Something real was found in 1998. It was not about terpenes, it was not about the plant, and for years I stood on stages and described it wrongly. This is the version I can defend.
A real result exists. It is from 1998, it came out of the human body rather than the plant, and it had nothing to do with terpenes. The version the industry sells you, cannabinoids and terpenes in a flower working together, was proposed as a hypothesis in 2011 and has never been demonstrated at a dose a person can actually consume. At the amounts a jar delivers, almost every terpene is pharmacologically silent. What survives is narrower than the marketing and more useful: the profile predicts the effect, it does not cause it. Here is the arithmetic, and the place where I got it wrong.
The term was coined in 1998, but not for terpenes and not for plants. A team including Raphael Mechoulam found that fatty-acid compounds the body makes, compounds that do nothing on their own and do not bind the cannabinoid receptor, amplified the activity of 2-AG, one of the body’s own cannabinoids (Ben-Shabat, Fride, Mechoulam, 1998).
A zero, added to a one, produced more than one. 1 + 0 > 1.
That is not what the industry means. The industry means lots of stuff working together, more is better. Mechoulam meant something stranger: the compound that does nothing is the one that changes everything.
That result stands. It has never been challenged and I am not challenging it. What I got wrong was what it was a result about.
I have said, from stages, for years, that the 1998 magic was an inactive molecule amplifying an active one at a receptor. Then I read the paper properly, which I should have done a long time earlier. The sentence is: “2-Lino-Gl ... significantly inhibits the inactivation of 2-Ara-Gl.”
The “inactive” compound was competing at the degradative enzyme. It was not amplifying a signal. It was protecting the active one from being destroyed. Metabolic protection, not pharmacology. Those are different things, and I taught the wrong one.
It gets worse for the story I was telling. THC is not broken down by those enzymes. Mechoulam’s mechanism does not exist for THC, so you cannot invoke the 1998 route for a terpene and THC interaction at all. It is structurally unavailable. REFUTED as a plant mechanism.
So 1 + 0 > 1 survives as a real observation about the endocannabinoid system. As a mechanism for the plant, it is dead. Anyone who says otherwise, and that includes me before this year, has not read the paper.
The version everyone means, cannabinoids and terpenes in the plant, arrived thirteen years later, in Ethan Russo’s “Taming THC” (Russo, 2011).
Notice what kind of paper that is, because almost nobody who quotes it does. It is a review proposing a hypothesis. It argues that terpenes and minor cannabinoids might synergise. There is no new experiment in it. Russo himself framed much of it as “if proven.” HYPOTHESISED
That is not an attack on Russo. A good hypothesis is worth a great deal, and this one organised a field. It is an attack on thirteen years of labels and training decks that quoted a review as a law.
If you want the long version of that correction with the numbers laid out, it is one of four places I correct myself in the free booklet.
At the doses a human gets from a flower, terpenes are pharmacologically silent. Not weak. Silent. Here is the arithmetic.
Tested directly, common terpenes largely did not modulate CB1 or CB2, or change how THC acted on them (Santiago et al., 2019; Finlay et al., 2020). REFUTED for terpenes tuning THC at the receptors. The whole set holds one nanomolar affinity, beta-caryophyllene at CB2, Ki = 155 nM (Gertsch, 2008). One out of nineteen, and that one is a one, not a zero. It is in the band, not the entourage.
Now put the whole flower on one scale. THC is roughly 22 percent of the mass. Everything else ever proposed as interesting, every terpene, minor cannabinoid and flavonoid, is perhaps two to three percent: 20 to 30 milligrams across more than a hundred compounds.
In plain terms: the arithmetic says there is no minor molecule. There is THC, and there is a rounding error. And the rounding error is still tracking the effect.
Holding both halves of that at once is the honest position. To see where these numbers sit on a real jar, learn how to read a certificate of analysis.
Somebody finally ran the experiment. 30 mg of vaporised THC produces anxiety and paranoia. Add 15 mg of d-limonene and both drop significantly. Limonene on its own was indistinguishable from placebo, a true zero, measured. Double blind, in humans (Spindle et al., 2024). SUPPORTED, and the best evidence anyone has in this field.
Now look at the dose. 15 mg, which the authors themselves call “unlikely to be encountered in unadulterated cannabis flower.” Your joint has about 1 mg. The wall is a budget constraint on the jar. It says nothing about the molecule. Engineer the dose and something happens. Grow it and it does not.
Two cracks in that study, which I would rather hand you than let a critic hand you later. The authors credit limonene’s independent action, not modulation of THC. And the abstract says THC pharmacokinetics were unaltered, while the Results report plasma THC significantly higher in the condition that worked (p < 0.05).
Patients hear “nothing in the jar arrives” and panic about their 1:1. Do not. CBD is not in the rounding error. It sits at percent-level mass, tens of milligrams, in the payload beside THC, and it has a proven mechanism: it is a negative allosteric modulator at CB1.
A negative allosteric modulator, in plain words: a compound that never pushes the lever itself. It rests on the arm of the one that does, and softens the pull.
So the ratio is a cannabinoid-dose question, and the wall is about compounds under a milligram. Confusing the two is how a patient abandons a product that was working. Choosing by the numbers rather than by a name on a jar is most of what separates medical use from recreational use.
Everyone writes the entourage effect as a sum, and that addition sign has been doing damage. A sum can only go up. It makes “more compounds” sound like “more effect”, which is the marketing claim, and it cannot describe a compound that makes the effect smaller. But we have one. Limonene brought THC’s anxiety down. Nothing was added. Something was turned down. Write it as a product instead:
E = f(C) · M(T, S)
The felt effect is what the cannabinoid dose does, f(C), almost entirely THC, multiplied by a modulator M set by the terpene profile T and the person-state S. Terpenes belong in the coefficient. They are not a second drug. M = 1 means you feel the THC. Above 1 is the classic entourage story. Below 1 the profile makes the effect smaller, and that is the only direction we have clean human evidence for. The same profile can run above 1 in one patient and below 1 in another, because S differs. Which is why the profile predicts the effect and never causes it.
Hence the slogan I use now, replacing the one I used to: 1 + 0 ≠ 1. The zero changes the outcome, and the direction is an open question, not a promise. What M actually is I cannot tell you. Artefact, marker, smell, aerosol physics: four candidates, none of them shown.
When a pharmacologist backed me into a corner with the dose wall, I used my escape route: the profile is a fingerprint, and something alongside it does the real work. A flavonoid. It is a reasonable thing to say, and it is also a place to hide, so I am taking it away from myself in public.
Cannflavins A and B do inhibit mPGES-1 and 5-lipoxygenase (Werz et al., 2014), a real mechanism nowhere near CB1 or CB2, and nobody is required to measure them, which is exactly what makes them comfortable. But their dose gap is worse than the terpenes’, not better, a hundred to a thousand times short even if you ate the whole gram. UNKNOWN for what reaches an inhaling patient, because nobody has measured it. And they are built from the same tank that fills the trichome with terpenes (Rea et al., 2019), so they co-vary with terpene content. A confounded marker cannot rescue a marker.
I would rather write this down than be asked for it. Five experiments would move me:
None of the five has ever been run. Four of them are cheap.
When someone tells you the entourage effect is proven science, the honest answer is short. It is a real and useful idea that is not fully proven. It gives us a way to reason, and then we check what happens in you.
That distinction is the whole job, and it is what separates a coach a physician trusts from one they do not. A profile is a prediction, and a prediction is worth nothing until you check it against one person over several weeks. Write down the product, the numbers on its label, the dose, the time, and what came back. The free WIZDOM journal exists for exactly that, and one filled-in log outranks every paper cited on this page, for you.
Please read this part. This is patient education, not medical advice, and I am not a physician and not a prescriber. Nothing here is a recommendation to start, stop or change a product or a dose. Anything involving a lab value or a prescribed medicine goes to your prescriber, and if something is getting worse, that conversation belongs today rather than after another experiment.
Something real was found in 1998, but not in the plant and not with terpenes. Mechoulam and colleagues showed that inactive fatty-acid compounds the body makes protected one of the body’s own cannabinoids from being broken down. The plant version, cannabinoids and terpenes working together, was proposed as a hypothesis in 2011 and has never been demonstrated at the doses a person actually consumes.
Tested directly, common terpenes largely did not modulate CB1 or CB2, or change how THC acted on them (Santiago et al., 2019; Finlay et al., 2020). Only beta-caryophyllene holds real affinity, at CB2, Ki = 155 nM. So a terpene profile may still track the effect you feel, but probably not by grabbing THC’s receptors.
A survey of 107 retail samples found mean d-limonene of 0.11 percent, which is about one milligram in a one-gram joint. The human trial that showed limonene reducing THC anxiety used 15 milligrams, a dose its authors call unlikely to be encountered in unadulterated cannabis flower. The papers people quote sit 100 to 10,000 times above what a lung can deliver.
Because CBD is not in the rounding error. It sits at percent-level mass, tens of milligrams, in the payload beside THC, and it has a proven mechanism as a negative allosteric modulator at CB1. The ratio is a cannabinoid dose question. The dose wall is a statement about compounds under a milligram.
Extracts have behaved differently from pure THC (Baram et al., 2019), but “always” is not established. Often the fuller profile helps, and sometimes it does not. The honest version is to match the product to the person rather than to a slogan, and then check what actually happened in that person.
A human trial of a terpene ratio, not a single terpene, producing an effect the components do not produce alone. Or the same chemistry with the smell masked. Or a fixed inhaled THC dose with the profile varied while plasma THC is measured. None of those has ever been run, and most of them are cheap.
Week 3 of the WIZDOM Cannabis Coaching Masterclass is terpenes, the entourage effect, and reading real labels. Twelve weeks, thirty seats, built for aspiring coaches and for the nurses, pharmacists and physicians who want to use this in practice rather than read about it.
See the Masterclass