I don't believe that hand alternation is the most important factor for speed. In fact many modern alternative layouts (e.g. Capewell, Maltron, Arensito, Qwerak) do not try to maximize hand alteration.

From the Arensito design principles:

From the Qwerak design principles:

From the Capewell design principles:

Putting all the vowels on one side means that you need to place vowels on the index finger, and thus highly increase same-finger ratio, or be forced to put common letters in weak positions. I've found that same-finger typing causes slowdowns more than any other factor.

I think a better way to look at looking at it is whether you can chord the keys at the same time, e.g. you can't really hit 'M','Y','P' (on QWERTY) as a chord, so it slows down typing. However you can easily hit 'J','K','L' as a chord, so you don't need to prepare the hand, because it's already prepared.

From my personal experience typing on Colemak is a little bit faster than Dvorak, but YMMV.

An essential thing to bear in mind is that design considerations that were important when Dvorak was putting together his keyboard back in the 1930s, are not necessarily so applicable today. I think that hand alternation is one of these.

Back in Dvorak's day, typewriters were mechanical devices and it took significant effort to press a key. In that context, hand alternation makes a lot of sense, since you needed quite lot of hand motion to press a key. Same hand could be seen as almost as bad as same finger in this light.

These days, almost no effort is required to press a key, hence whole hand motions are not necessary, and so I think hand alternation is not nearly so important. In fact, I'm pretty sure I can type two keys that are close together with the same hand (with different fingers of course), significantly faster than I can hit two keys with different hands. It's easier to "roll" two fingers on the same hand than two fingers on different hands. This kind of rolling probably wouldn't have worked with early mechanical typewriters.

Dvorak's layout has several problems, e.g. (a) excessive same finger on the left index finger, caused by putting a vowel on the index finger home key and surrounding it with consonants, and (b) awkward placement  of L and S, (c) R not being on the home row. In contrast I can't really find any serious deficiencies with the Colemak layout. The only perhaps controversial compromises to retaining compatibility with QWERTY are really very minor.

Cheers,

Sy

Yeah - I found it while I was browsing for something else and didn't notice the timestamp. Still it's a timeless debate!

Just getting started on this whole learning Colemak thing, after dabbling with Dvorak for a few weeks and feeling that it wasn't quite right. So far, I really like what I see.

Sy

I love the right-hand digraphs in Dvorak and I don't like the 3 and 4 letter one-hand combinations in Colemak during which the other hand has nothing to do. Dvorak of course has many problems and I am using Colemak now, but I find it somewhat weird for various reasons.

By the way, placing the vowels on one side also makes the layout easier to learn. Not only it's easier to remember the placement of the letters, but also (at least for me) it's easier to identify the one-hand combinations so you can type them almost simultaneously (by rolling your fingers) rather than separately one after the other. Of course, you can learn to do that with any layout with a lot of practice.

Sholes' typewriter was good for reducing jams. But not much else. A keyboard could be designed that both reduces jams and is efficient. If I wanted to write a program to do it, I'd just need a simple change: it costs points for two keypresses to be in the same column or in adjacent columns.

Yeah, I'll do that.

Here are the results of a quick run (i.e. not very accurate) with the cost of a potential jam weighted at somewhat more than same finger, and double that of jumping over the home row.

Here are the results.

score: 62.334, effort: 105190084.8, handcount: 123726.0, fingercount: 61636.0, jumphomecount: 2596.0, rowchangecount: 394605.0, inrollcount: 230720.0, outrollcount: 345499.0, jamcount: 5649781.0

 x p f u . ' c d g v
 n a e i h l r t s o
 ; , k y z w m b j q

I don't think the jam count is correct, because it was the same after every permutation. And yes, I know, this layout has a lot of issues like having a vowel on the index finger. I was doing a quick run. But this can give you a general idea.

Here is a version I made by moving around a few things.

 x f u p . ' c d g v
 i a e n h l r t s o
 ; , k y z w m b j q

The same finger is a lot better, and I don't think the jam count will be a lot higher.

Well jamcount is broken. I will see if I can fix it.