This study compares five input devices (mouse, mousepen, trackball, stylus, and touchscreen) on a series of goal-directed tasks using a drawing program. Striking performance differences are found for the touchscreen when compared with a previous study using a standard, isolated, laboratory task. The study also looks at the impact of device to screen mapping (absolute vs. relative) and device orientation (horizontal vs. vertical). Performance and preference data point towards an advantage for horizontal relative input devices.
KEYWORDS: input devices, input tasks.
Often times, the conclusions of input device studies are difficult to generalize. This is partly due to the fact that the studies usually test the chosen devices on a single task taken out of the context of goal-directed work. One way to overcome this problem is to test each device on a wider variety of tasks. Epps [2] conducted one of the few studies that included many tasks. Epps' study found that, on seven graphics editing tasks, the trackball and mouse were better than touchpads or joysticks on performance and preference measures.
The present study also employs multiple tasks and devices. The stimulus for this study is a continuous wave form that closely resembles the sine waves shown on oscilloscopes. For each device used, the subjects performed a variety of tasks: using menus and drawing tools, dragging guidelines, pointing at small targets, entering text and numbers and zooming. Another important feature of this study is that the tasks are performed in a goal-directed context of viewing and manipulating a waveform rather than the isolated trials typical of many lab experiments.
The devices used in this study were selected using two criteria. Four of the devices (mouse, mousepen, trackball, and touchscreen) were chosen based on their strong performance in a previous experiment [1]. Since the first experiment used a simple laboratory task, their inclusion in the current study will directly address the impact which the experimental task has on device performance. A fifth device (stylus with graphics tablet) was selected to permit the experiment to look at two important dimensions; (1) the mapping of the use of the device to the screen (absolute or relative) and (2) the physical orientation of the device (horizontal or vertical plane).
The touchscreen and stylus with graphics tablet are absolute mapping devices. For both of them there is a direct correspondence between positions on the device and the screen. The graphics tablet has the same aspect ratio as the screen and pointing to a location on the tablet jumps the cursor to a corresponding spot on the screen. This contrasts with the relative mapping of the mousepen, trackball and mouse. For all of these devices, moving the device causes the cursor to move relative to its current location.
The touchscreen is the only one of these devices which is commonly used in the vertical plane. All of the other devices are generally used in the desktop (horizontal plane). However, in order to balance out the experimental design and address some research questions of the corporate sponsor, the trackball was mounted in the vertical plane. The devices, thus implemented yield one example of each combination of mapping and orientation along with the mouse as the standard input device.
Nineteen computer literate college students (11 male, 8 female) served as paid subjects. All subjects had 20/20 corrected vision, at least two years of MacIntosh experience, and experience using drawing programs.
The experimental stimuli consisted of six continuous waveforms which closely resembled sine waves. The waves were black set on a white background. The six waves were created by using a MacIntosh drawing program. On each wave, there were two vertical guidelines and two horizontal guidelines that could be moved by the subjects. The waves themselves could not be moved. Each wave consisted of 16 peaks and 15 troughs. Subjects were only able to see 1/3 of waveform on the screen. In order to see more of the waveform they needed to scroll to the right.
Each waveform had six colored spikes which served as visual landmarks for the experimental tasks. The subjects could only see one of the colored spikes on the screen at a time. A written instruction sheet told the subjects what to do at each colored spike. The tasks included using menus and drawing tools, dragging guidelines, pointing at small targets, entering text and numbers, and zooming.
Each subject used all five input devices, and completed a single wave per device. Device and wave combinations were randomized, and the ordering was counterbalanced across the 19 subjects. Each subject came in on three successive days to complete the experiment.
The first day was used for training. Subjects were trained to use all of the input devices on a simple tracing task (see Cohen, Meyer, & Nilsen, 1993). They also went through one of the waveforms using the mouse. The instructions sheet was the same as for the real experiment, except in training, the experimenter went through the instructions one by one to make sure that the subjects understood.
The second and third days were used for data collection. Subjects used either two or three devices each day, completing one waveform with each device. At the end of day three, subjects were asked to complete a short questionnaire telling which device they preferred most and least. They were also asked to rate their perceived performance using each device on a seven point scale.
This short paper only reports the overall time for completing each waveform. The results for the individual tasks will be reported in future publications. A one-way repeated measures ANOVA indicated that there was a significant effect of device on total time (F 4,18=4.62, p<.005). The ordering of devices from best to worst were: mouse, mousepen, trackball, stylus, and touchscreen. The mean times and standard errors are shown in Figure 1. Post hoc analyses show that the mouse was significantly better than the trackball, stylus, and touchscreen. The mousepen was also significantly faster than the stylus and touchscreen ( all p's <.05). The mouse and mousepen were statistically equivalent.
When asked which device they liked the best, 14 subjects chose the mouse, 3 the mousepen, and 2 the stylus. 11 subjects chose the touchscreen, 6 the trackball, and 2 the mousepen for their least favorite device.
The questionnaire results for perceived performance indicate that there was a significant effect of device (F 4,18=16.87, p<.0001). The ordering of devices from best to worst were: mouse, mousepen, stylus, trackball, and touchscreen. The ordering of devices is identical to the performance data except for the reversal of the stylus and trackball. The trackball exhibits slightly better actual performance, but the stylus is rated higher on perceived performance. This observation is tempered by the fact that the differences between these two devices on both measures failed to reach statistical significance.
The finding that the mouse performed well on all of the imbedded drawing tasks is a surprise to no one. There are however two interesting findings from the study which bear comment and future research.
The horizontal placement of input devices was clearly favored over vertical placement. The most frequent comment in debriefing was the lack of accuracy and fatigue associated with using the touchscreen and the vertically mounted trackball. This was mirrored by the fact that these two devices also received the lowest subjective ratings for performance. Looking at the pattern of the performance data also suggests that the relative mapping is superior to absolute mapping. The touchscreen and stylus, the two devices with absolute mapping, were the slowest devices in the study.
When evaluating input devices, researchers should consider the wide range of tasks that will be performed with the device. This task dependency on the performance of the devices is best illustrated by the performance difference of the touchscreen in a prior study [1] and the current study. It moved from fast to last! For continuous tracing it was the best device while for tasks involving precise goal-directed selection it was the worst device. The performance of input devices is strongly dependent on the task characteristics.
1. Cohen, O., Meyer, S., and Nilsen, E. (1993) Studying the movement of high tech rodentia: Pointing and dragging. Adjunct Proceedings of INTERCHI '93. ACM: New York, 135-136.
2. Epps, B. (1987) A comparison of cursor control devices on a graphics editing task. Proceedings of the Human Factors Society, 31st Annual Meeting, 442-446.