Why would an organism voluntarily spend scarce metabolic resources on sustained running when no predator pursues it and no prey lies ahead? From an evolutionary standpoint, prolonged aerobic exertion is costly, and the brain is finely tuned to weigh costs against expected returns. Yet many humans report that exercise feels good. Laboratory rodents given free access to a running wheel will cover several kilometres a night, and they will work for the opportunity to do so.

The popular explanation, that endorphins flood the brain and produce euphoria, is incomplete at best. Beta-endorphin is a large peptide that crosses the blood-brain barrier poorly. Blocking opioid receptors also fails to abolish several key effects of running. The more compelling story involves endocannabinoid signalling, mesolimbic dopamine, and a cascade of plasticity mechanisms that unfold over weeks rather than minutes.

This distinction matters. Kent Berridge separates ‘wanting’ (incentive salience) from ‘liking’ (hedonic impact). Through that lens, exercise appears to engage both systems, through different molecular routes and on different timescales. Tracing those routes helps explain why physical activity shows real therapeutic promise in depression and addiction. It also helps explain why the patients who might benefit most often find it hardest to begin. What follows examines the acute neurochemistry, the chronic remodelling of reward circuitry, and the clinical evidence, noting where the data are strong and where they remain thin.

The Acute Response: Cannabinoids First, Dopamine Second

The strongest acute signal comes from the endocannabinoid system rather than from opioids. In 2012, Raichlen and colleagues showed that moderate-intensity treadmill running raised circulating anandamide in humans and dogs but not in ferrets. Humans and dogs are cursorial species, shaped by selection for endurance. Ferrets are not. The comparative pattern suggests that evolution may have favoured a neurobiological reward for sustained aerobic activity in lineages whose foraging depended on it.

Mechanistic confirmation followed in mice. Fuss and colleagues (2015) found that running reduced both anxiety and pain sensitivity. Blocking CB1 cannabinoid receptors abolished both effects, while blocking opioid receptors with naloxone did not. In humans, Siebers and colleagues (2021) reported that euphoria and reduced anxiety after running persisted under opioid blockade, while plasma endocannabinoids rose. Anandamide is lipophilic and crosses the blood-brain barrier readily, a property beta-endorphin lacks.

Opioids are not irrelevant. In a 2008 PET study, Boecker and colleagues detected opioidergic changes in prefrontal and limbic regions after a two-hour run, and these changes correlated with self-reported euphoria. The emerging picture is one of converging signals, with endocannabinoids carrying more of the weight than folk accounts suggest. Both systems act on hedonic hotspots in the nucleus accumbens shell and ventral pallidum, the regions Berridge identified as generators of ‘liking’.

Dopamine enters through both direct and indirect routes. Rodent microdialysis consistently shows increased striatal dopamine during treadmill running. One plausible indirect pathway runs through the ventral tegmental area, where CB1 receptors sit on GABAergic terminals. Endocannabinoid signalling there suppresses inhibitory input, which releases dopamine neurons from inhibition. Human evidence is less clear. A 2000 PET study by Wang and colleagues, using raclopride, found no significant striatal displacement after thirty minutes of treadmill exercise in trained subjects.

The most defensible reading is that acute exercise produces modest, regionally specific dopaminergic modulation layered onto more robust endocannabinoid effects. In Berridge's terms, the hedonic glow of a run may owe more to ‘liking’ mechanisms. Dopamine's contribution becomes clearer when we examine what repeated exercise does to the circuitry that generates ‘wanting’. Intensity matters as well. The effects appear strongest at moderate, sustained intensities rather than at maximal exertion, where aversive interoceptive signals from fatigue, lactate, and breathlessness may dominate the experience.

Takeaway

The runner's high appears to be primarily a cannabinoid phenomenon rather than an opioid one. It may be an evolutionary bribe for endurance that shaped how our lineage experiences sustained movement.

Chronic Remodelling: How Training Retunes the Reward Circuit

Repeated exercise does something a single session cannot: it reorganises the substrate on which reward is computed. In rodents, chronic wheel running causes ΔFosB to accumulate in the nucleus accumbens. This is the same long-lived transcription factor that builds up after repeated exposure to drugs of abuse and other natural rewards. Werme and colleagues showed that mice overexpressing ΔFosB in striatal dynorphin-containing neurons ran substantially more, implicating this molecule in the escalating motivation to run.

Changes in receptor availability offer a second mechanism. Treadmill training increases striatal D2 receptor expression in rodent models of Parkinson's disease. In humans, Robertson and colleagues (2016) used PET to study people in residential treatment for methamphetamine use disorder. Eight weeks of structured exercise increased their striatal D2/D3 receptor availability. Low D2 availability is a recurring finding in addiction and has been linked to impulsivity, so this result is mechanistically meaningful, though it comes from a modest sample.

Neurotrophic signalling provides a third route. Exercise reliably raises brain-derived neurotrophic factor, which supports dopaminergic neuron survival and synaptic plasticity. In 1999, van Praag and colleagues showed that running enhances hippocampal neurogenesis in mice. Erickson and colleagues later found that a year of aerobic walking increased hippocampal volume in older adults by roughly two percent. The hippocampus feeds contextual information into the mesolimbic loop, so its plasticity bears on how reward-predictive cues are encoded.

There is a further subtlety: the rewarding properties of exercise are themselves learned. Rats develop a conditioned place preference for environments paired with the aftereffects of wheel running. Schultz's prediction-error framework would predict that, over training, dopamine responses migrate from the outcome to the cues that precede it. The anticipatory pull a seasoned runner feels before lacing their shoes may reflect exactly this transfer of incentive salience to antecedent stimuli, though direct human evidence remains limited.

Not all of this plasticity is benign. In a minority of individuals, the same mechanisms that strengthen motivation can contribute to compulsive exercise, particularly alongside eating disorders. Garland and colleagues bred mouse lines selectively for voluntary wheel running. These high-runner lines show altered dopaminergic signalling, and their running responds differently to drugs such as methylphenidate. Exercise engages genuine reward machinery, with all the vulnerabilities that implies.

Takeaway

Regular exercise doesn't just deliver reward. It retunes the system that assigns value, gradually shifting incentive salience toward the act itself and the cues that predict it.

Exercise as Intervention: Depression, Anhedonia, and Addiction

Depression has the most extensive evidence base. In 1999, Blumenthal and colleagues found that sixteen weeks of aerobic exercise produced outcomes comparable to sertraline in older adults with major depression. Meta-analyses since then have generally reported moderate effect sizes. A 2024 network meta-analysis by Noetel and colleagues in the BMJ ranked walking or jogging, yoga, and strength training among the more effective treatments, with benefits generally larger at higher intensities.

The caveats are substantial. Trials are rarely blinded, and attention and social contact are difficult to control for. Earlier Cochrane analyses found that effects shrank when only methodologically rigorous studies were included. Schuch and colleagues argued in 2016 that the effects remain robust after correcting for publication bias, but heterogeneity across studies is considerable. The honest conclusion is that exercise is a credible intervention, not a universal one.

The neurobiological rationale is strongest for anhedonia and motivational deficits. Effort-based decision tasks developed by Treadway and colleagues show that depressed individuals are less willing to expend effort for reward, a pattern linked to striatal dopamine function. If exercise enhances dopaminergic tone and D2 availability, it should target precisely this deficit. The paradox is obvious: patients with the greatest impairment in valuing effort are the least able to start the intervention that might correct it.

Addiction research shows a similar gap between animal and human evidence. In rodents, access to running wheels reduces cocaine self-administration and weakens reinstatement after abstinence. Sex and timing modulate these effects, and Lynch and colleagues have shown that exercise during abstinence is particularly protective. Human trials are more equivocal. Acute exercise reliably reduces cigarette cravings, yet Cochrane reviews have found insufficient evidence that it improves long-term smoking abstinence. Methamphetamine trials show improvements in mood and dopamine receptor measures, but relapse outcomes are less conclusive.

Across conditions, exercise seems to work best as an adjunct that reshapes the neural terrain on which other treatments operate, rather than as a standalone cure. Clinically, this argues for scaffolding: starting at low thresholds, drawing on social and environmental cues, and accepting that motivation often follows activity rather than preceding it.

Takeaway

In motivational disorders, waiting to feel ready may get the causality backwards. The capacity to want can be partly rebuilt through the very action it currently resists.

Exercise engages the reward system through at least two partially distinct channels. The first is an acute hedonic response, largely mediated by endocannabinoids. The second is a slower dopaminergic and neurotrophic remodelling that changes how value and effort are computed. Seen this way, the runner's high is less a single phenomenon than the visible surface of layered adaptations.

This framing reaches beyond exercise. It supports the view that motivational systems are plastic, that incentive salience can be redirected toward effortful and delayed rewards, and that behaviour can reshape the circuitry that generates future behaviour. Few interventions show this two-way relationship so clearly.

Considerable uncertainty remains, particularly about human dopamine dynamics during exercise and how long clinical effects last. The evolutionary logic is nonetheless suggestive. A brain that rewards sustained movement was likely an advantage for our ancestors, and the circuitry that once drove endurance foraging may still be available, with practice, to support recovery.