willy@utcsrgv.UUCP (W. Buxton) (09/04/83)
Regarding chord keyboards, there are a number of interesting points, yet very little hard research. Their first use in computer systems (to my knowledge) was at SRI. This was in work by Engelbart and English, "A Research Centre for Augmenting Human Intellect", FJCC, 1968, pp395-410. This was, in fact, the same research project that brought us the mouse. The chord keyboards at XEROX PARC were from Engelbart's work, and were much like piano keys. As has been remarked, they did not achieve great popularity. Just because chording keyboards never caught on is no proof that they cannot provide powerful solutions to many UI problems. As stated earlier, however, there is very little compelling evidence in either direction. Much motivation for chord keyboards is to provide one-handed typing in cases where the other hand is otherwise occupied (as in using a mouse or tablet). In such cases, it should be noted that the real issue is one-handed typing, not chording. Often, where the repertoire of lexemes is low, the desired result can be obtained without chording. Some one-handed keyboards are available commercially. One is made by an English company, Maltron, 219 Sycamore Road, Farnsborough, Hampshire, GU14 6RQ, England. Little experimentation exists, however, to demonstrate how such two-handed, split- task input compare to more traditional input. Such tests should be run, and the keystroke model of Card, Moran, and Newell (CACM, 23(7), July 1980, 39pp. 396-410), could be adapted to enable such a comparison to be made. There is, I should add, as little in the literature concerning two-handed input as there is on chording keyboards. One little-cited reference, that gives no experimental results, but describes an early and effective implementation, is Pulfer,K. (1971), "Man-Machine Interaction in Creative Applications", IJMMS, 3, pp. 1-11. Concerning chording keyboards, themselves, there are a few important parameters that affect their use: o kinesthetics/industrial design: number of buttons, their layout, and action. o number of lexemes (especially when considered in combination with the number of distinct keys). o the encoding scheme used o interference from other concurrent tasks being performed. There are more, and these are not mutually exclusive. They do, however provide the basis for focusing in on some key points. First, the SRI chord keyboard was designed to be used by either the left or right hand. It also, consequently, was not optimal for either. Some devices have been designed to fit under the ball of the palm, and have the buttons layed out to recognize the natural curvature of the finger-tips across the hand. The Swiss mouse is one (that used on the Blit, available from Logitech, 165 University Ave, Palo Alto, CA, 94301), and a one-handed 6-button portable word processor called the Microwriter (Microwriter Limited, 31 Southampton Row, London WC1B 5HJ, England). One advantage to 5 or 6 button chording keysets is that the hand can effectively always sit on "home-row", so that if you can use them at all, you can touch-type. This is good for minimising interference with other tasks, such as pointing, reduces the time for button pushes (a la keystroke model), but makes the problem of quality of buttons (avoiding false key depressions since fingers rest on keys), and their placement, critical. Clearly, the more lexemes that are encoded on a keyset, the more difficult it is to use. The key cognitive problem with such keysets is that you CANNOT hunt-and-peck. You must memorize the encodings, and this means an increased learning curve, errors and slow-down for infrequent characters or functions, and re-learning problems for casual users. There are also real problems of how to minimize roll-over problems. The actual encoding used is critical in minimizing the problems of learning, retention, and operating error. In encoding text, various mnemonics have been used. The Microwriter uses several (often which interfere with one-another, thereby illustrating yet another issue). For example, the 2 most common characters. 'e' and ' ' are single keys, index finger and thumb, respectively. This is reasonable, since these are the strongest fingers. The mnemonic for the letter 's' is derived from the fact that it is a single key using the ring, or (s for 's' for) Signet finger. And yet others are based on spatial relationships. For example, 'L' is encoded with a 3-button chord, index finger, thumb, and baby finger, representing the 3 vertices of the graph. One example of how three types of mnemonics can interfere can be seen in considering how well they transfer from hand to hand. 's' would imply a mirror reflection, while 'L' wouldn't. Handedness is, in fact a, a real issue in one-handed key-sets, both among different users and even for the single user. In the former case, the problem of handedness is seen in that the keysets should fit the curvature of the hand used. It can also be seen as an issue, from the last example, where different mnemonics (and hence documentation) may be forced depending on handedness. More significant, but less obvious, is that if no other task is being performed, we will usually do one- handed keyboard operations with the major hand. Your calculator or Microwriter would be examples. When using a mouse or tablet, however, the major hand will nearly always be used for the spatial task, and the minor hand for the discrete button task. This suggests the potential for crossing in some cases. One point to fall out of the previous discussion is to emphasize the case where the chording button task is performed with the same hand as the spatial task. An example is in using chording on the buttons of the tablet puck or the mouse. The puck or mouse should be considered as a potential chording keyboard. This has, in fact, been practiced at several sites, including Blit users at Bell Labs. A key point to note in this regard, however, is the total unsuitability of nearly all pucks and mice to be used in this way. The optical mouse (regardless of manufacturer), and the Hawley mouse, for example, have kinesthetic properties that dictate against such use. The only mouse that lends itself well to such application is the Swiss mouse, mentioned earlier. Obviously the trend to one-button mice and pucks is a distaster from this perspective. Having experimented with a number of such devices, my suggestion is that one-handed typing in conjunction with the performance of spatial tasks has a great future. However, this is only true where the number of codes to be transmitted is small; cases where the chords are used as function keys. CAD systems are one type of system that could benefit. Tring to type alphabetic information (such as text), however, will not see much use. Try a Microwriter to convince yourself. Finally, there are two-handed chording keyboards. I have simply not considered them in this note, as they fit less-well into the framework and motivation of the discussion. For those with some more energy, here are a few additional references: Rochester, Bequaert & Sharp (1978). "The Chord Keyboard", IEEE Computer, December 1978, pp. 57-63. Montgomery, E. (1982). "Bringing Manual Input into the 20th Century", IEEE Computer, March 1982, pp. 11-18. See also follow-up letters in the May and, June & October 1982 issues. Norman & Fisher (1982). "Why Alphabetic Keyboards are not Easy to Use: Keyboard Layout Doesn't Much Matter", Human Factors (224(5), pp 509-519. Noyes, J. (1983). "The QWERTY Keyboard: a Review", IJMMS 18, pp. 265-281. Finally, I would be interested in obtaining photographs and documentation on any keyboard (chording or otherwise) that people have made in the course of their work. Any user comments would also be welcome. This is obviously related to my own research and writing. Bill Buxton CSRG University of Toronto Toronto, Ontario Canada M5S 1A4 (416)-978-6320 decvax!utcsrgv!willy
jfarrell@sun.UUCP (Jerry Farrell) (09/05/83)
Buxton's note was a welcome injection of facts & prior research into the discussion; thank you, Bill. I think two of his points about mice and chording deserve further discussion: It's true the dePraz (Swiss, ~hemispherical) mouse is well set up for chording its buttons. (The mouse is designed to be grasped with the whole hand in a grip which leaves the index, middle and ring fingers over the 3 buttons on its front edge.) Unfortunately, this aspect of its design conflicts with the mouse's use as a pointing device - the grip required is the "power grip"; it requires the large muscles of the upper arm to move the mouse, and loses the precision manipulation and pointing capabilities of the fingers. - the buttons are pushed in a plane parallel to the surface the mouse rests on; this makes it likely the mouse will be moved from its position by the act of pushing the button. His assertion that other mice (Hawley, Kirsh, etc.) are not suited to chording needs qualification. It's true that getting all 8 boolean combinations from 3 buttons is impractical if the mouse is only held in the fingers; the little finger is too weak to keep the mouse stable, and move it comfortably. But two-finger combinations can and have been used effectively. When NS-8000's with their 2-button mice replaced Altos in Xerox' development environment, it didn't take long before chording both buttons within n ms. was made an optional equivalent for pressing the missing middle button, even though that meant processing of >every< mouse button-push had to be delayed for those n ms. Similarly, Sun's window system uses two-finger chording as an expert-option for some common operations that are also available by a longer one-finger sequence. In both these contexts, the mouse is gripped between the thumb on the inside and the little & ring fingers on the outside; the other two fingers are free to range over the buttons. The heel of the palm rests on the table behind the mouse, which is moved almost entirely with the wrist and fingers -- canonical "precision grip."