Being sane in insane places
The Rosenhan experiment (1973) sent eight mentally healthy volunteers to fake a single symptom and seek admission to psychiatric hospitals.
Every one was diagnosed with a mental disorder and hospitalized, and staff never detected the deception.
The study became a landmark critique of the reliability of psychiatric diagnosis, questioning whether psychiatrists can tell the sane from the insane.
Key Takeaways
- Pseudopatient Study: Between 1969 and 1972, Stanford professor David Rosenhan sent eight pseudopatients (healthy volunteers posing as patients) to 12 hospitals without telling staff.
- Universal Admission: All 12 admissions resulted in a diagnosis of a mental disorder, and staff never detected the deception during hospitalization.
- Minimal Staff Contact: Staff gave a genuine verbal response to only about 2.94% of patients’ requests for information, and daily contact with psychiatrists and other clinicians averaged just 6.8 minutes.
- The Reversed Study: At a hospital warned to expect fake patients, staff rated 41 of 193 genuine patients (21.24%) as likely pseudopatients, even though Rosenhan sent none.
- Validity Questioned: Together, the two studies provided strong evidence against the accuracy and validity of psychiatric diagnosis.
- Data in Doubt: Journalist Susannah Cahalan (2019) could confirm only two of the eight pseudopatients, and historian Andrew Scull (2023) argued the study was scientific fraud.
- Still Debated Today: Psychiatric diagnosis remains at odds with some neurological findings, and no single reform proposal has yet won widespread support.
Aim
David L. Rosenhan was a professor of law and psychology at Stanford University. In the years leading up to 1973, he set out to test whether psychiatrists could actually tell normal and abnormal psychological states apart. As he put it:
At its heart, the question of whether the sane can be distinguished from the insane (and whether degrees of insanity can be distinguished from each other) is a simple matter: do the salient characteristics that lead to diagnoses reside in the patients themselves or in the environments and contexts in which observers find them?
Background and Rationale
The APA’s DSM-II, published in 1968, reflected a widely held belief among mental health professionals. Psychiatric conditions, they thought, could be clearly distinguished from one another and from normal mental health, just as physical illnesses are distinguished in medicine.
During the 1960s, growing criticism challenged this view. Critics argued that psychiatric diagnoses lacked the objectivity and scientific rigor of physical diagnoses. Such labels, they claimed, were more like professional opinions, vulnerable to bias even when offered by skilled clinicians.
Four lines of earlier work fed this doubt:
- Szasz (1961): Psychiatrist Thomas Szasz argued in The Myth of Mental Illness that mental illness is a metaphor, not a demonstrable disease, and that diagnosis medicalizes “problems in living.”
- Scheff (1966): Sociologist Thomas Scheff proposed in Being Mentally Ill that minor rule-breaking becomes a lasting mentally ill role once others react with a label.
- Beck (1962): Aaron Beck reviewed the reliability studies of his day and found agreement between diagnosticians disturbingly low.
- Cooper et al. (1972): The US–UK Diagnostic Project found New York psychiatrists diagnosed schizophrenia far more readily than London psychiatrists, who saw depression or mania in the same patients.
To test this claim, Prof. Rosenhan designed an experiment. He arranged for individuals with no history or symptoms of serious mental illness to attempt admission to psychiatric hospitals.
If these pseudopatients were consistently recognized as mentally healthy, that would support the idea that psychiatrists could reliably distinguish normal from abnormal mental states. The underlying assumption was that psychiatric normality is distinct enough for competent professionals to identify.
Sample
Recruitment and Exclusions
Nine participants, including Prof. Rosenhan, were recruited. All were deemed to have no present or past symptoms of serious psychiatric disorders. In eight cases, admittance was gained without the hospital’s staff’s foreknowledge.
One case was different. In Prof. Rosenhan’s case, the hospital administrator and chief psychologist knew of their hospital’s inclusion in the study. His data were not excluded on that basis.
Data from one participant were excluded for a protocol breach: falsifying personal history beyond the agreed name, occupation, and employment.
Between one and four of the remaining eight participants thereafter gained further admission to four other hospitals, producing 12 hospital stays in total.
Who Took Part, and Where
Data from these 12 stays, at 12 different hospitals, by eight participants were included in the study. Five of the included participants were male adults; three were female adults.
Five of the eight worked in psychology or psychiatry: three psychologists, a psychiatrist, and a psychology graduate student. The rest were not clinicians. One was a paediatrician, one a painter, one a homemaker.
The design combined a field experiment with covert participant observation. It took place in real hospitals, with real staff who did not know they were being studied.
One of the 12 hospitals was privately funded; the rest received public funding. An undisclosed number were “old and shabby” or “quite understaffed.” They were spread across five states on the East and West coasts of the US.
The hospitals were deliberately varied in this way so that the findings would not turn on the peculiarities of any one institution.
Procedure
The admittance, stay, and discharge process was as follows:
- Participants set up an appointment at one of the hospitals under a false name, occupation, and employment.
- At the appointment, participants complained of hearing unfamiliar, often unclear voices, seemingly from someone of their own sex, saying “empty,” “hollow,” and “thud.”
- Participants gave truthful information on everything else, changing only the names, occupations, and employment of friends and family to match their own assumed details.
- On admittance, participants stopped simulating any psychiatric symptoms, though there were a few cases of “brief[,] … mild nervousness and anxiety” which “abated rapidly.”
- In psychiatric wards, participants engaged with patients and staff as they would normally with colleagues in everyday life. When asked how they were feeling, they said they were fine and no longer heard the voice.
- Participants received but did not swallow their prescribed medication, except in one or two instances, and recorded their observations of the ward, staff, and patients.
- Participants were discharged when hospital staff saw fit. A writ of habeas corpus was kept on file for each participant in case any could not secure release.
- Between one and four participants repeated this process at four other hospitals.
Two further protocol breaches, which did not result in exclusion, were reported. One participant revealed their occupation in psychology to staff members.
Another “attempted a romance with a nurse,” revealed their occupation in psychology to staff members, and provided psychotherapy to fellow patients.
Nursing reports for an undisclosed (“most”) number of the 12 stays were obtained from the hospitals. Coding protocols and training were not reported.
Measuring Depersonalization
Two measures captured how staff treated patients.
- Responses to Requests: At four of the 12 hospitals, participants made a polite request of staff, never more than once a day to the same person. Responses were recorded.
- Daily Contact: At six hospitals, participants measured their daily contact with staff, in minutes, from the admission interview to the final discharge meeting inclusive.
A typical request was, “Pardon me, Mr. [or Dr. or Mrs.] X, could you tell me when I will be eligible for grounds privileges?” Others asked, “… when I will be presented at the staff meeting?” or “… when I am likely to be discharged?”
The response-classification criteria, recording and measurement protocols, and pre-study training were not disclosed.
Setting Up the Imposter Study
A second study was arranged at a research and teaching hospital in an undisclosed location. Its staff were aware of the first study’s results.
Staff were told to expect impostors. At some point in the following three months, one or more persons would attempt to gain admission. Each would have been previously instructed to simulate at least one symptom of a psychiatric disorder, i.e., a pseudo-patient.
Staff rated each patient seeking admission on a ten-point scale from most to least likely to be a pseudo-patient. A score of 1 or 2 meant most likely. No participants were, in fact, sent to the hospital. Scores were obtained for 193 patients.
No patients were independently evaluated by Prof. Rosenhan or affiliated researchers. No scoring protocol nor pre-study training was reported.
Findings
In eleven instances, participants were admitted on a diagnosis of schizophrenia and discharged with a diagnosis of schizophrenia in remission.
In one instance, a participant was admitted on a diagnosis of manic-depressive psychosis; their discharge diagnosis was not reported. Stays ranged from 7 to 52 days, with a mean of 19 days. Standard deviation was not disclosed.
Nursing reports uniformly indicated that the participants were “friendly,” “cooperative,” and “exhibited no abnormal indications.” Coding results and statistical significance were not reported.
Participants’ visitors “could detect no serious behavioral consequences [of hospitalization in participants].” No information on data collection from said visitors, nor statistical analyses on these data, were reported.
The errant diagnosis on admission, Prof. Rosenhan noted, could simply be attributed to physicians’ strong bias towards type II errors. As he put it:
The reasons [for this strong bias] are not hard to find: it is clearly more dangerous to misdiagnose illness than health. Better to err on the side of caution, to suspect illness even among the healthy.
Errant diagnoses after admission troubled Prof. Rosenhan more than the initial ones. Once diagnosed, participants could not shake the label, even though they had dropped the pretense the moment they were admitted.
A competent practitioner should eventually have recognized the initial diagnosis as a type II error and corrected it. No such correction happened in any of the 12 stays.
The admission diagnosis became, in Prof. Rosenhan’s words, “so powerful that many of the [participants’] normal behaviors were overlooked entirely or profoundly misinterpreted.” Three examples illustrate this:
- Note-Taking Read as Symptom: Nursing staff recorded a pseudopatient’s open note-taking as “the patient engages in writing behaviour,” treating an ordinary habit as pathology.
- Queuing Read as “Oral-Acquisitive”: One psychiatrist read patients’ ordinary, boredom-driven queuing outside the cafeteria as evidence of the “oral-acquisitive” nature of the illness. Pacing from boredom was read the same way, as anxiety.
- Family History Read as Cause: Even a truthfully reported, ordinary family background was written up as though it explained the supposed illness.
This attribution reinforced the diagnosis each time, fixing the type II error in place. Prof. Rosenhan argued that this pointed less to the incompetence of psychiatrists, psychologists, or nurses than to the “enormous overlap in behaviors of the sane and the insane.”
Tellingly, the people who regularly saw through the deception were other patients, not staff. Staff never did.
On the first three admissions, 35 of 118 patients on the admissions wards voiced suspicions, with challenges such as “You’re not crazy.” They guessed the pseudopatients were journalists or professors checking up on the hospital.
Depersonalization Observations
Staff Responses to Requests
On initiating contact with staff members, participants were overwhelmingly ignored by psychiatrists, nurses, or attendants at four of the hospitals. Only about 2.94% of 1,468 pseudopatient-initiated approaches received a genuine verbal response.
Inter-hospital differences were not deemed significant and were not reported. Statistical differences and significance were not reported.
Prof. Rosenhan read the near-total silence as revealing. He treated contact, or its absence, as a direct mark of concern versus avoidance.
Time With Staff and the Sense of Powerlessness
Six participants, totaling 129 days of hospitalization, measured their daily contact with staff. Daily contact with psychiatrists, psychologists, residents, and physicians ranged from 3.9 to 25.1 minutes. The mean was just 6.8 minutes.
Disaggregated data, contact time with other staff, statistical differences, statistical significance, and standard deviation were not reported.
Rosenhan drew a clear contrast. Contact, Prof. Rosenhan notes, indicates “concern and individuation,” its lack, “avoidance and depersonalization.” Staff showed overwhelming avoidance of patients.
Prof. Rosenhan argued that this resulted in an “overwhelming sense of powerlessness which invades the individual as he is continually exposed to the depersonalization of the psychiatric hospital.”
Table 1. Self-initiated contact by participants with psychiatrists, nurses, and attendants. Adapted from Table 1 of the 1973 Rosenhan study, p. 255. Copyright (c) 1973 AAAS.
The import of this practice, and the powerlessness it elicited, Prof. Rosenhan argued, could be readily inferred.
[H]ow many patients might be “sane” outside the psychiatric hospital but seem insane in it — not because craziness resides in them, as it were, but because they are responding to a bizarre setting … ?
Non-existent Imposter Study
Of the 193 scores obtained, 41 patients were judged highly likely to have been a pseudo-patient by at least one staff member. Twenty-three were judged so by at least one psychiatrist, and 19 by a psychiatrist together with at least one other staff member.
No pseudo-patients were sent by Prof. Rosenhan or affiliated researchers.
Of the 41 patients judged highly likely to have been pseudo-patients:
- Either they did not, in fact, have any abnormal psychiatric symptoms and were correctly judged pseudo-patients by the staff;
- Or they did, in fact, have abnormal psychiatric symptoms (as presumed by the fact that they wished to be admitted), in which case the staff misdiagnosed them as pseudo-patients and thereby made a type I error in seeking to avoid type II errors.
In either case, the results were telling. Prof. Rosenhan explained: “that the tendency to designate sane people as insane can be reversed when the stakes (in this case, prestige and diagnostic acumen) are high.”
But it is notable that, in the second case, the price for avoiding type II errors might just be a higher type I error rate.
Conclusion
It is clear that we cannot distinguish the sane from the insane in psychiatric hospitals. The hospital itself imposes a special environment in which the meanings of behaviour can easily be misunderstood.
Both practitioners and patients, the study reveals, seem caught in Catch-22s. Out of an excess of caution, psychiatrists and psychologists strongly tend towards type II errors on admission. Yet the trap cuts both ways.
Once that error is made, there is only a slim chance it will be caught during in-patient treatment.
Should practitioners instead try to avoid type II errors sticking to patients, they risk equally damaging type I errors. Patients face a parallel bind of their own.
Once admitted, they are likely to develop psychopathies, whether or not they truly had any on admission. The bizarre setting itself is enough. But should they avoid that setting altogether, they risk an untreated illness getting worse, if they truly had one.
A way out for practitioners and patients is not immediately clear to Prof. Rosenhan. Two promising directions he noted were:
- The avoidance of psychiatric diagnoses of the form encouraged by the DSM II in favor of diagnosing patients with “specific problems and behaviors” so as to provide treatment outside of psychiatric hospitals and to keep any diagnostic label from “sticking” to a patient;
- Increasing “the sensitivity of mental health workers and researchers to the Catch-22 position of psychiatric patients,” for e.g., by having them read pertinent literature.
Other Conclusions
A good number of the study’s shortcomings should give us pause when drawing conclusions.
Sampling, randomization, control, blinding, and statistical analysis methods were largely unreported and so likely not to have been up to present-day standards.
Participant training was not reported and so likely not undertaken before the study. No data on participants’ visitors and their evaluations were reported.
Study flaws aside, the observed effects were large enough to likely be both clinically and statistically significant:
- All 12 hospitalizations resulted in type II errors both on admission and discharge;
- 2.94% of the 1,468 recorded participant-initiated interactions with psychiatric staff resulted in verbal engagement with the participant;
- 9.84% of the 193 patients scored at a research and teaching hospital were judged highly likely to be pseudo-patients by both a psychiatrist and at least one other staff member.
The findings pointed to an unacceptable preponderance and persistence of type II errors by competent psychiatric staff. They also pointed to the danger of psychiatric harm to patients posed by then-current psychiatric practices.
Critical Evaluation
Rosenhan’s study has clear strengths and serious limitations. The sections below weigh each in detail.
Its main strengths:
- Ecological Validity: Real hospitals and unwitting staff captured genuine clinical behavior and avoided demand characteristics.
- Rich Data: Quantitative measures, such as length of stay and minutes of staff contact, sat alongside detailed qualitative observation.
- Lasting Impact: It exposed the unreliability of 1970s diagnosis and sensitized clinicians to label-driven interpretation.
Its main limitations:
- Appropriate Diagnosis? Reporting hallucinations is a legitimate reason to admit someone, and staff were deliberately deceived (Spitzer, 1975).
- Ethical Costs: Staff were deceived without consent, and healthy volunteers used scarce clinical resources.
- Weak Design: Sampling, coding and statistics are largely unreported, and the pseudopatients collected their own data.
- Improved Reliability: The documented unreliability was real for its era but was addressed by operational criteria from DSM-III onward.
- Data in Doubt: Cahalan (2019) and Scull (2023) question whether the findings can be cited as secure fact.
Was the Study High in Ecological Validity?
Field experiments have the major advantage of being conducted in a real environment. This gives the research high ecological validity.
However, it is not possible to have as many controls as would be possible in a laboratory experiment.
Participant observation allows the collection of highly detailed data without the problem of demand characteristics.
As the hospitals did not know of the existence of the pseudopatients, there is no possibility that the staff could have changed their behavior. They did not know they were being observed.
However, this does raise serious ethical issues (see below). The presence of the pseudopatient may also have changed the environment being observed.
Was the sample representative?
Strictly speaking, the sample is the twelve hospitals that were studied. Rosenhan ensured that this included a range of old and new institutions as well as those with different sources of funding.
The results revealed little differences between the hospitals. This suggests that it is probably reasonable to generalize from this sample and suggest that the same results would be found in other hospitals.
Prof. Rosenhan’s 1973 paper does not detail:
- How his sample size was determined, nor how his sample was selected;
- The study’s inclusion/exclusion criteria;
- How past or present serious psychiatric symptoms were diagnosed, nor by whom;
- Whether past or present mild to moderate psychiatric symptoms were diagnosed, nor by whom;
- How hospitals were selected;
- How participants were matched with false names, occupations, and employment information;
- How participants were matched with hospitals.
What type of data was collected in this study?
There is a huge variety of data reported in this study. Quantitative data detail how many days each pseudopatient spent in the hospital and how many times pseudopatients were ignored by staff. Qualitative descriptions cover the experiences of the pseudopatients themselves.
One of the strengths of this study is the sheer wealth of data reported. The conclusions Rosenhan reaches are well illustrated by the qualitative data he included, leaving little doubt about their basis.
However, the pseudopatients knew the hypothesis. They also gathered the data themselves. That leaves the observations open to experimenter and observer bias.
They were motivated to notice and record depersonalization. No coding protocol or inter-observer reliability check was reported.
Was the study ethical?
Strictly speaking, no.
The staff were deceived: they did not know they were being observed. Consider how they might have felt on discovering that the research had taken place.
Genuine hospital resources, including beds, medication, and clinicians’ time, were used up by healthy volunteers, potentially at the expense of real patients.
The pseudopatients themselves faced real risk too. They were exposed to the distress and powerlessness of involuntary confinement. Their only protection was a habeas corpus writ on file in case they could not secure release.
Was the study justified?
That is harder to say. There was certainly no other way to conduct the study, so you need to consider whether the results justified the deception. This is discussed later under the heading of usefulness.
What does the study tell us about individual/situational explanations of behavior?
The study suggests that once the patients were labeled, the label stuck. Everything they did or said was interpreted as typical of a schizophrenic (or manic-depressive) patient. The label, not the person, drove the judgment.
This means that the situation the pseudopatients were in had a powerful impact on the way that they were judged.
The hospital staff could not perceive the pseudopatients in isolation from their label and from the fact that they were in a psychiatric hospital. This raises serious doubts about the reliability and validity of the psychiatric diagnosis.
What does the study tell us about reinforcement and social control?
The implications of the study are that patients in psychiatric hospitals are ‘conditioned’ to behave in certain ways by the environments that they find themselves in.
Their behavior is shaped by the environment: nurses assume, for example, that signs of boredom are signs of anxiety. If the environment does not allow them to display ‘normal’ behavior, it will be difficult for them to be seen as normal.
Labeling is a powerful form of social control. Once a label has been applied to an individual, everything they do or say will be interpreted in the light of this label.
Across the 12 stays, staff gave pseudopatients a total of 2,100 pills, including drugs such as Elavil, Stelazine, Compazine, and Thorazine. Almost none were actually swallowed.
Rosenhan describes pseudopatients going to flush their medication down the toilet and finding other patients’ pills already there. This suggests that so long as patients caused no trouble, very few checks were made.
Was the study useful?
The study was certainly useful. It highlighted how hospital staff interact with patients and suggested many ways to improve hospital care and staff training. Its influence reached further:
- Diagnostic Reform: The paper became a rallying point for explicit, operational criteria. DSM-III (1980) introduced symptom checklists and structured interviews to raise reliability, widely read as a response to Rosenhan.
- Labeling Debates: It gave vivid evidence that diagnosis reflects context and expectation, alongside Scheff (1966) and Szasz (1961) and the labeling critique.
- Deinstitutionalization: Its exposé of ward depersonalization and powerlessness strengthened the case for community care over large asylums.
- Teaching Classic: It remains the standard illustration of why reliability underpins validity and how observer bias distorts clinical judgment.
There is an irony here. The task force that produced DSM-III was chaired by Robert Spitzer, Rosenhan’s most formidable critic.
However, some of Rosenhan’s conclusions can be questioned. Suppose you falsely told a doctor you had severe abdominal pain, and the doctor admitted you. You could hardly blame the doctor for the resulting diagnosis.
Isn’t it better for psychiatrists to err on the side of caution? The risk cuts both ways. Admitting someone who is not really mentally ill seems safer than sending away someone who might be genuinely suffering.
This does not excuse how long some pseudopatients spent in hospital acting perfectly normally. But it does support the actions of the staff who made the initial diagnosis.
Spitzer’s Rebuttal: Is This Really “Pseudoscience”?
Psychiatrist Robert Spitzer (1975) published the most famous response to Rosenhan, called “On pseudoscience in science, logic in remission, and psychiatric diagnosis.” It made three main points:
- Genuine Symptom: Auditory hallucinations are a serious complaint, so admitting someone who reports hearing voices was reasonable. Staff were deceived, and deception, not incompetence, explains the error.
- Discharge “In Remission”: Spitzer argued this label is rare in ordinary practice. It is the correct judgment for someone who shows no symptoms, so it shows staff responding to normal behavior.
- Overreach: Spitzer accused Rosenhan of drawing sweeping conclusions about psychiatric diagnosis from one study of determined liars fooling clinicians. He called this the “logic of pseudoscience.”
Spitzer also offered an analogy. Suppose a person swallows blood, feigns a hemorrhage in an emergency room, and is diagnosed with a peptic ulcer. That reflects the patient’s deception, not the validity of the diagnostic system.
Rosenhan did not leave this unanswered. In a reply in the same journal issue, he argued that diagnosis is shaped by the context in which a patient is met and by the diagnostician’s expectations (Rosenhan, 1975).
In his view, his critics’ emphasis on how sensible the admissions were illustrated this context-dependence rather than refuting it.
The exchange sharpened the dispute rather than settling it. Both sides agreed the admissions were a predictable response to the context the pseudopatients created. They drew opposite conclusions about whether that predictability vindicates or indicts the diagnostic process.
Has Diagnostic Reliability Improved Since the 1970s?
One defence is historical. The poor reliability of Rosenhan’s era was real and well documented. Beck (1962) found agreement between diagnosticians low across the studies he reviewed. Cooper et al. (1972) showed that the same patients were diagnosed with schizophrenia in New York but depression or mania in London.
Defenders argue this was substantially addressed from DSM-III (1980) onward. Explicit operational criteria and structured interviews raised agreement between clinicians to acceptable levels for most major categories.
Modern criteria also require a minimum duration. A schizophrenia diagnosis now needs continuous signs of disturbance for around six months, including at least a month of active-phase symptoms. A single hallucination that vanished on admission would not warrant the label today.
On this view, the study portrays 1970s practice, not diagnosis as a timeless whole. Later evidence complicates that defence (see Contemporary Research, below).
Outlook of Diagnostic Accuracy
Psychiatric diagnoses continue to be made much as they were in Prof. Rosenhan’s day. Clinicians still rely largely on inferences drawn from patient self-reports and their own observations of behavior, judged against criteria set by the APA’s DSM.
This points to two sources of diagnostic problems in psychiatry:
- the evidence used to reach a diagnosis, and
- the criteria by which said evidence is evaluated in reaching a diagnosis.
The evidence available to psychiatrists and psychologists in diagnosing mental disorders has long been much sparser than that available to other physicians.
There have been advances in understanding the biological basis of mental disorders in the decades since.
This growing corpus has yet to yield diagnostic tests, though.
In 2012, a group of three researchers, led by Prof. Shitij Kapur of King’s College London, examined why. They pointed to a number of reasons, including widespread methodological shortcomings and the DSM’s own classification system.
On that note, the DSM has left much to desire. As Thomas Insel, former director of the NIMH, put it:
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.
Since its first publication in 1952, the DSM has reached a classification of mental disorders without data on their biological underpinnings.
Its nosology is increasingly at odds with aetiological research, which increasingly suggests that mental disorders are rather gradual deviations from typical brain functions. The diagnosis, in short, may be tracking the wrong unit.
This, in turn, suggests that mental disorders should be classified as points or areas on spectra rather than the neat categories propounded by the DSM.
One effort at building such a nosology was begun by the NIMH in 2010. The project, dubbed the Research Domain Criteria (RDoC), is still confined to research and is not ready for clinical application. It remains a research tool only.
Myriad problems in psychiatric research and practice preclude any consensus on the accuracy of psychiatric diagnoses. They are likely to do so until resolved.
The field has not converged on a corrective program, though there exist a number of such programs competing for widespread support.
Contemporary Research
Journalist Susannah Cahalan’s reappraisal (see The Modern Reappraisal, below) casts doubt on Rosenhan’s data. The question he raised remains: how reliably do psychiatrists agree on a diagnosis?
That question resurfaced with the APA’s own DSM-5 field trials. Test-retest reliability for several major categories, including major depressive disorder and generalised anxiety disorder, fell into the “questionable” or “unacceptable” range (Regier et al., 2013).
A large study then asked whether that decline was genuine, or just an artefact of how reliability was measured.
- Aim: To test whether DSM-5’s disappointing reliability reflected worse clinical judgment, or the method used to measure it.
- Method: 339 patients were diagnosed by structured interview, then 218 were re-diagnosed independently in genuine test-retest interviews. A further 49 were assessed by audio-recording, where a second rater scores the same interview.
- Results: The audio-recording method showed good to excellent reliability (kappa = .80), close to DSM-IV estimates. Test-retest was poor to fair (kappa = .47), close to DSM-5’s own figures (Chmielewski et al., 2015).
- Conclusion: The same diagnostic criteria, applied to comparable patients, gave very different reliability scores depending purely on how reliability was measured. The method mattered more than the criteria.
Why the gap? Audio-recording inflates agreement because both raters score one interview. Independent test-retest interviews expose real variability across time and clinician.
On this reading, DSM-5 was not less reliable than DSM-IV. Its field trials were the first to measure reliability without that inflation.
On several major diagnoses, independent clinicians assessing the same patient on separate occasions still showed only “poor to fair” agreement (Chmielewski et al., 2015).
That gap echoes Rosenhan’s own findings. Marking the study’s fiftieth anniversary, a 2023 reappraisal argued that psychiatry’s classification project remains “a work half done.” The field still lacks an objective marker to verify a diagnosis, rather than merely agree on one (Agarwal, 2023).
The Modern Reappraisal: The Great Pretender
The most serious recent challenge to Rosenhan’s study targets not his interpretation, but his data. In The Great Pretender (2019), journalist Susannah Cahalan spent years trying to trace the eight pseudopatients using Rosenhan’s own papers. She is herself a survivor of a misdiagnosed organic brain disease.
What Cahalan Found
Cahalan’s investigation raised three main problems:
- Untraceable Pseudopatients: Cahalan could reliably identify only two of the eight: Rosenhan himself and graduate student Bill Underwood. Several others could not be confirmed to have existed, and some data may have been invented.
- Contradictory Admission Record: The paper says Rosenhan reported only a voice saying “thud.” His medical notes, filed under a pseudonym, describe distressing voices, suicidal thoughts, and copper over his ears to block the signals.
- An Omitted Ninth Pseudopatient: Harry Lando’s data were left out of the paper. Lando found his stay positive and humane, a result that did not fit Rosenhan’s thesis.
If accurate, the admission record changes the picture. That would undercut the study’s central claim that a trivial complaint led to needless confinement. Rosenhan himself would have been a genuine patient.
Lando’s omission raises further doubts about the data.
An earlier note of caution came from psychologist Lauren Slater (2004). In her book Opening Skinner’s Box, she described her own attempt to repeat the study by reporting a single hallucinated word at emergency rooms.
Her account was itself contested, but it foreshadowed the later doubts about how easily Rosenhan’s design could be reproduced or trusted.
Those doubts have since hardened. Historian of psychiatry Andrew Scull went further than Cahalan’s own cautious framing. Writing in History of Psychiatry, he drew on the records Cahalan shared with him. He concluded that Rosenhan’s paper is “a spectacularly successful case of scientific fraud” (Scull, 2023).
The charge is deliberate fabrication, not innocent embellishment. His verdict comes from within the history of psychiatry, not journalism alone.
What This Means for Citing Rosenhan Today
In brief, Cahalan could confirm only two of the eight pseudopatients, and Scull (2023) called the study scientific fraud.
For citing the study today, the practical lesson is to treat it as historically pivotal but evidentially fragile. The case is now genuinely contested. Its influence on psychiatric diagnosis and on the sociology of labelling remains a matter of record.
Its specific findings, though, can no longer be repeated as secure empirical fact. Any responsible account should flag Cahalan’s reappraisal alongside the classic narrative, rather than presenting the original data as settled.
FAQs
What did the Rosenhan study suggest in 1973?
The Rosenhan study in 1973 suggested that psychiatric diagnoses are often subjective and unreliable. Rosenhan and his associates feigned hallucinations to get admitted to mental hospitals but acted normally afterward.
Despite this, they were held for significant periods and treated as if they were genuinely mentally ill. The study highlighted issues with the validity of psychiatric diagnosis and the stigma attached to mental illness.
What did the classic study by Rosenhan reveal about the power of labels that are applied to individuals?
The classic study by Rosenhan showed the influential effect of labels on individuals, specifically psychiatric labels. By pretending to have hallucinations, mentally healthy participants gained admission to psychiatric hospitals.
The study demonstrated that once labeled as mentally ill, their behaviors were consistently interpreted in that context, even when they stopped simulating symptoms.
References
Agarwal, A. (2023). “On being sane in insane places” … it is a work half done. Indian Journal of Psychiatry, 65(5), 601–603. https://doi.org/10.4103/indianjpsychiatry.indianjpsychiatry_773_22
Beck, A. T. (1962). Reliability of psychiatric diagnoses: 1. A critique of systematic studies. American Journal of Psychiatry, 119(3), 210–216. https://doi.org/10.1176/ajp.119.3.210
Cahalan, S. (2019). The great pretender: The undercover mission that changed our understanding of madness. Grand Central Publishing.
Chmielewski, M., Clark, L. A., Bagby, R. M., & Watson, D. (2015). Method matters: Understanding diagnostic reliability in DSM-IV and DSM-5. Journal of Abnormal Psychology, 124(3), 764–769. https://doi.org/10.1037/abn0000069
Cooper, J. E., Kendell, R. E., Gurland, B. J., Sharpe, L., Copeland, J. R. M., & Simon, R. (1972). Psychiatric diagnosis in New York and London: A comparative study of mental hospital admissions. Oxford University Press.
Insel, T. R. (2013, April 29). Transforming diagnosis [Blog post]. National Institute of Mental Health. https://www.nimh.nih.gov/about/directors/thomas-insel/blog/2013/transforming-diagnosis.shtml
Kapur, S., Phillips, A. G., & Insel, T. R. (2012). Why has it taken so long for biological psychiatry to develop clinical tests and what to do about it? Molecular Psychiatry, 17(12), 1174–1179.
Regier, D. A., Narrow, W. E., Clarke, D. E., Kraemer, H. C., Kuramoto, S. J., Kuhl, E. A., & Kupfer, D. J. (2013). DSM-5 field trials in the United States and Canada, Part II: Test-retest reliability of selected categorical diagnoses. American Journal of Psychiatry, 170(1), 59–70. https://doi.org/10.1176/appi.ajp.2012.12070999
Rosenhan, D. L. (1973). On being sane in insane places. Science, 179(4070), 250–258. https://doi.org/10.1126/science.179.4070.250
Rosenhan, D. L. (1975). The contextual nature of psychiatric diagnosis. Journal of Abnormal Psychology, 84(5), 462–474. https://doi.org/10.1037/h0077123
Scheff, T. J. (1966). Being mentally ill: A sociological theory. Aldine.
Scull, A. (2023). Rosenhan revisited: Successful scientific fraud. History of Psychiatry, 34(2), 180–195. https://doi.org/10.1177/0957154×221150878
Sharp, C., Fowler, J. C., Salas, R., Nielsen, D., Allen, J., Oldham, J., Kosten, T., Mathew, S., Madan, A., Frueh, B. C., & Fonagy, P. (2016). Operationalizing NIMH Research Domain Criteria (RDoC) in naturalistic clinical settings. Bulletin of the Menninger Clinic, 80(3), 187–212.
Slater, L. (2004). Opening Skinner’s box: Great psychological experiments of the twentieth century. W. W. Norton.
Spitzer, R. L. (1975). On pseudoscience in science, logic in remission, and psychiatric diagnosis: A critique of Rosenhan’s “On being sane in insane places.” Journal of Abnormal Psychology, 84(5), 442–452. https://doi.org/10.1037/h0077124
Szasz, T. S. (1961). The myth of mental illness: Foundations of a theory of personal conduct. Harper & Row.
