Posttraumatic Stress and the Comprehension of Everyday Activity

Participants were recruited immediately after their participation in a cross-sectional study of the relationship between age, cognitive ability, and event segmentation [12]. Two hundred eight adults from the Sargent et al. [12] study were given the option to participate in this study and complete a packet of PTSD questionnaires at home. Sargent et al. [12] recruited these participants from the Volunteer for Health participant registry at the Washington University School of Medicine. Participants received $10 for participating in the study. Of the 208 adults who completed the cross-sectional study, 137 participants correctly completed and returned the questionnaires (51% female; 72% White; mean age = 50 years, age range 20–79, 18–28 participants in each decade; mean years of education = 15.1), and data from these 137 participants were included in all of the below analyses. The study received approval from the Washington University Human Resources Protection Office and participants provided written informed consent before participating in this study.

The mean age of those who declined to participate did not differ from that of participants, t(231) = –0.723, p = 0.42. Five participants did not provide their level of education, and these data were therefore imputed using data from the non-PTSD-related measures of the 208 participants who completed the cross-sectional study using the expectation maximization (EM) procedure in SPSS 19.0. Sixteen of those participants who declined to participate did not provide their level of education and did not provide enough other data for this information to be imputed; however, for the data available, mean levels of education differed between the included (mean = 15.1 years, sd = 2.5) and excluded (mean = 14.1, sd = 2.7) participants; t(215) = 2.76, p = .006.

The event segmentation task required participants to view three movies of everyday activities and push a button whenever they believed “one natural and meaningful unit of activity has ended and another has begun.” Each of the movies depicted an actor completing a neutral, everyday activity (making breakfast, preparing for a party, and gardening). The order of movie presentation was fixed across participants. The movies were filmed from a fixed, head-height perspective, with no zoom, and the movies ranged in length from 329 seconds to 376 seconds. Participants segmented each movie twice: at a coarse grain (push the button whenever you believe one large natural and meaningful unit of activity has ended and another has begun) and at a fine grain (push the button whenever you believe one small natural and meaningful unit of activity has ended and another has begun).

Performance on the event segmentation task was measured by calculating a segmentation agreement score for each participant. Segmentation agreement is a measure of the similarity between each participant’s segmentation and the segmentation of the group as a whole, and it was calculated using the methods provided in Kurby and Zacks [16]. First, time in each of the movies was divided into one-second bins. Then, the group norms for fine and coarse segmentation were calculated by determining the proportion of participants who identified an event boundary within each bin. Coarse and fine segmentation agreement scores for each participant were generated by correlating each participant’s coarse segmentation with the coarse group norm and each participants’ fine segmentation with the fine group norm, and scaling the resulting correlations to 0–1 variables based on the maximum and minimum correlation possible given how many boundaries each participant identified. Finally, segmentation agreement scores for fine and coarse segmentation were averaged to obtain a final segmentation agreement score for each participant¹.

Following the first viewing of each movie, participants completed a recall task, for which they were given seven minutes to write or type as much as they could remember from the movie. Recall was scored using methods similar to those described in Schwartz [17]. Three experimenters independently viewed each movie and listed every meaningful action performed by the actor in the movie. Each meaningful action agreed upon by the three experimenters was then included in the scoring template. Participants received one point for every phrase in their free recall data that matched one of the units on the scoring template. The final event memory score for each participant was calculated by summing the points obtained for each movie and then averaging across movies, inter-rater kappa = 0.84, p < .001, 95% CI [0.78, 0.90].

Participants also completed a psychometric battery that included three tasks in each of five domains of cognitive function: working memory (reading span, operation span, and symmetry span), laboratory episodic memory (selective reminding, verbal paired associates, and word list recall), executive function (reading with distraction, trail making, and Ruff figural fluency), processing speed (shape completion, letter comparison, and pattern completion), and general knowledge (information test, synonym vocabulary, and antonym vocabulary). (The event comprehension and psychometric measures were completed as part of the Sargent et al. [12] study, and are described in more detail in that paper.)

Participants completed a packet of questionnaires that included the Traumatic Life Events Questionnaire (TLEQ; [18]) and the PTSD Screening and Diagnostic Scale (PSDS²; [19]). The TLEQ is designed to determine exposure to twenty-two types of potentially traumatic events. Participants reported whether they had experienced each type of event, the number of times they had experienced the event, whether they experienced fear, helplessness, or horror during the event, and whether they were seriously injured due to the event. They then indicated which of these events currently caused them the most distress.

Participants were instructed to complete the PSDS using the most traumatic event they indicated on the TLEQ. Participants were instructed to complete the PSDS even if they indicated that they had not experienced any traumatic events. The PSDS is a thirty-eight item questionnaire designed to assess severity of PTSD symptoms based on DSM-IV criteria. It has high internal consistency (α = 0.93), test-retest reliability (r = .95), sensitivity (94–100%), specificity (63–86%), positive predictive power (78–90%), negative predictive power (83–100%), and convergent validity [19]. PTSD symptom severity scores were calculated using scores on questions four through twenty-three, for which participants rated on a five point scale the degree to which they had experienced each of seventeen symptoms of PTSD during the previous thirty days. Scores for each of the three DSM-IV PTSD clusters were calculated separately by summing scores on the questions corresponding to each cluster. Total PTSD symptom severity ratings can range from zero to eighty.

During the cross-sectional study, participants completed two 150-minute sessions in a private room in the laboratory. During the first session, participants segmented the three movies at a coarse grain and completed the recall task for each of these movies. They then completed the Reading Span, Operation Span, Symmetry Span, Shape Comparison, Reading with Distraction, and Synonym and Antonym Vocabulary tasks. During the second session, they segmented the three movies at a fine grain and completed the Selective Reminding, Verbal Paired Associates, World List Recall, Trail Making, Ruff Figural Fluency, Letter Comparison, Pattern Comparison, and the Information Test tasks. After completing the second session, participants were asked whether they were willing to complete the packet of PTSD questionnaires at home. If they agreed, they were given the packet of questionnaires and a pre-addressed and stamped envelope so that they could send the completed questionnaires back to the experimenter.

On the TLEQ, seven participants reported that they had experienced no traumatic life events, six reported one traumatic life event, and 124 reported two or more traumatic life events. PTSD symptom severity scores obtained from the PSDS ranged from 0 to 58. Scores ranged from 0 to 20 for the intrusive recollection cluster, from 0 to 20 for the avoidance and numbing cluster, and from 0 to 17 for the increased arousal cluster. All seven of the participants who stated that they had not experienced any traumatic events had a score of 0 on the PSDS. See Table 1 for additional descriptive statistics from the TLEQ.

For the event segmentation task, participants varied greatly in the number of segments they identified in the movies. In the coarse segmentation condition, the number of button presses ranged from 1 to 202 (mean = 24.32, sd = 23.74), while in the fine segmentation condition, the number of button presses ranged from 8 to 288 (mean = 61.59, sd = 48.37). However, participants tended to segment a similar number of times across movies: Cronbach’s alphas for coarse segmentation and fine segmentation were .89 and .94, respectively (Breakfast Coarse: mean = 27.76, sd = 26.94; Party Coarse: mean = 23.23, sd = 23.81; Gardening Coarse: mean = 22.19, sd = 18.70; Breakfast Fine: mean = 59.04, sd = 39.24; Party Fine: mean = 67.21, sd = 50.87; Gardening Fine: mean = 58.51, sd = 53.64). The number of meaningful units identified was not significantly correlated with PTSD severity (coarse segmentation: r = 0.05, p = .55; fine segmentation: r = 0.02, p = .82).

To reduce error and loss of degrees of freedom in later analyses, scores on the cognitive tasks were combined into five composite variables: working memory, verbal episodic memory, general knowledge, executive function, and processing speed. See Table 2 for descriptive statistics for these variables. To create the composite variables, we z-scored the scores for each of the tasks and then averaged the resulting scores. A confirmatory factor analysis found that all of the composite variables except executive function formed latent variables (see [12]). Therefore, the executive function variable was dropped from further analyses.³

Gender, age and education were included as covariates in the subsequent analyses. Gender was not correlated with PTSD symptom severity (r = 0.06, p = .24; males coded as 0, females coded as 1) or with segmentation agreement (r = –0.04, p = .64). We found significant negative correlations between age and PTSD symptom severity (r = –0.24, p = .004) and education level and PTSD symptom severity (r = –0.25, p = .003). However, neither variable was significantly correlated with segmentation agreement (r = 0.12 and –0.03, respectively).

To determine whether to conduct a standard hierarchical multiple regression analysis, the distributions of the education, PTSD symptom severity, and cognitive composite variables were checked for skew and kurtosis (see Table 2). Despite the non-normality of some of the predictors, the residual distributions were normal and homoscedastic, and parametric statistics were therefore used in the following analyses.

Participants with greater PTSD symptom severity had lower segmentation agreement scores; the simple correlation between PTSD and segmentation agreement was –0.26 (p = .002; See Figure 1 for a scatter plot of this relationship). To determine whether this relationship held after controlling for age, education, and general cognitive functioning, these variables were entered into a hierarchical regression analysis. Gender was entered first into the equation, followed by age and education. The four composite variables were then entered into the equation. Together, these variables accounted for 14.8% of the variance in segmentation agreement (p = .002). When PTSD symptom severity was added into the model after these variables, it accounted for an additional 3.8% of the variance in segmentation agreement (p = .02), and in the final model, only PTSD explained unique variance in segmentation agreement, See Table 3 for a summary of the hierarchical regression analysis that includes age, education, the composite variables and PTSD symptom severity.⁴

The simple correlation between PTSD and event memory was –0.16 (p = .05; See Figure 1 for scatter plot of this relationship), and the simple correlation between segmentation agreement and event memory was 0.46 (p < .001). When gender, age, education, and the four composite variables were entered into the model, these variables accounted for 36.3% of the variance in event memory (p < .001). PTSD symptom severity did not explain additional unique variance (p = .16). In the final model, only gender and working memory explained unique variance in event memory. See Table 3 for a summary of the hierarchical multiple regression analysis that includes gender, age, education, the composite variables, and PTSD symptom severity.

The PSDS provides symptom severity scores for each of the three PTSD symptom clusters separately. To test the hypothesis that increased arousal, including hypervigilance, is related to differences in perception of ongoing activity in people with PTSD, the symptom clusters were entered into three additional hierarchical multiple regression analyses. After accounting for the effects of age, education, and the cognitive composite variables, increased arousal accounted for an additional 3.0% (p = .03) of the variance in segmentation agreement and 1.6% (p = .07) of the variance in event memory. Reexperiencing accounted for an additional 3.2% of the variance in segmentation agreement (p = .03) and 0.1% of the variance in event memory (p = .29). Finally, avoidance accounted for an additional 1.7% (p = .10) of the variance in segmentation agreement and 0.1% (p = .17) of the variance in event memory.

The association of PTSD symptom severity with less adaptive event segmentation is consistent with findings of negative effects of PTSD on cognition (see [20], for a review). In addition, the finding that increased arousal negatively predicted segmentation agreement supports the hypothesis that this symptom cluster is relevant to the mechanism relating PTSD and the perception of ongoing activity. Of particular note, participants in the current study only viewed neutral movies that were unlikely to be related to their traumatic events, suggesting that the effects pervade everyday cognition.

What mechanism or mechanisms account for the relationship between PTSD and the perception of ongoing activity? One potential cause is a difficulty making accurate predictions. According to Event Segmentation Theory [13], event segmentation results from predictive processes during ongoing perception. As people experience the world, the brain makes predictions about future inputs, and there is evidence that the anterior cingulate cortex (ACC) is important for maintaining these predictions and calculating prediction error [21]. When prediction error spikes, the dopaminergic midbrain is involved in triggering a widespread brain response that includes perceptual orienting and memory updating; this response is experienced as an event boundary [13 22]. One possibility is that symptoms of PTSD compromise one or more of the steps along the pathway that includes perceptual prediction, error monitoring, orienting, and memory updating.

There is currently little research on the relationship between PTSD and prediction ability. However, neuroimaging studies on PTSD and on mechanisms involved in prediction provide some evidence for the existence of a prediction deficit in PTSD. First, a robust finding in the PTSD functional magnetic resonance imaging (fMRI) literature is that the dorsal ACC is hyperactivated in people with PTSD, even during tasks that are unrelated to participants’ traumatic events. This brain area is typically involved in tasks such as performance monitoring, response selection, and error detection [23]. For example, in studies very similar to the ERP studies discussed earlier, Bryant et al. [24] found hyperactivity in the dorsal ACC of people with PTSD in an auditory oddball task that was unrelated to trauma and emotional processing. Felmingham et al. [25] used the same auditory oddball task but also measured skin conductance response to target tones to determine autonomic arousal. They found that during target trials in which participants displayed a skin conductance response, participants with PTSD displayed greater dorsal ACC activation than controls. The authors of both of these studies suggest that people with PTSD display increased attention, vigilance, and processing of salient stimuli, consistent with the hyperarousal symptoms of PTSD. Because sensory oddballs are by definition unpredictable and likely to lead to prediction errors, it may be that these studies reflect hypersensitivity to prediction error amongst those suffering from PTSD.

These previous studies suggest two possible mechanisms that could lead to impaired event segmentation in PTSD. First, chronically elevated prediction error signals could mask transient prediction error spikes, rendering segmentation less reliable. Second, constant vigilance could rob processing resources from the encoding of ongoing activity, resulting in poorer event representations and therefore poorer prediction. In other words, inappropriate monitoring of the environment could function as a secondary task that impairs ongoing event encoding. Interventions directed at these mechanisms may be a promising avenue to address impairments of understanding and memory for everyday activity in PTSD. One possibility would be to train people with PTSD to segment activity more adaptively using feedback, thereby scaffolding the impaired error monitoring mechanisms and reducing the concurrent processing demands.

In a similar way, people with high levels of reexperiencing symptoms may suffer reduced processing availability for encoding the ongoing activity. At event boundaries, when adaptive comprehension already requires greater cognitive effort to reorient and update memory, higher levels of reexperiencing symptoms may reduce available processing capacity and result in the type of disrupted segmentation apparent in the present study.

On the other hand, PTSD symptom severity did not significantly predict performance on the event memory task, though the simple correlation between these measures was only just above the significance cutoff at p = .05. As discussed above, previous research suggests that people with PTSD experience memory difficulties (e.g., [15]) and that better performance on the event segmentation task is associated with better memory (e.g., [13]). Given that, in the current study, people with higher PTSD symptom severity displayed greater difficulty with the event segmentation task, it was expected that a similar relationship would hold for event memory. One possibility for the trending, though non-significant, finding in the current study is that the relationship between PTSD symptoms and event memory is strongest in people with clinical levels of PTSD and that a full clinical PTSD sample would be necessary to find a significant relationship between these measures.

It is important to note that the correlational design used here does not establish the directionality of the relationship between PTSD and difficulties with perception and memory of ongoing activity. On one hand, it is possible that PTSD causes deficits in event segmentation: Some people who experience a traumatic event may experience neural changes that alter the way they perceive both their initial traumatic event and later events. One potential mechanism of such an effect is that PTSD-related hyperactivation of the dorsal ACC may lead people with PTSD to be chronically hypervigilant in their search for danger, contributing to an increase in prediction error and therefore to deficits in event segmentation. On the other hand, people who have factors that lead to poorer event segmentation before they experience a traumatic event may be at a higher risk for developing PTSD. Again supposing that ACC hyperactivation produces poorer event segmentation, one possibility is that premorbid ACC dysfunction is a risk factor for PTSD. Recent work by Shin et al. [23] in twin pairs discordant for combat supports this suggestion. The authors found that dorsal ACC hyperactivation is a familial trait that is present even in non-combat exposed co-twins of a PTSD positive twin. If correctly processing and understanding events requires normative ACC activation and event segmentation performance, people who have deficits in these areas may be predisposed to develop PTSD, a syndrome that by definition involves a failure to interpret events in an adaptive manner.

In addition, participants in this study were not given a full clinically administered assessment, and it was therefore not possible to obtain clinical diagnoses for participants. However, the PSDS has high reliability and validity [19]. Furthermore, twenty-eight participants reported PTSD severity scores that were equal to or greater than 26, which is the most conservative cut-off score for clinical PTSD found in the literature [19], suggesting that our pattern of results likely will hold in clinically diagnosed samples. Nonetheless, future research should determine whether these results replicate when participants are diagnosed using a clinician administered assessment and display higher PTSD symptom severity levels.

Participants in this study were not asked whether they were taking medication or receiving therapy. Therefore, it was not possible to determine whether these factors influenced the results of this study. However, it is very unlikely that these factors completely drove the relationship between higher PTSD severity and difficulty on the event segmentation task. First, many people with PTSD symptoms do not seek treatment, whether medication or therapy. In fact, Roberts, Gilman, Breslau, Breslau, & Koenent [26] found that only 53.3% of Whites, 35.3% of Blacks, 42.0% of Hispanics and 32.7% of Asians seek any type of treatment for PTSD. Second, although there are likely fewer participants low in PTSD symptomology who take psychotropic medications or seek therapy for other clinical disorders, a portion of the participants with low PTSD scores likely were taking medication or participating in therapy when they participated in the study. Indeed, the National Comorbidity Survey Replication, which surveyed 9,282 participants representative of the United States population between 2001 and 2003 found a 12-month prevalence rate of psychiatric disorders of 26.2% [27], and found that 17.9% of the total respondents had utilized mental health services during the past year [28]. Therefore, it is unlikely that including medication and therapy seeking rates in the current study would have dramatically changed the patterns of results.

This study did not assess symptoms of clinical disorders other than PTSD. This raises the question of whether the current results are specific to people with higher PTSD symptom severity or whether we would have seen similar patterns of results in people with high levels of other types of psychopathology. Although replications of this study that assess for other psychopathology would be necessary to definitively answer this question, the current study provides some evidence that our results may, at least in part, be specific to people with PTSD. If many other clinical populations display similar deficits on the event segmentation task, we would expect the avoidance symptom cluster in the current study to significantly predict segmentation agreement because many clinical disorders, particularly anxiety disorders, include symptoms of avoidance. This was not the case. The fact that only two of the three symptom clusters predicted performance on the event segmentation task suggests that specific symptoms, not just any symptoms of psychopathology, predict disruptions in comprehension of everyday activity.

In sum, the results of this study suggest that greater PTSD symptomology, and, in particular, higher levels of increased arousal and reexperiencing symptoms, are associated with difficulty adaptively segmenting ongoing neutral activity into meaningful units. The novel finding that people with higher PTSD symptom severity experienced difficulty with event segmentation suggests the possibility for a new intervention for ameliorating some symptoms of PTSD: Training people to segment ongoing activity more adaptively could improve understanding and memory for everyday activity.

This research was supported in part by NIA R01 Grant 1R01AG031150 (PI Jeff Zacks), a grant from the National Science Foundation Graduate Research Fellowship Grant No. DGE-1143954, and the Mr. And Mrs. Spencer T. Olin Fellowship. The authors thank Todd Braver and Thomas L. Rodebaugh for advice and thoughtful comments regarding the results of the study, Taylor Beck for assistance with data collection, and Chris Kurby for assistance with data processing.

The authors declare that they have no competing interests.

Michelle L. Eisenberg, Department of Psychology, Washington University in St. Louis.

Jesse Q. Sargent, Department of Psychology, Washington University in St. Louis.

Jesse Sargent is now at Department of Psychology, Francis Marion University.

Jeffrey M. Zacks, Department of Psychology, Washington University.

The author(s) of this paper chose the Open Review option, and the peer review comments are available at: http://dx.doi.org/10.1525/collabra.43.opr