Practically-A-Book Review: Byrnes on Trance
...
Steven Byrnes is a physicist/AI researcher/amateur neuroscientist; needless to say, he blogs on Less Wrong. I finally got around to reading his 2024 series giving a predictive processing perspective on intuitive self-models. If that sounds boring, it shouldn’t: Byrnes charges head-on into some of the toughest subjects in psychology, including trance, amnesia, and multiple personalities. I found his perspective enlightening (no pun intended; meditation is another one of his topics) and thought I would share.
It all centers around this picture:
But first: some excruciatingly obvious philosophical preliminaries.
We don’t directly perceive the external world. Every philosopher has their own way of saying exactly what it is we do perceive, but the predictive processing interpretation is that we perceive our models of the world. To be very naive and hand-wavey, lower-level brain centers get sense-data, make a guess about what produced that sense data, then “show” “us” that guess. If the guess is wrong, too bad - we see the incorrect guess, not the reality.
So for example:
In the famous “checker shadow illusion”, Square A and Square B are the same color (see here if you disbelieve). Our lower-level brain centers guess that, given the shadowing, our sense-data about Square A must “actually” be produced by a “really” black square, and our sense-data about Square B must “actually” be produced by a “really” white square. Therefore, they “show” “us” a “picture” of Square A being black and Square B being white, even though these aren’t the real colors.
The “blind spot” is an even more famous example. The place where the optic nerve meets the eye lacks photoreceptors, leading to a 5-10 degree hole in the middle of the visual field - there’s a medium-sized spot in your vision where you can’t see anything at all. But our lower-level brain centers guess that probably there’s just, you know, normal stuff there. Therefore, they “show” “us” a “picture” of an intact world with normal stuff in the blind spot area - safe enough, unless there’s really an oncoming car.
Why did we go through these excruciatingly obvious philosophical preliminaries?
Sometimes, two or more models can explain the same data about equally well. For example:
This picture can be either a right-side-up staircase (with the blue side in front, and the green side as the back wall), or an upside-down staircase (with the green side in front, and the blue side as the back wall).
If you’re like most people, you probably don’t see it as ambiguous. You see one of these as immediately obviously and viscerally true - for me, it’s the right-side-up blue-in-front staircase. Then, if you stare at it enough and move your eyes in weird ways or whatever, it “flips”, so that it’s immediately obviously and viscerally an upside-down green-in-front staircase.
(if you can’t do this, try staring at the green part and imagining it gradually moving towards you, out of the computer, while thinking “this is in front, this is in front” really hard - I believe in you!)
This kind of picture is called a bi-stable image. You viscerally see your brain’s educated guesses as real. When your brain’s guess changes, your visceral perception changes too.
This illusion is usually described as “all the plates are right-side-up except one - when you find it, they will all turn upside-down”. I think that might be fake - they’re all right-side-up, but something about the process of looking for “the upside-down one” can make your brain flip from one model to the other and cause the plates to change sides. I find this is easiest to do looking at the square one in the top center, or the round one just below, but I don’t think that’s because they’re the “actually upside-down one”. If that doesn’t work, try viewing it from about ten feet away from your computer screen, but be careful - you might not be able to get them to flip back to right-side-up again!
The train is either going into the tunnel, or coming out of the tunnel. You can make it switch by quickly moving your eyes either left-to-right or vice versa, or by thinking very hard about the train going in or out.
This might be the toughest one to flip. If you start by seeing her spin clockwise, try focusing on her central foot to switch directions; if you start by seeing her spin counter-clockwise, try focusing on the reflection of her outstretched foot when it appears.
Which of these models - the clockwise dancer, or the counterclockwise dancer - is real? Trick question - neither is real. There is no dancer and no rotation; you’re actually viewing shifting pixels on a computer screen.
To belabor the excruciatingly obvious philosophical preliminaries: there is some sense in which our models of the world are very good. They usually correspond to reality exactly the way we think they do. The perception of world-models isn’t a reason for radical skepticism.
In another sense - not a very profound one - our models of the world are distorted. For example, they make us see rotation where there are really just shifting pixels. They’re also ambiguous enough to occasionally be bi-stable - sometimes, you can shift from one world-model to another, with an associated change in visceral perception.
From Models To Self-Models
Just as this is true for external senses like vision, Byrnes says this is true of our internal perceptions - perceptions of things like thoughts, desires, and conscious experience.
The “reality” of our inner experience is patterns of neurons firing in response to sensations or other neurons. This is a boring claim, like saying that the spinning dancer is “really” “just” “shifting” “pixels”, but let’s explore it a little more.
Sometimes, enough neurons representing similar concepts fire at the same time that they form some kind of temporarily stable pattern that takes over the global workspace.
Sometimes, a pattern like this knits together enough concepts to represent a world-state and give positive valence to that world-state.
Sometimes, those patterns reach a threshold where they cross over to the motor cortex and activate motor programs elsewhere in the body.
If this is the “shifting pixels” perspective, what’s the “looks like a dancer who is spinning around” perspective?
“Sometimes, enough neurons representing similar concepts fire at the same time that they form some kind of temporarily stable pattern that takes over the global workspace” → I thought about X
“Sometimes, a pattern like this knits together enough concepts to represent a world-state and give positive valence to that world-state.” → I want X
“Sometimes, those patterns reach a threshold where they cross over to the motor cortex and activate motor programs elsewhere in the body.“ → I decided to X
The “model” that people come up with to explain their inner life is the internal feeling of a separate “self” who reviews and signs off on the decisions of “the brain”. Referring to the philosophical tradition and pictures like this…
…Byrnes calls it the “homunculus” (Latin: “little man”).
The homunculus (“self”/”me”) is a useful tool for organizing internal experience. For example, if you have a seizure and your arm moves, you can say “I didn’t choose to move my arm - it just moved of its own accord!” (ie the homunculus isn’t doing the moving). If you have some kind of OCD or rumination disorder, you can say “I don’t want to keep having these thoughts about death, they just pop unbidden into my mind” (ie the homunculus isn’t doing the thinking). To actually analyze these situations would require a PhD in neuroscience, but we all understand the visceral experience of being stuck with thoughts that “we” don’t “want” and didn’t “cause”. Overall it’s very similar to the way I described natural intuitive “theory of mind” here:
The mind is an imaginary space containing things like thoughts, emotions, and desires. I have mine and you have yours. I can see what’s inside my mind, but not what’s inside your mind, and vice versa. I mostly choose the things that are in my mind at any given time: I will thoughts to happen, and they happen; I will myself to make a decision, and it gets made. This needs a resource called willpower; if I don’t have enough willpower, sometimes the things that happen in my mind aren’t the ones I want. When important things happen, sometimes my mind gets strong emotions; this is natural, but I need to use lots of willpower to make sure I don’t get overwhelmed by them and make bad decisions.
Byrnes makes this more concrete with a survey of homunculus beliefs across different cultures. We place the homunculus in the head, which happens to be correct (ie thoughts happen in the brain). But this is kind of a coincidence (or maybe downstream of knowing the real science); other cultures feel like the seat of consciousness is in the heart or the belly, and this feeling is about equally plausible and stable. Meditators say that with enough practice, they can imagine their consciousness being in their head, heart, belly, or outside their body entirely.
This is a subtle point (are you starting to see why we went through all the excruciatingly obvious philosophical preliminaries?) There is, in fact, a brain that has thoughts, located in your head. And your visceral experience includes a term for a self that has thoughts and is located in your head. But they’re not exactly the same thing. The trivial differences don’t matter in ordinary cases. But in edge cases, they can get pretty weird.
Trance And Spirit Possession
Okay, now the fun stuff.
Byrnes argues that “homunculus” vs. “trance” are two alternative bistable models for analyzing internal mental experience. The process of going into a trance (or being “possessed” by a spirit) is conceptually similar to the process of switching the dancer from clockwise to counterclockwise. The process goes:
Start with a strong background belief that the new model is plausible.
Relax.
Suppress all evidence favoring the old model.
Gather evidence favoring the new model.
First, start with a background belief that the new model is plausible. If you’re getting hypnotized, it helps to believe that hypnotism works. If you’re in a spirit possession ceremony, it helps to believe in spirits. Hypnotists and shamans should help this process along by being inherently believable - charismatic and confident, with lots of suggestive ritual that they perform correctly.
Second, relax. When you were trying to switch the direction of the dancer, you probably did this naturally - let your eyes get slightly out of focus, concentrated on the task in front of you.
Third, suppress all evidence favoring the old model. In the case of trance/possession, stop doing obvious voluntary actions. Watch a stage hypnotist show, and nobody is performing a running commentary: “Yeah, I’m focusing on the swinging pendulum . . . looks pretty normal . . . guess maybe I’m starting to feel sleepy . . . I wonder if this hypnotist is a fraud . . . “. They’re supposed to be quiet, immobile, and focus on the trance. Likewise, possession ceremonies often begin with hours of ritual dancing; by the end, it feels like your feet are moving of their own accord. Certainly you are not consciously choosing where to put your feet at each moment due to rational considerations.
Fourth, gather evidence favoring the new model.
In a typical stage hypnosis show, the hypnotist starts by making his subject watch a swinging pendulum (moving eyes back and forth tends to make people sleepy and exhaust their eye muscles). Then the hypnotist says “You are getting sleepy . . . your eyelids are getting heavy.” The subject is surprised! They are feeling unexpectedly sleepy! Their eyelids are getting unexpectedly heavy! It seems like the hypnotist is in control of their body!
Then the hypnotist asks them to do something simple, like hold their arm out. Is this part of the induction process? The first hypnotic suggestion? Hard to tell - in either case, the subject moves their arm out. Then the hypnotist might say something like “Raise your arm”. It’s a reasonable request - and also, when the arm is in a sufficiently unstable position, sometimes just considering movement will cause it to move a little, even without the mental motion that would usually be considered “volition”. Again, it seems like the hypnotist is in control and has creepy mind powers!
Then the hypnotist might ask the subject to do something simple, like jump. The hypnotist is a high-status person reputed to have creepy mind powers, giving a direct order. The audience is expecting the subject to jump. If the subject doesn’t jump, the show will be over and it will be awkward for everyone involved. So there are compelling reasons for the subject to jump, and no reason not to. The subject notices the amount of internal mental pressure that naturally corresponds to “there are compelling reasons to do this thing”. Why (they might unconsciously think to themselves) is there this mental pressure to jump? One answer is the story we just told - command from high-status person, not wanting to feel awkward, etc - but these are much more subtle and complicated than a simple alternative hypothesis - the hypnotist has creepy mind powers and is giving me a compulsion to jump. If all the previous steps have been completed correctly, there is a visceral flip in mental models, and the subject feels what was previously a working hypothesis as intuitive obvious ground truth - “I have lost control of my body and the hypnotist is puppeteering me”.
This flip itself reinforces the trance - my new evidence in favor of trance is not just background beliefs about possession, but the visceral feeling of losing control of my body, plus the undeniable fact that I just jumped when there was (seemingly) no reason to do so. The trance state is now a new attractor, not quite as strong as the old homunculus attractor (“I am in control of my mind” really does explain a lot, and has decades of experience/inertia behind it”), but strong enough to usually last for an hour-long hypnosis show without collapsing.
I couldn’t find that much about this in Byrnes, but the model flip itself must go back and affect ground-truth reality. “Hypnotist is compelling me” provides evidence for “I follow the hypnotist’s orders”, and therefore makes me slightly more likely to do so. It also frees me from some common reasons I wouldn’t follow the hypnotist’s orders - it’s too embarrassing, it would never work, it’s ‘not the kind of person I am’. Maybe the hypnotist orders me to do a silly dance on stage, and normally I’m too dignified, but since “the hypnotist is compelling me”, it doesn’t threaten my dignity and I can get away with it.
Byrnes also adds (we’ll see why later) a postulate that doesn’t really make sense to me on its own - something about the self-model assists in the formation of memory. Remembering “I did a silly dance on stage” is easier if there is an “I” concept active to “hang the memory” on. This could be related to the finding that people remember things better in the same context they learned it (eg you’ll do better on a test in the same classroom where you learned the material) or to the finding that emotions organize memory (eg when you’re angry at your spouse, it’s easy to remember all the times they’ve ever wronged you; when you’re happy with them, it’s easy to remember all the good times you’ve had together). “The self exists” is a pretty dramatic context cue, and maintaining memory between self-models is apparently a pretty tough task - hence the tendency for people to say they “don’t remember” what happened during a trance.
From Trance To Everything Else
Now we have the material we need to explain all sorts of weird mental phenomena:
Dissociative Amnesia
Someone with a desire that doesn’t make sense in the context of their normal personality eventually “flips” to an alternative personality that carries out the desire. When they recover, they have no memory of the incident.
Dissociative Identity (IE Multiple Personality)
If the homunculus-self is a mostly-accurate but not-directly-perceived-and-real model of mental processes, then a person whose mental processes often flip between two or more dramatically different states (for example, borderlines, who are notable for very strong emotional states and “splitting”) may gather evidence for and eventually flip to a model of themselves as multiple different homunculi. This is especially true if they’re primed with the suggestion that this is a likely way for the inside of their mind to be (for example, by a psychotherapist who believes in multiple personalities).
Ego Death (EG On Ketamine)
Remember, the ego (homunculus) is a model of mental processes. which says that thoughts arise in the brain because “I” “choose” to “have” thoughts, or actions happen because “I” “voluntarily” “make” the decisions.
From a god’s eye view, outside of the homunculus model, we might picture a decision as looking like:
A thought arises: “Maybe it would be a good idea to eat a taco”.
This thought spawns other thoughts: “Tacos are delicious”, “Tacos are expensive”, “I’m on a diet”.
All of these thoughts kind of battle it out until they turn into a unified analysis of the situation “Tacos are expensive, but I deserve a treat, so I’m going to have one”.
The basal ganglia and motor cortex implement an action program: I order a taco.
Within the homunculus model, this orderly progression of events is what we interpret as “I thought about it and decided to order a taco”.
On sufficiently weird drugs like ketamine, mental order breaks down. The relationship between one thought and the next is completely chaotic, or at least too complicated to model. The expectations of the homunculus-model, where thoughts naturally lead to consequences according to stable personality features, are profoundly violated. Since the homunculus model no longer credibly describes the data, the brain ditches it, and the drug user viscerally feels like they have “lost sense of self” or “experienced ego death”.
Buddhist Enlightenment
If you watch your own mental processes very very hard for a long time, you notice subtle ways that the homunculus model is incorrect. For example, if you carefully watch thoughts form, it’s obvious that “you” didn’t “decide” to “think” them; they just arose out of the void (or out of casual antecedents like sense-data or previous thoughts). If you watch mental decision very very carefully for a long time, you notice the same things the Libet experiment noticed, where they seem to often happen before “consciousness” is “aware” of the decision at all. These all provide evidence against the homunculus model. After enough evidence builds up, you suddenly flip from the homunculus model to some other model which is closer to the god’s-eye one mentioned above - there’s no self, thoughts arise on their own, you are part of the same nexus of causal processes which determine the rest of the world.
I like this because it explains something I’ve always found baffling - the claim that satori happens in a single instant (traditionally when you see a falling leaf, or your master hits you on the head with a stick, or something like that). Not many things in psychology happen instantaneously - but one of them is the flip in bistable perceptions!
Julian Jaynes
Jaynes was the psychologist and historian who gathered an exhaustive collection of sources suggesting that Bronze Age people didn’t experience consciousness the way we did - instead feeling like they were automata being commanded by the gods to do whatever they did.
Byrnes spends most of this section arguing against Jaynes (comparatively weak) claim that ancient people were incapable of deception and other basic theory-of-mind tasks, but seems mostly willing to grant the more sensational claim that they felt their actions more akin to a hypnotist’s compulsion than to self-motivated agency. None of this is especially surprising by the discussion of trance above - it’s just a whole civilization using the “spirit possession” model at scale.
See my previous review of Jaynes for more.
I’ll spare you the discussion of free will - which tends to make people really mad, and which is basically what you would predict given the background assumptions - but I recommend reading the entire series of essays, which goes into much more depth and belabors the excruciatingly obvious philosophical assumptions enough to make them really sink in on a deep level.
Byrnes also has a wide range of writing on other areas of neuroscience and on AI alignment.
Share this post