Commenter Rob has an observation in support of the point of view of the experimenter:

Seriously cool video. Two lessons were learned by a man who learned to ride an impossible to ride bicycle: knowledge does not equal understanding, and whether you believe it or not, you’re looking at the world with hard bias.

However, commenter Penax has this very astute reply to the topic of the video:

Riding a bicycle is mostly not a conscious process.  It involves a number of at least semi-autonomous feedback loops to manage what is actually a surprisingly complex (and somewhat nonlinear) control problem, and most of those feedback loops actually exist in the cerebellum, if I remember correctly.  Motorcyclists (and some high performance drivers) often refer to this semi-autonomous process (which is much faster than conscious analysis) as “lizard brain”, because it can react in surprisingly quasi-instinctive ways and sometimes get in the way of a conscious response. Most kinesthetic memory involves training the cerebellum in one way or another, and it does respond to changes in process models, albeit quite slowly and with a great deal of required repetition. (I run into this a lot in international folk dancing. There are some dances I know I’m doing wrong but my body learned then that way and it’s too much bother to retrain it at this stage..)

Cool stuff!  Thanks for making the video and to the commenters!