On vacation, I did a bunch of geometry / axes study, as well as drawing topview of X while viewing sideview.
This seems to have improved my visualization clarity somewhat, evidenced by easier adlibbing especially with nonlinear axes. This is interesting because it really fits into my earlier 'deconstruction needed' guess : larger deconstructive vocabulary seems to lead to better visualization.
Ptoing's "I know how letters look" comment is interesting because I just thought about this and realized: I know how letters look, but I draw them entirely procedurally; Despite lots of experience with making typography, I do not visualize them before I draw, except in two respects:
1. How the -current- stroke looks (now / when completed), especially the divisions of space and angles it creates.
2. The overall space a letter should occupy (and the padding it should NOT occupy). This latter is a bit vague and greatly aided by dual-wielding (adding another pencil in left hand, that just maintains a physical location for the origin; tying my proprioception into my spatial sense perhaps.)
At the moment I'm most interested in determining a exercise that will make #2 less vague -- possibly something like visualizing negative space of a mildly complex layout of boxes, and blind drawing it, in 2d and 3d.
Personally I would not be surprised to find lettering is a special/nonstandard case, since it ties in so strongly to language. Do deaf -- especially totally deaf -- people have a different relationship to lettering, for example? More directly visual (ie. experiencing an actual graphic, more than a symbol), or less directly visual?