The Telegraph Operators' Keyboard

The story is told in almost every account of the keyboard you are reading this on. The letters were scattered deliberately, to slow typists down, because the early machines jammed when two adjacent typebars were struck in quick succession. It is a satisfying story about how technology carries its old constraints forward.

It is also not true. The two researchers who went back through the machines and the patents, Koichi Yasuoka and Motoko Yasuoka of Kyoto University, put it about as bluntly as an academic paper allows. Several papers answer the question by saying the layout was made to slow the operator down, they wrote. "It's nonsense."

Who the machine was actually for

In November 1868 Christopher Latham Sholes and his colleagues Carlos Glidden, Samuel Willard Soulé and James Densmore shipped their first Type-Writer out of Milwaukee to Chicago. Their first customer was Edward Payson Porter, who ran Porter's Telegraph College.

That is the fact the received story leaves out. The early market for the typewriter was not authors or offices. It was telegraphy. An operator listening to Morse coming down a wire had to write it out in readable English at the speed it arrived, and a machine that could do that faster than a hand could was worth buying.

The first keyboard had 28 keys, looked like a piano, and closely resembled the Hughes-Phelps printing telegraph that was already in use in Porter's college. Its letters ran in alphabetical order, A to N left to right along one row and O to Z right to left along the next.

Then the users started asking for things.

What the operators wanted

Porter wanted numerals, because Morse traffic is full of them. In April 1870, with the help of Matthias Schwalbach, Sholes finished a machine with 38 keys carrying capitals, the numerals 2 to 9, a hyphen, a comma, a full stop and a question mark. The letters were still nearly alphabetical, with one oddity: U had moved next to O.

In September 1870 Sholes and Densmore took the machine to New York to show it to George Harrington and Daniel Hutchins Craig of the American Telegraph Works. One of Harrington's partners was Thomas Edison, who was not impressed. "The alignment of the letters was awful," he said later. "One letter would be one-sixteenth of an inch above the others; and all the letters wanted to wander out of line."

Harrington and Craig ordered machines anyway, on condition that a long list of things were changed, the keyboard among them. The changes that follow are the ones that produce the layout we have, and every one of them has a reason that comes from the wire rather than from the mechanism.

T, the commonest consonant in English, moved towards the centre. Q, one of the rarest, went out to the edge. I moved next to 8, because I was also used for the numeral 1 and an operator writing out a year needed to hit them together.

And S moved to sit between Z and E.

The reason for the S

American Morse, the code used on American landlines, is not the international Morse most people have heard of. In it, the letter Z is four dots with internal spacing, and that pattern is easily confused with the pair S followed by E, which turns up in English far more often than Z does.

An operator taking down a message frequently could not tell which one had just arrived, particularly at the start of a word, until more letters came through and settled it. So the layout put S within reach of both Z and E, and the operator's hand could go either way without moving.

That is why the third row of your keyboard begins A, S, D. It is a workaround for an ambiguity in a code that no commercial system has used for a century.

The rest of the shape arrived over the next year. On 10 August 1872 Scientific American put the machine on its front page with an engraving detailed enough that the keytops can still be read from it, which is how the layout of that year is known at all. By then it had 42 keys. In February 1873 Densmore and George Washington Newton Yost took a 43-key version to E. Remington and Sons at Ilion, New York, and on 1 March they signed a contract to manufacture it. The machine that went on sale that autumn sat on a stand like a sewing machine, with a foot pedal to return the carriage.

The second story, which is also wrong

There is a sequel myth, and it is the one that made QWERTY famous among economists. In 1985 Paul David used it as the standard illustration of path dependence: a demonstrably worse standard, locked in because everybody had already learned it.

The evidence for "worse" was the Dvorak Simplified Keyboard and a United States Navy study from 1944 reporting enormous gains from retraining on it. In 1990 Stan Liebowitz and Stephen Margolis went and read that study. Before presenting any results, its authors wrote that "indisputably, it is obvious that the Simplified Keyboard is easier to master than the Standard Keyboard," and later compared QWERTY to an ox and Dvorak to a jeep, adding that "no amount of goading the oxen can materially change the end result."

The man who held the patent on the Dvorak keyboard was Lieutenant Commander August Dvorak, who was the Navy's own senior expert on time and motion analysis during the war, and who had taken at least 130,000 dollars from the Carnegie Commission for Education for earlier studies of it. Earle Strong reported that Dvorak himself conducted the 1944 Navy experiment and Treasury experiments in 1946, a claim others have disputed. When Liebowitz and Margolis tried to obtain a copy of the Navy study, neither their own researchers nor the Navy's research librarian could find one.

Strong ran his own controlled experiment for the General Services Administration in 1956. Ten government typists retrained on Dvorak took well over twenty five hours of four hour daily training simply to get back to the speed they already had, and he concluded that retraining typists on Dvorak offered no advantage over retraining them on the keyboard they were already using. That study is missing from the histories.

So the layout was not designed to slow anyone down, and the case that it is holding everyone back rests on a report the Navy cannot produce, written by the man selling the alternative.

The engraving on the front of Scientific American in August 1872 shows a woman at the machine, hands on the keys. The letters under her fingers had been arranged around a code coming in over a wire, and they have not moved since.

Sources

• Koichi Yasuoka and Motoko Yasuoka, Kyoto University, "On the Prehistory of QWERTY", ZINBUN No. 42 (the rejection of the slowing-down explanation; the first Type-Writer shipped from Milwaukee in November 1868 to Edward Payson Porter of Porter's Telegraph College; the 28 key piano-like keyboard resembling the Hughes-Phelps printing telegraph and its alphabetical arrangement; Porter requiring numerals and the 38 key machine of April 1870 built with Matthias Schwalbach; the New York demonstration to Harrington and Craig and Edison's remarks on alignment; the moves of T, Q and I; the placing of S between Z and E because American Morse renders Z as a pattern confusable with the digram SE; the Scientific American front page of 10 August 1872 and the 42 key layout read from its engraving; and the visit to E. Remington and Sons in February 1873 with the contract signed on 1 March).
• Stan J. Liebowitz and Stephen E. Margolis, "The Fable of the Keys", Journal of Law and Economics volume 33 (1990) (the use of QWERTY as the standard case of lock-in; the 1944 Navy study's statements that the Simplified Keyboard is indisputably easier to master and its ox and jeep comparison; August Dvorak's position as the Navy's top time and motion expert, his patent and the at least 130,000 dollars received from the Carnegie Commission for Education; Earle Strong's report that Dvorak conducted the 1944 and 1946 experiments and Yamada's denial of it; the inability of the authors or the Navy's research librarian to locate a copy of the study; and Strong's 1956 General Services Administration experiment in which ten typists retrained on Dvorak needed well over twenty five hours to regain their previous speed).
• Scientific American, 10 August 1872, front page engraving of the Type-Writer, reproduced at "Sholes and Glidden typewriter" (the machine as it was sold to the public in 1872, with the keytops visible).
• Paul A. David, "Clio and the Economics of QWERTY", American Economic Review volume 75 number 2 (1985) (the original path dependence argument that Liebowitz and Margolis were answering).
• Hisao Yamada, "A Historical Study of Typewriters and Typing Methods: From the Position of Planning Japanese Parallels", Journal of Information Processing volume 2 (1980) (the survey that carried the slowing-down explanation into the literature, and Yamada's own disputes with Strong).