In all seriousness, my more-than-slightly-snarky hypothesis does basically suggest the same thing that Dude's more formal analysis does.

Higher SES kids get continuous reinforcement...

and the low SES kids don't.

But it's entirely consistent with madeinuk's observation that they didn't really "learn" anything. Neither did the higher SES kids, but that is masked by the continuous reinforcement of the conditioned responses, so it appears that they retain it all.

It's not really about retained learning. It's about non-extinguished operant conditioning.

Which is what NCLB has given us.

Schrödinger's cat walks into a bar. And doesn't.