top of page

Taking a second dive into perception

Thank you everyone for another warm and challenging discussion yesterday!

by Laura Rathbone

Tackling perception is not an easy topic for us to explore and every time we meet we each learn and grow that little bit more.

We were also really happy and thankful to have Prof Mick Thacker pop in and join us, answering some of our questions and kindly supporting us along the way!

This session we explored our thoughts and questions which arose from two papers:

These two papers both explore the Free Energy Principle as applied to human experience during ill health and why we might have the experiences we have.

The Free Energy Principle

The Free Energy Principle (FEP) is a principle of life and biology set forward by Prof Karl Friston and deals with the integration of action and perception.

The theory helps us to make sense of how organisms (from the single cell amoeba to complex organisms like mammals and of course humans) navigate and flourish within their environment.

There are several tenets:

  • Biological systems are self-organising (don't need an external source to 'organise them' and some kind of order occurs through local interactions of cells)

  • Must maintain their state by flexibly holding their internal processes within stable bounds in the face of a constantly changing environment - homeostasis.

  • becomes increasingly more skilful (or precise) at managing this internal state by predicting its environment and the state of its bodily integrity

  • Predictions are sent out to the sensory apparatus for comparison against the actual incoming sensory signals which is the error, or gap, between the prediction of the sensory signal and the actual sensory signal.

  • This amount of error is called Free Energy and presents the organism with a problem - how to reduce the free energy?

  • Organisms dynamically couple two energy minimising processes: perceptual inference and active inference.

  • Perceptual inference is the updating of the prediction to better match the external information

  • Active inference is the use of action to change the relationship of the organism with the environment to better meet the prediction

How do we differentiate between the internal and the external state?

This is a great question and led to a discussion about Markov Blankets which is explored in the Kiverstein et al paper.

A markov blanket is a model for understanding how cells, groups of cells and systems of groups of cells go about self-organising and minimising free energy.

It is said that biological organisms violate the 2nd rule of thermodynamics by resisting natural tendency towards disorder.

Markov blankets are a mathematical and metaphorical way of explaining this.

A single cell is both it's own markov blanket and part of a wider markov blanket. It's cell wall creates the boundary between it's own single cell internal state and the external environment.

This theory can be applied to neurons (tip from Mick when reading the paper - when you get to the part looking at markov blankets, it might help to temporarily replace the word 'node' with 'neuron' for your own sense-making). One neurone forms its own blanket, with an internal state based upon its predictions of the external environment - this produces a tonic activity which is the nerve's regular sampling of the environment (error) and the prediction being met (error minimisation). In this situation, the prediction is precise and matching the error - no need for further energy. However, if the external environment alters, in the synaptic cleft this could be a neurotransmitter, this could register as error on the post-synaptic site and require the cell to undergo updating to minimise that error.

So the markov blanket helps us to show that what is on the inside behaves in a certain way based on what the predictions are and will act in a certain way to minimise free energy (or error) produced when there is inaccuracy of those predictions.

Mick helped us to understand that within groups of cells, we will see a forming and breaking down of these connections as part of the usual behaviour of Markov blankets. cells will come together to create a blanket, developing ways of sharing and co-creating the environment within the boundary, and they will break up to form blankets with other cells. the blanket is formed when 2 or more cells self-organise to create a moment of shared order.

We can think of cells, groups of cells like organs perhaps or systems, and even wider like two or more people, groups of peoples and communities of people.

Is perceptual inference separate to active inference?

Perceptual inference is dynamically happening all the time within optimal functioning cells/organisms. we are constants acting on the environment to meet sensory predictions or change the sensory samples and also, undergoing alterations in the biochemical structuring of our predictions in order to better predict our sensory experience of the environment - Mick called this neuroplasticity and i think we all had an 'ah-ha' moment!

The question arose then in relation to pain and the Fotopoulou paper exploring anosognosia in hemiplegia, in ill health, is something happening within the dynamic coupling (action) of these two inferences that results in on-going suffering?

The Fotopoulou paper explores anosognosia (a neurological condition where the patient is unaware of their disability) through this framework and discusses that there may be a lack of updating based on error, meaning that the prediction of an intact body remains highly precise, even in the face of incoming error. It may be that that person is unable to undergo perceptual and active inference.

We also explored whether effective updating can ever fully be achieved if the person is unable to bring about action to sample their environment, for example if they have a plegia, amputation or are unable to move for other reasons?

Without error, the prediction can become highly precise, but what if the prediction becomes pain? As movement and environment exploration reduces, so too does the person's exposure to error, rendering the prediction more and more precise.

This then left the question open for us to think clinically...

Might the role of medicine and therapies be to explore how best to facilitate reducing precision, exposure to error and then ultimately, model updating?

This left us with a lot of scope for reflection and quite an exciting place to explore our practice!

If you are a subscriber to the platform, the recording will be sent out just after this mail and you will have 2 weeks from the date of recording to watch it back.


Friston, K., 2010. The free-energy principle: a unified brain theory?. Nature reviews neuroscience, 11(2), pp.127-138.

7 views0 comments

Recent Posts

See All


bottom of page