The one thing I don't see in stuff written by people who are committed to the idea that everyone can be taught to program is before-and-after mark distributions

The one thing I don't see in stuff written by people who are committed to the idea that everyone can be taught to program is before-and-after mark distributions, demonstrating how we can get rid of the apparently robust bimodal distribution that has dominated first year computing courses for decades. It's great to contend that everyone can be taught to program, but decades of research have yet to answer the question of "How?" Or even "Why?" Many people have very strongly felt opinions, but there is a surprising paucity of data. The idea that some people get it, some don't is not at all ridiculous: it is a perfectly plausible hypothesis given the robust bimodal mark distribution seen in first year computing, which differs from what we see in math or the sciences. The problem is no one has been able to find anything that correlates well enough with this to be predictive, and if "not getting it" and "getting it" were features of humans we might expect to be able to measure them in ways other than teaching first year computing. On the other hand, no one has ever been able to get rid of that bimodal shape (so far as I know), so there is an equal and opposite knock against that hypothesis. I lean toward the "inherent" hypothesis and have some ideas as to what might underlie it, but would be easily convinced otherwise if anyone could actually generate an intro computer curriculum that resulted in a mark distribution that looked more like what we see in first year math or physics.