RESNA 26th International Annual Confence
A Method for the Assessment of Alternative Keyboard Layouts
Although the QWERTY keyboard, the default pattern for virtually all computer and typewriter keyboards in the English speaking world, is generally recognized to be poorly designed in terms of ergonomics, attempts to demonstrate the superiority of alternative patterns have not been effective. This may be because experimental subjects have extensive prior experience with the QWERTY keyboard which overwhelms any ergonomic advantage of the proposed alternative.
This study presents an alternative pattern that maintains the inter-key relationships of the standard keyboard while controlling for prior experience. The authors propose that this reverse-QWERTY keyboard can be used to provide an improved estimate of relative performance by removing the effects of experience, while maintaining the mechanics of the standard keyboard.
The first commercially successful typewriter, Sholes & Glidden Type Writer(1) used a keyboard pattern that was designed not to facilitate typing, but to minimize the frequency of mechanical jamming by placing the frequently used letters far apart(2). This keyboard pattern, commonly known as the QWERTY layout, is the default keyboard pattern of typewriters and computers today.
Although this pattern has frequently been criticized as cumbersome, inefficient, and solving a problem that no longer exists (3, 4), it remains the most commonly used keyboard layout. Several explanations for the survival of an inefficient keyboard pattern have been proposed, from industrial inertia to economic "lock" (5,6), but the strongest reason seems to be the lack of compelling evidence that alternatives are demonstrably better than QWERTY. In repeated studies and critiques, the advantages of alternative patterns are minimal to non-existent. (7, 8, 9) Repeatedly, keyboard patterns that ought to be superior are not shown to be better for the able-bodied typist nor for the typist with a disability. When theoretical arguments consistently disagree with experimental findings, either the theory is wrong or there is a confounding factor operating in the experiments.
Since there has been no compelling theoretical argument for the superiority of the QWERTY pattern, it seems likely that a confounding factor is operating in the experimental design of keyboard layout comparison studies. One confounding factor may well be the pervasive exposure to the QWERTY layout. In the design of an experiment on learning or level of a learned skill, the sample should have equal exposure to both experimental conditions. The purpose of random assignment to groups is to equalize the uncontrolled variables. But, when virtually every available experimental subject has hundreds of hours of experience with one condition, and only the few hours of practice provided within the study with the second condition, there is an inherent bias toward the commonly used QWERTY keyboard pattern in such studies.
To overcome this experiential bias, it would be desirable to find a way to preserve the mechanics of the QWERTY keyboard pattern, while eliminating the effects of experience. One way to attempt this would be to perform a right-to-left inversion of the keyboard layout. Since the right and left hands are essentially mirror images, this would result in all key presses being performed by a digit of equivalent strength and mobility, with an equivalent movement, but in a novel pattern. Unfortunately, this optimal approach has been shown not to eliminate the effects of learning. Matias, MacKenzie, and Buxton demonstrated that individuals would quickly achieve nearly normal typing speeds with the "half-QWERTY" keyboard layout that involves such a left-to-right inversion. It appears that the motor plans of typing are accessible to the opposite hemisphere to control movement.
A more intense, but perhaps more effective approach would be to perform both left-to-right and a top-to-bottom inversion of the keyboard pattern. This inversion is not as mechanically similar to the standard QWERTY layout because of differences in the use of the top and bottom rows of the keyboard but may be cognitively more effective than the simple left-right inversion in negating the practice effect while maintaining the right-left hand usage levels (argued to be important in keyboard design by Norman and Rummelhart (9)).
Would a left-right/top-bottom of the standard keyboard eliminate the effects of experience, and allow more valid comparisons between alternative keyboard layouts?
Alternative keyboard layouts were prepared for the IntelliKeys keyboard using Overlay Maker. These layouts provided the standard QWERTY keyboard pattern, and a modified pattern with the keyboard keys inverted.
Six able-bodied college students who had no identified, uncorrected visual, auditory, or motor deficits participated in the study. These subjects typed sample texts for 20 minute intervals with each keyboard until they produced three consecutive trials with speeds within 7% of each other, at which point they were considered to have achieved fluency at this point, they began typing with the alternative pattern, and continued until achieving fluency. The order of presentation of layouts was balanced so that three subjects started with the QWERTY layout, and three started with the reverse-QWERTY (the inverted pattern).
All subjects used a single-digit typing method, though this was a matter of personal choice, not experimental design.
Although the spacing and relationships between the keys remains constant in the two layouts, all of the subjects typed slower on the reverse-QWERTY layout than they did on the standard QWERTY. A typical subject achieved fluency on the reverse-QWERTY layout (10 wpm) at 62% of the rate achieved with the QWERTY keyboard (16 wpm). Subjects reported finding the reverse-QWERTY pattern frustrating to use, because "it didn't make any sense." This result was consistent regardless of which keyboard was used first, and, at fluency, the error rates between the two layouts were similar.
Because the two keyboard patterns maintain the inter-key spacing and directions, the difference in typing speed between the two layouts is most likely because of the effects of familiarity. This proportional decrease is in good agreement with prior observations of the first author that the first level of fluency with a new input method tends to be at 150% of the initial speed with that method.
This finding suggests that the reverse-QWERTY keyboard can be used to provide a typing level that would be equivalent to a subject typing on the QWERTY keyboard without prior experience. Thus, comparisons of alternative keyboard layouts to the reverse-QWERTY pattern may provide a more balanced comparison of utility than is available using the QWERTY layout.