Another week, another internet generated furore to pick over, this time in regards to a now with drawn viral advert by the Korean car company Hyundai, albeit one that puts me in the rather unusual position of having to criticise Ben Goldacre for pushing a bit of dodgy science.

You probably saw it for yourself but if you didn’t the back story here is that Hyundai released a rather clever if highly unusual viral video advert to promote the environmental credentials of one of their cars which depicted a man trying, and failing completely, to commit to suicide by piping the car’s exhaust through in the passenger cabin.

The punchline of the ad is, of course, that the car’s exhaust emissions are so clean that that the ad’s fictional would be suicide was wasting his time trying that method, which is actually true of pretty much all new cars built since 1995, which have catalytic converters as standard. The science here is catalytic converters oxidise the noxious chemicals in car exhaust fumes, notably carbon monoxide and any unburned hydrocarbons, in to plain old carbon dioxide and water, in addition to reducing any nitrogen oxides to nitrogen and oxygen, so although its by no mean impossible to commit suicide by this method – one could still use a pre-1994 car without a catalytic converter, a car with a faulty converter or tap the exhaust system ahead of the converter – it is nevertheless much more difficult to commit suicide by this method as most modern cars do not produce anything like enough carbon monoxide out their exhaust to do the job.

There are a couple of things that have emerged out of this that merit comment. One is the Open letter to Innocean and Hyundai by Holly Brockwell, an advertising copywriter whose father sadly committed suicide via the exhaust fumes method way back in 1990 – i.e. before catalytic converters became standard equipment on all new cars.

The letter itself is moving and eloquent and has proved to be highly effective in propelling the Internet’s great waggy finger of tut-tut in the direction of Hyundai but, contrary to some of the opinions I’ve seen kicking around in the last day or so,  not of itself reason enough for Hyundai to remove the video from circulation. While it’s patently obvious that this letter will rapidly be encapsulated into the rapidly growing sphere of online folklore surrounding ‘trauma triggers’ this will ultimately prove only that the very concept of triggers has now become so detached from its correct clinical meaning as to have rendered almost entirely meaningless.

Put simple, there is a world of difference between a stress-induced dissociative episode, which is what those with PTSD experience when triggered, and finding something upsetting because it evokes unpleasant and unhappy memories, and blurring the line between these two very different phenomena – the latter being an entirely normal grief reaction – is likely to prove counter-productive in the long run by creating a culture which increasingly accepts the de facto medicalisation of essentially normal experiences while at the same time devaluing the experiences of those who, unfortunately, do develop PTSD.

That’s really an issue for another day and an another article, suffice to say for now that there are good reasons for both scepticism and caution in the face of the increasing ubiquity of so-called ‘trigger warnings’ on the internet which, however well-intentioned, are still very much subject to the law of unintended consequences.

The second thing I want to pick to pick up stems, as I’ve already mentioned, from Ben Goldacre’s response to this advert, which he describes as ‘almost surreally misguided‘.

Ben’s entitled to his opinion, of course, but as to his claim that there is “clear evidence that this kind of content increases the use of specific suicide methods”, well there I’m afraid neither the evidence nor the underlying issues are anything like as clear cut as Ben seems to be suggesting.

In this latest article, Ben cites, but forgets to link to, one of his own articles from 2009, an abridged version of which also appeared in his Bad Science column in the Guardian. Rather than quote Ben’s current Hyundai post, I’m going to quote  a little more fully from his unabridged 2009 article than he has on this occasion because this will help to clarify the issues I’m driving at – the additional section of Ben’s original article that I’ve added is given in italics:

In The Sorrows of Young Werthe by Goethe the hero shoots himself because his love is unattainable. It was banned after men throughout Europe were reported to be dressing like Werther, copying his affectations, and taking their own lives in the same style.

But a myth about a book is not enough: you need research. And it has been shown repeatedly that suicide increases in the month after a front page suicide story. There is also evidence that the effect is bigger for famous people and gruesome attempts. You may want to remember that fact for later.

Details matter, as ever. Overdoses increased by 17% in the week after a prominent overdose on Casualty (watched by 22% of the population at the time), and paracetamol overdoses went up by more than others. In 1998 the Hong Kong media reported heavily on a case of carbon monoxide poisoning by a very specific method, using a charcoal burner. In the 10 months preceding the reports, there had been no such suicides. In November there were 3; then in December there were 10; and over the next year there were 40. You may want to remember that story for later.

And it’s not pie in the sky to suggest that the media should be careful in how they discuss suicide. After the introduction of media reporting guidelines in Austria, for example, there was a significant decrease in the number of people throwing themselves under trains.

So organisations like the Samaritans take this seriously. They suggest that journalists avoid crass phrases like “a ‘successful’ suicide attempt”. They suggest that journalists avoid explicit or technical details of suicide methods, for reasons you can now understand. They suggest that journalists include details of further sources for help and advice, since an article about suicide represents a great opportunity to target people who are at risk with useful information. And they recommend avoiding simplistic explanations for suicide.

Okay, so the phenomenon we’re looking at here is one that is commonly referred to as ‘copycat suicide’, ‘suicide contagion’ or, more poetically, as ‘The Werther Effect’ after the eponymous hero of Goethe’s novel and the impact that it anecdotally had on the young men of Europe in the wake of its publication in 1774

Ben, as you’d expect, supplies a number of links to studies which back up his assertion that it has been shown repeatedly that short-term increases in suicide rate follow in the wake of high profile suicides that attract significant media coverage. Of these, the key paper in terms of accessibility, it being the only one where the full text is not behind a paywall, is the paper which noted an apparent 17% increase in overdoses in the week following a 1996 episode of the TV show “Casualty” – Hawton et al. (1998) “Effects of a drug overdose in a television drama on presentations to hospital for self poisoning: time series and questionnaire study” – in which a character committed suicide by taking a paracetamol overdose.

Now there are a few thing in this paper I could quibble about, particular in terms of the numbers – out of 484 patients who complete a research survey in the three weeks after the episode of Casualty was broadcast, 426 of whom had taken an overdose in which one of the drugs used contained paracetamol, just 69 had actually seen the programme and only 10 claimed that it had influenced their decision to take an overdose while 11 claimed it influenced their choice of drug. On those figures, it is evidence, yes, but rather weak evidence, particularly as the paper also notes that some patients claimed to have been put off the idea of taking a paracetamol overdose because the programme had also highlighted the dangers of liver damage, although how many patients took this view is not stated.

But rather than hack through the entire paper picking up its faults, what I want to draw your attention to is the very first paragraph of its introduction, which reads as follows:

The possibility that media representation of suicide and deliberate self harm may encourage suicidal behaviour in vulnerable individuals has attracted considerable attention, not least because it is a potentially modifiable factor. Studies of televised news reports of suicides have suggested associations with a short term increased incidence of suicide, especially if the reports are repeated and the deaths are highly publicised. Others have not shown such an effect. Investigations of the effects of fictional portrayal of suicidal behaviour on television have also produced varying results, with some studies indicating a strong influence on suicides or on referrals for deliberate self harm. Recent studies in the United Kingdom found either no evidence or equivocal effects. Most studies in this area have been retrospective so that it has not been possible to investigate whether subjects have seen the media stimulus.

Of the paper’s 25 references, 14 appear in that paragraph alone – check the original paper if you want to see which ones and where – and as you can see, the picture this paragraph paints of the overall state of research into ‘The Werther Effect’ is rather more mixed and complicated than the impression you might well get from Ben’s rather bald assertion that its has been “shown repeatedly that suicide increases in the month after a front page suicide story”. Some studies, it points out, ‘have not shown such an effect‘ while studies looking at fictional portrayals of suicidal behaviour on television have generated ‘varying results‘ and, indeed, recent – at the time – studies from the UK  were noted as having found ‘either no evidence or equivocal effects‘. Nowhere else in this paper is the existence of any apparently equivocal or contradictory evidence acknowledge.

Indeed, when we come to this paper’s discussion and conclusion sections only one of 14 papers cited in the introduction is referenced for a second time, with the authors noting that their findings are ‘in keeping with those of an earlier study in adolescents in the United States’  – Gould & Shaffer (1986). This may indeed be true but it is nevertheless a rather odd observation given that the Casualty study makes no mention whatsoever of age and is based on data for self-poisoning admissions in the general population and not adolescents, as was the case for the US study.

However this does neatly allow the authors to avoid drawing comparisons with another paper from its list of introductory references, one that is much closer fit for their own paper because it looked at the data for admissions for self-poisoning from UK hospitals following a prominent fictional suicide attempt on British television – an aspirin overdose taken by the character Angie Watts in the soap opera ‘Eastenders’ in February 1985.

That study – Platt (1987) – was also published in the BMJ and, indeed, also found evidence for an increase in self-poisoning admissions in the immediate aftermath of the programme being broadcast, but despite these similarities, it also arrived at a rather different conclusion:

The findings presented in this study are somewhat conflicting.

There was a 15% increase in the total number of cases of self poisoning during the day after the broadcast and an even larger increase (31%) three days later. When evaluating the significance of these changes, however, we must take into account the upward trend in the number of cases of deliberate overdose during the period under review; the equally or more dramatic changes that occurred outside the experimental period; and the failure to show that the increase in the experimental period was significantly greater than chance after controlling for trends in 1985.

…

How, then, can we summarise the evidence of an imitation effect after Angie’s overdose in the EastEnders soap opera? Within the limitations of the study design a conservative but fair conclusion must be “Case not proven.” Though the portrayal by or reporting of suicide in the mass media has been shown to increase suicidal behaviour in the general population, clear evidence that the presentation of fictional parasuicide has a similar effect is lacking. In the debate about the coverage of suicidal behaviour in the newspapers and on film and television the possibility of a differential impact according to the type of stimulus event (suicide v parasuicide) should not be overlooked.

To be clear, the results of the Casualty study are entirely consistent with those reported eleven years earlier by Platt; the key difference between the two papers is that Platt carried out an addition series of analyses which sought to place the observed effect following the Eastenders episode into the broader context of overall trends in self-poisoning admissions and its that comparison, which is entirely lacking in the Casualty study, that led him to report equivocal findings overall.

Notwithstanding the equivocal findings of his own study you see that Platt still not to question the general validity of the Werther Effect hypothesis. That his own study fails to provide clear evidence in support of that hypothesis is rationalised away in terms of the possibility of a differential impact between fictional portrayals of suicide in the media and the real world events found in newspapers and television news reporting.

Is that necessarily true?

Well, let’s return to Ben’s assertion that the ‘evidence is bigger for famous people and gruesome attempts’ by looking specifically at what was arguably the most high-profile celebrity suicide of the the last 30 years, that of Kurt Cobain on April 8 1994.

As high profile suicides go, Cobain’s death by way of a self-inflected gunshot wound to the head ticks all the Werther Effect boxes.

Following the release of Nirvana’s seminal 1991 album, ‘Nevermind’, and the global success of the first single from that album, ‘Smells Like Teen Spirit’, Cobain was hailed as the voice of Generation X, the  ‘poet emeritus‘ of a disaffected and alienated generation of young adults railing against what they perceived to be the rapidly decaying ‘me-first’ society constructed by the post-war baby-boomers – a generation that the comedian George Carlin memorably decried for having traded ‘cocaine for Rogaine’:

The Baby Boomers: whiny, narcissistic, self-indulgent people with a simple philosophy: “Gimme that! It’s mine!”

These people were given everything, everything was handed to them, and they took it all, sold it all; sex, drugs, and rock ‘n’ roll, and they stayed loaded for twenty years and had a free ride. But now they’re staring down the barrel of middle-age burnout, and they don’t like it. They don’t like it, so they’ve become self-righteous, and they wanna make things hard for young people. They tell em abstain from sex, say no to drugs. As for rock ‘n’ roll, they sold that for television commercials a long time ago so they can buy pasta machines and StairMasters and soybean futures.

You know something? They’re cold, bloodless people. It’s in their slogans, it’s in their rhetoric: “No pain, no gain,” “Just do it,” “Life is short, play hard,” “Shit happens, deal with it,” “Get a life.” These people went from “Do your own thing” to “Just say no!” They went from “Love is all you need” to “Whoever winds up with the most toys, wins”, and they went from cocaine to Rogaine.

And you know something? They’re still counting grams, only now it’s fat grams.

And the worst of it is we have to watch the commercials on TV for Levi’s loose-fitting jeans and fat-ass Docker pants because these degenerate, yuppie, Boomer cocksuckers couldn’t keep their hands off the croissants and the Häägen-Dasz and their big fat asses have spread all over and they have to wear fat-ass Docker pants. Fuck these Boomers, fuck these yuppies… and fuck everyone, now that I think of it.

So when Cobain committed suicide, the news of his death received blanket media coverage across the English-speaking world, not least because Cobain was 27 years of age when he committed suicide, invoking the legend of the ‘27 club‘ – although this has since been roundly debunked by a study published, again, in the BMJ, which found although musicians appear to be prone to an increased risk of death in their 20s and 30s there is nevertheless no evidence of a peak in those mortality rates at the age of 27.

Bearing all that in mind, I’m sure you’d agree that Cobain’s death looks to be ideal fodder for ‘The Werther Effect’ and, indeed, in the years following his suicide, researchers dutifully reported that the Seattle Crisis Clinic had received more than 300 calls on the day after his death had been announced in the media, over 100 more than usual (Toltz & O’Donnell, 1994) while there were also – according to Martin & Koo (1998) – at least “two well-publicized sets of suicide in young Australians” which were reputedly linked to Cobain’s death via suicide notes in which he was mentioned by name.

The 1998 study by Martin & Koo is of particular interest here because it’s one of two papers from Down Under critiqued by Australian sociologist Gerald Sullivan in a paper published in 2007 in “Remember Me: Constructing Immorality – Beliefs on Immortality, Life and Death” (ed. Margaret Mitchell) – the other paper (Baume, Cantor & Rolfe, 1997) is one of the earliest papers looking at suicide-related sites on the Internet.

As Sullivan’s paper – which I recommend you read in full – notes, following Cobain’s suicide Martin & Koo dutifully began a search for evidence in support of ‘The Werther Effect’ hypothesis in the Australian data on suicide rates in young people in the full expectation that they would find an increase in suicides*, and in particular in deaths by gunshot, associated with the media coverage of that of Cobain.

* For obvious reasons, here, Martin & Koo had no need to look also at hospital admissions as Cobain’s chosen method of suicide – a shotgun to the head – doesn’t tend to leave much room for failed attempts.

Sullivan reports Martin & Koo’s findings as follows:

They compared the suicide rate in Australia in the period following the media coverage of Cobain’s suicide with that in the same period in the previous five years, controlling for uneven variability in weekends, Mondays, and holidays. Contrary to their expectation, Martin and Koo found no evidence of an increase in deaths by gunshot, the method Cobain used. Further, they found that while the suicide rate for young people (15-24 years) had been increasing slightly in the April-May period under investigation each year from 1989 to 1993, it dropped (from 39 to 30 cases) in the same period in 1994, which led them to the conclusion that “celebrity suicide had little impact on suicide in young persons in Australia’’ (1997, p. 187).

Whoops… but that’s not the end of the story because Sullivan then goes on to document Martin & Koo’s response to their dismal failure to find evidence of a ‘Werther Effect’ associated with the death of Kurt Cobain:

Given the results of their study, it is interesting that Martin and Koo’s opening sentence summarizing previous research in the field is so uncritical:

The research evidence that newspaper stories about suicide may lead to imitation is convincing … despite equivocal finding [sic] …. Hassan (1995) has reinforced both the concerns about imitation or ‘copycat* suicide and contagion and the conventional wisdom that a very cautious approach should be taken to media reporting of such events (1997, p. 187).

Martin and Koo referred to an unpublished study by Martin in which he examined the impact of a suicide portrayed in a popular Australian television drama, broadcast in 1993. Although he found “no measurable effect” (1997, p. 189), they explain this as not the problem of the design, but rather that the audience penetration of the program was limited to 17% of the Australian viewing public … arguably the series may not have appealed to, nor been seen by, young people (1997, p. 189).

Once again, given the results, Martin and Koo’s conclusions seem extraordinary and indicate slavish loyalty to the media imitation thesis despite a lack of evidence. Martin and Koo restated the thesis and uncritically cite studies that support it:

The work of Phillips … stands out as showing that both newspaper reports and television stories may influence particularly young people to suicide. This evidence supports the general public view that if suicide is talked about, particularly if it is glorified, then vulnerable young people will consider it as an alternative when they are struggling with … problems which at the time appear to be insurmountable (1997, pp. 193-194).

Although they acknowledged dissenting data obtained in previous studies, consistent with their own, they dismissed or ignored these studies:

Apart from researchers who have had technical difficulties in discerning a clear effect of influence, the one dissenting voice comes from Kessler et al. (1988) who showed that during 1981-1984, teen-age suicides decreased after newscasts about suicide. As previously noted, this work has been attacked… (1997, p. 194).

In keeping with Platt (1987), rather than accept that their findings might raise significant questions about the general validity of the Werther Effect hypothesis, Martin & Koo went to considerable lengths to downplay and explain away their own findings and dismiss other studies that had similarly failed to generate evidence in support of the Werther Effect – anything other than admit the possibility that maybe, just maybe, the ‘media imitation thesis’, as Sullivan calls it,  isn’t anything like as clear-cut or reliable as its proponents would have us all believe, e.g. (from Sullivan):

Surprisingly, Martin and Koo referred to a fundamental design flaw in many studies about the effect of media reporting of imitative suicide, the ecological fallacy, as a way of dismissing their own findings, but again, they found this explanation implausible given the amount of media coverage of Cobain’s death. They considered whether the effect might have been undermined by a conspiracy theory published on the Internet that Cobain’s death was not suicide, but they also dismissed this explanation because it “was raised too late to have influenced events in the first month” (1997, p. 195).

… In the final paragraph, while conceding that “the expected copycat effect did not occur” (1997, p. 196), they also suggested that “given previous work it is possible that any increase in suicide was disguised by a marked increase in the MVA (motor vehicle accident] rate of other deaths” and recommended “further exploration … [of] what was special about the reporting in this case that may conceivably have reduced the likelihood of influence” (1997, p. 196).

The theoretical basis for The Werther Effect is supplied, in the main, by Social Learning Theory (warning – desperately one-sided and uncritical article) which argues that social behaviour is learned within a social context primarily by observing and imitating the behaviour of others – back in my university days I used to think of SLT as the ‘monkey see, monkey do’ model of human behaviour as an easy means of remembering it’s core principles.

As such, The Werther Effect shares its theoretical underpinnings with a number of other claimed media imitation effects which, by and large, tend to subjected to altogether greater levels of skepticism and scrutiny, notably those linking depictions of violent behaviour in film, television and video games with real world violence and violence against women with internet pornography, etc.

So, without launching its a detailed critique of SLT, what needs to be understood here is that as a research field it has a long history of generating findings that are unclear, unreliable, equivocal and contradictory and, amongst its proponents and supporters, and equally long history of these rather inconvenient findings being downplayed, ignored and dismissed, often for largely ideological reasons and particularly when it being used to support social policy interventions designed to curtail access to or censor certain types of media content – and what we have seen in this article is that the evidence for ‘The Werther Effect’ and the manner in which it is promoted by its proponents is entirely consistent with this track record.

It may well have been “shown repeatedly that suicide increases in the month after a front page suicide story” but never reliably or consistently and from Miller & Koo’s study (and Sullivan’s critique) it seems evident that back in 1994, Australia’s disaffected teenagers singularly failed to get the message that suicides involving famous people and gruesome methods – and a shotgun to the head is a pretty gruesome way of checking out – generate a bigger effect than other suicides.

As such, Sullivan’s critique and his overall conclusion, merits careful consideration:

Undoubtedly, there are many effects of suicide on the living, particularly for the family, friends, and associates of those who die by suicide. This chapter has examined an aspect of the effect of suicide on the broader community and concludes that based on careful assessment of the presented evidence, skepticism about the link between publicity about suicides and imitation is warranted — contrary to the conclusions of many studies and the common belief that there is a connection between suicide publicity and imitation. As mentioned earlier, owing to the conclusions reached in studies such as those discussed above, the energy and influence of their authors, and popular concern about the issue, health authorities in several countries have issued guidelines recommending curtailment of suicide reports in the media. The extent to which these guidelines alter reporting practices is a separate issue, but at least in some communities, they appear to have an effect. Therefore, this contribution to the understanding how the dead might influence the living concludes that there is an effect on the community. The effect simply is that of self censorship on the part of journalists and editors at the behest of public health authorities. That there is imitative suicide due to media publicity is yet to be established.

So, all things considered, what are we to make of Ben’s suggestion that the Hyundai ‘suicide’ advert was ‘almost surreally misguided’?

Well, based on the totality of the evidence for The Werther Effect I think the most that anyone can claim is that there could possibly be a risk that such an advert could conceivably trigger off a spate of copycat suicides but that that risk is currently unquantified and may be indeed be extremely small and unreliable. The truth is that we just don’t know enough about this phenomenon to justify such a description, let alone call for adverts of this kind to be censored, not least in that case because effectiveness of the method depicted in the advert is highly likely to prove ineffective, these days, as long as the car used was built after 1995 and is a reasonable state of repair.

That we are in such a position of ignorance is, unfortunately, due in no small measure to the widespread acceptance of The Werther Effect as a truism, amongst researchers, commentators and campaigning and support organisations, such as The Samaritans,  when the validity of this hypothesis remains highly questionable on an unbiased reading of the totality of the evidence. Sadly this is a far from uncommon problem is the social sciences, where researchers’ personal, and often ideological, investment is particular models of human behaviour can too often lead to poor quality, superficial, scholarship which is directed towards generating over-simplistic validations of pet hypotheses, often in an effort justify policy interventions that are, at best, dubious in their merits and, at worst, lacking in any merit at all if not wholly counter-productive.

As HL Mencken sensibly observed, almost a century ago:

Explanations exist; they have existed for all time; there is always a well-known solution to every human problem — neat, plausible, and wrong.