Skip to content

The Varieties of Suicidal Experience


Perhaps the hardest thing that I must do in my professional career is disagree with those whom I deeply respect.  In The Varieties of Suicidal Experience, I find that Thomas Joiner sometimes takes positions that I don’t believe fit the evidence – and I believe that there are important insights to be learned.  This is not the first of Joiner’s books that I’ve reviewed.  Why People Die by Suicide and Myths about Suicide are both his.  As I started my reading on suicide, they formed critical foundations for what I believed.  Let me start with the disagreements and move on to the insights.

Suicide and Mental Illness

Here, Joiner states, “I insist that any and all suicidal behavior is a manifestation of at least one mental disorder of some kind and of some above-zero level of acuity, an important and debatable point.”  (Acuity is the severity or complexity.)  The issue I take with the statement is that it is stated with an absolute nature.  It’s not most but rather all.

Some of Joiner’s research reviewed an old study done by Eli Robins at Washington University in Saint Louis.  The study was done in 1956 and 1957 and used 134 deaths.  The original research fell well short of the 100% mental illness that Joiner proposes, but his reevaluation of the data moved people from no discernable mental illness to having a mental illness.  This makes me feel uneasy.  The original researcher, one who did the first retrospective study approach that Shneidman would later call “psychological autopsy,” would seem the best positioned to evaluate his data.  As a sidebar, Robins’ work was roughly simultaneous with Shneidman’s work, so while Shneidman is credited with the approach, it’s perhaps best stated that it was simultaneously discovered or invented by Robins.  (See also Autopsy of a Suicidal Mind.)

There are two nits with the whole discussion.  The first is, what does Joiner mean by a mental disorder above zero-acuity mean? Second, why does this all matter?  To the first point, most psychological diagnoses require some set of conditions and/or duration.  What Joiner is saying is that there are signals – but they don’t rise to the level of an official diagnosis.  That requires understanding the difference between actionable signals and noise.

Signal processing (and the artificial intelligence that, in part, builds on that work) recognizes that everything has some part of what you want – the signal – and some part of what you don’t – the noise.  The goal in signal processing is to isolate the signal from the noise.  We’ve developed numerous techniques to do this – and I routinely use tools that implement them when producing videos.  However, this relies on the ability to distinguish the signal from the noise, and that’s not so easy.

Let’s assume the threshold of sensitivity for suicidal ideation is someone who no longer finds joy in things they once enjoyed (anhedonia).  It’s a key marker for depression.  As markers go, it’s a pretty good one.  However, the problem is that nearly everyone has experienced it – at least for a short time – at some point in their lives.  They’ve gotten bored with a hobby and either put it on pause or have given it up.  If we use even this very important signal as our minimum bar, we essentially flag everyone as having the kind of mental illness that Joiner is proposing.

So why does it matter?  Well, there are a few answers to this, but they revolve around the impact it has on people – both those left behind from a suicide loss and those considering it.  For those considering it, it establishes a shame cycle.  For those considering suicide, the thinking may be, “If I’m considering suicide, then I must have a mental illness – and I should be ashamed of that.”  I don’t believe that considering suicide means you have a mental illness – research points to one in three of us will at some point in our lives.  Nor do I believe being ashamed of a mental illness is fair or right.  Whether this is rational or not isn’t the point.  As Capture points out, it doesn’t have to be rational in a broader sense to make sense to the person in the moment.

For those left behind, they’re left with the belief that they missed something.  If the person had a mental illness, then it should have been detectable, and they should have seen it.  They missed it, so they’re responsible for their loved one’s death.  Let me be clear: this isn’t the case.  However, it’s the thoughts that many loss survivors have repeated to me.

Suicide Predictability

Joiner, perhaps rightly, takes issue with those that say predictability of suicide screenings is little better than a flip of a coin.  (I’ll admit I once had a room of suicide prevention-interested people flip a coin to make this exact point.)  The problem is that, while Joiner’s literal point is true that predictability is over a 50/50 split, it’s not by much.  He quotes George Murphy from 1972 as saying, “From the numerical standpoint, the prediction of no suicide in every case would be highly accurate. … It would also be entirely unacceptable clinically.”  This is the problem as Craig Bryan aptly points out in Rethinking Suicide. That is, it depends on what your goal is. If your goal is accuracy, then identify no one.  If your goal is prevention, you’ll need to identify (hundreds of) multiples of the people who would die by suicide.

He explains that we can’t predict the number of people who are going to be in an accident individually – though the statistics allow us to estimate the total number.  That’s what we have with our research, tools, and predictive models.  We can predict a rate within a population, but what we really want, what we desire more than anything else, is to be able to know who might die so we can intervene.

The tools we have now are excessively sensitive.  That is, they say that people are at risk when they are not.  The result is that we flood the mental health system with people who aren’t suicidal.  Lisa Horowitz at the National Institutes of Mental Health (NIMH) has argued that these people need help, too – even if they’re not suicidal.  I’m happy to accept this on face value.  They do.  However, the unresolved problem is whether they need it more critically or acutely than others.  The answer is no – and we’re back to the problem of prioritizing our available mental health resources to those with the greatest needs.  So, the harm in relying on screening tools for individual prediction is that they don’t work, and they flood the mental health system with patients who, though deserving of services, aren’t the most critical.

Joiner continues the discussion with, “In ‘reasoning backward,’ we have saved millions of lives by learning from specific events like car and plane crashes, though it remains difficult to prospectively predict individual accidents.”  This is quite right.  However, it’s also very different.  In the case of automobile accidents, the greatest gains came from a set of design decisions and approaches, as Ralph Nader points out in Unsafe at Any Speed.

We’ve poured untold millions into programs to encourage seatbelt usage over 50 years.  We’ve made amazing progress.  However, we’re still only seeing the low 90s percent utilization of seatbelts (based on miles traveled, information from the US Department of Transportation).  My point is that when it comes to changing human behavior – or understanding it – it’s not as easy as it seems.  I’m not convinced that the screening activity actually points to anything useful – and I’d rather see us invest our energy on potentially more useful strategies.

Suicide and Impulsivity

Joiner also puts forth the idea that no suicide is truly impulsive.  Specifically, he states, “A key refrain of this book is that suicidal people know what they are doing even if they do not reveal it to others, and that this applies with equal force to phenomena like murder-suicide, suicide-by-cop, and so on.”  Later, he says, “Some believe the main mechanism involves impulsivity—more specifically, that an impulsive resort to a lethal method may be prevented by distance from that method, allowing the impulse to pass. Perhaps occasionally this is so, but the role of spur-of-the-moment processes in lethal suicidality is dubious.”  What does this mean?  There is at least some allowance for impulsivity, but Joiner’s perception of the frequency of the phenomenon is radically different than the research.

Though there are many different studies with different rules, timings, and results, there’s more than a few interviews of suicide attempters (who didn’t die) where the large proportion hadn’t considered suicide prior to a few hours before their attempt – on the order of 70%.  Joiner properly criticizes the structure of some of these, because their questions were leading towards the answer that they hadn’t considered suicide.  However, other studies record rapid and dramatic fluctuation in suicidal thought intensity.  I’ve seen no evidence that this shouldn’t apply to people with no suicidal thoughts rapidly developing them.  In fact, Craig Bryan in Rethinking Suicide recites a personal story of a Marine whose suicidal thoughts came on quickly and intensely.

The other argument is that the attempter studies necessarily don’t precisely represent those who died by suicide.  I concur.  There will be some gap between those who have died and those who lived.  That being said, I don’t know that I can move the needle from majority (~70%) to a rare event.  Of course, we can’t know – but that’s a lot of movement when the impact of living versus dying is so close.

What I suspect may be happening is WYSIATI – What You See Is All There Is.  (See Thinking, Fast and Slow and Incognito.)  The clinical experience of suicidologists will necessarily exclude all impulsive types.  The only people that suicidologists will see are those people who are concerned about their suicidality – or those for whom others are concerned about their suicidality.  Therefore, the “feeling” that suicidologists and clinicians will have is that it is a rare (or never) event.  As Taleb notes in The Black Swan, failing to see something doesn’t mean that it doesn’t exist.

Let All Flowers Bloom

Joiner explains that he’s concerned about places where researchers provide alternative theories and approaches.  He suggests a philosophy of “Let all flowers bloom.”  In short, let people do what they want and let the evidence decide.  Conceptually, I concur.  Let’s try things and hope to find answers.  However, structurally, I disagree.  Sometimes, flowers are weeds, and they’re choking off other approaches.

Perhaps the greatest example of this is the preoccupation with screening for suicide.  So many hospital systems, forced by The Joint Commission, are implementing suicide screening programs.  The Joint Commission accredits hospitals, and not being accredited isn’t a real option.  The Centers for Medicare and Medicaid Services (CMS) represent a substantial portion of the healthcare system volume.  CMS doesn’t require The Joint Commission accreditation – but they have said that it meets all the CMS requirements.  So, if you’re accredited, you pass CMS requirements.

The Joint Commission requires suicide screening, so it effectively becomes required for most hospitals – and almost all that want CMS patients.

The problem, as indicated above, is that the approach breaks the mental health system.  It makes it impossible for the system to take care of patients who need care, because it’s flooded with patients who aren’t in crisis or even acute need.  (If you don’t believe this, call a random mental health and ask them how long until you can get in to see them.)

Deaths of Despair

Joiner makes the point that his arguments overlap those made in Deaths of Despair and the Future of Capitalism and Bowling Alone.  I’d add to these Loneliness and Social Forces in Urban Suicide.  At the root of this is the sense that people die by suicide because they’re alone and suffering.  However, despair isn’t evenly distributed.  He explains that the suicide rate in the world has generally been trending down, while that in the United States has climbed.  He did acknowledge that the US rate fell in 2019 and 2020.

I think it’s important to note that the relationship suicide rates have with turmoil is complicated.  It would surprise no one that suicide rates during the Great Depression were multiples of our current rates.  It might be odd, however, to consider that, during the World Wars, the rates seem to have been lower.  It seems that the focus on unity and helping others has a stabilizing effect.  This won’t surprise people who’ve been through a twelve-step program, as they’ve seen how those who serve others are more likely to succeed in the program.


Joiner explains the similarly complex relationship that suicide has with alcohol.  Many people believe that alcohol is implicated in most deaths by suicide, but that’s not the case.  In multiple studies, more than 60% of people don’t have any alcohol in their system at the time of their death.  That doesn’t mean that the remaining were intoxicated, but rather they only had some alcohol in their blood.

Contrast this with the longer term, and we can understand the perception that alcohol plays a larger role in suicide.  The problem with looking at the longer term is that there are interferences caused by socioeconomic factors.  A non-functioning alcoholic loses relationships, employment, and shelter.  Thus, when analyzing the impacts of long-term alcohol use, it has an effect – but at least in part because of the other outcomes of alcoholism.

Alcohol isn’t unique in this regard and in fact may be a slower, more muted driver.  As The Globalization of Addiction and Chasing the Scream both explain, other substances have similar downstream impacts.  Observationally, I’d say that they often occur much more rapidly.

Barriers Loom Large

While dismissing impulsivity, Joiner argues that suicide is difficult, and therefore even small barriers loom large in suicide prevention.  (See also Nudge for small barriers.)  Studies have shown that delaying access to or blocking suicide means reduces deaths.  Specifically, barriers on one bridge reduce the deaths at that bridge and don’t transfer those deaths to other bridges.

The public often believes there’s no point in blocking an attempt or access to a means.  After all, they’ll just find another way, they argue.  However, the data doesn’t support that.  First, evidence from changing gas formulation in the UK and fertilizer formulation in Sri Lanka to make them less lethal point to little means substitution.  In the case of those who were aborted from their attempt on the Golden Gate Bridge, a substantial majority never die by suicide.  Even with the elevated rate of suicide by those with previous attempts, roughly 90% of those who have attempted suicide never die by suicide.

It’s no secret that we fear death.  Whether we want to call it an evolutionary imperative to live or simply accept that most people fear death, the data supports our desire to live.  (See The Denial of Death and The Worm at the Core for more.)  Joiner rightly asserts that overcoming this bias towards life isn’t easy.  It’s important to counter a myth that no other species have individuals dying by suicide.  That’s simply not true, as there are numerous species with various forms of self-termination that occur at rates well better than chance.  (See Why Zebras Don’t Get Ulcers for some examples.)


Joiner asserts that murder-suicides should be thought of as suicides that decide to include murder rather than murderers who decide to die by suicide.  The framework of the argument is that people are suicidal, and then decide that it would not be fair to leave others around – either because they harmed the person or because it’s unfair to leave them in such a cruel world.  He cites the review of mass murders in Columbine, Aurora, and Parkland as examples.  (For more on Columbine, see No Easy Answers.)

This is an interesting argument – however, the number of murder-suicides that we face as a society is vanishingly small.  They’re tragic – but they aren’t frequent.  Generally speaking, it’s 2% of suicides.  When we’re already dealing with a number that’s 14 per 100,000, that is a very hard number to get good data on.

Staring Down Death

One interesting, sub-clinical, indication of suicidal intent is a lowered blink rate.  It seems that people who have suicidal ideation blink at a rate slower than the general population.  While this is an interesting correlation, there are many other causes for a lowered blink rate, and therefore it may not be particularly useful as a screening for suicidal ideation.  Joiner states that it is as if “they were in a sense staring down death.”

Suicide by Cop

The first time I heard the phrase “suicide by cop” was decades ago.  I was watching a news report with my brother-in-law, and he said it casually but also as a matter of fact.  He was a lifelong law enforcement officer.  I can’t remember the details of the event, but it wouldn’t surprise me if it were a man who brandished an unloaded firearm at an officer – forcing the officer to shoot.  Joiner explains that 14% of the weapons brandished to an officer were unloaded.

Final Exit

Joiner takes issue with an “exit guide” who helps guide people to suicide.  The guide in question seemed to be a part of the Final Exit Network – likely built on Derek Humphrey’s Final Exit work.  Certainly, we should not treat suicide as casually as we’d treat walking across the room.  However, simultaneously, we have to be careful about overarching statements that assisting someone in suicide is wrong.  Undoing Suicidism makes the point that we’ve gotten too paternal in our approach, and even in places where physician assisted suicide is an option, the constraints are prohibitive.

There are no clean answers here.  On the one hand, we have a responsibility to our brothers and sisters to ensure that their decisions that would remove them from our community of the living are properly considered.  On the other hand, we need to accept their right to make a choice.  I don’t know the answer here, but I do know that I don’t know the answer, just as I don’t understand all The Varieties of Suicidal Experience.