“The QWERTY problem” has been defined in economics textbooks as “an inferior industry standard that has prevailed possibly because of historical accident”. QWERTY is not just the first six letters of the standard English language keyboard, today it has also become the focal point for debates as to whether markets should be left to develop their own technological solutions, or whether governments should intervene to help stop them make mistakes.
QWERTY was invented by Christopher Latham Sholes in 1873 and here is a picture of his first QWERTY typewriter (thanks to the National Museums of Scotland for letting me photograph it).
Note the typebasket at the top of the device which arranged typebars in a circle. In these early machines the typebars swung up to hit the printing point above them. The arrangement of the keyboard and the typebasket in the 1873 QWERTY are shown in the second figure. One widely reported problem with these early machines was that typebars which were next to each other on the circular typebasket were prone to jamming if struck in rapid succession.
There are lots of undocumented stories (and alleged “myths”) that have grown up around why QWERTY was invented, for example one was that the seven letters that make up “typewriter” were put in the top row of the keyboard to make it easier for early typewriter salesmen to type and demonstrate ease of use.
Another story (and alleged “myth”) was that Sholes used a list of the most frequent letter pairs in the English language (like “th” and “es”) to separate them on the typebasket to reduce the risk and incidence of jamming.
First, what are the chances that the seven letters that made up “typewriter” would have landed on the top row by chance? Using basic probability theory, the answer is about one chance in 5,000. So we can reject the argument that this was an accident. The fact that the word TYPEWRITER was emblazoned in bold on the front of the early machines was also a helpful visual cue.
As far as the other supposed myth that Sholes used a list of frequent letter pairs to separate them on the typebasket, again basic probability theory was useful here. But this just led to another puzzle because it showed that random allocation of letters on the typebasket was almost as effective as such lists in separating frequent letter pairs.
So what did Sholes do? The answer is blindingly obvious once you see it, but initial commercial secrecy and Sholes premature death obscured the obvious.
It is quite difficult to prepare lists of frequent letter pairs. Sholes turned the problem on its head; it is very easy to identify infrequent letter pairs. With 44 typebars and 26 letters, some letters had to be next to each other on the typebasket. So look at the two strings of letters on the typebasket where that happened:
QA and SZCDXFVGBHNJ
Now try to find words that have letter pairs contained in these sequences read in either direction, like “FV” or “VF”. Not easy is it? I bet you cannot find more than a few, but let me know if I am wrong.
In fact, we have been able to show with electronic searches of PDF texts that with this “infrequency rule”, QWERTY was near-optimal in terms of separating frequent letter pairs on the typebasket, and this in turn would have solved much of the jamming problem.
QWERTY was not an “accident”, it was instead a brilliant design solution by Sholes to a real engineering problem, so there will have to be some tweaking at least of basic economics textbooks here.
But is it still fair to label QWERTY “inferior” and inefficient? Ah, that is a subject for another blog.
(Professor Kay’s paper “Rerun the tape of history and QWERTY always wins” is forthcoming in the journal Research Policy, accompanied by commentaries on the paper by Professors Brian Arthur, Stephen Margolis, and Jean-Phillipe Vergne, and a response by Professor Kay).