Deciding to see how several alternative keyboard layouts perform against the well-known QWERTY layout, I took a Java applet created by a Jon A. Maxwell for the purpose of testing keyboard layouts and attempted to add several extra layouts to the source code by looking at descriptions on Wikipedia and the creators' home pages. My modified source code can be found here, but readily usable .class files are unfortunately not included.
The applet gathers data on what keys would be pressed in order to type a given piece of text, and attempts to calculate the distance the user's fingers would travel, as well as what percentage of keypresses would be typed with the same hand or finger twice in a row. It also how many percent of the keypresses would be on each row on the keyboard.
The text sample used was from this page as rendered by the Mozilla Firefox browser (Opera would include an extra line break for each paragraph when copying the text), starting with the phrase "Chinese, Japanese and Korean" and up to and including "(e.g. Unicode)."
I've been advised that the sample size used was too small and that the layout data may contain errors (For instance, the ASSET layout had its minuscule d replaced by an extra minuscule f when I ran the test, which may have skewed the results). But in the absence of more accurate data, here it is:|
Layout |
Distance |
Same hand |
Same finger |
Total keys |
Number row |
Top row |
Home row |
Bottom row |
|---|---|---|---|---|---|---|---|---|
|
71.96 m |
38.03% |
4.985% |
2824 |
0.837% |
49.22% |
31.37% |
18.55% |
|
|
ASSET |
31.45 m |
32.25% |
2.764% |
2824 |
0.837% |
18.60% |
72.72% |
7.834% |
|
Capewell evolved (0.9.3) |
35.05 m |
39.21% |
2.471% |
2824 |
0.837% |
16.75% |
67.32% |
15.08% |
|
35.57 m |
33.43% |
1.089% |
2824 |
0.837% |
15.29% |
71.47% |
12.40% |
|
|
Klausler 2 |
35.90 m |
24.21% |
2.346% |
2824 |
0.837% |
21.36% |
71.17% |
6.619% |
|
Dvorak (I and U switched) |
36.00 m |
23.41% |
4.105% |
2824 |
0.544% |
23.16% |
68.07% |
8.211% |
|
Klausler 1 |
36.04 m |
32.34% |
1.340% |
2824 |
0.837% |
21.49% |
69.41% |
8.253% |
|
QWERAK |
36.19 m |
36.32% |
4.273% |
2824 |
0.837% |
22.91% |
68.16% |
8.085% |
|
Capewell Dvorak |
36.21 m |
24.00% |
2.513% |
2824 |
0.837% |
17.46% |
68.20% |
13.48% |
|
38.89 m |
23.41% |
4.105% |
2824 |
0.544% |
23.16% |
68.07% |
8.211% |
|
|
38.94 m |
42.73% |
6.828% |
2824 |
0.837% |
17.63% |
72.55% |
8.965% |
|
|
Arensito |
42.08 m |
44.57% |
6.116% |
2824 |
0.837% |
14.91% |
60.11% |
24.13% |
|
Capewell QWERF 2006 |
50.91 m |
38.03% |
4.985% |
2824 |
0.837% |
27.77% |
52.82% |
18.55% |
|
Capewell QWERF |
51.77 m |
38.03% |
5.152% |
2824 |
0.837% |
29.95% |
51.73% |
17.46% |
|
60.80 m |
29.66% |
6.954% |
2824 |
0.837% |
36.53% |
48.72% |
13.90% |
While I personally can't vouch for the accuracy of the results, they seem to confirm that several alternative layouts perform better than QWERTY. With several other text samples, the PLUM keyboard had the highest usage of the home row.