In 1973, psychiatrists were embarrassingly bad at distinguishing the mentally ill from the normal. Today, psychopathology is still very subjective and arguably as much of an art as it is a science, with questionable criteria as to what is considered a mental "illness" or not.
In the early 70s, just like today, there were many critics of the mental health industry, specifically in the area of psychiatry. Rosenhan was one of these critics. As a researcher and psychiatrist himself, he put together a team of eight perfectly healthy and sane "pseudo patients" (five men and three women) to have themselves committed in one of several psychiatric hospitals across the United States. In order to make this experiment work, the pseudo patients had to lie about their name, job, and the fact that they heard a voice in their heads.
Immediately after admission, the pseudo patients acted normally (asymptomatic) while discreetly taking notes for the experiment. They could not be released without staff corroboration of their sanity., which took from 7 to 52 days, with an average of 19 days to get.
When the pseudo patients were eventually discharged, none of them were seen as normal, but all with "schizophrenia in remission." A twist to this experiment is that 35 of the 118 patients voiced their observation that these pseudo patients were not mentally ill, yet none of the staff ever suspected anything (at least not enough to voice their concerns).
In another twist to this study, some of the normal behavior of the pseudo patients was seen as pathological. For example, the staff notes for three of the patients read "patient engages in writing behavior."
In a secondary study, a hospital was expecting some pseudo patients to be sent, and their staff identified 41 possible suspects whom they thought were not mentally ill. It turns out, the experimenter actually sent no pseudo patients that time.
This study was criticized on ethical grounds for the deceit to the different medical staffs, especially by the people who were tricked.
It was mentioned that the pseudo patients could have been displaying nervous or paranoid behavior as a result of the possibility of getting caught as fakers, thus making them look ill, although this kind of nervous behavior is well within the range of normal human behavior.
Modern critics point out that diagnoses were very different 40 years ago, and now, with the more recent Diagnostic and Statistical Manual, diagnoses are much more accurate.
Even today, there are many different models of mental illness, none of which are perfect or upon which are universally agreed. The line between sane and insane is a fine one, and a simple label can make all the difference via the expectancy bias. Different cultures tolerate what many would consider insane behaviors based on cultural or religious norms. Sometimes, this "helping" people get better is more about trying to change those who are just different from us.