How Cannabis Impacts Health & the Potential Risks | Dr. Matthew Hill

The conversation begins with a public scientific disagreement. Andrew Huberman had posted a clip from an earlier cannabis episode, and Matthew Hill, a professor of cell biology and anatomy at the University of Calgary, criticized its discussion of cannabis and psychosis, differences between indica and sativa, and several other points. Huberman invited Hill onto the podcast to examine the evidence. Hill’s laboratory studies cannabis in relation to stress, feeding, and the behavioral effects of exposure at different developmental stages.

Huberman frames the interview as an updated scientific discussion and announces questions about THC, CBD, product labels, psychosis, hunger, memory and anxiety. The full conversation does return to these topics: Hill distinguishes acute psychotic reactions from an unresolved schizophrenia-causation debate, questions indica/sativa as chemical categories, and separates animal mechanisms and self-reports from clinical anxiety evidence. Those later qualifications are part of the interview, rather than topics absent from the source.

Hill describes cannabis as a plant with centuries of medicinal, spiritual, and recreational use across cultures. In his account, its mainstream popularity in the West grew during the 1960s and expanded further from the 1990s onward. Its biological complexity comes from the many molecules it contains, whose proportions differ across types of cannabis.

The central intoxicating constituent is delta-9-tetrahydrocannabinol, or THC. Hill identifies it as the principal driver of the plant’s psychoactive and intoxicating effects and says the amount present influences how high someone becomes when consuming it. He estimates that cannabis also contains roughly 70 to more than 100 other cannabinoids, most at trace levels.

Cannabidiol, or CBD, illustrates why structural similarity does not establish functional similarity. Hill says CBD looks chemically similar to THC but behaves very differently and is not intoxicating. He is less categorical about whether it should be called psychoactive: people may not be able to tell whether they have taken it, yet reported effects on anxiety or mood could justify that term under a broader definition. This is a qualification about terminology and reported effects, rather than a demonstration of treatment efficacy.

Other cannabinoids, including cannabinol and cannabigerol, remain much less understood in Hill’s account. They might modify THC’s effects, or they might not. Cannabis also contains terpenes, volatile compounds found in many other plants. Hill associates limonene with citrus-like smells or flavors and pinene with an earthy, tree-like smell, and also names beta-caryophyllene and myrcene. Some terpenes have known biological activity, but their contribution to cannabis intoxication remains uncertain.

The proposed interaction among these constituents is often called the entourage effect. Hill explains the idea by contrasting nearly isolated THC, such as that found in some distillate pens, with THC accompanied by other plant molecules. He describes emerging experiments that add one terpene or minor cannabinoid to THC and measure the result. These studies offer a way to test individual interactions, but the number of possible combinations means that a comprehensive account could take decades. The phrase names a hypothesis about interactions; this excerpt does not establish how all those interactions work.

Asked to explain what being high actually means, Hill describes a collection of experiences rather than one uniform state. Cannabis can produce euphoria or a positive mood, although he distinguishes this from the kind of elevated state associated with cocaine or other stimulants. People may find things funnier, change their eating behavior, and perceive environmental stimuli differently. Some may also experience a degree of dissociation or feel somewhat outside their bodies.

Hill also characterizes cannabis intoxication as potentially introspective: a person may become more aware of internal bodily feelings while processing external information or directing attention differently. This combination helps explain why a simple description of heightened happiness captures only part of the experience.

In laboratory settings, researchers often operationalize these subjective experiences with visual analog scales, such as zero to 100 or one to 100. Participants rate whether they feel high, enjoy the experience, feel euphoric, or notice elevated mood. Hill notes that people receiving placebo cannabis can still report feeling somewhat high when they believe they have received cannabis. Such reports show that subjective ratings can reflect expectations as well as the administered substance.

Huberman asks whether this placebo response also occurs in people who have never used cannabis. Hill does not claim to know. He believes these drug studies generally require some prior experience, perhaps only once or twice, but explicitly says he does not run human clinical laboratory studies and cannot state the enrollment rules definitively. Consequently, the exchange does not establish whether a person with no previous exposure would report the same expectancy effect.

Time perception supplies a concrete example of altered experience. Huberman asks about the impression that a long interval has passed when little time actually has. Hill says temporal dilation does seem to occur: an intoxicated person may report that more time has elapsed than actually has. He does not regard the mechanism as well worked out.

Hill tentatively recalls older animal experiments involving temporal ordering, possibly in pigeons or rodents. Animals can learn to wait a particular interval—for example, 10 minutes—before performing a rewarded behavior. According to his recollection, cannabinoids change their responding on such tasks, suggesting altered timing behavior beyond human verbal reports. But he cannot confidently identify the species or specific work, says he is not especially familiar with it, and cannot recall modern studies. The example points toward an experimental approach rather than supplying a settled mechanism.

Huberman next asks about appetite stimulation, including its use among people whose cancer treatment has reduced their appetite. He proposes blood sugar, insulin, or glucose regulation as possible explanations for the recreationally familiar munchies. Hill pauses that question to explain cannabinoid signaling first. Hill returns to the appetite question later in the conversation. He describes cannabinoid receptors in hypothalamic feeding circuits, reduced inhibition of AgRP neurons, and interactions with reward and sweet-taste processing. These later explanations should not be confused with confirmation of Huberman’s initial blood-sugar hypothesis.

Hill says THC produces almost all its effects primarily through cannabinoid type one receptors, abbreviated CB1. He uses the familiar lock-and-key analogy: a receptor is a protein on a cell, and a binding molecule such as THC can activate it and trigger processes that alter the cell’s activity. CB1 is widely distributed in the brain; Hill describes it, with qualification, as one of the most widely expressed receptors, possibly the most widely expressed, apart from ion channels.

The existence of those receptors does not mean humans evolved to use cannabis. Hill attributes the overlap between plant molecules and human biology to the reuse of biological mechanisms, and he challenges the inference that a plant must be safe because it is natural. The receptors have functions involving molecules the body produces itself: endocannabinoids.

Hill distinguishes CB1 from cannabinoid type two receptors, or CB2. He describes CB2 as found mainly on immune cells, including macrophages, with expression in some brain immune cells and perhaps a limited distribution on neurons. Its principal role appears more closely related to regulating inflammation. In discussing cannabis effects on the central nervous system and behavior, however, he says the focus is almost entirely on CB1.

To explain endocannabinoids, Hill first describes conventional neurotransmission. A presynaptic neuron—neuron A—releases a chemical across a synapse to neuron B. Depending on the transmitter and its action, this can increase or decrease the receiving neuron’s activity. Endocannabinoids can reverse that direction: neuron B makes them on the postsynaptic side, and they travel backward to regulate transmitter release from neuron A.

Hill likens this retrograde signaling to a thermostat. If neuron A releases an excitatory transmitter such as glutamate and neuron B becomes too excited, neuron B can release endocannabinoids that signal back to reduce the incoming drive. He describes maintaining homeostasis as the primary physiological role that, in his view, most cannabinoid researchers would agree on.

Huberman connects this balancing function to CB1’s broad distribution: neural circuits need excitation and inhibition without either running away. He invokes epilepsy as an example of excessive excitation and impaired thinking or movement as consequences of excessive inhibition. Hill describes endocannabinoids as a circuit breaker that gates incoming signals, explaining that their mechanisms can suppress presynaptic electrical activity and transmitter release.

That feedback does not always reduce a circuit’s overall activity. Endocannabinoids can also reduce the release of inhibitory neurotransmitters. Suppressing an inhibitor removes restraint, which Hill compares to taking a foot off the brakes while a car moves downhill. The result therefore depends on which neuron or terminal is affected.

Hill connects this flexibility to synaptic plasticity. Reducing inhibition can help a synapse strengthen during learning, while acting directly at an excitatory terminal can weaken a synapse and support adaptive flexibility. In his account, the same signaling system can regulate excitation and inhibition differently according to its cellular target. Calling it homeostatic therefore does not imply that every action simply quiets the brain.

Hill identifies an exception to CB1’s broad neuronal distribution: dopamine neurons, which he says do not appear to express cannabinoid receptors, at least as characterized so far. Huberman asks whether this means cannabis-associated euphoria is independent of dopamine and instead relies on opioid or serotonin systems.

Hill rejects the inference that cannabinoids therefore do not affect dopamine. In the ventral tegmental area, dopamine neurons involved in motivation receive inhibitory input from other neurons. Those regulatory neurons are rich in cannabinoid receptors. Activating their receptors can reduce inhibitory control, allowing dopamine neurons to become more prone to burst firing and large dopamine releases.

Hill compares this indirect arrangement to how mu opioid receptors operate in the effects of morphine or heroin. The comparison concerns a circuit mechanism: reducing inhibition of dopamine neurons. It does not establish that the substances have identical effects, nor does the excerpt demonstrate that dopamine alone explains cannabis euphoria.

Dr. Matthew Hill describes endocannabinoids as regulators of neural flexibility: they help circuits adjust their activity to experience and environmental demands, supporting plasticity and homeostasis. Although cannabinoids influence dopamine transmission, he says they generally do so indirectly. He finds this unusual because, in his account, other neurotransmitter systems receive much more direct endocannabinoid regulation. Why dopamine is relatively spared remains an open question, and Hill does not offer an evolutionary explanation.

That influence on dopamine does not establish an explanation for cannabis-related pleasure. Hill says nobody has cleanly demonstrated whether the dopamine effects account for positive affect or euphoria, and he emphasizes that dopamine’s relationship to reward and motivation is complicated. This distinction sets up a recurring theme of the discussion: identifying a receptor interaction does not, by itself, explain a subjective experience.

Hill contrasts the organization of endocannabinoid signaling with serotonin and dopamine. In his simplified comparison, serotonin and dopamine each use one transmitter molecule whose actions diversify across receptor types. He loosely estimates serotonin has 15 or 20 receptors and says dopamine has at least five. Endocannabinoids instead act backward across the synapse, through retrograde signaling, with two principal molecules regulating the receptor under discussion. Their functional diversity therefore occurs partly at the level of the signaling molecule. Hill allows that additional bona fide endocannabinoids could exist.

The two principal molecules are anandamide and 2-arachidonoylglycerol, usually shortened to 2-AG. Hill explains that anandamide’s name comes from the Sanskrit word anand, meaning bliss. Rafi Mechoulam, who also isolated and discovered THC, chose the name to evoke inner bliss while referring to the molecule’s amide bond. Huberman asks whether the name implies personal experience with anandamide. Hill does not establish that; instead, he points to Mechoulam’s long involvement in cannabis research and characterizes the naming as somewhat tongue in cheek.

Hill identifies a landmark Mechoulam THC paper from 1964 and raises a possible connection between its publication date and the cannabis-associated reference to 4/20. He gives the date as April 20, 1964, but explicitly says he does not know whether the story is true. Huberman calls it a potential myth that the conversation may help propagate. Separately, Hill dates Mechoulam’s laboratory isolation of anandamide to 1990 and says 2-AG was discovered a few years later, through work by both Mechoulam and a Japanese group. These historical details are presented as Hill’s account, with the 4/20 connection expressly unverified.

The biological distinction between the molecules depends on separating affinity from efficacy. Hill describes anandamide as having high affinity but low efficacy: relatively little is needed to bind the receptor, yet its capacity to stimulate a biological response reaches a ceiling fairly quickly. By comparison, 2-AG has lower affinity and requires a higher concentration to bind, but once bound it produces a much stronger intracellular signaling response. Easy binding and strong activation are therefore different properties.

Hill proposes a useful, provisional division of labor. Anandamide may provide tonic regulation, with ongoing production and release helping keep a circuit’s resting activity within a desired range. Huberman recognizes this as a thermostat analogy, which Hill endorses. In contrast, 2-AG may provide more phasic regulation, arriving when activity demands a rapid adjustment. Hill says much of the endocannabinoid work involving neuronal recordings and synaptic plasticity is effectively work on 2-AG.

A seizure supplies Hill’s extreme example of demand-driven regulation. During very high neural activity, heavily activated neurons release large amounts of 2-AG, helping rapidly reduce the circuit’s activity. He also attributes most of the rapid endocannabinoid signaling involved in strengthening or weakening synapses in response to experience or environmental change to 2-AG. The seizure example illustrates a proposed regulatory mechanism; it is not presented here as evidence for a cannabis treatment.

Hill cautions that the tonic-versus-phasic distinction is an umbrella model rather than an absolute separation. Some evidence suggests 2-AG regulates circuits at rest, while anandamide can participate in plasticity. Both ultimately regulate neurotransmitter release through retrograde signaling, and the stimuli that recruit each remain incompletely understood. The clearer point, he says, is that 2-AG production depends on neuronal activity: as a neuron becomes more active, it makes 2-AG to regulate its inputs. Huberman notes that other neurotransmitter and neuromodulator systems also combine ongoing activity with transient increases, but the endocannabinoid system appears to achieve this in a distinctive way.

Hill describes cannabinoid receptors and endocannabinoid molecules as widely distributed across the brain, including the cortex, hypothalamus, striatum, hippocampus and cerebellum. He identifies a consequential exception: relatively little receptor expression in the brainstem cell populations responsible for unconscious cardiac and respiratory regulation.

He uses that distribution to explain a difference between cannabinoids and opiates despite similarities in their receptor signaling mechanisms. In his account, opiate receptor activation in cardiorespiratory brainstem regions can depress neural activity, allowing loss of consciousness to accompany suppression of breathing and cardiac regulation, with potentially fatal consequences. Cannabinoids do not produce the same suppression through those regions, he argues, because the relevant receptors are largely absent there.

Hill invokes the saying that no one has died from a cannabis or THC overdose, but qualifies his mechanistic claim: a fatal overdose in the same manner as opiates does not seem physically possible with cannabinoids as far as observations have shown so far. He also explicitly acknowledges that intoxicated behavior can result in death. His comparison concerns a particular overdose mechanism and does not establish an absence of other cannabis-related harms.

Turning to cannabis-associated appetite, Hill describes several interacting mechanisms rather than a single hunger switch. Cannabinoid receptors occur in hypothalamic feeding circuits. By regulating inhibitory inputs around AgRP neurons, cannabinoids can disinhibit those neurons, increase their activity and promote food seeking. Hill also mentions POMC neurons and multiple feeding-related processes in the arcuate nucleus. Huberman connects this discussion to an earlier interview with Zach Knight of HHMI and UCSF about hypothalamic neurons involved in food seeking, craving and eating.

Reward circuitry provides another route. Hill cites work by Steve Muller and Kent Berridge in which placing anandamide in the nucleus accumbens stimulated consumption of palatable food. He also recalls a Japanese study in PNAS from approximately 12 years earlier: after researchers gave rodents a cannabinoid, stimulation of sweet-responsive taste-bud populations produced an enhanced gustatory cortex response. He says the effect did not appear for salty, bitter or sour stimulation, and he is unsure whether umami was tested. This experiment concerns sensory responses in rodents, while the nucleus accumbens work concerns food intake; together, they suggest more than one pathway through which cannabinoids can affect eating.

For Hill’s laboratory, a central behavioral question is why eating resumes after fullness. He notes that satiated humans and animals can begin eating again after cannabis exposure. That observation could reflect disruption of satiety detection, interference with reward devaluation, or both. Reward devaluation is the process through which a food loses appeal as it is consumed: Hill illustrates it with pizza that initially tastes delicious after a day without food but seems greasy and unpleasant after five slices. The issue is therefore not simply whether food becomes attractive, but whether its attractiveness declines normally during eating.

Hill describes rat experiments using airtight vapor chambers that repeatedly fill with cannabis-extract vapor and then clear. Exposure lasts 15 minutes. He says his laboratory has titrated the procedure to produce THC blood levels matching those attainable in a person smoking cannabis. Following exposure, rats eat plain chow as well as fatty or sweet foods. Rats satiated beforehand resume eating, and animals required to press a lever for access to food also work to obtain it. Matching blood levels helps characterize the exposure, but the behavioral experiments he describes remain animal studies.

When Huberman jokes that the animals demonstrate hard work under cannabis, Hill limits the conclusion to working for food. He cites similar work by Elise Wirtz and Cassie Moore at Hopkins using progressive-ratio tasks, in which the number of lever presses needed for a reward rises. His examples move from one press to two, four and eventually 16, with the requirement increasing exponentially. He also describes one female rat in his laboratory making 300 lever presses for a single sugar pellet after cannabis exposure. That individual example illustrates the behavior; the broader argument is that food remains motivating enough to justify increasing effort.

Another experiment devalues food through conditioned taste aversion. Researchers pair a food with a nausea-producing substance such as lithium chloride, after which animals ordinarily avoid that food. Hill says THC can make animals resume eating food devalued this way, as it can after satiety. He interprets these findings as support for the possibility that cannabis preserves food’s reward value despite conditions that would usually reduce it. The responsible circuitry remains unresolved. He raises reward circuitry and the orbitofrontal cortex as possibilities, rather than claiming his laboratory has located the mechanism.