Paper presented at "Interface 89" The Sixth Symposium on Human Factors and Industrial Design in Consumer Products Human Factors Society POBox 1369 Santa Monica California 90406 Microsoft Mouse: Testing for Redesign Bill Verplank, Kate Oliver, IDTwo

As part of the redesign of the Microsoft mouse by Matrix Product Design, a series of user tests were performed by ID Two. We used artificial tasks representative of typical mouse use allowing repeated measures of time and error.

The initial questions were about the impact on performance and preference of different mouse shapes, asymmetric buttons and moving the ball from the back to the front.

The first tests compared four designs on four tasks: maze tracing, Fitts' tapping, precision selection, signature writing. Significant differences on the maze task showed an advantage for the mouse with the ball in front.

Each test led to design changes. Most tests led to other tests either because of measured performance or further design insights.

The first tests raised questions about people actually hold mice. The second test, a two-day photo survey of over 60 users, showed users holding the mouse with their fingertips not palms. Next, a test of three different switch types showed preference for a sharp click. Tests of left-handed users showed no disadvantage with asymmetric buttons but revealed errors is accidentally pressing both buttons. Finally, adding a ridge between the buttons was shown to eliminate this user error.

Key innovations either discovered or confirmed by the testing included: fitting the mouse to finger tips not palm, moving the ball from back to front for better feel and performance, using asymmetric buttons as a more obvious coding for frequency of use, and adding a ridge between the buttons to prevent accidentally pressing both.

The testing was aimed at providing the designers with timely quantitative results. In addition, Microsoft did more extensive and subjective "environmental" tests b y giving users a new mouse to use for a week and asking for opinions. The testing described here consisted of "abstract" tasks with repeated measures of time and error. They were typically one- or two-hour tests conducted in the studio.

Throughout the four-month design process we were able to do five quick tests. Because the design effort was basically a repackaging, we had the advantage of using working prototypes which were easily assembled around the existing mechanisms.

This first test compared four prototype mice with the existing Microsoft mouse. Twelve subjects were divided into two groups, each comparing the existing mouse (A) to two new mice (BC or DE). The prototypes had differences in placement of the ball, overall size and shape and button position and shape.



The twelve subjects included seven frequent and five infrequent mouse users. Half of the subjects were under 30 years old and the other half were over 35 years old. We included a range of users from student to office worker and from occasional to dedicated PC user.

The four tasks can be thought of a "stress tests". They included a Fitts' tapping task,* a maze, a precision positioning task, and a signature task. They were all composed on an IBM PC using Microsoft DOS version 3.1 and the Z-Soft Microsoft paint program.



Tapping Task. The task was to tap twenty times back and forth between a pair of targets clicking the mouse button, making a mark inside each target. The goal was to move a quickly as possible while making only one or two errors. Timing was done with a stop-watch. After each twenty taps, the subject was told to speed or slow down to keep the error rate consistent. Of all the tasks, tapping is most like common mouse tasks for menu and text selection.



Maze Task. The subject was asked to hold down the mouse button whil e tracing a mizze that included horizontal, circular and diagonal paths.

Touching the maze wall was counted as an error. Timing and subject feedback were the same as in the tapping tgask. The maze task is a test of continuous cmouse control and provide most sensitive to the ball positionoing.



Pixel Task. Given a three-pixel-high brush, the subject was asked to position the mouse within one-pixel precision, button-down and 'up without moving the mouse. We expected that we might see a disadvantage if pressing or releasing the mouse button tended to move the mouse.



Signature Task. The task was to sign one's legal signature in a given rectangle. This was an unlikely task for PC users but gave subjects a feeling for the "naturalness" of mouse use.

Procedure. Each subject filled out a non-disclosure form, answered background questions, then played with the first mouse while hand position was photographed. They performed each task three times in the order: tapping, maze, pixel, signature the tapping again. The subject repeated this same routine with the second and third mice.
The order of presentation was different for each of the six subjects in each group. After testing, the subject answered a questionnaire covering aesthetics, button preferences and comfort.



Results. The large prototype mouse with the ball at the front performed significantly better than the others in the maze task. This result validated the repositioning of the ball. The effect is to put the ball under the finger-tips, increasing the apparent moment of inertia as the mouse is pivoted from elbow or heel of the hand.

Some subjects reported cramping and weren't sure how to grasp the small mice. We captured in photographs examples of some very awkward attempts to wrap fingers around them.

There were many problems with the buttons. It was clear form observing errors that the flush-mounted buttons were too small. Some subject testing the larger mice (notably those with medium to small hands) like the mouse with the thinnest width.

We decided from these results that it would be useful to know more about how experienced users actually hold mice and if a tow-button mouse should be designed for one-finger or two-finger button use.

To gather more information about actual mouse usage, we conducted a tow-day photo survey of one-tow-and three-button mice. We observed and photographed over 60 mouse users in their normal working situations.



Most users favored a fingertip grip; there were vocal complaints about mice that didn't allow it. We also concluded that we should try to design the mouse to accommodate both one-finger and two-finger button use.

Microsoft had changed the switch mechanism on the second version of their mouse. User had complained of "mushiness". Switches with a sharp click were proposed as a replacement. It was not clear whether they would alter performance and if user would indeed prefer the crisper tactile feedback despite the louder sound.



A test with six subjects and three kinds of switch was designed. Three Microsoft mice were fitted different switches: the original Microsoft switches, the current "mushy" switches and the proposed switches. One of the original tasks and two new tasks, on involving double-clicking, were used. Opinions were collected in a questionnaire.

We found no significant performance differences. Subjects preferred the proposed switches with the loud but sharp click.

At this point, our preferred design had asymmetric buttons. This asymmetry raised the question of possible disadvantages for left-handers. Two mice were prepared, one with asymmetric buttons, one with symmetric buttons. Ten left-handed people tested the two mice on three screen tasks using their left hand. We added one task that required alternately clicking th4e left, then the right button.

Results showed no significant performance differences between symmetric and asymmetric mice. Our left-handed subjects did not consider the symmetric mouse as biased against them nor did they feel that the buttons presented any more noticeable stretch or discomfort.



A surprising result on the task that required alternating between buttons was a high rate of accidentally hitting both buttons at once. It was clear form the tests that some tactile differentiation or separation was needed. The design was modified by placing a slight ridge on the larger button.

The last test compared the final design including the ridge with the existing Microsoft mouse.



Four tasks were used: maze, tapping, signature and a two-button task. Ten people were tested. Each subject used both mice. The result was the errors were reduced by 90% showing that the ridge served it s purpose. The new mouse performed better or at least a well as the old mouse in all the tests.

Overall, the new mouse is an improvement over the original Microsoft mouse in performance and button feedback. It has subjective appeal - the subjects said it was more comfortable and more aesthetically pleasing. The product has met with a remarkable degree of customer acceptance. We were able to incorporate human factors at every stage of the design process. Our testing identified problems and pointed to solutions. We confirmed the advantages of moving the ball to the front, the acceptability of asymmetric buttons and the tactile differentiation provided by the ridge.

Finally, we recommend these quick, performance-oriented, quantitative "stress tests" as a useful adjunct to the more extensive and subjective "environmental" tests, and as a integral part of the comprehensive design process led by the industrial designers' good sense and sensitivity.

ACKNOWLEDGEMENTS

The mouse redesign was mansaged by Michael Cooper of Microsoft, the industrial design was by Mike Nuttall and Paul Bradley of Matrix Product Desgn, 660 High Street, Palo Alto, Califonia 94301.