In 1972 American psychologist David Rosenhan conducted an experiment to test psychiatric diagnosing. In what has become famous as the Rosenhan experiment, a number of mentally healthy people went to psychiatric hospitals around the country and applied for help. They told doctors they were hearing a voice in their head that said "Thud". Apart from this one symptom, they behaved completely normal.
All eight 'pseudopatients' were admitted and diagnosed as suffering from mental illness (seven as schizophrenic, one as bipolar), thus proving Rosenhan's suspicion that distinguishing sanity from insanity, and normal from abnormal, is much more difficult and contextually determined than had previously been assumed.
When the results of the experiment were published, one mental hospital challenged Rosenhan to repeat the experiment with them, confident that such mistakes would not be made in their institution. Rosenhan agreed to send one or more pseudopatients to them. And indeed, over the next months, the hospital identified 41 impostors from among their applicants. Rosenhan then revealed he hadn't sent a single pseudopatient to them, proving once more how fallable psychiatric diagnosis really is.
But to return to the first experiment, its greatest irony lies in how difficult it proved for the eight pseudopatients to get out of hospital again after they had been admitted. Once inside, they had been instructed to tell their doctors the voices had disappeared and they felt fine again. This, however, did not impress the staff, who proceeded to medicate and observe them. (As the pseudopatients took notes during their stay, many of the daily reports on their behavior contained the observation "Patient engaged in writing behavior.")
As none of them were able to convince their doctors they were not insane, it became clear, in a bizarre kind of Catch-22, that the only way out was to first admit they were insane, to affirm their psychiatrists' diagnosis, and then to show them they were getting better.
In the end, in one case after being confined for almost two months, all Rosenhan's pseudopatients were released as schizophrenics "in remission" (which is not to say they were sane again, just that their symptoms had sufficiently abated!).
...once labeled schizophrenic, the pseudopatient was stuck with that label. If the pseudopatient was to be discharged, he must naturally be "in remission"; but he was not sane, nor, in the institution's view, had he ever been sane.