Contraction bias: How previous experiences influence our perception
By Thomas Shallcross
Dr. Athena Akrami, of the Sainsbury Wellcome Centre, gave a seminar at the Centre for Developmental Neurobiology as part of the 2019-2020 “NEUReka!” seminar series. Dr. Akrami’s research focuses on understanding how incoming sensory information is integrated with our prior experiences, and how this process informs our choices and behaviours. During the seminar Dr. Akrami presented an elegant study, which she conducted during her post-doctoral research in the lab of Professor Carlos Brody (Akrami et al., 2018).
Our previous experiences and knowledge of the world shape not only our behaviour, but also how we perceive our environment. One example of how prior information affects our perception is a phenomenon known as contraction bias, which was discovered by a psychologist 110 years ago (Hollingsworth, 1910). Contraction bias describes the process by which our memory of the magnitude of a stimulus is biased towards the average of past observations. For example, if a person is shown multiple different balls, which vary in size from small to large, they will form a perception of an “average sized ball” in their mind. Then, when asked to judge the size a particular ball, they will underestimate the size of a larger than average ball and overestimate the size of a smaller than average ball. In real life, contraction bias may help to improve our estimation of objects in environments where the incoming information is noisy or incomplete. For instance, if a ball was only seen for a very short time, or was shown a long time ago, it may prove useful to rely on our previous knowledge of ball size to give a best guess estimate of the size.
Despite contraction bias being a known phenomenon, the exact brain mechanisms which give rise to it are not well understood. To answer this question Dr. Akrami developed an experiment, using rats, to study contraction bias of an auditory stimulus. In the experiment, rats were trained to distinguish between two tones which were played sequentially and differed in how loud they were. The rats were trained to report whether they perceived the first or second stimulus as the louder one; if they perceived the first stimulus as louder they would place their nose in a small hole on the left, and if they perceived the second stimulus as louder, they would place their nose in a small hole on the right. If they got the task correct the rat would get a reward. To make the task more difficult there was a delay of several seconds between the first and second tone, meaning that the rat had to hold in its memory the magnitude (loudness) of the first tone, and then compare it to the second. The rats underwent many such trials, one after the other, and Dr. Akrami demonstrated that the estimation of the first tone, which was being held in the rat’s memory, was biased towards the average volume of previous trials. This result is typical for what we observe in studies of contraction bias in humans. Furthermore, it was shown that the previous two trials affected the rat’s performance the most, whilst longer term history had a smaller but appreciable effect.
Whilst similar results had been seen previously in humans, the use of rats allowed Dr. Akrami to ask: where in the brain does contraction bias originate? A region of the brain, known as the posterior parietal cortex (PPC), receives input from the visual, auditory and somatosensory regions and is thought to be important for integrating these inputs with short term memory. To test whether the PPC is important in mediating contraction bias, Dr. Akrami inhibited neurons in this brain area using optogenetics, which allows an experimenter to artificially switch neurons on or off using light. Surprisingly, when the PPC neurons were switched off during the experiment, the ability of the rats to distinguish which stimulus was louder improved. Dr. Akrami showed that this was because the contraction bias effect was no longer taking place; when the rats were estimating which stimulus was louder, they were no longer influenced by the previous trials.
Although it may seem counter-intuitive that inhibiting a brain region can improve performance on a task, this can be explained by the design of the experiment. In real life there are many patterns and repetitions in our surroundings which we experience over and over again; very few things are truly random. This means that contraction bias, in general, can be beneficial, as using our recent memory of events to help us make a prediction will probably give us a good estimation of the current situation. In an experiment, however, these patterns and repetitions have been removed. This means that the previous trials don’t help the animal predict the current trial; if the first stimulus was louder in the previous trial, it doesn’t make it more likely that the first stimulus will also be louder in the current trial. Thus, ignoring the previous trials would, in general, be beneficial and this is exactly what is happening when the PPC is inhibited during the task.
Furthermore, Dr. Akrami also recorded the activity of neurons in the PPC during the experiment. A subset of neurons that carried information about which stimulus was louder were identified, and the activity of these neurons was shown to persist long enough to affect the next trial. Overall, these experiments indicate a neural mechanism underlying contraction bias in the PPC, contributing to the hypothesis that this brain region is important for relating prior experience with incoming sensory information.
Akrami, A., Kopec, C.D., Diamond, M.E., Brody, C.D. (2018) Posterior pariental cortex represents sensory history and mediates its effects on behaviour. Nature
Holllingworth, H.L. The central tendency of judgement. (1910) The Journal of Philosophy, Psychology and Scientific Methods
Find out about Scientifica's latest product releases, company news, and developments through a range of news articles, customer interviews and product demonstration videos.