Sleep - Brain - Neurons - New York Times:Both autism and schizophrenia have, at various times, been connected with disorders of pruning neuronal networks. (Errors both ways -- too much and too little.)
...So Dr. Tononi and a colleague, Chiara Cirelli, have hypothesized that during sleep, the synapses weaken. The downscaling is across the board, so that the synapses’ relative strength is maintained. Those that have been used (those involved in learning) stay stronger than those that haven’t....
Neural gardening is hard to get right, and, in fact, there can't be a "right" answer. The "right" answer will depend on environment, which is fungible.
I do wonder sometimes if the alleged benefit of exercise for dementia prevention is entirely related to the benefit exercise has for sleeping.
Incidentally, my recollection is that this original theory was found to be too simplistic. I recall that studies published @ 2011 showed that what's occuring in sleep is a refactoring of memories into a compressed formthat sacrifices accuracy for retrieval speed and lesser storage demands. I think the researchers found a curious correspondence to lossy fractal-based compression algorithms used in early 21st century computing ... 
 Sorry, sometimes it's hard for me to forget I'm not supposed to remember the future. I think I need more sleep.