[Ben_Dean]

There was a study in a CS dept. somewhere that had a pretty convincing test. I can't remember where I saw it or the school, so I'll just try to recount it.

The test was given to students before the first bit of instruction in computer science (and for the sake of simplicity, let's assume it controlled for those with previous experience). Basically, students were given a problem set of code snippets in a non-existent language, and were asked to write what the code evaluated to, i.e. 34 @@ 14 --> [2], followed by possible answers.

It turned out that the students who performed well in their subsequent CS courses were not those who got the right answers, or those who got the wrong answers, but those who answered consistently. The best single predictor of a good student of CS (not strictly equivalent to a programmer in a startup...) was forming a consistent model of what they were looking at, and sticking with it.