Interesting theory. Certainly seems plausible, despite JD's compelling argument.
Check out this nice photo of a beam spring board from Halvar on Deskthority. Note how the keycaps look basically like SA, but are mounted at an angle to the beam spring switches, giving them a much nicer profile:
Show Image
(http://deskthority.net/resources/3278-alphanumericals-profile/15303)
^^ I like this a lot. I'd love to try typing on such a keyboard.
Also, moved to key caps. :thumb:
Oh, ok, if you're publishing it somewhere else with more context, that's even better of course.
Here’s what a keyboard looks like that has steeply angled bottom row modifier keys:
Show Image
(http://i.imgur.com/USaQaIG.jpg)
Are you sure about the heavily angled bottom row being modifiers on this keyboard? That would seem less than ideal to me. I'd certainly prefer the thumb modifiers solution.
I don’t think the angle of the top of the switch matters all that much per se. I don’t expect I would have any problem using e.g. uniform DSA keycaps, assuming the switches themselves were positioned correctly. The problem is that there should optimally be a height step (of at least 2mm, but I think more is probably better) between the back of one keycap and the front of the next, for all rows further than the home row. Otherwise, when you depress the further-away key, your finger runs into the closer key.
Here’s my personal subjective ranking of various profiles (now with a slight negative tilt to the keyboard, used on a proper-height desk):
Show Image
(http://i.imgur.com/91dI4qM.png)
I plan to make a physical version of this chart in the relatively near future, so that I can make a video showing how a finger interacts with switches laid out in these various arrangements.
My suspicion is that people’s familiarity with laptops is biasing them toward uniform profiles. Laptops can get away with the uniform profile because (a) they have no space to do anything else, and (b) the switch travel is very short, so there’s no risk of a finger colliding with the next key down. The flat arrangement on a laptop is of course still suboptimal in the sense that it takes more hand movement and weaker finger positions to reach further away keys, and there’s additionally no grippy concave top for the fingers to rest on.
We have very different opinions on keytop angle and height steps. IMO, the angle matters a lot and no height step is necessary, in fact I consider it to be detrimental.
Finger collision with lower rows will only ever happen if you have a VERY flat finger angle when pressing the keys (and I have never experienced it). If you type correctly, with your wrists raised, this simply cannot happen with SA profile caps. For pressing number row and F row keys, you have to move your hand upwards anyway, so the angle never flattens enough to cause this. If you don't, you need to correct your typing technique, not your keycap profile.
The only case I could see a finger having a flat enough angle is with a large difference in finger lengths (or very small hands) and a long stretch of the index finger. Even then, it's more likely that the user will move their hand slightly to correct the finger angle than to get interference of the lower keycaps, since such a flat angle is both not comfortable and not correct. Again, in that case, change your technique not your keycaps.
The angles of the keycap tops affects the feel of the board a lot and IMO should be quite "neutral" in terms of being based around a flat top on the home row, with progressively more curve the further from the home row you move. This creates a surface with the least required effort to move from one row to the next and depress the key, without having to lift over either the home row key bottom edge or the upper row bottom edge depending on which row you're moving to. This creates less fatigue and allows the shortest finger movement path.
From your image, I consider the one you labeled "unacceptable" to be the most optimal. Also note that most people would use their board at a very slight positive angle, not the extreme you posted earlier or the negative angle in the latest image (actually they are all at slightly different angles in your last image if you take a line through the centre of each keycap top, where the finger would rest). For a good example of what I consider to be optimal or quite close to it (I prefer the steeper sides and larger spherical top of SA than the DSA type keycaps used here), take a look at the alpha area of the Maltron keyboard (a very well researched and tested ergonomic design):
(http://i82.photobucket.com/albums/j245/saxeharp/general10/engrave_zpsfe37d786.jpg)
Many find "chiclet" style keyboards "surprisingly comfortable" from the first time they start using them, not only after using them for a while, which makes me think that flat keycaps are not as bad as we are led to believe when used on a flat or very slightly angled board, although a curve and spherical keytops would certainly improve things.
Please note that both our views are merely opinions, but I hope we are both giving enough reasons for others to determine which they would prefer and why.
For a good example of what I consider to be optimal or quite close to it (I prefer the steeper sides and larger spherical top of SA than the DSA type keycaps used here), take a look at the alpha area of the Maltron keyboard (a very well researched and tested ergonomic design):
Show Image
(http://i82.photobucket.com/albums/j245/saxeharp/general10/engrave_zpsfe37d786.jpg)
Yep, I think this IBM design gets this wrong: http://geekhack.org/index.php?topic=63415
And I also think the Maltron gets it wrong. And I think the Kinesis Advantage gets it even more wrong than the Maltron.
Basically, the researchers (both at IBM, and Lilian Malt) did a good job tracing the motion of fingers in finding the tops of the keys. But they then failed to think about or account for the joints used to actually press the keys down. This is in my opinion a dramatic oversight, which poisons the design.
Fortunately, it’s only the number row that is excessively affected by this design choice, and numbers are typed relatively infrequently. On the Maltron, the letter layout has been arranged so that most typing happens on the home row, with occasional movements up or down by one row. For someone who needs to type many numbers or symbols though, I think the Maltron would be substantially improved by more carefully orienting the keys to align with the natural direction of finger motion when pressing them.
The guys making the IBM Selectric and IBM Beam Spring keyboards in the 60s/70s actually got closest to right about this, IMO, but their design is compromised by being in every other way the same as existing QWERTY typewriters (which are terrible), and therefore needing to make some other compromises: on a column-staggered board, it’s possible to get more aggressive with height steps between rows than on a standard row-staggered board. Also, some of their design principles are more aesthetic than functional. For instance, despite the tilt of the keyboard overall they arranged the bottoms of the switches to be parallel to the table, which is totally unnecessary; and they tried to get the switch tops to roughly follow a circular arc, when what they should have done is test every possible angle and keytop-height for each row and figure out which position and angle makes it easiest to acquire and press the keys, regardless of whether the resulting shape is the most visually harmonious. In my opinion having the letter and number key switches all oriented in roughly the same direction (by contrast to the Maltron which has them oriented radially) is approximately the most effective design, but I think they wound up with this design mostly by accident, because it also happens to be cheapest, rather than because they carefully considered which orientation was best.
* * *
Indeed, I have actually found no source (academic ergonomics research paper, patent, whitepaper, typing technique book, etc.) which properly describes the way fingers actually move when typing. No one starts from first principles and tries to reason things through. People do seem to have figured out some reasonable ways to hold arms and wrists for comfort, starting from that IBM patent in the 60s which is remarkably far ahead of its time from a “keyboard ergonomics conventional wisdom” point of view. But the motions of fingers themselves are never described adequately.
The only keyboard that I’ve seen evidence was created by someone with a deep understanding of human finger range of motion is Benjamin Rossen’s DataStealth, which only ever existed as a prototype. He didn’t explicitly write about this in any detail, but the proof is right there in the design. It makes some sense that he would have made a more anatomically relevant design, as he was an ergonomics and anatomy expert who explicitly started from scratch in his design process. (As far as I can tell Lilian Malt, while she did brilliant work, was not an expert on hand anatomy or physiology. Same goes for the inventor of the DataHand.) It’s really a bummer that Rossen’s keyboards never made it to production.
* * *
This is the kind of thing that an academic research group serious about ergonomics research would do a thorough investigation of. I think all the existing ergonomics labs and research programs have researchers who demonstrate poor understanding of human anatomy and poor understanding of scientific study design or statistical analysis. (I hope there are at least a few exceptions to this, but I haven’t seen much if any inspiring work in the field...)
It would be awesome to set a large number of people (at least several dozen) up with keyboards of 4–6 different shapes but otherwise identical construction and switches, and let them train on those keyboards for at least 1–2 months (or ideally 6). Then do careful motion studies of hand movement, impact forces, tension in all the joints of the fingers, etc., and try to compare from one keyboard to another. The products of such a program would include interactive animations of hands moving around the keyboard, vast tables of data comparing typists of different experience levels and hand shapes trying to type different passages, with sophisticated analysis of errors and particular slow combinations.
There are a whole bunch of different ways of setting up alternative keyboards to try, and various design constraints could be studied with different setups. For instance, they could start with standard layout (ANSI QWERTY) keyboards and test several keycap profiles, including both ones that commonly exist today and new designs tweaked in various ways. The ideal here would be to use a light-to-medium actuation force clicky switch with good tactile and audio feedback. Alternately keyboards could be made by arranging scissor switches in various 3-dimensional patterns, including both flat and sculptured designs. Since people are used to typing on laptops now, we could normalize to the travel distance, switch feel, and keycap feel of laptop keyboards while properly testing the effect of keyboard shape and switch orientation.
Beyond that, once we start looking at split designs, all the existing ergonomics research papers (at least the ones I’ve seen, mostly from the mid-90s through early 2000s) do an absolutely miserable job of properly exploring the full space of possible arrangements of two split halves, even assuming that the layout of each half has already been fixed and the positions are laterally symmetrical. This is a space that has 6 reasonable dimensions to vary (distance away from the body, vertical height, distance between keyboard halves, plus 3 dimensions of rotation), but every study I’ve seen picks at most 5 or 6 discrete points in that 6-dimensional space, via basically arbitrary selection criteria, and then try to draw conclusions from the differences between those unrelated points. There’s no model of human arm/hand/body motions involved in the analysis, no attempt to normalize for people of different hand or body shapes, etc. It’s a travesty.
Of course we should also have some studies properly looking at the effects of different types of switches on typing speed, accuracy, comfort, subjective preference, etc. The only ones I’ve seen do ludicrous things like compare two types of rubber dome to a Model M, and then make an analysis about whether “low” or “high” actuation force switches are better, without noticing that travel distance, actuation point, force curve on both downstroke and upstroke, and tactile and audio feedback are all dramatically different between the switches. To have even a hope of understanding the effects of different switch parameters, it’s necessary to control for all the other variables not being studied.
Anyway, I think a real serious research program (which would take man-power, equipment, time, and money) would be able to produce some great results about what actually works and doesn’t work in keyboard design (from an efficiency, comfort, and injury prevention point of view), and would be able to produce useful informational content teaching people what to look for in a keyboard and how to type. Unfortunately, I don’t think such a thing is ever going to happen, despite the incredible economic importance of typing (and incredible cost of RSI) to our modern society. Alas.
Rant over.