Monday 5 September 2016

Myths, mental illness and violence: Reply to Prof Lilienfeld

I am grateful to Prof Lilienfeld for responding to my blog post of 7th August.

Correct information about mental illness is an important matter, and worth debating, so it was good to be able post up his “Author replies”. Debate is the essence of empirical enquiry, so that arguments can be tested. I often let the reply stand without comment, but this time I am responding, albeit somewhat hesitantly and wanting to be sure of my interpretations, because the debate raises important issues about the public perception of mental illness.


To get an understanding of my arguments the following posts listed below would be helpful for new readers. For those in a hurry, the first post on “delusional numbers” would be sufficient.

Delusional numbers. 25 March 2013
This gives general background to the research, and discusses methods, and criticises some of the ways in which results have been presented. Summary: an increase in violence by schizophrenic patients over the population norm.

Bad Blood. 2 September 2013
This describes relative risk of getting a severe mental disorder depending on whether your parents have a severe mental disorder. Summary: 2.5 relative risk of getting a disorder. This is not directly related to violence, but relevant to the often-feared issue of hereditary madness.

Genetic story jumps ahead. 17 November 2015
Shows that schizophrenia and bipolar disorder exhibit a genetic correlation with cognition. Again, not directly about violence but about the genetic component.

Schizophrenia and violence. 30 December 2015
An early description of the just released A Sariaslan, H Larsson and S Fazel paper on violence in the mentally ill.

Myths dispelled, new myths propagated. 7 August 2016
Comment on Prof Lilienfeld chapter.

Detecting schizophrenia myths. 10 August 2016
Comment on the Large paper about detecting violent behaviour in schizophrenic patients; also includes more of the published A Sariaslan, H Larsson and S Fazel paper.

Postcodes and schizophrenia. 13 June 2016
Summary: Having psychotic experiences is very highly heritable; having the diagnosis of schizophrenia also highly heritable; and ending up in a deprived neighbourhood is also heritable.

Now, after that digression, back to the main story.

In his book Prof Lilienfeld disposes of the myth:

Myth 43 Most mentally ill are violent.

What I said in my post was: First, note the absolute nature of the myth: what is being dispelled is that a majority of mentally ill patients are violent. The myth is presented in exaggerated form, as a straw man, thus obscuring an important issue. What people fear is that mentally ill people are more violent than mentally well people, and on that basis it may be prudent to avoid them.

Prof Lilienfeld says this is not relevant to the myth he is describing, and that therefore my comments are a non sequitur. I felt that his putting up the “myth” in that strong form and then refuting it led to later misunderstandings, and that is why I worked through his supportive arguments and some of the references he gave in support. So, my apologies to Prof Lilienfeld. I should have made it far clearer that I agreed that Myth 43 was a myth. That was all he was claiming. On that basis Prof Lilienfeld is perfectly justified in reading no further.

Related matters

I do not know how many people actually think “Most mentally ill are violent”, which is the myth that Prof Lilienfeld and colleagues wished to refute in their book. As far as I can see, there is no confirmatory data from national surveys. I don’t think that most people go that far. They just assume or notice a link of some sort, enough to worry them somewhat. The Time Magazine article Prof Lilienfeld quotes in his reply does not confirm that the general public believe that most mentally ill people are violent, as Prof Lilienfeld himself readily concedes, but that they are more likely to be violent, particularly if they have schizophrenia. This is true. Media often exaggerate exceptional events. They are not always reliable sources of information about mental illness. The Time article links to a US government health website, which has this statement:
Myth: People with mental health problems are violent and unpredictable.
Fact: The vast majority of people with mental health problems are no more likely to be violent than anyone else. Most people with mental illness are not violent and only 3%-5% of violent acts can be attributed to individuals living with a serious mental illness. In fact, people with severe mental illnesses are over 10 times more likely to be victims of violent crime than the general population. You probably know someone with a mental health problem and don't even realize it, because many people with mental health problems are highly active and productive members of our communities.

The point of my original post was to question assertions of that kind. Namely, the 3-5% violence rate can only be interpreted if the rate of serious mental illness is also given. Being a victim of violence does not preclude also having been a more-frequent-than-usual perpetrator.

I was also somewhat critical about the “exposing myths” approach, while recognising that it is an engaging format, and that under that rubric some weak arguments have been exposed, and some popular beliefs revealed to have little research foundation. Some myths are myths. My concern is that once something has been called a “myth” then there is a risk it must be destroyed even if part of it is true.

I think that is happened in this case, because there is a middle ground where Prof Lilienfeld and I are in much closer agreement, namely about elevated levels of violence coming from some mentally ill persons, though we still differ in our emphasis as to how significant this is, and whether it explains a public perception of mentally ill persons being dangerous. We both reference work which shows elevated rates of violence among some mentally ill people. He sees those as modest. I see those as more significant, and the likely cause of some people’s over-generalised fear of mentally ill people.
For example, Prof Lilienfeld in his reply references the 2012 work of Van Dorn, Volavka and Johnson and quotes the authors’ conclusion: “..there is a statistically significant, yet modest relationship between SMI [serious mental illness] (within 12 months) and violence, and a stronger relationship between SMI with substance use disorder and violence.”
So, I looked up the figures which are shown in Table 1

The national household sample is a good one, though the period of reporting about violence is the past year only, which might produce lower overall estimates of a general tendency to offend. Violence rates are derived from reports during an interview, not an examination of convictions. The population base rate is 0.83, the Serious Mental Illness (schizophrenia, bipolar, major depressive) rate 2.88, or 3.5 times higher. Combined with substance abuse, the rate is much higher at 9.97, with a high relative risk of 11.45. Briefly, severely mentally ill persons who also take drugs are 12 times more violent than the mentally well. You may question why they take drugs, but those who do so are particularly more likely to be violent. Substance abuse is more common in people with mental disorders. (It is a bit like wayward youths getting into bad company: it is that sort of company many of them choose). Most people in the population are not violent. Those with severe mental illness are 3.5 more violent within one year, and those with severe mental illness abusing drugs, 12 times more violent within one year. As to whether these relative risk rates are modest is something the public can decide for themselves.

Personally, I disagree with the authors that these relative risks are modest (although the absolute risks are low). I judge them to be appreciable, and by the judicious avoidance of such persons the public can reduce their risks, if they wish to.
However, Prof Lilienfeld says that he and I are in agreement on this matter: “I (we) did and do not question Thompson’s conclusion that severe mental illness is tied to a heightened risk of violence, although at least some of this link appears to be mediated by substance abuse.”

(When researchers disagree only slightly, does this causes even more contention? Discuss!)

Prof Lilienfeld mentions that in his 1994 book he made comments on the Teplin (1985) paper, which chapter I did not know about. Ignorance on my part, pure ignorance, for which another apology. I have not been able to get the relevant book chapter yet. Prof Lilienfeld says: “There, I discussed some of the same sticky inferential issues regarding covariate control raised by Thompson.” So, any imputation that Prof Lilienfeld does not understand these issues is unwarranted. However, he did give Teplin (1985) as a support to establish that the rates of violence in severe mental illness were modest, and that was why I criticised that paper. It seems that both of us have reservations about the way that some covariates were controlled for.

Prof Lilienfeld has a further criticism: “Thompson seems to fall prey to the commonplace error of confusing relative with absolute risk ratios (see the following article for an excellent review of this and similar topics: Specifically, the fact that schizophrenia is associated with a several times higher risk of violence does not imply, or come remotely close to implying, that most people with schizophrenia are violent. To do imply thus would be to fall prey to the related error of base rate neglect, as the base rate of schizophrenia in the general population is probably a bit under 1 percent.”

For the avoidance of doubt, I was not suggesting that a higher relative risk meant that most people with schizophrenia are violent, merely that schizophrenia is associated with increased violence.

The link which Prof Lilienfeld has kindly given above is by the commendable Prof Gerd Gigerenzer. I have written 7 posts which reference Gigerenzer, including the one on Delusional Numbers above. Using the search bar on the blog will call them all up, including my commendation of him thus: “It can rarely be said of a psychologist that everything they write is worth reading. Gigerenzer is one such psychologist. He writes in plain English (presumably his second language) and understands his material so thoroughly that he can explain it simply, the sign of an intelligent and honest teacher. This straightforward approach means that you can follow this heuristic to make you smart: if you cannot understand him first time around, it is worth reading him several times until you do. With lesser writers, if you cannot understand them first time, turn elsewhere.”
Out of the 7 posts, here is one which particularly looks at absolute and relative risks.

So, it turns out that both Prof Lilienfeld and I are fans of Gerd Gigerenzer. That does not mean that I have got Gigerenzer right (please let me know if I have misunderstood relative risks or any other of his points, because I may need to read him several more times), but I suppose we could learn from him and move this debate away from percentages and relative risk figures to natural frequencies, which are better understood by the general public. A task for next time.

However, more recent work, not available when Prof Lilienfeld and colleagues were writing their book, suggests a more significant difference in violence rates for particular forms of mental illness. Let us use those findings as an up-to-date test case. How should we describe them in a way which is fair and balanced, and does not lead to misunderstanding and inadvertent myth making?

On 30 December 2015 I did a post on a recent population study about violence and schizophrenia in a full population sample of 1.8 million siblings drawn from 3.2 million men from Sariaslan, Larsson and Fazel (2016).

This massive study gets around most of the problems of representativeness encountered by convenience samples. Also, in typically Nordic fashion, the measure of violence is actual convictions drawn from the National Crime Register which includes custodial and non-custodial sentences and cautions, rather than report at interview, and not one year but lifetime since 1973, so a far more reliable indicator. The main findings are shown in Table 1 which I reproduce below, the findings of interest in the very bottom line:

Here is the link to the paper

Here are the results on violence in pictorial form:

A Sariaslan, H Larsson and S Fazel (2016)

The authors say: We observed that nearly one in four (23%) schizophrenia patients had ever been convicted of a violent crime, whereas the equivalent prevalence was 11% in patients diagnosed with bipolar disorder and 3% in controls.
This paper made me revise my earlier view that there is a fivefold increase in violence in schizophrenia. In this 3 million person population sample, schizophrenics are 7.5 times more violent, bipolars 3.5 times more violent. Some mental disorders lead to violence at rates which are higher than the general population norm. From the public’s point of view, the picture is clear enough.

In some ways this is all very simple: not every mentally ill person is violent, but the incidence of violence in mentally ill persons is raised in at least two conditions, schizophrenia and manic-psychosis (bipolar disorder) to above average levels. So the depiction of mental illness should be balanced, with most patients no more violent than anyone else, but in the case of those two conditions substantially higher, particularly when patients are in-medicated and using non-prescription drugs.
The risks should be given correctly so that people can come to their own conclusions. The link between mental illness and violence has some basis in reality. In some conditions such as schizophrenia the violence rate is raised appreciably. It would be an error to exaggerate the risk, and an error to over-generalize from one condition to all conditions, but in my view it would also be an error not to note that there is an elevated risk in the first place, which may account for some public alarm.

In summary, those who believe: “Most mentally ill are violent” are wrong. Not all the mentally ill are violent.

Those who believe: “A majority of the mentally ill are violent” are wrong. The majority of the mentally ill are not violent.

Those who believe: “Some of the mentally ill are more violent than mentally well people” are right. Some of the mentally ill are more violent than the general population. 


  1. Thank you, Dr.Thompson, for your patient refutation of the good Prof. Lilienfeld! Of course, a high percentage of people in the world, many without much or any education, are well aware that the seriously mentally ill are potentially dangerous and need to be treated with care. Lilienfeld's apparent notions would be regarded as absurd....As Thomas Sowell once wrote, "there are some ideas so dumb that only an intellectual could believe them."

  2. I do not know how many people actually think “Most mentally ill are violent”, which is the myth that Prof Lilienfeld and colleagues wished to refute in their book. As far as I can see, there is no confirmatory data from national surveys.

    My personal lifelong experience of life among the average people testifies to the myth being rather widespread.
    It is a double-sided myth.
    Mentally unsound people are believed to be more violent and dangerous because there is a strong wish to believe that dangerous violence is a product of lack of mental health.

    The average people will have a strong desire to believe he who has done something that unsettles them "crazy", since that is the most expedient way their mind has to regain comfort and keep their rosy, fancifully and neurotypical optimistic view of man.

    If all what you are afraid of is "crazy", you are bound to believe that the "crazy" are far more dangerous and violent and "bad" than the "normal".
    Comfort of mind requires that, and people — all people, of any intelligence, I am afraid — can't resist believing what assures their mental comfort.

    A particularly topical example is the concept of "delusion".
    As "mad" and "insane" have become politically incorrect categories, the socially fit will use "delusional" as their shield from anything that hurts them in the deep.

    Day 1: Girls begs boy to date her: indirectly, yet explicitly.
    Boy turns her suggestion down.

    Day 2: Girl's vanity has been hit hard. She, a part of hers actually, manufactures one or another bogus reason to bad-mouth Boy. She has to take out her frustration, hasn't she?
    Boy points out how chick-logical is to asperse they whose liking we did everything to achieve one day earlier.
    Girl: It was a joke! I never wanted to date you!
    Boy happens to have some time to waste. Thence, he proceeds to logically examine the before day's conversation, proving that, without a doubt, it was not a joke, not at all.
    Girl — as soon as no pretence of logicality can be given to her responses —: you're completely delusional!

    You give a look at people arguing on social media, and see it yourself.

    Foucault's Madness and Civilization: A History of Insanity in the Age of Reason does not cast light on anything we don't know, but it argues its points with majestic rhetorical skill.

  3. A nit-pick, or perhaps more: one can't read any sort of social science without coming across the term "significant" in the sense of "statistically significant", a technical term the meaning of which is entirely obscure to almost the whole population. So be it.

    In the context of social science it therefore seems to me bad judgement to also use "significant" in some sort of layman's sense, probably meaning "important". Thus I object to "we still differ in our emphasis as to how significant this is": nope, I argue that you differ in your assessment of its importance.

    In fact the word "significant" has come to annoy me a great deal, my having spent some years working with a colleague who seemed to use it, all too frequently, as a irritatingly inaccurate substitute for "important", "substantial", "salient", and many other words - once even "tall".

    I offer a rule of thumb: if one says that "A is significant" one should be prepared to explain what A signifies. If one doesn't generally have a good answer, one should drop the pompous - nay, affected - term from one's vocabulary.

    I do wish academics would suppress their snobbery, and say "big" when they mean "big". Do I imply that "significant" is usually an absurd genteelism? Indeed I do. Am I saying that the academic Emperor frequently has no intellectual clothes? You betcha. Am I implying that he's being lah-di-dah towards his less educated fellow citizens? Spot on.

    Nothing personal, old fruit, but deeply felt. Joe Bloggs pays his taxes to pay us; it's bloody rude to use words chosen to exclude him or make him feel inferior.

  4. You raise a significant point, which must be addressed, etc. Yep, agree with you that statistically significant does not mean actually important in the usual sense. As to the best way to use "big" in social science chatter, then I suppose effect size is what we are talking about. So, although I agree with your attack on affectation and snootiness, of which there is much in academia, we need to explain effect sizes, perhaps thus: an effect of size of 0.6 is big because two thirds of treated patients would get better.

  5. Thank you for the link - bookmarked. I realise I'm just preaching what Fowler or perhaps Gowers said. Orwell would be a good model.

    Naturally I have no objection to the technical jargon used within a profession: work would be impossible without. Where would the engineers be without their "debottlenecking" or the physicists without their "black holes"?

    I realise that I am probably over-reacting to the behaviour of an annoying twerp years ago, but for some reason his routine misuse of "significant" came to epitomise his silliness. Pretentious ass.