5.3.1 - Rosenhan’s diagnostic study

The question and pseudopatient method

Rosenhan's study asked whether psychiatric staff could distinguish people without a mental disorder from patients when they encountered them within a hospital setting.

The original title, On being sane in insane places, uses historical language. “Sane” and “insane” are not respectful modern diagnostic categories; the learning question concerns the validity of diagnosis and the influence of context. The study did not show that psychological suffering is unreal.

Rosenhan reported eight pseudopatients: people presenting as patients for research purposes. The reported group included three women and five men, including Rosenhan, psychologists and people in other occupations. Twelve hospitals across five US states were used, with varied facilities and funding. This was a small purposively assembled group and a range of selected institutions, not a random sample of the population or all psychiatric services. The pseudopatients were researchers/confederates; the staff whose decisions were observed and the real patients they encountered were also affected by the research.

The pseudopatients sought admission reporting unfamiliar voices that appeared to say “empty”, “hollow” and “thud”. They used false names and some concealed mental-health occupations, but the published account said that other important personal-history details were truthful. After admission, they stopped simulating symptoms, said they felt well and behaved ordinarily. They observed ward life and recorded notes, initially discreetly and later openly. In the first admission, the administrator and chief psychologist knew of Rosenhan's presence; it is therefore inaccurate to claim nobody in any institution knew. Retrieved original text, university-hosted reprint.

This is best understood as a field investigation with covert participant observation and a staged presentation of symptoms. The operational outcomes included admission, diagnosis, length of stay and reactions to subsequent behaviour. It was not a randomised clinical trial: there was no equivalent randomly allocated group giving the same history without the hallucination claim. Naming an “independent variable” alone would not remove that design limitation.

The procedure deliberately supplied misleading clinical information, so the interpretation must distinguish an admission decision based on reported hallucinations from how staff interpreted later ordinary behaviour.

Reported findings and the labelling interpretation

Rosenhan interpreted the continued diagnostic label as evidence that the hospital context influenced the meaning assigned to behaviour.

The published account reported admission of the pseudopatients, mostly with schizophrenia diagnoses and one admission with the historical label “manic-depressive psychosis”. Stays were reported as ranging from 7 to 52 days, averaging 19 days. Staff did not identify the reported pseudopatients as research impostors. In the schizophrenia admissions, discharge described the disorder as in remission, meaning symptoms were no longer apparent or had improved, rather than retracting the original diagnosis. These are the reported findings; later scrutiny challenges the trustworthiness of parts of the report.

One important observation concerned note-taking. Nursing records for three pseudopatients referred to writing as part of their behaviour, which Rosenhan interpreted as pathologising an ordinary act because it occurred in a psychiatric patient. He also described impersonal interactions and loss of privacy and control. These observations add a lived institutional perspective but are not automatically objective measures of every staff member's motives. Original discussion of writing and labels.

Labelling means that a category attached to a person can shape later interpretation of what they do. Once “schizophrenia” was accepted, ordinary activity might be read as a symptom, while absence of symptoms might be read as remission. This resembles confirmation bias: new evidence is interpreted to fit the existing expectation. Rosenhan concluded that diagnostic judgements could say as much about the setting and label as about the individual.

There is a competing interpretation. Clinicians normally take a report of hallucinations seriously; observing improvement after admission is compatible with a genuine episode improving. A remission label acknowledges the current absence of symptoms and is not itself proof of failure to observe it. Because the staff had been given false symptom information, the study cannot cleanly separate defective diagnostic criteria from a reasonable response to deception. A valid evaluation weighs this alternative against the reported misinterpretation of ordinary behaviour.

Reversing expectations in the second investigation

The second investigation changed the staff's expectation by warning a research and teaching hospital that pseudopatients might arrive during the following three months.