Keyboard Research

[hoosieree]

This turned up in the results of the search 'keyboard layout efficiency':

http://atri.misericordia.edu/Papers/Chubon.php

Figured this is a good place to share.

[jacobolus]

Thanks for the link.

For all those 1-finger typists out there, now there’s a credible alternative to QWERTY, as long as they’re willing to spend time learning it.

Specifically, they compared Chubon:

vs. Reverse QWERTY:


(Standard QWERTY was faster than Chubon, because the study participants were used to it, and there wasn’t time for long-term Chubon training as part of the study. It’s still untested whether Chubon would be faster than QWERTY after training.)

[Snarfangel]

The FOAD on the bottom left is useful!

[hoosieree]

I'm less interested in the layout than the use of a control (reverse qwerty) to evaluate its effectiveness.  I'd love to see similar scientific rigor applied to keyboards with "ergonomic" in their description.

[jacobolus]

To really evaluate effectiveness in a scientific way, you’re going to need to spend at least a few million dollars for a rigorous study with at least a few dozen participants in each group, and months of training/use on each keyboard layout. I don’t think there’s been any truly serious work on this in at least 50 years.

The study linked here is a toy, and doesn’t lead to any particularly meaningful conclusions IMO. (Thanks for linking it though, designing one-finger keyboards is a sort of interesting topic.)

Even reverse QWERTY is going to be tough to draw practical conclusions from in a 10-finger typing context. Muscle memory will apply pretty well if you’re just swapping hands, and while there will be many mistakes, training should be pretty easy compared to learning e.g. Dvorak, for a practiced QWERTY typist.

The best bet is to find people with zero touch typing ability and ideally no exposure to any keyboard layout, and train multiple groups of them on different layouts. Such people are probably pretty hard to find these days though.

[hoosieree]

Agreed, I don't think a meaningful "control group" study like this will ever be applied to keyboards in general.

But, in conjunction with the core idea presented here (namely, to measure efficiency via counting dictionary words that can be typed using only home keys), and heat maps like here, there are several things we (as users and/or creators of ergo-related keyboards) can have in our toolboxes to make more objective measurements of keyboards.

I'm not discounting subjective measurements either.  I'm a musician and understand very well that an instrument can be objectively perfect in any number of aspects, yet still totally miss the mark in aesthetics or other subjective ways.

Since I'm looking to build some sort of ergo keyboard for myself in the near future, I want to collect these ideas and share them here.  Hopefully they're useful or interesting to others.

[PieterGen]

Specifically, they compared Chubon:

Show Image

vs. Reverse QWERTY:

Show Image

This study was intended to test the following research hypotheses:

Hypothesis: The ReverseQWERTY keyboard will maintain the biomechanical relationships of the QWERTY keyboard while removing the effect of learning, allowing fair comparisons of alternative keyboard patterns.

So... you're comparing two layouts, in which one has an (upside down, wtf?) resemblance to the Qwerty letters, but puts the modfiers and Spacebar in very different locations;   versus a layout that has the letters arranged differently, but has the spacebar and modifiers in their usual locations.  What could you possibly conclude from such a study ?

[jacobolus]

I’ve actually never seen a research study which answered any of the numerous questions I have about keyboard letter layouts. Alas. (I’ve skimmed several hundred keyboard-related papers.) Most of my questions about general keyboard shape and positioning are also unaddressed by the literature, and my best answers are my own study of human anatomy mixed with reasoning and testing ideas on myself.

But, in conjunction with the core idea presented here (namely, to measure efficiency via counting dictionary words that can be typed using only home keys),

The author of that blog, Proword has written quite a bit of interesting stuff here at geekhack about his workflow/setup. Since he types using the Maltron keyboard and Malt’s layout, it stands to reason that he agrees with Malt’s design criteria. As he says though, “I don't for one moment hold that this is the only desideratum for assessing if a key distribution is ergonomic or not, but in my view it's certainly a very good starting point.”

Indeed, it’s one reasonable place to start, but I think the only way to make a fully optimized keyboard layout is to analyze multi-letter strings, breaking words down into chunks which are easy to type as a single motion, and then analyzing the difficulty of transition between chunks. None of the heuristics I’ve seen used to analyze keyboards get this right, focusing too much on individual letters and key distances.

Since I'm looking to build some sort of ergo keyboard for myself in the near future, I want to collect these ideas and share them here.

Poke around the Geekhack ergonomics subforum a bit. There’s been a lot of interesting discussion here in the past few years.

What could you possibly conclude from such a study ?

One conclusion I have: 180° rotated QWERTY on a grid layout keyboard is really bad for one finger typing. :-)

[vvp]

If I would be tasked with creating the best keyboard layout then I would minimize energy use. Create a set of equations which model human hands. Assign energy use to movements as well as maintaining static positions. Find the layout which minimizes the energy needed to write given test corpus.

Simple model taking into account bones, joints tendons, muscles (including partial sharing of muscle for e.g. fingers), mass of all of this plus mass of other tissue. Some friction estimates.
It may be useful to add some heuristic for blood flow when tissue is static/moving/under-external-preasure.

Probably the cheapest way how to find the least tiring layout because you do not need hundreds of study participants just a physician, physicist, and possibly a software developer.

[hoosieree]

Simulation in place of human tester's time would indeed be more efficient.  Checking the model's accuracy could be done with a thermal camera and just a few subjects.