qwerty, alphabetic, and dvorak keyboards
c70:editor-people
Sigh, the Sholes versus Dvorak myth rises again. I don't know where
Bruce Borden got his facts, but his story sounds to me as if it were a
combination of stories. I believe Borden is talking about the linotype
keyboard, which uses the "shrdlu" arrangement. The Sholes keyboard
(aka "qwerty") was designed for a typewriter so as to minimize the
jamming of typebars as they moved to the platen. This caused the
placement of frequent pairs as far from one another as possible. In
fact, this SPEEDS typing because typing on alternate hands is faster
than on the same hand (list of references and reprints of papers
available on demand: see, for example Rumelhart & Norman in the next
Cognitive Science). This point wasn't appreciated at the time because
nobody thought of using all ten fingers, and typing without looking at
the keyboard was unheard of; as someone else said, touch typing was a
heroic, unexpected invention (and required a national typing speed
contest to prove that it worked).
There have been hundreds of studies comparing Dvorak arrangements with
Sholes arrangements. Dvorak fans claim massive improvements in speed.
(We have an old movie -- made by Dvorak in mid 1900's that makes
remarkable claims.) However, experiments done by neutral parties tend
to put the improvement around the 5 to 10% range -- not worth the
effort. Card and Moran at Xerox Parc have a computational method of
computing speed that yields numbers in that range and Rumelhart and I
have a full fledged typing simulation model that, when given the Dvorak
keyboard, only speeds up by 5%. As others have pointed out, you can
get a far greater improvement in typing speed by moving the RETURN key,
either to where it can be reached without distorting the hand (say by
the left thumb which our studies show is not used by typists) or by
having automatic RETURNs (as in various text editors). Kinkead put it
this way: elimination of the RETURN key gives a minimum of 7%
improvement in speed and "up to 30% when the original copy is not
properly formatted."
A while ago, I decided that alphabetically arranged keyboards would
surely be better for first time typists, so we did some experiments. I
was wrong. Randomly arranged keyboards and alphabetically arranged
keyboards were equivalent. (Sholes arrangements were better, but that
is probably because everyone has had some exposure to keyboards, even
though we tried to study only non-typists.) On the typing simulation
model, alphabetic keyboards were all slower than Sholes, confirming the
fact that putting frequent pairs on opposite hands speeds up typing
rate. Why wasn't alphabetic better? Because the mental effort to make
use of the alphabetic arrangement is too much -- and most people don't
know the alphabetic that well anyway (how far away -- and in what
direction-- is "p" from "u"?, or even "e" from "i"?).
If you want to improve typing speed, don't tinker with the current key
layout, but do dramatic re-arrangements, as in the new 5 key hebrew
keyboard (by Gopher) or the various chord keyboards suggested by
others. (Why would you want the crazy, staggered key, assymetrical,
long key distances of the current keyboard? The hands are symmetrical
and this keyboard isn't. The distances one must travel are extreme.
The space bar is wasted. And so on.
Reprints of the simulation model paper and alphabetic studies are
available. Large, detailed study of typing available in a monograth,
but only for those who really care about detailed finger movements
and response time distributions.
Don Norman (norman@nprdc ucbvax!sdcsvax!norman).