the DSM

1. what it is

The Diagnostic and Statistical Manual of Mental Disorders is a classification manual. It lists categories of mental disorder and, for each, the criteria a clinician applies to decide whether a person meets it. In the American Psychiatric Association's own description, DSM-5-TR is “the standard classification of mental disorders used by mental health professionals in the United States” [2].

That is a narrower job than people assume. The manual does not explain what causes any of the conditions it lists, and it does not claim to. It says: when these things co-occur, call it this. The whole of the rest of this page follows from that one fact.

Its ancestry is administrative rather than clinical. The APA traces the American impulse to classify back to the census: “the recording of the frequency of ‘idiocy/insanity’ in the 1840 census”, and by 1880 seven categories — mania, melancholia, monomania, paresis, dementia, dipsomania, and epilepsy [1]. The first edition arrived in 1952 as an American variant of the World Health Organization's ICD-6 [1].

every edition, and what changed

Each row below is the APA's own account of its own manual, or a peer-reviewed history, quoted rather than summarised. Where a claim about an edition could not be verified from a primary source, it is not in the table.

the “explosion in diagnoses” number, and why it does not say what people think

You will constantly see a figure for how many diagnoses each edition contains, usually quoted as a rise from about 106 to about 300 and offered as evidence that psychiatry keeps inventing illnesses. There is a citable peer-reviewed series [23]:

“The number of diagnoses in DSM-III, however, climbed to 265 from the 106 diagnoses included in the first edition of the DSM criteria. This rapid growth in numbers of diagnoses slowed, however, yielding 292 diagnoses in DSM-III-R, 297 in DSM-IV and DSM-IV-TR, and 298 in DSM-5.”

Read the second sentence. The growth is a 1952-to-1980 story. Since DSM-IV in 1994 the count has moved by one. An independent history puts DSM-II at 182 and DSM-III at 265, consistent with the same arc [24]. Anyone using this statistic to describe modern psychiatry is using a thirty-year-old number to describe today.

And the caveat that has to travel with it. Neither source states a counting rule — whether subtypes, “other specified” residual categories and conditions-for-further-study are counted — and both trace upstream to the same 2005 paper, which is closed access and whose public abstract contains no numbers at all. So this is one count reported by two papers, not two independent counts, and it is not reproducible by you or by us. We print it because the shape of it corrects a widespread error, and we print this paragraph because the figure cannot carry more weight than that.

and it is still moving between editions

“Which edition are we on” stopped being a complete answer. The APA now runs a standing proposal-and-comment process: anyone can be invited to “propose changes, corrections and clarifications to the DSM-5-TR”, proposals approved by the DSM Steering Committee go out for public comment, and criteria and text updates are issued between editions. The most recent set on the record is September 2025; a proposal to clarify the Autism Spectrum Disorder severity specifiers was open for public comment until 30 April 2026 [2].

The practical consequence: any prevalence statistic you read carries an implicit date, and the definition behind it may have moved since it was measured.

2. who creates it

The American Psychiatric Association — a professional membership association of physicians, headquartered in Washington DC. Not the American Psychological Association, which is a different organisation with no role in the DSM. That confusion is the single most common error in writing about this subject, including in writing that is otherwise hostile to the manual, and it hands the argument away for free.

The structure is a task force plus disorder-specific work groups. In the APA's own words [1]:

“The work on DSM-5 began in 2000, work groups were formed to create a research agenda for the fifth major revision of DSM. … In 2007, APA formed the DSM-5 Task Force to begin revising the manual as well as 13 work groups focusing on various disorder areas. DSM-5 was published in 2013.”

Approval runs through the association's own governance rather than any external body. For DSM-5-TR the APA records that criteria clarifications “were reviewed and approved by the DSM Steering Committee, as well as the APA Assembly and Board of Trustees” [1]. There is no regulator in that sentence. The FDA does not approve the DSM; no government agency ratifies a criteria set. A professional association decides, and the decision then propagates into insurance billing, disability determinations, school accommodations, research eligibility and clinical practice.

how a criterion actually gets decided

Two mechanisms, and it is worth being precise about which does what.

Committee consensus sets the criteria. This is not a slur — it is the APA's own description of what DSM-III did, and a peer-reviewed history puts it plainly: “The essential change in DSM-III was to replace the idiosyncratic diagnoses of psychoanalysis with diagnoses that were consensus-based” [3]. Consensus was the improvement. Before it, two psychiatrists could look at the same patient and reason from different theoretical schools to different conclusions with nothing to arbitrate between them.

Field trials then test whether the written criteria work in a clinic — specifically whether two clinicians, interviewing the same patient separately and blind to each other, land on the same diagnosis. That is reliability. It is not a test of whether the category is real.

what the DSM-5 field trials found

The DSM-5 field trials ran at eleven academic centres in the United States and Canada. Here is the published summary of the result, quoted from the abstract [6]:

“There were a total of 15 adult and eight child/adolescent diagnoses for which adequate sample sizes were obtained to report adequately precise estimates of the intraclass kappa. Overall, five diagnoses were in the very good range (kappa = 0.60–0.79), nine in the good range (kappa = 0.40–0.59), six in the questionable range (kappa = 0.20–0.39), and three in the unacceptable range (kappa values < 0.20). Eight diagnoses had insufficient sample sizes to generate precise kappa estimates at any site.”— Regier, Narrow, Clarke, Kraemer, Kuramoto, Kuhl & Kupfer, American Journal of Psychiatry, 2013. Regier was vice-chair and Kupfer chair of the DSM-5 Task Force [6].

Read the labels, not just the numbers. The bands are the field trials' own labels, and “good” there means two clinicians agree at a kappa between 0.40 and 0.59.

the bands were set in advance, in public, by the same team

Twelve months before those results were published, the same journal carried a paper by the field-trial team — Kraemer, Kupfer, Clarke, Narrow and Regier, four of whom are also authors on the results paper — titled “DSM-5: How Reliable Is Reliable Enough?”. It set the target [7]:

“From these results, to see a κ_I for a DSM-5 diagnosis above 0.8 would be almost miraculous; to see κ_I between 0.6 and 0.8 would be cause for celebration. A realistic goal is κ_I between 0.4 and 0.6, while κ_I between 0.2 and 0.4 would be acceptable.”— Kraemer, Kupfer, Clarke, Narrow & Regier, American Journal of Psychiatry, January 2012 [7]. The results paper ran in the same journal in January 2013 [6].

The standard was published before the results, under the names of the people who would be measured by it. That sequencing is the substance of the criticism, and it can be demonstrated entirely from the two papers' own dates and author lists. No polemic required, and none offered.

Their reasoning deserves to be here too, because it is good. The same paper argues that the comparison everyone makes is against a fantasy version of the rest of medicine [7]:

“While one occasionally sees interrater kappa values between 0.6 and 0.8, the more common range is between 0.4 and 0.6. … The diagnosis of anemia based on conjunctival inspection was associated with kappa values between 0.36 and 0.60, and the diagnosis of skin and soft-tissue infections was associated with kappa values between 0.39 and 0.43. The test-retest reliability of various findings of bimanual pelvic examinations was associated with kappa values from 0.07 to 0.26. … many believe that the rates of reliability and validity of diagnoses in other areas of medicine are much higher than they are. Indeed, psychiatry is the exception in that we have paid considerable attention to the reliability of our diagnoses.”

Both things are true at once: the bar was set by the people it would judge, and the bar is roughly where the rest of medicine actually sits when anyone bothers to measure. Section 4 is about exactly that tension.

The limit on what we can show you here, stated rather than hidden. The field-trials results paper is not open access — Unpaywall returns is_oa: false, oa_status: closed, checked 2026-08-26 — and the per-diagnosis kappa values live in tables inside the paywall. So we quote the abstract's summary bands, which are public and checkable, and we do not quote a kappa for any individual diagnosis. You will see specific figures circulated widely; the most common is a 0.28 for major depressive disorder, which a peer-reviewed history does report while citing the field trials [23] — but that is a secondary report of a number we could not read in the primary source, so it appears here labelled and nowhere else on the page. The Kraemer paper [7] has no published abstract either; it is quoted above from an Internet Archive capture of the publisher's own full text, dated 12 May 2020.

3. based on symptoms, not underlying biology

The best statement of this is not from a critic of psychiatry. It is from the man running American psychiatric research at the time, published weeks before DSM-5 came out. Thomas Insel was Director of the National Institute of Mental Health from 2002 to 2015 [9]; this post is dated 29 April 2013 [8]:

“While DSM has been described as a ‘Bible’ for the field, it is, at best, a dictionary, creating a set of labels and defining each. The strength of each of the editions of DSM has been ‘reliability’ – each edition has ensured that clinicians use the same terms in the same ways. The weakness is its lack of validity. Unlike our definitions of ischemic heart disease, lymphoma, or AIDS, the DSM diagnoses are based on a consensus about clusters of clinical symptoms, not any objective laboratory measure. In the rest of medicine, this would be equivalent to creating diagnostic systems based on the nature of chest pain or the quality of fever.”— Thomas Insel, Director, National Institute of Mental Health, “Transforming Diagnosis”, NIMH Director's Blog, 29 April 2013 [8]

And the sentence from the same post that keeps the criticism honest, which almost nobody quotes beside it [8]:

“It became immediately clear that we cannot design a system based on biomarkers or cognitive performance because we lack the data.”

That is the whole register of this page in one line. Nobody chose description over explanation out of laziness or venality. The explanation is not available. We treat what we can describe, not what we understand.

Where we read this. The post is no longer on nimh.nih.gov; the original URL returns 404. We quote it from an Internet Archive capture dated 27 June 2021 [8]. That is the second time in this series we have had to go to a private archive for a primary US government document — the FDA's pre-2020 safety communications are the other. Worth noticing as a pattern.

the RDoC pivot — and what it has and has not delivered

Insel's answer was not to abolish diagnosis. It was to start building the data a different kind of diagnosis would need. In the same post he announced that “NIMH will be re-orienting its research away from DSM categories” under the Research Domain Criteria project, and gave the reason with an analogy worth keeping [8]:

“Imagine deciding that EKGs were not useful because many patients with chest pain did not have EKG changes. That is what we have been doing for decades when we reject a biomarker because it does not detect a DSM category.”

Fairness requires the next sentence too, and it is Insel's own [8]:

“RDoC, for now, is a research framework, not a clinical tool. This is a decade-long project that is just beginning.”

Thirteen years on, RDoC is still a research framework. Nobody is diagnosed by it. If you see RDoC described as psychiatry's replacement for the DSM, that is not what its author said it was.

Hyman on reification: the danger of forgetting the categories were written

Steven Hyman ran NIMH before Insel — Director from 1996 to 2001 [9]. His 2010 paper in the Annual Review of Clinical Psychology is titled “The diagnosis of mental disorders: the problem of reification”. Reification is the error of treating a description as though it were a discovered thing. From the public abstract [10]:

“Unfortunately, the focus on reliability came at a time when the scientific understanding of mental disorders was embryonic and could not yield valid disease definitions. … Yet DSM-IV diagnostic criteria dominate thinking about mental disorders in clinical practice, research, treatment development, and law. As a result, the modern DSM system, intended to create a shared language, also creates epistemic blinders that impede progress toward valid diagnoses.”

Note who is speaking, again: a former director of the institute that funds most of the research. The line of argument on this page has been made from inside the building for thirty years.

What we could not read. The Hyman paper is not open access (Unpaywall: closed, checked 2026-08-26), so everything quoted above comes from the public abstract, and nothing from the body of the paper appears on this page [10].

so is there a blood test? the precise answer

The popular version of this claim — no objective test exists for anything in the DSM — is false, and we are going to say so before someone else does, because the accurate version is stronger and survives contact with an expert.

Three things are true at once:

• The DSM does use objective measures where they exist. The sleep-wake chapter is the proof. DSM-5's narcolepsy criteria accept “Hypocretin deficiency, as measured using CSF hypocretin-1 immunoreactivity values … less than or equal to 110 pg/mL” as one route to the diagnosis [12] — a fluid biomarker with a numeric cutoff, written into the manual. DSM-5's obstructive sleep apnea hypopnea criteria go further and require polysomnography evidence on every route, including one that needs “15 or more obstructive apneas and/or hypopneas per hour of sleep regardless of accompanying symptoms” [13]. That is a DSM diagnosis made by measurement alone.
• The FDA has authorised two diagnostic aids in this space, and barred both from standalone use. A neuropsychiatric interpretive EEG assessment aid for ADHD was granted de novo authorisation in July 2013 (DEN110019); its classification regulation states it “is used only as an assessment aid for a medical condition for which there exists other valid methods of diagnosis”, that the design “must include safeguards to prevent use of the device as a stand-alone diagnostic”, and that labelling must carry “a warning that the device is not to be used as a stand-alone diagnostic” [14]. A pediatric Autism Spectrum Disorder diagnosis aid followed in June 2021 (DEN200069), whose regulation requires labelling to state “the device is not intended for use as a stand-alone diagnostic” and describes what it measures as “an interpretation of patient behavioral symptomology” [15] — a machine-assisted symptom rating, not a biological marker.
• And for the disorders anyone is actually arguing about, there is nothing. A 2012 paper in Molecular Psychiatry co-authored by Insel himself, titled “Why has it taken so long for biological psychiatry to develop clinical tests and what to do about it?”, opens: “Patients with mental disorders show many biological abnormalities which distinguish them from normal volunteers; however, few of these have led to tests with clinical utility” [11]. Note the word is few, not none — we are quoting it as written.

So here is the claim, stated so that it is true:

Every word there is doing work. Confirm or rule out is the standard, because supportive and risk-stratifying findings do exist. Validated matters because biological differences in these conditions are abundant and well replicated — what is missing is a test. And the four named diagnoses turn an unfalsifiable universal into four specific things a critic would have to disprove.

4. why this differs from the rest of medicine — carefully

This is the section most likely to be used against us, so read the concession first. “Psychiatry diagnoses by symptoms while the rest of medicine diagnoses by biology” is not true, and anyone who says it can be dismantled in under a minute by a doctor. Large parts of medicine diagnose exactly the way the DSM does. We are going to show you that first, because what is left over afterwards is the real claim and it is a stronger one.

how much of medicine is syndromic too

• Migraine. The International Classification of Headache Disorders, 3rd edition, defines migraine without aura as “At least five attacks” lasting “4-72 hr” with “at least two of the following four characteristics” and “at least one of” nausea/vomiting or photophobia and phonophobia [16]. Symptom counts and thresholds. No laboratory test appears anywhere in the criteria.
• Irritable bowel syndrome. Rome IV: “Recurrent abdominal pain on average at least 1 day/week in the last 3 months, associated with two or more of the following criteria” [17]. Again: frequency thresholds and symptom counts.
• Sepsis — and this one is the closest analogue of all. The Sepsis-3 consensus definitions were produced by “A task force (n = 19) … convened by the Society of Critical Care Medicine and the European Society of Intensive Care Medicine” through “meetings, Delphi processes, analysis of electronic health record databases, and voting”, and the paper contains a section headed “Assessing the Validity of Definitions When There Is No Gold Standard” which states plainly: “Because no gold standard diagnostic test exists, the task force sought definitions and supporting clinical criteria” [18]. It even records the vote: “The majority (n = 14/17; 82.4%) of task force members voting on this agreed that hypotension should be denoted as a mean arterial pressure less than 65mmHg” [18].
• Parkinson's disease. The Movement Disorder Society's clinical diagnostic criteria state that “The benchmark for these criteria is expert clinical diagnosis” [21].

and definitions move prevalence outside psychiatry too

The complaint that a committee moved a threshold and the number of sick people jumped is not a psychiatric complaint. In 2017 the American College of Cardiology and American Heart Association moved the definition of hypertension from ≥140/90 to ≥130/80 [19]. An analysis by the guideline's own chair and vice-chair reported the consequence [19]:

“the overall crude prevalence of hypertension among US adults was 45.6% … and 31.9% … respectively” — that is, “103.3 million US adults met the definition for hypertension according to the 2017 ACC/AHA guideline compared with 72.2 million US adults according to the JNC7 guideline”.

Roughly 31 million Americans acquired a disease overnight without anything happening to their arteries. And the honest counterweight, from the same paper, because it is the thing that stops this being a cheap point: “An additional 4.2 million US adults were recommended antihypertensive medication”, and the guideline “results in a substantial increase in the prevalence of hypertension but a small increase in the percentage of U.S. adults recommended antihypertensive medication” [19]. They were relabelled, not medicated. That distinction is exactly the one this whole page is about.

so what actually is different

Scope, and timing. Other specialties diagnose syndromically in places. The DSM does it everywhere. And other fields have been crossing over, one disease at a time — which is visible precisely because the crossing has to be announced.

In 2024 an Alzheimer's Association workgroup published revised criteria whose highlight line reads: “An abnormal Core 1 biomarker result is sufficient to establish a diagnosis of AD and to inform clinical decision making throughout the disease continuum” [20]. Its framing principle is stated in the abstract: “Defining diseases biologically, rather than based on syndromic presentation, has long been standard in many areas of medicine (e.g., oncology), and is becoming a unifying concept common to all neurodegenerative diseases” [20].

Read that as what it is: a field that had to write a paper in 2024 to establish that its central disease could be diagnosed biologically. The crossing is recent, hard-won and disease-by-disease. Psychiatry has not had its 2024. That is the difference — temporal, not categorical — and stating it that way is both more accurate and much harder to attack than the version people usually reach for.

One more thing we will not borrow, even though it would help us. Insel, in the passage quoted above, goes on to write that “symptom-based diagnosis, once common in other areas of medicine, has been largely replaced in the past half century” [8]. Migraine, irritable bowel syndrome, sepsis and Parkinson's each falsify that sentence. It is his overstatement and we are flagging it rather than repeating it, because a page that launders a convenient error is a page that deserves to lose the argument.

5. the money

The rule for this section. Everything below is a tax filing, a peer-reviewed count, or a document the American Psychiatric Association published about itself. Nothing is inferred. Where the APA has answered a criticism, its answer is printed beside the criticism at the same size — not in a footnote. And where a number people expect us to give does not exist in any public record, we say that instead of manufacturing one.

who owns the revenue, in the APA's own words

There is no ambiguity about this, because the association stated it in its 2010 annual report [29]:

“APA purchased the publishing business of American Psychiatric Publishing, Inc. on September 30, 2010 and now owns the rights to current and future proceeds of the DSM.”

The manual it writes is a product it sells. DSM-5-TR lists at $220.00 in hardcover and $170.00 in paperback, with member pricing below that [28].

what a new edition does to the books

DSM-5 was published in May 2013. The APA's fiscal year is the calendar year. Here is its Form 990 across the window, on two lines [27]:

Net income from book sales went from $5.7 million to $42.4 million and back to $14.8 million. Total revenue more than doubled and then fell. A revision of the manual is, among the other things it is, the association's largest single financial event.

What that line does and does not say. The Form 990 line is called sales of inventory, not DSM. We can tell you that the spike happened in the fiscal year DSM-5 was published; the 990 does not attribute it, and neither do we. Anyone writing “DSM-5 earned the APA $42 million” has added a word the document does not contain.

the number we are not going to give you

DSM revenue, as an isolated figure, is not computable from public records — for any year. The APA's Form 990 has no DSM line. Its FY2024 return reports Publications $14,845,018, which bundles the DSM with Psychiatric News, the journals, other books and CME products; and Royalties $4,528,572, a single undifferentiated figure that includes DSM licensing along with everything else [27]. Total revenue that year was $58,199,477.

You will see confident dollar figures for DSM income quoted in books and articles. We could not trace any of them to a primary source, so none of them appear here. If you see one, ask which line of which filing it came from.

who was on the panels, and who paid them

Two peer-reviewed counts exist. For DSM-IV, published in 2006 [26]:

“Of the 170 DSM panel members 95 (56%) had one or more financial associations with companies in the pharmaceutical industry. One hundred percent of the members of the panels on ‘Mood Disorders’ and ‘Schizophrenia and Other Psychotic Disorders’ had financial ties to drug companies. The leading categories of financial interest held by panel members were research funding (42%), consultancies (22%) and speakers bureau (16%).”

For DSM-5, published in 2012 after the APA had introduced a disclosure policy [22]:

“Currently, 69% of the DSM-5 task force members report having ties to the pharmaceutical industry. This represents a relative increase of 21% over the proportion of DSM-IV task force members with such ties (57% of DSM-IV task force members had ties).”

And by work group [22]: “67% (N = 12) of the panel for Mood Disorders, 83% (N = 12) of the panel for Psychotic Disorders, and 100% (N = 7) of the Sleep/Wake Disorders … have ties to the pharmaceutical companies that manufacture the medications used to treat these disorders”. The paper's conclusion: “The new disclosure policy has not been accompanied by a reduction in the financial conflicts of interest of DSM panel members.”

and the APA's published answer, which belongs at the same size

The American Psychiatric Association responded formally, three days after publication, in a statement issued for its then-president John M. Oldham [25]:

“In 2012, 72 percent of the 153 members report no relationships with the pharmaceutical industry during the previous year. The scope of the relationships reported by the other 28 percent of member varies: 12 percent reported grant support only, including funding or receipt of medications for clinical trial research; 10 percent reported consultations including advice on the development of new compounds to improve treatments; and 7 percent reported receiving honoraria.”— American Psychiatric Association, “APA Refutes Secondary Analysis of DSM-5 Disclosures”, 15 March 2012 [25]

And its methodological objection, in full [25]:

“Additionally, since there were no disclosure requirements for journals, symposia or the DSM-IV Task Force at the time of the 1994 release of DSM-IV, Cosgrove and Krimsky's comparison of DSM-IV and DSM-5 Task Force and Work Group members is not valid.”

the disclosure policy itself, and what it actually caps

The APA Board of Trustees adopted the rules in December 2006. They are specific [31]:

“Receive no more than $10,000 annually in the aggregate from pharmaceutical companies/device makers/biotechnology companies and similar industry entities for direct services, such as consultancies, advisory committee positions, forensic assistance, speakers’ bureau services, etc.”

“Not hold stock or shares worth more than $50,000 in the aggregate in pharmaceutical companies/device makers/biotechnology companies and similar health-care related commercial ventures…”

The APA argues this is a strict standard, and says so plainly: “This standard is more stringent than requirements for employees at the National Institutes of Health and for members of advisory committees for the Food and Drug Administration” [25]. That is a real claim and it is on this page.

Three carve-outs decide how much the cap actually bites, and all three are in the documents themselves:

• The $10,000 ceiling is for direct services; the policy states the industry entities covered “do not include clinical practices, hospitals, nonprofit organizations, managed care organizations, university-based lectures, and similar activities” [31].
• The $50,000 stock ceiling excludes “stock and shares held in mutual funds, pension or retirement funds, blind trusts, and similar arrangements” [31].
• And the one that matters most: unrestricted research grants sit outside the cap entirely. Cosgrove & Krimsky put it directly — “panel members are not required to disclose unrestricted research grants from industry” [22]. Research funding was the largest single tie category in the DSM-IV count, at 42% [26]. A ceiling that excludes the biggest channel is a real ceiling over a narrower doorway than it sounds.

Two further facts about what the disclosures could and could not show. They were categorical, with no dollar amounts — a $500 honorarium and a $9,000 consultancy are typographically identical on the form, and Cosgrove & Krimsky note that “APA's policy does not require disclosure of the amount of money received from industry” [22]. And the speakers-bureau finding [22]:

“None of the DSM panel members identified participation on a speakers bureau. When we did an internet search of the 141 panel members, we found that 15% had disclosed elsewhere that they were members of drug companies' speakers bureaus or advisory boards.”

Note that 15% uses the 141 work-group denominator, not the 29-member task force and not the 170 total. Three different denominators are in play in this section and we have labelled every one of them, because the alternative is a page that is technically sourced and practically misleading.

and the conduct that changed, because that is part of the record too

The APA had no disclosure requirement for DSM-IV; it created one for DSM-5. It also wound down a revenue stream. From its 2009 annual report, in the association's own voice [30]:

“We took a leadership role among medical specialties when we began to phase out industry support for continuing medical education. The APA felt it was important to take the lead to separate marketing from medical education…”

A reader who takes away only “the panels were conflicted” has taken away half the record. The fuller money picture around psychiatric organisations — the association's industry revenue, and the advocacy groups beside it — is a separate page: who funds the advocates?

“who benefits from more diagnoses?”

It is the obvious question and we are going to leave it a question. Nothing on this page shows a payment conditioned on a criterion, and no document we found states an intent to widen a definition for commercial gain. We are not going to assert one, because a page that asserts it can be dismissed and a page that shows the record cannot.

What can be shown is the documented widening itself — which changes were made, and what was measured afterwards. That is section 6, and it is less one-directional than you probably expect.

6. expansion, and the criteria that loosened

The strongest witness here is the man who chaired DSM-IV. Allen Frances led the task force that produced the 1994 edition [23] and then spent the next two decades arguing in the medical literature that the manual — including his own — had gone too far. This is from World Psychiatry in 2013, open access, in his own words [32]:

“But psychiatric diagnosis is now facing another serious crisis of confidence, this time caused by diagnostic inflation. The elastic boundaries of psychiatry have been steadily expanding, because there is no bright line separating the worried well from the mildly mentally disordered.

The DSMs have introduced many new diagnoses that were no more than severe variants of normal behavior. Drug companies then flexed their powerful marketing muscle to sell psychiatric diagnoses by convincing potential patients and prescribers that expectable life problems were really mental disorders caused by a chemical imbalance and easily curable with an expensive pill.”— Allen Frances, chair of the DSM-IV Task Force, World Psychiatry, 2013 [32]

Note the plural: “The DSMs”. He is not exempting the edition he ran. And in the same piece he lists what he thinks DSM-5 did [32]:

“The DSM-5 has introduced several new disorders at the fuzzy and populous border with normal and has also loosened requirements for many of the existing disorders. The biggest problems are removing the bereavement exclusion for major depressive disorder, adding a very loosely defined somatic symptom disorder, reducing the threshold for adult ADHD and post-traumatic stress disorder, adding a diagnosis for temper tantrums, introducing the concept of behavioral addictions, combining substance abuse with substance dependence, and adding mild neurocognitive disorder and binge eating disorder.”

And here is where we disagree with our own best witness, in public. The same article states that “attention-deficit/hyperactivity disorder (ADHD) has tripled in rates in the past twenty years” and that “rates of autistic disorder have increased by more than twenty-fold” [32]. We have built the full counting record for both, from federal data with every methodology break annotated — how childhood ADHD got counted and how autism got counted — and those pages show the story is considerably more complicated than a rate multiplier, with large parts of the movement sitting at survey redesigns and case-definition changes rather than at DSM editions. Frances is the authority on what the manual did. He is not, on this page, the authority on what the counts did.

the documented changes, in the APA's own words — in both directions

Everything in this table is quoted from a single 19-page document the American Psychiatric Association published to summarise its own revision [33]. We include the narrowing changes because they are in the same document, and a page that quoted only the widening ones would be doing to the APA what we are asking you not to do to us.

three places the simple story breaks

1. The anxiety chapter moves in both directions on one page. Separation anxiety disorder lost its before-age-18 onset requirement, admitting adults who were previously ineligible. Two pages earlier, specific phobia, social anxiety disorder and agoraphobia gained a six-month duration requirement at all ages, which the APA states is “intended to minimize overdiagnosis of transient fears” [33]. Same chapter, same revision, opposite directions.

A correction we are making before anyone else does. That six-month duration change is very frequently attributed to generalized anxiety disorder, including in earlier drafts of our own work. It is not GAD. We searched the full text of the APA's Highlights document: the phrase “generalized anxiety” appears in it zero times. The change belongs to specific phobia, social anxiety disorder and agoraphobia, and separately to separation anxiety disorder. If you see it cited as a GAD change, the citation is wrong.

2. Autism consolidated four categories, and the field predicted it would shrink counts. Asperger's disorder, PDD-NOS, autistic disorder and childhood disintegrative disorder — four, not the three usually listed — were folded into one diagnosis [33]. The published expectation was a reduction in ascertainment, and the measured effect is genuinely contested in magnitude. We built that whole record, with the retention estimates and their confidence intervals intact, at how autism got counted.

3. The “temper tantrum diagnosis” was added to reduce a different diagnosis. Frances lists disruptive mood dysregulation disorder as one of the biggest problems with DSM-5 [32]. The APA's stated reason for creating it was “to address concerns about potential overdiagnosis and overtreatment of bipolar disorder in children” [33]. Those are not the same claim, and both are on the record. We are showing you the stated intent, which is documented. We are not claiming it worked — we could not verify a measured reduction in paediatric bipolar diagnosis from a source we could read, so no such effect is asserted here.

and the step almost everyone skips

A criterion change only moves a number if the thing doing the counting implemented that criterion. The clearest documented case is the bereavement exclusion: a federal statistical agency has stated that its national survey never applied the criterion that changed, so its depression estimates could not move regardless. We set that out, with the agency's own words, in where diagnoses come from. Definition changes and count changes are joined by an implementation step that is usually invisible and sometimes simply absent.

What this section does not contain. Effect sizes for most of these changes. Frances also lists the merging of substance abuse and dependence, somatic symptom disorder, binge eating disorder and mild neurocognitive disorder [32]; we could not verify measured prevalence effects for those from primary sources within this pass, so no effect figures are given for them and none should be inferred. The claim on this page is that the criteria changed in documented ways, in both directions — not that any particular number moved because of it.

7. what this means if you have a diagnosis

You have read a page arguing that your diagnosis is a description agreed by a committee, written in a book sold by a professional association, without a laboratory test behind it. Here is the part that matters, and it is not a softening.

• Your distress is not in question here. Nothing about the DSM's construction is an argument that anyone is exaggerating. Insel's point was that the field lacks the data for a better system — “because we lack the data” [8] — not that the patients are mistaken.
• Reliability is a real achievement and it works in your favour. The reason a diagnosis means roughly the same thing to a clinician in one city and a clinician in another is that somebody wrote the criteria down. That is what lets research accumulate, and what lets a treatment plan travel with you.
• Treatments were not derived from the diagnoses, and they still work for people. Most psychiatric drugs were found before anyone could say how they worked, and the mechanism story was written afterwards — a history we set out in where diagnoses come from. A treatment that helps you does not become less helpful because the explanation arrived second.
• This is not a reason to stop anything. Stopping psychiatric medication abruptly can be dangerous, and several classes carry documented withdrawal or relapse risks. If this page has raised a question about your treatment, it is a question for your prescriber, with a plan agreed in advance — see stopping antidepressants: withdrawal and relapse.

what to ask your prescriber

• Which criteria did I meet, and were any of them borderline?
• What would have pointed you toward a different diagnosis?
• Does the diagnosis change what we would try, or is it mainly a name for what we are already treating?
• If the treatment does not work, does that make you reconsider the diagnosis?
• Would you expect this diagnosis to still apply in five years?

The fourth is the one clinicians tend to appreciate most, because it is the question they ask themselves. None of these are challenges. They are the questions a description invites and an explanation would not.