Milton Wood Raymond, aka "Milt" or "MWR", wanted to build a faired (streamlined) bicycle that would be safe and have as little frontal and side profile as possilbe, to get the best forwad speed and minimize tipping in side gusts.
One problem with conventional recumbents is a large wheel does not fit readily between the riders knees, leading to compromises such as a small wheel, high rider position, longer wheelbase, and so on.
Milt built "X-2" to experiment with two-wheel steering, to give quick handling and avoid spills in crosswinds; and to experiment with using the fairing as a "sail" to help propel the rider.
In summary, the wheels could be steered in the same direction for balance and opposite directions to change direction. It used handlebars with two degrees of freedom to get this mixture of steering.
X-2 had various steering arrangements over the years. The following is a brief summary, but the total implementation had many more nooks and crannies than described here.
The basic bike is a recumbent tandem with front wheel in front of the rider's feet. Any discussion of performance needs to be qualified that X-2 never had a drivetrain so was used either in coasting tests or with a putt-putt motor mounted to friction drive the rear wheel.
The front and rear head angles are adjustable, as is the fork offset. Adjustments have some minor effect on ride height.
The front and rear steering columns are adjustable, but as far as I know most experiments were done with the front head angle fairly conventional and the rear head angle nearly vertical.
Milt kept copious notes, but they were interspersed with all his other notes about daily life and are written in fairly difficult-to-read handwriting. The notes are currently in the barn at his late brother-in-law's house and occupy several boxes. He took several pages (3"x5" roughly) of notes most days but only a little has to do with X-2.
There were half a dozen steering mechanisms Milt built. The original goal was to build a steering mechanism that allowed the front wheel to be between the rider's knees and turn only a small angle, thus allowing a large wheel; the rear wheel would be used for gross steering. This was originally inspired by riding a Swing Bike(tm) around a small table at the Bicycle Exchange; Milt noted he could make gross steering with the rear wheel and small corrections with the front. This might have allowed a steering mechanism without some of the intrinsic problems of rear-only steering.
This developed in to the idea the front and rear wheels could both steer large angles with the steering used to place the vehicle "crabwise" to the direction of travel and allow a faired vehicle to run with lower wind drag and/or sail in a crosswind.
The basic mental model is the two wheels steer in the same direction, probably the same or similar amounts, to achieve balance without changing direction or without changing much; and the wheels steer in opposite directions equal or nearly equal amounts ("true circle") for changing direction, without changing balance.
For the short-wheelbase version with the front wheel between the knees, this model would not be followed. We debated, but I think never concluded, to what extent steering and balance would be isolated. You can imagine a handlebar with two degrees of freedom but which is constrained. If you try to move the front wheel past lock, the steering simply won't go there. You can also imagine a steering gear in which steering and balance are differential. If you try to steer the front wheel past lock, it steers the rear wheel instead. Neither of these was built, though the first could have been done by simply limiting the range of steering of the front wheel of X-2.
The first steering gear simply moved the wheels parallel; this allowed no change in direction but did allow balance. Despite the long wheelbase and low cg, it gave good balance except that it did not ride no-hands. As far as I know none of the designs acheived self-balance (no-hands). The first gear also tended to drift, but making the front steer slightly more than the rear allowed good ridability. I rode it at speed (40 kph?) with the front and rear wheels at opposite sides of the lane. Under hard braking (as hard as it would brake, not very) it was easy to keep it balanced despite traveling at rougly 45 degrees to a line between the two wheels.
The second steering gear was (IIRC) similar to the first, but a locking handgrip allowed steering, the idea being you would set course and mostly just need to balance.
A later version (maybe the third) used two handlebar halves that could pivot independently, with a centering spring and a lock. This still exists.
A yet later version (which may have been destroyed) used fore-aft sliders each carrying a post with sliders. Fore-aft motion steered, while vertical motion balanced. It suffered from too much friction despite liberal use of needle roller bearings and roughly 75mm pullies for the control cables.
The currently-installed steering gear uses more-or-less the same motions but using a conventional handlebar: pivoting about a vertical axis steers, while pivoting about a horizontal axis along the bike balances. This also uses various hydraulics whose nature escapes me, but I think it is to reduce handlebar forces in some situations.
The last version was never complited -- it uses a planetary-like set of interlocking bevel gears to effect the same motions as above but in a vastly more compact package.
Milt spent a lot of time building and not much time road testing, complicated by his use of carefule/time-consuming construction, and lack of a drivetrain. Also, as X-2 got heavier with adjustable steering, more complicated steering gear, etc. it got heavier with more complicated outriggers and even an air horn. I believe the starting weight was around 30-35 kgf and the final weight perhaps triple that. The lsat outrigger uses two automobile suspension springs! At any rate, this makes it hard to compare early riding test results and late results.
I have ridden only the earlier versions, though the last verion is, I believe, ridable. So any reports I can give only represent the truth insofar as my flaky memory of my understanding of Milt's sometimes-terse comments.
I think the early tests demonstrated two-wheel balance works for a vehicle with long wheelbase and low C, except that self (no-hands) balance was not demonstrated. Early steering experiments show the right mechanism for steering is important for it to be useful. Mid experiments show friction is higly important for a successful steering system (not a new result). Late experiments show an intutivie steer/balance system is possible, modulo that self-balance was (IIRC) never demonstrated. The last (never-installed) system should have replicated the functionality of the last-tested system at much lower weight, thus presumably demonstrating practicality (again, modulo no self-balance).
Milt also experimented with true-circle-only steering, in which the wheels steer equal and opposite directions. I'm not quite sure why: we concluded long before that the only balance effect you would get was from centrifugal force, so you would need to be going quite fast before there would be any substantial balance forces. Milt tried it anyway and confirmed that it was essentially unridable. When I asked why he bothered to test it, I never got an asnwer I understood.
I do not know why self-stability was never demonstrated. It is not obvious to me it is impossible or even difficult, though it is clear the geometries have to be set up so the net force from both wheels under lean is enough to steer the wheels. I expect this is possible, but may be tricky, as even a "neutral" rear axis develops varying caster as the rear wheel steers. However, as a "reducto ad absurdum" case, putting most of the balance motion up front and little motion in back degenerates to the usual front-wheel-only configuration, so configurations near that should be "safe".