Talk:QWERTY

The source used as evidence of the claim that Sholes intended to slow operators down is an apocryphal story is a hastily slapped together term paper. Do we have something real? 2601:281:CB00:B610:4861:1825:F155:F08E (talk) 18:29, 15 September 2024 (UTC)

Fortunately, the page now describes this story as dubious, It also uses a Japanese source for the existence of the myth.Jbohmdk (talk) 23:39, 30 September 2025 (UTC)

At least in some European languages where existing national keyboard layouts existed before the IBM PC, the AltGr key added by IBM computer keyboards isn't there to access diacritical marks needed for the national language, but to access ASCII graphical characters not commonly needed in the national language, such as the $ sign, backslash and angled brackets, and also to access characters from other languages (such as typing German text on a French keyboard). For instance, many pre-1980 European national layouts replaced the $ sign on Shift+4 by a generic "currency sign", on the PC version of those keyboards, AltGr+4 produces the ASCII $-sign for use in computer languages such as BASIC programs and Spreadsheet function names. Later, newly invented characters like the Euro sign were added as new AltGr keystrokes on existing keyboard layouts. Jbohmdk (talk) 23:55, 30 September 2025 (UTC)

I'm unfamiliar with content addition policies, so if the following proposal contains violations (such as lack of verifiability) please let me know.

@JMF: In the articles for QWERTY, the typewriter, and the Sholes and Glidden typewriter, it is shown that the theory of QWERTY reducing the frequency of typewriters jamming by separating common English bigrams is contested by sources, noting the adjacency of keys that form common bigrams, such as ⟨e⟩ and ⟨r⟩. Despite this counterargument, typewriters have typebars arranged in a semicircular or circular typebasket, with row staggers allowing typebars on the same column to strike a shared point using parallel lines. This means that the ⟨E⟩ and ⟨R⟩ typebars aren't adjacent to each other, as can be seen on Figure 3 of U.S. patent 207559 and Figure 1 of U.S. patent 558428, where for circular-basket typewriters (such as the Sholes and Glidden typewriter) the typebars in the front of the basket alternate between keys of the top two rows (Q2W3E4R5T6Y7U8I9O-P...) and typebars in the back of the basket alternate between the bottom two rows (AZSXDCFVGBHNJMK,L...). Despite there being few reliable sources covering this order of typebars, there seems to be none that explicitly claim otherwise, the patent figures can probably serve as primary sources, and there are many images of typebars on the Internet that show their order (e.g. https://commons.wikimedia.org/wiki/File:Typewriters.jpg (note that this is QWERTZ on a semicircular-typebasket typewriter)); so I believe this isn't likely to be challenged for verifiability.

Although this disproves the assumption that the arrangement of typebars directly correspond to that of keys, of which many sources against the theory rely on, On the Prehistory of QWERTY shows that many common English bigrams have their letters a low distance apart on the typebasket, such as ⟨E⟩ and ⟨R⟩ with only one typebar of separation. For this to act against the theory of reducing typebar collisions being a design consideration of QWERTY, it relies on the assumption that having higher amounts of separation makes a noticeable difference and that it is possible to do so for more common bigrams.

On the other hand, the paper QWERTY and the search for optimality instead assumes that "all Sholes needed to do was separate the letter pairs by at least one type bar" to reduce the frequency of typebar jamming. These two sources (I couldn't find other reliable sources that have as many accurate and relevant points; please reply if there are more) arrive at opposite conclusions due to having opposite assumptions, so I believe in this regard Wikipedia should arrive at a neutral point of view, mentioning both assumptions but implying the accuracy of neither.

As for arguments supporting the theory of typebar jam reduction, QWERTY and the search for optimality mentions that the vowels ⟨E⟩ and ⟨I⟩ aren't included in the alphabetical "DFGHJKL" string on the third row since it could otherwise result in those vowels forming common adjacent-typebar bigrams with other letters, and that the ⟨A⟩ typebar is adjacent to ⟨Q⟩ and ⟨Z⟩, with "AQ"/"QA" and "AZ"/"ZA" being relatively uncommon bigrams. (The following wrapped in parentheses is not mentioned in the source, but might be logically deducible: All vowels excluding ⟨A⟩ and the common consonants ⟨R⟩ and ⟨T⟩ are also placed on the second row, where typebars are only adjacent to numbers and special characters on the first row.) Additionally, adjacent typebars in the string of the typebar arrangement on the back of the typebasket QAZSXDCFVGBHNJMK only form infrequent bigrams.

The source also demonstrates that another layout (starting with the sequence "XPMCHR") included in U.S. patents 558428 and 568630 filed by Sholes is nearly optimal in terms of reducing adjacent-typebar usage, with non-letter characters fully occupying the first row and the majority of the fourth row, separating letters on the second and third rows, except for two sequences of adjacent letters XQJVB and GZKY.

When analysed with a dataset consisting of Life on the Mississippi, the source also shows that QWERTY and XPMCHR result in around 43× and 173× less occurrences of adjacent-typebar bigrams compared to a layout of the alphabetical order.

Lastly, the source found that no modifications Remington made to the layout increased the frequency of adjacent-typebar usage, such as (the following examples of changes are unspecified in the source, but are mentioned in other sources such as Smithsonian Magazine and On the Prehistory of QWERTY which are both cited in the Wikipedia article): swapping ⟨C⟩ and ⟨X⟩ which changes SCDX to SXDC, avoiding SC; swapping ⟨M⟩ and ⟨N⟩ which changes HMJNK to HNJMK, avoiding NK/KN moving ⟨M⟩ to the right of ⟨N⟩ which forms JMK but no common adjacent-typebar bigrams; and swapping ⟨.⟩ and ⟨R⟩ which does not increase the frequency of adjacent-typebar letter bigrams since typebars of keys on the second row are only adjacent to typebars of the numbers and special characters on the first row. Thus, the source suggests that Remington was also aware of this design principle due to the high chance of random modifications breaking the principle by increasing frequency.

There are also sources that suggest Sholes had problems with typebar jamming when using alphabetical-order layouts (layout not specified in source) (e.g. The Early History of the Typewriter (1918) p.29), which could corroborate the jamming reduction theory, but I won't go in-depth since it might not be relevant.

In conclusion, based on the points above, I believe this and other related articles should describe this theory with neutral wording, rather than the current wording that implies it as an incorrect minority view. (To clarify, I don't know how to prove this theory's popularity, but out of around 50 (including potentially non-reliable) sources I found from several Google searches, 31 of them supported the jamming reduction theory. Also, despite the arguments listed above, I would still prefer neutral wording over implication of either side, since in-depth information about this seems to be scarce, especially when limited to reliable sources). Thanks for reading. had0j (talk) 10:33, 10 October 2025 (UTC)

Postscript: (I'll try not to make more edits to this post afterwards) As a further note, I think whether or not QWERTY successfully reduces typewriter jams should be accounted separately to whether or not it was designed with that intent. For example, this source I found after making this post (https://web.archive.org/web/20210802144615/https://conference.druid.dk/acc_papers/rluar3kk4i5t4gelr7j36gx9hma5.pdf) shows that semicircular-basket typewriters significantly increase the frequency of adjacent-typebar use on QWERTY, since it turns the order of typebars into QA2ZWS3XED4CRF5VTG6BYH7NUJ8MIK9... or QA ZWS XED CRF VTG BYH NUJ MIK OL (only including groups of letters); though it mentions that Sholes and James Densmore likely designed QWERTY for earlier circular-basket typewriters instead of later semicircular-basket ones, where the string of typebars had a 1212... (front) and 3434... (back) row pattern with the second row being the safest due to being adjacent only to numbers on row 1, rather than 23142314... which makes the second row the most dangerous due to being adjacent to letters on rows 3 and 4. Regardless of whether or not one typebar of separation is enough in practice and whether or not QWERTY reduces jams effectively, I think it's hard to disprove the unlikeliness of QWERTY's low adjacent-typebar frequency occurring by coincidence, unless the arrangement of typebars these papers assume is disproved or primary-source information that goes against this is discovered.

I also found another paper (https://strathprints.strath.ac.uk/92487/1/Galbraith-Kay-BH-2025-QWERTY-the-search-for-optimality-and-IP-complementarity.pdf, 2025) although it seems to built on QWERTY and the search for optimality but with a larger focus on path dependence rather than analysis of QWERTY's design considerations. Note that these three aforementioned papers all include the same author, so I'm unsure whether or not they should be treated like one source in terms of verifiability. — Preceding unsigned comment added by Had0j (talk • contribs) 16:14, 11 October 2025 (UTC)

So we have a single source (Yasuoka, Koichi; Yasuoka, Motoko,2011) that says it is an urban myth that is not supported by any contemporaneous records, despite meticulous searches. We have another single source that disagrees. Most of your essay above seems to based be your own analysis (aka WP: OR) which, despite being eminently logical and reasonable, is a kind of post hoc ergo propter hoc fallacy: this is where he started, this is where he ended up, these conditions would logically explain how it might have happened – but crucially does not show that this is what did happen. For our purposes, all that is irrelevant and we only have two choices:

1. give equal status to both perspectives and let the reader decide
2. evaluate the Galbraith-Kay papers to see if they are conjecture or evidence-based. Another general principle is that we don't question reliable academic sources, just report what they say. But we need at least to understand the foundations for its different view so that we can describe them. Which will probably take us back to option 1 but we give readers an informed choice.

So with that in mind, would you please study the Galbraith-Kay papers and let us know why they reach a different conclusion to Yasuoka and Yasuoka? And indeed read the Yasuoka and Yasuoka papers with the same challenge. --𝕁𝕄𝔽 (talk) 17:48, 11 October 2025 (UTC) revised to clarify my post hoc challenge. --𝕁𝕄𝔽 (talk) 18:13, 11 October 2025 (UTC)

Although some parts of the essay (such as the reason for ⟨R⟩ and ⟨T⟩ being on the second row, the swapping of ⟨C⟩ and ⟨X⟩ avoiding the "SC" bigram) are continuations based on original research, many of the points above are explicitly noted in QWERTY and the search for optimality (2013), such as:

• patent 558428 showing a figure "connecting the characters from the top two rows of the keyboard to the top half of the typebasket in alternating fashion ... Similarly, the levers connecting the lower two rows of the keyboard to the typebasket also do so in alternating fashion on the lower half of the typebasket" (pp.11–12)
• it being "extremely difficult at first sight to think of many words that contain letter pairs associated with that sequence [QAZSXDCFVGBHNJMK] read in either direction" (p.5)
• the exclusion of ⟨E⟩ and ⟨I⟩ from the alphabetical string DFGHJKL being "consistent with deliberate development and application of the infrequency principle" (p.8)
• non-letter characters acting as typebar buffers being "shown to great effect along the top of the typebasket where all the vowels were buffered from immediate proximity with any other letters (except for A which was largely neutralized by being buffered between Q and Z)" and "buffering in this fashion ensured that these vowels were not adjacent to any other letter on the typebasket" (pp.10–11)
• "by populating most of the bottom row of the keyboard with non-letters (mostly punctuation), one result is that every non-letter is fully employed as a buffer preventing letters from being adjacent to other letters on the typebasket, including the five vowels A, E, I, O and U. That still means that some letters have to be next to each other on the typebasket in two sequences XQJVB and GZKY." (p.13)
• "We compared how QWERTY on the Remington No. 7 would have performed against QV [XPMCHR] on Sholes's elliptical typebasket in these terms using a number of test texts. We show the results for LotM [Life on the Mississippi] as indicative text in Table 1. Table 1 also shows the results for the alphabetic or ABCDE format that Sholes originally started with." (p.14)
• "the ABCDE format had 10,214 occurrences of letter pairs on the printed page that were also next to each other on the typebasket, or about once every 14 words. By contrast, QWERTY would have encountered such events about once every 609 words while QV [XPMCHR] would have met the same fate about once every 2458 words." (p.14)
• changes by Remington made between 1872 and 1878 being "explicable in terms of consistent and systematic application to the rule" (p.8)
• "it [the "infrequency principle"] was that underlying principle embodied in QWERTY and ingeniously integrated with the hardware in Sholes's devices that really helped win the battle of the standards" (p.18)
• "this paper adds further weight of evidence that this [the "infrequency principle"] was indeed the principle that he [Sholes] and his partner Densmore used to design QWERTY in the first place" (p.19)

though most of these aforementioned points rely on one source, so there may be little weight to it.

Regarding the effects of the post hoc ergo propter hoc fallacy, I'm afraid I don't understand which parts of the argument you are referring to. If you mean the argument as a whole, I believe the conclusion I made may have caused a misunderstanding due to being vague; my argument is meant to demonstrate that there is a possibility (not guarantee) of the theory holding, thus I believe the article should describe this theory neutrally, and neither have a favoring nor disfavoring connotation.

As for the reason these two main sources come to different conclusions, after reading most of On the Prehistory of QWERTY pp.162–173 and QWERTY and the search for optimality pp.1–19 (but not yet QWERTY, the Search for Optimality and IP Complementarity or The QWERTY Problem), I believe the foundation of the argument in On the Prehistory of QWERTY pp.162–173 is that certain common English bigrams have their two typebars a low distance from each other, which would be unideal for jamming reduction if the frequency of jams with typebars of a lower distance apart is substantially higher than that of typebars of a higher distance apart; whereas for QWERTY and the search for optimality, the foundation is that based on the arrangement of typebars in a circular-basket typewriter, the QWERTY and XPMCHR layouts show remarkably low adjacent-typebar frequency and aspects that may potentially be intentional design choices, suggesting this "infrequency principle" (take the following with a grain of salt since I don't fully understand and it doesn't seem to be as relevant) may have been a trade secret complemented with patents to conceal the advantages discovered, demonstrating a case of intentional path creation, rather than a case of a suboptimal accident of history leading to path dependence. had0j (talk) 10:20, 12 October 2025 (UTC)

But fundamentally this is all conjecture. Very reasonable, credible and logical conjecture but still conjecture. No actual contemporaneous evidence. That's the big problem with a single source (excluding your google search that just confirms that the "clashing keys" theory is widely believed). If we had the support of other researchers, then it would be easy – we follow the expert consensus.

So it still seems to me that we must acknowledge foremost that there is no evidence for the clashing keys theory, citing the Japanese source that asserts very firmly that they searched diligently and found none. We can also add that the clashing keys theory is widely believed and appears to make sense in engineering terms (citing the Scottish source) but is not supported by any contemporaneous evidence. But let's leave it a few days to see if anyone else has a better idea. --𝕁𝕄𝔽 (talk) 12:41, 12 October 2025 (UTC)

While there seems to be no contemporaneous evidence supporting the explanation, there seems to be none that are against it either. I agree with citing both sources in summarising the basis of the conjecture and mentioning that no firm evidence regarding the accuracy of it is known of. I also agree with leaving this discussion open to see if there are other viewpoints before determining a conclusion. had0j (talk) 14:49, 13 October 2025 (UTC)

"Absence of evidence is not evidence of absence" strikes again. 𝕁𝕄𝔽 (talk) 15:31, 13 October 2025 (UTC)

This edit request has been answered. Set the |answered= parameter to no to reactivate your request.

At the top of the page it is stated that the qwerty keyboard is the "de facto standard on computers, as of 2025". We are now in 2026 so the year just needs to be changed. Thanks Keyboardwarrior2006 (talk) 06:17, 15 January 2026 (UTC)

 Done GearsDatapack (talk) 11:42, 15 January 2026 (UTC)

I was sent here by this Wiki link, but there is no section about Spanish: https://en.wikipedia.org/wiki/QWERTY#Spanish

The article mentions the Greek letters by dead keys, but no idea how, and the figure suggests only one minuscule---so I find using plural about 470% exaggerated. : ) ~2026-46488-8 (talk) 23:14, 21 January 2026 (UTC)

I have fixed the article so that QWERTY#Spanish now works as designed. It refers you to List of QWERTY keyboard language variants, which I suspect is what you really want, see List of QWERTY keyboard language variants#Spanish and List of QWERTY keyboard language variants#Greek.

The reference you describe to Greek in this article is about a specific type of keyboard, EurKEY: if you are interested, see that article. 𝕁𝕄𝔽 (talk) 12:10, 12 February 2026 (UTC)