The brain is a loom of mind-boggling complexity, its 100 billion neurons the struts and frets for the cascading electrical signals which ping around our minds. The threads we select to use on this loom, woven into the fabric we call our lives, are our experiences – the books we read, the music we hear, the nature we enjoy, the places we go – but above all the people who love us and whose love we return.
I have been fascinated for many years by the task of becoming bilingual in the language of the brain as well as the language of the mind. I have tried to learn about the brain as a physiological entity of neurotransmitters, myelinated axons and ion channel functions and to know the mind as a lived realm, a kingdom of the symbolic, the conscious, the imaginative and the creative.
I have learnt how frontal lobe damage from a stroke takes away a person’s capacity to speak, know their own limbs for their own, gamble away their savings, fail to plan a dinner party or find their way around a strange city. I have discovered how damage to specific loci in the brain creates problems in interpersonal relating or how hyperactivity in dopaminergic circuits results in someone thinking they are Jesus.
Science tells us we have a brain. Consciousness tells us we have a mind. We know they are one and same thing, but what you see depends on what you use to study it. The mind/brain debate often reminds me of the Indian story about the blindfolded sages who are led to an elephant. Each sage feels a different part of the elephant and then they meet and discuss their discoveries. They argue, passionately, decisively, dramatically late into the night, none of them realizing their subject of enquiry is one and the same. The mind/brain continuum is a warren of staggering size and depth and theoretical rabbit holes are there to fall down at every turn. Exploring them has humbled me in how much more there is to know.
I started with psychoanalytic theories of repression and transference (training as a psychoanalytic psychotherapist), then followed attachment-based rabbit holes such as the puzzles of interoception and the insula, epistemic trust and oxytocin, default mode networks and basic affective systems (through my encounters with neuropsychoanalysis). I have marveled at the emerging domains of psycho-neuro-immunology which seeks to understand how the mind and body are jointly acted upon by the molecules of emotion and I have discovered those domains so necessary to marketeers for manipulating consumers such as priming and cognitive bias. What all this falling down rabbit holes has taught me is quite how much of our ‘rational’ behaviour is in fact anything but, since the vast majority of our mental activity is governed by unconscious processes.
A particularly fruitful rabbit hole has been my recent discovery of the Free Energy Principle and it is this principle I wish to make the focus of this blog post because it throws light on the guiding principle of the entire warren’s architecture. A wonderful book by attachment-based psychotherapist, Jeremy Holmes, called The Brain has a Mind of its Own (2019) has helped me understand it better. In his insightful eight chapters, appended with an excellent glossary, Holmes first defines Free Energy.
Free Energy is the force of entropy where unbound ‘Free Energy’ spells disorder and chaos for a biological agent, be it a single-celled bacteria or a human being. Thus all living things must work to ‘bind’ this energy and minimize the entropy of their sensory states by maintaining homeostasis. In simple terms, when I find myself shivering, I put on a jumper. Holmes is drawing on mathematician Karl Friston’s work using Bayesian mathematics (the precise workings of which I cannot claim to entirely understand) that human brains are engaged in using the Free Energy Principle at every level of our functioning to keep entropy at bay.
The reach of the Free Energy Principle is exciting and the Maths matters because it is an ambitious way to model the core principle of brain activity, namely that the 3lbs we carry around inside our skull functions like a mighty prediction machine. Learning matters because it allows us to make generative models about the world; eg. ‘When I hear the sound of an engine and see something large moving down the street, it is probably a vehicle of some sort.’ But if I want to change my prediction, which has been generated unconsciously and therefore exists in my brain outside conscious awareness, then I must bring my attention to the ‘thing’ that I am making a prediction about. To return to my example, I might hear the engine, look at the vehicle then think to myself, ‘Oh no! It isn’t a car, it is a low-flying U.F.O. and it is heading straight for me!’ The great insight I get from Friston is that his use of mathematics explains how coping with a failed prediction is a parsimonious method of explaining why we are conscious.
What the maths explores is essentially this: when my prediction system doesn’t work, my brain must work to bind the Free Energy released. Consciously, I will experience ‘surprise’. My surprise is the signal that tells me a gap has emerged between my generative model and reality. My reaction to this will always involve action of some sort. A little terminology follows: the work of ‘binding’ Free Energy is called Prediction Error Minimisation (PEM) and the brain energy expended in finding the required action to do that is Active Inference. Imagine for a second you are sat in front of a magician. He pulls a rabbit out of a hat. How did he do that? You experience a Free Energy surge because you do not have a generative model for rabbits appearing from hats – your prediction machine is ‘surprised’. You use your sight to verify it really is a rabbit and you insist on examining the hat to see where the rabbit might have been hiding (Active Inference). You might even beg the magician to show you the trick because you want to minimize your prediction error (PEM) by discovering a secret compartment where rabbits are born! But the magician maddeningly replies, ‘It’s magic’. You are fascinated and feel a pleasurable frisson one could call amazement or wonder.
But let too much Free Energy loose and our nervous system is overwhelmed. Trauma is an example of when this happens. God forbid that gunmen in balaclavas burst into my house and shoot my family dead. For no pre-existing generative model in my brain would be able to encapsulate what had happened and the gaping hole between actual reality and how I assumed the world to be until that moment would be rent wide open. Essentially my PEM would fail and I would be in a dysregulated emotional state I couldn’t escape from until I made new generative models that could encompass such a horrific trauma. Boredom, on the other hand, would be an example of too little Free Energy in my system, as would frustration and depressive feelings.
The point is that all our mental effort nearly all of the time is expended on keeping ourselves within the comfortable bounds of homeostasis. Our environment is dynamic so the attempt to remain in the Goldilocks Zone (not too much boredom, nothing traumatic) is eternal; ‘I am hungry, so I eat; I am bored, I do something; I am tired, so I sleep; I am lonely, I call a friend. Then I am hungry again’ and so on until we die at which point we give up the effort to keep Free Energy at bay and let microbes break us down into our constituent atoms. Holmes explains how the Free Energy Principle operates at every level of brain activity (in mathematical terms) and how as a consequence, it can help us understand how the ‘bottom-up’ processes of automatized unconscious brain activity interact with ‘top-down’ consciousness in the endless, ongoing iteration otherwise known as being alive.
How is this relevant to you, I hear you ask? Well, an example of this iterative process is made manifest in the therapeutic relationship. When patients come to me with emotional problems which manifest as patterns interfering with their capacity to make and maintain relationships, essentially they are presenting with prediction problems. This is usually because their predictions about the Other have been generated in their early pre-conscious experience and then automatized (ie. repeated ad nauseam until they become habitual). This makes the predictions and the behaviour they generate axiomatic, unconscious and implicit. The work of therapy is to bring these predictions to mind (i.e. make them conscious) and then use the therapeutic alliance to make and test novel predictions about the Other and, over time, automatize more helpful predictions for making fulfilling relationships.
We could say in other words that the therapeutic intervention offers a novel and moderately stressful situation (like the Strange Situation in Attachment Studies) and this causes the release of Free Energy. When the patient attempts their habitual Prediction Error Minimisation, the therapist reflects back to them what they are doing and helps them understand why. I might say, for instance, to a patient: ‘It is interesting that you told me about something terrible happening to you this weekend in the last five minutes of the session so that we don’t have time to explore it; perhaps that is you re-experiencing a ‘fact’ that I won’t be able to help you, just as your mother couldn’t when she was depressed when you were little’. By re-framing the situation, I push the patient to use different methods of Active Inference – literally to ‘see’ differently. With my help, the patient can be nudged into re-framing their habitual perspectives much like saying a word out loud can sometimes release other associations to a clue in a crossword.
The Free Energy Principle is so interesting because it is the most elegant way of explaining why so much of our brain function is unconscious (most neuroscientists agree about 95% of brain activity in adults is beyond conscious awareness). Essentially, it is only when Prediction Error Minimisation is required that we ‘become’ conscious of a situation because it doesn’t match our prediction (our generative model has proved lacking). It also makes sense of recent neuroscience findings that everything we think we are experiencing we have actually already predicted, then used our sensory apparatus to confirm; hence we take from our optic nerve or auditory nerve only the gist of a scene and you might be surprised to learn were you sat in an audience focused on a task to count down from 100 to 0 in 7s, you would not notice a man in a gorilla suit walking across the stage! If you don’t believe me, check out www.theinvisiblegorilla.com. This is an example of selective attention, handsome proof of our capacity only to be conscious of what we expect to see. I find the Free Energy Principle, underpinned by its powerful mathematics, a brilliant explanation of why!
Rather than seeing conscious and unconscious processes as distinct in the brain, the Free Energy Principle unifies the two, explaining how it is that processes which don’t contradict our generative models stay unconscious while events that do, come into consciousness because they have loosed upon our mind Free Energy which we cannot ‘bind’. The psychoanalytic discourse around the notion of ‘containment’ is an interesting parallel to the word Friston uses for the work of coping with Free Energy – that of ‘binding’.
In Chapter 4 of his book, Holmes explores how the Free Energy Principle can make sense of psychopathology. Put simply, he identifies early childhood adversity in terms of ‘chronic unminimised prediction error’ (p.73). Think for example of how a child would struggle to ‘bind’ the Free Energy released by one parent beating up the other in front of them, or the terror unleashed by a parent losing their temper and hitting the child. As Holmes explores, during therapeutic encounters, the out-of-mind abuse/neglect of the past is jointly re-enacted with the therapist and therefore available for shared Prediction Error Minimisation. “Mutually mentalised, the unbound energy associated with loss or violence becomes containable” (p.73).
This explanation is a powerful way to understand how therapy works and offers a welcome ‘scientific’ justification for the method. It also helps explain mental health problems in other domains as Prediction Error Minimisation problems.
In autism, for instance, excessive precision is attributed to sensory input at the expense of generative models of the world. The autistic person cannot screen out their sensory input in order to let a generative model operate. Mark Haddon describes this ever so evocatively in his novel, The Curious Incident of the Dog in the Night-time when his protagonist, Christopher, travels to Paddington Station. On arrival, he is assaulted by the plethora of visual stimuli from the rotating information on the live departure boards, to shop signs, adverts, station signage etc. Poor Christopher is compelled to read every single word of it (his PEM puts too much emphasis on bottom-up sensory input) rather than select the appropriate visual stimulus that would tell him which platform his train leaves from. Were he able to do this, his PEM would be using a top-down generative model AND a visual stimulus rather than just the bottom-up sensory stimulation.
Meanwhile, in psychosis, the reverse situation is operating. When a psychotic person believes he is Jesus and can perform miracles, the sensory input that in fact he cannot raise the dead or feed the 5,000 is disregarded (bottom-up sensory input is ignored) and the psychotic person’s PEM over-prioritises a top-down generative model (“No I can’t raise the dead but I know I am Jesus because I know so”). This underscores the psychiatric maxim that it is counterproductive to try to argue people out of their delusions. Some interesting fMRI research by Ermakova et al (2018 – Journal of Neuropharmacology 43: 1691-1699) has shown signalling errors between the midbrain and the right dorsolateral prefrontal cortex in psychotic patients revealing how bottom-up (midbrain) conversations with top-down (prefrontal cortex) cognitions are impaired, especially in the dopaminergic pathways which we know are implicated in psychosis.
I have a patient, L., who has a generative model that the Other will not be able to see who she is. It may well be that the area of the brain involved in this kind of almost-conscious rumination is the Default Mode Network, the store of generalized self-other representations, active when you are not apparently thinking about anything in particular. These areas of the brain are involved in how our limbic-mediated ‘emotions’ become ‘feelings’, i.e. where our mammalian brain butts up against our prefrontal cortex where language-mediated, social interactions with other humans is underpinned by brain activity.
Early experience taught L. that simply ‘being’ was not an effective strategy to keep the Other close to her, so I believe she automatized early on in her infancy how to carefully observe the Other, then be what was understood to be wanted (Echo to their Narcissus, you might say). The problem is that this operation has nothing whatsoever to do with her feelings or idiosyncrasies. Her experience/identity are frankly irrelevant to proceedings. The consequence is that she lives with barely any idea of what she likes and therefore who she actually is. Since she basically isn’t ‘there’, any experience she does have is wiped off like chalk from a blackboard (we call it ‘Teflon’). This is because her generative model for being alive says her only task is to ‘bind’ to the person she happens to be with and therefore what has just happened to her, or might happen when this person isn’t with her, is not deemed relevant. The result of this, as you can imagine, is a very miserable existence in which L.’s dominant affect is ‘I am irrelevant’.
In L.’s system, Free Energy is released when the Other responds to her particularity (be the Other another person or a book/film etc.). This is exemplified by the fact that if she manages to pick up a book and start reading she will find herself putting it down the moment she has consciously noticed that she’s enjoying it. Her PEM requires her to cut off from the thing she is enjoying to reduce the release of Free Energy. If, on a rare occasion, an interaction with a friend resonates with her as a real instance of connectedness, she notices that her subsequent contact with that person drops off dramatically.
The mechanism which keeps the whole charade in place is so automatized, so perfect in its operation, it has taken us nearly a decade meeting three times a week to see and hear it whirring away. Who knows how much longer it will take to render obsolete.
However, the point is that now we have ‘it’ (this generative model which does its very best to erases her) in our sights and have evolved a way to talk about how her system is entirely set up to have her at the back of the queue – and why she sometimes ends up contrasting herself with what she calls, ‘Real people’.
Optimistically, over the last few years, she has had flashes of knowing who and what she feels and very, very rarely what she might want to do, but always and only when she is on her own. Much more recently she has begun, very tentatively, to be able to feel ‘real’ in my company too. One day, hopefully, she will take these revised predictions out into the world and recast her relationships so that she can actually thrive and remain herself as opposed to temporarily survive, perfectly bonded to the Other, while in their company.
Thanks for reading all the way to the end! Please do also enjoy this brilliant 15-minute TedX talk by Dr. Jared Horvath from the University of Melbourne to experience his brain puzzles which reveal just how powerful a prediction machine your brilliant brain really is.
I want to thank my helpful reader, Sam Levine, for her judicious editing of this piece.