(Not) a world of pure imagination: seeing dreams with neuroimaging

People want to remember their dreams so badly that one popular request circulating around the internet is for someone to invent a pillow that magically collects them and plugs into a computer to watch them back. Most people would consider dreaming to be an amazing experience, where outrageous fantasies appear real. After awakening, however, dreams sadly all but disappear from our memories. Several factors influence dream recall, such as age and cognitive functioning, but for the most part, we have trouble remembering them. To aid in dream recall, experts studying dreams suggest actively thinking about the things we see in our sleep once we wake up. Researchers have employed methods such as making participants write about their dreams in a “dream journal” soon after waking up, or simply having participants retell the dream out loud (Schredl, 2007). These methods are obviously not guaranteed to work, as they rely on someone’s short-term memory. For how eager we are to remember our fascinating dreams, our natural capabilities for dream recollection are pretty shoddy. There is surprisingly little that can be done to actually facilitate dream recollection.

The idea of a rectangular sack of cotton absorbing your detailed bedtime hallucinations in order to remember their contents seems a bit far-fetched. Fortunately, there is another method in development that hopes to uncover the contents of our dreams in real time! This method involves reading brain scans and determining which brain regions are most active, and when. Regions of the brain activated are specific enough to indicate that someone is seeing, for example, a face, as opposed to a house, in their dream.

If that seems preposterous, rest assured. This method has also been used on conscious people to check their neurological responses to visual stimuli, and because of this, it is confirmed to work. The key is to understand that the brain is specialized, and brain cells, called neurons, respond to a variety of things in our environments. Within the population of neurons that carry visual information, there are smaller groups of neurons. These smaller groups are activated for specific things that we see (Huffman, 2021). Neuroimaging techniques that measure brain activity, like MRI or fMRI, can be used to assess active neuronal groups and therefore determine what someone is looking at. In studies looking at brain responses to visual stimuli, conscious participants would be hooked up to a neuroimaging machine. Using the faces versus houses example, participants would be shown a neutral image on a screen, and then shown pictures of faces and houses to determine how brain activity changed depending on what they were seeing. Across dozens and dozens of trials, patterns would emerge to demonstrate that some brain activity specifically occurs when we are presented with an image of a face, but not a house. This example is not merely a figment of the imagination; around twenty-five years ago, using neuroimaging, researchers discovered that neurons in an area of the brain called the fusiform gyrus are implicated in facial, but not object, recognition (Kanwisher, McDermott, & Chun, 1997).

Image from Wikipedia

Applying these concepts of neuroimaging to the visual contents of our dreams can be very promising, and gets us closer to that hope of developing a plug-in pillow. If our brains respond depending on what we see in our environments, then they should also have responses specific to what we see in our heads. Since 2013, researchers in Kyoto have been looking into fine-tuning this method to determine what is dancing across our eyelids, even when we cannot remember. The difficulty when compared to testing conscious participants is when participants are awake, researchers can actually control what participants are seeing. There is a lot more guesswork involved when people can conjure up absolutely ridiculous thoughts in their sleep, leaving researchers unable to control which visual inputs lead to which brain imaging outputs.

To combat this issue, researchers have connected participants to an fMRI machine and let them sleep for a little while. Participants are then woken up and asked to give reports about what they saw in their dreams. After being asked to go to sleep again, the process would cycle, yielding lots of data. Working backwards in this manner, researchers have been able to piece together what people see in their dreams (Horikawa et al., 2013). However, there is a flaw in this method, which is that people might be incorrectly reporting what they dreamt. Study participants are unlikely to be pathological liars, but misremembering and forgetting things are part of the human experience, especially in dreaming.

A more conceptual issue for the neuroimaging approach to gather dream contents is that, while neurons can activate in response to very specific stimuli as described, they can also be activated to more general things. For instance, during viewing experiences in the environment, the average neuron in your brain’s medial temporal lobe codes for a considerable range of objects. You might be surprised to learn that, while still different, your neuronal firing patterns from seeing either a cow or the Statue of Liberty are rather similar, even though the objects are obviously nothing alike (Valdez et al., 2015). Although your fusiform gyrus lights up at the vision of a face, it is harder to pinpoint exactly who the face belongs to. Therefore, when analyzing dream contents, even a neuroimaging expert would have trouble figuring out if you were dreaming about, say, Bill Clinton or David Letterman (for your sake, I hope you are not dreaming about either of them).

There is much left to be desired when it comes to decoding the contents of our forgotten dreams. Promising headway (pun intended) has been made with neuroimaging, although the technique can certainly be improved upon. The function of dreams and their components is unclear, but learning more about our dream contents would be helpful to uncover more information. However, the ethics of revealing someone’s dreams are worth considering. Would dream neuroimaging be a violation of privacy? Akin to the third episode of Black Mirror, being able to reveal someone’s innermost visualizations with technology may be too invasive and personal for the average Joe. This particular Black Mirror episode explores the recording of someone else’s memories, and although dreams should not be confused with reality, humans may be better off leaving them in a world of pure imagination.


Horikawa, T., Tamaki, M., Miyawaki, Y., & Kamitani, Y. (2013). Neural decoding of visual imagery during sleep. Science, 340(6132), 639-642.
Huffman, D. (2020). Electrophysiology: Single cells and representations. Lecture.
Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face area: a module in human extrastriate cortex specialized for face perception. Journal of Neuroscience, 17(11), 4302-4311.
Schredl, M. (2007). Dream recall: Models and empirical data.
Walker, M. (2017). Why we sleep: Unlocking the power of sleep and dreams. Simon and Schuster.
Valdez, A. B., Papesh, M. H., Treiman, D. M., Smith, K. A., Goldinger, S. D., & Steinmetz, P. N. (2015). Distributed representation of visual objects by single neurons in the human brain. Journal of Neuroscience, 35(13), 5180-5186.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s