User Studies with the Touch-Sensitive Mouse Device

4.4.1

First User Study – Testing the Concept

The goal of the first user study was to compare the touch-sensitive mouse and gaze positioning against the classical pointing with a standard mouse device. The program written for the user study sets the target, a big red circle, and the mouse pointer at random positions and records the time until the first mouse click into the target. A pilot study did not show the speed benefit hoped for although subjectively it felt faster with the touch-sensitive mouse. The reason for the good performance of the classical mouse lies in the presentation on an empty background. At the beginning of the task there are only two changes on the display at the positions of the target and the mouse pointer. Both can be spotted with pre-attentive perception. To simulate the common situation, where the user is not aware about the position of the mouse pointer, the program for the user study was extended to provide a second mode – target and mouse pointer are displayed together with a background picture (see Figure 47). The pilot study now showed a speed benefit because the users started to stir the mouse to locate the mouse pointer position.

Figure 47: Displaying target and mouse pointer without and with background.

Ten participants in the age from 23 to 46 years, nine male and one female, all from European countries and regular computer users, participated in the user study. Every person had to do four runs with 50 clicks per tasks. The size of the target was 100 pixels (≈3°) in order to avoid accuracy problems and calibration shift. See 2.6.2 for the geometry of the setup.

Every user got a one-minute explanation and one minute to try the system. The first run was on white background using a classical mouse. In the second run the participants had to use the touch sensitive mouse. After this the users were asked with which input method they think to be faster and whether it is convenient not having to search the mouse pointer. The next two runs were a repetition of the first two runs, but this time with the background picture displayed. Again, the participants were asked both questions.

Normally, a user study should use a random order for the tasks to average possible learning effects and indeed the data show effects of learning. The reason for a fixed order lies in the observation made in the pilot study. Gaze positioning on a blank background felt faster than the classical mouse positioning although the measured

times were equal. Because on complex background gaze positioning is faster the participants had to answer their subjective speed impression on a blank background first without bias from the complex background task. Table 9 summarized the results of the study. The medians are chosen to eliminate outliers due to irritations of the participant during the task. A typical reason for outliers was confusion in the sequence of movements – lift the finger, look, touch the mouse key, click the mouse key. Calculations with the arithmetic means do not change the qualitative results. The mean of the distance to the target centre was about 400 pixels.

Table 9: Medians for total time in milliseconds of all participants

Median for total times (in milliseconds)

blank background with background

Participant mouse positioning gaze positioning mouse positioning gaze positioning

P1 1097.0 751.0 1382.0 826.5 P2 971.5 1007.0 1066.0 806.0 P3 1121.0 716.0 1276.5 791.0 P4 1096.5 1066.5 1352.0 896.5 P5 932.0 871.0 1191.0 696.0 P6 926.5 1256.5 1066.0 1121.0 P7 1111.0 957.0 1217.0 891.0 P8 1211.5 1327.0 1412.0 1327.0 P9 1062.0 1482.0 1266.5 1277.0 P10 976.0 881.0 1101.0 656.0 Mean 1051 1032 1233 929 Std. Dev 94 253 128 234

Table 10: T-tests for four different task combinations (p-values)

t-test

gaze vs. mouse pointing with background vs. blank background blank background with background gaze positioning mouse positioning

0.823686 0.002036 0.020738 0.000014

Table 10 shows the results of paired Student’s t-test on the medians from Table 9 for four different combinations, which means the probabilities that the compared data sets are from a distribution with the same mean. There is strong significance that gaze positioning is faster than classical mouse positioning when using a background i.e. in the case the user is not aware of the mouse pointer position. The values also show that classical mouse positioning takes longer on a complex background. The effect of (significant) better performance for gaze positioning on a background compared to no background is most probably an effect of learning, as the order of the runs was not randomized. For a strict statistical evaluation it is necessary to show that the data obey the preconditions (Gaussian distribution) for the validity of the t-test. However, this is not worth the effort as the user study was designed with a poor understanding of the situation. The analysis, presented in 4.5, will give a

deeper insight which does not need statistical values as a proof and does not even need a user study. Nevertheless, this user study was an important step on the way of understanding.

In the interviews after the blank and complex background task all but two users stated to be faster with the touch sensitive device under both conditions (see Table 11). However the measured times showed that this was not objectively true in the case of the blank background (see Table 9). One of the two users who did not perceive gaze positioning as faster insisted on equal performance time. This answer was not valid within the design of the study and finally his or her answer was counted in favour for mouse positioning.

Table 11: Judgments of speed after the tasks with blank and complex background.

Which method is faster? Mouse Positioning Gaze Positioning

After blank background 2 8

After complex background 0 10

The timeline diagram in Figure 48 gives a potential explanation. It shows typical timings for the condition with a white background. Even though the overall time needed is the same, the time users move the hand or finger is shorter when using gaze. This suggests that the time required moving the eye is not perceived as time needed to perform the task.

Figure 48: Timing analysis that explains the perceived speed-up with a touch mouse.

Time

Time