Introduction

In 2016, The University of Chicago publicly announced its opposition to trigger warnings in academic and social settings [1]. The University argued that the presence of trigger warnings allowed individuals to “retreat from ideas and perspectives at odds with their own [2].” This position contributed to ongoing discussion regarding the potential benefits and costs of trigger warnings, and questioned the psychological and intellectual consequences for college students [3–5].

In academia, trigger warnings are statements that warn students, particularly those with clinical mental health issues, about potentially disturbing material [1, 6]. They are common in academic settings and are frequently used to preface content that could be emotionally upsetting. The warnings may be fairly brief, for example, simply stating the type of experience referred to (i.e., “Trigger Warning: Domestic Violence”). Conversely, they can be extensive, alerting the student to potential negative reactions as well as mental health resources in the community. Most frequently, trigger warnings have been used to preface material related to significant traumas such as sexual assault, but they have also been extended to cover a wider range of issues, such as racism and misogyny, that individuals might find upsetting [4].

Ideally, trigger warnings allow individuals who have experienced trauma to be forewarned about material that may elicit unwanted memories of past traumatic events [7]. These memories may prompt strong physiological reactions and unpleasant emotions, responses with enough research support to be codified into the DSM-5 symptoms of posttraumatic stress disorder (PTSD) [8]. By forewarning students about triggering material, instructors allow individuals who have experienced trauma time to prepare or avoid altogether potential emotional distress. For example, a student may 1) choose to sit in the back of the class so they can easily slip out if they become upset, 2) read or watch material at a time or a place that is most advantageous for them or 3) set an appointment with a counselor to prepare or debrief about the material. Wyatt has argued that giving trigger warnings is an empathic act that helps minimize distress for students with trauma-related histories [9].

However, some have hypothesized that trigger warnings have more costs than benefits for students, even for those with relevant traumas [4, 5]. While a trigger warning may allow individuals to benefit from avoiding potentially distressing material, arguably they also preclude coping with their trauma, potentially adding to their psychological distress [7]. In addition, theory related to expectancy effects and self-fulfilling prophecies suggests that trigger warnings may make individuals more likely to respond poorly to upsetting material or lead them to interpret normal, brief, and appropriate distress as pathological [3, 5, 10]. In this scenario, trigger warnings would actually make students worse rather than better.

In addition, it appears that both instructor and student attitudes towards trigger warnings are mixed [11–14]. For example, when asked whether they supported the use of trigger warnings in their training, medical students were evenly split among the options [13]. Psychology students across both public and private institutions showed similar ambivalence, as some appreciated their potential value but were concerned that avoidance of triggering topics by instructors might limit their learning [13, 15]. In the end, the use of trigger warnings by instructors vary considerably and, short of decrees such as that by the University of Chicago, there is no consistent pattern across institutions or instructors regarding their use.

Research could inform this debate, but the research is only emerging. In the first of two studies, Bellet, Jones, and McNally assigned online participants to distressing passages with and without trigger warnings [3]. They found that participants who received the trigger warning believed themselves to be more vulnerable if they were to experience a trauma in the future than those who did not receive a trigger warning. Those individuals in the trigger warning group also reported more anxiety related to the passage, but only if they endorsed the idea that words can cause harm. However, this effect was not replicated by the authors in a subsequent paper [16]. In fact, the authors argue that the most reliable finding was that trigger warnings did not work as intended, in that they may actually cause small increases in anxiety consistent with previous work [17]. In sum, this emerging body of work suggests some small and subtle effects regarding the impact of trigger warnings on emotional responses.

Kimble, Koide, and Flack (2017) reported on a sample of 123 undergraduates, including individuals with trauma histories, who received either a positive, a negative, or a neutral trigger warning before reading a passage that involved physical and sexual assault [18]. All participants reported higher levels of distress (Subjective Units of Distress Scale) [19] right after reading the passage as compared to 2 and 14 days later. This suggests that the effects were fairly brief and that they occurred across all participants regardless of what type of trigger warning was received. The primary significant effect was that those who met criteria for a provisional PTSD diagnosis reported higher distress at all three time points, more likely a reflection of their ongoing PTSD symptoms than a function of their study participation. This finding was corroborated by an extensive and thorough report by Sanson and colleagues [20]. In six separate studies that looked at student reactions to readings and film clips, the authors found that participants responded similarly to the material regardless of whether trigger warnings were presented or not. The authors conclude that trigger warnings have little effect, either positively or negatively.

These studies leave many questions unanswered including the central question inherent to the use of trigger warnings: Does “triggering” material have a negative and lasting effect on individuals with a trauma and PTSD? Do such readings in fact “trigger” these individuals, cause significant distress, and make them think more about their own trauma than they otherwise would? Research has shown that trauma-related stimuli can reliably produce physiological and psychological effects in trauma survivors, but these studies typically use materials that are very evocative and/or idiographic [21]. In the research literature on PTSD, such priming paradigms have not included material that are typically at issue in academic circles.

The answer to whether or not one should give trigger warnings should be based at least in part on how individuals typically respond to such material. While most prior research has investigated the effects of this type of material and trigger warnings on an average student, there has yet to be a study that investigates how students with PTSD symptoms and relevant traumas respond to potentially triggering material. To some extent, this is the population of most interest as they would mostly likely be the ones who would be triggered. In addition, with the exception of the study by Kimble and colleagues, there have been no data on what this looks like over a period of weeks [18]. Psychological effects that are lasting should be given more consideration than effects that are brief.

The term “trigger” has another inherent assumption: that individuals with a trauma related to the presented material will respond more strongly and have an exacerbation of symptoms that is greater than those who experienced a trauma that is not related to the passage. If individuals are indeed having their own traumas triggered by the material, then an individual who has experienced an assault should respond more strongly to a passage about an assault than some other type of traumatic event. However, some have argued that traumatized individuals may be more broadly responsive to threat and, while responding strongest to material related to their trauma, they may still be responsive to a wide range of negative material [22, 23]. This generalized effect could be even greater in those diagnosed with PTSD.

Finally, there have been no data on whether students avoid this material in academic settings, possibly through requests for different assignments or absences from class [4, 5]. There is some suggestion, typically in less formal venues such as blogs and posts, that students will use trigger warnings to excuse themselves from academic responsibilities or retreat to “safe spaces.” While there is ample evidence to suggest that many individuals with PTSD use cognitive and behavioral strategies to avoid reminders of their trauma, there are only sporadic and anecdotal reports that students do this in the context of their courses. Whether students use warnings with any frequency to avoid triggering material is entirely unknown.

In the current study, students from four different universities read a passage that contained a depiction of physical and sexual assault and were tracked in their emotional responses and symptoms over a two-week period. The design of the study allowed us to answer the following questions:

1. Q1: What percentage of students avoid potentially triggering material when they have the opportunity to do so?
2. Q2: Does trauma type affect the amount of distress experienced by the student after reading a triggering passage?
3. Q3: Do individuals with PTSD or a triggering trauma have an exacerbation of PTSD-related symptoms after doing the assignment?

Method

This work was approved by the Institutional Review Boards at Middlebury College, Bucknell University, California Polytechnic State University San Luis Obispo, and Coastal Caroline University. In all cases, the consent form was in written format. All analyses were completed using SPSS 27 (IBM SPSS Statistics for Windows, Version 27).

Participants

Three hundred and fifty-five undergraduate students at four different universities participated in the study. All were taking an Introductory Psychology class and received course credit for participation. Only basic demographics were collected on the participants given that the Introductory Psychology participant pool was quite small at two of the institutions and demographics beyond gender and race/ethnicity could be identifying to the investigators and their research assistants. Sixty-eight (68%) percent were female and thirty-two percent (32%) were male. Primary racial and ethnic identification were as follows: White 69%, Black 10%, Asian 6%, Multi-racial 8%, Hispanic 6%, and Other 2%. Provisional PTSD rates (PCL-5 scores greater than 33) were 12.3% (44/355) and the rate of endorsement of at least one of the 17 traumas having “Happened to me” on the LEC-5 was 87% (308/355). The participating institutions ranged from small private residential liberal arts schools (Middlebury College, enrollment of 2,500; Bucknell University 3,600) to mid-size public universities (California Polytechnic Institute, San Luis Obispo, enrollment of 20,000; Coastal Carolina University, enrollment of 9,000). Table 1 includes demographics at these institutions with respect to Gender, Race, SES, and selectivity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Institutional characteristics^*.

https://doi.org/10.1371/journal.pone.0247579.t001

Results

Q2: Does trauma type affect the amount of distress experienced by students after reading a triggering passage?

RRPQ-R: Condition, trauma, and P-PTSD effects.

Those participants who were assigned and read the triggering passage demonstrated a trend with slightly higher RRPQ-R scores than those assigned to the control passage, t(331) = -1.89, p = .06, Cohen’s d = -.33. RRPQ-R scores did not differ between groups in response to the triggering passage (n = 298) based on whether they had a triggering trauma, other trauma, or no trauma, F(2,295) = 2.14, p = .12, eta squared = .014). Those with P-PTSD had significantly higher RRPQ-R scores (n = 40) to the triggering passage than those with PCL-5 scores below 33 (n = 252). Equal variances using the Levene’s test were not met so the reported t values do not assume equal variances between groups: t(76.52) = -6.38, p < .001, Cohen’s d = -.77.

SUDS scores: Condition, trauma type, and P-PTSD effects.

SUDS scores over time were assessed with a repeated measures ANOVA with Condition as the between-subjects variable (Trigger v. Control) and Time (Day 1, Day 2, Day 14) as the within-subjects variable. Greenhouse-Geisser corrections were used in this mixed model ANOVA when Mauchly’s Test of Sphericity were significant. The trigger condition had an n = 249 and the control condition had an n = 32. SUDS scores differed as a function of Time, F(1.7, 465.5) = 46.50, p < .001, partial eta squared = 0.25 and Condition, F(1,279) = 23.02, p <. 001, partial eta squared = .076, with lower scores occurring later and higher SUDS scores in response to the triggering passage. However, these main effects were modified by a Condition x Time interaction, F(1.67,465.5) = 9.96, p < .001, partial eta squared = .034. The Condition effect differed at Day 1, with those in the triggering condition reporting significantly higher SUDS than those reading the control passage (p < .001), an effect that dissipated by Day 2 (p = .074) and remained non-significant at Day 14 (p = .059).

In response to the triggering passages (n = 248), SUDS scores did not differ as a function of Trauma Type nor was there a Trauma Type x Time interaction, F(3.4,416) = 0.25, p = .88, partial eta squared = .002. SUDS scores did differ as a function of P-PTSD, with those with scores above 33 on the PCL-5 reporting higher SUDS scores at all three time points, F(1,242) = 42.89, p < .001, partial eta squared .15. There was no P-PTSD x Time interaction, however, as both groups’ scores decreased over time, F(1.7,411) = 0.29, p = .71, partial eta squared = .01 (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. SUDS by trauma and time.

Average subjective units distress shown with standard error bars as a function of day and trauma type (Trigger, n = 105; Other, n = 109, None, n = 34).

https://doi.org/10.1371/journal.pone.0247579.g001

Alternate PCL scores: Condition, trauma type, and provisional PTSD effects.

To test condition effects, alternate PCL scores over time were submitted to a repeated measures ANOVA with Condition as the between-subjects variable (Trigger v. Control) and Time (Day 2 v. Day 14) as the within-subjects variable. Analyses were based on an n = 250 after non completers and outliers were removed. There was a statistical trend for Condition, with individuals assigned to the Trigger Condition reporting more symptoms related to the reading across time but not significantly so, F(1,248) = 3.45, p = .064, partial eta squared = .014. Analysis by symptoms cluster (Reexperiencing, Avoidance, Hyperarousal) demonstrated that this trend was driven by a statistically significant effect for the hyperarousal cluster, in which those assigned to the trigger condition reported more hyperarousal symptoms related to the reading than those assigned to the control condition, F(1,250) = 4.51, p = .035, partial eta squared = .018. In no analyses was there a main effect for Time, nor was there a Condition x Time interaction. Using Trauma Type as a between-subjects variable in ALT.PCL scores to the triggering passage (n-217) revealed that those with a trigger trauma reported higher Alt.PCL scores regardless of Time, F(2,214) = 3.35, p = .037, partial eta squared = .03 (see Fig 2). Tukey’s post hoc tests showed that the trigger trauma group had trended toward higher alternate PCL scores compared to the other trauma group (p = .051) and but were not signficantly different from the no trauma group (p = .145). Further analyses of the clusters indicated this pattern was present for primarily due to the hyperarousal cluster. There was no case in which the other trauma and no trauma groups differed from each other. There was no overall main effect for Time nor was there a Trauma Type x Time interaction. To the triggering passage, using Provisional PTSD (n = 213) as a between-subjects factor showed a main effect for P-PTSD, F(1,211) = 56.94, p < .001, partial eta squared = .013, but not a P-PTSD x Time interaction, F(1,211) = 2.82, p = .095, partial eta squared = .013. There was no main effect for Time.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Alt.PCL by trauma and time.

Average alternative PTSD checklist scores shown with standard error bars as a function of the control (n = 32) and triggering conditions (n = 218).

https://doi.org/10.1371/journal.pone.0247579.g002

Q3: Do individuals with PTSD or a triggering trauma have an exacerbation of PTSD related symptoms after doing the assignment?

PCL-5 scores over time: Trauma type and PTSD effects.

PCL scores over time were submitted to a repeated measures ANOVA with Trauma Type as the between-subjects variable and Time (Day 1 and Day 14) as the within-subjects variable. Only individuals who received the triggering passage and who reliably completed the PCL at both time points were included in this analysis (n = 236). There was a main effect for Trauma Type, with those with a triggering trauma reporting higher PCL scores at Time 1 and Time 2 compared to both of the other groups, F(2,233) = 10.93, p < .001, eta squared = .086. Tukey’s post hoc tests indicated that the other trauma group did not differ significantly from the no trauma group, p = .59. There was also a main effect for Time, with scores for all three groups lower at Time 2 than Time 1, F(1,233) = 52.14, p < .001, partial eta squared = .183. There were no significant interactions with time, F(2,233) = 1.59, p = .21, partial eta squared = .13, indicating that all three groups had lower scores on the PCL at Time 2 than Time 1. (Fig 3). P-PTSD participants did not differ over time in their reactions to the triggering passage either; there was no significant P-PTSD x Time interaction, F(1,234) = 3.09, p = .080, partial eta squared = .013. (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. PCL scores at Day 1 and Day 14 as a function of trauma.

Average PTSD checklist scores shown with standard error bars as a function of day and trauma type (Trigger, n = 95; Other, n = 107; None, n = 34).

https://doi.org/10.1371/journal.pone.0247579.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. PCL scores at Day 1 and Day 14 as a function of P-PTSD.

Average PTSD checklist scores shown with standard error bars as a function of day and provisional PTSD (n = 24) and No PTSD (n = 212) groups.

https://doi.org/10.1371/journal.pone.0247579.g004

Discussion

The purpose of this study was to investigate how university students reacted to a passage from literature containing depictions of physical and sexual assault. In particular, the study was designed to investigate how one’s trauma history and current symptoms affect an individual’s responses to such material and how that might change over time.

In this study, the vast majority of participants were willing to read potentially triggering material. Right before reading the passage, all participants who were assigned to the triggering passage were given the offer, made in writing, to read an alternative passage that did not have any content related to physical or sexual assault. At the conclusion of Day 1, participants were asked to indicate, by checking a box, whether they read the assigned or alternate passage. It was clear to the participants that there was no penalty for making this change nor, given the anonymity of the study, was there any way to know if they made the change.

The majority of student participants decided to read the triggering passage even when they were aware of the content. Ninety-six percent of the students who were assigned the triggering passage decided to read it. These percentages were nearly identical for those who had identified themselves as having experienced a triggering trauma (96.9%) or those with a provisional PTSD diagnosis (97.6%). In the qualitative responses, only one participant specifically said they read the alternative passage because “I was afraid that I would be triggered emotionally by the material.” Three indicated that “I preferred not to read the more difficult passage because I expected it to be unpleasant”. One endorsed both of the previous responses. The most frequently endorsed reason participants chose the alternative passage was it “seemed more interesting”. One read the alternative passage because “I could not understand which was the assigned reading”, one read the alternative passage because they “didn’t want to read something depressing or sad”, and two gave no reason for reading the alternative passage.

These data run counter to arguments that suggest that trigger warnings and triggering material will lead students to avoid classes, assignments, or academic responsibilities in numbers that complicate teaching or interfere with instructor choice of assignments. While we could not investigate the question in this study, it is also possible that the de facto trigger warning (via the informed consent procedure) may have increased participation as students felt prepared for the material or were curious about the assigned content. Given the low numbers in this study of students who avoided the material, it would seem that the avoidance of assignments due to possibly triggering material in classrooms would be relatively small particularly due to the pressure some students might feel to complete assignments.

We were surprised to find that there were no differences in avoidance between those with triggering traumas and/or P-PTSD. At a minimum, we expected that individuals with P-PTSD, which include symptoms of behavioral avoidance, would avoid triggering material in a context where there was no penalty for doing so. Perhaps, consistent with research on conformity, they proceeded with the study simply because they were asked to do so. They may have perceived reading the alternative passage as somehow not complying with the request of the experimenters. It is also possible that their general familiarity with the text of a well-known novel may have ameliorated their concern or piqued their interest. It also should be considered that participants may not have avoided the material because it was written as opposed to material that is audio-visual (i.e., movie) or material presented orally in class (i.e., lecture). A participant would know that a written document can be easily avoided if needed, either by skimming the material or avoiding it altogether by turning to the alternative passage. Some may have started the passage with an intention of not finishing if it became too emotionally intense.

It is probable that avoidance of such material would be even lower in real academic situations. Authorized avoidance of an assignment is typically associated with having to talk to an instructor or a dean, a labor-intensive task that may be emotionally challenging. Avoidance of assigned material without accommodation would typically be associated with a penalty that would impact one’s grade, which would further discourage avoidance. For these reasons, it seems reasonable to conclude that avoidance in academic settings is likely to occur less frequently than occurred in this study (~3%) and therefore fairly small in comparison to other reasons students do not go to class or complete assignments.

Therefore, the data suggest that the large majority of students, even those with trigger traumas and P-PTSD, are willing to read triggering material. How students responded to that material on the day they read it however varied as a function of PTSD status but not trauma type. In particular, those with more PTSD symptoms reported higher scores on the RRPQ-R than those without, but having the triggering experience itself was not sufficient to result in higher RRPQ-R scores. These RRPQ scores are largely consistent with the pattern in the SUDS scores. Those with triggering traumas did not necessarily report more distress after reading the triggering passage. In fact, all three groups reported more distress (range 5.1 to 5.6) right after reading the passage then they did two days (range 2.4–2.9) and fourteen days (range 1.9–2.1) later. In addition, the groups did not differ in their reported distress at any time point. All participants, regardless of group status, reported some distress right after reading the passage and then significantly less later. And while this is true of those with triggering traumas, it was also true of the other groups; i.e., the upsetting passage was equally distressing to all of those who read it.

The more robust finding was the significant overall Time effect in SUDS scores. All participants reported being more distressed right after doing the reading than they reported 2 days and 14 days later. (See Fig 1). SUDS scores dropped by the largest amount within the first 48 hours, although the difference in SUDS scores between Days 2 and Days 14 was also significant. Within 48 hours, the SUDS of all three groups were below 3 on an 11-point scale. It is reasonable to surmise that this would be the SUDS among a population of college students at any time point and thus might reflect a return to a value close to “normal”. Future designs should incorporate a SUDS assessment before reading the passage to get a sense of the baseline and/or how much distress is associated with anticipatory anxiety. In addition, the slope of the distress within the first 48 hours would also be informative. For those concerned about student health, distress that lasted well after the session, perhaps into the next day, would be more concerning than distress that dropped almost immediately after leaving the experimental session.

While the trauma type had little differential effect on SUDS scores, PCL-5 scores did. Those with PCL scores above 33 reported significantly more distress throughout the study at all three time points. The most parsimonious explanation is that those with PTSD, regardless of their participation in the study, would regularly report distress as a result of the ongoing PTSD symptoms they are experiencing.

However, by using the alternative PCL, and asking participants specifically about any symptoms that resulted from the reading itself, we were able to get some sense of whether they experienced any ongoing intrusive or hyperarousal symptoms that the participants attributed directly to the reading. On average, participants reported a trend (p = .064) toward more symptoms to the triggering passage as compared to the control passage, an effect that may not have been significant due to the small number of participants assigned to the control condition (n = 32). Those with a trigger trauma, at least compared to those with other traumas, as well as those with provisional PTSD scores reported more symptoms tied to the passage than did those without PTSD or other traumas. This suggests that those with trigger traumas and/or PTSD found themselves thinking about the passage after they had left the session. However, it is important to interpret these findings in light of the magnitude of the effect. For example, those with a trigger trauma reported a mean of 3.54 two days after the reading. This would suggest that they endorsed a few items on the 13-item scale corresponding with “a little bit” or “moderately”. A mean of 3.54 would suggest that the majority of items were endorsed “not at all”. Given that the total score could be as high as 52, this number represents little ongoing rumination about the passage.

However, when scholars, clinicians, and school personnel use the word “trigger,” they are more often referring to the notion that a passage might exacerbate ongoing psychiatric symptoms from an individual’s personal trauma that is closely related to the triggering material. While it seems there may have been a small increase in thinking about the passage for those with a triggering trauma, the data also suggest that the passage did not trigger an exacerbation of symptoms related to their own trauma. This is reflected in little change between Day 1 and Day 14 PCL scores.

The primary effect was one of time. All participants reported fewer PCL symptoms at Day 14 relative to Day 1 regardless of provisional PTSD status or type of trauma. Over the 14 days of this study, participants who read the triggering passage did not get worse. Those with provisional PTSD scores and assaultive traumas did report more PTSD symptoms at Day 14 than those in other groups, but this is likely a function of their ongoing PTSD and not related to the passage they read.

In fact, PCL scores went down significantly for all groups from Day 1 to Day 14. While one possibility is that the participants decreased in their symptoms as a function of their study participation, this is not the most likely explanation and the effect is more likely due to procedural issues. The first PCL was given in the lab, in person, on paper, and under circumstances in which they were aware, via informed consent, that they would be reading a passage from literature that would contain assault. In comparison, the Day 14 assessment was taken online at the convenience of the participant. Given that there were no significant interactions with time and that there was a drop that occurred across all groups, this suggests that the difference may be due to the differential administration format of the instrument across two time points. In future designs, the Day 1 PCL could be taken on a computer in the lab and thus be similar in format to the PCL given at Day 14. It also possible that Day 1 assessments were elevated due to anticipatory anxiety associated with knowing they were to read a triggering passage, fill out a trauma history, as well as complete a psychological instrument that inquires of their psychological well-being.

Acknowledgments

We are particularly grateful for the research support of Lynn Korsun and Kaitlin Morehead at Bucknell University, Spencer Knowlton, Melissa Margolis, and Rachel Semple at California Polytechnic State University, San Luis Obispo and Rebecca Doucet, Anna Kim, and Edith Lopez at Middlebury College.