Do-It-Yourself Keyboard Study

It is pretty easy to demonstrate that the Dvorak layout is much more logical than QWERTY in several respects. I developed a quick litmus test that anybody can repeat:

I started with some keyboard goals that seem acceptable to all sides of the discussion:

Typing words on the home row is easier.
Typing whole words with one hand is difficult.
Typing whole words that never touch the home row is ridiculous.

And, since one correspondent speculated the top two rows are about equally easy (perhaps favoring QWERTY):

Typing on the top two rows is pretty easy.

To compare Dvorak, QWERTY, and an alphabetic layout against these criteria, I used the hidden-word feature of a shareware Mac anagram program (games/word/anagram-crossword-15.hqx on Info-Mac) to see how many unique 3-letter or larger words it could construct from different groups of keys.

For example, to test the Dvorak home row, I fed the program "aoeuidhtns." It listed all the words it could create with those letters. Here are the word counts for all the key sets I tested:

                      QWERTY    ABCD    Dvorak
                      ------    ----    ------ 
(More is better...)
Home-row-only           49        40      618 
Top-two-rows-only     5810      6977    10448 

(Fewer is better...)
Left-hand-only         687      1183       58  
Right-hand-only        101        49       29  
Off-home-row-only     1138       596       15

Dvorak comes out well ahead in all these tests using the default dictionary. (Even the alphabetic keyboard looks better than QWERTY in some respects.) I also tried some tests with a list of 1000 most-used words that I found on the web; the ratios were similar.

The numbers look small overall because the strings I used had only one instance of each tested letter, so the test ignores words with two or more of any letter (like "test" and "letter"). I accepted this limitation to keep run times down.

Later I found that the program only accepts 32 character strings; I did a quick test with a couple of padded strings (for example, "aoeuidhtnsetaoinshdetaoinshdetao") and I didn't see much difference in the ratios.

Someday I hope to write a program that will measure actual text against these (and perhaps other) criteria. That would be a more true-to-life test; but frankly I don't expect QWERTY to catch up.

Back to Introducing the Dvorak Keyboard.

Last update: 8 December 1996
Original page established: 25 February 1996

Marcus Brooks: HTML Home Weblog