Human Behavioral Pharmacology of Cannabis: Insights on Cannabis Use Disorder

Abstract

Objective:

Cannabis policy, public perception, and consumption patterns have undergone enormous shifts, yet the scientific understanding of these changes has lagged. Our objective is to describe how human behavioral pharmacology studies provide a rigorous framework for modelling Cannabis Use Disorder (CUD) and for informing the development of treatment approaches.

Methods:

This narrative review synthesizes insights from over 25 years of human laboratory research, highlighting how behavioral pharmacology, e.g., dose, placebo control, behavioral specificity, inform the development of evidence-based therapies and targeting the behavioral drivers of CUD.

Results:

Key takeaways: (1) Cannabis potency is not Δ⁹-tetrahydrocannabinol dose. Participants adjust inhalation patterns as a function of cannabis strength. (2) Placebo responding is robust even among experienced cannabis smokers; placebo-controlled studies are limited by regulatory barriers but are essential to distinguish pharmacological effects from expectancy. (3) How medications alter CUD-relevant behaviors (reinforcement, misuse-related ratings) and non-specific behaviors (e.g., sedation) portend clinical efficacy and tolerability and vary as a function of participant phenotype.

Conclusions:

Rigorous, placebo-controlled human laboratory studies of the behavioral effects of cannabis are essential to informing a quickly evolving field complicated by the complexity of cannabis as a plant with a myriad of active constituents, regulatory hurdles, and industry narrative.

1. Introduction

Over the past two decades, the United States has experienced substantial shifts in cannabis policies, public attitudes, and patterns of use. More than half of Americans reside in states where recreational cannabis use is legal, and these changes coincide with an increased prevalence of both cannabis use and Cannabis Use Disorder (CUD) among adults (Jayawardhana et al., 2025). There has also been an explosion in the availability of new cannabis products with limited regulation or understanding of their effects on public health and CUD risk (NASEM, 2024). Cannabis has a highly unusual status as drug that has been ‘legalized’ within individual states while remaining a federally illegal Schedule 1 drug with concomitant regulatory restrictions for empirical study (Haney, 2020). The disconnect between policy, commerce, and research has generated a persistent gap between empirical evidence and industry narratives, complicating efforts to characterize the determinants, prevention, and treatment of CUD, as well as the potential therapeutic value of cannabis and its constituents.

As part of this Special Issue on Behavioral Pharmacology, our objective with this narrative review is to highlight key insights from placebo-controlled, human laboratory studies administering cannabis or its constituents that leverage the principles of behavioral pharmacology to inform our understanding of cannabis, CUD, and its potential pharmacological treatment (e.g., Cooper and Haney, 2009; Haney et al., 2008, 2015, 2016, 2023). Specific behavioral pharmacology topics include misuse liability, positive and negative reinforcement, dose dependence, placebo control, tolerance, and behavioral specificity

2. Behavioral Pharmacology Framework

Behavioral pharmacology, an approach to precisely measure the effects of controlled drug administration within the context of environmental contingencies, offers key insight into understanding and treating substance use disorders. As recently described by Witkin and colleagues (2025), the field was developed in the 1950s by Peter Dews using the principles of B.F. Skinner to provide precision in studying the behavioral effects of drugs, environmental stimuli, and their interaction.

Using a behavioral pharmacology approach to assess the effects of cannabis, a plant containing hundreds of chemical constituents, presents unique challenges. Herein, we provide a framework essential to consider when analyzing cannabis’s behavioral effects:

2.1. Cannabis and its Constituents.

Cannabis is a hemp plant containing over 100 unique components or phytocannabinoids (ElSohley et al., 2017), of which Δ⁹-tetrahydrocannabinol (Δ⁹-THC) and cannabidiol (CBD) are the most well-studied in comparison to placebo. Δ⁹-THC, a partial agonist at the cannabinoid type 1 (CB1) receptor, produces classic cannabis effects, including increased appetite and positive subjective and reinforcing effects (Huestis, 2007; see Cooper and Haney, 2008). CBD, by contrast does not readily bind to the CB1 receptor and does not produce classic cannabis intoxication. The pharmacology of CBD is complex; it binds to multiple receptors, including serotonin (5-HT1A) and transient receptor potential vanilloid receptors (Ibeas Bih et al., 2015), and it can enhance, reduce or have no effect on Δ⁹-THC intoxication, depending on dose, route of administration, and individual cannabis use history (e.g., Solowij et al., 2019; Haney et al., 2016).

In addition to Δ⁹-THC and CBD, there are numerous minor cannabinoids (e.g., cannabigerol, cannabinol, tetrahydrocannabivarin) typically present in low levels in the cannabis plant but which could potentially alter specific effects of THC or produce effects in isolation. For example, relative to placebo, oral cannabinol (300 mg) improved objective measures of sleep in patients with insomnia (Lavender et al., 2026), and oral cannabigerol (20 mg) reduced ratings of anxiety in healthy volunteers undergoing a laboratory stress procedure (Cuttler et al., 2024).

The plant also comprises hundreds of non-cannabinoid compounds, such as flavonoids, which contribute to its color, and terpenes (e.g., limonene, pinene, myrcene, β caryophyllene; ElSohly et al., 2017), which give cannabis its distinctive flavor and aroma and which may also contribute psychoactive or therapeutic effects (Banister et al., 2019; Laaboudi et al., 2024). One example: d-limonene (1, 5 mg) inhaled with vaporized Δ⁹-THC (15, 30 mg) reduced Δ⁹-THC-induced anxiogenic effects in a dose-related manner relative to placebo (Spindle et al., 2024).

For behavioral pharmacology, a discipline defined by precision, characterizing the behavioral effects of a plant with a myriad of potential active components is complex. There are over 700 plant varieties or chemovars of cannabis, and an almost infinite combination of cannabinoid and non-cannabinoid constituents. Although these chemovars may vary in their psychoactive or therapeutic effects, disentangling why is difficult. The exceedingly limited placebo-controlled clinical data on individual cannabinoids or other plant constituents, alone or in combination, stand in stark contrast to the expansive therapeutic claims of a billion-dollar cannabis industry, which has incentives to conflate even in vitro data for commercial interest (Cogan, 2020). In the United States, extensive regulatory barriers to conducting human studies with Schedule 1 compounds further the disconnect between empirical evidence and industry marketing (Haney 2020, Vandrey 2018). There is a critical need for a scientific contribution to a narrative almost entirely dictated by a poorly regulated industry with an obvious motivation to exploit even meager evidence to their advantage.

2.2. Dose-Response Relationship.

Characterizing a dose-response relationship is fundamental to understanding a drug’s behavioral profile, yet dose is complicated to define for cannabis even when focusing on Δ⁹-THC. The potency of cannabis products available, defined by the percentage of Δ⁹-THC, has steadily increased over the past several decades (ElSohly et al., 2016). Yet, when evaluating the behavioral effects of smoked or vaped cannabis, potency is not equivalent to Δ⁹-THC dose. In fact, demonstrating ‘dose dependence’ for different cannabis potencies is difficult, even when attempting to control for smoking topography because experienced cannabis consumers adjust their inhalation patterns to achieve a desired subjective effect. For example, we guided research participants through smoking 3-puffs of cannabis (0.0, 1.8, 3.6% Δ⁹-THC) using a standardized paced-puff smoking procedure (Foltin et al., 1987), where we controlled inhalation duration (5-seconds), time spent holding smoke in the lungs (10-seconds), and inter-puff interval (40-seconds). We found that the two active cannabis conditions resulted in comparable plasma Δ⁹-THC levels and subjective effects, because participants inhaled more in the 1.8% condition relative to the 3.6% condition (Cooper and Haney, 2009). Another study attempted to test four ‘dose’ conditions: participants took 6-puffs of cannabis, but we varied how many puffs had active vs placebo cannabis: the goal was to have four Δ⁹-THC dose conditions: 0, 18, 35, 53 mg Δ⁹-THC. Again, the study found that cannabis effects did not significantly vary as a function of active cannabis condition; levels of expired carbon monoxide, an indication of inhalation strength, showed that participants dose-dependently inhaled less cannabis as the amount of active cannabis increased (Ramesh et al., 2013).

In order to get a better estimate of Δ⁹-THC dose (mg), several recent studies have guided participants through smoking a defined portion of a cannabis cigarette (800 mg) using the paced-puffing procedure, so that even if inhalation patterns varied as a function of cannabis strength, we could ensure that all inhaled a comparable amount of Δ⁹-THC. Fig. 1, portraying peak ratings of ‘High’ as a function of estimated Δ⁹-THC dose (mg) shows that dose-dependent cannabis effects can be observed using these procedures.

Fig 1.

Peak ratings of ‘High’ as a function of delta-9-tetrahydrocannabinol (mg)

Notes: Data points were compiled from published studies where the amount of cannabis inhaled was controlled (Cooper & Haney, 2010; Cooper, Comer, & Haney, 2013; Cooper et al., 2018; Haney et al., 2015; Haney et al., 2016; Kearney-Ramos et al., 2022). Maximum score for visual analog ratings of ‘High’ = 100 mm. Pearson correlation analysis was conducted between Δ⁹-THC (mg) and Peak ‘High’ ratings. A significant positive correlation was observed (r = 0.85, t(12) = 5.47, p< 0.001, 95% CI [0.57, 0.95]), indicating that Δ²-THC levels were positively associated with ‘high’ ratings. Abbreviations: Δ⁹-THC, delta-9-tetrahydrocannabinol; CI, confidence interval.

In the United States, federally funded researchers have been restricted to obtaining cannabis from a single source, and the products available to study scientifically are lower potency than what is available commercially. For a study testing whether a medication reduced cannabis intoxication, it was important that the participants, who on average smoked >2.4 grams cannabis/day, achieved a robust ‘high,’ so we guided them through smoking two cannabis cigarettes (7% THC), and their ratings of ‘good cannabis effect’ were 75% of the maximum achievable rating (Haney et al, 2023). This again illustrates that cannabis potency is not equivalent to Δ⁹-THC dose. Although federally funded researchers certainly need access to higher potency cannabis products, individuals who smoke a substantial amount of cannabis daily still report robust positive subjective effects from low potency cannabis when sufficient quantities are administered.

2.3. Participant Phenotype.

When studying the behavioral effects of controlled drug administration, a goal is to minimize confounds, i.e., factors that affect behavior but that are unrelated to the drug per se. For example, we have developed a laboratory model of CUD to assess the effects of medications alone and in combination with cannabis across a range of outcomes that define CUD: positive cannabis reinforcement (self-administration under non-abstinent cannabis conditions), negative reinforcement (self-administration following several days of abstinence), measures of misuse liability (e.g., good effect, liking, high), mood and physical symptoms of associated with cannabis withdrawal (e.g., irritability, disrupted sleep and food intake), and non-specific effects that may impact medication tolerability (e.g., cognitive task performance, sedation). We enroll individuals who smoke cannabis regularly (e.g., ≥6 days per week; ≥1 g of cannabis per day), who are not seeking treatment for their cannabis use, and who do not have major psychiatric diagnoses and/or substance use disorders (other than CUD). Mood, sleep, cannabis self-administration, and food intake are primary behavioral outcome measures, and baseline fluctuations in any of these endpoints would impact our ability to isolate the behavioral effects to cannabis.

Although we exclude participants who meet criteria for substance use disorders other than CUD, we have not excluded individuals who smoked tobacco cigarettes and who may have met criteria for Tobacco Use Disorder. Participants who smoked tobacco cigarettes were permitted to do so in the laboratory, as we did not want to introduce the confound of nicotine withdrawal. However, in a secondary analysis, we discovered that tobacco use is, in fact, a significant confound when assessing cannabis-related behavioral outcome measures. The goal of this secondary analysis was to see if we could identify factors that predicted who would ‘relapse’ to cannabis use in the laboratory, operationalized as a return to cannabis self-administration, at a financial cost, following days of cannabis abstinence. We found that those individuals who smoked both cannabis and tobacco cigarettes were far more likely to relapse to cannabis than non-cigarette smokers (odds ratio: 19.02; 95% CI: [2.18–165.95], p < 0.01); Haney et. al., 2013). These data show that even among participants who were homogenous in their patterns of cannabis use, concurrent tobacco cigarette smokers had a markedly distinct response to our primary behavioral endpoint: cannabis self-administration.

It is also essential to carefully consider how much cannabis participants use in the natural ecology, as tolerance to cannabis’s behavioral effects will directly impact outcome and thereby introduce a confound. As an example, in studies asking whether there are sex differences in the behavioral effects of cannabis, we were careful to match men and women on their patterns of cannabis use. If men used more cannabis than women and had a reduced behavioral response to cannabis than women, one could not determine if this was due to sex or due to tolerance to cannabis’s effects (Cooper and Haney, 2014; Lake et al., 2023).

Naturalistic patterns of cannabis use also may affect how medications interact with cannabinoids. A study investigating how the opioid antagonist, naltrexone, influences the subjective effects of oral THC found that naltrexone (12 mg) blunted THC intoxication in regular cannabis smokers but enhanced intoxication in non-cannabis smokers (Haney, 2007). Chronic cannabis use alters not only the endogenous cannabinoid system but also the endogenous opioid system (see Haney, 2007) and a range of other neurotransmitter systems, illustrating the importance of controlling for cannabis use frequency when investigating cannabis effects either alone or in combination with medications.

This study, using cannabis use frequency as an independent variable, also demonstrates how researchers can design studies to interrogate the influence of specific phenotypes. In another example, we conducted a pilot study comparing the effects of inactive and active cannabis in individuals matched for cannabis use frequency but differing in their risk for developing a psychotic disorder, i.e., a clinical high-risk population vs healthy controls. We found that relative to inactive cannabis, active cannabis produced significant increases in anxiety and paranoia in the high-risk participants but not the healthy sample; both groups had similar ratings of intoxication (Vadhan et al., 2017).

A final comment about participant phenotype: Although behavioral pharmacology studies designed to test potential treatment medications for CUD attempt to reduce the influence of confounds by limiting the heterogeneity of those enrolled, patients presenting for CUD treatment vary in their patterns of cannabis, other drug use, and psychiatric co-morbidities so it is reasonable to question the generalizability of behavioral pharmacology findings. Yet for the development of pharmacotherapies, laboratory models need not mimic all aspects of the natural ecology to predict behavior in the natural ecology (Haney and Spealman, 2008), and there is considerable consistency in the effects of medications on cannabis use in either setting (see Arout et al., 2019; Brezing and Levin, 2018). For example, oral Δ⁹-THC (dronabinol) decreased symptoms of cannabis withdrawal without significantly reducing cannabis self-administration in the laboratory or cannabis use in the clinic. Although the predictive validity of the laboratory model of CUD cannot be confirmed in lieu of a medication proven efficacious to treat CUD, the consistent correspondence between the two settings suggest that behavioral pharmacology studies provide a critical understanding of the effects of a medication alone (e.g., its potential for abuse, effects on mood, physical symptoms, cognition, sleep), as well as on how the medication alters the positive and negative reinforcing effects of cannabis (Arout et al., 2019)

2.4. Placebo control.

Controlling for the effect of expectation of a drug effect by using a placebo condition is an essential principle of behavioral pharmacology. When research participants are blind to cannabis strength and are told that they will be smoking cannabis with potencies ranging from weak to strong, even near-daily cannabis smokers demonstrate a mild, time-dependent intoxication from placebo cannabis (e.g., Haney et al., 2015, 2016). In fact, a subset of individuals (>15%) reliably pay their own money to self-administer placebo cannabis (e.g., Haney et al., 2015; Lake et al., 2023). Thus, when even highly experienced cannabis smokers are told that they might receive a range of cannabis potencies to smoke, they demonstrate a reliable placebo response.

This expectation of an effect is precisely why a placebo cannabis condition is essential when testing the potential therapeutic effects of cannabis. In an 8-week, placebo-controlled pilot study testing the effects of cannabis capsules (100 mg CBD: 5 mg Δ^{9 −}THC) for chemotherapy-induced neuropathic pain, participants maintained on placebo capsules requested dose reductions to the same degree as those maintained on active cannabis, demonstrating that double-blind dosing can be achieved with proper instruction. In this case, we told participants that they could receive a range of cannabis conditions, including a CBD-only condition and therefore they might not feel any intoxication even if receiving an active medication.

This pilot study demonstrated that both the placebo and active cannabis conditions had time-dependent reductions in pain and neuropathy ratings, with active cannabis not out-performing placebo (and in fact, in some conditions, worsening outcomes relative to placebo). Changes from baseline ratings are insufficient for demonstrating cannabis efficacy; a placebo condition is essential.

The vast consumption of cannabis and cannabinoids for pain paired with exceedingly limited placebo-controlled data leaves both clinicians and patients without the evidence needed to make critical decisions about the benefits vs. risks of cannabis use (Haney, 2020, 2022). Public policy decisions regarding therapeutic cannabis are ideally guided by empirical evidence (rather than observation/anecdote/industry pressure/politics) and the current regulatory constraints on conducting placebo-controlled research on a Schedule 1 drugs hamper the ability to disentangle pharmacological effect from expectation (Haney, 2020).

2.5. Behavioral specificity.

A key behavioral pharmacology principle when testing the effects of a potential pharmacotherapy for CUD is behavioral specificity, i.e., does a medication alter key features of CUD specifically or is it acting on behaviors that would be poorly tolerated clinically? Preclinical behavioral pharmacology models assess whether a medication decreases drug self-administration without altering the reinforcing effects of non-drug reinforcers, such as food. A medication rendering animals too sedated to press a lever for either drug or non-drug reinforcers is unlikely to be useful as a pharmacotherapy.

It is similarly important to assess a potential treatment medication’s effect on cannabis’s positive subjective and reinforcing effects within the context of a range of behaviors, as this can reveal both specificity and tolerability. As an example, using our inpatient CUD model, we found that the alpha-2-noradrenergic agonist, lofexidine, combined with dronabinol reduced symptoms of cannabis withdrawal and reduced cannabis self-administration relative to placebo (Haney et al., 2008). These promising human laboratory findings led to a randomized placebo-controlled trial in patients seeking CUD treatment. Unfortunately, patients poorly tolerated this medication combination, in part due to its sedating effects, and only a subset achieved the target medication dose (Levin et al., 2016). In the human laboratory study, we also found that the medication combination produced substantial sedation relative to placebo, but this was tolerable to inpatient participants who had no work or family responsibilities (Haney et al., 2008). This example of reverse translation highlighted the necessity of considering the constellation of a medication’s behavioral effects in a controlled setting in order to advance medications with fewer adverse effects into the clinical setting.

Models of controlled cannabis administration can also be reverse translational in terms of preclinical models. For example, one theory of addiction based on opponent-process theory suggests that with repeated drug use, there is an allosteric shift in the brain reward system, where baseline mood has shifted toward a negative emotional state, and that use of the drug no longer produces robust positive mood but rather returns mood to a normal set point (Koob, 2021). Our data with cannabis certainly support the shift in negative mood state: The physical and emotional discomfort comprising cannabis withdrawal (negative mood, disrupted sleep; Haney, 2005) contribute to high rates of relapse and sustained cannabis use. Yet, the same participants undergoing cannabis withdrawal, i.e., individuals smoking cannabis daily for many years, still have a robust positive subjective response to controlled cannabis administration (Haney et al., 2023). These findings suggest that daily cannabis use is maintained by both positive and negative reinforcement for individuals with CUD.

3. Conclusions

Human behavioral pharmacology approaches have provided essential evidence of how cannabis influences behavior, how medications alter CUD-related behaviors (reinforcement, misuse-related ratings) and non-specific behaviors (e.g., sedation) that portend efficacy and tolerability in the clinic, how these effects vary as a function of participant phenotype, and finally whether cannabis has therapeutic efficacy relative to placebo. The consistent placebo response observed even among individuals who regularly use cannabis highlights the need for double-blind, placebo-controlled dosing to rigorously differentiate pharmacological action from psychological effect.

Behavior is a highly integrated activity of the nervous system (Witkin et al., 2025), and so a careful understanding of how a drug or combination of drugs affects behavior increases the probability of developing effective pharmacotherapies while reducing risk to humans. Unlike other drugs of misuse, preclinical models do not readily model cannabis’s misuse-related and reinforcing effects (Bedillion et al., 2026) so human behavioral pharmacology approaches are uniquely informative in guiding the development of potential pharmacotherapies and demonstrating a sufficiently robust signal to justify a large randomized controlled clinical trial.

Characterizing cannabis’s behavioral effects requires careful consideration, given the hundreds of cannabinoid and terpene combinations, and given that participants adjust their inhalation patterns as a function of cannabis potency. Cannabis administration studies in humans are not easy to conduct from a regulatory standpoint yet are essential to counteract industry narratives that conflate the limited clinical data available for commercial gain. Ultimately, by grounding our understanding of cannabis in the foundational rigor of behavioral pharmacology, we can move beyond anecdotal claims toward the development of evidence-based therapies that effectively target the behavioral and biological drivers of CUD.

Highlights.

• Behavioral pharmacology provides a rigorous framework to assess the effects of cannabis

• Cannabis potency is not equivalent to Δ⁹-THC dose

• Cannabis smokers adjust their smoking topography as a function of cannabis strength

• Placebo responding is robust even in near-daily cannabis users