I think you might be right, and interesting, I did not know it had been retracted. Is the retraction trying to retract some claim that people who lack a consistent model are incapable of programming, or something more? (As my original read of the paper was more that the lack of a consistent model was the cause behind poor performance, but not an unsolvable problem, and more of a great hint as to how we could or should approach teaching CS. However, while TA'ing, I did occasionally notice such a behavior, and tried to point it out to the student, and that in and of itself turns out to be harder and trickier than one thinks to communicate, like trying to describe the color blue to someone who hasn't seen it. And, of course, TA'ing is hemmed in by the real-world constraints of too many students, too little time.

> It's just that because these courses are self-selected, the second hump has a small number of people.

By second hump, you mean the weaker group (the "inconsistent" group, from the papaer)? At my college, the early CS classes — which was where I spent most of my time as a TA — were also taken by some students of non-CS majors, as some majors required two semesters of CS. (These were primarily major whereby one might need to say, write a computer model to help validate/invalidate something more related to one's field.) So, they weren't all their completely voluntarily, I suppose?

But we definitely had some self-selected students, those who had clearly been self-teaching themselves computers and programming from early in their teenage years. However, our college also permitted one to voluntarily skip one or both of the first two semesters, if one felt one was strong enough to do so. (There was a grace period at the start of the semester whereby you could still transition between the classes, if you found you had bit off more than you could chew.) So, those clearly strong students who had already been independently studying in the field prior to university could skip over some of the courses.