Sane in Insane Places: A Theoretical Review

David Rosenhan sought to challenge common psychiatric diagnoses, using a controlled experiment to test whether the sane can be distinguished from the insane—wanting to shed light on a problem of type 2 error and misdiagnosis that often has serious social and psychological consequences (1973). The experiment placed eight “pseudopatients,” or normal, clinically sane individuals, in various mental hospitals throughout the country under the pretense that they had heard voices as seen in those with schizophrenia diagnosis (and once admitted, ceased displaying any symptoms). The reactions and diagnoses of these pseudopatients were used to verify misdiagnoses, and Rosenhan ultimately concluded that psychiatric professionals have a bias for diagnosing the sane as “insane,” and use this determination to interpret behavior as a product of that diagnostic label. This paper will review this theory and experiment using the evaluation criteria set out by Akers and Sellers (2012), in order to determine its validity in a methodological, scientific fashion.

Logical consistency. While Rosenhan points out from the very beginning that the concepts “sane” and “insane” are problematic—due to cultural differences and interpretations—he nonetheless adheres to a simplified understanding of deviant and non-deviant behavior. This is quite ambiguous, but he accounts for this by asking instead if “the salient characteristics that lead to diagnoses reside in the patients themselves or in the environments and, contexts in which observers find them” (Rosenhan, 1973, p. 236). In this way, the terms “sane” and “insane” do not necessarily have to be defined—as they can be measured in perceived sanity or insanity, through diagnosis. In this way, the experiment is turned to perceived illness as compared to whether the behaviors associated to that illness are acted out. He claims that normality (and therefore abnormality) “is distinct enough that it can be recognized wherever it occurs” (Rosenhan, 1973, p. 236). This is also ambiguous, and contradicts his previous statements concerning definitions of “sane” and “insane.”

Scope and parsimony. This is a very small-scale study, using only eight test subjects. The theory looks to apply to an aggregate of decisions made by psychiatric professionals, so the unit is appropriate—but the scope of the study seems inadequate to give significant results. Rosenhan’s argument is far from parsimonious, difficult to find through lots of explanation. He did sum up his theory well when he states: “Psychiatric diagnoses, in this view, are in the minds of the observers and are not valid summaries of characteristics displayed by the observed” (1973, p. 236). He does not go on to explain why he feels this is the case until the end of the paper, though, when he explains that failure to detect sanity in the sane could be due to bias to type 2 error prevalent throughout the medical industry on whole and a need for psychiatric professionals to give a label in the face of uncertainty.

Testability. While the concepts Rosenhan used may have been a bit unclear, the testability of the study was quite sound. The experiment could certainly be repeated (albeit, at different hospitals where they would not suspect a repeat and alter their behavior). The theory was certainly open to empirical falsification—the pseudopatients could have been diagnosed as sane and no type 2 errors would have occurred, disproving his theory. No tautology was present, as the independent variable (the behavior of the patients) was distinct and different from the dependent variable (the diagnosis of the psychiatric professionals involved). A clearer definition of “normal” vs. “insane” behavior could strengthen the validity of the study, but was clearly sufficient for measurement in this case.

Empirical validity. Though the results reported by the study seemed to demonstrate empirical validity, this analysis finds this problematic. There were several issues with the setup of the experiment. All of the pseudopatients were diagnosed with “schizophrenia in remission.” While this is clearly a misdiagnosis, the psychiatric professionals were operating with the belief that the patients had heard voices in the past—no meaningless symptom. If the symptom was self-reported without any incentive for having a diagnosis, then there is no reason for the psychiatrist to suspect the patient of falsifying this information. And though Rosenhan seems to take issue with it, he would have been aware of the permanence of the diagnosis while designing the methodology for the study. Schizophrenia is the most common diagnosis for individuals who hear voices, so it is reasonable to think that he would know going into the study that these individuals would be diagnosed as such, and released as schizophrenics “in remission.” The independent variable (behavior) really wouldn’t have much bearing on the dependent variable (diagnosis)—because the results would have been determined even prior to the behavior being observed.

That being said, other observations and measurements taken throughout the course of the experiment do give validity to Rosenhan’s theory. He mentioned that the “writing behavior” of the patients was never questioned, and attributed to the illness—as was the perceived “nervous” pacing that was actually due to boredom, and several other examples. This puts the independent variable (behavior) in a better position to be validly correlated with the dependent variable (perceived cause of the behavior). There seems to be a lack of variation, however, that would allow us to examine whether the independent variable influences the perceived cause of the behavior to be more or less due to the believed condition. Perhaps if more and less “serious” persistent mental health illnesses were feigned, there would be more variation (or perhaps not). In general, Rosenhan would have a more valid test if he more clearly defined the hypotheses he was attempting to prove or disprove, use additional measurements that are more clearly defined, and use controls to weed out bias.

Using Akers and Sellers’ evaluation methodology (2012), we see that Rosenhan’s study was logical and fairly consistent. It measured on the right level, but wasn’t large enough necessarily to be significant in determining the prevalence of type 2 error among psychiatric professionals. It also had some fundamental issues that infringed on its validity, but still demonstrated to some extent that perceived illness does cause these professionals to interpret their behavior through a lens that is narrow and limiting of the spectrum of human emotion. In terms of usefulness and policy implications of the study, it certainly does bring to light the gravity of this error in some cases. It really is a critique of the ways we approach the diagnosis of mental illness, and perhaps could shape the ways in which psychiatric professionals approach the interpretation of behavior. It could have interesting application to understanding the diagnosis of criminals aiming to use an “insane” diagnosis as a way to escape punishment—but more research would need to be performed on that particular segment of the population for it to be useful. In any case, this study can greatly be improved upon, but sheds light on an important issue to research.

References

Akers, R. L., & Sellers, C. S. (2012) Criminological theories: Introduction, evaluation, and application (6th ed.). USA: Oxford University Press.

Rosenhan, D. L. (1973). On being sane in insane places. Science, 179, 235-242.