Sanity vs Insanity
The distinction between sane and insane is very often arbitrary. There is no clear borderline. These two terms are relative, i.e. they depend on cultural and social surroundings and can be at best statistically defined. But nevertheless there are cases, especially in law, in which there has to be a definite decision on the sanity of a person. This decision can be a matter of being guilty or "not guilty by reason of insanity".
D. L. Rosenhan's 'On being sane in insane places' from 1973 shows alarmingly how delicate and tricky such decisions can even be for specialists. He performed an interesting test. Eight completely sane people were admitted to 12 different hospitals. He wanted to find out, if these people could be diagnosed as sane by the stuff. These eight people acted completely normal inside the mental hospitals. They tried their best to prove their sanity. Rosenhan summarizes: "The pseudo-patient's sanity went undetected. They spent an average of 19 days (range of 7 to 52 days) on the ward, before being released. When released, they were diagnosed as being `schizophrenic in remission' not as being sane. Some visitors and patients detected the pseudo-patients' sanity (35 out of 118 patients)." This is a type-one error, which is more likely to happen, i.e. a doctor is more likely to diagnose a healthy person as sick than a sick person as healthy. A type-two error is the other way round: A doctor diagnoses a sick person as healthy.
To see, how susceptible the stuff of a mental hospital is towards a type-two error, he arranged a follow up experiment at another hospital. Rosenhan writes in his article: "They doubted that they would commit type-two errors. They were warned that over the next three months one or more pseudo-patients would present themselves for admission at their hospital. The staff was asked to make judgements on a 10-point scale. Forty-one patients were detected by at least one staff member. Twenty-three were suspected by at least one psychiatrist. A psychiatrist and one other staff member suspected nineteen. In fact, no pseudo-patients had been presented! So it seems that type-one errors can be made when the stakes are high (prestige or diagnostic ability)."