Why do humans hallucinate on little sleep?

Driven by bewilderment, a hunch, and a sense of purpose, I set out to determine how sleep deprivation causes visual hallucinations. My quest turned up a key paper (written entirely in German), a new hypothesis and a vow to get more sleep. Below is an exposition of my journey, beginning about a decade ago.

The Quest

The spring after I graduated from college, I did one of those things you only do when you’re young and mad: a friend and I drove from the mountains of North Carolina to the coast and back in a single day. Nearing the end of our six hour return trip, about 30 miles outside our town, my friend gazed to the side from the driver’s seat and saw a polar bear on the side of the road. To be clear, western North Carolina, even in the chilliest of springs, isn’t hospitable to polar bears. My friend certainly hallucinated.

Fast-forward a few years. I was sitting in a seminar one evening after work and inadvertently dozed off. When I opened my eyes, I noticed that the coffee stain on my jeans had grown legs and started crawling away. It took a good 10 seconds for that coffee stain to stop moving. What trick of visual perception caused this illusion of motion where there was none?

It is well documented that a common side effect of sleep deprivation is hallucination, as evidenced by, but not limited to, long-distance swimmers, ultra marathon runners, and overly ambitious college students. According to one Stanford researcher, at least 80% of people will hallucinate if severely sleep-deprived, (“severe” meaning anything from getting only a few hours of sleep in a single night to going days without sleeping). Given such a high frequency, there must be a physiological basis for those all-nighter induced visions of things not really there. Right?

The Questions

Surprisingly, despite the regularity with which we hallucinate, there has been little research to explain what’s going on in our brains when we do so, particularly during states of sleep deprivation or sleep-wake transitions. Many people describe hallucinating as dreamlike, as if part of the brain is dreaming while the rest of the body is awake. When I observed the crawling coffee stain as I awoke in the dimly lit lecture hall, I thought I might be dreaming. But I knew I was awake and that what I was seeing was really there. Or at least partially there.

Hallucinations and dreams are both so close to sleep, but in his book, Hallucinations, the late renowned neurologist Oliver Sacks, attempts to differentiate them: “Hallucination is a unique and special category of consciousness and mental life,” distinct from imagination and dreams, writes Sacks. He goes on, however, to compare hallucinations and illusions: While illusions are misperceptions of objects that are already there, hallucinations are misperceptions “conjured out of thin air.”

But nothing in neuroscience is ever that simple. If you squint your eyes enough, you’ll blur the line separating hallucination and illusion. In other words, visual misperceptions often exhibit hallmarks of both. When I suddenly awoke during that seminar, did I hallucinate a spider crawling across my leg, or was it an illusion of a coffee stain in motion? Alas, to the severely sleep-deprived, maybe it doesn’t matter whether that baby dancing to “Hooked on a Feeling” is a hallucination or an illusion. As it turns out, lack of sleep disturbs visual processing, which results in false perceptions that can manifest as hallucination, illusion, or both. Or, as I found out, failed illusions.

Ironically enough, it’s the failed illusions that offer the most insight into the hallucinating brain. In some ways, being “bad” at perceiving an illusion is (and I’m taking some liberties here) an index of being “good” at hallucinating. Whatever failure of visual inference inhibits a person from seeing an illusion could also cause that person to hallucinate. But before I could reach this conclusion, I had to brush up on my German.

The German Nurse Experiment

My naïve searching of PubMed and relentless reading of Hallucinations initially turned up zero clues suggesting what in our brains makes us hallucinate when sleep-deprived. Eventually, it was an unlikely source that clued me in to a possible connection: a German article in an obscure journal with no English translation to be found. Fortunately, a lab mate of mine speaks fluent German, and, in the spirit of collaboration, agreed to translate the findings for me.

In short, the study showed that when nurses working the night shift were given a test of visual perception after the first, third and seventh shifts, they repeatedly failed it. The test they were given, called the Binocular Depth Inversion Illusion Test or “BDII,” makes use of a basic principal of visual processing -- that we normally see what we expect to see based on prior experience regardless of what’s actually there. The nurses failed this test when shown flowers (blumen), a house (haus) or a patio chair (gartenstuhl), but not when shown a face (gesicht), and they eventually passed the test after catching up on sleep for a week.

In order to understand these findings, we first have to dive into the BDII and what it can tell us about visual perception.

The Binocular Depth Inversion Illusion Test

In the vision field, the detection of visual signals, or photons of light, hitting the retinas is known as “bottom-up” processing. The brain must then interpret and make inferences about these signals. Incidentally, our brains are impressively good at doing this. Without your even being aware of it, visual centers in your brain are right now receiving impulses generated by photons of light reflecting onto your retinas and making predictions about what you’re sensing based on your expectations of what’s real and possible, correcting for anything that seems impossible. This correction is a fascinating feature of visual perception and is known in the field as “top-down” visual processing. Normal visual perception is the optimal combination of bottom-up input and top-down correction. Normally, what we expect to see, what we “think” we see, matches what’s actually there in front of us. But sometimes it doesn’t.

What do psychosis, psychedelics and sleep deprivation have in common? They make you really bad at perceiving visual illusions. And really good at hallucinating. 

The BDII is a beautiful example of your brain overriding actual visual input. This illusion demonstrates that when you’re presented with a familiar 3D object (think of a face) that has been inverted so that it’s now concave (now visualize a hollow mask), your brain’s visual system so strongly expects the object to appear convex that it automatically “corrects” the unusual input, creating an illusion more consistent with past experience. There’s no better way for you to understand this illusion than to try it yourself.

The BDII test was developed to check your ability to perceive this illusion -- an illusion you’re supposed to have. During periods of prolonged sleep deprivation, the German nurses weren’t able to correct inverted objects, such as a house or a patio chair, but their brains were still able to compensate for inverted faces. For these sleep deprived but otherwise healthy nurses, top-down correction had become compromised.

A failed BDII test -- or failure at perceiving the inversion -- suggests defective visual perception. But does it correlate with a susceptibility to hallucinate, to see polar bears on the side of the highway or coffee stains moving moving against denim? Were any of those nurses hallucinating?

Schizophrenia, Psychedelics and Sleep

Neuroscientists are into sleep deprivation in the same way they’re into psychedelics. Both are used as a means to disrupt cognitive function in order to better understand normal and abnormal cognition. Sleep deprivation and psychedelic-induced trips are reversible whereas psychosis isn’t. But like psychosis, both psychedelics and sleep deprivation can cause hallucinations, so functional studies of the brain in any of these states can hint at the neural processes that contribute to the susceptibility to hallucinate. Functional imaging studies of a hallucinating schizophrenic brain are complicated and somewhat rare. Data from BDII tests, however, are more straightforward and surprisingly common.

The next logical question, then, is: What do people with schizophrenia see when given the BDII test? It turns out, they see a hollow mask (or fail to see the inversion), especially during acute states of the disease. Like the sleep deprived nurses, they are not able to perceive the illusion. And people high on cannabis? They can’t see the illusion either.

There must be something to this: people who characteristically hallucinate fail to perceive the exact same illusion. So do they all hallucinate in the same way? 

To answer this question, I searched for evidence that, given a certain task, a sleep deprived brain will look similar to a schizophrenic brain. Fortunately, the BDII is so commonly used as an early indicator of psychosis that it wasn’t hard to find imaging data of the brains of schizophrenic patients experiencing (or “resisting”) the hollow mask illusion. When examined via fMRI, schizophrenic brains undergoing the BDII test showed different connectivity patterns than non-schizophrenic brains. This means that the diseased brain has different strategies for minimizing visual errors. It “overweights sensory evidence” at the expense of internal correction. To use more technical language, the schizophrenic brain tends to rely more heavily on bottom-up processes, while the normal brain relies on top-down counter-balancing to correct implausible visual stimuli, making it “see” the illusion.

In the psychotic state, the brain has trouble balancing the interpretation of sensory input and the generation of correct “guesses.” This suggests that something similar may be happening in the sleep deprived state, leading to the susceptibility to hallucinate, or “guess” something to be there that’s not actually there. So, risking the rabbit hole of scientific literature, I searched for evidence that the sleep deprived brain also has trouble balancing bottom-up and top-down processes. 

The Sleep Deprived Brain

Sleep deprived college students are a common bunch. They also prove valuable test subjects for visualizing what the brain looks like on little sleep. In a study at the National University of Singapore, twenty-seven undergraduates underwent an fMRI scan while performing a visual attention task. In the first stage of the experiment, the students were well rested, and in the second stage, they had just undergone a night of total sleep deprivation. Perhaps not surprisingly, the students performed worse at the task when sleep deprived. Results from the fMRI indicated that this was due to defective top-down visual processing.

In another study at Brookhaven National Laboratory, fourteen healthy adult men were asked to perform a similar visual attention task while undergoing fMRI in the well rested state and again in the sleep deprived state. In the well rested state, the harder the task was, the more the top-down network was activated and the bottom-up network was deactivated. But in the sleep deprived state, this pattern fell apart: top-down biasing was no longer strengthened in comparison with bottom-up visual processing. Like schizophrenic patients taking the BDII test, sleep-derived subjects “overweighted” bottom-up sensory evidence.

Ways of Seeing

A healthy brain is complex; a hallucinating brain is mysterious, perplexing. No two hallucinating brains are exactly alike. In mental states that are characteristically prone to hallucinations, however, the brain consistently has trouble making predictions about visual input. Whether high, sleep-deprived or psychotic, the visual system can’t invert hollow faces or perform well at visual attention tasks. I still can’t say exactly what about a sleepy brain induces hallucinations, but this could partly explain how hallucinations occur: the brain becomes bad at “guessing” based on visual clues and, just maybe, sometimes guesses incorrectly.

In their book, Sleights of Mind, neuroscientists and amateur magicians Susana Martinez-Conde and Stephen L. Macknik explain, “your brain is constantly making up its own reality whether it receives actual reality-driven input from your senses or not. In the absence of sensory input, your brain's own world making machinations keep on truckin' nevertheless."

Beyond a doubt, we “see” with more than just our eyes. And what we see is affected by, among other things, how much we’ve slept. So that pirouetting polar bear that appears at your bench in the middle of an all-nighter in the lab? That’s just your brain telling you to get more sleep.