Possible explanation for YouCantLearnSomethingUntilYouAlreadyAlmostKnowIt
Assume (at least) two brain modules: one for 'fuzzy' processing and one for 'symbolic' processing.
The fuzzy module might work basically more or less like the well-known NeuralNetwork
s and recognize patterns in the input (which might be experiences or knowledge). The symbolic module can take single input items (which can again be knowledge or, for example, output of the fuzzy module) and toss them around (what we call conscious thought). Compare them, infer other bits, combine them.
If we take 'learning' to be 'understanding on the symbolic level' (which it surely isn't for everyone; consider RoteMemorization
), you need some actual structure formed by the fuzzy module to do so. If you haven't been exposed long enough for the fuzzy module to pick up any patterns, any explanations will be dry and without life (possibly connecting knowledge but without the underlying connection to reality by the fuzzy module).
And if, on the other hand, all the connections are already there (the fuzzy module has picked up all patterns there are), then learning on the symbolic level can become uninteresting too, because no new patterns about reality can be added (you feel that you already know everything there is).
So both levels must be trained, but with the fuzzy module earlier. As a diagram (cool AsciiArt
| symbolic ___________
| .--´´ __.-´´
| fuzzy ______/__.--´
| .-´ /
See also: ZoneOfProximalDevelopmentTheory