The recent acquisition of Twitter by Elon Musk has led to a centering of several interrelated discourses regarding online speech and social media. Many of these conversations have to do with the intersection of free expression and community safety. Now in part this is because Musk, prior to and during his court-enforced acquisition of Twitter, talked a lot about how his motivation for acquiring the social media platform was to foster freer speech on Twitter. This largely seems to have been taken as a rallying cry for right-wing voices who ran afoul of Twitter’s hate-speech moderation policies to return to the site although, in the chaos following Musk’s acquisition, it appears that many of the reactivated right-wingers were promptly banned again as many of these content moderation policies remained in place after the start of Musk’s chaotic tenure.

However this brings in a second thread. Because the period of Musk’s onboarding has been incredibly chaotic: marked by mass firings and haphazard policy decisions that seem to have been cooked up on the fly by the incoming owner. In all this chaos many users have looked to alternatives and one stood out as appearing, on the surface, close enough to Twitter to attract attention: Mastodon. Now, of course, many of these similarities are skin-deep. There are massive structural differences in that nearly anyone can host a Mastodon instance and nearly anyone can register on a Mastodon instance and communicate with others on that instance.

However the biggest social grouping within Mastodon is the Fediverse: a series of interconnected (federated) Mastodon instances that allow cross-communication and that agree to certain shared content moderation standards. The Fediverse, as with most large social websites, has a highly distinct culture and one element of Fediverse culture is widespread use of content tagging. And when I say widespread I mean it’s considered, within the Fediverse, good form to provide brief descriptive content warnings for a vast panoply of potential media from things you might expect (such as discussions of sexual assault and suicide) to those that might, on the surface, seem more benign (such as food photos).

This is largely an emergent property of a system that was an early adopter of content warning tagging and a culture very interested in users being able to customize their social media experience. The culture of Mastodon does not view content warnings merely as a tool to alert users to material that might cause a trauma response but also simply as a tool to allow users to opt into what sorts of things they see and engage with as opposed to Twitter’s more opt-out system of blocking and muting. Many Twitter users accessed Mastodon only to be confronted with a wall of (TW: Food), (TW: Nudity), (TW: specific-type-of-body), etc. and found this disconcerting to say the least. And many Twitter users have reported back to complain about this cultural difference on Twitter where this discussion of content warnings has found ample attention within the writing community.

Within this community there has been a recent flurry of discussion regarding whether authors have a duty to provide explicit content tagging in their books. The opposing views here on one hand are that this will allow readers to make informed decisions about the sort of material they read, allowing them to avoid books that engage with subjects that are likely to retraumatize them. On the other hand some people, including myself, have been quick to point out that content warnings may be fine as a voluntary provision but should not become an industry standard, citing examples such as film rating systems and parental advisory warnings on music as having had a significant censorious impact that was particularly born by marginalized artists including people of colour and LGBTQ+ artists. A third group largely consists of racists who want to be able to say slurs online but we can disregard them from this conversation because, frankly, neither side of the argument I care about here has anything good to say about those sorts of people. However this discussion ends up at something of an impasse when one assumes both that content warnings serve a purpose for protecting readers who have experienced trauma and that standardization of content tagging will lead to censorship.

As such, in order for this conversation to progress it behooves us to ask a few questions.

1. Do content warnings have a positive impact on the reading experience of traumatized readers?
2. Do content warnings lead to censorship?
3. Do artists have a moral duty to furnish content warnings?
4. Can artists meaningfully create effective content warnings that can serve traumatized readers?
5. Are content warnings really for traumatized readers at all or do they serve some other function?

Now before we dig into these five questions there’s a few questions that will not be entertained. These include whether freedom of speech, as an abstract principle, is more important than the safety and access of any marginalized group and whether it is morally permissible for authors or classroom instructors to voluntarily disclose content that they deem might give their audience problems. Frankly there is nothing wrong with a teacher, upon assigning Lord of the Flies to schoolchildren, telling the schoolchildren that certain material is contained within the book. In fact it is somewhat critical to do so in order that the children can be made alert to this content and how it communicates the themes of the book. In other words disclosure of content is positive in a classroom setting because it allows for anticipation of content and attention to content rather than because it allows for avoidance. A student who is studying a text should anticipate material in it so that they can learn how to identify components of the text that might not be as obvious as the plot.

And as for authors: if an author wants to disclose this sort of material voluntarily, as the academic research we will review later supports quite clearly, this isn’t likely to be overly harmful to anyone and may be helpful to the author in marketing. I think it’s important to set these limits up front because while this discussion will explore some discursively fraught questions it will not at any point be saying that authors must not include content warnings nor that instructors should not disclose information about a fraught text to their class.

But what I will be doing here is ultimately asking the question of what content warnings do and who they serve.

And a good place to start that is in the academic literature on content warnings. Now I will start by referring to a paper from within the class of research called meta-analysis. Meta-analyses are a form of research paper common in medicine and social science: fields of research that are both highly dependent on statistical analysis of data to make conclusions and that also suffer from systemic limitations on sample size and composition. These papers will do a survey of extant research on a topic and will seek out recurring methodologies, themes and limitations. Having established, using these points, that disparate studies are exploring the same topic they will then conduct a statistical analysis of the results of findings across studies to ascertain the replicability of findings. The study I will be working with most here is, “A Meta-Analysis of the Efficacy of Trigger Warnings, Content Warnings, and Content Notes” by Bridgeland, Jones and Bellet.

Now the first thing I want to do is to focus on their discussion of limitations because Bridgeland et. al. raise a very important point here that we will be returning to. “Although the current study provides evidence that trigger warnings are broadly inert as applied writ large, it does not provide information on whether trigger warnings have differing effects in specific subpopulations or contexts.” This is because, due to both reasons of access and ethics, most studies of trigger warnings do not consist of people who have experienced traumatic events nor of people suffering from PTSD. Instead they mostly draw from a general population. This makes the findings of the majority of academic work on the topic inapplicable to our first question. Do content warnings have a positive impact on the reading experience of traumatized readers? Honestly, according to this analysis we don’t really have strong evidence one way or the other.

However what we can say is that content warnings do almost nothing one way or the other to people who don’t have mental health needs surrounding trauma. In fact Bridgeland et. al. found only one measurable category in which content warnings did anything statistically significant at all: “trigger warnings appear to reliably increase anticipatory anxiety about upcoming content. This finding is supported by both subjective (e.g., rating scales) and objective (e.g., psychophysiological measures) markers of distress. Moreover, this finding appears to be consistent across the different trigger warning types used across studies, attesting to the robustness of this effect.

In theory, this anticipatory period could indicate that forewarned individuals are bracing themselves for a negative emotional experience. However, as discussed in the section on response affect, whatever bracing might occur during this anticipatory period is apparently completely ineffective.”

Of particular note here are the psychophysiological measures used for anticipatory anxiety: heart rate, respiratory rate and skin conductance. These measures were used in only one study and represent the most statistically significant variance from the general support of the null hypothesis found in almost all studies.

But those of us who engage with horror media know about this sort of anticipatory anxiety all too well. It pretty closely maps to the tension one might feel during a stalking scene or some other moment of peril before a horrific event occurs. It’s something horror artists actively court, building mood in order to entice the audience to become anxious and it’s an affect the audience of horror generally seeks out. People who don’t like being scared don’t generally like horror.

But what this does, when read in light of Bridgeland et. al. comments on the limitations of these studies, is point to the fact that we cannot do a straightforward read of, ‘content warnings increase anxiety and therefore are bad,’ because what they do is give non-traumatized audiences a taste of the forbidden pleasures to come. Far from being a tool for avoidance, “cw:incest” allows a reader to anticipate that they will be reading a book that contains incest and it’s worth noting that this anxiety could actually sweeten the reading experience for them in much the same way that a horror fan enjoys a kill better when it’s been built up properly (I’m thinking of the perfectly executed build-up to the kill of Wes Hicks in Scream (2022) as an example.)

So this now points us not toward an answer to our first question but rather to our fifth. Are content warnings really for traumatized readers? Certainly they aren’t just for traumatized readers although the advocates of them rarely seem interested in pitching them as a tool for marketing and discovery it does appear that, when not engaging with a traumatized audience but rather with a general audience, that this is just about the only thing content warnings do at all.

But if we want to look at the impact on traumatized readers in particular we have to turn our attention to a different study. “Helping or Harming? The Effect of Trigger Warnings on Individuals with Trauma Histories,” by Jones, Bellet and McNally is one of a very small number of studies that actually looks at the effect of content warnings specifically on traumatized populations. Now again we should start with limitations here as it’s very important, when working with academic research, to be alert to the scope of the research. In this case the principal limitation is a methodological weakness of depending on self-report for all participants. While steps were taken to ensure subjects had experienced trauma there does not appear to be much the authors could have done to prevent a person from providing misleading information regarding their past experiences. However a clinical study, which might have validated the trauma experience of subjects more cleanly, would certainly not have been able to achieve the sample size of n=600 that this paper managed.

Further limitations, however, included a dependence on English fluency and a requirement for US residency that should not be overlooked as challenges to replicability. However, at the end of the day, the paper came to conclusions that very closely mirrored Bridgeland et. al. saying, “For individuals who met a clinical cutoff for severity of PTSD symptoms, trigger warnings slightly increased anxiety. Trigger warnings were not helpful for individuals who self-reported a diagnosis of PTSD. Perhaps most convincingly, trigger warnings were not helpful even when they warned about content that closely matched survivors’ traumas. That is, when considering only the passages which participants reported as reminding them of past trauma, trigger warnings were still unhelpful.”

Perhaps more damning still was the confirmation of the information Bridgeland et. al. had found regarding anticipatory anxiety, saying, “We found evidence that trigger warnings increase the narrative centrality of trauma among survivors, which is countertherapeutic (Boals & Murrell, 2016). We also found that trigger warnings increase anxiety for those with more severe symptoms of PTSD. Although these effects were preregistered and found in a large sample, the size of the effects were small and have not yet been rigorously tested across multiple studies.” Now the authors are quite right to point out that this impact was small and that replication is required so I think it would be hasty to say that content warnings are actively harmful to people who meet the clinical cutoff for PTSD (among people who do not meet that requirement the null hypothesis was observed).

And so we can now answer the question of whether content warnings have a positive impact on the reading experience of traumatized people – they don’t. According to the best research available, for most people who have undergone trauma, content warnings do nothing much at all. For those people whose suffering is particularly severe there is even a small risk a content warning might harm more than it helps. We can also say with certainty that authors cannot meaningfully create content warnings that will serve a traumatized audience because the reality is that, to the extent content warnings serve anyone at all, they serve non-traumatized audiences and authors best via their use as a discovery tool and for the deliberate assumption of anxious affects in the course of engaging with material that touches on taboo subjects.

But all this talk of null hypothesis means this does become a question of personal taste and courtesy. You can be assured that you won’t much hurt a person by omitting a content warning but courtesy certainly goes beyond avoidance of explicit harm. It doesn’t harm someone to eschew “please” and “thank you” but if you never use either phrase people still might rightly call you an asshole.

This is where our second and third questions are still relevant. Having established that content warnings have nearly no clinical impact on traumatized populations our next question is whether content warnings might in fact be harmful to marginalized audiences. For this we might want to visit the advent of the parental advisory label on music.

The response from the record industry has been that it, much like content warnings, had very little impact at all. It was useful as a marketing tool both for albums with it and albums without. For example, manager Danny Goldberg pointed out that while the sticker did allow stores like WalMart to brand themselves as “family-friendly” by declining to sell any album with a PA label most children had very little difficulty acquiring material that was marked as parental advisory. Overall the stickers didn’t much effect record sales one way or the other.

This was less the case in film where the Hays Code and subsequent film and television ratings systems inordinately targeted queer narratives, largely driving LGBTQ+ artists and themes out of cinema. Now books are not movies nor are they records. But in both cases we see how ratings and advisory systems have been deployed as a method of exclusion. Unsuccessfully in the case of music, where the exclusion of “PA” content by WalMart was countered by the willingness of record stores to sell “PA” material and successfully in the case of cinema where the ratings system created a series of economic incentives for self-censorship. As such those people who have concerns about systematic industry standards in content reporting leading to censorship have a point. While the attempts to censor material based on industry-set content warnings has been haphazard and has certainly not been universally successful the use of industry standard content warnings has, in other artwork, been used to censor that art.

This then finally establishes the full framework for commenting on an author’s moral responsibility. Content warnings are useless for traumatized people to manage trauma reactions but useful as a discovery tool among general populations. They may lead to censorship but it’s not clear the extent to which such censorship would be effective. Certainly there has been a lot of proactive attention from reactionaries on book censorship of late and giving these reactionaries extra tools with which to discover books to ban would be counter-indicated. But the advocates of content warnings like to point out a key outlier circumstance as a justification for the moral argument: the deployment of sexual assault in fiction. Now they’ve certainly got a point that if some inconsiderate person runs around shouting “rape” we would consider that rude, immoral, and alarming behaviour. And we cannot just handwave away every deployment of sexual assault in fiction as being beyond reproach. There are boundless examples of rapes that were included to titillate and, even among those works with something critical to say about sexual assault, there’s no guarantee that these themes will be approached well by the author. If I had a nickel for every book or movie that tried to deconstruct some concept only to reify it out of incompetence I’d be a wealthy man.

Giving people a heads-up about this might, then, be a gesture of common courtesy. And that would suggest it is, at least, the polite thing to do to provide content warnings for this (along with other broadly questionable content). Except we need to recontextualize this call for courtesy in light of some of the academic findings about content warnings. Specifically: there is no indication that people who read a content warning are any more likely to avoid that content but again we need to bring up that increase in anticipatory anxiety. We cannot count on people to use this courtesy to decide to read something else. Academic research suggests they probably won’t – at least in aggregate – but we can count on people to get excited by it. Their heart rates rise, they breathe faster. This then raises an opposite question: is an author morally responsible to tell people that their kinks are in a book? I mean it’s a bit of a silly question, isn’t it, because if I’m an author writing kink I want an audience who want to read kink to find it. But is that a moral imperative or is it just good marketing? Any regular reader of this blog will be aware that I’m quite apprehensive about assigning moral imperatives to art. The aestheticization of morality is a dangerous tool and a favourite one of reactionaries. People generally have a hard time separating out, “this is beautiful” from “this is good.” It’s quite easy to look at the deployment of something in a work of art, such as a sex assault, and say, “wow that was ugly and no good.” But ugly is not the same as evil and I think it’s important for critics and artists, especially, to learn to differentiate between an ugly work of poorly executed art (like a Dresden Files novel) from an evil work of propaganda (like the Turner Diaries).

Bring libido into the picture and it becomes even more of a landmine. Because once you go from “this is beautiful” to “this excites me” moral questions immediately become far more tangled. After all, who doesn’t likely feel shame about their own arousal in some form? This is especially so if we’re dealing with darker erotic themes. While I can certainly understand that some readers might prefer not to interrogate that it seems unlikely that content warnings will actually help them with that.

In the end I think part of the problem is one of form. The advocates for content warnings want short, broad, concise tags at the front of a book. This is excellent for marketing and discovery because it’s very algorithm friendly. I would propose a more graceful method is to actually use back-flap space to describe what a book is about rather than reserving it for blurbs. I don’t think anybody benefits from reading The Story of the Eye without understanding what they’re in for ahead of time. But I think that a clinical, “CW: Masturbation, Rape, Dubious Consent, Violent Orgies, Necrophilia,” doesn’t particularly do justice to the affect being pursued. Context matters and should be communicated in such a discussion. But an exploration of the context being one about Bataille grappling with the interplay of sex and death in European thought, expanding upon the legacy of Sade (of whom he was a principal scholar) and raising questions of limit experiences and madness isn’t going to give you that easy-to-search list of tags they prefer.

But what it does is improve discoverability. Ultimately there’s no sincere debate about whether it should be possible to find out what a book is about. The question is one of form: should this look like a voluntary and often community-driven process of resource sharing or should it be a system of brief and concise tags an author is supposed to put on their book.

Certainly tags share a few benefits: They’re good for marketers and for censors for precisely the same reason – they make it easy to find content you’re looking for. But this is where the benefit ends. The preference for context-driven back-cover notes and third-party disclosure (such as instructors discussing challenging material prior to reading in class and community driven efforts to surface potentially upsetting content) is less friendly to marketing efforts and raises the risk that things might be missed but it does provide the necessary context to identify the difference between a work that deploys charged material for libidinal reasons from those that do the same for critical reasons – something tagging systems are necessarily mute on.

And so, here at the end, we have our answers:

1. Content warnings have nearly no impact on traumatized readers.
2. Content warnings may or may not lead to censorship but do lead to increased discoverability which is a useful tool both for marketers and censors.
3. Artists do not have a moral duty to make work more discoverable. But if they want to find an audience to sell their work to it’s probably not a bad idea. Furthermore it’s good from a critical readership perspective that context concerning the content of a work be known prior to reading so that a reader may be alert to it.
4. Because content warnings do not meaningfully serve traumatized readers artists cannot meaningfully create content warnings that serve traumatized readers.
5. Content warnings appear to increase anticipatory anxiety but not avoidance among non-traumatized populations, making them effective for getting a person excited for upcoming material. This, combined with the concise and tag-like method preferred by advocates make them excellent marketing and discovery tools.

So in the end who do content warnings really serve? The people who want to read that content and the writers who want to sell it.