Home / Sleep News / Dream On: A New Theory on Why Strange Dreams May Help Us

Dream On: A New Theory on Why Strange Dreams May Help Us

Written by

Olivia Murillo

author
Fact Checked

The purpose of the vivid, narrative hallucinations that most people have every night 一 dreams 一 has evaded philosophers and neuroscientists alike. Leading theories suspect that dreams, may support memory, emotional regulation, creativity, problem-solving capabilities, or some combination thereof. Another popular idea is that dreams actually serve no purpose at all.

“While this theory dubbed the Overfitted Brain Hypothesis (OBH) requires further research, it suggests that the strange nature of dreams may in part serve to make our perceptual and cognitive systems better equipped to handle new situations.”

Recently, Tufts University neuroscientist Erik Hoel threw a new theory into the ring in a recently published review. He borrowed ideas from the rapidly advancing field of artificial intelligence to explain why dreams might indeed serve an important function.

While this theory dubbed the Overfitted Brain Hypothesis (OBH) requires further research, it suggests that the strange nature of dreams may in part serve to make our perceptual and cognitive systems better equipped to handle new situations.

What Is The Overfitted Brain Hypothesis (OBH) Theory?

Hoel uses concepts of artificial intelligence to explain this theory. Highly sophisticated algorithms, like those behind self-driving cars and virtual assistants, are trained on a lot of data to perform tasks. Hoel argues that this is not unlike our daily experiences, which teach our cognitive and conceptual systems to interact with the world. For example, it becomes second nature to commute to work, interact with people, and navigate our local grocery store.

What these familiar experiences and routines do not prepare us to handle, however, are unexpected situations that diverge from them. Computer algorithms can run into this problem, known as overfitting: if the data used to train it is too similar, an algorithm can end up performing significantly worse data that it has not seen before.

For example, if a hypothetical algorithm designed to identify dogs is trained on images of golden retrievers, German shepherds, labradors, and jack russells, it may mistakenly exclude Boston terriers and bulldogs because it was lead to believe from the images it was given that all dogs must have long tails.

One solution to avoid overfitting and promote generalizing abilities is to make the training data less self-similar by including some poor quality data, known as “noise.” In other words, throw the algorithm a few curve balls to help it better understand more situations.

Hoel argues that dreams may serve a similar purpose in providing poor quality data to our cognitive and perceptual systems to prevent them from overfitting to our daily experiences. The “poor quality data” of our dreams might appear as random locations or items in a dream sequence, or perhaps the lack of certain elements, like missing text or details on a phone screen.

“Through these vivid dreams that provide us with exposure to outside-of-normal sensory information, we can improve how we respond to new experiences, because we have enhanced our ability to generalize and adapt to them.”

This is Hoel’s overfitted brain hypothesis (OBH). Through these vivid dreams that provide us with exposure to outside-of-normal sensory information, we can improve how we respond to new experiences, because we have enhanced our ability to generalize and adapt to them.

OBH Theory Builds on Other Research

Hoel draws support for the OBH from existing research and what has been discovered while attempting to replicate some human capabilities with deep-learning. He notes that one of the most effective ways to trigger specific dream content is to overtrain someone on a repetitive task (which even worked in patients with amnesia trained on the game Tetris), and cites examples where dreaming about a task was shown to improve performance on that task, such as reading with inverted goggles.

Additionally, there is evidence of never-before-experienced patterns of brain activity during this rehearsal during sleep that is not the exact same as the experience itself, as demonstrated in rats in a study from Carnegie Mellon University.

He also finds that the anecdotal and scientifically-supported association between dreams and creativity fits well with the OBH, for example, with the melody for “Yesterday” coming to Paul McCartney in a dream as well as evidence that rapid eye movement (REM) sleep, which is when the most vivid dreams taken place, has been shown to shown to improve creativity via associative networks in the brain.

OBH Theory Requires Further Research

Critically, the OBH does not necessarily conflict with other ideas of the functions of sleep, such as the clearing of cellular metabolic waste or memory consolidation. Nor does it compete with all other theories on dreams. Rather, Hoel explains that the OBH should be considered as an “umbrella” hypothesis, simply proposing that experiences of the dreaming brain may improve the daily task performance of the waking brain.

Further empirical studies and theoretical considerations are required to approach any solid answers explaining these strange nightly escapades. As reviews like this one attempt to connect the dots of the studies and theories that are dramatically and rapidly expanding our knowledge of the diverse functions of sleep, dreams themselves, similar to trying to remember one in the hours after awakening, remain elusive.

  • Was this article helpful?
  • YesNo

About Our Editorial Team

author
Olivia Murillo

Contributing Writer

About Our Editorial Team

author
Olivia Murillo

Contributing Writer

References

Learn more about Sleep News