The joy of magic mushrooms Science is finally explaining how hallucinogenics can trick your mind

First, though, the study. It took 59 patients diagnosed with moderate-to-severe depression, and gave 30 of them two quite large doses of psilocybin three weeks apart, as well as a course of placebo capsules every day for six weeks; it gave the other 29 a normal daily dose of escitalopram, and a tiny dose – so small as to be ineffective – of psilocybin at the same three-week interval as the others. Then they measured the subjects’ mood, using four standard scales of measuring depression; two questionnaires filled out by the patients, and two by their doctors.

The results were intriguing. The researchers preregistered one of the scales as the “primary outcome” of the study, and the other three (among other things, including measures of well-being, suicidality, work and social functioning, and anxiety) as “secondaries”. That’s to avoid “p-hacking” or “hypothesising after results are known” – getting your data and chopping it up until you can make it say whatever you like.

On the primary outcome, psilocybin did not outperform escitalopram. Or, rather, it did, but not by enough to reach “statistical significance”, so, by convention, scientists say that the result could have been a fluke. Still, though: magic mushrooms did at least as well as the best antidepressant in treating depression, which is quite exciting.

And it did reach statistical significance on several of the secondary measures, and it also had a faster onset and fewer and less troubling side-effects. Preregistration is there for a reason, and we should pay more attention to the primary outcome than the secondary, but it’s all interesting and suggests more research is worthwhile.

Some caveats: First, it’s a small study, and study leader Dr Robin Carhart-Harris of Imperial College London pointed out to me, correctly, that by the standards of exploratory Phase II trials it’s a pretty fair size (and we’ve all been spoiled by these N=30,000 Covid vaccine trials). But still, it’s small and should be treated with caution. 

Second, RCTs like this are supposed to be “double-blind” – that is, neither the subject nor the doctor know who’s had the real drug and who’s had the control. Obviously it’s hard to properly hide from people whether they’ve had a psychedelic trip or not, but they hoped the way it was administered would put some doubt in people’s minds. Dr Carhart-Harris said that even in clinical trials into “normal” SSRI antidepressants, about 80% of subjects can correctly guess whether they’re in the treatment group or the placebo one, so this isn’t completely unusual.

And third, the people on the trial were not randomly selected from the population but were self-referred, and were probably all people who wanted to try psychedelics. Dr Carhart-Harris agreed that those who didn’t get the “big, mystical experience you get with psilocybin” would likely have been disappointed.

These are good reasons to be a bit cautious. Certainly the study hasn’t “proved” anything. But I think it’s one more decent data point in a growing body of evidence that psychedelics can have a positive effect on depression, and as long as everyone is upfront about the fact that it’s exploratory and should simply point us towards the value of doing more research, then it’s fine. Other scientists broadly seem to say that it’s definitely good new evidence, if not some huge breakthrough.

But the reason I find it really interesting is the theory behind it. The idea is that depression is caused by our mental models of the world going wrong, and that psilocybin relaxes the hold those models have on our sense of reality.

Here’s what I mean by that. We feel like we’re looking out through a clear, unvarnished window onto the world, but (obviously) that’s not how it is. We only see colour and sharpness in a tiny area in the centre of our field of vision; the image on your retina is distorted by the curve of the eyeball and half-hidden by blood vessels; there’s a big missing bit where the nerve cables to the brain leave the eye. It blurs and moves constantly as our eyes, head and body move around. But our brain is constantly using that information to create what feels like a pin-sharp, three-dimensional image of our surroundings, in real time.

It does that by combining the information from your senses with an existing model. If new information doesn’t meet expectations, it is either ignored or used to update the model. So, to use an example from an earlier paper of Carhart-Harris’s, I have a strong prior belief that walls don’t move rhythmically as though they’re breathing. The unfiltered sense data coming in from my senses is noisy and weird and sometimes it might contain information that could be interpreted as “the walls are breathing”. But my top-down model of the world has that strong prior belief – walls don’t breathe! – so the information is ignored as an anomaly.

Of course, if I had some weaker belief, such as “I used the yellow cup when I made coffee,” and then I looked down at my cup and it seemed blue, then the sense data would override my prior belief, updating it (and it would register as a surprise: my attention would be called to it). 

You can see your priors at work when you interpret ambiguous illusions, such as the McGurk Effect or this video of a mask. You don’t tend to see inside-out faces, so your brain really struggles to see one: your strong prior belief is that faces come out, not in.

What psychedelics seem to do is reduce the strength of your top-down models, your prior beliefs, about the world. That means that the sense data coming in can update them more easily. Your rock-solid assumption that walls don’t breathe is rendered much weaker, so random noisy fluctuations in the data coming from your eyes override it, and (as often happens with psychedelic experiences) still objects seem to move: walls seem to breathe, patterns on wallpaper seem to crawl. Things that are utterly familiar which your attention would normally ignore altogether – things like your hand – become strange and alien. I assume you’d be much more capable of seeing that the mask was inside-out if you were on mushrooms.

(People may recognise that this is just Bayes’ theorem: you have a prior belief, of some given level of confidence; new information comes in, and you update that level of confidence appropriately. If your prior belief is strong enough, then it will take a lot of new information to shift it.)

I’ve been talking about visual “beliefs” here – walls don’t move, my coffee cup is yellow, my hand is a pretty normal thing which I don’t need to pay too much attention to. But this system of prior beliefs updated with new information applies to much higher-level concepts as well. Imagine I have a prior belief that I am a productive and decent member of society, and then some new information comes in that I am in fact a chiselling little crook: say, I get convicted of fraud. Depending on the confidence of my prior belief and the strength of the new information, I will reduce my confidence in my belief or even change it altogether.

What Carhart-Harris and Karl Friston suggested in their previous paper (see here for a good writeup) was that in depression, our prior beliefs get stuck in a sort of “hole” that is too deep for new information to shift it out of. We form pathological, incorrect beliefs about how terrible everything is or how worthless we are, and the belief is so confident that new information coming in – good news or positive feedback or whatever – is unable to shift it.

They have a metaphor of a sort of belief “landscape”, with hills and valleys. You are a little car or something on the landscape; you naturally roll downhill, but you can go a little way uphill, with work. The further down you are, the “truer” your beliefs are, or – more precisely – the more accurately they match your experiences. So your goal is to get as low as possible. Strong beliefs like “walls don’t breathe” are very deep valleys, and it takes huge amounts of new information to “drive” up the sides of them to get out; weaker beliefs such as “my coffee cup is yellow” are shallower, and just a little bit of information lets you get out and roll down into the deeper, more accurate valley, marked “my coffee cup is blue”.

The trouble is, you might end up getting stuck in a local dip, unable to climb out into the deeper, better valley next door. That’s a delusional state: you have some belief, “the US government is run by paedophiles out of a pizza restaurant”, say. The world would make more sense if you didn’t hold that belief, but the sides of the valley are too steep to get out into the next one, in which the US government isn’t run by paedophiles out of a pizza restaurant. So even though new information comes in, it’s never enough to shift you out, and you keep rolling back down.

This is what is apparently going on with depression, according to Carhart-Harris. Your prior beliefs have formed a “landscape” in which you are stuck in a delusional belief about your own worthlessness or the terribleness of the world. If you could get enough information to get you out of it, you’d roll down into the deeper, more accurate valley next door in which you’re not depressed. But you can’t.

Now, psychedelics. Remember how they weaken your Bayesian priors? In this metaphor, that means they reduce the height of the hills and the depths of the valleys: they flatten the landscape. Suddenly, while on psychedelics, pessimistic or negative beliefs are weakened (along with beliefs in non-breathing walls). So your little car can drive out of the valley and by the time the psychedelics wear off, and the hills and valleys become steep again, you have successfully moved into the non-delusional bit and stay there.

This Bayesian model seems to be the closest thing there is to a working theory of how the brain works; it seems to explain decision-making and willpower quite effectively, among many other things. We have prior beliefs and update them all the time with new information. Our brains are constantly generating a model of the world, and hallucinations happen when our confidence in that model is reduced so we pay more attention to anomalous sensory fluctuations.

Of course, the fact that there’s a convincing-sounding theory behind psychedelic therapy for depression doesn’t mean it’s true; one scientist friend says that “When people repurpose a drug/activity that they enjoy taking/doing, and say it cures depression, I’m highly sceptical. You might even call it my … prior.” So let’s remain cautious.

And, of course, even if it does work, psychedelics relax your priors on true beliefs as well as false ones – so it is possible that you would climb your way out of some real belief and then settle into a delusional one. Carhart-Harris agrees this is a risk, albeit a rare one, and that’s why it’s important to use psychedelic therapy only in carefully controlled situations.

And there’s a further problem that even if it is shown to work, there’s a strong societal and regulatory resistance to the use of recreational drugs as medicines. Professor David Nutt, a professor of neuropsychopharmacology, also at Imperial and who also worked on the study, agreed that that is a difficulty, but he felt that the growing evidence of their effectiveness and the overwhelming evidence of their comparative safety should overcome it. (“I wouldn’t care if they ended up in the same category as morphine,” he said, “as long as they can be used in medicine. But they shouldn’t have been illegal in the first place.”) Other psychoactives, such as ketamine and MDMA, are also being studied for use on depression and PTSD.

I hope psychedelics turn out to be effective against depression, if only because almost nothing else is – antidepressants do “work”, as in have more effect than placebo, but it’s extremely hit-and-miss and any given SSRI will probably have no therapeutic effect in any given patient: doctors will probably have to try several before finding an effective one. But I also hope they’re effective because I think the theory is lovely and elegant, making good clinical use of this theory that our brains are Bayesian reasoning machines, and that sometimes our priors can get too strong and need a bit of massaging to relax them.

And I also hope they start selling them in Camden Market again. I’m too old for that stuff these days, but I don’t get why the young people shouldn’t be allowed to have fun.

---

Tom Chivers is a science writer. His second book, How to Read Numbers, is out now.

TomChivers