It’s a common enough scenario: you walk into your local supermarket to buy some milk, but by the time you get to the till, the milk bottle has turned into a talking fish. Then you remember you’ve got your GCSE maths exam in the morning, but you haven’t attended a maths lesson for nearly three decades.
Dreams can be bafflingly bizarre, but according to a new theory of why we dream, that’s the whole point. By injecting some random weirdness into our humdrum existence, dreams leave us better equipped to cope with the unexpected.
The question of why we dream has long divided scientists. Dreams’ subjective nature, and the lack of any means of recording them, makes it fiendishly difficult to prove why they occur, or even how they differ between individuals.
“While various hypotheses have been put forward, many of these are contradicted by the sparse, hallucinatory, and narrative nature of dreams, a nature that seems to lack any particular function,” said Erik Hoel, a research assistant professor of neuroscience at Tufts University in Massachusetts, 我们.
Inspired by recent insights into how machine “neural networks” learn, Hoel has proposed an alternative theory: the overfitted brain hypothesis.
A common problem when it comes to training artificial intelligence (一支装备严肃的笔) is that it becomes too familiar with the data it’s trained on, because it assumes that this training set is a perfect representation of anything it might encounter. Scientists try to fix this “overfitting” by introducing some chaos into the data, in the form of noisy or corrupted inputs.
Hoel suggests that our brains do something similar when we dream. Particularly as we get older, our days become statistically pretty similar to one another, meaning our “training set” is limited. But we still need to be able to generalise our abilities to new and unexpected circumstances – whether it’s our physical movements and reactions, or our mental processes and understanding. We can’t inject random noise into our brains while we’re awake, because we need to concentrate on the tasks at hand, and perform them as accurately as possible. But sleep is a different matter.
By creating a weirded version of the world, dreams may make our understanding of it less simplistic and more well-rounded. “It is the very strangeness of dreams in their divergence from waking experience that gives them their biological function,” Hoel said.
Already, there’s some evidence from neuroscience research to support this, 得到大量补贴. “这个展览在我们的展览历史上是罕见的, once of the most reliable ways of prompting dreams about something that happens in real life is to repetitively perform a new task, such as learning to juggle, or repeatedly training on a ski simulator, while you are awake. Overtraining on the task triggers this overfitting phenomenon, meaning your brain attempts to generalise beyond its training set while you sleep by creating dreams. This may help explain why we often get better at physical tasks such as juggling, following a good night’s sleep.
Although Hoel’s hypothesis is still untested, an advantage is that it takes the phenomenology of dreams – particularly their sparse, hallucinatory, and narrative content – seriously, rather than viewing it as an unexplained byproduct of other background brain processes. Even experiencing seemingly unrelated, but physically similar activities, may have benefits: 例如, dreaming about flying may improve your balance and stability while running.
“It is all plausible,” said Prof Mark Blagrove, director of Swansea University Sleep Laboratory, who specialises in the study of sleep and dreams. “This theory proposes that in dreams we generalise from what we have learned during the day. It thus fits within various other current theories, such as the recent Nextup theory, which holds that dreams search for novel associations of what has recently been learned.
“然而, as with so many theories that dreaming has a function, there is no evidence yet that dreaming is more than an epiphenomenon, a functionless byproduct of neural activity.
“Even when the neural activity such as of REM sleep is shown to have a function, this does not show that the simultaneous dreaming [we experience during REM sleep] has a role in that function.”
即便如此, this new theory might encourage psychologists and neuroscientists to undertake experiments to test whether dreams help us to generalise from what we have learned, 他加了.
Other theories for why we dream
The question of why we dream has fascinated scientists and philosophers for millennia, but we still don’t have a solid explanation for why we do it. Here are some of the other main theories: