QWERTY Keyboard History: It Was Not Designed to Slow You Down, and the Study Proving Dvorak Was Faster Was Run by Dvorak

Two things are repeated about the keyboard under your hands. The first is that Christopher Latham Sholes scattered the letters to slow typists down, so the typebars would stop jamming. The second is that a better layout exists, was proven faster, and lost anyway - which makes QWERTY the standard example of a market getting stuck with an inferior standard.

Both claims have a documentary record. Neither survives it.

The jamming story does not fit the customers

Koichi Yasuoka and Motoko Yasuoka (2011), of Kyoto University’s Institute for Research in Humanities and Kyoto University of Foreign Studies, published On the Prehistory of QWERTY in ZINBUN (No. 42, pp. 161-174). They traced the arrangement through patents and correspondence from 1868 to 1882.

Their account is that QWERTY was not designed at all, in the sense of being laid out to a specification. It accumulated. The keyboard descended from the Hughes-Phelps printing telegraph, and it was reshaped repeatedly - by Sholes, by George Harrington and Daniel Hutchins Craig, by Jenne and Clough, and finally by Wyckoff, Seamans and Benedict - each responding to a different requirement at a different moment. They cite US Patents 207,559 (Sholes, 1878) and 199,263 (Clough and Jenne, 1878), and date the modern arrangement to the machine Wyckoff, Seamans & Benedict released in August 1882.

The decisive argument against the slow-down story is not about mechanics. It is about who was buying the machines. In November 1868 Sholes and his colleagues shipped their first typewriter to Edward Payson Porter, the principal of a telegraph college. The early customers were Morse receivers - operators transcribing incoming telegraph traffic in real time.

The Yasuokas’ objection follows directly:

“It’s false… early customers of Type-Writer were Morse receivers. The speed of Morse receiver should be equal to the Morse sender, of course. If Sholes really arranged the keyboard to slow down the operator, the operator became unable to catch up the Morse sender.”

A machine deliberately built to be slower would have been unsaleable to the only people buying it.

They dispose of a second legend on the way. The claim that R was moved so that salesmen could type “TYPE WRITER” from a single row fails on dates: the brand name Sholes & Glidden Type-Writer was settled before Jenne and Clough repositioned R. Their aside is the kind of question that ends an argument: “Why wasn’t SHOLES on one row?”

The alternative, and who ran the studies

August Dvorak patented the Dvorak Simplified Keyboard in 1936. The case for it rests on two studies, and both were examined in detail by Stan Liebowitz and Stephen Margolis (1990) in The Journal of Law and Economics (33(1), 1-25) - a paper written specifically because economists had adopted QWERTY as the textbook case of an inferior standard locking in.

The 1944 US Navy study. This is the one usually cited. Fourteen Navy typists were retrained on Dvorak keyboards for two hours a day; the report claimed they recovered their original speed within two hours of training and then reached an average of 56 net words per minute against an original 32 - a 75% increase.

Liebowitz and Margolis obtained the report with some difficulty and describe what it does not say about itself. Dvorak was a Lieutenant Commander and “the Navy’s top expert in the analysis of time and motion studies” during the war. Earle Strong subsequently reported that the 1944 experiments “were conducted by Dr. Dvorak.” Dvorak held the patent on the keyboard being tested and had received, the authors state, “at least $130,000 from the Carnegie Commission for Education for the studies performed.”

Beyond the conflict, the design problems they list are serious: participants “do not appear to be based on a random process”; the typists began at an average of 32 words per minute, below the Navy’s own competency standard, leaving a great deal of room to improve; measurement was inconsistent between conditions in ways that inflated the Dvorak figures; and “the Qwerty group was not a reasonable control.” Their summary of the pattern: “All, suspiciously, tended in favor of the Dvorak design.”

The report’s own framing gives the game away before the results appear: “Indisputably, it is obvious that the Simplified Keyboard is easier to master than the Standard Keyboard.”

The 1936 comparisons. Dvorak’s own book set typing speeds from “different and completely separate experiments, conducted by different researchers for various purposes” side by side - one on Dvorak, three on QWERTY - with students of different ages in different school systems taking different tests in classes of different lengths. Seventh and eighth graders learning Dvorak were compared with students in conventional high schools. Liebowitz and Margolis note that Dvorak “did not stipulate whether their choice of studies was a random sample or the full universe of available studies.”

The study nobody quotes

In 1956, Earle Strong ran the experiment for the US General Services Administration that the Navy study should have been: a controlled comparison designed to answer the question a government agency would actually face - is retraining worth paying for?

Ten typists were retrained on Dvorak. In the second phase, those typists continued Dvorak training while ten QWERTY typists were given equivalent retraining on QWERTY.

Two results:

• It took “well over twenty-five hours of four-hour-a-day training for these typists to catch up to their old speed” - against the Navy’s claim of two hours.
• “the Dvorak typists improved less quickly with further Dvorak training than did Qwerty typists trained on Qwerty keyboards.”

That second finding is the one that matters, and it is the reason a straight speed comparison is the wrong test. Retraining anybody improves them. The question is whether retraining them onto a new layout beats spending the same hours on the layout they already know. Strong found it did not.

He concluded that Dvorak retraining “would never be able to amortize its costs” and recommended continuing with QWERTY. Liebowitz and Margolis credit that study with “putting Dvorak to rest as a serious alternative.”

What is actually true

The honest version is less dramatic than either story, and it is a better example than the myth was.

QWERTY was not designed to slow anyone. It emerged from fourteen years of incremental modification by a shifting cast of people solving different problems, in a machine whose first serious buyers needed to keep pace with Morse code.

Dvorak may well be marginally better. The evidence that it is decisively better comes from studies its inventor ran, holding its patent, funded by a grant to him. The one independent controlled test found the retraining cost was never recovered.

Which means the keyboard is a poor illustration of the thing it is usually used to illustrate. It is not obviously an inferior standard that got locked in. It might just be a standard.

There is a general lesson here that has nothing to do with keyboards. When a claim about an interface is repeated for a century, it is worth asking who ran the study, what they stood to gain, and whether anyone independent ever ran it again.

Limitations and disputes

• Liebowitz and Margolis had a position. Their paper was written to contest the economics of lock-in, and their reading of the Navy report is adversarial. We have reported their specific documented findings - Dvorak’s rank, his patent, the Carnegie funding, the design flaws they enumerate - rather than their broader theoretical argument.
• The 1944 Navy report and the 1956 GSA study were not read in the original. Both are described here as reported by Liebowitz and Margolis, who state they obtained the Navy report directly.
• The $130,000 Carnegie figure is as stated by Liebowitz and Margolis and we have not independently verified it against a Carnegie record.
• The Yasuokas’ account is a historical reconstruction from patents and correspondence. It is the most thoroughly documented account available, but historical reconstruction is not experiment.
• None of this establishes which layout is faster today, on modern keyboards, for modern users. That question is not answered here and the studies discussed are from 1936, 1944 and 1956.
• We have not printed any modern typing-speed comparison between layouts, because we did not verify one.