Psychophysiological reactions during the trauma-film paradigm and their predictive value for intrusions

ABSTRACT

Background: Adequate adaptation of the autonomic nervous system (ANS) is crucial in potentially life-threatening situations. The defence cascade provides a descriptive model of progressing dominant physiological reactions in such situations, including cardiovascular parameters and body mobility. The empirical evidence for this model is scarce, and the influence of physiological reactions in this model for predicting trauma-induced intrusions is unresolved.

Objectives: Using a trauma-film paradigm, we aimed to test physiological reactions to a highly stressful film as an analogue to a traumatic event along the defence cascade model. We also aimed to examine the predictive power of physiological activity for subsequent intrusive symptoms.

Method: Forty-seven healthy female participants watched a stressful and a neutral film in randomized order. Heart rate (HR), heart rate variability (HRV), and body sway were measured. Participants tracked frequency, distress, and quality of subsequent intrusions in a diary for 7 consecutive days.

Results: For the stressful film, we observed an initial decrease in HR, followed by an increase, before the HR stabilized at a high level, which was not found during the neutral film. No differences in HRV were observed between the two films. Body sway and trembling frequency were heightened during the stressful film. Neither HR nor HRV predicted subsequent intrusions, whereas perceived distress during the stressful film did.

Conclusions: Our results suggest that the physiological trauma-analogue response is characterized by an orientation response and subsequent hyperarousal, reaching a high physiological plateau. In contrast to the assumptions of the defence cascade model, the hyperarousal was not followed by downregulation. Potential explanations are discussed. For trauma-associated intrusions in the subsequent week, psychological distress during the film seems to be more important than physiological distress. Understanding the interaction between physiological and psychological responses during threat informs the study of ANS imbalances in mental disorders such as post-traumatic stress disorder.

HIGHLIGHTS

• We used a trauma-film analogue to examine the defence cascade model and to investigate the influence of psychophysiological response on subsequent intrusions.

• While we found an orientation phase, upregulation phase, and coactivation indicated by heart rate, no downregulation phase was observed.

• None of the physiological parameters examined predicted subsequent intrusions, whereas subjective distress during the film did.

Antecedentes: La adaptación adecuada del SNA desempeña un papel crucial en situaciones potencialmente mortales. La cascada de defensa proporciona un modelo descriptivo de las reacciones fisiológicas progresivas y dominantes en tales situaciones, incluidos los parámetros cardiovasculares y la movilidad corporal. La base empírica de este modelo es escasa, y del mismo modo, la influencia de las reacciones fisiológicas dentro de este modelo para predecir intrusiones inducidas por traumas no está resuelta.

Objetivos: Utilizando un paradigma de una película de trauma, nuestro objetivo fue probar la reacción fisiológica a una película altamente estresante, que funciona como análogo a un evento traumático a lo largo del modelo de cascada de defensa. En segundo lugar, nuestro objetivo fue examinar el poder predictivo de la actividad fisiológica para intrusiones posteriores.

Método: 47 participantes femeninas sanas vieron una película estresante y otra neutral en orden aleatorio. Se realizo medición de la frecuencia cardiaca (FC), variabilidad de la frecuencia cardiaca (VFC) y el balanceo corporal. Las participaron registraron la frecuencia, la angustia y la calidad de las intrusiones posteriores en un diario durante 7 días consecutivos.

Resultados: Para la película estresante, observamos una disminución inicial de la FC, seguida de un aumento, antes de que la FC se estabilizara en un nivel alto, lo cual no se encontró durante la película neutra. Para la VFC, no observamos diferencias entre ambas películas. El balanceo del cuerpo y la frecuencia de los temblores aumentaron durante la película estresante. Ni la FC ni VFC, pero sí el malestar percibido durante la película estresante, predijeron intrusiones posteriores.

Conclusión: Nuestros resultados sugieren que la respuesta fisiológica análoga al trauma se caracteriza por una respuesta de orientación y una hiperactivación posterior que alcanza un alto nivel fisiológico. En contraste con los supuestos del modelo de la cascada de defensa, la hiperactivación no fue seguida por una regulación a la baja. Se discuten potenciales explicaciones. Para las intrusiones asociadas al trauma en la semana siguiente, el malestar psicológico durante la película parece ser mas importante que el malestar fisiológico. Comprender la interacción entre las respuestas fisiológicas y psicológicas durante una amenaza informa sobre los desequilibrios del SNA en los trastornos mentales como el trastorno de estrés postraumático.

KEYWORDS:

• PTSD
• defence cascade
• fight
• flight
• freeze
• autonomic nervous system
• trauma analogue
• heart rate variability (HRV)

PALABRAS CLAVE:

• TEPT
• cascada de defensa
• lucha
• huida
• congelarse
• sistema nervioso autónomo
• análogo de trauma
• variabilidad de la frecuencia cardiaca (VFC)

1. Introduction

While the majority of people (70–90%) report having experienced at least one traumatic event in their life (Benjet et al., 2016; Kilpatrick et al., 2013), the lifetime prevalence of post-traumatic stress disorder (PTSD) is less than 10% (Kilpatrick et al., 2013). Not all, by far, traumatic experiences provoke post-traumatic distress. Events evoking such distress are more likely to be distinguished by certain event characteristics, such as sexual violence, or by recurrence (Kessler et al., 2017; Kilpatrick et al., 2013; Maercker et al., 2018; Santiago et al., 2013). Some individuals are more prone to post-traumatic distress than others, and higher vulnerability relates to, among other things, enhanced intensity of the psychological and physiological response during the traumatic event (Ozer et al., 2003; Yehuda, 2004).

The peritraumatic interaction of the psychological and psychological response is theorized in the ‘defence cascade’ model (Lang et al., 1998; Schauer & Elbert, 2010). This model proposes a temporally fine-tuned interplay of cognitive and emotional factors with the autonomic nervous system (ANS) in peritraumatic responses. The ANS is composed of the upregulating sympathetic nervous system (SNS) and the downregulating parasympathetic nervous system (PNS), where the latter usually acts as a brake for the sympathetic tone (Zoladz & Diamond, 2013). However, to clarify a misleading assumption, SNS and PNS can interact not only antagonistically, but also synergistically or independently (He, 2020).

According to Schauer and Elbert (2010), the defence response consists of six progressive reactions, depending on defence possibilities and on the proximity of danger. The first stage is an initial ‘orientation phase’, where threat is distal and could be avoided. With increasing proximity of the threat, ‘flight’ and ‘fight’ start as active defence responses. These are characterized by upregulating SNS until an arousal peak is reached. The following reaction, ‘fright’, can be classified as a passive defence response (Terpou et al., 2019), which marks a turning point from rising hyperaroused activity to hypoarousal. Subsequently, the PNS dominates, leading to ‘flag’, which can culminate in a final shutdown response, where afferent sensation and perception, and efferent motor control, are impaired, which ends in ‘fainting’. This model has been modified and extended over the past decade, with variation in the names and descriptions of the phases (Kozlowska et al., 2015; McKinnon et al., 2016). For our purposes, it may be intuitive to call fight and flight ‘upregulation’ and to call fright ‘downregulation’. While the original defence cascade model assumed a termination of downregulation with fainting, an alternative outcome has been hypothesized, which may be called ‘coactivation’. This coactivation describes that fight or flight is put on hold until an opportunity arises to escape the situation (Kozlowska et al., 2015). Thus, along with the SNS activation, activation of the opposing PNS increases, and the dominance of SNS or PNS activation may vary over time (Roelofs, 2017). The incoming PNS activation can cause a sudden drop in heart rate (HR) or a reduced HR acceleration (Kozlowska et al., 2015; Roelofs, 2017). Some support for the coactivation hypothesis comes from animal studies, indicating a switch between upregulation and downregulation activation, depending on the threat situation, which is accompanied by a continuous change in physiological parameters (Hagenaars et al., 2014a; Kozlowska et al., 2015). The defence cascade model proposes that not all phases need to be passed through. For instance, if fight or flight results in a successful escape from the threat, the transition into fainting is no longer necessary. However, if there is no opportunity for escape, the individual progresses to downregulation. This is caused by the overwhelming input and is characterized by reduced sensory responsiveness, dissociative reactions, uncontrollable shivering, and/or simultaneous tonic immobility (Terpou et al., 2019). Survivors describe the latter as an inability to move their body, despite having an intense urge to flee. These characteristics of the downregulation phase are especially reported by survivors of sexual assault (deMello et al., 2023; Gbahabo & Duma, 2021; TeBockhorst et al., 2015). The defence cascade model predicts post-traumatic reactions induced by certain trauma-associated triggers from responses during a life-threatening event (Schauer & Elbert, 2010), and hypothesizes that peritraumatic stages of the cascade are re-enacted with their physiologically and psychologically responses.

For ethical reasons, the assumptions of the defence cascade model cannot be experimentally tested in real life. Therefore, prospective studies focus on times immediately after the traumatic experience or before a potential threat, e.g. during military service. Such studies have found that higher HR directly after the traumatic experience and lower heart rate variability (HRV) before the event predict the development of post-traumatic symptoms (Minassian et al., 2015; Morris et al., 2016; Pyne et al., 2016; Shalev et al., 1998). Post-traumatic stress is often tested with retrospective studies, which examine the autonomic reactivity – typically assessed by HR and HRV – in affected patients. These studies revealed increased HR and decreased HRV under resting, symptom, and stress provocation conditions (Ge et al., 2020; Nagpal et al., 2013; Pole, 2007; Schneider & Schwerdtfeger, 2020). These results are indicative of SNS hyperreactivity, facilitated by increased noradrenaline release (Pitman et al., 2012; Zoladz & Diamond, 2013) and PNS hypoactivity. Moreover, post-traumatic core symptoms relate to impaired allostatic recovery (Pole, 2007). Norte et al. (2013) confronted trauma-exposed people with and without PTSD with their personal trauma scripts. Although both groups showed increased HR during exposure, only the PTSD group failed to recover to baseline levels and only this group showed decreased HRV during and after trauma exposure. It is assumed that the prolonged interoceptive SNS signals danger to the organism and leads to a perceived lack of safety (Porges, 2022).

An approach to studying peritraumatic physiological and psychological factors during stress-inducing threat is provided in the trauma-film paradigm. This method can be used to study presituational and perisituational factors in analogy to traumatic events, by having participants watch very stressful, trauma-associated video films in a well-controlled situation (Holmes & Bourne, 2008; James et al., 2016; Lau-Zhu et al., 2018). Similarly to real traumatic experiences, the stressful film evokes intrusive thoughts that relate to the content of the seen material. Both the frequency of intrusive thoughts and the degree of distress of intrusive experiences vary as a consequence of the trauma-film paradigm and emerge in different quality types, such as sounds or pictures (Holz et al., 2016).

Studies using trauma-associated films to investigate cardiovascular reactions during the stressful film in relation to subsequent post-traumatic stress-related symptoms yield inconclusive results. Holmes et al. (2004) found that a drop in HR during watching live footage of car accidents was associated with later intrusions. Furthermore, they found that the mean HR during the specific scenes that led to intrusions was lower than during scenes which were non-intrusive later. Chou et al. (2014) replicated these content-related findings but found no significant correlation between intrusions and overall HR changes. The authors explain this by a design issue due to the added startle probes. In contrast to this, Weidmann et al. (2009) found that higher, but not lower, HR predicted more intrusions. This study also compared films with different content, and found that the depiction of a rape scene elicited higher HR and more intrusive memories. The authors state that films with a straight storyline may result in a higher degree of involvement and produce more intense reactions than the live footage of car accidents, mostly consisting of unconnected scenes and showing the aftermath of a traumatic event and not the event itself. Although different film themes may have different impacts on psychological and physiological outcomes (Arnaudova & Hagenaars, 2017), this seems not to be the only reason for diverging results. Using the same rape scene, Holz et al. (2016) found significant increases in HR from prefilm to perifilm, but neither HR during the film nor HR at baseline correlated with subsequent intrusion frequency or distress. Furthermore, Hagenaars et al. (2014b) used a stabilometric platform to assess body sway, and identified a freezing-like response during an unpleasant film depicting the aftermath of a car accident, indicated by decreases in body sway and HR deceleration. Moreover, spontaneous tonic immobility in a picture-viewing paradigm was positively correlated with an increased frequency of intrusions (Kuiling et al., 2019).

Taken together, there is evidence that HR during watching stressful films predicts later intrusions in the trauma-film paradigm, but results vary. This is not sufficiently explained by sample or film characteristics. Most studies examined healthy participants, who were predominantly students and females, while half of the studies excluded participants with previous film-related traumatic experiences (Chou et al., 2014; Holz et al., 2016; Weidmann et al., 2009). While a few studies worked with picture footage, most studies used trauma-analogue films, which seem better suited to elicit absorptive effects. (For a general discussion of dissociative absorption by movies, see Butler & Palesh, 2004.)

Only a few studies have examined the phases of the defence cascade in time-related progress. Using a virtual reality (VR) horror scene, Nackley and Friedman (2021) detected an initial orientation response, and later a ‘flag’ interpreted downregulation, but no upregulation. The authors argue that the VR experience, with, for example, flickering lights, was not close enough to a real threat. While this may be the case, the elicited downregulation, rather, indicates a successful simulation of an inescapable threat, but not of a threatful situation evoking fight or flight. Hagenaars et al. (2014b) used the trauma-film paradigm with film footage of the aftermath of a car accident. The authors found a downregulation in HR over the time-course of 60 s of unpleasant film footage compared to the pleasant film scene. They acknowledge that this study is a first step in the exploration of defence-related psychophysiological reactions and that further research with a prolonged time-course is needed. This appears reasonable, as the rather short film sequences may only evoke an orientation response.

With our study, we aim to fill the gap outlined by the two study approaches of Nackley and Friedman (2021), who used a novel VR approach, but may not have implemented the natural course of a threatful situation within their horror scenario, and Hagenaars et al. (2014b), who used a time-relevant approach with film footage, but failed to embed aspects such as proximity of the threat or perceived ability to escape into their film material, which could have enabled the participants to pass through the consecutive stages of the defence cascade model.

First, we aim to empirically test the physiological reaction to a stressful film, which works as an analogue to a traumatic event. Following the defence cascade model, we expect an initial HR increase and HRV reduction mirroring fight and flight (upregulation), followed by HR decline and HRV increase during the passive defence response (downregulation). Thereafter, we expect a pattern of SNS and PNS coactivation, which reflects a presumed switching between upregulation and downregulation, characterized by high HR and high HRV (coactivation). Differently from the original defence cascade model, we expect no fainting. Besides these cardiac effects, the whole-body movement is reduced in the face of a traumatic or potentially threatening picture (Azevedo et al., 2005; Roelofs et al., 2010) and during stressful films (for a reviews see Hagenaars et al., 2014a; Hagenaars et al., 2014b). Thus, we expect reduced body sway and higher trembling frequencies during the stressful film. Secondly, we aim to examine the predictive power of physiological activity for intrusions. We hypothesize that HR change and HRV change during the stressful film predict the frequency of intrusions, distress, and variety of quality.¹ Furthermore, we explore whether HR and HRV during the different phases differ in the predictive power of intrusion.

2. Methods

The study presented here is part of a larger examination, aiming to investigate psychophysiological markers and peritraumatic dissociation. The whole project is preregistered under https://drks.se/search/de/trial/DRKS00027822.

2.1. Participants

We enrolled healthy female participants between the ages of 18 and 65 years. We excluded women who (1) had current clinically relevant mental symptoms, assessed with the short version of the Diagnostic Inventory of Mental Disorders (Mini-DIPS) (Margraf et al., 2017; Margraf & Cwik, 2017); (2) had ever received treatment for mental disorders; (3) had sustained a severe head injury; (4) had a neurological disorder; (5) had used long-term medication, other than contraceptives, within the past 4 weeks; and (6) had a history of sexual violence, with associated thoughts and memories, that impaired daily life, on a self-reporting basis. To increase the internal validity of this study, we recruited only female participants and tailored the trauma-film paradigm to women. This was done for two reasons. First, women and men differ in their cardiophysiological differences in baseline values (Koenig & Thayer, 2016) and in their cardiac reactions to mental stress (Adjei et al., 2018). Secondly, there is evidence that men and women tend to react differently to stress on a psychological and biological level (Seligowski et al., 2021; Verma et al., 2011).

The power calculation was based on our first main aim and was calculated using the G*Power software (Faul et al., 2009). To statistically ensure a significant slope of cardiovascular parameters during eight phases/conditions (see Section 2.4) with medium effect size (f = 0.25), the optimal sample size was calculated to be n = 34 participants [repeated measures analysis of variance (ANOVA), eight measurements, one group, α = .05; power = .9, correlations among repeated measurements = 0.1, non-sphericity correction = 1].²

The ethics committee of the University of Dresden approved the study (SR-EK-407092020), which was conducted in accordance with the Declaration of Helsinki (World Medical Association, 2013). During the recruitment process and via study information, participants were informed about the aim, procedure, and distressing nature of the film. Participants were included after providing informed written consent. They received an expenses allowance of €50 for full participation or €10 after the first appointment, if they matched the exclusion criteria or withdrew from the experiment, which happened once each.

2.2. Procedure

Participants were recruited via a local online study platforms and public notice boards. Preceding an in-house appointment, interested participants completed an anonymous online screening assessing the above-mentioned exclusion criteria. People passing the inclusion and exclusion criteria received further instructions for the first appointment. They were instructed to not take non-prescription drugs for at least 6 h and not to smoke, eat, exercise, or consume caffeine for at least 2 h prior to the experiment.

For the first laboratory appointment (Figure 1), participants answered the Mini-DIPS (Margraf et al., 2017; Margraf & Cwik, 2017) to exclude participants with mental health issues. As the study is part of a larger investigation, further questionnaires were filled in, e.g. the PTSD Checklist for DSM-5 (PCL-5) (German version) (Krüger-Gottschalk et al., 2017; Weathers et al., 2013b), including the Life Events Checklist for DSM-5 (LEC-5) (Weathers et al., 2013a). Thereafter, we started the experimental phase using a repeated measurements design. Each participant watched a stressful and a neutral film, presented in randomized order, while standing on a stabilometric platform with approximately 30 cm distance between the feet, arms hanging relaxed, at a distance of approximately 1.8 m from a 1 m × 1.8 m screen, in a dark room. Movie sound was played over headphones. To ensure that participants kept engaged and did not close their eyes or turn off the headphones, a research assistant sat by their side but not within the visual field of the participant.

Figure 1. Schematic visualization of the study procedure.

First, there was a 5 min adaptation phase for HR normalization while participants were standing without any further instruction. Subsequently, a fixation cross was presented for the 2 min baseline assessment, using E-Prime 3.0 software (Psychology Software Tools, Pittsburgh, PA, USA) (Psychology Software Tools, 2016). Immediately afterwards, either the neutral or the stressful film started. The film presentation was followed by a 30 min interval, during which participants sat and filled in additional questionnaires assessing dissociation, stress, and immobility, as part of the broader investigation. Then, the adaptation and baseline phases were repeated, followed by the second film (Figure 1).

Subsequently, participants were instructed to fill out a paper-based intrusion diary for the 7 days following the experiment (Supplemental Figure 1). After 1 week, participants were invited back, when they handed in their diaries and were debriefed by the psychological staff.

2.3. Materials and measures

2.3.1. Postinterventional intrusions

Participants were instructed to note each intrusion directly after it occurred in the diary, and to rate the intrusive distress (11-point scale from 0 ‘not at all’ to 10 ‘extreme’). We also asked them to categorize intrusive re-experiencing into four different qualities: (1) recurrent, sudden, spontaneous, and involuntary memories of the film scene; (2) spontaneous re-experiencing, e.g. through vivid images; (3) other sensory perceptions, e.g. sounds or bodily sensations; and (4) nightmares.

2.3.2. Film scenes

For the stressful film, we chose a scene (12 min, from 0:41:51 to 0:53:51) from the movie Irréversible by Noé (2002). This movie depicts a rape scene and was used in previous research, where it reliably induced intrusions (Holz et al., 2016; Rombold-Bruehl et al., 2019; Schultebraucks et al., 2019; Weidmann et al., 2009). For our analysis of the defence cascade response, we defined four segments in chronological order and, hence, orientated us according to the film’s content and its dramaturgic curve. Each of the first three scenes lasts for 45 s, beginning with the ‘entrance scene’, the subsequent ‘incipient threat scene’, which we related to the upregulation phase, followed by the ‘trapped scene’, which we related to the downregulation phase, and the ‘assault scene’, which contains the last 9 min and 45 s and which we related to the coactivation phase (Figure 1). Note that for the entrance scene, we had no special expectations according to HR or HRV, because it does not cover a phase of the defence cascade and was assumed to ease the participants into the stressful film. A neutral film from a different video depicting a female instructor during strength training with a male client [(Fe)MaleASMR, 2018] was shown as the control condition. This scene was balanced in length (12 min), actors (one male, one female), and the camera cut as a one-shot scene. The authors gave permission to use this video for study purposes. Owing to its low emotionally affecting content, we did not expect a relevant change in HR or HRV during the neutral film. Hence, we refrained from a content-based division of the neutral film. For comparison with the stressful film scenes, we chose the same time-related segments for the neutral film. In addition, we controlled distress induction using subjective ratings (‘How much stress did you feel?’, rated from 0 ‘not at all’ to 10 ‘extreme’) after each film.

2.3.3. Assessment of heart rate

We recorded the electrocardiogram (ECG) with the SUEmpathy^® System (SUESS Medizin-Technik, Aue, Germany) with one channel measuring HR and HRV, using electrodes (Ambu^® WhiteSensor, disposable 42 mm surface) that were positioned on the left costal arch and on the left and right subclavicular chest. The ECG sensor recorded at 512 Hz with 12-bit resolution. Using the SUEmpathy Scientific software (SUESS Medizin-Technik, 2016), the ECG raw signal was transformed into interbeat intervals (IBIs), measured in milliseconds. Missed or incorrectly detected R-peaks, which serve as markers for IBI detection, were manually corrected in the SUEmpathy software. The ECG signal was analysed in ARTiiFACT software (Kaufmann et al., 2011), with an artefact detection algorithm developed by Berntson et al. (1990), and corrected by cubic spine interpolation (see Supplemental Table 1a and b for an overview of all ECG data corrections). We calculated beats per minute (bpm) from the IBIs, as bpm is more intuitive to understand. We used the root mean square of successive differences (RMSSD) as a recommended measure for HRV, which reflects PNS activity (Laborde et al., 2017) and has a good applicability within ultra-short-term measurements (Shaffer & Ginsberg, 2017). The RMSSD was calculated with the PhysioNet Cardiovascular Signal Toolbox (Vest et al., 2018).

2.3.4. Assessment of body sway

Postural body sway was assessed with a force plate for biomechanics (Type 9260AA6, 0.6 m × 0.5 m; Kistler^® Instrumente, Winterthur, Switzerland), with a 100 Hz sampling rate, and filtered with a second order Butterworth, 0.1–20 Hz band-pass filter. With specialized software (Kistler MARS by S2P, Winterthur, Switzerland) (Šarabon, 2011) we obtained the centre of pressure (COP), determined by excursions in the anterior–posterior (AP) and the mediolateral (ML) directions. The total sway path, calculated as a sum of the point-to-point Euclidian distances, gave direction-independent information about the length of the trajectory of the COP. To assess shivering, we used the mean frequency of the 2–10 Hz spectrum band, defined as the frequency of the oscillations of the COP, calculated as the mean frequency of the power spectrum for the frequency band 2–10 Hz for the AP and ML directions. We chose this frequency band because the normal unconstrained standing frequency is < 1 Hz (Duarte et al., 2000), while involuntary trembling occurs in a frequency band between 3 and 10 Hz (Maatar et al., 2012).

2.4. Data management, preprocessing, and statistical analysis

Statistical analyses were conducted using IBM SPSS Statistics software (IBM Corp., Armonk, NY, USA) (IBM Corp., 2020). Because of technical issues or premature termination of the experiment by two participants, we were not able to include data for all participants. Therefore, the actual sample size is indicated in the corresponding reported results.

First, we examined the variables with regard to outliers. For normally distributed data, we defined outliers as 3 standard deviations (SD) apart from the mean group value. In line with Holz et al. (2016), outliers were converted into new values, which were one unit above or below the next most extreme value in the distribution, while one ‘unit’ was defined as the mean deviation of the values in the corresponding variable, without considering outliers (Supplemental Table 2). Before performing outlier analysis for the HRV parameter RMSSD, we transformed it into its natural logarithm, lnRMSSD (Laborde et al., 2017; Shaffer & Ginsberg, 2017).

We compared distress induction between the neutral and the stressful films using a one-tailed Wilcoxon signed-rank test for these non-normally distributed data.

2.4.1. Heart rate and heart rate variability during the upregulation and downregulation phases

To test whether HR increases in the upregulation phase and decreases in the downregulation phase, we correlated bpm with time (as a serial number within the given timeframe) for each participant and scene. To depict the whole cascade, we performed this calculation for the last 45 s of baseline before the stressful film, for the entrance scene, incipient threat scene, and trapped scene, and for the corresponding four segments of the neutral film. The resulting coefficients per participant and scene were Fisher z-transformed, to approximate the distribution of the correlation coefficients to the normal distribution. Thereafter, we tested the acceleration and deceleration of HR per scene using one-sample t-test against zero (two-tailed) (see Supplemental Results). To compare mean slopes, we computed a scene (4) × condition (2) repeated measures ANOVA, modelled the interaction effect, and corrected for violations against sphericity. Differences in means between scenes were tested by Bonferroni-adjusted pairwise comparison of the estimated marginal means. We report associated confidence intervals (CIs).

To test whether HRV is lowest in the upregulation phase and higher during the trapped scene downregulation phase, we computed a single mean lnRMSSD score for each 45 s segment. This was based on the methodological consideration between the recommended 1 min acquisition time (Shaffer & Ginsberg, 2017) and the minimum required 10 s epoch for ultra-short assessments (Munoz et al., 2015; Nussinovitch et al., 2011). Again, a scene (4) × condition (2) repeated measures ANOVA was conducted. To control for potential confounders, we reran both repeated measures ANOVAs with the covariates age, body mass index (BMI), menstrual phase, and presentation order (Supplemental Table 3a and b).

2.4.2. Heart rate and heart rate variability during the coactivation phase

To map SNS and PNS more reliably during the coactivation phase, we extended the sequence duration to 120 s. We calculated the mean bpm and lnRMSSD for the 120 s baseline, the part in the middle of the neutral film (seconds 368–488), and the part in the middle of the assault scene (seconds 368–488). Thereafter, we conducted two repeated measures ANOVAs, one for mean bpm and one for lnRMSSD, with the within-factor scene (3). We corrected for violations against sphericity. Differences in means between scenes were tested by Bonferroni-adjusted pairwise comparison of the estimated marginal means. We report associated CIs. To control for potential confounders, we reran both repeated measures ANOVAs with the covariates age, BMI, menstrual phase, and presentation order (Supplemental Table 4a and b).

2.4.3. Body sway and trembling

We compared total sway paths as well as trembling using mean frequencies in the 2–10 Hz spectrum band over the whole 12 min between the two films with a one-tailed Wilcoxon signed-rank test, because the data were not normally distributed.

2.4.4. Predictive power of physiological activity for later intrusions

Intrusions were counted if they elicited at least some distress (score above zero). To control for individual variations in psychophysiological baseline, we subtracted the mean values of bpm and lnRMSSD observed during the entire stressful film from the respective mean values of the entire previous baseline. These bpm and lnRMSSD change scores, the perceived distress, and the subsequent intrusion frequency, distress, and variety underwent Spearman’s rank correlation because the data were not normally distributed. In addition, we repeated correlational analysis for intrusion frequency and distress for each day.

According to the expected defence cascade reaction, we performed an exploratory analysis into whether the predefined scenes (upregulation phase, downregulation phase, and coactivation phase) correlated with subsequent intrusions. For each of the independent variables (intrusion frequency, distress, and variety), we calculated alpha levels adjusted by the number of correlation coefficients (k = 11) (Curtin & Schulz, 1998), resulting in the adjusted alpha level of α = .004. If we could not identify linear relationships, we tested for quadratic, inverse, and cubic distributions using curve estimation, as implemented in SPSS.

3. Results

3.1. Sample and film characteristics

The sample comprised 47 healthy women. As we wanted to model the cascade without any pre-biographical influences and only in relation to the film, it was crucial not to included prestressed participants. In line with this, the PCL-5 scores of participants who reported at least one traumatic event were negligibly low. See Table 1 for sample characteristics and Supplemental Table 5 for an overview of trauma history.

Table 1. Sample characteristics.

	Mean (SD, range) or n (%)
Age (years)	29.8 (10.7, 18–64)
BMI (kg/m^2)	24.1 (4.4, 17.6–40.4)
Menstrual phase
Follicular	18 (38.3)
Luteal	20 (42.6)
Other^a	9 (19.2)
Smoking
Never	45 (95.7)
Occasionally (< 1 cigarette/day)	2 (4.3)
Trauma history
Self-experienced a traumatic event (LEC-5)	15 (31.3)
PCL-5 score	2.3 (2.8, 0–9)

Note: ^a Six had no cycle owing to the menopause or contraception, two were in the ovulation phase, and one was 4 months pregnant.

BMI = body mass index; LEC-5 = Life Events Checklist for DSM-5; PCL-5 = PTSD Checklist for DSM-5.

Distress during the stressful film (M = 6.89, SD = 3.50) was significantly higher than during the neutral film (M = .20, SD = .62; z = −5.79, p < .001, n = 46, Cohen’s d = 3.23).

3.2. Heart rate and heart rate variability during the upregulation and downregulation phases

Compared to baseline (r = .07), HR decreased significantly during the entrance scene (r = −.23, p = .005) and the incipient threat scene (r = −.20, p = .021), but not during the trapped scene (r = .10, p = 1), while no such pattern was found during the neutral film. Hence, there was a significant condition by scene interaction [F(3, 135) = 6.17, p < .001, η²_p = 0.121, CI_one-sided = 0.04, 1.00; n = 46], showing a downregulation during the first two scenes, but no trend during the trapped scene (Figure 2, Supplemental Table 6, and Supplemental Figure 2a). After controlling for confounders, this interaction did not remain significant. For HRV, no significant condition by scene interaction [F(2.14, 96.35) = .732, p = .493, η²_p = 0.016, CI_one-sided = 0.00, 1.00; n = 46] was observed (Supplemental Figure 2b), and this remained non-significant after controlling for confounders (Supplemental Table 3a and b)

Figure 2. Heart rate (HR) during the trauma-film analogue experiment. Analysing the predefined scenes from the movie Irréversible, we observed that HR during the entrance and incipient threat scenes pointed towards an orienting reaction, while there was no significant change in HR in the trapped scene. During the assault scene, HR increased significantly until it plateaued at a high level. Displayed are the mean HR and the corresponding 95% confidence interval for baseline and the full 12 min stressful film. Black frames indicate predefined scenes, while the grey box indicates the sequence of the ‘assault scene’ used for the coactivation calculation. Please note that there was a short break of several seconds between baseline and the entrance scene.

3.3. Heart rate and heart rate variability during the coactivation phase

HR during the assault scene (M = 91.02, SD = 17.01) was higher than during the corresponding neutral scene (M = 85.74, SD = 14.34, p = .009, Cohen’s d_{RM, pooled} = 1.31, CI = 0.87, 1.75) and during baseline (M = 80.66, SD = 13.99, p < .001, Cohen’s d_{RM, pooled} = 1.83, CI = 1.37, 2.29). Hence, there was a significant main effect of scene [F(1.45, 62.90) = 25.60, p < .001, η²_p = 0.379, CI_one-sided = 0.22, 1.00; n = 43] (Figure 3a). HRV also varied significantly between scenes [F(1.59, 64.93) = 12.13, p < .001, η²_p = 0.224, CI_one-sided = 0.09, 1.00; n = 43] (Figure 3b) and post-hoc analysis revealed a significantly lower HRV during the assault scene (M = 2.83, SD = .64) than during baseline (M = 3.12, SD = .62, p < .001, Cohen’s d_{RM, pooled} = −0.84, CI = −1.23, −0.41), but failed to show significant difference between the assault scene and the neutral scene (M = 2.97, SD = .47, p = .082, Cohen’s d_{RM, pooled} = −0.50, CI = −0.93, −0.07). In contrast to HR, for HRV this main effect remained significant after controlling for confounders (Supplemental Table 4a and b), and the post-hoc comparison between the assault scene and the neutral scene also reached significance (p = .021).

Figure 3. Heart rate (HR) and heart rate variability (HRV) during the ‘assault scene’ of the stressful film, compared to baseline and the neutral scene. Coactivation during the assault scene is depicted by the highest HR during this scene, while HRV shows no difference between the neutral scene and the assault scene. Differences in mean are shown for (a) HR in beats per minute (bpm) and (b) mean HRV as the natural logarithm of the root mean square of successive differences (lnRMSSD). Each of the scenes had a duration of 120 s. Error bars indicate the 95% confidence interval.

3.4. Body sway and trembling

Total sway path during the stressful film was significantly longer than during the neutral film (median = 6.89 m, range = 3.67–13.53 m vs median = 6.40 m, range = 3.31–12.66 m; z = −1.77, p = .038, n = 41, r = −.28). Trembling frequency during the stressful film was significantly higher than during the neutral film (median = 3.04 Hz, range = 2.59–3.85 Hz vs median = 2.83 Hz, range = 2.59–3.6 Hz; z = −4.05, p < .001, n = 42, r = −.62).

3.5. Predictive power of physiological activity for later intrusions

All participants together reported in total 369 events of intrusions, of which 342 evoked at least some distress. Each participant experienced on average M = 7.28 (SD = 7.77, range = 0–30) intrusions, with a distress level of M = 3.91 (SD = 1.86, range = 1.00–7.75). Frequency and distress decreased over the week. Given the opportunity to report multiple qualities within one intrusion, most of the participants reported at least one involuntary thought (sum = 262, n = 38), followed by flashbacks (sum = 67, n = 20), pictures (sum = 51, n = 18), bodily sensations (sum = 29, n = 10), sounds (sum = 15, n = 8), nightmares (sum = 4, n = 4), and smells (sum = 2, n = 2) (Figure 4).

Figure 4. Reported intrusions over the subsequent week in (a) mean number of intrusions and (b) mean perceived distress, and (c) an overview of all reported qualitative categories, with the corresponding 95% confidence interval.

We found no significant correlation between psychophysiology (Supplemental Table 7) and reported intrusions in the subsequent week (Table 2) or for each day (Supplemental Table 8), and curve fits were also not significant. The same was true when we examined the predictive power of HR or HRV within specific scenes (Supplemental Table 9). The subjective stress experience during the stressful film, however, did relate to the frequency, distress, and variety of subsequent intrusions, although the significance of intrusion distress did not withstand alpha-level correction (Table 2).

Table 2. Spearman’s rho correlations for physiological change scores (film minus baseline) during the stressful film: intrusion frequency, distress, and variety in the subsequent week and distress during the stressful film (two-tailed).

	Intrusion frequency	Intrusion distress	Intrusion variety
HR change score (bpm)	−.12	−.18	−.11
HRV change score (lnRMSSD)	.18	−.09	.04
Distress during film	.43*	.36†	.49**

Note: we used adjusted alpha levels to account for alpha error cumulation. p-Values ≥ .004 are considered as non-significant. *p < .004, **p < .001, †p < .05.

HR = heart rate; bpm = beats per minute; HRV = heart rate variability; lnRMSSD = natural logarithm of the root mean square of successive differences.

4. Discussion

In this study, we aimed to empirically model physiological reactions predicted by the defence cascade model along the fight and flight reaction (upregulation phase; high HR and low HRV), followed by the passive defence response (downregulation phase; low HR and high HRV), and terminating in switching between upregulation and downregulation (coactivation phase; high HR and high HRV). The phases were chosen according to predefined film scenes.

Contrary to these assumptions, we observed an alteration of HR over time, which was distinctive for the stressful film and not present in the neutral film. The entrance scene was characterized by a decrease in HR, which dropped even more during the incipient threat scene, reached a plateau during the trapped scene, and thereafter switched to HR acceleration. For HRV, we found no effects depicting changes during these scenes. The later coactivation phase was characterized by higher HR compared to the neutral film and lower, but non-significantly, HRV. However, a significant reduction in comparison to baseline could be observed for the coactivation phase as well as for the neutral film. Accordingly, it can be assumed that participation in the experiment already led to heightened physiological arousal, represented by lower lnRMSSD. Comparing physiological reactions during coactivation, higher HR during the stressful film, without a significant decrease in lnRMSSD, corresponds to our assumption of high SNS activity with simultaneous PNS regulation. Although the RMSSD correlated with HR activity, a decoupling in the data indicates that SNS and PNS were coactive at the same time (Nackley & Friedman, 2021). In line with this, we found heightened body sway and trembling during the stressful film, indicative of hyperaroused movement patterns dominating during the stressful film.

Nevertheless, the results showed an opposite pattern to that predicted by the defence cascade model. We can offer several explanations for this rather surprising result. The first is that the defence reaction may be more dynamic and does not necessarily imply passage through a cascade.

For instance, the response to a threat depends on the spatiotemporal distance of the threat (Mobbs et al., 2020), meaning that distant treats can simply be avoided, and the fight or flight response may not be initiated. The degree to which the victim is in the focus of the perpetrator is also relevant (Bastos et al., 2016). These circumstances lead to different adaptation of physiological processes (Low et al., 2015).

Therefore, the orientation response may take longer than we had assumed. Schauer and Elbert (2010) describe the orientation response as a decrease in HR accompanied by attentional alertness. This reaction typically persist for few seconds directly after the threat is perceived, and therefore empirical studies using fearful threats, such as gun attack simulations, mainly focus on only short time intervals (Gladwin et al., 2016; Hashemi et al., 2019; Low et al., 2015). Our data indicate that this reaction can also persist for longer. The fear-related bradycardia that we observed lasted for approximately 2 min before the HR started accelerating. Although the study by Hagenaars et al. (2014b) showed an early onset of the orientation response within the first 2 s, based on an analysis of body sway data, the decline in their HR data as a marker for the autonomic system persisted for the entire film clip.

The lack of a subsequent downregulation phase can be explained by a misattribution of the coactivation. One could assume that the physiological reaction that we predicted as the coactivation phase is actually a prolonged freezing, also described as the ‘fright phase’ Schauer and Elbert (2010). According to the authors, this phase occurs as a peak of the upregulation phase before switching into downregulation, when the active actions of fight and flight are no longer an option, and is characterized by HR acceleration and high emotional arousal. In our experiment, participants were asked to stand still and watch the film, while they were hooked up to the instruments. This set-up may have conveyed a sense of being trapped, but was not powerful enough to induce feelings such as anguish and hopelessness, which would emotionally escalate into physiological downregulation. In addition, the transition to downregulation depends on the distance of the threat and occurs when the situation is inescapable, while sensory and emotional input is too overwhelming (Terpou et al., 2019). Although this scenario is depicted in the stressful film, the participant’s body is not in danger and requires no defence, which seems to prevent them from further escalating in the defence cascade. In line with this, other empirical studies primarily evoke the stages of orientation phase, flight, and fight (Marx et al., 2008). This notwithstanding, while content-related scenes were predefined, physiological rush may vary between people. This implies that not all individual reaction patterns elapse within the same time. Still, we found different physiological reactions in our predefined scenes, and previous studies also showed that emotional film content can evoke physiological changes across individuals (Golland et al., 2014; Perez et al., 2021).

Taken together, our data revealed an extended orientation phase (indicated by a prolonged HR drop), followed by upregulation. In contrast to the predicted model, this was not followed by downregulation but culminated in the coactivation of SNS and PNS, although SNS activation was more dominant. To explain this, first, we assume that the lack of downregulation is explained by the observation that humans can reach high levels of ongoing physiological stress, without a subsequent downregulation, but, rather, staying hyperaroused (Jenks et al., 2020; Lieberman et al., 2016). Furthermore, the defence cascade is like most other physiological models based on animal observations. Although animal models support some findings for PTSD in humans (Deslauriers et al., 2018), they cannot be directly translated to humans. Besides neglecting the subjective threat experience, such models also typically do not consider age or sex (Richter-Levin et al., 2019; Shansky, 2015) and are therefore not generally transferable to clinical practice. A second potential explanation relates to the laboratory character of this study. Experiments using stress inductions without trauma-related content regularly report similar results to those found in our study; namely, HR acceleration without any downregulation (Hellhammer & Schubert, 2012; Reinhardt et al., 2012). Hence, our threat induction may have borne more resemblance to stress than to real traumatic events. Thus, the limited potential of trauma-analogue studies to evoke the orientation phase (Chou et al., 2014; Holmes et al., 2004; Nackley & Friedman, 2021) and upregulation phase (Danböck et al., 2021; Holz et al., 2016; Weidmann et al., 2009), but no downregulation, should be taken into consideration in future study design.

As our second aim, we investigated whether physiological reactions during the trauma-film paradigm effect subsequent trauma-analogue intrusion. As expected, the trauma-film paradigm was successful in evoking intrusive involuntary thoughts and sensations related to the stressful film on moderate distress levels and a broad range of qualities up to nightmares. However, there was no correlation between subsequent intrusions and physiological measures, either during the stressful film or with predefined scenes. More predictive was the subjective experience of distress during the movie, which explained 12% of the variance. The distress during the film was not related to the physiological parameters. This leads us to conclude that physiological reactions during stress are not accountable for following post-traumatic intrusions. Using the same film scene, Weidmann et al. (2009) found an acceleration of HR, subjective stress, and intrusions, which is comparable to our findings, but the authors did not report correlations between these measures. Examining relationships, Holz et al. (2016), similarly to us, found no correlation with physiological markers, but did find a correlation between heightened negative affect after the film and the distress of subsequent intrusions. Therefore, physiological arousal may be of minor relevance, while subjective experience has more weight in developing PTSD-associated symptoms (Norr et al., 2016; Ozer et al., 2003). In contrast, using a similar paradigm and film material, Danböck et al. (2021) found that HR acceleration predicted subsequent intrusions within the following 24 h.

To provide a more comprehensive picture of the origins of intrusive experience, further studies should also investigate psychological responsiveness in the recovery stages; hence, after watching the film. This seems crucial as we omitted to show the downregulation phase, while participants had continuously high HR during the second half of the stressful film. Furthermore, physiological recovery is impaired in PTSD (Norte et al., 2013; Pole, 2007; Porges, 2022).

Our study is subject to several limitations. First, we underrated the impact of the entrance scene as a hint for the upcoming stressful film. Because it did not contain explicit threat stimuli, we did not expect a stress-related adaptation. Nevertheless, the dark, oppressive atmosphere presented in the entrance scene was seemingly sufficient to inform the participants about the condition. One may interpret our entrance scene as part of a prolonged orientation phase. This interpretation is supported by the decline in HR, and is consistent with the assumptions of the defence cascade model (Schauer & Elbert, 2010). Freezing is a part of the orientation response and occurs immediately after threat detection (Hagenaars et al., 2014a). This first physiological response depends on the individual trauma history. People with aversive life events show greater HR reduction in response to unpleasant pictures than people who reported no such events (Hagenaars et al., 2012). Therefore, we did not expect to detect such a reaction in a sample without a trauma history, using a scene that does not contain concrete threat cues. Secondly, other factors besides film content may have contributed to the HR acceleration. Those are primarily standing for a long time and the stress evoked by the experimental situation. This is plausible as we observed this acceleration also in the neutral film sequence, although to a much smaller degree than in the stressful film. Thirdly, owing to adaptation of our preregistered analytics plan, the study may be underpowered. This is especially the case for the analysis including confounders. After controlling for confounders for HR, neither the scene × condition interaction effect during upregulation and downregulation, nor the main effect of scene during coactivation remained significant. In contrast, the effect for HRV during coactivation became significant after controlling for confounders. Thus, it should be considered that the shrinkage of effect sizes to non-significance – for almost all effect sizes, except for the scene × condition interaction for HRV – may be caused by inflation of dependent and confounding variables. To encounter this, we reported CIs of the relevant effect sizes. Although these results should, therefore, be interpreted with caution, they serve as a direction for normative physiological behaviour. Fourthly, we decided to include only women, which does not allow the data to be generalized to men, owing to physiological and psychological differences in values (Adjei et al., 2018; Koenig & Thayer, 2016; Verma et al., 2011). Although we do not close the gap on this relevant aspect in differentiating psychophysiological stress reactions, our focus on women, especially in the context of sexual violence, unfortunately, still reflects a social reality (Borumandnia et al., 2020; Jina & Thomas, 2013). Gaining more clarity on basal psychophysiological mechanisms in this group may help to expand the findings to other populations in the future.

All things considered, the trauma-film paradigm does not support the defence cascade model. Rather, the trauma-analogue response in humans is characterized by an orientation response with a subsequent hyperarousal upregulation and a constantly heightened SNS-dominated, but PNS-regulated, physiological plateau. However, this paradigm, which was performed under highly controlled, yet artificial laboratory settings with healthy female participants, is not generalizable to more naturalistic conditions. The observed effects may be more pronounced in people with a previous trauma history (Hagenaars et al., 2012) and in patients with PTSD, who show a greater reduction in HRV due to trauma cues than healthy controls (Rabellino et al., 2017). While physiological reactions do not seem to play a major role in the development of subsequent intrusions, subjective distress during the stressful experience seems to be more relevant.

Acknowledgements

The authors would like to thank Marcel Franz for his help in automating data preprocessing, and Theresa Quaas and Sophie Scharff for their help with study recruitment, conduction, and data transition. The authors also express their gratitude to all women who participated in the study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability

The data that support the findings of this study are openly available in Open Science Framework at https://osf.io/r4g9p/. Please note that five participants decided not to agree to their data being shared publicly; therefore, data for 42 participants are available. With the intention of ensuring privacy rights, we further anonymized the data (see the attached document in the OSF project).

Additional information

Funding

This research was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) [grant number CR 479/9-1] (http://www.dfg.de/). The work by Sarah Beutler was supported by a fellowship from the Heinrich Böll Foundation. The Article Processing Charges (APC) were funded by the joint publication funds of the TU Dresden, including Carl Gustav Carus Faculty of Medicine, and the SLUB Dresden as well as the Open Access Publication Funding of the DFG.

Notes

1 The exploratory hypothesis regarding variety of quality as an outcome variable was added after preregistration. As planned, we assessed different qualities of intrusions using a multiple-choice format. Hence, participants reported multiple qualities for one event, with different qualities over the day. Therefore, we were unable to assign a singular quality type to the subjects, contrary to our best efforts. To test for relationships between physiological parameters and quality components, we implemented the outcome variable ‘variety of intrusions’. This reflects the diversity of perceived qualities during the week but does not set different qualities in a hierarchical order.

2 We changed the preregistered analytical plan. Instead of the preregistered repeated measurements ANOVA with five repetitions (first baseline, 45 s in the middle of the neutral scene, ‘entrance scene’, ‘incipient threat scene’, ‘trapped scene’), the use of timewise analogue sequences for the neutral and the stressful films was deemed more appropriate and caused a 4 (scene) × 2 (condition) design. The power calculation is based on a repeated measurement design with eight repetitions. This is similar but not congruent to a 2 × 4 repeated measurement design, for which the power calculation is more complex and not supported by regular statistical software. Therefore, the above calculated sample size serves as a benchmark. For critical evaluation of the obtained effect sizes, we report confidence intervals (CIs). For the main and interaction analyses we report one-sided 95% CIs for the partial eta square (CI_one-sided). Within this approach, the upper CI is fixed to 1.00 to avoid overestimation of the effect size. For subsequent post-hoc comparisons, we used Cohen’s d for pooled repeated measures (Cohen’s d_{RM, pooled}) and related CIs, which takes intercorrelation of the repeated measures into account (Lakens, 2013; Lenhard & Lenhard, 2016).