Large-Scale Brain Scan Study Reveals Cognitive Effects of Long-Term Cannabis Use

The headlines arrived with predictable certainty. "True impact of weed on brain according to hundreds of thousands of scans." The implication: science has finally settled the question. But the actual research tells a story far more complicated than any headline can convey — one where the same substance appears to harm young brains while preserving older ones, where moderate use looks different from heavy use, and where the most honest answer to "what does cannabis do to the brain?" remains "it depends."

Two landmark studies, published within a year of each other, form the core of the current debate. Together, they represent the largest neuroimaging investigations of cannabis ever conducted. And they point in seemingly opposite directions.

The Studies: Scale and Scope

In January 2025, researchers at the University of Colorado Anschutz Medical Campus published in JAMA Network Open what they called the largest study of its kind: functional MRI data from 1,003 young adults aged 22 to 36 [1]. Participants were classified as heavy users (1,000 or more lifetime uses, n=88), moderate users (10 to 999 uses, n=179), or non-users (fewer than 10 uses, n=736) [2]. Each underwent brain imaging while performing seven cognitive tasks testing working memory, reward processing, emotion, language, motor skills, relational reasoning, and theory of mind [1].

In early 2026, a separate team — also from CU Anschutz — published results from 26,362 participants in the UK Biobank, aged 40 to 77 (mean age 55), examining structural brain volumes and cognitive performance [3][4]. This dwarfs the earlier study in sample size and draws from a population-level dataset of over 500,000 adults.

For context, the Adolescent Brain Cognitive Development (ABCD) Study — the largest long-term study of brain development in U.S. youth — follows approximately 11,876 adolescents with neuroimaging and cognitive assessments [5]. The Amen Clinics database, frequently cited in popular press, contains SPECT scans from approximately 62,454 brain scans across over 30,000 patients — though critics have raised serious concerns about its clinical methodology and sample bias [6][7].

Source: OpenAlex

Data as of Jan 1, 2026CSV

Academic research on cannabis and neuroimaging has surged, with over 11,600 papers published since 2011 and a peak of 1,784 in 2023, reflecting both expanded legalization and growing scientific interest [8].

What the Imaging Shows: Younger Adults

The JAMA Network Open study's central finding: 63% of heavy lifetime cannabis users showed reduced brain activity during working memory tasks, and 68% of recent users showed a similar pattern [1][2]. The affected regions — the dorsolateral prefrontal cortex, dorsomedial prefrontal cortex, and anterior insula — are involved in decision-making, attention, emotional processing, and memory [1].

The researchers applied false discovery rate (FDR) correction across all seven task domains to minimize false positives. Lead author Joshua Gowin stated the team set "rigorous thresholds for statistical significance" [2]. Reduced activation was specific to working memory; the other six cognitive domains — reward, emotion, language, motor, relational, and theory of mind — did not show statistically significant differences between cannabis users and non-users [1].

This specificity matters. The study does not show global brain impairment. It shows that one cognitive domain — working memory, the ability to hold and manipulate information in real time — is affected during active use and heavy cumulative exposure.

What the Imaging Shows: Older Adults

The UK Biobank study found the opposite pattern in an older cohort. Lifetime cannabis users had larger brain volumes in CB1 receptor-rich regions, including the hippocampus, caudate, putamen, and amygdala [3][4]. Greater lifetime use was associated with better performance on tests of learning, processing speed, short-term memory, and attention [3].

Moderate users showed the most consistent benefits across measures. Heavy users outperformed on some specific metrics — right amygdala volume and visual memory — but did not beat moderates overall [4]. One region, the posterior cingulate cortex, showed lower volume with higher use [3].

"The story is nuanced. It's not a case of cannabis being all good or all bad," said Dr. Anika Guha, lead researcher on the UK Biobank study [3].

The divergence between these findings — harm in younger brains, apparent benefit in older ones — may reflect developmental timing. The prefrontal cortex does not fully mature until the mid-20s, and adolescent-onset use has consistently shown worse outcomes than adult-onset use across multiple studies [9][10].

Effect Sizes: Clinically Meaningful or Statistically Noisy?

A critical question is whether the observed differences are large enough to matter in daily life. A 2022 meta-review in Addiction synthesized multiple meta-analyses and found that acute cannabis exposure produced moderate cognitive deficits: Cohen's d of 0.69 for verbal learning and memory, 0.51 for working memory, 0.38 for processing speed, and 0.37 for executive function [11].

Source: Dellazizzo et al. 2022, Addiction; Scott et al. 2018, JAMA Psychiatry

Data as of Jan 1, 2022CSV

Residual effects — cognitive differences persisting beyond intoxication in regular users — were smaller: Cohen's d of roughly 0.20 to 0.25 for attention, executive function, and processing speed [11][12]. In neuropsychology, a Cohen's d below 0.30 is generally considered small and may not translate to noticeable real-world impairment, though it can still be meaningful across populations.

The acute effects, particularly on verbal memory (d=0.69), are clearly in the moderate-to-large range. The residual effects are closer to the boundary where statistical significance and practical significance diverge — measurable in a laboratory, but harder to detect in someone's everyday functioning.

The Dose-Response Question

Research consistently shows a dose-dependent relationship, though the curve is not linear across all outcomes. A 2021 review in Frontiers in Psychiatry reported that mild use (fewer than 10 joints per month) was not associated with measurable cognitive decline, while regular use (multiple times per week) was linked to mild deficits, and daily use to mild-to-moderate deficits [12]. Memory performance decreased progressively with each additional five-year period of use [12].

Potency compounds frequency. Average THC content in retail cannabis has risen from 1-4% in the 1970s to 15-30% in modern dispensaries [13]. The American Heart Association's 2020 scientific statement noted linear dose-dependent effects of THC on driving impairment and cognitive function, with higher-THC products producing roughly double the effect sizes of lower-potency products [13].

Some evidence suggests a ceiling effect: once cannabis exposure crosses a threshold of heavy, sustained use, additional consumption may not produce further measurable brain changes [14]. This would mean the relationship between use and harm is not a straight upward line, but a curve that flattens — with implications for how risk is communicated to the heaviest users.

Confounders: The Persistent Problem

Cannabis research is plagued by confounding variables, and critics have long argued that observed brain differences may reflect pre-existing conditions rather than drug effects.

The JAMA Network Open study is observational and cross-sectional, meaning it captured a single time point and cannot establish causation [1][2]. The researchers applied FDR correction and set high statistical thresholds, but the study has not published detailed information about adjustment for alcohol use, other drug use, psychiatric diagnoses, socioeconomic status, or sleep quality [2].

The UK Biobank study has similar limitations. Cannabis potency and composition were unknown — participants were categorized by lifetime use estimates, not by what they actually consumed [3]. Most participants had used cannabis decades ago, when products differed substantially from those available today [3].

The ABCD Study, which follows nearly 12,000 adolescents longitudinally, represents the gold standard for controlling confounders over time [5]. Its 2026 findings, combining self-reported use with objective hair toxicology, found that teenagers who began using cannabis showed slower cognitive development — but acknowledged that disentangling cannabis effects from co-occurring risk factors (poverty, trauma, other substance use) remains methodologically difficult [5].

Dr. Daniel Amen's clinic-based SPECT research, which claims cannabis use accelerates brain aging by 2.8 years, has drawn sharper criticism. A commentary in the American Journal of Psychiatry argued that Amen's work "appears to prioritize marketing over science" and that his scans are drawn entirely from clinical patients seeking treatment — a selection bias that limits generalizability [6][7].

Reverse Causation and Self-Medication

The strongest counterargument to "cannabis harms the brain" is that the causal arrow may point in the wrong direction. People with pre-existing anxiety, depression, or neurological differences may be drawn to cannabis, and their brain differences may predate their use.

Meta-analytic evidence on this question is mixed. A 2020 literature review found a "robust positive association" between cannabis use and anxiety symptoms but concluded that minimal research supported a causal or predictive relationship in either direction, with most studies failing to adequately adjust for confounders [15]. A study using propensity score matching — a statistical technique that compares cannabis users and non-users who are otherwise similar on measured variables — found that brain activation differences persisted after matching for anxiety, depression, alcohol, nicotine, and demographic variables, reducing but not eliminating the reverse-causation concern [16].

Research testing the self-medication hypothesis directly has produced mixed results. One study found that people with prior depression did not use cannabis to relieve symptoms; under cannabis influence, they more often reported increased depression, sadness, anxiety, and paranoia [17]. But this does not rule out that people with subclinical mood or attention differences — below the threshold for diagnosis — might still gravitate toward cannabis at higher rates.

The brain regions most affected in the JAMA Network Open study — the prefrontal cortex and anterior insula — are also implicated in anxiety and mood disorders independently of substance use [15][16]. This overlap makes clean causal attribution difficult without longitudinal designs that image brains before first cannabis exposure.

Reversibility: What Happens After Quitting

The evidence on recovery is among the most clinically relevant — and most reassuring — findings in this literature.

CB1 cannabinoid receptors, which THC binds to and which are downregulated by chronic use, appear to normalize within about four weeks of abstinence. Positron emission tomography studies found a 20% decrease in cortical CB1 receptor availability among chronic users at baseline, with differences no longer detectable after one month without cannabis [18].

Cognitive improvements begin within the first week of cessation for most domains [18]. However, recovery varies sharply by age of onset. Adult-onset users show more complete recovery than adolescent-onset users, and heavy users require longer abstinence periods [18]. A longitudinal New Zealand cohort study found that neuropsychological functioning was "not fully restored" in adolescent-onset persistent users even after a year or more of reduced use [19].

Structural changes present a more mixed picture. Hippocampal atrophy was observed even after six months of supervised abstinence in young adults with heavy use histories [14]. This suggests that while receptor function bounces back relatively quickly, some tissue-level changes may be slower to reverse — or may not reverse fully in all individuals.

The Legal-Market Gap

One of the most policy-relevant questions — whether regulated cannabis markets produce different cognitive outcomes than unregulated ones — remains largely unanswered. No large neuroimaging study has directly compared users in legal versus illegal markets.

What is known: federally produced cannabis used in research (from the University of Mississippi's NIDA-funded program) contains THC levels averaging 27-35% of those available in state-legal dispensaries, while containing 11-23 times more CBN, a degradation product [20]. This means decades of federally sanctioned research may have used products that poorly represent what tens of millions of Americans actually consume.

The American Heart Association has recommended that legal markets adopt potency caps and mandate balanced THC:CBD ratios to reduce cognitive risk [13]. Some researchers have proposed that lower-potency, CBD-inclusive products — more feasible in regulated markets where labeling is required — could minimize harm, but this hypothesis lacks direct neuroimaging validation.

What Remains Unknown

The honest summary of the evidence is that cannabis affects the brain, the effects differ dramatically by age of first use and frequency of use, moderate acute effects on memory are well-established, residual effects in non-acute users are small, and some changes reverse with abstinence while others may not.

What the evidence does not yet support:

• That any amount of cannabis use causes brain damage in adults
• That the brain changes observed in heavy users are irreversible across the board
• That cannabis use in middle-aged and older adults is harmful — the largest study of this population found the opposite [3]
• That observed brain differences are entirely caused by cannabis rather than partly reflecting who chooses to use it

What the evidence does support:

• Heavy, frequent use — especially beginning in adolescence — is associated with reduced working memory function and structural brain changes [1][5]
• These effects show a dose-response pattern, with mild use below a clear harm threshold [12]
• THC potency matters, and today's products are substantially more potent than those studied in earlier decades [13]
• Receptor-level changes are largely reversible within weeks; structural changes may take longer [18]

The temptation, in both directions, is to cherry-pick. Advocates for legalization can point to the UK Biobank study. Opponents can point to the JAMA Network Open findings. The research, taken whole, supports neither blanket reassurance nor blanket alarm. It supports a more difficult message: that the effects of cannabis on the brain are real, dose-dependent, age-dependent, and at least partially reversible — and that the largest studies ever conducted still cannot tell us precisely where the line between acceptable risk and measurable harm falls for any individual user.