I would not say they did this wrong, but that they hold one specific variable /table height/ as static (and varying other variables). Which is legit and normal methodological approach.
This is only a valid approach if all of the people tested are the same size.
The important feature determining the appropriate keyboard slope is the relative position of the torso and the keyboard surface. For different sizes of people, this will be different, even if the chair and table are held fixed.
It seems like in this study they set the height of the keyboard to always be either 4 or 8 cm above the height of the elbow, which is fairly reasonable (hopefully there’s not *too* much variation in the length of the subjects’ forearms).
The big problem with this study was that their “default” configuration A from which they made various other modifications was not very good, and left people’s arms in non-neutral positions with static muscle load. If they had started from a configuration where all of the joints were in neutral positions they could have seen how modifying it in any direction would make subjects trade-off various types of discomfort to compensate.
In their test condition where the keyboard is above the elbow, the keyboard should by default have a slight “positive” tilt, i.e. should be tilted up slightly at the back. They also should have removed the palmrest, which caused all of their subjects to have slight wrist extension even in the negative tilt condition. (As compensation for the negative tilt, instead of flexing their wrists, their subjects lifted their elbows up and out to the side increasing forearm pronation and load on the shoulders.) If the palmrest is there, people tend to use it, because they think they are supposed to, even though it’s not as comfortable as ignoring the palmrest.
If they wanted to test the effect of a palmrest that should have been a separate experimental condition, and they should have picked a taller palmrest which didn’t require wrist extension to use.
It’s not clear why they set the seat backrest angle to 105° and told people to lean back. That’s not really a great long-term typing posture, IMO. I guess it’s better than what most office workers do in practice.
The best of the conditions (by empirical measures rather than subjective preference) they tested was E, with 0° tilt, height of 4 cm, and opening angle of 12°. When they increased the tilt (which subjects subjectively preferred), the subjects extended their wrists an uncomfortable amount, and also bent them outward (“ulnar deviation”). Some of their subjects liked configuration C, but that was because of the increased opening angle not the negative tilt.
They would have gotten a more comfortable configuration by adding some “tenting” to reduce forearm pronation, further increasing the opening angle to reduce ulnar deviation, moving the left half of the keyboard to the left (increasing distance between sides), picking a positive tilt of more like 4°, and dropping the palmrest.
It’s ridiculous that they only tested the 8cm height condition with a –8° negative tilt. When raising the keyboard higher it should be tilted up at the back, not down at the back.
A negative tilt is fine for a keyboard used on a low table (or for a standing desk) where the keyboard surface is slightly *below* the level of the elbows.