The Setup
“You were told it was fake. You took it anyway. Why did you get better?”
An inert pill can win even when you know it’s inert.
The old objection to placebo research is that it runs on deception — that the effect needs the lie. A 2026 randomized trial cut the lie out and kept the effect.
Barbiani D., Antonietti A., Pagnini F. et al. — “Placebo mechanisms in aging: a randomized controlled trial comparing deceptive and open-label placebos…” Int. J. Clinical & Health Psychology (2026). DOI 10.1016/j.ijchp.2026.100673. Università Cattolica, Milan. ✓ Sourced
90 healthy older adults, three arms, three weeks. One arm got nothing. One got a deceptive placebo — told the pills contained active ingredients. One got an open-label placebo (OLP): told plainly the pills were inert, but that they could still trigger beneficial mind–body responses. paraphrase, not a verbatim quote
| Outcome | Deceptive placebo | Open-label (told it’s inert) |
|---|---|---|
| Physical performance | +7% | +9.2% |
| Cognitive performance | +12.6% to +14.6% | +6.9% to +21.5% (test-dependent) |
| Perceived stress | — | lower than BOTH other groups |
| Short-term memory | — | improved vs. control |
The finding: the honest placebo matched or beat the lie. Pagnini put the size of it as “comparable to those seen in some experimental studies on physical activity… especially with regard to memory.” ✓ Sourced
n = 90. Three weeks. Healthy volunteers. The authors themselves call it preliminary and flag mechanism, durability and generalizability as unconfirmed. Limited It shows the effect survives honesty. It does not yet show how big, how long, or for whom.
You didn’t report relief. You made it, out of your own opioids.
Levine J.D., Gordon N.C., Fields H.L. — “The mechanism of placebo analgesia.” Lancet 1978;2(8091):654–657. PMID 80579. ✓ Sourced
Patients had impacted wisdom teeth extracted. Hours later they got naloxone — a drug that blocks opioid receptors — or a placebo, double-blind. The logic that came out is exquisite:
- Naloxone patients reported significantly greater pain than placebo patients. ✓ verbatim
- Placebo-first patients split into responders (pain reduced or unchanged) and nonresponders (pain increased). Naloxone given second produced no further pain rise in nonresponders but did raise pain in responders. ✓ verbatim
- Reported detail: nonresponders’ final mean pain equalled that of responders after naloxone — i.e. naloxone took the relief away and left responders exactly where the non-responders already sat. Unverified — figure not independently loaded
And it has a dose.
Levine J.D., Gordon N.C., Smith R., Fields H.L. — Pain 1981;10:379–389. PMID 7279424. ✓ Sourced
In the same post-surgical pain model, the paper reports that mean placebo relief fell between that obtained following hidden administration of 4 and 6 mg of morphine. a range, 4–6 mg — not a point estimate That is the honest number, and it is extraordinary without any help: a real analgesic dose, self-manufactured, for the price of a sugar pill and a story. We print the range everywhere it appears. The ceiling is not the finding.
The field fought about naloxone for twenty years. Both sides were right.
“Naloxone blocks placebo, QED” is too clean. The pushback was immediate and it was real.
Goldstein A. & Grevert P., Lancet 1978;2:1385 — immediate same-year pushback. Unverified — letter not independently loaded
Gracely R.H., Dubner R., Wolskee P.J., Deeter W.R. — “Placebo and naloxone can alter post-surgical pain by separate mechanisms.” Nature 1983;306:264–265. 🔴 Contested
Gracely reported, in the abstract, that “placebo analgesia can occur after blockade of opioid mechanisms by naloxone,” and that naloxone can cause hyperalgesia independent of any placebo effect. ✓ verbatim That is a counterpoint aimed squarely at the story this very tab is built on. We keep it on the page.
The resolution — and it’s subtler than “expectation vs. conditioning.”
Amanzio M. & Benedetti F. — “Neuropharmacological dissection of placebo analgesia…” J. Neuroscience 1999;19(1):484–494. PMID 9870976. ✓ Sourced
There isn’t one placebo effect. There are at least two systems — but the switch that decides which one you get is which drug you were conditioned with, not simply “expectation vs. conditioning.” The paper reports:
Morphine conditioning alone (without expectation cues) induced a naloxone-reversible placebo effect. Opioid in, opioid out — naloxone blocks it. ✓ verbatim
Ketorolac conditioning alone (a non-opioid) induced placebo responses that were completely insensitive to naloxone. Different machinery, untouched by the blocker. ✓ verbatim
So placebo analgesia dissects into opioid and non-opioid components depending on the induction procedure. Levine saw the opioid arm. Gracely saw the non-opioid one. Both camps were correct about different machinery — and the variable nobody had named was the drug in the conditioning history.
The Hard Truth
“Your mind changed. The disease did not.”
Two columns. Read them against each other.
Wechsler M.E., Kelley J.M., Kaptchuk T.J. et al. — “Active albuterol or placebo, sham acupuncture, or no intervention in asthma.” NEJM 2011;365(2):119–126. PMID 21751905. ✓ Sourced
A double-blind crossover: 46 asthma patients randomized, 39 completed, each receiving all four conditions — real albuterol, a placebo inhaler, sham acupuncture, and no intervention at all.
| Intervention | Subjective improvement (VAS) | FEV₁ — actual lung function |
|---|---|---|
| Albuterol (real drug) | 50% | +20% |
| Placebo inhaler | 45% | ~7% |
| Sham acupuncture | 46% | ~7% |
| No intervention at all | 21% | ~7% |
All three interventions beat no-intervention subjectively (p<0.001). But for lung function, albuterol beat all three others at p<0.001 — and placebo, sham and no-intervention were indistinguishable at ~7%. ✓ table verbatim
The paper also reports clinically-meaningful FEV₁ response (≥12%): albuterol 77% vs placebo 24%, sham 20%, none 18%; the albuterol-vs-placebo subjective gap was not significant (p=0.12, d=0.21). Unverified — full-text responder figures not independently loaded
Subjectively, the placebo tied the drug. Objectively, the placebo did not open a single airway. In a real asthma attack, the placebo group dies feeling fine.
The no-intervention arm also reported 21% improvement. The authors use it to dismiss the “patients just want to please the investigator” objection: those patients had no expectation and no treatment to praise, so wanting to please would have made them report less, not more. The no-intervention arm is what separates a real placebo effect from regression to the mean and the natural course of the disease. It’s the same move SAMSON makes in Tab IV.
Limits: n = 39 completers. Pilot. Crossover. Limited
Placebo is not the poison. Placebo instead of medicine is the poison.
Do not write “placebo makes the disease worse.” It doesn’t. The disease does what it was always going to do. The killer is what you took instead. The accurate version is more damning anyway.
Johnson S.B., Park H.S., Gross C.P., Yu J.B. — “Use of alternative medicine for cancer and its impact on survival.” JNCI 2018;110(1):121–124. PMID 28922780. ✓ Sourced
From the National Cancer Database: 281 patients with non-metastatic breast, prostate, lung or colorectal cancer who chose alternative medicine as their sole anticancer treatment — no chemo, radiation, surgery or hormone therapy. Matched 2:1 to conventionally-treated patients (840 analyzed). Result: significantly worse survival overall. Lung cancer: HR 2.17 (95% CI 1.42–3.32) — more than twice as likely to die. ✓ verbatim
One of the four cancers showed no significant survival difference: prostate cancer, HR 1.68, 95% CI 0.68–4.17, P = .36 — a confidence interval that crosses 1. 🔴 Contested The finding is real for the pooled group and for lung; it is not uniform across sites. If we only flagged the other guy’s null results and hid our own, we’d be running the preacher’s play. So here it is.
The alternative-medicine group was, at baseline, more likely to be younger, female, lower-comorbidity, higher income and better educated — and also of higher cancer stage. stage predicts WORSE survival — it does not cut our way Most of those favor better survival; higher stage does not. So the honest answer to “weren’t they just sicker?” isn’t to hide the stage variable — it’s that the survival analysis was stage-matched (2:1 propensity matching), and the gap survived the matching. That is stronger than the omission would have been.
Johnson S.B. et al. — “Complementary medicine, refusal of conventional cancer therapy, and survival.” JAMA Oncology 2018;4(10):1375–1381. Cohort 1,901,815. ✓ Sourced (cohort corroborated via secondary)
The companion paper isolates the mechanism: complementary-medicine users were more likely to refuse at least one component of conventional treatment, and that refusal carried a ~2-fold greater risk of death. The complementary medicine wasn’t the poison. The refusal it traveled with was.
The one nobody expects — and the one that makes the lab.
The placebo problem is not a problem for tent-revival preachers and cranial hucksters alone. It is a problem inside real hospitals, run by real surgeons with real credentials.
Moseley J.B. et al. — arthroscopic surgery for osteoarthritis of the knee. NEJM 2002;347(2):81–88. ✓ Sourced — quotes verbatim
180 patients, three arms: arthroscopic debridement, arthroscopic lavage, and placebo surgery — verbatim, “patients in the placebo group received skin incisions and underwent a simulated debridement without insertion of the arthroscope.” Patients and outcome assessors blinded. Two-year follow-up.
The authors’ own line: “if the efficacy of arthroscopic lavage or débridement… is no greater than that of placebo surgery, the billions of dollars spent on such procedures annually might be put to better use.” ✓ verbatim
Kirkley A. et al. — arthroscopic surgery vs optimized therapy. NEJM 2008;359:1097–1107. ✓ Sourced
Surgery added nothing over optimized physical and medical therapy at two years. The surgery group did do better for the first three months — and the authors wrote that this “transient benefit was anticipated, since sham surgery is associated with a large, short-term placebo effect.” They predicted the placebo and watched it decay on schedule. Note: Kirkley had NO sham arm — the authors attribute the 3-month bump to placebo by reference to Moseley; they did not measure it.
Laupattarakasem W. et al. — Cochrane 2008, CD005118. ✓ Sourced (full quote restored)
Cochrane graded it “gold-level evidence that arthroscopic debridement has no benefit for un-discriminated OA (mechanical or inflammatory causes).” The qualifier is part of the quote — we keep it.
Vertebroplasty — two NEJM trials, same issue, opposite hemispheres.
Buchbinder R. et al. NEJM 2009;361:557–568 (Australia, n=78). · Kallmes D.F. et al. NEJM 2009;361:569–579 (Mayo, n=131). ✓ Sourced
Both trials found no significant benefit of cementing a fractured vertebra over a sham procedure. Both groups improved — cement or no cement. Kallmes, the Mayo lead, said of his own data: “I simply couldn’t believe what I was seeing.” ✓ verbatim (Buchbinder’s conclusion is corroborated in substance but we did not confirm it as a verbatim quotation. quote unverified)
Critics (AJNR letters, 2009) argued the sham wasn’t inert: both trials injected a short-acting local anesthetic into the periosteum, which carries pain fibers. ✓ the periosteal-injection point verified via secondary And VAPOUR / VERTOS II did favor vertebroplasty — but in a much narrower population (acute inpatients, pain under ~6 weeks, high opioid use); VERTOS IV, blinded and sham-controlled, sided with Buchbinder and Kallmes. VAPOUR / VERTOS II / IV population specifics unverified — not independently loaded
Honest read: vertebroplasty is probably placebo for the broad chronic population and possibly real for a narrow acute one. Say exactly that. Do not flatten it.
FIDELITY — the arrow points the other way.
Finnish Degenerative Meniscus Lesion Study — arthroscopic partial meniscectomy (APM) vs placebo surgery, 5-year follow-up. PMC7606577. ✓ Sourced — numbers verbatim
At five years there was “a consistent, slightly greater risk for progression of radiographic knee osteoarthritis in the APM group as compared with the placebo surgery group”: Kellgren–Lawrence grade increase ≥1, adjusted absolute risk difference 13% (95% CI −2% to 28%); OARSI sum score, adjusted mean difference 0.7 (95% CI 0.1 to 1.3).
Tab II’s closer, with the caveat riding in the same breath: read carefully, it is not the placebo that makes you worse — on this signal-strength evidence it is the unnecessary real intervention you took instead of the placebo.
The Ritual
“The crutches on the stage. The hands on the skull.”
The relationship is a bigger active ingredient than the ritual.
Kaptchuk T.J., Kelley J.M., Conboy L.A. et al. — “Components of placebo effect: RCT in patients with irritable bowel syndrome.” BMJ 2008;336(7651):999–1003. PMID 18390493. ✓ Sourced
262 adults with IBS. Kaptchuk isolated the three components of a healing encounter and added them back one at a time:
| Arm | Components | “Adequate relief” |
|---|---|---|
| Waiting list | assessment + observation only | 28% |
| “Limited” | sham acupuncture + businesslike practitioner | 44% |
| “Augmented” | sham acupuncture + warm, empathic practitioner | 62% |
p < 0.001 for trend. Monotonic. Dose-dependent. ✓ 28 / 44 / 62 verbatim from the primary a widely-circulated secondary source misprints these as 27/43/62 — the primary is right
What “augmented” meant, nearly verbatim: a 45-minute initial visit, active listening, touch, warmth, expressed empathy, confident positive expectation, thoughtful silence, and asking the patient how they understand their condition and how the illness relates to their life. the 45-min figure, the 76%-women/age-39 sample descriptors, and the “dose dependence” conclusion quote were not independently loaded — Unverified
Replication / extension: Fuentes et al. (2014), chronic low-back pain — augmented interaction beat the identical sham delivered coldly. Barnes et al. (2024), Appl. Psychol. Health Well-Being, DOI 10.1111/aphw.12497 — warmth enhances placebo and attenuates nocebo. both Unverified — not independently loaded
Separate the two claims ruthlessly.
Claim A — the stated mechanism: not what they say it is.
An earlier version of this page ran the header “the stated mechanism: false” and told you flatly that the cranial rhythm does not exist. That was too strong, and a primary source says so. We are not going to fix it in the dark.
Rasmussen T.R. & Meulengracht K.C. — “Direct measurement of the rhythmic motions of the human head identifies a third rhythm.” J. Bodyw. Mov. Ther. 2021;26:24–29. DOI 10.1016/j.jbmt.2020.08.018. ✓ Sourced — quotes verbatim from the primary
Fifty heads. Roughly 42 minutes of recording each. Instruments clamped to the skull at 10 g of contact — no palpation anywhere in the design. The result, verbatim: “In all individuals, a third rhythm was distinguished as separate from the arterial and respiratory rhythm at all times… with a mean of 6.16 cycles/minute (4.25–7.07).” ✓ verbatim The authors conclude that this gives “rational scientific evidence documenting the existence of a rhythmic movement different from arterial and respiratory rhythms.” ✓ verbatim
So we withdraw “false” and print the two claims that actually survive contact with the evidence. They are stronger than the one we cut.
A1. The rate they measured matches none of the rates the schools teach.
Rasmussen found a rhythm. It is not their rhythm. Line the numbers up:
| Where the number comes from | Cycles / min | Provenance — in the same breath |
|---|---|---|
| Upledger & Vredevoogd — the normal range therapists are taught to presume | 6–12 | ✓ Sourced Wirth-Pattullo & Hayes 1994, p.914, verbatim: “Therapists presume the normal range of craniosacral rates to be 6 to 12 cycles per minute.” |
| Woods & Woods 1961 — the first palpation study | 10–14 | ✓ Sourced Rasmussen 2021, Discussion p.28, verbatim: “reporting a CRI range of 10–14 cpm.” |
| “8–12” — a rate this lab printed in an earlier pass | 8–12 | ⚪ Unverified — and we went looking. It appears in none of the six primary sources loaded for this pass. We are not deleting it quietly and we are not laundering it: we cannot source it, so it does not count as evidence here. |
| What the instruments actually recorded (Rasmussen 2021, n=50) | 6.16 (range 4.25–7.07) | ✓ Sourced — verbatim Direct measurement. No hands involved. |
The measured mean, 6.16, sits on the very bottom edge of Upledger’s 6–12 and nowhere near Woods & Woods’ 10–14. The measured range is worse for them: its ceiling — 7.07 — is below the floor of Woods & Woods (10). Rasmussen’s own Discussion makes our point for us: palpated rates show “a wide range from which it has been challenging to create a normative range,” and “it is possible that different studies may report on different rhythms under the same name, the CRI.” ✓ verbatim
Rasmussen does not agree with us. We quoted the two lines of his Discussion that help our case and we were about to walk away from the paragraph that doesn’t. That is the preacher’s play, and the house rule says it goes on the page — first, and unsoftened. The same Discussion, one paragraph later, verbatim:
“The rate of the third rhythm measured in this study is similar to the rate reported in a large palpation study (Sergueef et al., 2011), and with palpations simultaneously measuring the Traube-Hering-Mayer oscillations (Nielson et al. 2001, 2006). Further, the reported rate here is similar to the experimental study by Fryman (1971) using a similar direct measurement of the head movements.”
“Thus, the third rhythm reported here may be the same as the CRI reported in the studies above (Sergueef et al., 2011; Nielson et al. 2001, 2006; Nielson et al., 2001, Fryman 1971).” ✓ verbatim — Rasmussen 2021, Discussion, p.28
Read that against the header of this section. A1 says the measured rate “matches none of the rates the schools teach.” The authors of the study we are leaning on say their rate matches a large palpation study, and that what they measured may be the very thing the palpators call the CRI. That is the strongest single sentence against A1 in the literature we hold, and it is in our own source. 🔴 Contested
What the lab has left, honestly: not much, but not nothing. The claim that survives is the narrow one, and only the narrow one — the measured ceiling of 7.07 sits below Woods & Woods’ floor of 10, so whatever Rasmussen recorded, it is not the rate that study taught. Upledger’s 6–12 is a different matter: 6.16 sits inside it, at the bottom edge, and Rasmussen’s own reading of the palpation literature is that this may be the same phenomenon. So the honest version of A1 is narrower than the header we wrote: the measured rate is inconsistent with some of what is taught, not with all of it, and the man who measured it thinks the palpators may have been feeling something real. We are not deleting the header. We are printing the sentence that shrinks it, in the same breath, where you cannot miss it.
A2. And two trained practitioners palpating the same skull cannot agree on when it’s happening.
This is the part that does survive, and it survives hard:
| Study | Finding — with its flag |
|---|---|
| Wirth-Pattullo V. & Hayes K.W., Phys Ther 1994;74(10):908–916 (PMID 8090842) ✓ Sourced — primary now loaded, quotes verbatim | Three examiners, craniosacral rate measurements. ICC = −.02 ✓ verbatim — greater variance among the examiners than among the subjects, which is what drives an ICC below zero. Interexaminer correlations: r = −.33, −.60, +.49. Agreement within ±1 cycle: 44%. The authors, verbatim: therapists “were not able to measure it reliably”; the error “may be sufficiently large to render many clinical decisions potentially erroneous”; and, in the Discussion, “Craniosacral motion may not exist and might be imagined by the evaluator.” |
| Moran R.W. & Gibbons P., JMPT 2001;24(3):183–190 (PMID 11313614) ⚪ ICC figures Unverified — primary NOT among the sources loaded this pass | Two registered osteopaths with postgraduate cranial training, simultaneous head + sacrum palpation: “interexaminer reliability… was poor to nonexistent.” Reported ICC range in this paper: −0.09 to +0.31. carried with the flag on — see the warning below |
| Alvarez L.A. et al., Cureus 2024;16(8):e68219 (PMID 39347206) · DOI 10.7759/cureus.68219 ✓ Sourced — primary now loaded | 44 subjects; cranial vault hold vs. direct palpation. “The positive kappa values, all between 0 and 0.2, indicate the inter-rater reliability for diagnosis with the vault hold is above random chance but has none to slight reliability” — and “none of the positive kappa values were statistically significant (p>0.05).” ✓ verbatim κ = −0.970 — now sourced, and stated precisely. ✓ Sourced It is one cell of one table: Table 2, right fibular head, cranial physician 1 vs. the confirmatory group — 0/44 (0%) agreement. Not the study’s headline result; a single landmark. The paper’s own gloss on it, verbatim: “The kappa coefficient suggests almost perfect disagreement (−0.970) between cranial physician 1 and the confirmatory group.” |
Wirth-Pattullo’s ICC = −.02 and Moran & Gibbons’ ICC = −0.09 are different numbers from different studies with different designs. They are both negative and they are both small, which is exactly what makes them easy to merge by accident into one tidier-sounding figure. We hold them apart on purpose. The −.02 is sourced from the primary, which we now hold. The −0.09 is not — Moran & Gibbons was not among the papers loaded this pass, so it keeps its ⚪ flag until somebody puts the paper on the table. A number that gets promoted by standing next to a sourced one is how a citation goes bad.
Alvarez contains a result the tidy version of this argument would rather you didn’t see. Comparing the two cranial physicians to each other, they agreed above random chance (κ>0) on 6 of 11 segments — OA, AA, C4, sacrum, left innominate and right fibular head. The authors’ own reading, verbatim: “This finding may support the idea that they were palpating the same phenomenon.” 🔴 Contested So do not write “they never agree with each other” — this page won’t. The honest counter-counter is the authors’ own: the effect sizes were small, none were significant, and they attribute the agreement to “shared bias among the evaluators” — two people trained in the same school, taught to expect the same thing, expecting it together. That is a live argument, not a settled one. It stays on the page.
Sourcing corrections, printed: a “JOSPT 2006 review (Rogers et al.)” reporting an ICC range of −0.09 to 0.59 could not be located at all and has been cut, not printed — and both of its headline numbers turn out to be corruptions. The founder’s own reliability study is Upledger J.E., J Am Osteopath Assoc. 1977;76:890–899 — 1977, not 1997, now settled from Wirth-Pattullo’s primary reference list, ✓ Sourced — the year is no longer in question and its real reliability figure is .57 (Wirth-Pattullo, p.910: “The reliability of his data was .57, which we believe is too low to support his conclusion…”), not the 0.59 the phantom review claimed. ✓ verbatim The cut was right, twice over.
Claim B — the delivery system: this lab’s hypothesis, and the trial that went after it.
Fifty minutes of quiet, sustained, gentle touch, from a confident person giving you their complete undivided attention, in a calm room, who asks how you understand your own suffering. That is Kaptchuk’s “augmented” arm at maximum dose. So the natural move — the move this lab made, in a box exactly like this one — is to say the ritual is a wrapper and the time is the drug:
That line used to sit on this page in a green box, flagged as a framing line and nothing more. It is not “nothing more.” It is a falsifiable claim about what is doing the work — and somebody built the experiment that tests it.
Haller H., Lauche R., Cramer H., Rampp T., Saha F.J., Ostermann T., Dobos G. — “Craniosacral Therapy for the Treatment of Chronic Neck Pain: A Randomized Sham-controlled Trial.” Clin. J. Pain 2016;32(5):441–449. NCT01526447, Kliniken Essen-Mitte / Univ. Duisburg-Essen. ✓ Sourced — primary now loaded, quotes verbatim
This page previously disarmed Haller by saying a real effect with a false mechanism “may even beat a cold sham.” That sentence was false, and it was false in the one way that matters: Haller’s sham was not cold. Read the design, verbatim from the Methods:
“The sham protocol was designed to be credible but not specifically effective. Therefore, light touch was applied on standardized anatomic areas, equal to those treated with CST, for 2 minutes each time. In addition, body awareness instructions were given to simulate CST dialog techniques.” ✓ verbatim
“Standardized treatment protocols comprised 8 units of CST or sham treatment once a week lasting 45 minutes each.” ✓ verbatim
Line that up against our own hypothesis. Same duration. Same hands-on touch, on the same anatomy. Same warm practitioner. Same weekly ritual, eight times. Even the conversation was scripted in — body-awareness dialogue, deliberately built to simulate the CST talk. Haller held constant every single thing this lab said was doing the work — and then measured what was left.
What was left was not nothing.
| CST vs. the matched sham | Result — all verbatim from the primary |
|---|---|
| Pain intensity, week 8 (100-mm VAS) | −21 mm group difference; 95% CI −32.6 to −9.4; P = 0.001; d = 1.02 ✓ verbatim |
| Pain intensity, week 20 (3 months post) | −16.8 mm; 95% CI −27.5 to −6.1; P = 0.003; d = 0.88 ✓ verbatim |
| Reaching a minimal clinically important difference at week 20 (≥20% VAS reduction) | 78% of the CST group ✓ verbatim — and print the comparator, or the number is a lie by omission: the sham group hit it too, 51.9% (Table 4: CST 77.8% vs. sham 51.9%, P = 0.046, significant). More than half the sham arm cleared the MCID. |
| “Substantial clinical benefit” at week 20 (≥50% VAS reduction) | 48% of the CST group ✓ verbatim — vs. 25.9% of the sham group, and P = 0.091 — NOT statistically significant. n.s. — Haller’s own χ² does not call this a between-group win, and neither do we The 48% is a within-group figure. Standing alone under a “CST vs. sham” header it reads as a victory the trial did not report. This one overstated the case against our own thesis — and it comes down anyway. |
| The authors’ conclusion | “CST was shown to be specifically effective and safe in reducing neck pain intensity…” ✓ verbatim |
A d of 1.02 against a sham that already contained the 50 minutes. That is not a waitlist. That is not a cold sham. That is the exact control group our thesis demanded — and our thesis did not survive it cleanly. The house rule says if we only flag the other guy, we’re running the preacher’s play. So the flag goes here, on us.
We do not get to make this go away, and we are not going to try. But one trial is one trial, and these are real:
- n = 54. Single center. Haller’s own Limitations section, verbatim: “the sample size was relatively small and consisted of 81.5% of female patients, which may reduce the representativity and the generalizability of the results.” ✓ verbatim
- The therapists were not blinded — and they were not even the same people. Patients and outcome assessors were blinded; the practitioners delivering the treatment necessarily knew which arm they were in. And Haller says so himself, verbatim: “the comparability of the CST and the sham groups may be limited due to the allocation to the therapists. Whereas 3 therapists performed CST, only 1 therapist performed the sham treatment.” ✓ verbatim — this is the authors’ own caveat, not ours A three-versus-one therapist split is a live confound in a trial whose entire subject matter is what the practitioner brings into the room.
- The primary outcome is a subjective pain VAS. Which is precisely the axis on which Tab II showed placebo does its best work — the asthma patients tied the drug on the questionnaire and lost on the spirometer. Haller measured the questionnaire. That does not make the relief unreal (this lab has spent two tabs insisting it is real), but it means the finding lives on exactly the axis where we should expect it to.
- No replication. Not one. Haller’s own closing sentence asks for it: “Further studies with rigorous methodological designs and long-term follow ups are” needed. ✓ verbatim
None of that is a refutation. Every one of those objections could be raised about a trial whose result we liked, and we would call them nitpicking. They are open questions. They are not an answer.
The claim “it was the 50 minutes” is HELD OPEN on this page. It is not closed, and it is not settled in our favor. It is now the author’s contested hypothesis, sitting against one direct experimental test that came out the other way.
What would kill our hypothesis: a well-powered, multi-center, pre-registered replication of Haller — same matched-touch, matched-duration, matched-dialogue sham, therapists balanced across both arms — that reproduces a large specific CST effect. Do that and “it was the 50 minutes” is dead, and this lab will say so on this page in this box.
What would save it: the same replication coming back null, or the effect collapsing once the therapist allocation is balanced and the outcome measure is something you cannot talk yourself into.
As of this writing, neither experiment has been run. So we sit in the uncomfortable chair. A page about verification does not get to hold a hypothesis that survived only because nobody tested it — and then hide the one time somebody did.
One thing does hold, and it is the smaller, harder claim: whatever is in that room, nobody has shown it is the cranial rhythm. The palpation of it is unreliable (A2). The rate the schools teach does not match the rate the instruments record (A1). Haller demonstrated that something specific to the CST protocol outperformed a matched sham; he did not demonstrate what, and no rhythm was ever measured in that trial. the author’s reading — and the narrowest version he can defend You still have to correctly name what is in the bottle. We just have to admit that we do not currently know what is in it either.
A real mechanism, a population it couldn’t help, and a false explanation sold on top.
Design a maximum-strength placebo delivery system from scratch and you build a revival service: sustained touch, a supremely confident authority, ritual, emotional arousal, a crowd, music, public expectation, and a story that makes sense of your suffering. Every variable Kaptchuk isolated, cranked to the ceiling, all at once. So yes — some of those people genuinely felt better, genuinely walked, and were not lying.
William A. Nolen, MD — Healing: A Doctor in Search of a Miracle. Random House, 1974. publication year given as 1974/1975 across sources — unresolved primary book not independently loaded
Nolen, a surgeon, spent two years investigating faith healers. He attended a 1973 Kathryn Kuhlman service in Philadelphia, identified 23 people who claimed a cure, and followed them long-term. His finding: no patient with organic disease was cured.
A woman said to have been cured of spinal cancer threw away her brace and ran across the stage at Kuhlman’s command. Her spine collapsed the next day. She died four months later. recorded sequence corroborated via secondary (matches the Kuhlman record); the 1974 book itself was not loaded, so treat as reported, not verbatim-from-Nolen
She was not lying — she really did feel able to run. That is the whole thesis of this lab. As to cause: the record shows order in time (she ran, the spine collapsed the next day, she died four months later); the disease was advanced spinal cancer. We state the sequence and stop short of a proven causal claim beyond it. sequence, not a controlled cause
Nolen sorted conditions into functional (autonomic malfunction), hysterical (mind-induced), and organic (broken bone, cancer, gallstones), and held that the first two respond to faith-healing’s suggestion and the third does not. He also named the two confounds this lab has been circling — self-limited conditions (a cold goes away anyway) and cyclical ones (regression to the mean, named in 1974). His line: “being able to breathe deeply on stage is not evidence that lung cancer has been cured.” Nolen’s taxonomy and quoted lines Unverified — not confirmed from a loaded source
Somebody got there fourteen years earlier — and he was the chaplain.
Dahl, Robert A., A.B., B.D. — “A Brief History of Faith Healing.” Quarterly Bulletin of the Northwestern University Medical School 1960;34(1):64–71. Delivered 12 January 1960 as the second lecture in the 1960 Lecture Series on the Growth of Medicine. The author’s footnoted title: “Chaplain, Chicago Wesley Memorial Hospital.” ✓ Sourced — primary loaded, quotes verbatim
Not a skeptic. Not a debunker. A working hospital chaplain, standing in a medical school in 1960, describing the revival circuit — and landing the exact objection this lab is built on, fourteen years before Nolen went and checked:
“In very few instances have the case histories of the patients been recorded or authenticated in any way.” ✓ verbatim, p.69
That is the missing control group, named from the pulpit side of the aisle, in a medical-school lecture hall, in 1960. Nobody wrote the outcomes down. That is the whole game, and a chaplain said it first.
Dahl is sympathetic to faith healing, not skeptical of it. It would be very convenient to quote his two good lines and walk away. That is the preacher’s play, and we don’t get to run it. His own numbered conclusions, p.71, verbatim:
“There are a few examples of pure faith healing which must be reckoned with. They cannot be ignored. They should be studied and not tossed aside as unscientific.”
“The part which medicine or faith play in the healing of any particular person would be difficult, if not impossible, to determine.” ✓ verbatim
So our best 1960 witness against the revival tent is also a witness for taking faith healing seriously, and he explicitly says the medicine-versus-faith split cannot be untangled in an individual case. 🔴 Contested Both halves of the man go on the page or neither does. We took the halves we liked and we are showing you the ones we didn’t.
And then somebody ran the confound as an actual experiment.
Nolen named the self-limited-disease confound in 1974. Naming is not proving. In 2020 two researchers built it in a lab and watched it happen.
Blanco F. & Matute H. — “Diseases that resolve spontaneously can increase the belief that ineffective treatments work.” Soc. Sci. Med. 2020;255:113012. DOI 10.1016/j.socscimed.2020.113012. University of Deusto. ✓ Sourced — primary loaded, quotes verbatim
Three contingency-learning experiments. A fictitious disease, a fictitious medicine that was completely inert — zero effect, by construction, and the experimenters knew it because they built it that way. The only thing they varied was whether the disease got better on its own. Verbatim conclusion:
The authors go further, and it is the hinge of this entire lab: patients treating a self-limited disease with a pseudotherapy “would incorrectly attribute the improvement in symptoms to the effect of the treatment, rather than to the natural course of the disease” — which, in the authors’ hedge, could make self-limited illness “a particularly dangerous and prevalent opportunity for pseudotherapies to acquire undeserved trust and be used in the future for more serious health problems.” ✓ verbatim
That is Tab II’s substitution harm, with its fuse lit in Tab III. The cold that was always going to clear buys the credibility that gets spent on the cancer. And Blanco & Matute cite, for that exact step, the same JAMA Oncology paper this lab uses on Tab II (Johnson et al., 2018 — their reference list, verbatim: “JAMA Oncol. 4, 1375–1381”) — the two halves of this lab were already shaking hands in the literature before we got here. ✓ the citation is in their reference list
This is the experimental proof of the line this whole lab ends on. Hold that thought until Tab IV.
Nolen’s method was criticized and we don’t get to hide it: Lawrence Althouse said he attended only one service and didn’t follow up with everyone who claimed healing; Richard Casdorph published counter-evidence; Craig Keener wrote that “significant recoveries” are hard to dispute; Kuhlman’s NYT obituary noted a Johns Hopkins cancer researcher testifying to healings; a Tulsa World retrospective estimated ~2 million reported healings over her career. counter-claims Unverified — not independently loaded Sample of 23. One service. Contested methodology. And a woman whose spine collapsed the day after she ran across his stage. All of it is true at once. Print all of it.
The Nocebo
“Expectation runs backwards too.”
Without this tab the lab is dishonest. You cannot claim the upside of expectation without owning the downside. Same wiring. Reversed. And the strongest instrument in the whole lab is here, because it comes with an empty bottle.
Run it yourself. Then break the blind.
SAMSON (Self-Assessment Method for Statin Side-effects Or Nocebo) — Wood, Howard et al., Imperial College. NEJM 2020 (research letter); follow-up JACC 2021. ✓ Sourced — result table verbatim JACC DOI, mean age 65, “60% muscle aches”, and the Howard quotes not independently loaded
60 patients who had already quit statins over intolerable side effects took, over 12 months in randomized double-blind order, one bottle per month: 4 bottles of atorvastatin, 4 of placebo, and — the beautiful part — 4 empty bottles. Each day they logged a symptom score, 0–100.
Your 12 months. The bottles are blinded — you can’t see which is which. Open them one at a time and log the month’s symptom score. Then, and only then, break the blind.
The published trial: mean symptom intensity 16.3 on statin, 15.4 on placebo, 8.0 with no tablet. Both pill conditions significantly worse than no-tablet (p<0.001); statin vs. placebo, no meaningful difference. ✓ verbatim Your run draws from those means with month-to-month noise, so your numbers wobble — but the shape holds.
Howard’s framing: side effects from statins are very real, but they are mainly caused by the act of taking tablets, not by the statin inside them. Howard quotes Unverified — not independently loaded
StatinWISE says the same thing.
StatinWISE — BMJ 2021. 200 randomized double-blind n-of-1 trials in UK primary care; 151 analysed. PMID 33627334. ✓ Sourced — verbatim
No difference in muscle-symptom scores, statin vs. placebo: mean difference −0.11 (95% CI −0.36 to 0.14), p = 0.40. Withdrawals for intolerable muscle symptoms: 9% on statin, 7% on placebo. Two-thirds of completers intended to restart statins.
1. Selection bias, acknowledged by critics (Welty). SAMSON enrolled only people who developed symptoms within 2 weeks of starting. Non-drug side effects are often greatest in the first weeks and then abate (metformin→diarrhea, beta-blockers→fatigue — people adapt). The design may bias toward finding nocebo. Limited — a real caveat aimed at our own strongest tab
2. Statins are not side-effect-free. Rhabdomyolysis is real. In a meta-analysis of ~30 RCTs / 83,858 patients the counts were 5 cases among non-statin users vs 7 among statin users. this figure carries no author/journal/year in the source packet — Unverified citation Note honestly: 7 vs 5 out of 83,858 is numerically higher but cannot by itself establish “more common.” The underlying pharmacology is sound; these particular numbers are too small to prove it. Do not oversell this tab into “all drug side effects are fake” — that would be doing the exact thing this lab exists to stop.
The control group wasn’t an experimental nicety. It was the therapy.
SAMSON’s empty-bottle arm let patients see their own data. Because the design built in no-tablet periods, participants could see as clearly as the researchers could how powerful the nocebo effect was — and half of them happily restarted statins. ✓ “half restarted” verbatim
That is the Jenkins Method as medicine. Ground truth first. Then overlay the claim. Let the person see the contradiction themselves. It is the same move Wechsler’s no-intervention arm makes in Tab II — you just ran it, above, with your own hand on the bottles.
And it closes the loop on Tab III in the ugliest possible way.
Barnes et al. 2024 makes it clinical: practitioner warmth boosts placebo and blunts nocebo; coldness and blame do the reverse. Unverified — not independently loaded The person delivering your care is an active pharmacological variable in both directions.
Every informed-consent conversation, every drug label, every “this may cause dizziness” is a live nocebo dose. Medicine has to deliver it anyway. Nobody has solved this. We’re not going to pretend otherwise.
How to read the flags on this page.
| ✓ Sourced | Named study, journal, year, DOI/PMID; number checked against the primary or a verbatim quote. |
| 🔴 Contested | Sources genuinely disagree. Show both, name both, resolve neither. (Naloxone mechanism · JNCI prostate null · vertebroplasty · Nolen’s methodology · Dahl’s pro-faith-healing conclusions · Alvarez’s cranial-vs-cranial agreement · Rasmussen’s own “may be the same as the CRI” against this lab’s A1 · and Haller vs. this lab’s own “50 minutes” thesis.) |
| Limited | Real result, real caveat — small n, CI crossing zero, selection bias. The limit rides on the claim. |
| ⚪ Unverified | Could not be loaded and confirmed this pass, or has a sourcing problem. Flagged, not dropped and not promoted. (Moran & Gibbons’ ICCs · the “8–12 cycles/min” rate · Nolen’s quoted lines · the Kuhlman counter-literature.) |
Where the Contested flags fall matters: several land on our side of the argument — the prostate null, the naloxone counter-evidence, the vertebroplasty fight, Nolen’s critics, the 1960 chaplain who won’t dismiss faith healing. And the biggest one lands on the lab’s own headline framing: Haller ran a sham that contained the 50 minutes, and craniosacral therapy beat it anyway (d = 1.02). That claim is held open, not closed. If we only flagged the other guy, we’d be running the preacher’s play.
Promoted ⚪ → ✓ Sourced (primaries loaded): Wirth-Pattullo & Hayes 1994 (all three quotes + ICC = −.02) · Upledger 1977 (year settled; real figure .57) · κ = −0.970 (Alvarez Table 2, right fibular head, 0/44) · Haller 2016 (full effect sizes) · CRI rates 6–12 and 10–14.
Defects we opened against ourselves: the “cold sham” rebuttal to Haller was factually false and has been rewritten, not deleted · “the stated mechanism: false” was overstated and has been withdrawn in favour of a stronger, narrower claim · the “8–12 cycles/min” rate could not be sourced and now says so out loud.
One correction to the correction. The repair brief that drove this pass told us “almost perfect disagreement” was a fabricated quotation and ordered the quote marks removed. We went to the PDF. The brief was wrong. Alvarez writes, verbatim, in the Discussion: “The kappa coefficient suggests almost perfect disagreement (−0.970) between cranial physician 1 and the confirmatory group.” ✓ verbatim — the quote marks stay, and they are earned We do not remove a real quotation because an instruction told us to. The instruction gets checked against the paper too.
The pass above cleared every quotation. It did not clear the prose around the quotations. A cold verification — fresh eyes, builder not consulted, every figure matched against the primary — found three defects, and all three are the page’s own sentences, not its sources. They are printed here, named, and fixed.
- We put a ✓ Sourced flag on the wrong journal. This page said Blanco & Matute cite “the same JNCI paper this lab uses on Tab II.” They don’t. Their reference list reads “Johnson, S.B., Park, H.S., Gross, C.P., Yu, J.B., 2018… JAMA Oncol. 4, 1375–1381” — the JAMA Oncology paper, which this lab also uses on Tab II. Corrected in both places it appeared. The true version is the stronger one. A ✓ flag on a false attribution is the exact defect this page exists to prevent, and we shipped one.
- We invented a statistic. This page claimed Rasmussen’s measured rate had “most of its mass sits below 6.” No such figure exists in the paper. Rasmussen reports a mean of 6.16, a min of 4.25, a max of 7.07 and “a very narrow range… with few outliers in both ends of the scale” — no median, no distribution, nothing that supports the sentence we wrote. It was fabricated to make the argument land harder. Deleted, not replaced. The honest claim was always sufficient: the measured ceiling of 7.07 sits below Woods & Woods’ floor of 10.
- We ran the preacher’s play on our own best source. We mined Rasmussen’s Discussion for the two lines that helped A1 and left out the paragraph, on the same page, in which he says his measured rate is similar to a large palpation study and may be the same phenomenon the palpators call the CRI. That is the single strongest sentence against A1 in the literature we hold and it is in our own primary. It is now printed in the A1 box, in full, first, before our answer — and our answer concedes that A1 is narrower than the header we wrote.
Also corrected on this pass: Haller’s 78% and 48% were printed bare under a “CST vs. sham” header with no sham column — the comparators (51.9% and 25.9%) are now in the same breath, and the 48% is flagged not significant (P = 0.091), which is what Haller’s own Table 4 says. · Four page citations were off by exactly one, all in the same direction: Wirth-Pattullo 913→914 and 909→910, Rasmussen 27→28, Dahl 68→69. Every quote was verbatim; only the folio was wrong. Harmless to the argument. Corrosive to a page whose whole claim is that it checked.
What this means: two passes of this page carried a fabricated statistic and a false citation while wearing ✓ Sourced flags, and a third reader had to go to the PDFs to find them. The flags are only worth what the last check was worth. This is the third check. It is on the page.
The facts are sourced. The framing is a person’s.
What’s real
Every study, every number, every quotation carrying a “Sourced” flag: Levine, the 4–6 mg range, Amanzio & Benedetti, Wechsler’s asthma table, the JNCI/JAMA-Oncology cancer data including the prostate null and the higher-stage baseline, Moseley’s and FIDELITY’s verbatim lines, Kaptchuk’s 28/44/62, SAMSON’s 16.3/15.4/8.0 and the 0.90 nocebo ratio, StatinWISE.
Newly sourced from primaries this pass: Wirth-Pattullo & Hayes’ ICC = −.02; Alvarez’s κ = −0.970 (one table cell — right fibular head, 0/44); Rasmussen’s measured 6.16 cpm; Haller’s d = 1.02 against a matched sham; Blanco & Matute’s self-limited-disease experiment; Dahl 1960 — both halves. Anything marked ⚪ Unverified is exactly that — carried with the flag on, never promoted to fact. Moran & Gibbons’ −0.09 is still Unverified and is NOT the same number as the −.02.
What’s mine — and one of them is now on the ropes
Opathorlokan University; the four-tab spine; the framing lines — “it moves symptoms, it does not move disease,” “in a real asthma attack the placebo group dies feeling fine,” “that is a successfully administered nocebo,” “he just never built a control group.”
“It was never the cranial rhythm. It was the 50 minutes” is no longer filed as a harmless framing line. It is the author’s contested hypothesis — a falsifiable claim about mechanism, tested directly by Haller’s matched-duration sham, and it did not come through that test cleanly. It stays on the page because it is still arguable. It stays flagged because it is not a finding. Read Tab III before you quote it at anybody.
The SAMSON instrument is a teaching model that draws from the trial’s published means with added noise; your run is illustrative, not a re-analysis of patient data. The word “placebo” on the operating-room tab is an argument, not a diagnosis — and it is labeled as one.
You are not imagining it. You built the relief yourself, out of your own opioids — a dose between 4 and 6 mg of morphine, for the price of a sugar pill and a story. It is as real as anything in the bottle.
It just never touched the thing that is actually killing you.
And the man who told you it did — he wasn’t necessarily lying. He just never built a control group.
That last line is the author’s. The experiment behind it is not. Blanco & Matute built a disease that healed on its own and a medicine that did nothing, and watched people come to believe in the medicine anyway: “self-limited diseases can produce strong overestimations of effectiveness for treatments that actually produce no effect.” Soc. Sci. Med. 2020;255:113012. ✓ Sourced — verbatim You do not need a liar. You only need a disease that was going to end anyway, and nobody writing down what happened.