The thing that's really great about all our acoustic musical instruments is that when you push on them, they push back. I don't mean this simply in the normal force sense. Consider a guitar. You've got a string which you press over a fret with your left hand. This can be naively emulated by a switch. But with the guitar, the timber and pitch change depending on how you fret the note, on the finger pressure and position and motion. With your other hand, you might be plucking the string in any of a number of different places, with your finger or with a pick. The sound is affected by your attack angle, how hard you pick, the pick's composition, and so on.
A guitar string is clearly a complicated system. There are lots of variables at play. But more importantly, it's a coherent system. It makes sense to us a physical object that can be manipulated. When you pluck the string, you can feel it vibrate in your fretting hand. When you bend the string its tension increases. If you amplify it, you get the sense that you are physically touching the sound.
(This is incidentally, why audio latency absolutely KILLS when doing amp simulation)
The experience playing a wind instrument is similar. While a saxophone may appear to be something you blow into that has keys, things are really far more complicated than that.
Keyboard-based instruments are a little different. Unlike most any symphonic instrument, the piano actually has relatively few parameters per key. There's note velocity... and that's about it. The various sustain pedals also apply. The piano's design trades single note expressivity for the ability to play ten of them at once.
It should be noted that computer synthesis (procedural or sampled) of keyboard-based is very convincing. They same cannot be said for any other instrument.
Now, what about new kinds of control systems? Most of them tend to fall into two categories. One tries to improve on the piano harmonically, by coming up with a better arrangement of where the notes go. Here's an overview of some: http://sequence15.blogspot.jp/2010/03/alternative-keyboards..... They try to fix the fact that it's hard to play in different keys on a piano. Whereas on the guitar you can learn a single scale or chord and move it up and down the neck to transpose, things change radically on a piano keyboard.
The second category is those like the Seaboard, which try to add new dimensions of control to a regular piano keyboard. Another example is the Contiuum (http://www.hakenaudio.com/Continuum/) It's very common now to have both velocity and continuous pressure sensitivity (aftertouch) on a regular keyboard as well as various side controllers for dealing with pitch or an abstract "modulation" parameter.
These controllers nearly always buy into the separation of control from synthesis. It makes perfect technical sense. But most of the instruments we would consider to be "expressive" don't work that way! In fact piano-style instruments are pretty much the only ones that do.
Which leads my to my point: A control mechanism should be considered together with the instrument it controls. It's fantastic that this new keyboard has all these new dimensions that you can map to sound, but what is it REALLY good for? What is the instrument that wants to be controlled in this way? The spiffy new control surfaces nearly always leave this problem unsolved and thus remain little more than novelty items.
The classical pipe organ, often referred to as "the king of instruments", is a synthesiser. The organ console is an electrical or pneumatic controller, with no direct connection to the pipes. Most large organs have considerable "latency", due to the distance from the console to the pipe room. Some pipes may be as much as a hundred feet away from the player, so the sound will take over 90ms to reach them - several orders of magnitude more delay than a modern computer system.
The organ is still regarded as a highly expressive instrument, in spite of the relatively modest control a player has beyond simple pitch and duration. Although most organs have many stops which provide similar timbres to existing instruments, it is understood that the organ is an instrument in its own right and should be played as such, and is not merely a tool for emulating other instruments.
The idea that an electronic instrument should imitate acoustic instruments is simply a poverty of imagination. Electronic instruments, used in a manner that is sympathetic to their natural properties, can be absolutely as expressive as any acoustic instrument. The theremin has perhaps the worst user-interface of any musical instrument as the player has no physical contact with the instrument whatsoever, but is utterly beautiful when played by a master.
The challenge for electronic musicians is that they are often both performer and instrument-maker. A monosynth of any quality can be configured in a near-infinite variety of ways, many of which were completely unforeseen by the designer. Musicians working with modular systems or DSP programming environments have a blank canvas. We do not as yet have a good theoretical framework for this task, but electronic music is extremely young - no more than ninety years old at most. We are only just beginning to scratch the surface of what is possible.
Additional points of interest which do not affect your actual argument, but which are good for clarification: the most responsive pipe organs use tracker action, which being mechanical allows much more control than the more "modern" but less responsive electropneumatic actions necessary for the very largest instruments. IT makes the action from pressing key to opening the pipe faster - much of the latency in the organ is often in the pneumatics - but of course cannot change the distance issues.
also Attack makes a huge difference to the sound on a well-voiced instrument, and this makes a huge difference. Subtleties in duration are of course also as vast on the Organ as they are with the piano and other instruments.
You don't even need to go as far as organs to see latency.
Low-pitched stringed instruments such as fretless/upright bass guitars have latency due to the mechanics of the strings themselves; e.g. jazz bass players have to account for this when playing. Most good jazz bass players do this unconsciously.
Polyphonic reed instruments (e.g. harmonicas, melodicas, accordions) all have this issue as well. Reeds of significant mass (i.e. lower-pitched reeds) can take tenths of a second to sound (more on older instruments).
Also: if you can feel the string in your fretting hand vibrate on a fretted instrument, you're doing it wrong. (The string does not vibrate past the fret -- that's the point of frets! If it does, you're not pressing firmly enough and you get fret buzz. Either that or your finger's on the wrong side of the fret and you're muting the sound.)
Generally the entire body of an acoustic guitar vibrates with the sound of a plucked string - the fretting hand (being the only hand currently attached to the guitar) would likely feel the vibration carrying from the body, through the neck of the guitar. It's faint, but nevertheless perceivable.
> The organ console is an electrical or pneumatic
I've no criticism of your theremin example though.
When I depress the fingerboard of a Continuum I can feel it vibrate in my ears, which is quite enough feedback to manipulate a sound expressively. Sure, playing the guitar is a beautiful, unique, rich sensory experience (which I love), but it does not follow that the lack of the "guitar experience" leads to a lack of musicality. Each instrument has its own mode of interaction, from the guitar to the piano, and talented people seem to find ways to be expressive with all of them.
> These controllers nearly always buy into the separation of control from synthesis. It makes perfect technical sense. But most of the instruments we would consider to be "expressive" don't work that way!
So? Why does what already exists matter? There are plenty of people willing to experiment with a new input surface to find out what it's good for. You may not be one of them, but why do you need to be "suspicious?" These experimenters don't take away your ability to play a guitar.
> but what is it REALLY good for?
I can't imagine this being played on any other instrument:
> What is the instrument that wants to be controlled in this way?
You could also ask, "what is the music that wants to be made by a stringed instrument?" People have been exploring that question for thousands of years, and we're still finding out new answers. Electronic instruments are very, very new compared to that, and there's been comparatively very little time to learn about them. I say let's go wild and create myriad new instruments and find out what works.
Personally, I think the decoupling of input surfaces and sound generators is one of the all-time best developments in music. As a brass musician, it's nontrivial for me to produce the sound of a oboe. However, given a very expressive input surface I can produce an extraordinary range of timbres without dedicating another decade to practicing each individual instrument.
This is exactly my point! In the abstract the continuum doesn't have much to say, musically. Paired with this sound source and played by Jordan Rudress, it works. (Who also has something to do with this new company, it seems)
Pat Metheny's approach to the guitar synth speaks to this, I think:
Unlike many guitar synth users, Metheny limits himself to a very small
number of sounds. In interviews, he has argued that each of the timbres
achievable through guitar synthesis should be treated as a separate
instrument, and that he has tried to master each of these "instruments"
instead of using it for incidental color. One of the "patches" that Pat
used often is on Roland's JV-80 "Vintage Synth" expansion card titled
"Pat's GR-300". 
> You may not be one of them, but why do you need to be "suspicious?" These experimenters don't take away your ability to play a guitar.
Point taken, I'm probably not one of them. But I used to be, and there's a lot of snake oil out there. My experience with newfangled instruments is as follows:
- Korg Padkontrol: This was my only real contact with an MPC-style interface. I wanted to use it to sequence drums in real time, and it was decidedly mediocre for that. When using it for other things, part of the musical task turned to the configuration of the controller. It's creativity, but a different kind to be sure. It blurs the line between performance and composition.
- Zendrum: Seeing that people were able to play live on these things somewhat convincingly led me to try it. It never really clicked for me. I spent too much time configuring and never enough actually practicing the instrument. There are many reasons this could be my fault, not the least of which is that I'm not a drummer.
- Chapman Stick knockoff: I was never able to get beyond just piddling around on this thing. My imagination was captured by a video I saw on the web at some point, and I guess the ad copy closed the sale. But some weeks after I got it, I was left with a distinct feeling of "now what?" It was an instrument without any useful context. I've been told that the real thing is far more compelling than the knockoffs; perhaps I'll try one of those some day.
This is far more likely a commentary on myself than on these three instruments. For me, searching for the perfect instrument was something like creating a new programing language before writing your program. It's a never-ending task that inevitably fizzles out. I have since been better served by my Telecaster.
I think the video does a reasonable job showing how it can be used creatively. I get the point you're trying to make about harmony between the input device and sound source. However, I would argue in modern days with really incredible software synthesizers both software and hardware (I own one of these: http://www.studioelectronics.com/products/synths/omega8/) the inverse is more common.
Input devices do a poor job of exposing the features of the instruments.
This is the first device I've seen that seems to actually care about the performability and feel without making you look like a douchebag. That's what makes it interesting to me.
There's actually a recurring pattern for anyone who cares to notice. Most new instruments are developed to emulate something else and they do it badly. Pipe organs were developed to emulate choirs. Now we love pipe organs precisely because of their limitations. They have a "distinctive sound."
The same thing happened with the Fender Rhodes piano, the Mellotron, the Moog, the Fairlight Synthesizer, and (believe it or not) MIDI. They started as cheap substitutes (masked by a "wow" factor), then people bemoaned their limitations relative to what they emulated and ditched them, then they were resurrected for the uniqueness of their sound.
Instruments are defined by their limitations. No matter how much we claim to hate those limitations, we end up loving them for them.
I agree that keyboard instruments are the most convincing, and certainly the easiest virtual instruments to play (since playing a digital piano is the same as playing an acoustic piano), but there are some pretty expressive and convincing virtual instruments out there:
Bass guitar (acoustic and electric): http://www.spectrasonics.net/products/trilian-audio.php
Acoustic guitar: http://www.youtube.com/watch?feature=player_embedded&v=q...
Electric guitar: http://www.vir2.com/instruments/electri6ity
Orchestra (perhaps not the best solo instruments, but pretty great results for a whole symphony): http://www.vsl.co.at/en/67/702/703/413.htm and http://www.soundsonline.com/Symphonic-Orchestra and http://www.garritan.com/products/personal-orchestra-4/
And, just to mention it, the best piano virtual instrument I've heard so far: http://www.synthogy.com/demos/grandpiano.html
I'll have to agree that it might just have to remain as a novelty for now.