, all metaphors are not created equally. Some metaphors may not be appropriate and therefore not provide any advantage.
  2. The function labels (terminology) carries much of the weight in conveying the interface metaphor to the user, and therefore should be carefully chosen. The icons while visually appealing and practical in terms of conserving screen real estate do not appear to effectively convey the metaphor.
  3. When possible, use an integral metaphor rather than a composite metaphor.

ACKNOWLEDGMENTS

I wish to thank Drs. Roger Schvaneveldt, Michael Darnell, Nancy Cooke, Douglas Gillan, James McDonald, and William Ogden for their continued help and support during this project. This work was supported in part by a grant from the Council of Higher Education, New Mexico State University.

REFERENCES

  1. Apple Computer, Inc. Human Interface Guidelines: The Apple Desktop Interface. Addison-Wesley Publishing: Reading, MA, 1987  
  2. Carroll, J.M., Mack, R.L., and Kellogg, W.A. Interface metaphors and user interface design, In Handbook of Human-Computer Interaction, Helander, M. (Editor), Elsevier Science Publishers B.V.: North-Holland, 67- 85, 1988.  
  3. Carroll, J.M. & Thomas, J.C. Metaphor and the cognitive representation of computing systems, IEEE Transactions on Systems, Man, and Cybernectics, 12(2), 107-116, 1982.  
  4. Gentner, D. The structure of analogical models in science. BBN Technical Report 4451, 1980.  
  5. Gentner, D. Structure mapping: A theoretical framework for analogy. Cognitive Science, 7, 155-170, 1985.  
  6. Gentner, D. and Gentner, D.R. Flowing Waters or Teeming Crowds: Mental Models of Electricity, In Mental Models, Gentner, D. and Stevens, A.L., Editors. 1983, Lawrence Album Associates: Hillsdale, NJ.  
  7. Gick, M. & Holyoak, K. Schema induction and analogical transfer. Cognitive Psychology, 15, 1-38, 1983.  
  8. IBM Common User Access Guidelines: Object-Oriented Interface Design, First Edition, Que Corporation Publishers: Carmel, IN, December 1992.  
  9. Jones, W.P. and Dumais, S.T. The spatial metaphor for user interfaces: Experimental tests of reference by location versus name. Association of Computing Machinery Transactions on Office Information Systems, 4(1), 42-63, 1986.  
  10. Mayer, R.E. Some conditions of meaningful learning for computer programming: Advance organizers and subject control of frame order. Journal of Educational Psychology, 68(2), 143-150, 1976.  
  11. Rummelhart, D. E. & Norman, D. A. Analogical processes in learning. In Cognitive Skills and Their Acquisition, Anderson, J.R. (Editor), Erlbaum: Hillsdale, New Jersey, 1981.