I see. However, never trust a layout maker's claims. If we were to choose layout going by the strength of claims, I'd be recommending Workman or something. ／(⁎≧∇≦)＼
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
I agree wholeheartedly, hence my interest in experiments and publications that discuss experiments.
The people behind Maltron actually did conduct some experiments and included error-rate counts in an article that I can't find online anymore.
To obtain the error counts for Maltron, errors were located through a DEC "Runspell" program on 5,326,470 characters. "Runspell" is accurate to 96%, and locates all errors except where a recognized word results. e.g. 'there' for 'their', or 'than' for 'that'. With a 96% accuracy level in error location "Runspell" is more accurate than manual methods. A total of 1,139 errors, (0.021% of the keystrokes analyzed) were located on Maltron keying. Of these 328 were substitution errors, 28.80% of all errors, and 0.006% of the total keystrokes analyzed. 63,916 substitution errors were given for Sholes, and 3,329 for Dvorak. The total number of keystrokes in which the # substitution errors occurred is not given for Sholes and Dvorak layouts.
Get yourself an ergonomic keyboard. Learn Maltron. Or stenography.
Such an experiment would be massively subject to selection bias. Maybe the typists trying out Maltron were inherently more accurate than the others tested, for instance. As an example, if you were to test the Colemak layout based on the Colemak Discord Speedtyping channel denizens, your test group would be very biased indeed – and have an overall lower error rate than the outside population, I should think.
*** Learn Colemak in 2–5 steps with Tarmak! ***
*** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***
