H567/03 Section A - Rosenhan (1973): being sane in insane places

The question and the field study

Rosenhan's 1973 paper asked whether clinicians could distinguish people without serious psychiatric disorder from patients in psychiatric hospitals, and how the hospital setting influenced interpretation.

Its background was debate about the validity and reliability of psychiatric categories. An individual explanation locates behaviour in characteristics of the patient. A situational explanation considers the institutional setting and expectations. If ordinary behaviour is interpreted as evidence of illness after a label is applied, the judgement may reflect more than the behaviour itself.

The first phase was a field study involving planned deception to gain admission and covert participant observation inside hospitals. The observers participated as patients while recording staff and patient behaviour. It also used structured requests to staff and examined clinical records. There was no randomly allocated unlabelled control group sharing otherwise identical conditions, so describing the entire first phase as a clean laboratory experiment would misrepresent its control.

The reported sample comprised eight pseudopatients: five men and three women, including Rosenhan. Their occupations were a psychology graduate student, three psychologists, a paediatrician, a psychiatrist, a painter and a housewife. Across them there were 12 admissions to hospitals in five US states. The hospitals varied in age, resources and staff ratios; one was private. Eight people and 12 admissions are different units. The paper's note describes excluding a ninth person's data because that person's falsified history differed from the specified procedure.

The materials included a standard account of hearing voices, notebooks for observations, hospital/nursing records, courteous request prompts and later a ten-point staff rating scale. These were tools for examining diagnosis and treatment, rather than a psychological ability test. Most staff and genuine patients were unaware of the research. For Rosenhan's own admission, the administrator and chief psychologist knew of his participation, creating an exception to complete concealment.

This is historical research: the paper was published on 19 January 1973, during the DSM-II era. DSM-IV was not published until 1994; ICD-10 also belongs to a later period. The study cannot directly demonstrate the reliability of those later systems or contemporary diagnostic practice.

Admission and behaviour on the ward

The admission procedure created an apparent psychiatric symptom, followed by ordinary behaviour in an institutional setting.

The pseudopatients called for an appointment, then reported unfamiliar voices of their own sex, often unclear, apparently saying “empty”, “hollow” and “thud”. They used false names, and those in mental health occupations disguised their professions. The paper says other important life-history details were presented truthfully. The symptom report was deliberately false, so the initial decision to admit them must be interpreted in that context.

Once admitted, they stopped simulating symptoms and behaved cooperatively: talking normally, joining available activities and saying they felt well when asked. They responded to medication calls but generally did not swallow the pills. They wrote observations secretly at first and then openly when the writing attracted little attention. They did not know the discharge date and had to convince staff they could leave. The later admissions had legal safeguards, including an attorney on call, although stress and loss of freedom remained significant risks.

Quantitative measures included admission/discharge diagnosis, stay length, staff contact and responses to requests. Qualitative observations described privacy, treatment and interpretation of behaviour. Combining these makes the report richer than a discharge count alone, but the observers' experiences and expectations also influenced what they recorded.

Rosenhan reported that all gained admission. Of the 12 admissions, 11 were diagnosed as schizophrenia and one as manic-depressive psychosis, the latter in the private hospital. The paper describes discharge with schizophrenia “in remission” and no recognition by staff that the symptoms had been simulated. Hospitalisation ranged from 7 to 52 days, with a mean of 19. Remission describes reduced or absent symptoms under an illness label; it is not the same as recognising that the admission symptom was feigned.

During the first three hospitalisations, 35 of 118 genuine patients voiced suspicion that the observers were not genuinely ill, sometimes linking this to their note-taking. This does not show that every patient recognised them or that patients were formally more accurate diagnosticians: there was no equivalent systematic diagnostic test of patients and staff.

Labels, contact and depersonalisation

The reported ward experiences suggested that an existing label changed how ordinary actions were interpreted.

Rosenhan described note-taking recorded as a clinical behaviour and an ordinary family history reinterpreted as evidence of disturbed relationships. A pseudopatient pacing from boredom was asked whether they were nervous. These examples support labelling: expectations associated with a diagnosis can shape the meaning given to subsequent behaviour. However, an isolated observation of writing in a record does not independently prove that every clinician considered all writing pathological; the argument depends on the wider pattern and interpretation.

Depersonalisation means being treated as a category or object rather than recognised as an individual. Powerlessness concerns reduced control over one's activities, privacy and choices. The report included restricted movement, inspections of possessions, limited privacy and staff discussing patients without including them. Nearly 2,100 pills were dispensed to the pseudopatients, with only two swallowed. These are reported observations, not a recommended practical exercise.

Contact measures add numerical evidence. Attendants spent a mean 11.3% of their time outside the staff area. This was not a direct percentage of time interacting with patients: it also included chores. Across six pseudopatients and 129 days, daily contact with doctoral staff averaged 6.8 minutes, including interviews and meetings. Operational definitions matter when judging what these numbers show.

In four hospitals, pseudopatients made polite relevant requests about privileges, meetings or discharge, approaching each staff member at most once daily. They recorded responses. Psychiatrists stopped and talked on 4% of 185 attempts involving 13 respondents; nurses and attendants did so on approximately 0.5% of 1,283 attempts involving 47 respondents. Attempts are repeated encounters, not independent people.