- Reputation: 17
- From: CZ
- Registered: 14-Feb-2012
- Posts: 438
I'm just dumping my thoughts here, because I think some people may find it particularly interesting. What's up? I finished my thesis on keyboard layouts a while ago, and I've run some analysis on the Maltron THOR layout. (I have to figure out the best way to share these results online, but that's off-topic here.)
What's Maltron THOR? It's a keyboard layout (see patent US4244659 by Lillian Malt) designed for a dual-hand ergonomic keyboard (based on IBM's patent GB1016993A). (Note that there's a layout for one-handed typing described in Malt's patent as well.) You can read more about it in Malt's paper available from Maltron's website. The key design principles are following:
Balance finger/hand use.
Utilize the shape of the ergonomic keyboard, esp. thumb clusters and column placement, that reflects inconstant finger length.
Maximize the use of home (resting) finger positions.
Favor sequences of adjacent keystrokes over extreme hand alternation (preferred by Dvorak et al.), because such sequences are presumably faster on electric keyboards, esp. if finger length is taken into account.
Separate vowels, because confusing them is the most common source of typos on DSK (see Malt's paper on error types and patterns).
A scheme of the layout over at Wikipedia. I call it THOR, because of the right-hand finger-resting positions; the left-hand side (ANIS) sounds more like, ahem, "anus".
I've coded a new layout analyzer, based on Carpalx, but easier to use or extend: it supports Unicode, layers and dead keys; the reported values are mostly the same as in Carpalx, but I calculate the keystroke count and some sort of hit direction (both inspired by Marsan's criteria by Wagner et al.) on top of that. It's been used to compare layouts on Czech and English corpora. The base English corpus is the bundle of ~19th century novels from Project Gutenberg, distributed with Carpalx. My Czech corpus is a collection of novels and newspaper articles published between WWI and WWII; letter frequencies more or less match the SYN2010 literature corpus (Czech National Corpus).
It's interesting, that tested layouts (QWERTY, DSK, Colemak, Maltron THOR) maintain their main properties (esp. hand alternation frequency and relative same-finger ratios) between the corpora, except for finger balance (due to high degree of inflection and heavy use of diacritic marks in Czech, as opposed to English, which is an analytic language and has only 26-letter alphabet).
Anyway, back to Maltron. It turns out that the THOR layout scores almost exactly the same as Colemak in terms of hand alternation and rolls, same-finger ratio, etc. In Czech, THOR has a much nicer finger balance without obscurities, such as the WRX column on Colemak, and a lot of typing shifts from the bottom row to the top row, because of CVBM placement. All in all, the main distinction is obviously the size of the home row (i.e., E under left thumb) and that cognitive (i.e., not biomechanical) factor related to error patterns.
I've been considering THOR (and DSK) as my next layout recently—as a replacement for Colemak—but the results are quite a twist. Is relearning the layout justifiable? Reduced error rate would be obviously nice (assuming we trust Malt); the controversial part is E in a thumb cluster. I'm not bothered by limiting myself to keyboards with a lot of thumb keys, because I'd learn DSK basics for backwards compatibility anyway. However, isn't the physical key better used for some modifier/toggle or control symbol? Consider that we don't only type, but also control user interfaces or edit data nowadays…
Get yourself an ergonomic keyboard. Learn Maltron. Or stenography.