Well, it _should_ come from there - non-blind people tend to "cheat" and use their vision instead. (or more generously: as humans we have a lot of overlapping and redundant systems to manage our body in space - another fascinating example is depth perception - actual stereo vision accounts for only a tiny percentage of it, while things like color gradient, relationship with the horizon, relative size, overlap and so on are much more important, even though they are technically not depth perception at all (though it is probably a good thing- movies would be really confusing if actual depth perception was more important))
Similarly I think as dancers this is one of the big problems in learning (and teaching) - in general we see how a movement looks, and then recreate the look, and then bootstrap ourselves into an idea what the alignments and muscles and balance might underlie this. Just like balance we don't have a good way to communicate internal perceptions. Some of my most helpful lessons were lessons where my teacher danced with my partner, and i shadowed him with my hands on his back and waist and spine to be able to feel what and how he actually moved.
