Who's Zooming Whom?
Attunement to Animation in the Interface

Michael Chui
Computer Science Department and Cognitive Science Program
Indiana University
Lindley Hall 215
Bloomington, Indiana 47405
mchui@cs.indiana.edu

Andrew Dillon
School of Library and Information Science
Indiana University
Bloomington, Indiana 47405
adillon@indiana.edu

ABSTRACT

INTRODUCTION

In attempting to build more usable cognitive artifacts, much research has attempted to apply an information-processing theory of cognition to the analysis of human-computer interaction (HCI). For example, the Goals, Operators, Methods, and Selection Rules (GOMS) model of Card, Moran, and Newell (1983) is a well-known formalization of the production rule approach to modeling cognition. While this approach has achieved modest success, it has been argued that a strong theory of HCI cannot be drawn from this approach (see Landauer, 1995).

While the precise role of cognitive theory in design practice remains open to interpretation, it is possible that insights can be gained from psychological theories outside of the information processing tradition. The ecological, or Gibsonian, approach to perception/action research provides a complementary, or perhaps alternative, framework upon which we can begin to conceptualize process of interaction. The ecological approach emphasizes the shaping forces of environments in influencing cognition, and rather than seeking to establish the parameters of information processing inherent in cognitive architecture, it seeks to understand the problems that must be solved by a cognitive system and the resources that exist to aid solving.

One output from this line of research is the view that temporally extended events in the environment are fundamental sources of stimulus for the perceptual system. Living organisms become attuned to various events that are specified by trajectories through time. This view is distinguished from the stereotypical information-processing base case of perception being inference about the world, derived from a series of punctate, perceptual "snapshots." The information-processing framework begins from the assumption that interpreting a static pattern of energy (e.g. while reading) is the fundamental case for perception, rather than a continuously changing flux (e.g. while walking). In the ecological approach, the perception of temporally-extended events is fundamental.

The ecological approach has been applied only sparingly in the HCI literature to date. Rasmussen developed a conceptual tool, called a means-end or abstraction hierarchy, that has been used to describe the functional structure of complex work domains (Vicente & Rasmussen 1990). Abstraction hierarchies have been used to describe the hierarchical structure of affordances, the Gibsonian notion of perceivable properties of the environment that are relevant to a particular organism's goals. From means- end hierarchies of affordances, Vicente and Rasmussen developed a framework for interface design called Ecological Interface Design (Vicente & Rasmussen 1992). However, much of this work has been applied in the domain of real-time control systems, upon which the notions of perceiving affordances and exploiting them in the environment can be mapped fairly directly.

Another possible application of the ecological perspective is the investigation of designs in which temporally-extended displays, i.e. animations, are informative components of an interface that is not used for real-time control. In the studies detailed below, we examine the effect of the presence and absence of the zooming effect displayed when a folder is opened or closed in the Macintosh user interface. This effect carries the information that a particular window is associated with a particular folder on the Macintosh desktop. In some other windowing systems, the default behavior, when an icon is opened into a window or a window is closed into an icon, does not include such an effect. From an ecological perspective, such temporal-extensions might aid the user at a perceptual (visual cueing) level, and at a cognitive (memory enhancement) level.

There are a number of references to the benefits of this particular effect in the HCI literature. For example, Baecker and Small (1990) state:

"The outline zoom that accompanies the opening (and closing) of an icon orients the user to the location and origin of the new window that appears on the desktop. This is particularly helpful in a crowded environment. If the new window were to appear without the opening zoom, it would be more difficult for the user to determine that he had indeed opened the correct icon. The closing zoom assists in informing the user where he was working before he started the process that has just been completed. This type of animation is already in widespread use on the Macintosh, perhaps because of its relative ease of implementation and uncontroversial nature." (Baecker and Small, 1990, p.259)

May and Barnard (1995) assert: "Salient information or objects should not just appear or disappear from the screen," and cite as a rare commercial application of this phenomenon the Macintosh Finder, where windows 'zoom' in and out of their parent folder or application icons. Chang and Ungar (1993) make reference to the absence of such an animated effect accompanying the opening of a window, stating that in some windowing systems, "Much of the screen changes suddenly and without indication of the relationship between the old state and the new state."

However, despite these assertions of the enhanced usability brought by the inclusion of this feature in the interface, and its continued inclusion (indeed, elaboration, in Apple's new Copland interface) in the Macintosh interface, there is a paucity of empirical studies investigating these assertions. The following studies explore this territory. It was hypothesized that enabling the zooming effect will increase the subjects' memory association between folders and windows, and/or the speed with which they accomplish tasks that require knowledge of this association. In so saying, these experiments are tests of a common-sense interpretation of the value of temporal extensions in interface design (a necessarily conservative approach that seems justified given the history of failure associated with other common-sense ideas of good interface design).

Experiment 1.

In the first instance, subjects were required to perform a search task involving opening and closing of multiple windows seeking a target match.

METHOD

Subjects

Ten graduate students (5 female, 5 male), from Indiana University and Georgetown University, volunteered to participate in this study. Five of the subjects were regular (daily) users of Apple Macintosh computers, but all ten subjects were competent users of the Macintosh user interface, i.e. all knew how to use a mouse to open and close folders.

Materials

Nine different displays were created on a Macintosh "desktop." Each display was similar, consisting of six folders arrayed along the bottom of the screen, and six folders arrayed along the right side of the screen. This configuration was chosen because informal observations of typical Macintosh desktops showed that users would commonly place often-used folders in those positions. Folders were numbered from 1 to 12, beginning in the bottom-left corner, and ending in the top-right corner.

Figure 1.

Each of the twelve folders opened into a window, displayed in a random location on the screen, containing a single file. Nine categories of words were used to generate names for the files: months, fonts, Macintosh programs, Macintosh control panels, computer companies, male names, female names, academic disciplines, and sports. All of the files in a given display were given names from the same category. Two of the files, in each display, were given identical filenames. These matched pairs were selected at random, with the only constraint being that they were not placed in adjacent folders.

These displays were presented on an Apple Macintosh IIci, with a 21-inch Radius color screen. Using a utility called 7Tuner1.7, two copies of the System 7.1 Finder and System files were created - one with the zooming effect enabled, the other with the zooming effect disabled.

Design and Procedure

After a practice trial, subjects performed the task on the eight remaining displays grouped into four blocks of two with order of display counterbalanced across subjects.

Subjects were seated individually in front of the computer, and shown the practice display. It was explained that each folder only contained a single file, and that the subject's task was to discover the two folders containing files with identical names. The subjects were instructed that they were to open only one folder at a time, i.e. they had to close each folder's window before opening the next folder. However, they were instructed to open both of the folders that contained files with identical names when they had found the match. Subjects were informed that their performance would be timed, and that they would be asked to recall the folder in which specific files were located, but that their primary goal was to locate the match as quickly as possible.

Before each trial, subjects were told from what category the filenames were generated, e.g. "months of the year". Timing of the task, by stopwatch, began when a subject clicked on a folder, and ended when the two folders with matching files were both visible on the screen. A record of every folder the subject opened and closed while searching for a match was also kept. Finally, after each trial, the subject was presented with a sheet of paper, listing the names of all of the files in the display, and was instructed to write the numbers of the folders in which the files could be found, if they could recall them.

RESULTS

Speed

Mean time to open and close a folder in the zooming condition was greater (mean = 3.96 seconds, SD = 0.67) than in the no-zooming condition (mean = 3.87 seconds, SD = 0.6), as predicted. However, the difference does not reach statistical significance (F(1) = 0.77, p = 0.388). [1]

Number of folders required to find match

Using the record of folders visited, the number of folders the subject visited after they have seen the matching files was calculated. If the subject had perfect memory for location, then the minimum number of folders required to find the matching folder would have been one (i.e., when the first matched (second seen) item was noted, the subject would then select the folder containing the first seen item of the pair). With less than perfect memory, the subject would need to search for the matching target, and the number of folders required to find a match would increase. The results are shown in Table 1.

Table 1. Number of folders required to find match

The mean number of folders required to find a match was greater in the non-zooming condition than in the zooming condition. However, this difference was not statistically significant (F(1) = 0.24, p = 0.626).

Distance to first response

As a measure of the accuracy of memory for location, we calculated the distance in ordinal location from the first folder opened after the subject had seen a match, to the folder which actually contained the matching file. Hence, the minimum value of this statistic, given that a subject has perfect memory for location, is zero. If the subject selected a folder adjacent to the correct folder, than this statistic will be scored a one. It should be noted that this is not a direct spatial measure but an ordinal one.

Table 2. Ordinal distance to first response

The difference is not in the hypothesized direction, but it is not significant (F(1) = 0.32, p = 0.575).

Post-trial memory

Users were given a sheet with the names of all the files, and were asked to write the numbers of the folder in which the files were located. We calculated the ordinal distance (as above) from reported location to the actual location of the folder visited immediately after the subject visited the first of the matching pair of folders.

Table 3. Ordinal distance error from first folder visited after match

Subjects more accurately reported the location of the folder when zooming was enabled. Although not significant statistically (F(1) = 1.51, p = 0.229), the difference is in the hypothesized direction, indicating that subjects might have more accurate memory for location in the zooming condition.

While the results from this experiment suggest that there is a limited advantage afforded by the zooming effect, these findings are statistically not significant. Clearly, the simple provision of the zooming effect is not a straightforward benefit to users as seems to have been assumed in the literature. From a design perspective, the general preference for zooming suggests it may be worth keeping for most users but at a theoretical level, the reasons for retaining or advocating its use are not so clear.

In post-task interviews, all subjects reported that they had begun rehearsing either the names of the various files, or the first letter in each filename, as they proceeded through a trial. Subjects also reported that when they were required to memorize more than about seven filenames, their ability to rehearse the entire list of filenames was markedly decreased, and then they relied on some general sense of memory for location. Subjects reported that the salience of various name categories varied, though they weren't sure what effect that had on their memory performance.

Interestingly, none of the subjects reported having noticed the zooming windows effect appearing and disappearing. Even one subject who had been told a week before what the authors were researching did not report that they recognized the manipulated variable, until they were reminded. This would seem to indicate that whatever effect the presence or absence of the zooming event has on cognition is at least subconscious at this speed.

The use of numbered folders enabled subjects to use numbers as cues for memorization. It became clear that subjects were employing a variety of mnemonic techniques in order to improve their probability of succeeding at the task. The variability in subjects' ability to use these techniques could have swamped the lower-level perceptual effect we were attempting to study. In fact, much of the variability in the samples was accounted for by individual differences.

In order to pursue this issue further, a revised methodology was developed and further data collected. In the following experimental design, a less cognitively complex task, involving a higher rate of presentation, was employed to prevent the effective deployment of higher-level memorization strategies. If zooming effects were likely to have any effect on user cognition, such a task was more likely to expose them.

Experiment 2.

Subjects

A total of twenty-six students at Indiana University volunteered to participate in this study, thirteen male and thirteen female.

Materials

Forty different displays were created on a Macintosh desktop. Each display was similar, consisting of four folders arrayed along the right side of the screen. This configuration was again chosen because the default location for new folders created on the Macintosh desktop is along the right side of the screen. The names of the folders were concealed by using naming them with varying numbers of blank spaces, and using a white background.

Figure 2

Each of the four folders in a display opened into a window, displayed in one of the four equally-sized quadrants of the screen not used to display the four folders. These displays were presented on an Apple Centris 610, with a 12-inch color screen. As before, two versions of the experimental folders were created.

Design

Each subject was tested on all forty displays. The displays were grouped into four blocks of ten, alternating between 'zooming enabled' and 'zooming disabled'. Order of presentation was counterbalanced across all subjects.

Procedure

Subjects were seated individually in front of the computer. For the first two blocks of ten displays, the four folders in each display were opened in random order. The experimenter then pointed to one of the opened windows and asked the subject to which folder that window corresponded.

For the second two blocks, the trials began with the four folders already opened into four windows, one in each quadrant. The four windows were closed in random order. The experimenter then pointed at one of the folders, and asked the subject in which quadrant the corresponding window had been displayed.

Results

Table 4. Accuracy matching folders with windows

Subjects performed both tasks with greater accuracy, and less variance, when the zooming effect was enabled. However, as before, these differences are not statistically significant.

In order to explore these differences further, several post-hoc analyses were performed. Subjects could be distinguished in terms of their experience with Macintosh computers (half of the subjects reported that 50% or more of their uses of personal computers were of a Macintosh). Therefore, incorporating Macintosh experience as a factor operationalized in this manner enabled a three-way ANOVA (zoom x task x user) to be performed. The results indicated that the subject's experience with Macintosh systems is a significant factor in performance.

Table 5. ANOVA

The interaction between type of user and zooming effect was highly significant, accounting for 48% of the variance in the sample.

Figure 3.

Table 6. Number Correct in Opening Task

Figure 4.

Table 7. Number Correct in Closing Task

While the accuracy of the Macintosh users on both tasks appeared to have been increased by the presence of the zooming effect, analysis revealed that the only statistically significant difference between the zooming and non- zooming conditions occurred with Macintosh users on the opening task (one-tailed t = 3.397, df = 24, p = 0.001).

Qualitative data

All subjects reported that the second "closing" task was more difficult than the "opening" task. The display time required to close four windows was less than the time required to open four windows. Furthermore, 23 out of the 26 subjects (88%) reported preferring the presence of the zooming effect, rather than its absence.

Subjects reported using a number of different strategies in order to improve their odds of successfully completing each trial. These strategies included numbering either the folders or the quadrants and attempting to memorize the order in which they were involved in the display; attempting to memorize the spatial sequence in which the folders and windows were involved in the display; and focusing one's attention on the spatially more compact group of folders, while using peripheral vision to observe the location of the relatively larger windows. In the reportedly more difficult closing task, some subjects abandoned all hope of correctly memorizing the entire sequence of events, and simply focused their attention on specific parts of the display (e.g. the first and last folders closed, or a specific folder), hoping that they would be queried first about one of those parts.

DISCUSSION

Although the data did not indicate a significant overall performance advantage for the zooming effect, our preliminary findings suggest a highly significant interaction between the zooming factor and the type of user, with respect to computer use. While we intend to conduct further analysis of this effect with experience on zooming interfaces manipulated a priori, only the performance of regular users of the Macintosh interface seems to have been enhanced by the presence of the zooming effect in the current experiments.

In the ecological approach to perception, organisms become attuned to various temporally-extended patterns that specify information. We can understand the differential effect of the zooming factor on Macintosh users as a type of temporally-extended, information-bearing pattern to which they have become attuned through repeated exposure. The discovery of such attunements is characteristic of the ecological approach to psychology, which emphasizes the interaction between the organism and the environment, rather than just the abilities of the organism (Vicente 1995). While our results do not directly resolve any controversy between the ecological and information-processing approaches to perception research, they do illustrate the value of the ecological perspective in suggesting empirical examination of a neglected aspect of the interface.

It should noted that we did not attempt to operationalize the possible affective benefits of the zooming effect. It is possible that the inclusion of such animations in the interface could lend some reassurance to users that an event had occurred, resulting in a higher level of user satisfaction, even if it did not result in an advantage by performance measures.

In terms of interface design, it is clear that apparently advantageous features might not be enhancing performance as expected. Indeed, there is some suggestion in the present findings that the zooming effect might lead to improvements or disimprovements as a function of user experience. Interface designers should adopt a cautious attitude to 'common-sense' views of good design where there is little evidence to support their view. Researchers and theorists of HCI need to develop a more critical attitude to claims for the cognitive compatibility of animation in user interfaces. Further research could usefully explore the relationship between animation and experience.

REFERENCES

Baecker R. & Small, I. 1990. Animation at the Interface. In The Art of Human-Computer Interface Design. Edited by B. Laurel. Reading, MA: Addison-Wesley. pp. 251-267.

Card, S.K., Moran, T. P., & Newell, A. 1983. The psychology of human-computer interaction. Hillsdale, NJ: Lawrence Erlbaum.

Chang, B. & Ungar, D. 1993. Animation: From Cartoons to the User Interface. In UIST '93: User Interface Software and Technology. Atlanta, GA, November 3-5, 1993. pp. 45-55.

Gibson, J. J. 1986. The Ecological Approach to Perception. Hillsdale, NJ: Lawrence Earlbaum.

Landauer, T. K. 1995. The Trouble with Computers: Usefulness, Usability and Productivity. Cambridge, MA: MIT Press.

May J. and Barnard P. 1995. Cinematography and interface design. In K. Nordby, P.H. Helmersen, D.J. Gilmore and S.A. Arnesen (eds) Human Computer Interaction: Interact '95. Chapman and Hall: London. pp. 26-31.

Vicente, K. J. 1995. A Few Implications of an Ecological Approach to Human Factors. In Global Perspectives on the Ecology of Human-Machine Systems. Volume I. Edited by Flach, J., Hancock, P., Caird, J., and Vicente, K. Hillsdale, NJ: Lawrence Earlbaum Associates. pp.54-67.

Vicente, K. J. and Rasmussen, J. 1992. Ecological Interface Design: Theoretical Foundations. IEEE Transactions on Systems, Man, and Cybernetics. SMC-22, pp. 589-606.

Vicente, K. J. and Rasmussen, J. 1990. The ecology of human-machine systems II: Mediating "direct perception" in complex work domains. Ecological Psychology, 2, pp. 207-250.

FOOTNOTES