The keyboard beneath your fingers is the most famous exhibit of "market failure" in economics textbooks: QWERTY is said to be an inferior design invented to prevent jamming, locking humanity into a suboptimal layout for over a century—unable to switch to the superior Dvorak layout. This story entered the *American Economic Review* in 1985 via Paul David and was amplified by textbooks. Yet Stan Liebowitz and Stephen Margolis, after meticulously examining historical sources in 1990 and 1995, reached the opposite conclusion: the exhibit itself is a fable. This article audits the evidence and proposes a three-condition framework specifying what must be proven to establish that "lock-in to an inferior standard" has occurred.
## The Standard Version of the Story
Christopher Latham Sholes, a newspaper editor and printer from Kenosha, Wisconsin, filed a typewriter patent in October 1867 and received it on July 14, 1868 ([Wikipedia QWERTY entry](https://en.wikipedia.org/wiki/QWERTY)). [The full text of *The Fable of the Keys*](http://web.archive.org/web/20010109130300id_/http://www.utdallas.edu:80/~liebowit/keys1.html) (Liebowitz & Margolis, 1990; hereafter "Fable") acknowledges that Sholes addressed the problem of typebars jamming when certain key combinations were struck rapidly in succession, and that separating high-frequency key pairs was a "partial solution." However, Wikipedia notes that the popular explanation is suspect: E and R are adjacent on QWERTY, yet "ER" is a high-frequency English letter pair—contradicting the jamming rationale of "separating frequent pairs."
In early 1873, the patent rights were sold to E. Remington & Sons, with commercial production beginning by year-end (Fable); the machine went on sale in 1874. The Remington No. 2, released in 1878, was the first typewriter with a Shift key enabling both upper- and lower-case characters—and QWERTY spread through its success (Wikipedia). A challenger appeared in 1932: August Dvorak, professor of educational psychology at the University of Washington, and his brother-in-law William Dealey completed a new layout; they received [U.S. Patent 2,040,248](https://patents.google.com/patent/US2040248A/en) on May 12, 1936. Their design claimed approximately 70% of keystrokes on the home row, 22% on the top row, and 8% on the bottom row ([Wikipedia Dvorak entry](https://en.wikipedia.org/wiki/Dvorak_Simplified_Keyboard)); in 1982, ANSI designated it an alternative standard layout (Wikipedia).
David’s version of the story (as paraphrased in full in L-M 1995; original behind a paywall, unread) holds that QWERTY is "dramatically inferior" to Dvorak but became locked in due to coordination failure: "No one learns Dvorak because Dvorak machines are hard to find; Dvorak machines are hard to find because no one learns Dvorak" ([*Path Dependence, Lock-In, and History*, full text](http://web.archive.org/web/20001208011800id_/http://www.utdallas.edu:80/~liebowit/paths.html), 1995; hereafter "L-M 1995"). Pillar One: WWII Navy experiments—David claims retraining costs could be recouped within ten days after returning to full-time work; L-M notes this citation is "unattributed" (Fable). Pillar Two: The Cincinnati typing contest of July 25, 1888—Frank McGurrin, a court stenographer from Salt Lake City, is believed to have been the first to memorize the keyboard and type blind; his victory is treated in popular history as the watershed moment proving QWERTY’s superiority (Fable).
## First Audit: The Witness’s Interest
Dvorak held the keyboard patent and received at least $130,000 in funding from the Carnegie Foundation for the Advancement of Teaching while at the University of Washington (Fable footnote 30); during WWII he served as a Navy lieutenant commander and was the Navy’s top expert in time-and-motion analysis. Earle Strong reported that the 1944 Navy experiment and the 1946 Treasury Department experiment were both conducted by Dvorak himself (Fable). Defender Hisao Yamada claimed Dvorak only supplied the typewriters and did not conduct the experiments; L-M admits in the original text that "we do not know whom to believe," but expresses skepticism based on indirect evidence (Fable footnote 29).
Dvorak et al.’s 1936 book *Typewriting Behavior* compares experiments described as "four completely independent studies, conducted by different researchers for different purposes": only one tested Dvorak, while three tested QWERTY; subjects differed in age and ability (e.g., Chicago University Laboratory School grades 7–8 versus ordinary high school students); tests and instructional hours varied; and it is unspecified whether the studies cited constitute a random sample or merely all available studies (Fable)—making cross-study comparisons invalid. An Apple advertisement claimed the Dvorak Simplified Keyboard (DSK) "lets you type 20 to 40% faster"; David quoted this ad verbatim in his paper (Fable).
## Second Audit: The Navy Itself Disavowed the Record
The most dramatic figure from the Navy experiments is 180 words per minute. Fable cites an Associated Press report dated October 7, 1943: the new keyboard enabled typists to "zip along at 180 words per minute." But the very next sentence in that AP report states that Navy Lieutenant Commander W. Marvin McCarthy, in a letter dated October 14, 1943, denied the Navy possessed any record of such a speed test and also denied issuing any official announcement about it; *Business Week*’s October 16, 1943, report gave the figure as 108 words per minute, not 180 (Fable). Arthur Foulke—the biographer of Sholes and a believer in Dvorak’s superiority—was the source of the conflicting reports (Fable). Wikipedia further notes that Dvorak "discarded at least two prior studies," whose results remain unknown (Wikipedia, secondary source).
## Third Audit: The 1956 Controlled Experiment
The U.S. General Services Administration commissioned Earle Strong to conduct a controlled experiment, published in 1956 as *A Comparative Experiment in Simplified Keyboard Retraining and Standard Keyboard Supplementary Training* (Fable bibliography). First, ten government typists were retrained on Dvorak until they regained their original QWERTY speed—a process requiring "far more than 25 days of four hours each," i.e., over 100 hours—far exceeding David’s claimed ten-day recovery period. In the second phase, the Dvorak group continued training, while another group of ten QWERTY typists received parallel supplementary QWERTY training; the Dvorak group progressed more slowly (Fable). Strong concluded that retraining on Dvorak conferred no advantage over continuing QWERTY training, and recommended that the government provide additional QWERTY training to QWERTY typists; Fable notes this study has been ignored in both David’s economic history and in popular accounts.
Wikipedia states Strong harbored personal animosity toward Dvorak and had publicly opposed the new keyboard before conducting the study (Wikipedia); Fable concedes that skilled typists cannot answer the question of whether beginners learn Dvorak faster (Fable). Yamada cites two supportive studies, revealing their scale: in 1973, six Western Electric typists trained for 104 hours achieved a 2.6% speed increase over their original QWERTY rate; in 1978, Oregon State University typists trained for 100 hours recovered to 97.6% of their original QWERTY speed (Fable). Both durations approximate Strong’s, neither included a parallel QWERTY control group, and both show advantages far weaker than the Navy narrative.
Beginner-side positives: Wikipedia records that in the 1930s, the Tacoma, Washington, school district taught Dvorak to 2,700 high school students, who learned it in one-third the time required for QWERTY—yet the newly elected school board terminated the program (Wikipedia, secondary source): institutional decision-making, not market coordination, killed it. Counter-evidence: Wikipedia notes subsequent experiments showed many layouts—including alphabetical order—yielded similar speeds after training (Wikipedia); Fable surveys human factors engineering literature and finds no scientifically reliable evidence of significant Dvorak advantage—methodologies are imprecise, the null hypothesis of equivalence cannot be rejected, and observed speed advantages fall far short of the advertised 20–40% (Fable).
## Overlooked Contemporary Reporting
Fable also unearthed contemporaneous *New York Times* reporting omitted from the dominant narrative. Just months after McGurrin’s win, the *Times*, under the headline "Wonderful Typing," reported a Brooklyn demonstration: Thomas Osborne typed 142 words per minute in a five-minute test and 198 words per minute in a 30-second burst using the non-QWERTY Caligraph typewriter (72 keys, separate upper- and lower-case keys); George McBride, typing blind, achieved 129 words per minute in the same session (Fable). At the New York contest on August 2, 1888, McGurrin won with 95.8 words per minute, followed closely by May Orr at 95.2 and M. C. Grant at 93.8 (Fable): victory was not a landslide, and non-QWERTY machines were not slow.
Fable details: Dvorak’s measured home-row keystroke share was 67%, with left-right hand load distribution at 47:53; QWERTY’s load distribution was 57:43, with most keystrokes on the top row—but its design of "separating jam-prone keys" precisely satisfies Norman and Rumelhart’s "C factor": alternating hands, so the idle hand is already preparing for the next keystroke (Fable).
## Three-Degree Path Dependence: Separating Two Things
L-M 1995’s analytical framework distinguishes degrees of path dependence: First-degree sensitivity to initial conditions, carrying no implication of inefficiency. Second-degree choice made under incomplete information, where inferiority was unknowable at the time and recognized only retrospectively—in light of then-available knowledge, it was not meaningfully inefficient. Third-degree path dependence arises when initial sensitivity leads to inefficiency that is "remediable": a feasible arrangement existed—or once existed—that could have identified and implemented a superior outcome, yet failed to do so (L-M 1995). Only third-degree dependence challenges the neoclassical paradigm. L-M explicitly state that first- and second-degree dependencies are commonplace, whereas third-degree dependence rests on stringent assumptions: someone must have known B was better, and that informed party must have been unable—through coordinating buyers or selling information—to convert that knowledge into profit.
The economic foundation of this framework is appropriability. Fable states: "An owner with the prospect of appropriating substantial benefits from a new standard would have an incentive to share some of the costs of switching to a new standard." Applied to keyboards: Dvorak held a patent and received at least $130,000 in funding—if DSK were truly significantly superior, he had the strongest motive and means to subsidize conversion; the market’s failure to self-correct is itself evidence that "superiority" remains unproven. L-M also point out the rhetorical slippage in David’s narrative: attributing error to "decentralized decision-making" implicitly suggests centralized alternatives could correct it (L-M 1995)—the fable smuggles in policy prescriptions.
## Three-Condition Test and Modern Comparisons
This article proposes a three-condition, empirically testable framework for "inefficient lock-in" (this article’s synthesis + judgment): (1) A feasible alternative exists; (2) The net benefit differential is measurable and requires comparative evidence—not mere assertion; (3) Switching costs and appropriability are insufficient to explain persistent failure to correct. The QWERTY case fails condition two by its own proponents’ evidentiary standards: Navy records disavowed by their source, the 1936 book’s cross-experiment comparisons invalidated, the 1956 controlled experiment showing no advantage, and convergence across multiple layouts.
VHS vs. Betamax: Wikipedia lists three mechanisms unrelated to quality (video rental store inventory network effects, manufacturer expectations, Sony’s stricter licensing), while the "pornography theory" is labeled speculative and contested ([Wikipedia Path Dependence entry](https://en.wikipedia.org/wiki/Path_dependence)). Liebowitz’s 2002 *Re-thinking the Network Economy*, p. 41, cites counterevidence: "It was the inferior playing time that led to the demise of the Betamax, not the fact that it was first or second or third." (Wikipedia quote). Recording duration is a measurable benefit differential: condition two holds, so the outcome can be explained by quality—not lock-in.
HD DVD vs. Blu-ray: Warner Bros. announced abandonment of HD DVD on January 4, 2008; Best Buy and Netflix followed on February 11; Walmart on February 16; Toshiba conceded defeat on February 19 ([Wikipedia High-definition optical disc format war entry](https://en.wikipedia.org/wiki/High-definition_optical_disc_format_war)): Toshiba’s surrender came less than two months after Warner’s defection—"lock-in" was broken by a shift in focal attention. USB-C: The EU revised its Radio Equipment Directive in 2022, mandating USB-C for new smartphones by end-2024 and laptops by 2026 ([Wikipedia Common external power supply entry](https://en.wikipedia.org/wiki/Common_external_power_supply), citing BBC, December 12, 2022): when benefit differentials are measurable, regulation can resolve coordination failure.
## Three Judgments
Judgment One (this article’s judgment): The QWERTY case fails the evidentiary standard for "inefficient lock-in"; the correct conclusion is that layout differences are far smaller than the narrative claims—not that QWERTY is optimal. Existing evidence includes: L-M’s archival audit, the Navy’s disavowal letter, Strong’s controlled experiment, Yamada’s cited studies, and convergence across layouts. This article’s contribution: declaring the claim "unproven" rather than "disproven," and articulating the burden-of-proof structure.
Strongest rebuttal: L-M are not disinterested—Fable’s front-page snapshot link explicitly lists the Microsoft antitrust case as their writing focus; Strong harbored animosity; skilled-typist limitations; the Tacoma beginner experiment favored Dvorak. Response: counter-evidence is already embedded in the main text; the framework does not depend on L-M’s motives—the Navy’s disavowal and the 1936 book’s methodological flaws are independently verifiable. Evidence that would overturn this judgment: a new beginner experiment with a parallel QWERTY control group demonstrating Dvorak’s sustained, significant advantage—and where switching costs alone are insufficient to explain persistent non-adoption.
Judgment Two (this article’s synthesis + judgment): The three-condition framework transforms lock-in from narrative into an audit checklist; appropriability is its economic foundation: those who stand to profit from correction have incentive to subsidize it, and persistent failure to correct is itself evidence. Existing evidence includes: L-M’s third-degree definition and stringent assumptions, Fable’s appropriability quotation, and Dvorak’s patent and funding. This article’s contribution: compressing third-degree dependence into three testable conditions and calibrating them with three modern cases.
Strongest rebuttal: Appropriability is incomplete—patents do not guarantee capture of all social benefits; Dvorak was an educator, not an entrepreneur; funding was not commercial capital. Response: the framework only requires that appropriability be "insufficient to explain" persistent failure—not that profit opportunity guarantees correction. Evidence that would overturn this judgment: a case where net benefit differential is large and measurable, the patent holder possesses capacity to subsidize, yet long-term failure to correct persists without regulatory or focal intervention.
Judgment Three (this article’s judgment): The fable’s true lesson concerns the transmission chain and evidentiary standards: compelling stories, amplified by textbooks, can outrun archival verification. Existing evidence includes: Fable notes that papers by Farrell & Saloner and Tirole’s textbook treat the keyboard as a canonical market-failure example—all tracing back to David 1985; Diamond 1997 and *The Economist*’s April 1999 article "The QWERTY myth" (Wikipedia, secondary source; original unread). This article’s contribution: the QWERTY case is the newest addition to the market-failure fable series, following beehives and lighthouses.
Fable’s concluding sentence calls the Dvorak keyboard "what beehives and lighthouses were for earlier market-failure fables," and "an example of market failure that will not withstand rigorous examination of the historical record" (quoted verbatim). Strongest rebuttal: textbook simplification is a necessary cost of dissemination; fables retain heuristic value. Response: heuristic value does not exempt narratives from audit; the fable smuggles in the policy prescription that "decentralized decisions err, centralized decisions can correct." Evidence that would overturn this judgment: if David’s original paper (unread) contained properly attributed Navy experimental citations and comparative evidence, the David account paraphrased here via L-M would require revision.
## What You Can Take Away
Standards and procurement decision-makers: Upon hearing 'We are locked into an inferior standard,' first conduct a three-condition audit: (1) Does a viable alternative exist? (2) Is the performance gap measurable? (3) Are switching costs insufficient to explain persistent adoption? Distinguish coordination lock-in (real, not necessarily inefficient) from inefficient lock-in (claimed, rarely proven); regulation and shifting focus can break coordination lock-in—but only after the performance gap has been demonstrated. General readers: When encountering the narrative 'We are trapped by history,' ask three questions: Where is the evidence? Who is the witness? Does the witness have a stake?
The keyboard is not a monument to market failure—it is a reminder of evidentiary standards.
If you’ve tried Dvorak or Colemak, how long did the switch take—and was it worth it? If you haven’t tried them, will this article change your mind? Share your thoughts in the comments.
## Scope and Boundaries
Access date: November 21, 2026. David’s 1985 'Clio and the Economics of QWERTY' (American Economic Review 75(2):332–337, JSTOR stable URL 1805621) remains behind a paywall and unread; its arguments are relayed via Liebowitz & Margolis (1995) and Wikipedia. The full text of L-M (1995) is drawn from Wayback Machine snapshots (January 9, 2001; December 8, 2000); the authors state the original manuscript is lost and the scans contain typographical errors. English quotations are transcribed verbatim from the snapshots; Chinese quotations represent translations. 'Rummelhart' is standardized as 'Rumelhart'; McCarthy’s letter date, OCR’d as 'October I4', is standardized as 'October 14'.
Wikipedia’s citation for 'Path dependence' incorrectly lists the Journal of Law and Economics (JLE) reference as '30:1–26'; Crossref (DOI 10.1086/467198) and the journal’s header ('vol. XXXIII (April 1990)') confirm volume 33, issue 1, pages 1–25. Wikipedia’s paraphrase 'about as good a design as any' appears nowhere in the full text and is therefore omitted. The ending page number for the Journal of Law, Economics, and Organization (JLEO) 1995 article remains unverified; we cite page 205 onward (confirmed by the Oxford University Press PDF). Strong’s '100 hours' figure follows Wikipedia’s phrasing, whereas the full text states 'far more than 25 days at 4 hours per day' (>100 hours); we adopt the full-text phrasing. The Dvorak home-row usage rate—70% (Wikipedia’s design claim) versus 67% (Fable’s empirical measurement)—is reported separately, not merged.
Tacoma’s 2,700 students, one-third time allocation, Strong’s personal grievances, and convergence across multiple layouts all derive from Wikipedia (secondary sources). The USB-C directive number and exact date remain unverified (eur-lex retrieval failed); only Wikipedia’s version is used (citing BBC, December 12, 2022). Betamax vs. VHS recording duration is unverified and omitted; only Liebowitz (2002, p. 41) is cited directly (via Wikipedia’s secondary citation; ISBN 978-0-8144-0649-6). The L-M link to the Microsoft antitrust case appears only on the Fable snapshot’s front page; no verification of historical testimony is attempted, so it is not mentioned.
Diamond (1997), The Economist (1999), Arthur (1989 and 1990) are all secondary citations via Wikipedia; original texts remain unread. Hossain & Morgan (2009), 'The Quest for QWERTY' (American Economic Review 99(2):435–440, DOI 10.1257/aer.99.2.435), bibliographic details verified but content unread; no arguments cited. The 'one-reversal condition' judgment remains untested—no data exists to support it.
## Sources
The Fable of the Keys, Stan Liebowitz and Stephen Margolis, Journal of Law and Economics 33(1):1–25, 1990, DOI https://doi.org/10.1086/467198; full text (Wayback snapshot, January 9, 2001) http://web.archive.org/web/20010109130300id_/http://www.utdallas.edu:80/~liebowit/keys1.html
Path Dependence, Lock-In, and History, same authors, Journal of Law, Economics, and Organization 11(1):205 onward, 1995, DOI https://doi.org/10.1093/oxfordjournals.jleo.a036867; full text (Wayback snapshot, December 8, 2000) http://web.archive.org/web/20001208011800id_/http://www.utdallas.edu:80/~liebowit/paths.html
Clio and the Economics of QWERTY, Paul David, American Economic Review 75(2):332–337, 1985, JSTOR stable URL 1805621 (original unread, arguments relayed via L-M 1995; bibliography verified against Wikipedia’s 'Path dependence' reference list)
Typewriter keyboard patent US2040248A, Dvorak and Dealey, issued May 12, 1936, https://patents.google.com/patent/US2040248A/en
QWERTY, Wikipedia, https://en.wikipedia.org/wiki/QWERTY
Dvorak Simplified Keyboard, Wikipedia, https://en.wikipedia.org/wiki/Dvorak_Simplified_Keyboard
Path dependence, Wikipedia, https://en.wikipedia.org/wiki/Path_dependence
High-definition optical disc format war, Wikipedia, https://en.wikipedia.org/wiki/High-definition_optical_disc_format_war
Common external power supply, Wikipedia, https://en.wikipedia.org/wiki/Common_external_power_supply
Re-thinking the Network Economy, Stan Liebowitz, 2002, New York, p. 41, ISBN 978-0-8144-0649-6 (book; quotation cited via Wikipedia’s 'Path dependence' reference list)