So, to summarize, the original typewriter layout (1868) was alphabetical, with the top row (A-N) going left-to-right and the bottom row (O-Z) going right-to-left. Then (1870), the vowels (including Y) were pulled out and put in a separate top row. After that (by 1872), changes were made in order to better support the use case of people receiving Morse code, and that's when we finally start to see something that looks like QWERTY. Additional changes later got it to the modern form, but by 1872 something QWERTY-like was in place.
And yeah -- if you look at the bottom two rows of a QWERTY keyboard, you can still see what remains of that alphabetical ordering, being left-to-right on one row followed by right-to-left on the row below!
In the next century, researchers will discover that the GUI wasn't designed to make computing harder by forcing people to find cryptic little symbols, randomly arranged on the screen, and break routine operations into tiny sequences of manual steps. And it wasn't called a "personal computer" because it turned each person into a computer.
"A common misunderstanding is that GUIs were invented to give ancient computers enough time for processing by slowing down user input speed. However our research shows that counterintuitively input latency was better at the time when GUIs were invented and then gradually got worse..."
Ha. so the reason that I is next to 8 is that early typewriters used the I as a 1(no independent 1 key) and the morse transcription company wanted to type years(1871) quickly. I love it.
"The legend was referred by Prof. James V. Wertsch,[22, 23] a professor of the Department of Psychology, Clark University, then it was regarded as an established theory in the field of psychology. "
The reproducibility crisis struck early, it seems.
Humans do not have fins, but Micheal Phelps can still cut through water. That elites can thrive is not a compelling argument when most people just want technology to get out of the way.
An alternative layout with commonly used symbols on the home row makes the QWERTY deficiencies immediately apparent. Significantly less effort required for writing prose when using something like DVORAK.
I really don't think people who type slowly do so because of QWERTY. Anecdotally, my dad basically isn't able to develop muscle memory for the key locations and will frequently revert to the "scan and then press with one finger" method. You could give him any layout and he would still type slowly. Pretty sure even an alphabetical order would trip him up, because it'd need to be broken into multiple rows, so he'd need muscle memory again.
And while this is speculative, given how close typing speeds seem on a cursory search between layouts, this suggests to me that the vast majority of the performance comes not from the layout, but from touch typing and effective use of multiple hands and fingers at the same time. All layout agnostic skills.
This is not to say that on an input method level, things cannot be further improved. I sometimes see stenography [0] related software and demonstrations on YouTube for example. It also isn't to say that there cannot be a benefit health wise (i.e. ergonomically) to alternative layouts. It's just that for speed I'm not convinced it affects much, and so I think it's the wrong thing to try and change. Especially considering that sometimes things that are suboptimal can be better by being the standard.
Stenographers have special keyboards that type phrases, not keys, so their speed is also extremely fast due to only a subset of character combinations needed.
I would agree with something like that, that e.g. it's only useful when you have a set text ready that you want typed in.
Personally I clock in around 110 wpm at my fastest (in English), and I'll only reach that in practice when I have intense disagreements with people over the internet (colleagues, friends or randoms). Outside of that, it's barely useful.
Dvorak is way more comfortable and is very fast for me, but its emphasis on alternating hands causes me to frequently invert pairs of letters when typing fast, probably because my left and right brain isn't perfectly coordinated.
I have considered switching to Colemak, which is supposed to have less of that, but as a 100+ wpm typer of Dvorak it really is diminishing returns.
Colemak has flaws of its own, bandaged over with "Mod-DH", but I'd say just skip it and go to Workman. It aims for 50/50 hand usage unlike other layouts, but it does not encourage the constant alteration like Dvorak. Common bigrams like "th" and "en" are easy to type with one hand, encouraging smooth finger-roll motions. It doesn't have Dvorak's `ls -l` problem.
I personally went QWERTY -> Dvorak -> Workman. While I spent under 2 years with Dvorak, I have been using Workman for over 3 years now and think I'll likely stick with it. There are other crazy layouts like QGMLWB and Halmak, but I think their higher on-paper efficiency comes at the cost of other things, and Workman seemed like it struck the best balance to me. I had comfort/ergonomics in mind and was trying to get away from some hand pain which seemed to only switch hands going from QWERTY to Dvorak, but is now gone with Workman.
Thanks for sharing your experience. Hearing that you went from being proficient at Dvorak to Workman and in the end prefer Workman motivates me to try it.
Dvorak is awful for the shell, but actually not bad for vi. Do you have any experience with Workman and vi?
Yes, I do use neovim as my main editor, as well as applications like ranger that use similar binds. While the hjkl placement is a bit nonsensical like in most non-qwerty layouts, in the end you're just pressing letters to do actions, and you get used to it the same way you get used to typing in the new layout in general. I don't do any special rebinding with the layout in mind. I also tend to argue that if you're advanced in vim, you won't use hjkl as much anyway since you'll be moving by words or paragraphs or using the search instead.
> think we're I to adopt my split columnar ergonomic keyboard now I would just stick with qwerty. I think the keyboard helped me more than the layout.
Doing just this is actually why I changed layouts again. I built a Pinky4 split columnar stagger keyboard. A few months later I was used to it, but still had hand pain using Dvorak. The layout really is what made a difference for me. I still use that keyboard, but with Workman now.
I got into alternate keyboard layouts and developed my own (roughly an optimized NIRO). When I tried using it on my small Surface Go I found that my fingers would 'jam' typing letters close together, so I leave that in QWERTY so it happens much less.
Everyone has their own style I would assume. I've been on a PC for 25ish years now and I still type with 3 or 4 fingers almost all the time. I've tried to be more disciplined but I manage 100ish corrected WPM and it gets the job done, I never find that I'm limited by my typing speed.
I can believe it. For a physical keyboard, I would much prefer a layout with a DVORAK-like home row, but probably not for mobile. The imprecision of touch and swiping text entry likely do markedly worse when high frequency characters are on top of each other.
I believe software does that, in that characters are disambiguated in part based on previous characters, even without swipe typing. That means each character is mapped to a secondary keyboard layout where certain keys' "hitbox" are enlarged.
We've had a few conversations on HN about how QWERTY may not be optimal for mobile swiping, but it's better than the layouts optimized for typing. Dvorak, with all the vowels on the left hand home row, would be terrible for swiping, what with all the English words just a vowel away from each other.
As it is the UIO vowel complex is a frequent problem with swiping. I'd like to see U and O moved away from each other, in theory.
Yes, normal Dvorak is not great for use on a smartphone, unless you usually type with both thumbs. For one thumb, especially swiping, I've found one-handed Dvorak for right hand to be a great fit. Works well for me on Multiling O Keyboard with vowel omission enabled.
I found the opposite when I went through a split keyboard phase and did Colemak-DH layout. As far as english goes I found that a lot of words had a smoother "flow" when it comes to finger location.
In the end I went back to a regular qwerty because my WPM on split keyboards/colemak-DH was considerably worse even after many months of practice.
Even at a bit less than 200WPM, my physical ability to press keys is seldom the limiting factor for composing prose, let alone for coding. I would suspect that this is the case for most halfway competent keyboard users using more than three fingers, and that QWERTY is “good enough”.
The keyboard arrangement was incidentally changed into QWERTY, first to receive telegraphs, then to thrash out a compromise between inventors and producers, and at last to evade old patents.
Interesting article. The connection to Morse code makes a lot of sense (C being similar to S). The requirement to move I below 8 to type 1870 or 1871 quickly is hilarious in retrospect. At the time who could have known the decision to focus on efficiency for the coming decade could be so enduring?
This still doesn’t explain one of my biggest peeves with keyboards:
Why are the keys angled up and to the left — for both hands?
Was this to solve the type bar jamming?
Or is that also an urban legend?
I know there are modern keyboards that solve this, by either splitting the keyboard and angling in the natural direction of your fingers (so to the right for the left hand, and vice versa), or just ortholinear keyboards that have straight rows of keys (but still angled ergonomically).
But that ridiculousness has lived on, such that even “economic” split keyboards will still angle both sides to the left.
The key levers had to be staggered on their way back to the type bars so they wouldn't hit each other. The angle isn't ergonomic in origin but mechanical. You probably haven't ever seen a mechanical typewriter, but maybe you can find one to experiment with, or at least a YouTube video showing the workings.
Very interesting article. I don’t understand, however, how shorthanders used typewriters for short-writing. The figure on p.168 (above fig. 9) is not explanative.
The numbers above the words indicate which finger (index, middle, ring) is used to press a key, the letters below indicate the hand (left or right). Basically a precursor of touch typing that doesn't use the little fingers and doesn't always use the same finger for the same key.
The actual shorthand would be written on paper, with the typewriter being used to expand it to a more readable form.
In original Morse code, S was "dot dot dot", E was "dot", and Z was "dot dot dot space dot". So if you were transcribing Morse code back then, it was ambiguous if you received a "Z" or "SE". To handle this, the "S" key was moved from the right side to a position between "Z" and "E" so you could quickly type either "Z" or "SE" once you figured out what was sent. Modern (i.e. 1865) Morse code changed "Z" to "dash dash dot dot", eliminating the ambiguity.
Yes, but how does it help to have the ambiguous characters close. I could understand if they had left out Z completely similar to early typewriters having no 0 but using the O character instead. The ambiguous Z would always be written as SE.
If they wanted to resolve the ambiguity it makes sense to move SE and Z further apart so the operator does not easily make a typing mistake in addition to the recognition mistake easily.
They had to wait for additional characters to make a good guess as to whether it was Z or SE. So they thought that having your fingers already positioned close to both outcomes would allow you to handle that delay better. Maybe it allowed you to catch back up faster, or made it easier to handle both of those threads in your brain and nerves? Perhaps whatever letters came after Z/SE were unlikely to also be Z/SE, so it made it more likely your right hand would be typing the next character?
In Morse telegraphs, some words are abbreviated and so characters have a different frequency distribution. For example RECEIVED may be shortened to RX, making X much more frequent than otherwise.
Because the smart way (with "M" there instead) was patented. From TFA (emphasis mine):
> In order to evade the patents that were assigned to the Type Writer
Company, WS&B slightly changed the design of No. 2 including the keyboard
arrangement (Fig. 9), where M was moved next to N, and C was exchanged
with X. It was the QWERTY keyboard arrangement as seen nowadays.
If we consider 100-year-old prose, we used the semicolon in places you see a dash now; like this - not this. It was important; it was located with period and comma and apostrophe.
Isn't this based on letter frequency? In languages like German, Z is a fairly common letter. Y not so much. Whereas in English it's the opposite. They put the more commonly used letter in the center.
So, to summarize, the original typewriter layout (1868) was alphabetical, with the top row (A-N) going left-to-right and the bottom row (O-Z) going right-to-left. Then (1870), the vowels (including Y) were pulled out and put in a separate top row. After that (by 1872), changes were made in order to better support the use case of people receiving Morse code, and that's when we finally start to see something that looks like QWERTY. Additional changes later got it to the modern form, but by 1872 something QWERTY-like was in place.
And yeah -- if you look at the bottom two rows of a QWERTY keyboard, you can still see what remains of that alphabetical ordering, being left-to-right on one row followed by right-to-left on the row below!
In the next century, researchers will discover that the GUI wasn't designed to make computing harder by forcing people to find cryptic little symbols, randomly arranged on the screen, and break routine operations into tiny sequences of manual steps. And it wasn't called a "personal computer" because it turned each person into a computer.
"A common misunderstanding is that GUIs were invented to give ancient computers enough time for processing by slowing down user input speed. However our research shows that counterintuitively input latency was better at the time when GUIs were invented and then gradually got worse..."
This reminds me of Motel of the Mysteries :^)
And paid subscriptions are not even necessary for offline apps
Ha. so the reason that I is next to 8 is that early typewriters used the I as a 1(no independent 1 key) and the morse transcription company wanted to type years(1871) quickly. I love it.
"The legend was referred by Prof. James V. Wertsch,[22, 23] a professor of the Department of Psychology, Clark University, then it was regarded as an established theory in the field of psychology. "
The reproducibility crisis struck early, it seems.
Whatever its intent, QWERTY definitely hasn't impeded the fastest typists, who can regularly exceed 200wpm these days.
Odd to see no mention of the Linotype layout, also known as the "Etaoin Shrdlu", given that was also a common competing keyboard layout in that era.
Humans do not have fins, but Micheal Phelps can still cut through water. That elites can thrive is not a compelling argument when most people just want technology to get out of the way.
An alternative layout with commonly used symbols on the home row makes the QWERTY deficiencies immediately apparent. Significantly less effort required for writing prose when using something like DVORAK.
I really don't think people who type slowly do so because of QWERTY. Anecdotally, my dad basically isn't able to develop muscle memory for the key locations and will frequently revert to the "scan and then press with one finger" method. You could give him any layout and he would still type slowly. Pretty sure even an alphabetical order would trip him up, because it'd need to be broken into multiple rows, so he'd need muscle memory again.
And while this is speculative, given how close typing speeds seem on a cursory search between layouts, this suggests to me that the vast majority of the performance comes not from the layout, but from touch typing and effective use of multiple hands and fingers at the same time. All layout agnostic skills.
This is not to say that on an input method level, things cannot be further improved. I sometimes see stenography [0] related software and demonstrations on YouTube for example. It also isn't to say that there cannot be a benefit health wise (i.e. ergonomically) to alternative layouts. It's just that for speed I'm not convinced it affects much, and so I think it's the wrong thing to try and change. Especially considering that sometimes things that are suboptimal can be better by being the standard.
[0] https://en.wikipedia.org/wiki/Stenotype
I can average over 85wpm when typing English sentences in typing games.
But outside that, I find I can’t really think meaningfully and simultaneously type that quickly for a sustained amount of time.
Maybe very fast typing speeds are more useful to stenographers?
Stenographers have special keyboards that type phrases, not keys, so their speed is also extremely fast due to only a subset of character combinations needed.
There are a small number of people who use standard keyboards for stenography: https://www.openstenoproject.org/plover
That said, I don't know if any work as professional stenographers.
I would agree with something like that, that e.g. it's only useful when you have a set text ready that you want typed in.
Personally I clock in around 110 wpm at my fastest (in English), and I'll only reach that in practice when I have intense disagreements with people over the internet (colleagues, friends or randoms). Outside of that, it's barely useful.
[dead]
Dvorak is way more comfortable and is very fast for me, but its emphasis on alternating hands causes me to frequently invert pairs of letters when typing fast, probably because my left and right brain isn't perfectly coordinated.
I have considered switching to Colemak, which is supposed to have less of that, but as a 100+ wpm typer of Dvorak it really is diminishing returns.
Colemak has flaws of its own, bandaged over with "Mod-DH", but I'd say just skip it and go to Workman. It aims for 50/50 hand usage unlike other layouts, but it does not encourage the constant alteration like Dvorak. Common bigrams like "th" and "en" are easy to type with one hand, encouraging smooth finger-roll motions. It doesn't have Dvorak's `ls -l` problem.
I personally went QWERTY -> Dvorak -> Workman. While I spent under 2 years with Dvorak, I have been using Workman for over 3 years now and think I'll likely stick with it. There are other crazy layouts like QGMLWB and Halmak, but I think their higher on-paper efficiency comes at the cost of other things, and Workman seemed like it struck the best balance to me. I had comfort/ergonomics in mind and was trying to get away from some hand pain which seemed to only switch hands going from QWERTY to Dvorak, but is now gone with Workman.
Thanks for sharing your experience. Hearing that you went from being proficient at Dvorak to Workman and in the end prefer Workman motivates me to try it.
Dvorak is awful for the shell, but actually not bad for vi. Do you have any experience with Workman and vi?
Yes, I do use neovim as my main editor, as well as applications like ranger that use similar binds. While the hjkl placement is a bit nonsensical like in most non-qwerty layouts, in the end you're just pressing letters to do actions, and you get used to it the same way you get used to typing in the new layout in general. I don't do any special rebinding with the layout in mind. I also tend to argue that if you're advanced in vim, you won't use hjkl as much anyway since you'll be moving by words or paragraphs or using the search instead.
When I was deciding on a layout many moons ago, I looked at Colemak and Workman but decided on an offshore of those... Norman.
It seems that keyboard layouts have evolved quite a bit since those days.
I think we're I to adopt my split columnar ergonomic keyboard now I would just stick with qwerty. I think the keyboard helped me more than the layout.
> think we're I to adopt my split columnar ergonomic keyboard now I would just stick with qwerty. I think the keyboard helped me more than the layout.
Doing just this is actually why I changed layouts again. I built a Pinky4 split columnar stagger keyboard. A few months later I was used to it, but still had hand pain using Dvorak. The layout really is what made a difference for me. I still use that keyboard, but with Workman now.
> causes me to frequently invert pairs of letters when typing fast
I am a Dvorak user and this happens to me too. I thought it happens when I'm stressed out.
I got into alternate keyboard layouts and developed my own (roughly an optimized NIRO). When I tried using it on my small Surface Go I found that my fingers would 'jam' typing letters close together, so I leave that in QWERTY so it happens much less.
Yeah, there is a conflict between "use your very best fingers for everything" versus "spread the work across fingers so that it gets done sooner."
I tend to favor the latter.
Everyone has their own style I would assume. I've been on a PC for 25ish years now and I still type with 3 or 4 fingers almost all the time. I've tried to be more disciplined but I manage 100ish corrected WPM and it gets the job done, I never find that I'm limited by my typing speed.
I can believe it. For a physical keyboard, I would much prefer a layout with a DVORAK-like home row, but probably not for mobile. The imprecision of touch and swiping text entry likely do markedly worse when high frequency characters are on top of each other.
I'm sure somebody has created a software keyboard where the more used characters are simply larger than the less used ones.
I believe software does that, in that characters are disambiguated in part based on previous characters, even without swipe typing. That means each character is mapped to a secondary keyboard layout where certain keys' "hitbox" are enlarged.
We've had a few conversations on HN about how QWERTY may not be optimal for mobile swiping, but it's better than the layouts optimized for typing. Dvorak, with all the vowels on the left hand home row, would be terrible for swiping, what with all the English words just a vowel away from each other.
As it is the UIO vowel complex is a frequent problem with swiping. I'd like to see U and O moved away from each other, in theory.
Yes, normal Dvorak is not great for use on a smartphone, unless you usually type with both thumbs. For one thumb, especially swiping, I've found one-handed Dvorak for right hand to be a great fit. Works well for me on Multiling O Keyboard with vowel omission enabled.
I found the opposite when I went through a split keyboard phase and did Colemak-DH layout. As far as english goes I found that a lot of words had a smoother "flow" when it comes to finger location.
In the end I went back to a regular qwerty because my WPM on split keyboards/colemak-DH was considerably worse even after many months of practice.
Pretty sure Micheal Phelps has fins though
Even at a bit less than 200WPM, my physical ability to press keys is seldom the limiting factor for composing prose, let alone for coding. I would suspect that this is the case for most halfway competent keyboard users using more than three fingers, and that QWERTY is “good enough”.
> efficiency for the coming decade
They were future proofing though: the "I" is between "8" and "9", so they were optimising for the next 130 years, not merely coming decades.
This still doesn’t explain one of my biggest peeves with keyboards:
Why are the keys angled up and to the left — for both hands?
Was this to solve the type bar jamming?
Or is that also an urban legend?
I know there are modern keyboards that solve this, by either splitting the keyboard and angling in the natural direction of your fingers (so to the right for the left hand, and vice versa), or just ortholinear keyboards that have straight rows of keys (but still angled ergonomically).
But that ridiculousness has lived on, such that even “economic” split keyboards will still angle both sides to the left.
The key levers had to be staggered on their way back to the type bars so they wouldn't hit each other. The angle isn't ergonomic in origin but mechanical. You probably haven't ever seen a mechanical typewriter, but maybe you can find one to experiment with, or at least a YouTube video showing the workings.
Very interesting article. I don’t understand, however, how shorthanders used typewriters for short-writing. The figure on p.168 (above fig. 9) is not explanative.
The numbers above the words indicate which finger (index, middle, ring) is used to press a key, the letters below indicate the hand (left or right). Basically a precursor of touch typing that doesn't use the little fingers and doesn't always use the same finger for the same key.
The actual shorthand would be written on paper, with the typewriter being used to expand it to a more readable form.
Thank you. I mistakenly thought the typewriter was used to type shorthands.
Or shorter: Qwerty wasn't designed.
I don't follow the connection to Morse. Can someone summarize their argument in a comprehensive way?
In original Morse code, S was "dot dot dot", E was "dot", and Z was "dot dot dot space dot". So if you were transcribing Morse code back then, it was ambiguous if you received a "Z" or "SE". To handle this, the "S" key was moved from the right side to a position between "Z" and "E" so you could quickly type either "Z" or "SE" once you figured out what was sent. Modern (i.e. 1865) Morse code changed "Z" to "dash dash dot dot", eliminating the ambiguity.
Yes, but how does it help to have the ambiguous characters close. I could understand if they had left out Z completely similar to early typewriters having no 0 but using the O character instead. The ambiguous Z would always be written as SE.
If they wanted to resolve the ambiguity it makes sense to move SE and Z further apart so the operator does not easily make a typing mistake in addition to the recognition mistake easily.
They had to wait for additional characters to make a good guess as to whether it was Z or SE. So they thought that having your fingers already positioned close to both outcomes would allow you to handle that delay better. Maybe it allowed you to catch back up faster, or made it easier to handle both of those threads in your brain and nerves? Perhaps whatever letters came after Z/SE were unlikely to also be Z/SE, so it made it more likely your right hand would be typing the next character?
The first typewriters were for telegraph operators turning morse code into written letters.
In Morse telegraphs, some words are abbreviated and so characters have a different frequency distribution. For example RECEIVED may be shortened to RX, making X much more frequent than otherwise.
Now why was fricken Semicolon given the prime real-estate of being an unshifted key?
Because the smart way (with "M" there instead) was patented. From TFA (emphasis mine):
> In order to evade the patents that were assigned to the Type Writer Company, WS&B slightly changed the design of No. 2 including the keyboard arrangement (Fig. 9), where M was moved next to N, and C was exchanged with X. It was the QWERTY keyboard arrangement as seen nowadays.
If we consider 100-year-old prose, we used the semicolon in places you see a dash now; like this - not this. It was important; it was located with period and comma and apostrophe.
It's still used for complex lists: https://www.grammar-monster.com/lessons/semicolons_in_lists....
To make it easier to code in C and Pascal ;-)
I'm more interested why some of us have qwertZ instead of Y there :D
(Z an Y are swapped... mostly a non issues, except with some games, where Z and X are some gameplay controls, and we have a Y down there)
Isn't this based on letter frequency? In languages like German, Z is a fairly common letter. Y not so much. Whereas in English it's the opposite. They put the more commonly used letter in the center.