Skip to main content

Verified by Psychology Today


Dreams Are Even Weirder Than You Think

Dreams aren’t random, and what does and doesn't happen in them is quite strange.

Key points

  • Dreams have peculiar constraints. Certain perceptual experiences are highly unlikely to occur, such as smells.
  • Visual imagery in dreams does not adjust to your head's orientation during sleep. This has implications for why we dream.
  • Far from being utterly random, it is common for dreams to be quite repetitive.
Andrea Piacquadio/Pexels
Source: Andrea Piacquadio/Pexels

It seems like anything can happen in a dream, and this invites creative interpretations. Dream analysis is nearly universal among human cultures, and features prominently in Indigenous American and Australian cultures. We want to find order and meaning in the chaos. But if dreams were totally random, any analysis wouldn't be very meaningful.

In fact, dreams are not totally random. Our perceptual experience during dreams is constrained in fairly predictable ways. Certain experiences are rare or nonexistent. And dreams can be quite repetitive. These features make dreams all the more strange, and examining them may help us from a neuroscientific perspective to understand why we dream at all.

Dreams Rarely Involve Smell

For people born with sight, dreams are primarily visual. It is common for dreams to also include the experience of sound and, less often, a feeling of touch and body movement. But experiences of taste and especially smell are quite rare. Less than half of people report having experienced an olfactory dream in their lives. In one study, much less than 1 percent of men reported olfactory experiences upon waking, while for women the figure was slightly above 1 percent.

In dreams, you might integrate stimuli around you into the dream, such as a ringing phone. This does not seem to be the case with smells. A particular odorant present while you sleep is highly unlikely to appear in your dream. However, we can integrate the valence of an odorant into a dream: Putting something nice-smelling (a rose) under a dreamer’s nose tends to elicit more positive reported experiences than smelling something noxious (rotten eggs), even though the dreamer recalls no experience of the particular odor.

The rarity of smell in dreams may be because smell circuits, being a very ancient function, follow distinct pathways into the cerebral cortex and have less overlap with vision and auditory networks. Interestingly, though, people born blind, who do not experience visual dreams, are much more likely to report experiences of touch, taste, and smell.

If dreams really are about shoring up our recent memories and “consolidating” them for long-term storage, as some argue, it is odd that smell would be rare. Smell is intimately related to memory for places and navigation in those places (spatial memory). We all know the experience of a particular odor transporting us back to a specific place in our past; grandma’s kitchen, for example. This implies that dreams may not be primarily about memory consolidation, at least with regard to spatial memory. If this is the case, it also helps explain why our sense of space is distorted and illogical in dreams.

Dreams Are Never “Tilted”

A curious limitation of visual imagery in dreams has not been noted, as far as I know. We always experience a visual dream upright, regardless of how our heads are positioned while we dream. In normal, wakeful vision, our sense of which way is “up” changes as our head moves and turns in space. That dreams are not like this is self-evident. If you lie with your right ear on the pillow, your dream should be tilted 90 degrees, but this is not what happens. It feels upright.

To me, this suggests that we do not just carry on “seeing” with our eyes closed during dreams. Nor are dreams just snippets of daily experience, some of which would necessarily be tilted. Instead, dream imagery seems to operate in an abstract space of knowledge, and understanding that discounts what our bodies know about their position in space, during both dreams and wakefulness.

The uprightness of dream imagery is related to one recent proposal of the purpose of dreams. Consciousness researcher Erik Hoel argues that our lives are rather predictable, and we need the seemingly random experiences of dreams to help us generalize our knowledge beyond what we have already seen. He takes as inspiration the current trend for comparing brain operations to “deep learning” artificial intelligence (AI) systems. Though they can be trained to perform certain tasks like detecting faces in an image almost perfectly, these computer systems are notoriously prone to getting stuck in a rut. They may perform extremely well in contexts they have seen before but fail utterly when faced with the unfamiliar. An object that the system previously recognized with no problem in thousands of previous images may fail to be identified when the object is in a context never seen before, or when parts of the object are blocked by other objects. In these situations, improvement can be gained by shaking up the AI and feeding it random images for a while. The random images serve to nudge the system in new and previously unexplored directions.

Hoel argues that dreams do something similar for our perceptual systems. Random dream imagery, produced by random neural activity, serves as a kind of offline training program for our perceptual systems so we are more ready for things we haven’t yet experienced during wakefulness. All this should happen in an abstract space of idealized objects rather than based on snippets of sensory input, which don’t necessarily take account of where your head is currently pointed. The fact that dreams are always upright was not noted by Hoel but is consistent with his proposal.

Dreams Can Repeat

However, there is one crucial fact about dreams that is usually ignored: This blind spot afflicts Hoel’s hypothesis, as well as most other neuroscientific theories about the purpose of dreams. We ignore the fact that they can be highly repetitive. Dreams that recur in seemingly minute detail are common—and they can be terrifying. People with posttraumatic stress disorder are known to experience intense periods of REM sleep and highly repetitive dreams. War veterans report having the same dream about battle experiences long after the events. This is a problem for Hoel’s idea: If dreams are the result of random neural activity, the chances are vanishingly small of generating the same pattern more than once across millions or billions of neurons.

Dreams Are Weird, and Not Just in the Ways We Typically Notice

We may never be able to understand why we dream specific imagery or experiences. But accurately assessing the perceptual characteristics of dreams may help us understand why we dream at all. From this perspective, we can see that dreams are still decidedly weird, but in interesting and potentially informative ways.

We can also take an evolutionary view. We can ask: What other species have dreams? When might dreams have first appeared in evolution, and what does this tell us about our dreams?

Copyright © 2022 Daniel Graham. Unauthorized reproduction of any content on this page is forbidden. For reprint requests, email


Hoel, E. (2021). The overfitted brain: Dreams evolved to assist generalization. Patterns, 2(5), 100244.

Hurovitz, C. S., Dunn, S., Domhoff, G. W., & Fiss, H. (1999). The Dreams of Blind Men and Women: A Replication and Extension of Previous Findings. Dreaming, 2(9), 183-193.

Ribeiro, S., & Nicolelis, M. A. (2007). The evolution of neural systems for sleep and dreaming. Evolution of Nervous Systems, Vol 3. Eds. J. Kaas and L. Krubitzer. Elsevier.

Schredl, M., Atanasova, D., Hörmann, K., Maurer, J. T., Hummel, T., & Stuck, B. A. (2009). Information processing during sleep: the effect of olfactory stimuli on dream content and dream emotions. Journal of Sleep Research, 18(3), 285-290.

Stevenson, R. J., & Case, T. I. (2005). Olfactory imagery: A review. Psychonomic Bulletin & Review, 12(2), 244-264.

More from Daniel Graham, Ph.D.
More from Psychology Today