Enabling devices, empowering people: The design and evaluation of Trackball EdgeWrite

Abstract

Purpose. To describe the research and development that led to Trackball EdgeWrite, a gestural text entry method that improves desktop input for some people with motor impairments. To compare the character-level version of this technique with a new word-level version. Further, to compare the technique with competitor techniques that use on-screen keyboards.

Method. A rapid and iterative design-and-test approach was used to generate working prototypes and elicit quantitative and qualitative feedback from a veteran trackball user. In addition, theoretical modelling based on the Steering law was used to compare competing designs.

Results. One result is a refined software artifact, Trackball EdgeWrite, which represents the outcome of this investigation. A theoretical result shows the speed benefit of word-level stroking compared to character-level stroking, which resulted in a 45.0% improvement. Empirical results of a trackball user with a spinal cord injury indicate a peak performance of 8.25 wpm with the character-level version of Trackball EdgeWrite and 12.09 wpm with the word-level version, a 46.5% improvement. Log file analysis of extended real-world text entry shows stroke savings of 43.9% with the word-level version. Both versions of Trackball EdgeWrite were better than on-screen keyboards, particularly regarding user preferences. Follow-up correspondence shows that the veteran trackball user with a spinal cord injury still uses Trackball EdgeWrite on a daily basis 2 years after his initial exposure to the software.

Conclusions. Trackball EdgeWrite is a successful new method for desktop text entry and may have further implications for able-bodied users of mobile technologies. Theoretical modelling is useful in combination with empirical testing to explore design alternatives. Single-user lab and field studies can be useful for driving a rapid iterative cycle of innovation and development.

Keywords:

• Text entry
• text input
• gestures
• unistrokes
• area pointing
• goal crossing
• word prediction
• word completion
• word-level stroking

Notes

1. ‘Clutching’ occurs when a user reaches the limit of allowable motion and must lift a device or finger to regain freedom to move in a particular direction. With a mouse, this happens when the mouse reaches the edge of the mouse pad or table. With a touchpad or trackball, it happens when the user's finger or thumb can no longer move in a given direction and therefore must be lifted.

2. This would turn out to work quite poorly, but the investigation was illuminating and so it is described here.

3. For example, ‘ad’ is a prefix for words ‘added’, ‘addition’, ‘administration’ and ‘additional’, all of which are more common than ‘ad’.

4. During the test, Jim transcribes simple English phrases presented to him in both methods. Uncorrected errors are those that Jim leaves in his transcribed strings. Corrected errors are those that Jim initially makes but fixes (using backspace) during entry. See Soukoreff and McKenzie [47] for more details.

5. Total error rates are simply the sum of the uncorrected and corrected error rates.