A few days ago, my Twitter timeline included references to a new initiative aimed at reducing suicide amongst students. One of the projects was described in the media as looking for suicide risk on social media. I think such a project makes very little sense. This post contains quotes from genuine suicide notes.

The project will presumably draw upon two long traditions of research into suicide notes. First, it’s the tradition of assuming that suicide notes give access to the suicidal mind, the other compares genuine suicide notes with simulated ones and in a such a way wants to predict suicide risk. Let me briefly deal with both.

The first tradition was started with the work of Edwin Shneidman who proposed that a person about to commit suicide is in a ‘special state of mind’, characterised by close-mindedness and fixation on the purpose. Indeed, in the first ever study of suicide notes Shneidman and Farberow count ‘thought units’ in the notes, suggesting that language is a mere representation of what a person thinks, their mind. It is, however, Antoon Leenaars, who took research into suicide notes much further. He looked for unbearable pain, interpersonal relations, inability to adjust, indirect expressions and other aspects of notes. Incidentally, Leenaars (and others following him) never identified what ‘indirect expressions’ might be or what unbearable pain might look like in language.

For a linguist such statements about accessing the suicidal mind are extremely problematic. Language is not and cannot be taken as a transparent medium to view anything, and especially the mind of who is speaking or writing. If you take the following three notes, quoted in their entirety:

• I am sorry.
• Forgive.
• It’s my decision. I love you.  

I challenge anyone to reach the mind of the person who wrote them! What emotions can be read off these notes? Where are the pain, pitiful forlornness or other indicators of the suicidal mind postulated by Shneidman, Leenaars and other researchers studying suicide notes? I am not disputing, of course, that the authors might have felt all those things, I am only claiming that the notes offer no insight at all into such states of mind.

I also reject the notion that the use of ‘never’ in suicide notes indicates ‘all-or-nothing thinking’. Even in suicide notes, ‘never’ can be used in emphasis, just as family members say things such as “I’ll never speak to you again.”, happily chatting a few hours later.  We use such general quantifiers for rhetorical purposes all the time and, incidentally, ‘all the time’ in the previous clause is not indicative of my ‘all-or-nothing’ thinking, but, rather, is meant as a device making the point more forcefully.

Can one read the suicidal mind off suicide notes? No.

So, let’s take the other strand of research. There is much research comparing genuine suicide notes with simulated notes. It mostly uses a version of thematic analysis and there is quite a lot of agreement between studies. Let me take one such study. The authors suggest that genuine notes contain more of themes such as ‘planned escape’, ‘negative affect and self-mitigation’, ‘positive affect and failed relationship’, ‘lack of self-acceptance’, while simulated notes contained more of ‘escape’, ‘positive affect and self-blame’, ‘purposeless life’. And I am somewhat disappointed, because the overlap is quite significant, just the wording is different.

One of the problems with such studies is that they are quite simplistic in their approach to language. For example, is the word ‘love’ simply representative of positive emotions? It tends to be assumed as such in studies of farewell letters. And yet, when I was reading my corpus of suicide notes, time and again I thought that love is spoken of as a curse. Something like: “You don’t love me anymore, but I still love you and I can’t live without you. So I must die.” Indeed, people can say/write things like “Unfortunately, I still love you.” quite explicitly positioning their love as something unwanted and, in the context, negative. I am also quite sceptical of the assumption that emotions are simply expressed by ‘emotions words’, just like, say, anxiety only by ‘anxiety words’, a term actually used in one of the articles. I would also say that I doubt that ‘anxiety words’ are used exclusively to express anxiety, much like ‘anger words’ are unlikely to be used only to express anger (see my points about emphasis above).

But a recent Polish study, in which the author focuses on the linguistic form of the letters takes the research to a new level. The author suggests that there are some significant differences between genuine and simulated letters. Without claiming insight into the mind, she shows results of a corpus analysis of a significant number of genuine and simulated letters and it’s hard to challenge it. So, it is very likely that we can, with some probability, recognise whether a suicide note is genuine or not. So, the project to go through students’ social media is, really, a great one. We have the means to recognise the differences. Do we? No, not really.

There are two reasons for it. The first concerns the problem of suicidology in general. Suicidology is very good as recognising all sorts of risks or protective factors in suicide, yet, it is still very far from being able to say anything about an individual.

Similarly, to my knowledge, no one has so far claimed that analyses such as the one by Zasko-Zielinska can lead to identification suicidal intent. It’s even worse, I don’t think there are even claims about operationalising the results of such studies into an intervention. The author is very clear that textual characteristics of suicide notes that would have to  be taken into account are extremely too complex. She lists 15 major characteristics of such letters with a number of dimensions in each of them. It is not clear at all what should be taken as boundary criteria for considering a note genuine.

Moreover, we don’t know why, for example, there are differences in the use of pronouns/grammatical person in genuine/simulated notes. This is important because such text characteristics can have a multitude functions and their mere presence is very far from claims of suicidality.

But, I’m afraid, that’s not all. The crucial bit in all these considerations is that the research I have discussed above was carried out on suicide notes, that is to say, on a particular ‘literary’ genre. The suicide note is a text written by a person in a very particular situation. You cannot translate how people write farewell letters into social media output.

There are two aspects here. The fact that people write suicide notes in a particular way doesn’t mean that what they write reflects suicidality, that is to say, a particular state of mind. Suicide notes are not transparent windows into any mind, indeed, they are more than likely to have very varied and complex communicative goals. Notes such as

• You killed. You killed. You killed. You killed.
• You have got what you wanted.

can be seen as accusations, apportioning blame, or just expressions of anger. Indeed, note authors confessed love, accused, apologised, thanked, sometimes all at the same time in one note! I would also argue that a note such as

• I will haunt you and [name] till you are alive.

should first and foremost be understood as a threat with a goal to scare the addressee! That is the point of what the author of the notes wrote and as a competent user of language, he is likely to have known it. It still cannot and should not be taken simply as transparent insight into what the author thought.

The second point is that even if people write in patterns similar to those of suicide notes, one cannot assume that the communicative goal of such writing is similar. Indeed the accusatory character of “You killed” is more than likely to have a very different communicative intent behind when said in a face-to-face interaction, and different when said in a tweet. To simply assume suicidality behind it is nonsense.

Conclusion? Well, in my view, we’re very unlikely to discover a way in which to read suicide or suicidality off text, just about regardless of what it says. That doesn’t mean, of course, that one should not err on the side of caution when people explicitly talk about suicide, but that’s very far from textual clues to suicide.

Is there a ‘language of suicide’? Apart from the fact that I don’t really know what it might be, I don’t think any such thing exists. Suicide is too contextual, just as are its textual contexts, such as suicide notes, let alone other acts of communication. And while we can study how people talk/write, changing it into an effective preventative intervention is, in my view, unlikely. Alys Cole-King’s safety plan is much more likely to be helpful.

So, why do the project? I’m afraid I have a very cynical view of it. I think it offers an easy way out of the mental wellbeing predicament. As institutions collect data, it will be data collection which will quickly become the measure of caring. You will not need to do much more.