I've got another guest post over at Discover magazine: Is the Purpose of Sleep to Let Our Brains “Defragment,” Like a Hard Drive?
It's an expanded version of two Neuroskeptic posts(1,2) about the theory that the job of slow-wave sleep is to prune connections in the brain, connections which tend to become stronger while we're awake and might become too strong without periodic resetting.
One of the commenters on the Discover post pointed out that this idea a bit like a much older idea about sleep, from Francis Crick (of discovering-the-structure-of-DNA fame). Back in 1983, Crick and Graeme Mitchison proposed that dreaming sleep serves to help us "unlearn": The Function of Dream SleepTheir idea was a bit different, but it was really very elegant.
The sleeping brain, they said, is cut off from real sensory input, and is subject only to essentially random activity variations. However, sometimes these meaningless inputs may be 'interpreted' as having meaning, activating representations (concepts, thoughts, memories) that we've learned to recognize when awake.
These "patterns" that the brain wrongly "sees" in noise are what we experience as dreams. Crick and Mitchison's point is that, ideally, a pattern recognition system (like the brain) shouldn't be picking up patterns from random noise because that would be a sign that it was biased in favor of those patterns - "obsessed" with them, as it were, and liable to see them everywhere. So it would be good if there were some way of identifying the brain's overlearned biases and (partially) unlearning them.
That's what dreams do, somehow, according to Crick. It's like if dreams are a self-administered Rorschach test that the brain uses to work out what's "weighing on its mind"! Incidentally, this is an idea I once suggested (much less clearly) myself.
It's a beautiful and ingenious theory, although as the authors admitted, it would be very hard to test and it leaves wide open the question of how dreams could cause memories to be "unlearned" or indeed whether unlearning is even possible in the brain. It's also not very similar to the modern "defrag" theory, because Crick was talking about the dreaming rapid eye movement stage of sleep, not slow-wave sleep.
Francis Crick and Graeme Mitchison (1983). The Function of Dream Sleep Nature, 304, 111-114