01 / Frames05 / SummaryA / About
18432026
“Crisis” is a claim before it is a condition. This archive follows who declared it, what counted as evidence, where the claim traveled, and what it made possible.
One cell represents each year from 1843 to 2026. Start with the 151-record claims archive, then isolate the 37-study evidence layer or inspect both together. A year appears brighter when the selected layer contains more records from it.
18431 record · claims-making text
What it measures. The field plots archive composition and nothing else. The default Claims layer preserves the historical claims record; Evidence isolates publication years; Combined shows both. None measures illness, prevalence, or how often the crisis claim was made in the world.
How it is built. The field has 184 equal cells, one for every year from 1843 to 2026 inclusive. They are arranged 23 across and 8 down and read from left to right, then down. Every year gets the same area whether it holds seven documents or none, so brightness represents density and color represents source type: claims-making text, state, organization, or scholarship. Mixed years blend those colors. The soft glow extends beyond each cell, so exact values should not be estimated by eye. The readout and the Timeline ledger provide the year, record count, and source mix.
What it shows. The Claims layer holds 160 records across 79 years; the Evidence layer holds 37 studies across 25 publication years; Combined occupies 84 years. The scholarship search is recent by design, so its density must never be read as historical frequency. The 1956 claims cell is the first to hold the phrase “mental health crisis” itself.
How to use it. Choose a layer, then tab into its lit years and use the arrow keys, Home, or End. Hover or focus prints the selected year’s exact count and source mix; activation opens its first visible record in the Timeline. Every record remains available at full size in that chapter.
The archive does not estimate a single prevalence trend. It maps studies of symptoms, diagnoses, mortality, service use, access, and classification. These outcomes are not equivalent. The archive asks what each can and cannot establish about a durable national-crisis claim. Four findings organize the five chapters below.
Timeline · the claim accumulates
The archive opens a century before the word crisis, with Dorothea Dix petitioning a legislature in 1843. The phrase itself first appears in a 1956 newspaper headline about Mike Gorman’s Every Other Bed, reaches Congress in 1998, and by 2022 has a State of the Union, a poll putting agreement at nine in ten, and a three-digit number anyone can call.
Explore the timelineFrames · what the claim authorizes
In 1946 it supported federal investment; in 1963, moving care out of the asylum; in the 1990s, parity; and in 2022, permanent national infrastructure. By 2025–26 it was being used to support phone restrictions, reduced prescribing, and the closure of part of 988. The label remains the same even as the proposed actions reverse.
See what the claim authorizedMeasurement · the case changes
Hospital beds, diagnostic categories, four-question symptom screens, self-rated stress and population surveys are not points on one line. Each defines a case differently and identifies a different population, so the four headline figures of the 2020s cannot be assembled into a trend.
Inspect the measuresClaims-makers · authority is divided
In the archive’s coding, the psychiatric profession is the only claimant with an unbroken ownership bar, beginning with the APA’s predecessor in 1844 and later exercised through the DSM. Government can fund, regulate, withdraw, or redirect care without controlling that definition. Authority is usually shared or contested rather than transferred outright.
Map the authorityThe chapters move from chronology to synthesis through frames, measurements, and claims-makers. Links to the complete method appear where its rules are applied.
Follow claims and scholarship in one chronology while reading every affirm, qualify, or dispute judgment at the specific link where it applies.
Compare the public grounds, warrants, conclusions, and counter-claims, then inspect the separate evidence register beneath each era.
Align construct and threshold, instrument, population, recall window, survey mode, and field frame before asking whether anything changed.
Trace ownership across people, professions, government, advocacy, courts, media, and industry.
Follow every selected record through one five-stage trace without treating a financial relationship as proof of motive or influence.
Bring the Timeline, Frames, Measurement, Claims-makers, and Money into one bounded conclusion, then follow the connections across all five.
Public claims, evidence and responses in the United States, 1843 to 2026.
The question the archive asks, the ground it stands on, and the shape it takes.
This archive follows one public claim, that mental illness in the United States amounts to a crisis, from 1843 to 2026.
It asks how that claim was made rather than whether the condition behind it is real, because those are two different questions and only the first leaves a documentary record.
The approach comes from the constructionist tradition in the study of social problems, which runs from Spector and Kitsuse through Joel Best. That tradition asks of any public problem who made the claim, in which arenas, on what grounds, and with what consequences, and those four questions are what each chapter here puts to the record.
Therefore the archive is neither a history of psychiatry nor a history of mental health itself, and it never sets out to establish whether Americans have in fact become more or less unwell. What it does instead is watch one claim travel, and leave every record it used where a reader can check the source.
It is also a passion project, which I built in my own time and continue to pay for myself. The archive carries no affiliation with any individual, company, organization or institution, and nobody outside the project has commissioned, sponsored, reviewed or approved a word of the work.
The reading runs through six numbered rooms, with three lettered companions that stay open beside them.
Six rooms, read in order
Three companions, open at any point
Where the work began, what I did, and what a machine did.
The project began in a research seminar I took in college, Social Problems: Theory and Methods, and the notes I kept in it are the base of everything here. Nobody else has worked on it since, so every decision in the archive is mine, and AI is the instrument I used to build the archive rather than a collaborator on the work.
AI assigned every claim to one of ten analytic frames and five explanatory schemas, and I checked every code against its record before a chapter could use that claim.
Four hundred and ninety survey and vital-statistics estimates were transcribed with their intervals and their breaks, while forty-two claims-maker profiles were written with fifty-one documented ties between them.
Because the selection is mine as well, every record names the source it came from, and the archive opens its selection and coding methods to inspection, so a reader who doubts a reading is able to go and check the source. Read how records were selected.
I am a psychology major rather than a trained engineer, and an archive on this scale is more than one person can hand-code in the hours a personal project gets. Therefore AI is what carried the work the rest of the way, from a folder of documents into an interface a reader can work in.
What a model did not do is decide what the archive claims. Every admission, every code, every question a chapter puts and every limit it states was settled by me, and those decisions were made, unmade and made again across every chapter here.
That is also where the time went, because reading the sources, judging what counted as a record, and settling how a reader should meet them are the parts no model could do for me.
Every inquiry has a boundary. These are part of this one.
This is a historical and interpretive research artifact. It does not assess anyone’s mental health or provide medical advice, diagnosis or treatment guidance.
The archive is a selected body of records and evidence, shaped by explicit research choices and by which sources have survived, so it does not represent every experience, institution or public claim.
Historical diagnoses and contemporary surveys do not measure one stable thing, and changes in definitions, populations and methods limit what can be compared across time. Read the limits of inference.
Recent psychology graduate fascinated by interactive information and research design.
I grew up in Shanghai, China, went to the University of Chicago for my undergraduate degree, and have lived in Chicago since graduating.
I majored in psychology, and the work I care about runs between mental health advocacy, research design, information design and interaction design. Through this archive, I am exploring how the way I organize and present research can help readers examine mental health claims for themselves, while keeping the evidence and its limits in view.
I am a complete amateur in web design and design more broadly, since my formal training was entirely in psychology and psychological research design. After graduating, I taught myself Python and data analytics through Coursera’s Google Advanced Data Analytics Certificate program.
I am looking for full-time opportunities in Chicago, or a remote position anywhere in the world.
I'd love to hear from you.
Research & design · Oliver Zeng.
Asking how a crisis is defined raises a different question: how that distress becomes publicly visible, who gets to explain it, and which responses follow.
How the corpus was assembled, coded, and bounded, including what it cannot establish.
The sections below document the full method: the sources of the records and numbers, the four analytic frameworks, the claims-maker coding, and the limits of the resulting inferences. The rules are presented in full so readers can inspect how the archive reached its conclusions.
The corpus began with ProQuest title searches of U.S. newspapers, using title(mental health crisis) OR title(mental health epidemic). Results were filtered to newspapers and historical newspapers and sorted oldest first, across US Newsstream and the historical files for the Los Angeles Times, New York Times, Wall Street Journal, Washington Post, USA Today, and Chicago Tribune.
An addition had to bear directly on mental health policy or on the public framing of mental illness and take one of these forms:
Legislative history was traced through congressional records, the NIH almanac and NIMH’s own history, presidential proclamations and executive orders, and White House archives. Organizational analysis drew on IRS Form 990 filings for the Kennedy Forum and the American Psychiatric Association, alongside founding histories and institutional structure.
Fifty-six evidence records were added through a structured seed search and an independent saturation pass across three primary streams: definition and classification, population condition, and cause. Forty-five are primary adjudication records; seven care-system and three ontology records remain contextual sensitivity checks. This is an evidence map rather than an exhaustive multi-database systematic review.
Every conventional journal article is coded as scholarship. Books remain claims-making texts, and the CDC’s MMWR surveillance publication remains state evidence. Arena follows the same division: every journal article enters academia, and the MMWR bulletin carries no arena because a surveillance publication is not a journal. Arena still records where an entry entered, never whether its finding holds.
The study and the public event are separate units. Each of the 30 assessed links records the target event; the claim tested; affirms / qualifies / disputes relative to that target; known then / known now relative to the target date; direct, comparative, or contextual scope fit; and confidence. One study can therefore carry a different judgment at different links. Relationship counts describe this coding architecture, not a vote among scholars or a rating of study quality.
Evidence roles remain visible throughout the archive. Forty-six primary records directly adjudicate definition, population condition, or cause. Seven care-system and three ontology records are contextual sensitivity checks: they can qualify interpretation but do not independently drive the project’s population narrative.
The statistical claims in the archive were traced back to their source instruments, each read for how it defines and counts a case:
The headline figure must be read with its definition: what counts as a case, the recall window, and the population. Those choices determine the estimate each instrument can produce.
Four framework families do the archive-wide analytic work. The Frames chapter places Stone’s causal stories inside the Best / Toulmin sequence so the move from evidence to responsibility is explicit.
They analyze claims, not study validity. Best identifies offered grounds; Stone classifies the causal story told; Gusfield and Hilgartner & Bosk trace ownership and arenas; Haslam describes category expansion. None determines whether a study is true or comparable.
A claim has grounds, the evidence offered; a warrant, the value that makes that evidence matter; and a conclusion, what should follow. The causal-configuration loom inserts Stone’s causal bridge between ground and warrant, then keeps competing stories visible inside each era.
A public problem has an owner, the party with authority to create and influence its public definition; a party held causally responsible for the condition; and a party held politically responsible for acting on it. The three need not be the same actor. Ownership is contested rather than settled, so assertion, struggle, transfer and disowning are all part of the process.
A social problem must find room in a limited set of institutional arenas with finite carrying capacity: the press, the legislature, the courts, and the professions. The model also supplies two positions: the operative, who actively works the problem, and the stakeholder, whose interest is at stake in it without a claim being made.
Best supplies the general phenomenon, domain expansion: a problem, once established, becoming a foundation for further claims. Haslam gives it two directions. A concept creeps vertically when it reaches down to milder instances, by lowering the severity threshold or loosening the criteria; it creeps horizontally when it reaches out to contexts it did not previously cover.
Claims-makers are coded by the role they play in the problem’s structure rather than by type. Three roles are Gusfield’s, set out in the method: a public problem has an owner, a party held causally responsible, and a party held politically responsible, and these need not be the same actor. Two further positions come from Hilgartner and Bosk: the operative who works the problem, and the stakeholder whose interest is at stake in it. Coding by position rather than by type lets an organization and an individual sit in one frame, so whether a claims-maker is a person or an institution is recorded on the card rather than treated as an analytic axis.
The five roles are defined in full beneath the role filter in the claims-makers section, where they can be read next to the cards they apply to.
The roles are analytic positions, not fixed actor types, so one claims-maker may hold two at once. For example, the White House can define the crisis while also accepting political responsibility for it. The coding records a primary role and, where supported, a secondary role. The position map shows professional standing and material stake as separate scales.
The horizontal axis records an actor’s standing within the profession that defines the condition. This is distinct from authority within the state. As a result, a government department sits at the outer edge while a professor of psychiatry sits within the profession.
The map also orders nothing by time; the persistence rails on the cards carry duration, which neither the roles nor the map records. A small number of the biographical details on the claims-maker cards come from general historical record rather than from the study’s own primary sources.
The evidence map includes studies of symptoms, diagnoses, mortality, utilization, and access, but those outcomes are not interchangeable. The archive can report bounded findings within a study; it cannot combine them into one estimate of how much “mental health” changed.
The evidence map was built for conceptual saturation around selected propositions, not exhaustive effect estimation. Its affirms, qualifies, and disputes counts depend on the claims chosen, the search strategy, and the link coding; they must never be read as the balance of all scholarship.
Nineteen studies are directly U.S.-scoped; many others are international, global, or from other national settings. Comparative evidence can test whether a mechanism or measurement problem travels, but it cannot by itself establish a U.S. population trend.
The claims corpus began with title searches of six newspapers and under-samples clinics, statehouses, and churches. The scholarship layer is a structured evidence map rather than a completed PsycINFO, MEDLINE, Web of Science, and historical-index systematic review. Era 6 also remains outside the original 2022 claims-search endpoint.
A 188-record ledger follows the claim through ten eras and places 22 linked evidence studies beside the events they support, complicate, or contradict.
No records match these filters.Every one of the 170 entries is still in the archive. The filter is hiding them.
A curated archive, not a frequency sample.
Claims records and evidence records share a ledger but retain different analytic labels.
Nothing leaves the page except the ↗ links.
Equal decades keep the unified chronology legible while distinguishing its two record families. Every cell reports claims and evidence independently, then shows the selected source mix. Select a decade for representative visible records and open any one in the full ledger.
This figure returns to the 151 claims records within the unified corpus. It asks where public claims entered, after Hilgartner and Bosk, and excludes the 37 evidence studies. They are excluded on sampling, not on principle: the evidence layer was assembled by systematic literature search rather than by the historical claims frame, so folding its 37 academia entries in here would draw a migration toward academia that is a property of the search and not of the period. Each column is one era read as a share of itself, so unequal eras remain comparable. Hover, focus, or select a band to follow one arena across all ten bands.
Claims only; counts, not rates. The eras cover unequal spans, so a taller band means more claims records gathered, never a faster rate. No record is shown moving: arena is coded per claim record, not per claims-maker, so change here means distribution, not one claim relocating.
Select a card to read its definition.
A frame names the version of the problem a claim brings into view. The word crisis can point to illness, institutional harm, unequal access, or something else entirely.
A causal story follows the explanation inside a claim: who or what is said to produce harm, and whether the result is described as intended or unintended. It maps the argument, not whether the cause is true.
Locus records where a claim places the problem: in a person, an institution, a population, other people, the environment, or the response. That placement affects which actions the claim presents as relevant.
The table brings together 30 arguments from the archive. Each is coded by its frame, its account of cause, and where it locates the problem. Switch between these three hinges to see which ideas recur, when they appear, and how they combine.
All thirty configurations
No records match these selections.
The ten frames compare arguments about what has gone wrong in mental health, why, and who should act. They are this archive’s analytical categories; some supporting records do not explicitly name a crisis. Even when speakers share the language of crisis, they can propose opposing remedies.
Construct, instrument, population, outcome, and time must align before estimates can form a trend.
six dimensions must remain comparable
An estimate can indicate change only when the construct and threshold, instrument, population, recall window, survey mode, and field frame remain sufficiently stable, or when a defensible method explicitly models the break. Publication year alone does not establish comparability.
Read the six comparison checks in Measurement.
Patel et al. find substantial PHQ-9 comparability across several U.S. demographic groups; Hlynsson et al. fail to establish convincing temporal invariance in a treatment sample. Together they prohibit both blanket dismissal and blanket comparability.
how a population became countable
The diagnostic manual did not make mental illness countable for the first time. The 1940 draft had already counted 1.8 million neuropsychiatric rejections. DSM-III made criteria more reproducible and portable across clinicians and surveys. That is a reliability achievement; it does not by itself prove that each category is a valid natural disease entity. Category totals are retained because claim-makers cite them, not because they measure prevalence or severity.
| Edition | Categories | What changed | Effect on the threshold |
|---|---|---|---|
| DSM-I · 1952 | 106 | Disorders described as reactions to circumstance, in the vocabulary of wartime psychiatry. | Diagnosis rests on a clinician reading a person in a situation. |
| DSM-II · 1968 | 182 | The manual drops the term “reaction” while retaining broad clinical descriptions. | Explicit diagnostic criteria arrive with DSM-III. |
| DSM-III · 1980 | 265 | Explicit diagnostic criteria specify symptoms, counts, and durations. | Diagnosis becomes more standardized and portable across studies and institutions. |
| DSM-III-R · 1987 | No tally supplied | The revised third edition corrects inconsistencies and clarifies criteria. | Publication history is recorded separately from category counts. |
| DSM-IV · 1994 | 297 | The manual contains nearly three times as many categories as DSM-I. | Category count alone cannot show whether severity thresholds moved. |
| DSM-IV-TR · 2000 | No tally supplied | The fourth edition receives a separately published text revision. | Publication history is recorded separately from category counts. |
| DSM-5 · 2013 | 541 entries; 151 with diagnostic criteria | Some categories merged and the bereavement exclusion was dropped. | Grief after a death can meet depression criteria. These totals include administrative entries and cannot establish a contraction in diagnostic scope. |
| DSM-5-TR · 2022 | No tally supplied | The fifth edition’s text revision adds prolonged grief disorder and clarifies criteria. | Publication history is recorded separately from category counts. |
Reliability / validity. Robins and Guze describe a multi-stage validation program; Dohrenwend shows that standardized field methods retain validity problems; Jablensky separates clinical utility from natural-kind validity. The published tallies use different counting conventions, so they cannot establish expansion or contraction without examination of individual diagnostic rules. See the counting notes.
The manual also expanded across populations. PTSD entered DSM-III in 1980 at the urging of Vietnam veterans rather than clinicians, and was later applied to Holocaust survivors, survivors of sexual violence, and, by 2020, a country described as living through collective trauma. DSM-5 adapted ADHD criteria for adolescents and adults. The 1980 Timeline entry provides the detail and links this change to the 1973 vote.
Era 5 · Universalization and infrastructure · domain expansion
Two kinds of expansion are read from the 2020s corpus. Haslam's vertical creep reaches down to less extreme cases; the archive separately tracks an established crisis claim reaching populations it did not previously cover.
Vertical expansion reaches milder cases through lower severity thresholds or looser criteria. The archive begins with severe illness and institutional failure; by 2020 one source of evidence is a four-question symptom screen asking whether a respondent has felt little interest or pleasure “several days” in the past week. Dropping the bereavement exclusion is the second mechanism: the criteria loosen because a rule is removed, not because a threshold falls. Either change can increase who qualifies, but this archive does not estimate how much of any observed increase it explains.
The archive tracks the established crisis claim reaching young children, LGBTQ youth, college students, Asian Americans amid pandemic-era racism, and post-Roe patients. This is domain expansion across populations, not automatically Haslam's horizontal creep, which requires a qualitatively new phenomenon or context.
four different constructs
Among adults 18–24, 56% reported symptoms. These figures are not directly comparable with the much lower 2019 NHIS benchmark: the instruments, mode, and recall windows differ.
19.9% considered attempting suicide; 9% attempted. 37.1% reported “poor mental health” during the pandemic.
Stress falls with age: 61% of Gen Z adults vs. 40% of boomers. Yet only 19% said their mental health was worse than in 2019.
The same report also publishes a 350:1 patient-to-provider ratio and treatment gaps. Those are care-capacity outcomes, not interchangeable parts of this prevalence estimate.
The profiles record who defined the problem, who was assigned causal responsibility, who was expected to act, and who participated in advancing the claim.
Gusfield locates ownership and responsibility; Hilgartner and Bosk distinguish active participants from the interests they represent. The first listed role drives the filters and maps; actor type remains a descriptive card label rather than an analytic axis.
Persistence is the interval between a claims-maker’s first and last documented presence in this corpus. During that interval, the archive records the actor making or carrying the claim, or being assigned a role in it. Cards are ordered from the longest to the shortest interval. Every rail uses the same 1843–2026 scale.
The bar records duration within the corpus, not influence or continuous activity. Gorman appears for one year despite his role in naming the crisis. The exposé tradition spans 1946 to 1972 through three discrete events, which appear as separate ticks. Ticks count entries that cross-reference the card. If a bar continues beyond its last tick, the card asserts the longer interval and the rail leaves that distinction visible.
A solid tick marks an entry within the declared span. A hollow tick marks an entry outside that span that names the actor. Vatz is mentioned in entries from 1961 to 1998 but makes his own claim once, in 1993. Beers’s organization acts in 1917 through Salmon, which the archive attributes to Mental Health America rather than to Beers. The two marks distinguish an actor’s documented participation from a later or earlier mention.
Spans are taken from each card’s own declared dates. Where a card declares an open end the span closes at 2026, the corpus boundary, not at today’s date; where a card declares no dates at all the span was assigned by judgment, and every such judgment is recorded in the archive’s provenance. The order is an artifact of what was collected: it would change if the corpus did.
The association operates four subsidiaries: a charitable foundation (APAF), a political action committee (APAPAC, described as “our voice on Capitol Hill”), an insurance trust established in 1962, and a property LLC. Some organizational information is available only to members. Stakeholder basis. The association receives revenue from industry financing and sales of the diagnostic instrument it defines.
The archive groups Congress and the White House as the federal claimant responsible for legislation, appropriations, and commissions. It assigns them political responsibility but no standing within the psychiatric profession; federal scientific agencies are coded separately. A 1981 budget law replaced much of the categorical federal program with block grants and shifted responsibility toward the states. The archive codes this as a partial disowning because the federal role continued. Administrations from both parties asserted the claim, while the Reagan administration narrowed the federal role. Executive orders in 2025 and 2026 extend the span to the corpus boundary even as HHS contested the cause. This record supports the finding that the frame persisted institutionally, but it does not establish whether the underlying condition changed. Ownership basis. The 1981 change transferred responsibility without eliminating the federal role.
Three claims pressed by people the system had already classified. Elizabeth Packard, confined in 1860 on her husband’s word alone under a law that required no hearing of a married woman, was found sane by a jury in 1864 and carried the 1867 Illinois act that made commitment depend on a jury trial. The Insane Liberation Front organized in Portland in 1969, followed by groups in New York, Boston and San Francisco, publishing Madness Network News and running an anonymous column on the overuse of antipsychotic medication. Patricia Deegan, a psychologist in recovery from schizophrenia, defined recovery in 1988 as “a transformation of the self.” She distinguished it from a cure or being “simply stabilized or maintained in the community.” Federal policy adopts the word by 1999 and the New Freedom Commission builds a report on it. This card exists because the corpus went outside its own sampling frame: all three records are later context, and a title search of six newspapers returns none of them. Owner basis. Deegan’s is a definitional claim made by a person the diagnostic system had classified, rather than by the system that classified her.
Mental Health America is the oldest advocacy organization in this archive. Its medical director, Thomas Salmon, studied war neuroses in 1917 and recommended psychiatric screening of every recruit. The federal draft applied screening at scale in 1940, twenty-three years after the report. The organization now runs prevention campaigns and publishes The State of Mental Health in America, using data publication as part of its advocacy.
The six newspapers provide records for the corpus and publish claims within it, including the earliest 1913 result, the 1956 headline that first uses the crisis label, the 1993 challenge to that label, the 2020 shadow-pandemic metaphor, and a 2019 letters page. The archive therefore codes the press as both a venue and a claimant. Stakeholder basis. The newspapers also have a commercial interest in the material they publish.
The archive treats the public as the principal stakeholder in claims made on its behalf. Within this corpus, members of the public appear mainly as survey respondents rather than as speakers, with a 2019 letters page and 2022 polls as exceptions. The position map therefore places the public outside the profession and at the values end of the participation axis. This pattern may reflect the sampling frame: the corpus began with title searches of six newspapers. Three ex-patient records were later added as context, and none appeared in that title search. The limited number of direct public claims should therefore be read as a property of the corpus, not as evidence that such claims did not exist.
Federal science agencies exercise authority through research, certification, and treatment approval. NIMH has directed a biomedical research agenda since 1949. In 2013, its director announced that the institute would move research beyond DSM categories because symptom-based diagnosis lacked validity for that purpose. The Surgeon General’s reports of 1999 and 2021 describe the scale of the problem. The FDA approved fluoxetine in 1987, relaxed direct-to-consumer advertising rules in 1997, declined to approve MDMA-assisted therapy in 2024, and finalized psychedelic guidance in 2026. Its authority concerns treatments rather than case definitions. Causal-responsibility basis. Beginning in 2021, the Surgeon General also named potential causes through advisories on youth mental health, social media, and loneliness, followed in 2024 by a call for warning labels on social-media platforms.
Seven rulings from Buck v. Bell through Olmstead upheld compulsory sterilization, established a right to treatment, created a duty to warn, set an evidentiary standard for psychiatric confinement, and classified institutional isolation as discrimination. Courts exercise definitional authority only when a case is brought before them.
Chlorpromazine entered the U.S. market in 1954, shortly before the resident hospital population began to decline from its peak. For most of the archive, the industry appears through its financing of organizations that make public claims. Drug companies supplied 29% of the APA’s $62.5 million in 2006 financing, and NAMI reported about $23 million from pharmaceutical companies between 2006 and 2008, roughly three-quarters of the fundraising reviewed in the Senate inquiry. In May 2023, the industry’s trade association published Fighting the growing mental health crisis, the archive’s only societal crisis claim made by the industry in its own name. Operative basis. The public statement changes the coding from stakeholder alone to stakeholder and operative for that instance.
John F. Kennedy signed the Community Mental Health Act; Edward Kennedy sponsored the 1980 and 1992 acts; Patrick Kennedy led the 2008 parity act and later founded the Kennedy Forum; and Joseph P. Kennedy III advanced the claim in 2018. In 2025, Robert F. Kennedy Jr. entered HHS and identified treatment as a cause. The archive codes that claim separately under HHS because it was made from federal office.
NAMI includes 49 state organizations, a youth advisory body (NAMI Next Gen), corporate and influencer partners, and public events such as Mental Illness Awareness Week and NAMIWalks. Stakeholder basis. Roughly three-quarters of the 2006–08 fundraising reviewed in the Senate inquiry came from pharmaceutical companies.
Across thirty-seven years, these researchers helped establish population measurement as a tool of advocacy and government. Thomas W. Salmon compiled national mental-illness statistics in 1912 for a voluntary organization. C.-E. A. Winslow argued in 1933 that the field should include “those incipient maladjustments with which the physician qua physician never comes into contact.” In 1934, a Johns Hopkins unit began a study of one Baltimore health district; Paul Lemkau, Benjamin Tietze, and Marjorie Cooper published a prevalence estimate for that defined population in 1936. Lemkau later helped place mental hygiene within public health and served on the committee that designed NIMH. Owner basis. Some clinicians rejected symptom inventories as a basis for prevalence. Similar objections later appeared in debates about the PHQ-9.
SAMHSA was created in 1992 to separate service administration from research. It administers the block grant established in 1982, grant programs reauthorized in 2000, the early-psychosis set-aside introduced in 2014, and the 988 crisis line launched in July 2022. In July 2025, the agency ended 988’s specialized LGBTQ+ youth service, stating that the appropriated funds had been spent and that it would no longer “silo LGB+ youth services.” Four months earlier, HHS had announced a proposal to consolidate SAMHSA into an Administration for a Healthy America. The archive codes SAMHSA separately from federal science agencies because its formal role is service administration. Political-responsibility basis. Its responsibilities include both implementing and withdrawing services.
The industry opposed coverage mandates on cost grounds. It gave about $175 million to federal candidates from 1990 onward, while the CBO estimated that the 2008 parity act would increase premiums by about four-tenths of one percent. The 2025 suit that suspended enforcement was filed by The ERISA Industry Committee, an employer trade group, not by an insurer. In 2026, the Labor Department settled with a health plan over inadequate provider networks for $28.3 million. The archive treats this as conduct rather than a public claim and therefore retains the stakeholder code.
Life’s “Bedlam 1946,” The Shame of the States, and the Willowbrook broadcast were three exposés that located the cause in the institution and supported reform claims with images rather than statistics.
Harriet Shetler and Beverly Young helped found NAMI, and Alison K. Malmon founded Active Minds. Each drew on personal experience in organizing a grassroots network.
The Kennedy Forum developed from the Kennedy family’s policy work, while Alison K. Malmon’s campus group became a national nonprofit with replicable chapters and named programs. The archive codes both as claims-making through standardized organizational networks, but it does not infer a shared brand strategy, because the public record does not establish that intention.
Carter chaired the 1977 commission and endowed mental health journalism fellowships; Gore answered Vatz in print and chaired the 1999 conference. Both used proximity to federal office as a basis for public authority without holding elected office.
Szasz described mental illness as a metaphor, Goffman analyzed the “total institution,” and Rosenhan placed eight pseudopatients in twelve hospitals. Together, their work formed a counter-tradition that moved from conceptual criticism to an empirical challenge between 1961 and 1973.
The archive begins with a private citizen presenting first-hand evidence of people confined in cages, closets, and cellars to a legislature. Dix gained congressional support on her fourth attempt, but President Pierce vetoed the bill. The archive codes this as a claim made outside the state and refused by it.
Meyer’s 1906 paper on aftercare argued that early treatment could reduce the severity and recurrence of mental illness. By 1908, he had extended the idea from hospital aftercare to community intervention before illness began. In 1915 he proposed dividing the country into mental hygiene districts that coordinated schools, playgrounds, churches, law-enforcement agencies, and mental-health personnel. The archive treats this proposal as an antecedent of the 1963 Community Mental Health Act, which is why the rail shows a hollow mark at that date. The mark records a later use of Meyer’s idea, not an action he took in 1963.
Beers survived institutional maltreatment, founded the organization that became Mental Health America, and helped establish early outpatient and child-guidance work in 1913.
The book’s subtitle states the claim and names the proposed cause. Within months, Los Angeles Unified voted to ban phones in schools and other states followed. The academic evidence is disputed. This profile rests on one Timeline entry.
HHS Secretary Robert F. Kennedy Jr. raises antidepressants as a possible cause of mass shootings despite the absence of credible supporting evidence. The department later announces a plan to curb psychiatric prescribing and encourage deprescribing among children, and it terminates the 988 line’s LGBTQ+ youth service. This is the archive’s first federal claim to identify psychiatric prescribing as part of the problem rather than only as a remedy. Causal-responsibility basis. The secretary’s statement is a causal claim made from a position of definitional authority; the archive records it without endorsement.
In 1948 she named the mother as the cause of schizophrenia, extending earlier work on the over-protective mother into a doctrine of the schizophrenia-inducing family. Therapy that separated patients from their mothers, and treated mothers to change their behavior, was commonly recommended into the 1970s. Parents organized against this form of blame, and the family movement that founded NAMI in 1979 adopted a strongly biological account of schizophrenia in response. This is the only causal claim in the constellation made from within the profession about the families of people it treated. The archive records the claim without endorsing it.
Gorman was a journalist and lobbyist whose book produced the earliest “mental health crisis” headline found in the corpus and prompted an immediate professional counter-review.
Veterans and their counseling groups presented the case for a post-Vietnam syndrome directly to the APA task force revising the manual, and PTSD entered DSM-III alongside the new operational criteria. The archive codes this as an instance of non-professionals gaining definitional authority. The profile rests on one Timeline entry.
Vatz, a rhetoric professor and ally of Szasz, argued that the crisis was “phony.” The archive uses this as evidence of a dispute over the label itself.
One week after the shooting, the NRA identified people with mental illness as the problem and called for a national registry. The archive codes this as a crisis claim used within a separate gun-policy dispute. The organization had no documented mental-health role elsewhere in the corpus, which is why the frame is analyzed separately from the institutions that developed it.
Three professional bodies representing about 77,000 clinicians jointly declared a national emergency in child and adolescent mental health five months before the State of the Union. This was a collective professional claim made outside the APA’s diagnostic manual. The profile rests on one Timeline entry.
A pollster-authored op-ed described the crisis as bipartisan in January. Nine months later, a CNN/KFF poll found that nine in ten adults agreed the country was in a mental health crisis. The survey produced a measure of public agreement that was then cited as evidence for the claim.
The horizontal axis records a claims-maker’s standing within the profession that defines the condition: no standing, lay petitioner, adjacent expertise, inside the field, or authority to set the rule for what counts as a case. The chapter uses the profession as its reference point because it is analyzing authority over the condition’s definition. This axis replaces an earlier measure that combined professional standing with ownership. The vertical axis records the basis for participation, from values alone to material interest alone. Marker color retains the role coding above, and each name links to its card.
Eight of the thirty-one profiles have no standing within the profession, tying the lay-petitioner category for the largest column. They range from the public at the values end of the vertical axis to insurers and the NRA at the material-interest end. The bench is also in this column, although Tarasoff directly affected clinical practice. These actors use legal, media, survey, budget, and executive instruments rather than professional authority.
A professor of psychiatry, a sociologist of institutions, and a psychologist who conducted an experiment appear in the inside-the-profession column, with Vatz one column to their left. They share that column with Meyer, Fromm-Reichmann, and the survey researchers whose work they criticized. The column therefore measures credentialed participation, not agreement. After 2012, the NRA and HHS reassign the cause without rejecting the crisis label.
The American Psychiatric Association and the World Health Organization occupy the rule-setting category because they maintain the diagnostic classifications used by the field. State actors occupy four other positions. NIMH and the FDA sit inside the profession, while SAMHSA is placed under adjacent expertise. Congress, the White House, and HHS in 2025 have no professional standing. These placements distinguish scientific, administrative, and political authority within the state.
Ownership is not a fixed label here. The map asks who held the authority to define the problem, how that authority was challenged, and where it moved.
Bars show the duration of authority held, claimed, or relinquished. Forty-seven linked marks show the coded operations that changed it. The unbroken professional bar contrasts with the shorter intervals in every other lane.
The political state, federal science agencies, and HHS under the second Trump administration occupy separate lanes because obligation, professional standing, and a claim made over the profession are not the same kind of authority.
The industry lane has one qualifying claim in its own name and no duration bar. Its near-emptiness describes the corpus first. Time is compressed before 1940 because five operations occupy the first century and forty-two occupy the last eighty years.
The lane carries one qualifying mark and no duration bar. Its sparse coverage reflects
the sources this corpus searches and the claims its inclusion rule accepts. The seed corpus is a
title search for mental health crisis; the method’s own test (see
§05) separates a societal claim, that a population is in crisis, from
an individual one, that a person needs urgent help. Only the societal
claim, made about the nation in the claimant’s own name, would populate this lane.
Most of what industry says fails that test on one axis or the other.
Advertising is the individual sense, not the national one. Pfizer’s 2001
Zoloft campaign told a mass audience that depression may be due to a chemical
imbalance
the drug corrects. That is a claim about an individual condition.
The archive tracks disease definition elsewhere; this lane requires a claim that
the country is in crisis. The campaign names its sponsor but does not make the
national claim, so it would not qualify even if advertising were part of the corpus.
The disease-branding campaigns are the profession’s name, not industry’s. The 1999 campaign that built social anxiety disorder into a mass condition was funded by SmithKline Beecham but issued under a coalition carrying the American Psychiatric Association and two advocacy bodies. Where the corpus samples it, the claim is attributed to the profession and advocacy because those organizations issued it. The lane records the public claimant; the financial record identifies the industry sponsor.
Industry’s national claim in its own name. One instance qualifies:
the pharmaceutical trade
association PhRMA published a 2023 statement titled Fighting the growing mental
health crisis
, a societal claim about the nation issued in its own name.
It is admitted as a context document and placed as the industry lane’s one mark,
an assertion with no bar beneath it because a single statement is not a duration.
A second, adjacent case is already carried inside the 2026 MAHA entry: the APA’s
press release speaks of the nation’s mental health crisis
as settled fact
while contesting the cause. This is a professional organization making the same
kind of national claim.
What the PhRMA statement changes. Stakeholder coding still depends on interest rather than whether an actor speaks. Advertising and the 1999 coalition do not qualify for this lane because they concern individuals or appear in another actor’s name. PhRMA’s statement does qualify and corrects the earlier finding that industry never makes a national crisis claim in its own name. The corpus searches six newspapers’ title fields, which do not index trade-association statements, comment-letter dockets, or advertising. PhRMA therefore enters as an added context document. Broader coverage would require a change to the inclusion rule to admit those source classes; this archive has not made that change.
The federal government took up the problem in 1946 and enlarged its role as late as 1980. A year later, the Reagan budget law replaced much of that categorical program with block grants. Every later federal act partly restores a role narrowed in one budget reconciliation.
The second break is quieter. In 2013 NIMH announced that its research program would move away from DSM categories because diagnosis based on symptoms alone lacked validity. The profession kept the manual, while a federal research agency stopped treating its categories as the only starting point.
The profession is the only claimant with an unbroken bar from 1844 to the present. Courts, press, families and the state each hold a claim and then recede. This makes the 2025 move onto psychiatry’s ground the sharpest ownership contest since 1973.
The 501(c)(6) classification is a legal fact. Fourteen consecutive years of filings show its financial operations over time. The manual carries definitional authority inside a large publishing business. In 2013, the year DSM-5 appeared, aggregate publication sales produced the association’s largest revenue spike in this series.
Across the series the association’s net assets rise from $41.1 million in 2012 to $128.8 million in 2024, a 3.1× increase. Program service revenue is the steady floor; publication sales are the spike. The five streams decompose total revenue exactly, so each year’s percentages sum to 100. The two revisions in the window, DSM-5 and DSM-5-TR, are the two years the spike appears.
In April 2013, the director of the National Institute of Mental Health announced that the institute would move research away from DSM categories because symptom-based diagnosis lacked validity. DSM-5 was published the following month. In that calendar year, publication sales reached $42.4 million, or 52% of the American Psychiatric Association’s revenue. The filing does not identify sales by title, so it cannot assign the entire amount to DSM-5.
The archive codes these as two entries in two lanes. They are one moment: the federal science agency moving its research beyond the instrument in the same year that publication revenue peaked for the body that owns it. Insel challenged the categories’ research validity; nothing in the filings suggests the manual’s institutional authority moved at all.
This bears on where the association sits on the position map. It was placed at chiefly material interest rather than only, reasoning that 29% industry financing leaves 71% that is not industry financing. That reasoning weighed the wrong stream. The material interest includes both industry financing and publication revenue. The publication stream includes the manual the association defines and sells, although the filings do not isolate that title. The placement is unchanged here; the evidence for a material interest is broader than the original justification claimed.
The revenue chart above shows where the APA’s money came from by category, and by that measure the association looks commercially independent: overwhelmingly earned, barely donated. A Senate inquiry two years earlier shows what those categories can conceal.
| Organization | Period | Pharmaceutical share | Source |
|---|---|---|---|
| American Psychiatric Association | 2006 | 29% of $62.5m in financing | Figures confirmed during Sen. Grassley’s 2008 inquiry into industry funding of medical associations. |
| NAMI | 2006 – 08 | ~$23m, about three-quarters of fundraising reviewed | Senate Finance Committee inquiry, 2009. NAMI’s own reply put drug-company money at an average 56% of its national budget for 2005 – 09. |
| 33 health advocacy groups | 2009 – 10 | disclosure requested | Grassley widened the inquiry beyond psychiatry after the NAMI findings. |
The stakeholder category looked empty because the pharmaceutical industry was not usually making a rival public claim beside NAMI and the APA; for several of these years, it was funding them. NAMI received roughly three-quarters of the fundraising discussed in the 2006–08 inquiry from companies that sold treatments for the conditions it publicized. Its executive director told the New York Times that “the percentage of money from pharma has been higher than we have wanted it to be.” Those disclosures immediately precede the 2008 parity act and the 2013 founding of the Kennedy Forum, as the claim moved from advocacy into statute. The insurance industry appears on the other side of the ledger. Its documented objection was cost: the CBO estimated that the 2008 act would raise premiums by about four-tenths of one percent.
“Mental health crisis” was searched on each organization’s own website. The phrase carries two senses, and the distribution between them is the finding. A societal use claims that a population is in crisis; an individual use denotes an episode in which one person needs urgent help. The test applied to each was whether the sentence still works when the phrase is replaced with “a national problem” or with “a psychiatric emergency.” One page per organization was then read in full and coded against a rule written beforehand.
| Organization | Form | Sense | The phrase in its own words |
|---|---|---|---|
| NAMI | 501(c)(3) | Individual | Fourteen uses on its crisis-response policy page, all of them episodes: “a mental health crisis deserves a mental health response,” and “if law enforcement encounters someone in a mental health crisis.” Its figure of one in ten adults reporting a crisis in the past year counts individual episodes rather than describing the country as being in crisis. |
| Mental Health America | 501(c)(3) | Individual | Its position statement on responding to crises uses the phrase only of persons: mobile crisis teams, peer respite, avoiding handcuffs. Unemployment, poverty and homelessness appear, but as causes of individual episodes. |
| Active Minds | 501(c)(3) | Societal, youth-scoped | “The mental health crisis experienced by youth and young adults has become one of the most pressing public health issues of this time.” A population claim, though scoped to the young rather than to everyone. |
| American Psychiatric Association | 501(c)(6) | Societal | Its statement on the 2022 State of the Union uses the possessive in its headline, “Our Mental Health Crisis,” and its president adds that the crisis “is, indeed, something that affects us all, regardless of politics, geography, race or ethnicity.” |
The split is not between charities and trade associations: Active Minds is a charity and still makes the societal claim. Instead, NAMI and Mental Health America use the phrase almost entirely for an individual episode. Their pages are built to help a person in crisis, and that person is the subject of the phrase. The societal claim comes from the professional association and the youth organization, neither of which centers its work on a helpline. An earlier search-result pass coded Active Minds as mixed; reading the full page changed the result. That correction shows what the snippet-level method missed.
What the chapters establish together, and where the evidence leaves questions open.
Explore the history of crisis claims, the arguments behind them, the measures used to support them, the people and institutions making them, and the money committed in response. The four findings that follow draw these perspectives together.
00Timeline01 / 05
The Timeline brings together 160 records of public claims and their context, alongside 37 research records, from 1843 to 2026. Twenty of the public records use the phrase “mental health crisis”. The archive assesses 30 connections between studies and specific claims: the evidence adds qualifications in 22, supports the claim in seven and challenges it in one.
01Frames02 / 05
Frames asks what each argument identifies as the problem, what it says caused that problem, and whom it holds responsible. Half of the 30 arguments analyzed attribute the problem to unintended consequences of earlier actions. Only two locate the problem within the individual.
02Measurement03 / 05
Measurement traces how definitions and survey methods shape the numbers, with 38 series and 509 published estimates from nine systems. In one comparison across 44 studies, 24.6 percent of participants screened positive for depression on a questionnaire, while 12.1 percent met the criteria in a diagnostic interview.
03Claims-makers04 / 05
Forty-two profiles and 51 documented relationships show who made, promoted or contested crisis claims. The American Psychiatric Association (APA), including its predecessor, is the only organization shown claiming authority throughout the ownership map, which tracks who asserts or gives up responsibility for the problem. The federal government takes on responsibility in 1946, reduces its role in 1981 and expands it again from 1990.
04Money05 / 05
Money examines 34 records of spending, payment rules, organizational resources and funding withdrawals. It asks where money came from, who controlled it, where it went, how it was used and what changed. Treatment for mental disorders accounted for 3.4 percent of national health spending in 2000 and 4.1 percent in 2021. Growth in the number of people treated accounted for most of the increase in spending.
Read the findings alongside the evidence from each chapter. Follow the links to explore the records and see what supports each conclusion.
Together, the findings show how the crisis is defined, measured and addressed. Comparable measurements can reveal change in a particular population. Measures of different conditions and populations cannot be combined into a single verdict on the nation’s mental health.
01/ 04
The archive’s earliest crisis headline dates from 1956. Its Congressional Record search first finds the exact phrase in 1998; the phrase appears in the State of the Union in 2022. Within weeks in 2026, the Department of Health and Human Services invoked the crisis both to accelerate access to psychedelic treatment and to reduce psychiatric prescribing.
Calling the problem a crisis does not settle what caused it or how to respond. Even within one department, the label can justify expanding some treatments and reducing others.
A shared goal has supported conflicting responses before. The prevention argument behind the child-guidance clinics that opened in 1922 was used in 1927 to justify compulsory sterilization. To assess a proposed response, ask what problem it is meant to solve and whose interests it serves.
The archive began with title searches of six newspapers. Its selected records document uses of the phrase; they cannot establish its first use anywhere, how often Americans used it, or how much distress lay behind it.
Twenty of the 160 public records use the phrase “mental health crisis”; 18 of those date from 1998 or later.
Read the 1956 headlineSeven of the 30 arguments analyzed date from 2023 onward. They include claims that treatment is being withheld and claims that it is being overused.
Inspect the frame register02/ 04
From 2021 to 2024, the share of adolescents reporting a major depressive episode fell from 20.8 to 15.4 percent. Persistent sadness or hopelessness among high-school students fell from 42 percent in 2021 to 33 percent in 2025. In the National Survey of Children’s Health, the share of adolescents aged 12–17 whose parents reported a current anxiety diagnosis rose from 14.8 to 16.1 percent between 2021 and 2023.
The measure and year matter when assessing a crisis claim. The federal Make America Healthy Again Commission’s 2025 report cited the 2021 figure of 42 percent. That figure describes students’ responses in 2021 rather than conditions at the time of the report.
In the earlier survey, reported adolescent depression nearly doubled between 2010 and 2019, while the adult rate rose by one percentage point. The research in the archive provides stronger evidence for that increase than for any single explanation of its cause.
These surveys ask different questions and rely on different respondents. The depression survey was redesigned in 2021, limiting comparisons with earlier years. The 2025 high-school survey had a 30.6 percent response rate, which raises concerns about how well respondents represent all students.
Reported adolescent depression rose from 8.0 to 15.7 percent between 2010 and 2019. In the redesigned survey, it fell from 20.8 percent in 2021 to 15.4 percent in 2024.
Open the youth seriesThe federal Make America Healthy Again Commission’s 2025 report cited a 2021 survey in which 42 percent of high-school students reported persistent sadness or hopelessness. The survey for 2025 found 33 percent.
Read the 2025 survey03/ 04
The American Psychiatric Association (APA) publishes the Diagnostic and Statistical Manual of Mental Disorders (DSM), which defines psychiatric diagnoses. Governments and courts shape how care is funded and delivered. Medicaid’s 1965 funding exclusions encouraged care outside state psychiatric hospitals. From 1971, court decisions required treatment, not just confinement.
In 2026, the U.S. Department of Labor reached a settlement with Kaiser Foundation Health Plan over delays and barriers to mental-health care. It provided at least $28.3 million for members forced to seek care outside the plan’s network. The underlying duty to cover mental-health care on comparable terms to medical and surgical care remained enforceable, even after enforcement of a newer parity rule was suspended.
The boundaries of these roles are contested. Since 2025, the federal administration has argued that excessive diagnosis and treatment contribute to the crisis. Its 2026 prescribing plan includes payment guidance for reducing or stopping medication. The APA’s response disputes the plan’s account of psychiatric treatment.
The archive documents institutional roles and relationships. A funding connection or proximity on the network map does not establish that one institution caused another’s decision.
The American Psychiatric Association and its predecessor claim authority throughout the map of institutional responsibility. Of 41 recorded actions, one is classified as a transfer of responsibility: to the federal government in 1946.
Inspect the ownership mapIn 1981, federal block grants gave states more discretion over mental-health spending, shifting much of the responsibility for services to state governments.
Read the Community Care sequenceThe Labor Department’s 2026 settlement with Kaiser Foundation Health Plan addressed barriers to mental-health care after enforcement of a newer insurance parity rule had been suspended.
Read the 2026 settlement04/ 04
Growth in the number of people treated accounted for most of the increase in mental-disorder treatment spending between 2000 and 2021. The share of adults receiving counseling rose from 9.5 percent in 2019 to 14.0 percent in 2024. The Money records document changes in capacity, rules and responsibility, but do not establish who gained access to care as a result.
The amount committed to a crisis response cannot, by itself, show whether it reached the people it was meant to help.
The 2021 American Rescue Plan and 2022 Bipartisan Safer Communities Act committed federal money to mental-health services. The archive documents funding transfers and proposed uses. Assessing what those funds achieved requires further evidence about the care people received.
Rising treatment use may reflect changes in access, need or decisions about when to offer treatment. These figures do not separate those influences. Missing evidence of an outcome does not mean there was no benefit.
Of the atlas’s 36 entries, 25 document a change in capacity, rules or responsibility. These changes do not establish whether access to care improved.
Inspect the five-stage tracesThe share of adults receiving counseling rose by nearly half between 2019 and 2024, measured with the same questionnaire design.
Open the service-use seriesThe chapters show how “mental health crisis” gains authority through the people and institutions that define problems, publish figures and organize responses. To judge a claim, ask what is being measured, whose experience it captures and how it has changed. To judge a response, follow the money and the decisions through to the care people received and the results it achieved.