# Lesson 5: The Psychedelic Experience: Mystical States, Set and Setting, Expectancy and Blinding

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Here's an experiment that two researchers proposed in a 2021 paper. Give a person a psychedelic while they're under deep anesthesia, so they're fully unconscious and remember nothing of it. If they still get the full, lasting benefit, what was the experience for?

**Kiffer:** Yaden and Griffiths proposed that test. They said they wouldn't be surprised if some benefit remained, though smaller or shorter-lived, and that if the full, lasting benefit survived, they'd concede that the subjective effects were irrelevant. A trial with a different drug has come close. Forty people with depression received ketamine or saline during anesthesia for routine surgery, and over the next three days ketamine did no better than saline.

**Sarah:** It's tempting to read that as proof that the experience is the medicine. I'm less sure, because the anesthesia took away a lot more than the experience.

**Kiffer:** That's a fair objection, and it's a question you and I see a little differently.

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. Whether the experience itself helps people heal runs through all of Lesson five, The Psychedelic Experience: Mystical States, Set and Setting, Expectancy and Blinding, so let's start with the evidence that it predicts who gets better.

**Sarah:** Question one. If the people with the strongest mystical-type experiences improve the most, why can't we say the experience caused the improvement?

**Kiffer:** Let's put real numbers on it. In a Johns Hopkins trial of psilocybin for major depressive disorder, with twenty-four adults and a waiting-list control, the peak rating of mystical-type experience across the two sessions correlated minus 0.45 with depression scores four weeks after the second session. Peak psychological insight correlated minus 0.75. And the therapeutic alliance, rated by participants at their final preparation session, before they'd taken any psilocybin, correlated minus 0.65.

**Sarah:** Let me turn those into something easier to compare. If you square a correlation, you get the share of the variation in one measure that goes with the other. Minus 0.45 squared is about 0.2, so the mystical rating goes with about twenty percent of the variation in depression. Insight, 0.75 squared, is about fifty-six percent. And the alliance, 0.65 squared, is about forty-two percent.

**Kiffer:** So in this trial, the feature with the longest research history is the weakest of the three, and a relationship measured before dosing predicts nearly as strongly as insight. Each correlation rests on twenty to twenty-four people, so every one of these numbers is uncertain, and I wouldn't rank them with any confidence. What they do show is that several parts of the treatment move with the outcome, and the mystical experience is one of them.

**Sarah:** And the alliance has a twist, doesn't it?

**Kiffer:** It does. The same trial measured the alliance again a week after the second psilocybin session, and that later rating correlated minus 0.85 with depression at four weeks, which is about seventy-two percent of the variation. So here's a question for everyone. Why would the alliance measured after dosing predict depression so much more strongly than the alliance measured before dosing? Take a few seconds, and pause the audio if you'd like more time.

*(Pause)*

**Sarah:** Because by a week after the second session, many people may already have started to feel better, and someone whose depression is lifting is likely to rate the relationship with their therapists more warmly. So part of that minus 0.85 could run backward, from improvement to the rating.

**Kiffer:** That's one explanation, and the sessions themselves are another, since a powerful shared experience can change how a person feels about the people who sat with them. The trial's authors raise that possibility themselves. Either way, it's the time-order problem from the distinction between prediction and mediation. A predictor has to be measured before the outcome starts moving, or the correlation can't tell you which way the influence runs. The preparation-session rating passes that test, and the post-session rating fails it. And with about twenty people, chance could explain part of the gap too.

**Sarah:** The alliance could also be one of the dashed paths in the experiential model. It correlated 0.49 with the mystical rating as well as with depression, so a good relationship might produce both a better session and a better recovery.

**Kiffer:** It might, and the trial comparing psilocybin with escitalopram suggests the influence runs both ways. In its psilocybin arm, a stronger alliance predicted more emotional breakthrough and mystical-type experience, which predicted lower depression, and a breakthrough in the first session then predicted a stronger alliance before the second. When the experience, the relationship and the outcome feed one another like that, a correlation between any two of them can't tell you which one is doing the work.

**Sarah:** Question two. If the dose is the strongest predictor of what happens in a session, how much can the room really matter?

**Kiffer:** The dose part is well supported. When researchers in Zurich pooled twenty-three laboratory studies with two hundred sixty-one healthy volunteers, dose was the strongest predictor of every response they measured. Among the factors that weren't pharmacological, though, a session in a brain scanning centre was one of the strongest predictors of an anxious reaction, and the authors suggested that its clinical atmosphere, with technical equipment, white walls and staff in lab coats, might partly explain that.

**Sarah:** And then there's the McGill study, where nobody got a drug at all. Thirty-three students received a placebo in a room set up like a psychedelic party, and sixty-one percent, about twenty of them, said they felt some effect. Some had altered-state scores of a size usually seen with moderate or high doses of psilocybin.

**Kiffer:** Which shows how far context and expectation can go by themselves. What it can't show is whether changing the room changes how much people recover in therapy. As far as I know, no randomized trial has compared session rooms directly, so most of the evidence on the physical environment comes from comparisons across studies and from clinical experience.

**Sarah:** And comparisons across studies can mislead. In the 1960s, the Toronto group tried to repeat the Saskatchewan results in a deliberately detached setting, with observers told not to interact with patients, and found lower improvement rates. It's tempting to read that as an experiment on setting, but the two research groups may also have differed in their patients, their doses and how they measured outcomes.

**Kiffer:** Right, and that's why reporting matters. A systematic review of psychedelic trials in psychiatric disorders, published up to December 2023, found that only fifty-two percent of the twenty-five included studies reported on monitor training. That works out to thirteen studies. If a paper doesn't say who sat with the participant and how they were trained, nobody can compare it with another trial or repeat it.

**Sarah:** Which is what the 2025 reporting guidelines are for. Eighty-nine experts from seventeen countries agreed on thirty items, grouped into four categories: the physical environment, the dosing session procedure, the therapeutic framework and protocol, and participants' subjective experiences.

**Kiffer:** So let's use them. Here's a short methods paragraph I've made up. "Dosing sessions took place in a comfortable, living-room-style room. Participants wore eyeshades and listened to music through headphones. Two trained facilitators were present throughout." If you were reviewing this paper, what would you ask the authors to add? Take a few seconds.

*(Pause)*

**Sarah:** For the physical environment, I'd ask what comfortable means: the lighting, the decorations and whether medical equipment was visible. For the procedure, how long the session ran, when vital signs were taken, how interruptions were handled, what happened if someone became distressed, whether touch was used and how consent to it was obtained, and how the session ended. For the therapeutic framework, how many preparation sessions there were, how long they lasted and what approach the facilitators followed. And I'd ask what was recorded about participants' experiences, such as their trust and their sense of safety.

**Kiffer:** Those are the right questions, and I'd add two. The word "trained" tells a reader nothing by itself, and it's exactly the gap that systematic review found. I'd also ask about the music: which playlist, who chose it and whether participants could change it. In a study of nineteen patients with treatment-resistant depression, how they experienced the music predicted their improvement a week later, while the overall intensity of the drug didn't.

**Sarah:** Question three. If almost everyone in a trial can tell which arm they're in, should we throw out the result?

**Kiffer:** I'd keep the result, read it as an upper estimate and ask what drove the guesses. The guesses can be extreme. In a trial of psilocybin for alcohol use disorder with diphenhydramine as the comparator, participants guessed correctly after about ninety-four percent of sessions.

**Sarah:** And in the second phase three trial of MDMA-assisted therapy for post-traumatic stress disorder, the blinding index was 0.92 in the MDMA arm and 0.55 in the placebo arm. Unblinding can also change who stays in a trial. In that MDMA trial, eight of fifty-one people in the placebo arm dropped out, compared with one of fifty-three in the MDMA arm. The authors attributed the gap partly to the less effective treatment in the placebo arm and partly to disappointment from ineffective blinding.

**Kiffer:** There's a caution from the lesson I want to push further, though. A correct guess can come from feeling better, and a drug that works makes more people feel better.

**Sarah:** So how much can that alone do?

**Kiffer:** Let's test it with a harder example, and the numbers are hypothetical. Imagine a trial with fifty people in each arm and a drug with no side effects anyone can notice. Thirty of the fifty on the drug improve, and fifteen of the fifty on placebo improve. Everyone guesses from one thing only: if they feel better, they guess drug, and if they don't, they guess placebo. Nobody says they're unsure. What's the blinding index in each arm? Take a few seconds.

*(Pause)*

**Sarah:** In the drug arm, the thirty who improved guess drug, which is correct, and the twenty who didn't improve guess placebo, which is wrong. Thirty minus twenty is ten, and ten divided by fifty is 0.2. In the placebo arm, the thirty-five who didn't improve guess placebo, which is correct, and the fifteen who improved guess drug. Thirty-five minus fifteen is twenty, and twenty divided by fifty is 0.4.

**Kiffer:** There are a few traps in this one. The placebo arm looks more unblinded than the drug arm, because most people on placebo didn't improve. The drug arm sits at exactly 0.2, which fails the stringent rule, because that rule needs a value below 0.2, though it meets the less stringent one. And none of the guessing came from side effects. Every guess came from whether a person improved, and some of that improvement would have happened on placebo anyway. If the drug did nothing at all and fifteen people improved in each arm, the drug arm would sit at minus 0.4 and the placebo arm at 0.4.

**Sarah:** So a high blinding index tells you that people guessed well, and it can't tell you what they based their guesses on.

**Kiffer:** Right. The FDA's 2026 guidance suggests blinding questionnaires with certainty ratings, and I think the timing of the question matters too. A guess made soon after the first dose, before most improvement, mostly reflects how the drug feels. The ketamine trial in Dublin that used midazolam as its comparator shows the pattern. The midazolam arm's index was minus 0.07 after the first infusion and 0.2 by the end of treatment. The trial can't tell us how much of that rise came from how people's symptoms changed and how much from noticing the effects over several infusions, and asking at both times at least makes the change visible.

**Sarah:** And expectancy is the other half. A 2021 review of psychedelic trials argued that their effect sizes are likely overestimated because of unblinding and high expectations. Then an analysis of the psilocybin and escitalopram trial found that expectancy predicted improvement on escitalopram and didn't predict improvement on psilocybin.

**Kiffer:** Those findings conflict, and I'd hold both loosely. The second analysis had fifty-five participants, and it measured expectancy before treatment, which misses the expectations that form once someone recognizes the drug. The FDA guidance suggests a practical step: consider measuring expectancy before randomization and again at the end of treatment, and ask participants and investigators what they think was given. I'd add that those guesses should be reported for each arm.

**Sarah:** Before question four, a short content note, because this part discusses suicidal thoughts and suicide attempts. If you'd like support, you can call or text nine eight eight, the Suicide Crisis Helpline, at any time. The Hope for Wellness Help Line for Indigenous people is at one eight five five, two four two, three three one zero. So, question four. If most people say a challenging experience was good for them, should we count it as harm or as part of the therapy?

**Kiffer:** Start with the most reassuring figure. In an online survey of one thousand nine hundred ninety-three people describing their single most difficult experience with psilocybin mushrooms, eighty-four percent said they had benefited from it. The same survey found that eleven percent had put themselves or others at risk of physical harm during the experience, and 2.7 percent had received medical help. Among people whose experience was more than a year old, 7.6 percent had sought treatment for lasting psychological symptoms, and across the whole survey, three cases appeared to be associated with attempted suicide.

**Sarah:** And everyone in that survey chose to answer a study about difficult experiences. That's the denominator problem. The survey can describe the range of what happens, and it can't tell us how often it happens.

**Kiffer:** For frequency, the nationally representative sample in the United States is better. Of six hundred thirteen adults who had ever used a classic psychedelic, about forty-one percent had had at least one challenging experience, 8.9 percent reported impairment lasting more than a day, and 2.6 percent sought help afterward.

**Sarah:** And in clinical and research settings, a 2024 meta-analysis found serious adverse events, including worsening depression and suicidal behaviour, in about four percent of participants with existing neuropsychiatric disorders and in none of the healthy participants. Lesson six takes those risks up. But I want to try something with that eighty-four percent. Suppose, and this is purely hypothetical, that people who felt they benefited were twice as likely to answer the survey as people who didn't. What share of everyone with such an experience would say they benefited? Take a few seconds.

*(Pause)*

**Sarah:** My first instinct is to halve it. Eighty-four divided by two is forty-two percent.

**Kiffer:** Check it before you go on. If forty-two percent benefited and they answer twice as often, what share of the people answering would report benefit?

**Sarah:** Okay. Out of a hundred people, forty-two benefited and answer at double the rate, so they count as eighty-four. Fifty-eight didn't, and they count as fifty-eight. Eighty-four out of one hundred forty-two is about fifty-nine percent. That's nowhere near eighty-four, so halving was wrong.

**Kiffer:** The doubling acts on the odds of benefit. Try converting the percentage to odds first.

**Sarah:** Eighty-four to sixteen is odds of 5.25. Halve that and I get 2.625. To turn odds back into a share, divide by one plus the odds, so 2.625 divided by 3.625, which is about 0.72. So about seventy-two percent. Checking it, seventy-two times two is one hundred forty-four, compared with twenty-eight who didn't benefit, and one hundred forty-four out of one hundred seventy-two is about eighty-four percent. That works.

**Kiffer:** Good. So a twofold difference in who answers would turn a true seventy-two percent into an observed eighty-four. The benefit figure holds up fairly well under that assumption. I'd worry more about the harm figures, because people who were harmed and don't want to revisit the experience may be the least likely to answer at all.

**Sarah:** And benefit and harm can sit in the same person. In interviews with fifty people in Norway, the stories people told about frightening experiences seemed to help them cope, and the researchers concluded that this storytelling may also make continued use more likely. A story can make a harmful experience easier to live with and leave the harm in place.

**Kiffer:** The worst-experience survey adds one more piece. Greater difficulty went with lasting gains in well-being, while a longer difficulty went with smaller gains. And in a survey of six hundred eight people with difficulties after use, about a third had problems lasting more than a year, and six percent reported suicidal thoughts or thoughts of self-harm. So my answer to your question is that it can be both. The honest place for that is consent, preparation and follow-up, where people hear about difficulty and the chance of lasting problems as plainly as they hear about benefit.

**Sarah:** Okay. Now the part where we disagree, which takes us back to the anesthesia experiment. Is the experience necessary?

**Kiffer:** Let me set out the two positions. Yaden and Griffiths argue that the subjective effects are necessary and account for most of the benefit, partly because mystical-type experience and insight still predict outcomes after statistically controlling for the overall intensity of the drug's effects. Olson argues that the evidence is mostly correlational, that mystical scores may track how strongly serotonin 2A receptors were activated, the way blood pressure rises with dose, and that compounds engineered to avoid hallucinogen-like effects in rodents keep their effects on plasticity. He disclosed that he was president and chief scientific officer of a company, Delix Therapeutics.

**Sarah:** I lean toward Olson's side, at least on what the evidence shows so far. The anesthesia trial removed the experience, and it also kept people from knowing what they'd received and had no therapy around the dose, so a null result there fits the expectancy explanation as well as the experiential one. And the access argument matters to me. A drug that worked without the experience could reach people who are excluded now, such as people with a family history of psychosis, and it could be given without long supervised sessions.

**Kiffer:** I agree with you about the anesthesia trial. It concerned ketamine, it followed people briefly, and it can't separate experience from expectancy. Where I land differently is that I think the experience probably contributes. The associations generally hold after controlling for intensity, and a small study in eight healthy volunteers found that greater memory impairment from midazolam given with psilocybin tended to go with lower ratings of insight and well-being. The setting evidence also shows that the content of a session varies with things that have little to do with receptor activation.

**Sarah:** But "probably contributes" is a long way from "necessary".

**Kiffer:** It is, and I wouldn't claim necessary. The study that would come closest to settling it is a blinded trial in people comparing a psychedelic with a non-hallucinogenic analogue, and Lesson four described those compounds, which so far have mostly animal evidence behind them. Until then, I think psychedelic-assisted therapy should be designed as if the experience matters, because with the drugs we have, people will have one either way.

**Sarah:** I'm more open to the drug-only account than Kiffer is, but we agree on the practical rule. Read a link between experience and outcome as prediction until a study changes the experience while holding the drug constant, and as long as the drugs in use produce an experience, prepare and support people for it.

**Kiffer:** Let's pull it together with three things to take away.

**Sarah:** First, a feature of treatment can predict improvement without causing it, and when it's measured matters. The alliance rated after dosing predicted far more strongly than the alliance rated before, possibly because improvement had already begun.

**Kiffer:** Second, the blinding index describes how well people guessed, and it can't say why they guessed that way. Guesses can come from side effects or from whether people improved, so ask early and late, with certainty ratings, and report each arm separately.

**Sarah:** And third, every figure on challenging experiences comes with a denominator and a way of recruiting people. Eighty-four percent reporting benefit, forty-one percent with at least one challenging experience, and four percent of patients with serious adverse events answer three different questions, and benefit and harm can occur in the same person.

**Kiffer:** If you'd like more practice, rework the hypothetical blinding example from this episode with forty of the fifty people on the drug improving, and work out which arm's index changes and by how much.

**Sarah:** Next time, it's Lesson six, Safety, Screening and Risk Management, where the psychiatric risks we touched on today meet the screening and support that trials use to reduce them.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson six.
