Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

The Rosenhan Experiment

3 Post author: chaosmosis 14 September 2012 10:31PM

I haven't seen any links to this on Lesswrong yet, and I just discovered it myself. It's extremely interesting, and has a lot of implications for how the way that people perceive and think of others are largely determined by their environmental context. It's also a fairly good indict of presumably common psychiatric practices, although it's also presumably outdated by now. Maybe some of you are already familiar with it, but I thought I'd mention it and post a link for those of you who aren't.

There's probably newer research on this, but I don't have time to investigate it at the moment.

http://en.wikipedia.org/wiki/Rosenhan_experiment

Comments (20)

Comment author: DanielVarga 16 September 2012 11:20:10AM 9 points [-]

The second part of his study involved an offended hospital challenging Rosenhan to send pseudopatients to its facility, whom its staff would then detect. Rosenhan agreed and in the following weeks out of 193 new patients the staff identified 41 as potential pseudopatients, with 19 of these receiving suspicion from at least 1 psychiatrist and 1 other staff member. In fact Rosenhan had sent no one to the hospital.

Conducting an important scientific experiment, teaching a painful lesson to those who challenge your authority, and showing your good sense of humor. Without any work whatsoever. It must have felt good.

Comment author: Trevor_Blake 15 September 2012 04:30:38AM 3 points [-]

A good post to raise a toast to the recently departed Dr. Thomas Szasz.

Comment author: Vaniver 15 September 2012 01:26:15AM 2 points [-]

Note that 1973 was a long time ago, and that the deinstitutionalization movement got a significant boost in support from the Rosenhan experiment. I imagine things are different now, although diagnostic criteria for mental illness are still primitive at best.

Comment author: novalis 15 September 2012 05:37:51PM 3 points [-]

Things are not entirely different.

Here's another one

(Note that the main thing you'll find when you search for replications is Lauren Slater's book. Don't trust it; she won't share any evidence and has a history of lying).

Comment author: MixedNuts 15 September 2012 06:08:01PM 2 points [-]

First link: Patients were psychiatric nurses, faking specific illnesses during their whole stay. (Rosenhan had various people faking atypical symptoms until admission.) Doctors believing them aren't to blame.

Among the findings of the project were that patients frequently found it difficult to get information on their treatment

Agrees with my experience. A computer for patients and access to Crazy Meds help, but I don't know an easy fix for less geeky patients.

Second link: That's a pretty good test. However, the doctors were shown patients who had been treated and were doing well. It's harder to diagnose short-sightedness if your patient is wearing contact lenses.

So neither of these tests are nearly as stringent as Rosenhan's.

Comment author: Dentin 15 September 2012 04:39:43AM 3 points [-]

Things may in fact be better now; however, based on an experience I had approximately ten years ago, they are not substantially better, and certainly not good enough. The patient in my anecdotal evidence was only freed upon the same conditions as those in the Rosenhan experiment, and the behaviour of the staff and doctors was spot on. If anything, the patient left the institution with more problems than she entered with.

Comment author: MixedNuts 15 September 2012 09:20:09AM 2 points [-]

Staff behavior is easiest to judge (patients can't see hospital notes, and if you're not faking you can't judge the diagnosis much). Here's my experience in the acute psychiatric wing of a Swedish public hospital:

I am not psychotic (as far as I know, ha) but had a symptom a few years back that I mentioned when asked. (I think they screen everyone for psychosis.) They focused on that a lot, but did not medicate me for it.

the pseudopatients were not able to obtain their release until

I was voluntarily hospitalized. I don't know if demanding to leave would have worked. I was able to get day permissions and then released basically by being visibly happy then telling doctors so.

they agreed with the psychiatrists that they were mentally ill

Nobody asked me to self-diagnose!

and began taking antipsychotic medications

That one's true - medication (not antipsychotics in my case) was not optional.

No staff member noticed that the pseudopatients were flushing their medication down the toilets

Taking meds is the only time we were actively watched (except for patients on suicide watch).

Their possessions were searched randomly, and they were sometimes observed while using the toilet.

often discussing patients at length in their presence as though they were not there, and avoiding direct interaction with patients

Some attendants were prone to verbal and physical abuse of patients when other staff were not present.

Nuh-uh.

Contact with doctors averaged 6.8 minutes per day.

30 minutes every weekday morning with two doctors and a nurse, in a private room. I think this is standard procedure in Sweden.

Comment author: Vaniver 15 September 2012 02:02:30PM *  0 points [-]

they are not substantially better

They have to be better along the "how long were they detained?" axis because for decades there were about 40% less asylum beds than there had been previously. I think it's likely that the underlying thought processes and biases- i.e. normal people looking crazy enough to diagnose if you already think they're crazy- are not significantly different, and that's what most people care about anyway. (Being able to convince a doctor that I'm sane is more important to me than whether it takes 2 days or 19!)

Comment author: MixedNuts 15 September 2012 05:28:18PM 1 point [-]

Being able to convince a doctor that I'm sane is more important to me than whether it takes 2 days or 19!

Why is that? That's the case if you need psychiatric certification for a job or a medical procedure or something. But generally, being locked up somewhere people can make you take arbitrary medication is bad, and grows worse over time (medication kicking in and unsafe to quit too quickly, effects of detention, damage to your outside life), whereas a piece of paper telling you to take some medication can be ignored. Is there something I'm missing, like insurance premiums?

Comment author: Vaniver 15 September 2012 09:40:06PM 1 point [-]

My preference ordering:

Declared sane and released after 2 days > declared sane and released after 19 days > declared insane and released after 2 days > declared insane and released after 19 days.

I agree that some people might switch the ordering of the second and third outcomes.

Comment author: MixedNuts 15 September 2012 10:02:26PM 1 point [-]

I understand that's your preference ordering, I'm asking why you find being declared insane worse than 17 days of captivity.

Comment author: Vaniver 15 September 2012 10:16:17PM *  2 points [-]

My type 1 systems expect the total discounted long-term costs to be higher. My type 2 systems aren't prepared to do the calculation themselves and are having trouble coming up with justifications, but it seems like self-image and social standing are the most visible concerns.

Comment author: Curiouskid 22 September 2012 06:05:19PM 1 point [-]

Some clippings from my personal document on the experiment:

http://frontierpsychiatrist.co.uk/the-rosenhan-experiment-examined/ (Counter-arguing the conclusions)

http://en.wikipedia.org/wiki/Rosenhan_experiment#Impact_and_controversy (The wikipedia article seems a bit biased in that it omits many of the counterarguments.

(I'm staring a blog soon and the other portions of the document are quite cryptic at this point.)

Comment author: MixedNuts 22 September 2012 07:06:20PM 2 points [-]

Link the first:

Spitzer also writes that ‘schizophrenia in remission’ was a diagnosis rarely used by psychiatrists at the time of the experiment, and as such this indicates that the diagnoses given were a function of the patients’ behaviours

Agree that if all Rosenhan had observed was "discharged with an 'in remission' diagnosis" that would prove hospitals can detect sanity well. But the stays were long - maybe psychosis is much sneakier than depression or hypomania and requires longer observation? And Rosenhan observed more - accepting treatment and agreeing with diagnoses as conditions for release, for example.

Rosenhan reports that the psychiatrists did not spend much time with the pseudopatients. [...] Whilst the medical staff’s lack of engagement with the pseudopatients is regrettable, it does point towards poor clinical skills rather than an indictment of psychiatric classification.

Not sure what difference that makes in practice.

He is struck by what he sees as Rosenhan’s actual failure to provide data demonstrating where normal hospital experiences were categorized as pathological.

Okay for the nurse report mentioning "engages in writing behaviour", though I'd like to know what is reported in that weird style and what isn't. But how about " A group of bored patients waiting outside the cafeteria for lunch early were said by a doctor to his students to be experiencing "oral-acquisitive" psychiatric symptoms."?

The ease with which the pseudopatients gained admission on the basis of what are reported to be mild symptoms

Yup. I don't think that's bad if beds aren't scarce. It's only bad if patients aren't released easily.

Spitzer remarks that the doctors should have been wary of making a diagnosis of schizophrenia in a previously unknown patient presenting without any history of insidious onset. However he is more lenient toward the pseudopatients’ psychiatrists, writing that, given the information available, schizophrenia was the most reasonable diagnosis.

How about referring the patient to someone who knows more about schizophrenia, and can decide if the diagnosis and hospital stay are needed?

Hunter takes exception to Rosenhan’s assertion that the pseudopatients acted ‘normally’ in the hospital:

The pseudopatients did not behave normally in the hospital. Had their behaviour been normal, they would have talked to the nurses’ station and said "Look, I am a normal person who tried to see if I could get into the hospital by behaving in a crazy way or saying crazy things."

Oh, excuse me, I didn't realize the study was supposed to apply to faking experimenters. I thought it was about misdiagnosed patients who figure clamoring they're sane won't help, and might believe the diagnosis.

the pseudopatients would likely not have been, unlike Rosenhan’s assertion, admitted on the basis of their hallucinations solely. Their presentation to hospital and request for admission may also have carried diagnostic weight as it suggested much greater distress.

Good point, but no one's criticizing admission, they're criticizing

throughout their stay, the pseudopatients do not appear to have been assessed in detail.

which pretty much implies that hospitals' function is to lock away loonies, not treat them.

The poor diagnostic skills and apparent lack of curiosity of the psychiatrists that the pseudopatients met is not an indictment of the classification per se, rather its application.

Again, if it's not fulfilling its purpose in practice, who cares?

Rosenhan would favour a classification system based on behaviours:

It seems more useful … to limit our discussions to behaviours, the stimuli that provoke them, and their correlates

Yet despite this early on in the paper he writes that “Anxiety and depression exist”

It's not hard to describe those in terms of behavior. Most likely the benefit is that diagnosing patients would require actually examining them.

his study consisted of only eight subjects

There wasn't that much variation in outcomes, but fine, do a replication.

Overall, this picks some valid nits, but sweeps most interesting data under the carpet. An average of 19 days to notice someone has no symptoms is not negligible!

Link the second:

If I were to drink a quart of blood and, concealing what I had done, come to the emergency room of any hospital vomiting blood, the behavior of the staff would be quite predictable. If they labeled and treated me as having a bleeding peptic ulcer, I doubt that I could argue convincingly that medical science does not know how to diagnose that condition.

Rosenhan replied that if they continue thinking that you still have an ulcer during x weeks despite having no other symptoms of ulcer, that makes for a big problem.

Also, doctors sometimes have to detect malingering outside of experiments.

Kety also argued that psychiatrists should not necessarily be expected to assume that a patient is pretending to have mental illness, thus the study lacked realism.

Yvain does:

When it gets cold or rainy, the hospital fills up with homeless people. Word has spread on the streets that if you go to the emergency room and tell the nurse that evil spirits are telling you to kill everyone, you will get a nice bed and three warm meals a day

Comment author: Curiouskid 22 September 2012 08:28:43PM 2 points [-]

Thanks for spending the time to respond point by point. I'd love to do the same, but this thread would become a bit unwieldy. However, of all the argument mapping software I've looked at, this one seems to be the best: http://workflowy.com/

I transferred your points and counter-points into this and then responded to a few of them (I'll finish responding when I've got a bit more time):

This document can be edited by anybody with this link, so please feel free to chime in. As I mentioned earlier, I'm starting a blog. The goal being to crowdsource ideas on how to make better argument mapping software from the LW community (rather than having discussion isolated to scattered posts). A huge part of this is sketching out example argument maps like the one aove.

For the most part, I agree with the sentiment expressed by Rosenhan. In fact, I agreed so much that I failed (I blame wikipedia :) ) to look at counter-arguments until recently as part of an effort to re-examine my old beliefs and formalize them into argument maps. Thanks for posting this. I wouldn't have been motivated to formalize this into bullet points otherwise.

Comment author: chaosmosis 24 September 2012 05:13:00PM 0 points [-]

Did someone delete all of your arguments? I got there, and nothing was there. Maybe I'm doing something wrong and so I can't see what you're talking about?

Comment author: Curiouskid 25 September 2012 12:10:39AM 1 point [-]

I just checked, and nobody deleted all of my arguments. If you click on this link:

https://workflowy.com/shared/c9e57ddb-d684-ede5-0511-8b8d11c561e0/

Comment author: chaosmosis 25 September 2012 03:28:18AM 0 points [-]

I still see nothing.

Do you see the "meta" note that I made on there?

Comment author: Curiouskid 26 September 2012 05:21:03PM 0 points [-]

I do see your meta comment... I'm not sure what you mean by it though.

Comment author: chaosmosis 26 September 2012 10:33:30PM *  0 points [-]

I mean that by separating the pro and con arguments it becomes more difficult to trace the lines of argument and counterargument. Rebuttals are harder to follow if you have to sort through a list of bullet points to find the one that's relevant to them.

I also made the comment to test whether or not you could see my comments. You can. That probably means that I'm doing something wrong or that you're writing in some equivalent of invisible e-ink. I can't think what I might be doing wrong though. I'm not too concerned about seeing the site though, so don't worry about it.