DreymaR,

I have been lurking here for a bit but your post finally motivated me to join the board.  As this is my first post, I will say a litttle about myself first.  I am a Dvorak user and I would like to congratulate Shai for his layout, as it is the first one I have seen that comes close to Dvorak and Maltron.  It is certainly better than Dvorak in some respects, such as the lower same-finger ratio.

Now on to some responses to DreymaR's comments:

I don't understand the obsession with these shortcuts.  Personally, I don't use them that often.  And when I do use them, I remember them by their letter, rather than by their position.   I think that a change in the position of these shortcuts is just part of switching to another layout.  Even Colemak changes many control shortcuts.  If one set out as a goal to not change a single shortcut, then one would have to stick with Qwerty.

The concern many seem to have with this is that they want to be able to type these shortcuts with one hand.  However, it is unergonomic to type shortcuts with one hand.  The proper typing technique is to type control+combos with two hands, as this reduces strain on the hands.

Further, according to some research Apple computer did, keyboard shortcuts were actually slower than performing the same action with the mouse, even though the perception of users was that shortcuts were faster.

Dvorak must have placed the U, instead of the I, under the left index finger in order to facilitate the typing of the most common vowel-vowel digram in English: OU.  In English, the vowel-vowel digrams involving U are considerably more common than those involving I.  If you will note, the Dvorak layout appears to be based more on facilitating the typing of digrams than on facilitating the typing of individual characters.  This is why U is on the Dvorak home row, even though it is about the 12th most common letter in English.

Most typists type by typing combos of digrams and trigrams rather than by typing individual characters.  Dvoraks's design may be well-justified on account of this.

The only hard evidence I have seen for this comes from Lillian Malt (co-designer of the Maltron keyboard and layout).  She claimed that maximizing hand alternation was optimal for mechanical typewriters, but not for electronic keyboards.  However, she only used a sample size of two, and only examined a very limited set of digrams.

On the other hand, Peter Klausler's data suggest that hand alternation maximizes typing speed:

http://klausler.com/evolved.html

Now, his results may have been biased by his being a Dvorak user.  But then again, Lillian Malt's results could have been similarly biased, since she used Qwerty typists as her test subjects.

So overall, I consider the matter of hand alternation to be an unresolved question for electronic keyboards.  But everyone agrees that too little alternation is a bad thing.  It is clear that Shai tried to maximize hand alternation (within the limits of his other design constraints) by placing all the vowels, except A, under the right hand.

But Dvorak actually has a lower incidence of the sub-optimal outward rolls than do Colemak or Qwerty.  In this respect Colemak is actually worse than Qwerty.

I fed the following test-site the first two chapters of "The Adventures of Huckleberry Finn" by Mark Twain (from Project Gutenberg):

http://www.siteuri.ro/dvorak/index.php

I got the following results for hurdles/row-jumps and for smooth rolls:

Hurdles/row-jumping:

Outward rolls:

Note that Colemak has about a 33% higher incidence of row-jumping than Dvorak and about a 165% higher incidence of outward rolls. 

Thus, if you think inward rolls are important, Dvorak is your best choice in this regard, and Colemak is actually worse than Qwerty.  Row-jumping is another difficult maneuver that is more common in Colemak than in Dvorak. 

The higher row-jumping and outward roll ratios in Colemak are a by-product of both, the lower hand alternation, and the higher percentage of keystrokes assigned to the bottom row.

Colemak indeed is easier to learn for someone who who is already a Qwerty typist.  But Dvorak is easier to learn for someone new to typing.  It is quite neat to have all the vowels under the left home row for the purposes of learning. 

Personally, I quit Qwerty cold-turkey and did not find Dvorak hard to learn.  I was typing slowly, but acceptably, after 2 days.  I surpassed my Qwerty speed in like 3 weeks, although I was not a proper Qwerty touch typist, while the switch to Dvorak forced me (and guided me) to learn proper touch-typing.

I don't see any point in trying to make a new layout be easier to learn for Qwerty users, or in trying to make it compatible with Qwerty.  The way I see it, if you are going to go through the trouble of learning a new layout, you might as well go all the way and go for a complete optimization.

I read in a mailing list/forum, that according to Dvorak's book, the middle finger is strongest, but Dvorak placed more keystrokes on the index finger, because that finger is assigned 6 keys (versus 3 for the middle finger).

I actually did some finger exercises where I lifted weights with my fingers.  I found that my middle fingers were actually stronger than my index fingers.  But overall, my index, middle and ring fingers were fairly close in strength.  The pinky was significantly weaker, but not as weak as you might imagine.  My index fingers could do about 50% more repetitions than my pinkies could.  This is actually comparable to the finger ratios in Dvorak, where the index fingers are assigned about a 50% greater burden than the pinkies.

So I think the biggest issue with the pinkies may not be so much the strength, but the length.  This may be the real reason why the L key can be uncomfortable in Dvorak.

One additional consideration about the index fingers is that nowadays we use mice.  But the most common mouse clicks are also assigned to the index fingers, while the pinkies are usually unused.  Thus, Colemak may lead to worse finger balance when one takes mousing into account.

Shai cites mousing as something favoring Colemak's more even hand balance, as most people mouse with the right hand.  But as I stated above, Colemak is worse in terms of mousing finger-balance.  And personally, I am ambidextrous with mice and alternate my mousing hand from left to right, in order to even out the mousing burden.

I haven't really noticed this while typing since the distance to the top row is relatively small, although it seems to be true for longer reaches.  This is probably because most fingers actually share muscles with other fingers, but this only supports the idea of maximizing hand alternation, don't you think?  Maximizing hand alternation is the simplest way to reduce the incidence of complex moves by fingers on the same hand (such as row-jumping).

I don't find the R in Dvorak to be that annoying.  Only the L annoyed me, but this annoyance has faded since I started to use the Kinesis contoured keyboard.

And Qwerty places about 57% of the keystrokes on the left hand, including E! If what you say about the staggered rows on the traditional keyboard is true, it may be more reason to give a slight bias to the right hand (as in Dvorak), aside from the fact that most people are right-handed.

I agree that Colemak and Dvorak are both close to optimal for English.  Whether one or the other is better may depend on the question of hand alternation, which is apparently unresolved for electronic keyboards.

That aside, for me Dvorak wins on account of its superior compatibility.  As dominant as Qwerty is, I can at least count on Dvorak being pre-installed on Windows, Linux and Mac.  I don't need to install any drivers or layout files.  I can just switch the built-in layouts.   One can also get hard-coded Dvorak keyboards.

Perhaps someday Colemak will be built-in some operating systems, but until then Dvorak is less troublesome on this count.

To say 33% higher is to exaggerate the difference which in both cases is quite low compared to QWERTY. 
It would make more sense to say that incidence for row-jumping is  9.7 % of QWERTY for Dvorak and 12.9 % of QWERTY for Colemak.  Then you would need to estimates of likely standard deviation to say whether that 3% difference was significant or not.  In both cases a QWERTY touch typist is likely going to notice a difference in switching to Dvorak or Colemak, but between the alternate layouts I bet probably not.

So in you example Dvorak had about half the out outward rolls of QWERTY and Colemak had about 20 % more.  I actually like some of the outward rolls more than the inward rolls, so this may come down to "in the eye of the beholder".  Again comparing numbers without some estimate of variation is actually a fairly meaningless exercise.  With some understanding of variance and shape of the distribution you actually don't know if a few percent is really that significant in this case. 

After all the point of Colemak is not convert Dvorak users but to make the transition for good QWERTY touch typists to a more optimal keyboard much easier and faster.  To quote Shai

Colemak considers a lot of issues that Dvorak was not concerned about but they are both much more optimal when comparing to QWERTY.   Saying which is more optimal is going to probably end up as personal preference because variation of usage will I suspect always lead to a high enough variance that really good layouts will always beat each other in certain situations especially if you are considering languages other than English which is apparently one of Colemak's strengths.

aaackk.  I was in hurry and completely scanned over that it was just the first two chapters of Huck Finn that he was talking about.  I somehow got it in my mind that is was a larger data set.   I no expert in this area but I would thing you would need to have an ensemble of data sets to address the variance issue.  Without that all these comparisons of numbers is meaningless.  Also how do the moments and variance change when going from late nineteenth century novels to mid-twentieth century science fiction to current autobiographies or technical manuals, etc.

I mentioned this myself ;)

All I was saying is that the notion that maximizing hand alternation is not optimal lacks evidence.  The only hard evidence came from Lillian Malt.  Here is Malt's study:

http://www.ergo-comp.com/articles/keyboarddesign.html

I'm sure Shai will find it interesting.

But, just as Klausler was a Dvorak user, Malt used 2 Qwerty typists. 

So my point is that the question of whether hand alternation is good or not with electronic keyboards is currently unresolved, contrary to what the original poster suggested:

The above is what I was referring to.

No, when you use a large enough corpus, Colemak consistently beats Dvorak on some measures, such as the same-finger ratio and the percentage of keystrokes on the home row/home keys.  I was just pointing out that Dvorak seems to beat Colemak on row-jumping, on outward rolls (if you think outward rolls are sub-optimal), and hand alternation (if you think this is good).

Actually, Colemak is very similar in its general characteristics to the Maltron layout, with the exception that it doesn't place E under the thumb, for obvious reasons.  This is why Colemak is an impressively good layout.

Since my last post I tried a few other samples and found a few where colemak had a lower row jumping incidence than Dvorak.  I also saw one where Dvorak won over Colemak by a large margin.  I guess we just need a bigger corpus.

By the way, those are not my statistics as I do not run that site; I am not Andrei Stanescu.

Like you, I am interested in keyboard design.  I was able to derive some interesting statistics from the mentioned site.

From a sample of text entered I got the following results for outwards finger rolls:

Same hand:

Same finger:

Now, I wondered what would be the ratio of inwards rolls, and the ratio of inwards to outward rolls.  If you think about it there are only a few possible outcomes for the typing of a digram:

1) They alternate hands
2) They are typed with the same finger
3) They roll inwards
4) They roll outwards

Since the site provides us with the hand alternation, same finger and outward roll rations, we can thus calculate the inward roll ratio, since the four factors should add up to 100% (barring rounding errors).

Thus I calculate the following ratio for inward rolls using the equation: (same hand) - (same finger) - (outward rolls) = (inward rolls).  This equation must hold, because if the digram does not change hands, there are only three possible outcomes: an inward roll, an outward roll, or same finger.  Thus, the inward rolls, outward rolls, and same finger ratios must add up to the same hand ratio.

I obtain the following results for inward rolls:

If I have not made a mathematical error, it would appear that Colemak and Dvorak have comparable rates of inward rolls, but Dvorak has far fewer outward rolls (5% vs 11%). 

We can then calculate the ratio of inwards to outwards rolls by dividing the inward roll ratio by the outward roll ratio:

Can't say I've read this whole thread (yet), but someone brought this part to my attention:

Vowel-vowel digrams with "I" are at least as common as those with "U"

OU is very common, but the other U-digrams are not.  Overall, I found I-digrams to be 30% more common.  Maybe OU would be even more common if my source text had more British English in it (since British English uses OU a lot, whereas American English uses just O in words like "colo(u)r").

Data from when I counted digrams for a typing-tutor program:

rank  digr count
---------------
21   (ou, 670)
159   (ue, 103)
193   (au,  73)
261   (eu,  38)
250   (ua,  44)
636   (uo,   2)
---------------
SUM:       930

rank  digr count
---------------
40   (io, 393)
59   (ie, 309)
106   (ai, 187)
139   (ia, 133)
143   (ei, 129)
208   (oi,  66)
---------------
SUM:      1217

I/U = 1217/930 = 1.308

If we were even to declare a tie based on just the digrams, "I" is still several times more common than "U" overall.  So.. I think Dvorak made a mistake by putting the U under the index finger (though I'm sure his intentions were good..).

Update: Andrei Stanescu has put my emails to him on his site with links from his Dvorak page, in a gesture of very polite disagreement!  :)

No idea, fsiefken. I have a feeling the august Mr. Dvorak must've had a reason for it although I feel it must've been a wrong one.

I don't think you're calculating the digram frequencies right though. You'd have to account for all digrams and not just vocal-to-vocal ones. And that'd make for a rather hefty egrep command line!

Are you aware about mouse-only editing? Without pop-up menus, I mean. Yes, you can copy/cut/paste using just your mouse and its buttons. I tried it once, but I don't remember the procedures anymore, sorry.

That Apple quote was originally written in 1989 about people that were still refusing to use mice at all.  It's quoted often, but usually with little understanding of what it actually means.

Jeff Atwood clears it up nicely: https://blog.codinghorror.com/revisitin … ouse-pt-1/