Hallucinations: It’s not what you hear but expect to hear

Hallucinations are a cue word for schizophrenia and severe mental illness. Yet, in themselves, hallucinations are neither unique to schizophrenia, nor are they uniquely related to ill people. Instead, they are common in other diseases, such as Parkinson’s, and even occur in some people without any mental illnesses. Have you ever heard your name in a place where you knew nobody could have said it? Have you ever had a creepy-crawly feeling without anything bugging you? Or have you ever thought you’ve seen someone you know at first glance, yet realised later it was someone else? Chances are high that you’ve had one of these experiences because hallucinations lie on a spectrum. On one end of the spectrum, people suffering from severe diseases like schizophrenia might hear vicious voices they can’t turn off, that are terrifying and aggressive, while, on the other end, there may be such common everyday misperceptions. In fact, if you regarded my very minor examples as hallucinations, we all might be susceptible to hallucinating to an extent and a recent study published in the renowned journal Science does indeed suggest that this is the case. The study led by researchers Powers, Mathys, and Corlett has looked closely at how hallucinations come to life and suggests they may be failures in how we integrate our previous knowledge with the actual perceptual input.

Perception as a two-way street

To understand hallucinations, it is important to understand one fundamental aspect of our perception: When we see, hear, smell, or feel something, this does not only result from the input we get from our environment but also from what we have learnt before and what is important to us. Take for example the so-called cocktail party effect: When we’re at a noisy party, we’re very much able to hear and notice our name being said somewhere in the room while other names might not draw our attention. This is because our name is relevant to us and our brain has neural pathways that are fine-tuned for its perception. So while the volume of our name among other words might be enough to stick out in the crowd, our brain wiring amplifies it because of its importance. This dependence of our perception on previous experiences is not specific to our names but a more general phenomenon, where personally important knowledge can influence perception. So rather than perception being a one-way street from the outside world into our consciousness, it can be viewed as a two-way street, in which our sensory input combines with our previous experiences to form our perspective on the world.

The researchers in the Science article hypothesised that this two-way combination of previous experiences with sensory input might go wrong in people who hallucinate. Specifically, they estimated that hallucinators’ brains overly rely on previous experiences and therefore “expect” to see, hear, smell, or feel something that isn’t there, thereby eventually resulting in a genuinely altered perception. To test this hypothesis, the study authors attempted to induce hallucinations in an experiment that exploited one of the most famous psychological techniques of all time, a technique called classical conditioning.

Classical conditioning or “Psychology 101”

Classical conditioning was initially described by Ivan Pavlov, whose name will be haunting introductory psychology books for the unforeseeable future. Pavlov conducted a by-now infamous experiment with dogs whose salivation he measured. First, he would give food to the dogs, which, unsurprisingly, resulted in the dogs starting to salivate in anticipation of their dinner. Second, he rang a bell, which, equally unsurprisingly, did not result in increased salivation since sounds generally don’t make for a good dinner. In a third and crucial step, Pavlov then paired the bell with giving food, so always ringing the bell just prior to feeding his dogs. When he measured the dogs’ salivation after doing this several times, he noticed that the dogs would start salivating even when he just rang the bell but withheld the food. This means Pavlov’s dogs had learnt to associate the bell with food, just as we might associate the ping-sound of our microwaves with a warm dinner.

Although this finding of learned associations might seem very simple and dull, it has revolutionised psychology. Not only can classical conditioning offer some helpful explanations of how symptoms of mental disorders arise but clinical psychologists also use classical conditioning techniques in their treatment. Take insomnia as an example: During exam period, you might be too stressed out to sleep in bed, worrying about the exams to come. When such worrying prevents you from falling asleep over several days, lying in bed (corresponding to Pavlov’s bell) might now become falsely associated with staying awake (corresponding to Pavlov’s dogs salivating) because it had been repeatedly paired with worrying (corresponding to Pavlov giving food to his dogs). Even when exam time ends, these problems with worry-related wakefulness might persist, thereby resulting in insomnia. A therapist might advise you to pay attention to so-called sleep hygiene, only lying down in bed when you’re super-tired and sure to faint once lying down or, if you cannot fall asleep, to get up again. Just as lying in bed had been associated with being awake beforehand, this would then reverse the conditioning to “train” you to sleep well again.

Learning to hallucinate

Coming back to hallucinations after this detour, Powers’, Mathys’, and Corlett’s experiment was set up quite similarly to Pavlov’s initial study with dogs; the differences being (1) that the dogs were human participants, (2) the food was a high-frequency sound, (3) and the bell was a visual stimulus on a computer screen. On the computer, participants were instructed to indicate whenever they had heard the sound, which the researchers made very hard to hear. Crucially, the visual stimulus was frequently paired with the tone, so that participants learned that whenever there was a visual stimulus, the tone was more likely to be there as well. After pairing the visual stimulus with the sound many times, the researchers would then “withhold” the sound and sometimes only show the visual stimulus. If the association between visual stimulus and sound was strong enough, this “previous experience” might be strong enough to trigger the sound in participants’ heads as well. Strangely enough, this is exactly what happened and participants would oftentimes still confidently report to having heard the sound, so they had effectively been trained to hallucinate!

The researchers went further by looking at participants brains when they heard sounds that were present, when they hallucinated the sounds, and when they rightly identified no sound during presentation of the visual stimulus. Comparing the areas that were active between real sounds, hallucinated sounds, and no sounds showed that brain activity was indeed very similar during hallucinations and actual sounds, both showing activity in the auditory cortex. Yet, when only the visual stimulus was present and no sound perceived, brain activity was different. This demonstrates that the mere expectation of a sound was sometimes strong enough to create the actual sound in our heads!

Although all participants reported these trained hallucinations, the study participants had been specifically selected to belong to either of four groups that can be subdivided into those seeking and not seeking treatment and those normally hearing and not hearing voices:

  Hearing voices
  Yes No
Treatment-seeking Yes 1.     psychotic illness & hearing voices 2.     psychotic illness & not hearing voices
No 3.     no diagnoses but hearing voices 4.     no diagnoses & not hearing voices

When comparing trained hallucinations between these groups, the researchers noticed that people from groups with regular voice-hearers also reported more of the trained hallucinations. This has two important implications: First, some people may be more prone to hallucinations. They rely more than others on their previous experiences rather than on their sensory input and this makes them more likely to experience hallucinations. Second, hallucinations are not synonymous with mental illness. As this study shows, all people have the potential to hallucinate and it’s not only those seeking treatment and diagnosed with an illness who do so. Depending on how hallucinations are defined, previous studies have suggested that about 4-25% of all people experience hallucinations and most of them don’t actually suffer from a mental disorder. So rather than the mere presence of hallucinations, clinicians need to focus on what these hallucinations mean to any single person. This involves both looking at the type of hallucinations as well as how people deal with them because only then the clinical relevance can become clear. Regarding type of hallucinations, a persistent research finding has been that hallucinations with negative content are more likely to be impairing, which is to-be-expected because who would want a voice shouting “you’re awesome!” to disappear? Regarding dealing with hallucinations, the movie A Beautiful Mind provides a good example, which tells the story of nobel price-winning mathematician John Nash: Nash suffered from schizophrenia and while his hallucinations had dictated many of his actions during the most severe periods of his illness, he was able to simply acknowledge their presence in the end, thereby enabling him to live a full life despite his hallucinations. This should also be the take home message from this blog: While hallucinations can be sinister and disabling symptoms of severe mental disorders, more often than not they happen to healthy individuals and can, in fact, happen to all of us.



  1. de Leede-Smith, S., & Barkus, E. (2013). A comprehensive review of auditory verbal hallucinations: lifetime prevalence, correlates and mechanisms in healthy and clinical individuals. Frontiers in Human Neuroscience, 7, 367. doi:10.3389/fnhum.2013.00367
  2. Fénelon, G., Mahieux, F., Huon, R., & Ziégler, M. (2000). Hallucinations in Parkinson’s disease: Prevalence, phenomenology and risk factors. Brain, 123(4), 733-745. doi:10.1093/brain/123.4.733
  3. Johns, L. C., & van Os, J. (2001). The continuity of psychotic experiences in the general population. Clinical Psychology Review, 21(8), 1125-1141. doi:10.1016/S0272-7358(01)00103-9
  4. Powers, A. R., Mathys, C., & Corlett, P. R. (2017). Pavlovian conditioning–induced hallucinations result from overweighting of perceptual priors. Science, 357(6351), 596-600. doi:10.1126/science.aan3458
  5. Also check out another blog and cool video on this study: http://www.sciencemag.org/news/2017/08/how-your-mind-protects-you-against-hallucinations
The following two tabs change content below.

Nils Kappelmann

I'm a PhD student at the Max-Planck-Institute of Psychiatry in Munich, Germany, investigating potential biomarkers in the psychotherapeutic treatment of depression.