Oct 15 2015

The Nature of Hallucinations

A new study tests a hypothesis about the nature of hallucinations, which is a sensory perception not connected to reality (an illusion, by contrast, is a misperception of something that is there). The question the researchers were addressing is whether hallucinations are best understood as a shift in the balance of a normal brain function.

This is an interesting general question in trying to understand brain function and disorders. Many brain disorders are understood not as something in the brain being “broken” but simply out of its optimal parameters. Anxiety is a great example. The phenomenon of anxiety is a healthy and necessary brain function. If you completely lacked anxiety you would lack the motivation to look both ways before crossing the street, or even to have a job or take care of yourself. Too much anxiety, however, can become uncomfortable or even crippling.

The question is, does this same logic apply to hallucinations? Researchers from Cambridge and Cardiff Universities in the UK think the answer is yes.

The basic brain function they think is altered in hallucinations is the relationship between what they call top-down vs bottom-up perception. Bottom up perception involves sensory information being processed by primary sensory systems in the brain, then being passed onto higher and higher systems for analysis, construction, and assigning of meaning.

However, sensory information is inadequate to uniquely identify all things in our environment and all of their properties (size, distance, movement, shadows, etc.) and so the perception system also works top down, starting with our prior knowledge of the universe. This prior knowledge influences how the lower level processing constructs the inadequate or even ambiguous bottom-up stimuli.

This is all a fancy way of saying that expectation affects what we think we see. But it is important to note, this is not just how we consciously interpret what we think we see, but how our brain actually constructs the image we perceive (or the sounds we hear).

This top-down system is necessary, otherwise we could not make sense of the ambiguous sensory input our brain receives. For example, everything in our visual field outside of a small area in the middle is really just a blur. Our brains average images over time to maintain continuity (a process call the continuity field), even when images are morphing due to changes in lighting, angle of viewing, and distance. Our brains also project a bit into the future to compensate for the processing time delay (otherwise we could never catch an object thrown to us).

The current researchers are addressing yet another function of our sensory perception system – the balance between bottom-up processing based on sensory information, and top-down processing based on prior knowledge in constructing our perceptions. Their hypothesis is that hallucinations may arise from this balance shifting toward prior knowledge.

To test this they did two experiments, one with subjects with early symptoms of psychosis, but without a formal diagnosis and untreated with anti-psychotic medications. The second study involved psychosis prone healthy individuals. They compared them to healthy and not psychosis-prone controls on a task that involves constructing an image from ambiguous stimuli.

They showed subjects a black and white image that lacked enough information to construct into a full image, but was based on an image. They also showed them a black and white image that was just noise. They exposed the subjects to priming images, containing color versions of the black and white images. The subjects then had to identify the black and white image using the prior knowledge of the color images.

As predicted, subjects with early psychosis or psychotic tendencies performed better on these tasks – they were better at using prior knowledge to construct the ambiguous images into a coherent pattern.

As always, such studies have to be interpreted with caution. Brain function and behavior are complex, and it is notoriously difficult to truly isolate one factor. We have to slowly build our neuropsychological models from many experiments, testing ideas from multiple angles. No one study can establish any specific model.

A competing, or perhaps complementary, hypothesis is that psychosis (hallucinations and delusions) represents a failure of reality testing. Our brains over-generate pattern recognition, and then filter out the patterns that don’t make sense (do not stand up to reality-testing). One hypothesis is that psychosis may result from either hyperactive pattern recognition, and/or underactive or impaired reality testing. The current study supports a role for hyperactive pattern recognition, but does not rule out an important role for impaired reality testing.

There also may yet be other important circuits in the brain influencing this balance that we have not identified.

Despite this complexity, the authors did make specific predictions based upon one hypothesis as to the nature of hallucinations (at least in some individuals) and their predictions were confirmed by their experiments. Their hypothesis is also consistent with other research.

We are, in fact, starting to form a fairly coherent scientific understanding of how the brain constructs our perception of reality, and the ways in which this process can go wrong.


The basic notion that our brains construct our perceptions remains a dominant and important part of our understanding of brain function. The current study supports this idea – that we use top-down mechanisms in which our prior knowledge influences how lower-level brain regions construct sensory information into a coherent narrative of reality.

This, I think, is the important take-away I would like everyone to understand and remember – our brains construct our perceptions of reality. This is a humbling notion. Once you truly grasp this you can no longer honestly use phrases such as, “I clearly remember,” or, “I know what I saw.” You have a fluid reconstructed memory of what was essentially a functional hallucination.

14 responses so far