4 ms·

Reading through his blogs, he keeps saying that one of the advantages of his layout is intentionally alternating keys between hands, and slags colemak for being

by gwd 7y ago

Reading through his blogs, he keeps saying that one of the advantages of his layout is intentionally alternating keys between hands, and slags colemak for being "really bad" at that. But when I was reading up about the colemak, they claimed that was done on purpose -- that in fact, at high speeds, coordinating alternating keystrokes between hands is very difficult, and that it's actually a lot more efficient to have "rolling" patterns where a single hand can type a sequence of characters in a smooth motion. I can definitely say that the "inter-hand coordination" is the source of an awful lot of my typing errors.

equalunique 7y ago
>at high speeds, coordinating alternating keystrokes between hands is very difficult It's for this reason I believe the traditionally wide spacebar should be split up into a number of different keys. Having two sets of Shift, Control, Alt, and the Windows key on either side of the keyboard seems less sensible than one set that can be accessed by either thumb. I for example use a custom layout that splits the Spacebar into both Space and Shift. Now I don't have to shift my hands to the left/right to capitalize a letter, make an exclamation/question mark, make a double quote, etc.
vages 7y ago
Made the same observation! And the rolling patterns are better if they go from in the pinkie-to-index direction than the other way around. An example of how Halmak does not take this into account is how the h and t keys are placed. It immediately jumped out at me as a Dvorak user.
networkimprov 7y ago
The myth of alternating-hand advantage probably dates to the era of mechanical typewriters, when hammers arriving from opposite sides of the machine were less likely to jam.
- mkl 7y ago
  Alternating hands was certainly better for my hand joint issues. I found Dvorak not really faster, but more comfortable, and the alternating hands is a big part of that. Alternating hands is much better than consecutive keys far apart on the same hand, and way better than consecutive keys on the same finger.
- candiodari 7y ago
  Actually they were more likely to jam. They hit the same place and if you hit a left and right key at the same time, they'd invariably jam, which necessitated going in and manually prying them back free (the keys sort-of fit into each other)
logicallee 7y ago
>that in fact, at high speeds, coordinating alternating keystrokes between hands is very difficult, and that it's actually a lot more efficient to have "rolling" patterns where a single hand can type a sequence of characters in a smooth motion. This seems like a particularly easy hypothesis to test (if I weren't on mobile I would try coding it up). From a dictionary of words ordered from most to least common, start bagging them into three lists: all right, all left, and mixed. There are 2 ways to do the next part: 1. When you have, say 30 of each, start A/B/C testing how long it takes to do a random one from one of them (first choose a random number 1 to 3 and then based on that choose a random word from the associated list.) 2. Instead of the above, I prefer to add them all into one big bag, randomize the bag, then pop elements from the bag (without replacement) until you're done, timing how long it takes the user to type each one. Don't look up which list each element was from until you're done collecting the timing information. (By grepping the three lists to figure out where you got it.) This makes things a bit "double blind" in that your user timing algorithm just gets a mixed bag to pop through. It doesn't know the sources. At any rate, once you do either the first or the second one, you can see if there is such an effect. It is worth checking via print statements and an iterator variable, how far down your list each one came from, just as a sanity check. (I could imagine that due to some historical accident in qwerty placement, most of the common words are mixed, and very uncommon ones are the first ones to be done all with one hand. They could be very long as well, something to watch out for.) Someone doing this in python should be able to complete this in about 20 minutes (if they're not used to saving timing information, and including finding the list of words). It seems easy to test. The only single-character common words are a and I - I would remove these two words because they'll add to single hand typing speed. Additionally it may be worth removing the "enter" penalty. (Maybe sending an enter is slower than pressing space.) Finally, perhaps it is worth seeing how long it takes to type one of the most common 20 words that are all left, all right, or mixed, of a given length. (such as 5 or 7 characters.) I would have to look and see how far down the frequency chart the words are, to see if the user suffers an "unfamiliar" penalty. Data science and testing are fun :) Maybe I'll do it when I get home.

by gwd 7y ago

equalunique 7y ago

>at high speeds, coordinating alternating keystrokes between hands is very difficult It's for this reason I believe the traditionally wide spacebar should be split up into a number of different keys. Having two sets of Shift, Control, Alt, and the Windows key on either side of the keyboard seems less sensible than one set that can be accessed by either thumb. I for example use a custom layout that splits the Spacebar into both Space and Shift. Now I don't have to shift my hands to the left/right to capitalize a letter, make an exclamation/question mark, make a double quote, etc.

vages 7y ago

Made the same observation! And the rolling patterns are better if they go from in the pinkie-to-index direction than the other way around. An example of how Halmak does not take this into account is how the h and t keys are placed. It immediately jumped out at me as a Dvorak user.

networkimprov 7y ago

The myth of alternating-hand advantage probably dates to the era of mechanical typewriters, when hammers arriving from opposite sides of the machine were less likely to jam.

mkl 7y ago
Alternating hands was certainly better for my hand joint issues. I found Dvorak not really faster, but more comfortable, and the alternating hands is a big part of that. Alternating hands is much better than consecutive keys far apart on the same hand, and way better than consecutive keys on the same finger.
candiodari 7y ago
Actually they were more likely to jam. They hit the same place and if you hit a left and right key at the same time, they'd invariably jam, which necessitated going in and manually prying them back free (the keys sort-of fit into each other)

logicallee 7y ago

>that in fact, at high speeds, coordinating alternating keystrokes between hands is very difficult, and that it's actually a lot more efficient to have "rolling" patterns where a single hand can type a sequence of characters in a smooth motion. This seems like a particularly easy hypothesis to test (if I weren't on mobile I would try coding it up). From a dictionary of words ordered from most to least common, start bagging them into three lists: all right, all left, and mixed. There are 2 ways to do the next part: 1. When you have, say 30 of each, start A/B/C testing how long it takes to do a random one from one of them (first choose a random number 1 to 3 and then based on that choose a random word from the associated list.) 2. Instead of the above, I prefer to add them all into one big bag, randomize the bag, then pop elements from the bag (without replacement) until you're done, timing how long it takes the user to type each one. Don't look up which list each element was from until you're done collecting the timing information. (By grepping the three lists to figure out where you got it.) This makes things a bit "double blind" in that your user timing algorithm just gets a mixed bag to pop through. It doesn't know the sources. At any rate, once you do either the first or the second one, you can see if there is such an effect. It is worth checking via print statements and an iterator variable, how far down your list each one came from, just as a sanity check. (I could imagine that due to some historical accident in qwerty placement, most of the common words are mixed, and very uncommon ones are the first ones to be done all with one hand. They could be very long as well, something to watch out for.) Someone doing this in python should be able to complete this in about 20 minutes (if they're not used to saving timing information, and including finding the list of words). It seems easy to test. The only single-character common words are a and I - I would remove these two words because they'll add to single hand typing speed. Additionally it may be worth removing the "enter" penalty. (Maybe sending an enter is slower than pressing space.) Finally, perhaps it is worth seeing how long it takes to type one of the most common 20 words that are all left, all right, or mixed, of a given length. (such as 5 or 7 characters.) I would have to look and see how far down the frequency chart the words are, to see if the user suffers an "unfamiliar" penalty. Data science and testing are fun :) Maybe I'll do it when I get home.