The Science of Mental Health — What We Actually Know

Mental health conditions are among the most common health problems in the world, and among the most misunderstood. The research literature is enormous, genuinely informative in places, and genuinely unsettled in others — and public discussion tends to flatten that difference, either overstating how much neuroscience explains or dismissing the conditions as not real.

This article tries to hold the line between those errors: what the numbers are and where they come from, what brain research has and has not established, which treatments have good trial evidence, and which widely repeated claims do not survive scrutiny.

Educational content, not medical advice. This article describes general research findings. It cannot diagnose anyone or tell any individual what treatment to use. Decisions about medication, therapy or any change to an existing treatment belong with a qualified clinician who knows your situation. If you are in crisis, see the crisis resources section near the end.

Key takeaways

• WHO estimated about 1.2 billion people living with a mental disorder in 2023 — close to 1 in 7 people worldwide.
• The treatment gap is large: WHO reports that only about one third of people with depression, and 40% of people with psychosis, receive formal mental health care.
• The "chemical imbalance" account of depression is not supported by the evidence as a complete explanation, which is a separate question from whether the conditions are real or whether treatments work.
• Several treatments have strong randomised-trial support, including cognitive behavioural therapy, antidepressant medication for moderate to severe depression, and specific protocols such as exposure and response prevention for OCD.

How common are mental health conditions?

Global estimates come mainly from the World Health Organization, drawing on the Global Burden of Disease programme. The figures below are the ones WHO published for 2023 in its mental disorders fact sheet.

Condition	People affected (2023, WHO)	Notes
All mental disorders	~1.2 billion (nearly 1 in 7)	Includes all diagnoses below and more
Anxiety disorders	470 million, incl. 146 million children and adolescents	Most common single category
Depressive disorders	322 million, incl. 43 million children and adolescents
Bipolar disorder	36 million, incl. 3.3 million aged 10–19
Schizophrenia	~27 million (about 1 in 300)
Eating disorders	18 million, incl. 4.7 million children and adolescents

National survey data gives a different and complementary view. In the United States, the National Institute of Mental Health reports, from the 2022 National Survey on Drug Use and Health, that 23.1% of adults — about 59.3 million people — had any mental illness in the past year, and that about half of them received any mental health treatment. For major depressive episodes specifically, NIMH's figures from the 2021 survey put past-year prevalence at 8.3% of adults, around 21 million people, rising to 18.6% among 18-to-25-year-olds and falling to 4.5% among those aged 50 and over.

Two methodological cautions apply to all of these numbers. First, prevalence estimates depend on the diagnostic thresholds and survey instruments used; a different instrument produces a different count for the same population. Second, NIMH itself flags that US survey estimates from 2021 and 2022 are not directly comparable with earlier years, because data collection changed during the COVID-19 pandemic and response rates fell sharply. Cross-country comparisons, examined further in our piece on how mental health disorder rates differ between countries, reflect differences in screening and reporting as much as differences in underlying rates.

The treatment gap

The most robust single finding in global mental health is not about biology. It is that most people who meet criteria for a disorder do not receive care for it.

WHO reports that only about one third of people with depression, and about 40% of people with psychosis, receive formal mental health services. The reasons differ by setting and stack on top of one another: too few trained clinicians, concentration of what services exist in capital cities, cost, waiting lists, and stigma that delays the first appointment by years in many cases. In high-income countries the bottleneck is more often waiting time and cost; in low- and middle-income countries it is often the absence of any specialist service within reach.

What is actually happening in the brain

Mental health conditions involve the brain, but "involve the brain" is a weaker claim than it sounds — so does learning to ride a bicycle. The useful question is which specific mechanisms have evidence behind them, and at what strength.

Neurotransmitter systems

Serotonin, dopamine, noradrenaline, GABA and glutamate are all implicated in different conditions, and drugs acting on them change symptoms. That much is solid. What does not hold is the simple version: that depression is caused by a serotonin deficit and antidepressants top it up.

A 2022 umbrella review in Molecular Psychiatry led by Joanna Moncrieff at University College London examined the main lines of evidence for the serotonin hypothesis — metabolite levels, receptor binding, tryptophan depletion, genetic association — and concluded that the theory is not empirically substantiated. The paper attracted substantial criticism, including a published comment signed by dozens of researchers arguing its exclusion criteria were too aggressive. The reasonable reading is the one most psychiatrists already held: serotonin is involved in mood regulation, but "low serotonin causes depression" is not a supported mechanism.

A point that is frequently confused: whether a drug works and why a drug works are separate questions, answered by separate evidence. Aspirin relieved pain for decades before prostaglandins were identified. The failure of the serotonin hypothesis as an explanation says nothing about trial results on antidepressant efficacy. Our article on the neuroscience of depression goes through this distinction in more depth.

Brain circuits

Rather than single chemicals, current research focuses on networks. Altered activity and connectivity have been reported in the default mode network (associated with rumination), the salience network (threat detection) and executive control networks across several conditions. These findings are reasonably reproducible at group level, which matters for understanding mechanisms, but the effect sizes are small relative to individual variation — which is why no brain scan can diagnose depression or anxiety in a particular person, and why anyone selling such a scan is ahead of the science.

Structure

MRI studies report average differences in regional volumes — hippocampal volume in depression and PTSD is the best-known example. These are group averages with heavily overlapping distributions, and the direction of causation is often unclear: chronic stress can reduce hippocampal volume, so a difference observed after years of illness may be consequence rather than cause.

Stress and the HPA axis

Dysregulation of the hypothalamic–pituitary–adrenal axis — the system that governs cortisol release — is documented across depression, anxiety disorders and PTSD. Chronic activation of this system has downstream effects on sleep, immune function, appetite and hippocampal plasticity, which makes it a plausible mechanism linking adversity to illness. Our explainer on cortisol and the stress response covers the normal system that these conditions distort.

Inflammation

A subset of people with depression show elevated inflammatory markers, and inflammatory illnesses raise depression risk. Whether inflammation causes depressive symptoms, results from them, or both, is unresolved, and anti-inflammatory treatments have so far produced mixed trial results. This is an active research area rather than an established mechanism.

Genetics

Heritability is real and polygenic. The largest schizophrenia genome-wide association study to date, published by the Psychiatric Genomics Consortium in Nature in 2022, analysed up to 76,755 people with schizophrenia and 243,649 controls and identified common-variant associations at 287 distinct genomic loci, with the signal concentrated in genes expressed in excitatory and inhibitory neurons. That is a landmark result and also a demonstration of how the genetics works: hundreds of variants of individually tiny effect, no single "schizophrenia gene," and nothing deterministic. Most people carrying risk variants never develop the condition.

What the honest summary looks like

Real biological differences exist; they are statistical rather than diagnostic; genetics contributes substantially but probabilistically; environment and life history matter enormously; and no condition in psychiatry currently has a complete causal account. Both "it's just a chemical imbalance" and "it's all in your head" are wrong, in opposite directions.

Major conditions at a glance

Prevalence figures below are WHO 2023 global estimates where available; others are widely cited epidemiological ranges and should be read as approximations.

Condition	Scale	Core features	Evidence-based treatments
Depression	322 million globally (WHO 2023)	Persistent low mood, loss of interest, sleep and appetite change, cognitive slowing	CBT, behavioural activation, antidepressants, exercise as adjunct
Anxiety disorders	470 million globally (WHO 2023)	Persistent worry or fear out of proportion to circumstances	CBT with exposure, SSRIs/SNRIs
Bipolar disorder	36 million globally (WHO 2023)	Alternating depressive and manic or hypomanic episodes	Mood stabilisers, psychoeducation, structured routines
Schizophrenia	~27 million globally (WHO 2023)	Hallucinations, delusions, disorganised thought, negative symptoms	Antipsychotics, family intervention, supported employment
Eating disorders	18 million globally (WHO 2023)	Disordered eating and body image with serious physical risk	Specialist protocols; among the highest mortality in psychiatry
OCD	~1–2% of adults	Intrusive thoughts and compulsive rituals	Exposure and response prevention; SSRIs at higher doses
PTSD	Varies sharply with trauma exposure	Re-experiencing, avoidance, hyperarousal	Trauma-focused CBT, prolonged exposure, EMDR
ADHD	~5% of children, ~2.5% of adults	Inattention, executive difficulty, impulsivity	Stimulant and non-stimulant medication, behavioural support

This is not an exhaustive list, and the categories themselves are contested. Diagnostic manuals describe clusters of symptoms, not confirmed disease entities; comorbidity is the norm rather than the exception; and the boundary between a severe but ordinary reaction to circumstances and a clinical condition is defined by duration, severity and impairment rather than by any biological test.

Two conditions worth a closer look elsewhere: how anxiety works in the brain covers the threat-detection circuitry and why avoidance maintains the problem, and our dopamine explainer addresses the reward-system claims that circulate around ADHD and addiction.

What the treatment evidence shows

Psychological therapies

Cognitive behavioural therapy has the largest trial base of any psychological treatment, with good evidence across depression, anxiety disorders, OCD, PTSD and insomnia. For mild to moderate depression it performs comparably to medication in many head-to-head trials, with some evidence that gains persist better after treatment ends. Effect sizes in the research literature have fallen somewhat over time as trial quality improved and publication bias was corrected for — a normal and healthy pattern, not a refutation.

Several condition-specific protocols are more effective than generic therapy for their targets:

• Exposure and response prevention (ERP) for OCD.
• Prolonged exposure and trauma-focused CBT for PTSD; EMDR also has supportive trial evidence.
• Dialectical behaviour therapy (DBT) for borderline personality disorder and recurrent self-harm.
• Interpersonal therapy (IPT) for depression, particularly where interpersonal loss or role change is central.
• Behavioural activation for depression — simpler than full CBT and, in several trials, comparably effective.

Medication

The most informative single source on antidepressants is Andrea Cipriani and colleagues' 2018 network meta-analysis in The Lancet, which pooled 522 randomised trials covering 116,477 participants. All 21 antidepressants studied outperformed placebo on response rate, with odds ratios ranging from 2.13 for amitriptyline down to 1.37 for reboxetine. Differences between the active drugs were modest, and the authors rated the certainty of much of the evidence as moderate to very low.

What that supports is a measured statement: antidepressants work better than placebo for major depressive disorder, the average benefit is real but not dramatic, and there is no clear best drug for everyone. Benefit is larger in more severe depression, which is why most guidelines recommend psychological therapy first for mild cases.

Other medication classes have their own evidence profiles. Antipsychotics substantially reduce relapse in schizophrenia and are essential in acute mania, at the cost of metabolic and neurological side effects that require monitoring. Lithium has the strongest long-term evidence in bipolar disorder, including on suicide risk, and requires blood-level monitoring. Stimulants are among the most effective medications in psychiatry for ADHD symptoms. Benzodiazepines relieve acute anxiety reliably but carry tolerance, dependence and rebound risks that make them poor long-term choices.

None of this constitutes advice about any individual's treatment. Drug choice, dosing, interactions and monitoring are clinical decisions, and stopping a psychiatric medication abruptly can cause discontinuation symptoms or relapse.

Lifestyle and adjunctive factors

• Exercise. Meta-analyses support a moderate antidepressant effect for structured exercise in mild to moderate depression. Trial quality is variable — blinding is impossible and control conditions differ — so the honest framing is that exercise is a well-supported adjunct, not a demonstrated replacement for treatment. See the science of exercise.
• Sleep. The relationship runs in both directions. Insomnia predicts later depression, and CBT for insomnia improves depressive symptoms in trials that treat sleep alone, which suggests the link is partly causal rather than purely symptomatic. Our piece on how sleep affects the brain covers the underlying mechanisms.
• Social connection. Loneliness and social isolation are consistently associated with worse mental and physical health outcomes. Most of this evidence is observational, and reverse causation is difficult to exclude.
• Diet. Observational evidence favours Mediterranean-style patterns; the small number of randomised trials is suggestive but not conclusive.

Newer and emerging treatments

Repetitive transcranial magnetic stimulation is approved in several countries for treatment-resistant depression, with modest average effects. Ketamine and esketamine act within hours rather than weeks, which is clinically important in severe cases, but durability and long-term safety remain under study. Psychedelic-assisted therapy has produced encouraging early trial results for depression and PTSD, in small samples, with unresolved problems around blinding — participants almost always know which arm they are in — and strong expectancy effects. Treat these as promising and unfinished, not established.

That last point generalises. Expectancy is a real, measurable contributor to symptom change across medicine, and especially where outcomes are self-reported, as explained in our piece on how placebos change real symptoms. It does not mean treatments are fake; it means trial design has to work hard to isolate specific effects.

Established, contested and unsupported

A rough map of where claims sit:

Well established

• Mental health conditions are common, impairing and measurable.
• Heritability is substantial and highly polygenic.
• Several psychological and pharmacological treatments beat placebo or waitlist in randomised trials.
• The majority of affected people worldwide receive no formal care.

Genuinely contested

• Whether inflammation plays a causal role in a subset of depression.
• The durability and real-world effect sizes of psychedelic-assisted therapy.
• How much of the measured rise in adolescent distress is attributable to social media versus better detection, reduced stigma and other social changes.
• Whether current diagnostic categories carve the territory at useful joints.

Not supported

• That depression is caused by a serotonin deficiency.
• That a brain scan or blood test can diagnose a mental disorder in an individual.
• That mental illness reflects weakness, laziness or a character failing.
• That any single treatment works for everyone with a given diagnosis.

Common misconceptions

"Mental illness is a weakness." It is not. Heritability, documented physiological differences and the response to specific treatments all point to real disorders of a biological organ, shaped by circumstance.

"Antidepressants are happy pills." They do not elevate mood in people without depression. In trials they reduce depressive symptoms in people who have them; the effect is partial and takes weeks.

"Therapy is just talking." Evidence-based therapies are structured protocols with defined components, homework and measurable targets. ERP for OCD, for example, involves deliberately confronting triggers while refraining from the compulsion — closer to rehabilitation than to conversation.

"You should be able to snap out of it." Willpower is not a treatment for a condition that affects motivation, concentration and sleep regulation. Expecting it to be one is part of what delays help-seeking.

"Mental illness is increasing because people are soft." Measured increases are substantially driven by better detection and reduced stigma. Where genuine increases appear, particularly distress in adolescents, the causes are debated rather than settled.

"Conditions are either chronic or cured." Many follow a relapsing–remitting course. A first depressive episode raises the probability of another, which is why maintenance treatment and relapse prevention matter — a management model closer to asthma than to appendicitis.

What we still do not know

• No biomarkers. There is no laboratory test for any common psychiatric diagnosis. Diagnosis is clinical, based on history and observation.
• No reliable treatment matching. Clinicians cannot yet predict which patient will respond to which antidepressant or which therapy, so practice remains partly trial and error.
• Mechanisms of action. Why antidepressants take weeks to work, when receptor occupancy happens within hours, is still not fully explained.
• Category validity. Whether disorders such as "major depressive disorder" are single conditions or heterogeneous collections sharing a symptom profile is an open question with major consequences for research.
• Prevention. Evidence on preventing first episodes, as opposed to treating them, is comparatively thin.

When to seek help, and crisis resources

General signs that professional assessment is worthwhile: symptoms persisting beyond a couple of weeks; interference with work, study, relationships or self-care; escalating alcohol or drug use; withdrawal from people you normally see; or any thoughts of self-harm or suicide.

Suicide is a global public health problem, not a rare event. WHO's suicide fact sheet reports that more than 720,000 people died by suicide in 2021, that it was the third leading cause of death among 15-to-29-year-olds worldwide that year, and that about 73% of suicides occurred in low- and middle-income countries.

If you or someone else is in immediate danger, contact local emergency services.

• United States: the 988 Suicide & Crisis Lifeline is reachable by calling or texting 988, or by chat at chat.988lifeline.org.
• United Kingdom and Ireland: Samaritans can be called free on 116 123, 24 hours a day, every day of the year.
• Elsewhere: most countries operate a national crisis line; the International Association for Suicide Prevention maintains a directory of local services.

Reaching out on someone else's behalf is appropriate. Staying with the person, removing access to means where possible, and contacting a crisis service together are reasonable steps when you are worried.

Frequently asked questions

Q: How many people have a mental health condition? A: WHO estimated about 1.2 billion people, nearly 1 in 7 worldwide, were living with a mental disorder in 2023, including 470 million with anxiety disorders and 322 million with depression. In the United States, NIMH reports 23.1% of adults had any mental illness in the past year, based on the 2022 national survey.

Q: Is depression caused by a chemical imbalance? A: Not in the simple sense. A 2022 umbrella review in Molecular Psychiatry found the serotonin-deficiency account unsupported, and that conclusion has itself been contested. Current research points to interacting contributions from genetics, stress-response systems, circuit-level changes and life circumstances rather than one deficient chemical.

Q: Do antidepressants actually work? A: For major depressive disorder, yes, better than placebo on average. The 2018 Lancet network meta-analysis of 522 trials and 116,477 participants found all 21 drugs studied outperformed placebo, with modest differences between them. Benefit is clearer in more severe depression. Whether medication is appropriate for a particular person is a clinical decision.

Q: Is therapy as effective as medication? A: For mild to moderate depression and for many anxiety disorders, cognitive behavioural therapy performs comparably in head-to-head trials, with some evidence of better maintenance after treatment stops. For severe depression, mania and psychosis, medication plays a central role and therapy is usually an addition rather than an alternative.

Q: Can a brain scan diagnose a mental illness? A: No. Imaging studies find average group differences that overlap heavily between individuals with and without a diagnosis. No scan, blood test or genetic panel can currently diagnose a psychiatric condition in one person.

Q: Is mental illness really becoming more common? A: Reduced stigma, better recognition and broader screening account for a large share of the measured increase. There is genuine debate about whether distress among adolescents has risen beyond that, and about what would be driving it.

Q: Can mental health conditions be cured? A: Some episodes resolve completely. Many conditions are better described as managed than cured, with a relapsing course, which is why ongoing treatment and relapse prevention are standard in psychiatry rather than a sign of failure.

Sources

• World Health Organization, "Mental disorders" fact sheet (2023 figures) — https://www.who.int/news-room/fact-sheets/detail/mental-disorders
• World Health Organization, "Suicide" fact sheet (2021 figures) — https://www.who.int/news-room/fact-sheets/detail/suicide
• National Institute of Mental Health, "Mental Illness" statistics (2022 NSDUH) — https://www.nimh.nih.gov/health/statistics/mental-illness
• National Institute of Mental Health, "Major Depression" statistics (2021 NSDUH) — https://www.nimh.nih.gov/health/statistics/major-depression
• 988 Suicide & Crisis Lifeline — https://988lifeline.org/
• Samaritans (UK and Ireland) — https://www.samaritans.org/
• Cipriani A. et al. (2018). Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder. The Lancet 391:1357–1366.
• Moncrieff J. et al. (2022). The serotonin theory of depression: a systematic umbrella review of the evidence. Molecular Psychiatry.
• Trubetskoy V. et al. (2022). Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 604:502–508.

Related Reading

• The Science of Sleep — Why You Need It and What Happens Without It
• The Science of Focus — Attention in the Age of Distraction
• Intermittent Fasting — What the Science Actually Says
• How Learning Actually Works — The Science of Getting Better at Anything