Uncovering the Temporal Dynamics of Negative Thought in Posttraumatic Stress Disorder

Examining negative thought dynamics in PTSD might is important for several reasons. First, although longitudinal studies have found that PTSD is a chronic condition characterized by relative stability (Lee et al., 2020; Marmar et al., 2015), PTSD symptoms also vary on short timescales, including day-to-day and hour-to-hour (Bridges-Curry, 2025; Greene, 2021; Reeves & Fisher, 2020). Because negative thoughts are part of the symptom constellation of PTSD (i.e., negative beliefs about oneself, others, and the world [Criterion D2], self- and other-blame [Criterion D3]; Diagnostic and Statistical Manual of Mental Disorders-5-TR; American Psychiatric Association [APA], 2022), it follows that negative thought may also vary on short timescales. Further, research on the temporal course of PTSD symptoms suggests that negative posttraumatic cognitions and emotions are central symptoms to PTSD and strong predictors of variation in other symptoms over time (Bridges-Curry, 2025; Reeves & Fisher, 2020) Understanding the degree to which negative thought is a static versus dynamic component of PTSD could help inform conceptualization and treatment efforts.

Second, despite consistent findings that overall levels of negative posttraumatic thought are elevated in PTSD and represent a mechanism of change in trauma-focused psychotherapies, research has also found that certain subdomains of negative thought are differentially related to PTSD risk and treatment. For example, the PTCI reflects not only the overall degree of negative thought that one espouses, but breaks down negative thought into subdomains of self-blame, negative self-attributions, and negative other- and world-attributions (Foa et al., 1999). Some research has shown that changes in negative thoughts about the self (i.e., self-blame, negative self-attributions) predict PTSD symptom change in CPT (Schumm et al., 2015), whereas others have found that changes in self-blame predicted PTSD symptom change in CPT (Dillon et al., 2020; Holliday et al., 2018). Still, another study found that changes in negative self-, other-, and world-attributions predicted symptom change in PE (Kumpula et al., 2017). Though it is difficult to draw conclusions from these conflicting results, they suggest that (1) subdomains of negative thought might be uniquely responsive to treatment and that (2) different people may struggle with different subdomains of negative thought. Thus, examining the dynamics of negative thought overall and at the subdomain level is important.

Finally, alterations in the temporal dynamics of negative thought—namely, variability, inertia, and reactivity—might be important markers that provide insight into PTSD. The DSM-5-TR defines negative thoughts in PTSD as persistent and exaggerated (APA, 2022), which hint at the dynamic properties of these symptoms. For example, persistence infers resistance to change over time, which can be captured via inertia. Exaggeration reflects increased frequency and intensity of negative thought, providing more opportunity for variability in negative thought over time and particularly during moments when negative thought is “triggered” by environmental stress, consistent with cognitive reactivity (Beck, 1967). Though properties of variability and stability seem mutually opposing, meta-analytic research shows that psychological dysfunction is associated with emotions that are both more variable and more inert (Houben et al., 2015). This pattern of dynamics suggests that psychological dysfunction is characterized by larger and more extreme emotional shifts that once activated take longer to dissipate. Together with research documenting links between these cognitive dynamics and other forms of psychopathology, it is theoretically and empirically plausible that these dynamics might also emerge in PTSD.

Consequently, this study sought to use ecological momentary assessment (EMA) to examine the temporal dynamics of negative thought in a sample of trauma-exposed adults—approximately half of whom met criteria for a current PTSD diagnosis—and to test whether individual differences in these characteristics are associated with PTSD severity. PTSD was modeled dimensionally because symptom severity scales demonstrate better reliability and validity than categorical diagnoses (Shankman et al., 2018) and because recent work suggests that PTSD is likely best represented as a dimensional construct (Klein et al., 2024). We posed four primary hypotheses. First, we expected that PTSD severity would be positively associated with mean levels of momentary negative thought, consistent with the larger trait literature (H1). Second, we expected that variability in negative thought would be predominantly explained by state (i.e., within-person) rather than dispositional (i.e., between-subjects) factors, reflecting the time-varying nature of negative thought, and that PTSD severity would be positively associated with variability in negative thought in daily life (H2). Third, we expected that negative thought would be inert or resistant to change across time, and that inertia would be higher in people with more severe PTSD (H3). Fourth, to index reactivity, we hypothesized that perceived stress would predict concurrent shifts in negative thought, and that these shifts would be greater in people with more severe PTSD (H4).

We tested all hypotheses using a composite that includes four subdomains of negative thought often studied in PTSD: Self-Blame, Other-Blame, Negative Self-Attributions, and Negative Other- and World-Attributions. These types of negative thought were selected to mirror commonly used self-report measures (e.g., PTCI; Foa et al., 1999) and DSM-5 symptoms. We should note that the level of cognition examined here (i.e., negative thought) refers to surface-level thoughts that are negatively valenced and readily accessible to participants. We do not label instances of negative thought as distorted because that would require a level of contextual information that is difficult to achieve with the brief, repeated nature of daily life assessment. Due to possible differences in negative thought type, we examined subdomains of negative thought by separately analyzing Self-Blame, Other-Blame, Negative Self-Attributions, and Negative Other- and World-attributions. We considered these analyses fully exploratory. Findings from the current study are expected to inform the cognitive underpinnings of PTSD and the development of ambulatory interventions to address negative thoughts in real time.

Eighty-five trauma-exposed, adult community members were recruited from a mid-sized city in the Southeastern United States for a study examining the impact of trauma reminders on psychophysiology in daily life. Eligible participants were English-speaking, 18–40 years with a body mass index (BMI) between 18.5 and 34.9, and had lifetime exposure to at least one traumatic event consistent with DSM-5 (APA, 2022). BMI was an inclusion criteria due to physiological data collected as part of the parent study that are not analyzed here. Exclusion criteria were past-month trauma exposure, psychosis, and factors known to affect physiological data acquisition including dissociation, pregnancy, cardiovascular disease, and medications known to affect cardiovascular functioning (e.g., antidepressants, antihistamines, blood pressure medication). Five participants were excluded due to current antidepressant use (n = 1), active suicidal ideation (n = 2), loss to follow up before EMA (n = 1), or withdrawal from the study (n = 1).

The final sample included 80 participants– 41 trauma-exposed controls (TECs) and 39 participants who met criteria for a current PTSD diagnosis based on the Clinician Administered PTSD Scale for DSM-5 (CAPS-5; Weathers et al., 2013a). Demographic characteristics are shown in Table 1. Briefly, participants were M = 21.79 years old (SD = 4.21), 75% female, and completed at least some college (51.2%). Participants were: 35% White; 28.7% Black; 21.3% Hispanic/Latino; 7.5% Multiracial/Other; 5% Biracial; and 2.5% Asian/Pacific Islander.

Participants were recruited from the local community using online advertisements, flyers, and through a repository of previous research participants who consented to future contact. Participants completed an eligibility prescreening questionnaire using Qualtrics, which assessed for demographic and health-related information (e.g., medication use, BMI), psychosis (PRIME Screen; Miller, 2004), lifetime trauma history (Life Events Checklist for DSM-5, LEC-5; Weathers et al., 2013b), and past-month PTSD symptoms (Posttraumatic Checklist for DSM-5, PCL-5; Weathers et al., 2013c), including two questions to assess for the dissociative subtype (Herzog et al., 2020). Eligible respondents were then contacted via phone to confirm that the index (i.e., “worst”) trauma reported was at least one month ago and met for DSM-5 Criterion A trauma exposure (e.g., the event involved actual or threatened death, serious injury, or sexual violence; APA, 2022).

Eligible participants with “probable” PTSD were identified using a cutoff score of 33 on the PCL-5 (Bovin et al., 2016) and over-sampled to obtain an approximately 1:1 ratio of those with and without current PTSD. The PCL-5 was used only for sampling purposes; diagnostic status was confirmed with the CAPS-5 (described below). Recruitment was stratified based on age, sex, race/ethnicity, and trauma type (e.g., sexual assault, transportation accident, unexpected illness or death). The study was comprised of two laboratory sessions that flanked three days of ambulatory assessment, including EMA and ambulatory physiological assessment. In the first laboratory session, participants underwent structured clinical interviewing to assess for PTSD and comorbid DSM-5 mood and anxiety conditions. Next, participants completed three days of ambulatory assessment, described below. Finally, participants returned for a second laboratory session where they listened to two standardized study scripts as part of a script-driven imagery procedure and completed self-report questionnaires. Only the structured clinical interview and EMA data are analyzed here. Participants were compensated $150 USD after completing the study.

Within one week of the first laboratory session (M = 5.48 days, SD = 4.50), participants began the first of three days of ambulatory assessment. Ambulatory assessment days were often non-consecutive because participants had to come into the lab in the mornings to get outfitted with a mobile physiological acquisition device, which noninvasively and passively measures physiology using disposable electrodes placed on the skin. Participants were then oriented to the EMA survey administered using Qualtrics on a Lenovo tablet that was time-synced to the mobile device. Participants completed a practice survey and had the opportunity to ask questions. Participants then left the lab wearing the mobile equipment. Tablets were pseudo-randomly configured to administer surveys within 90-min blocks up to 17 times per day beginning at 9am and ending at approximately 11:30 pm, or when participants went to bed. The mean time between prompts was 53.67 min (SD = 13.34, range: 31–73). Participants were prompted by a tablet alarm at each survey to complete questions about their experiences in the past 10 min and had up to 20 min to respond. Participants removed the equipment at the end of each day and returned to the lab on the next scheduled day of assessment. The average number of days in between the first and third day of ambulatory assessment was 7.04 (SD = 4.36).

PTSD The Clinician Administered PTSD Scale for DSM-5 (CAPS-5; Weathers et al., 2013a) was used to determine current PTSD diagnosis and severity. The CAPS-5 is a structured clinical interview comprised of 20 items reflecting the DSM-5 PTSD symptom criteria. Items are coded for severity on a 5-point Likert scale (0 = Absent to 4 = Extreme/Incapacitating) based on the frequency and intensity of that symptom and are then summed to generate a PTSD severity score ranging from 0 to 80. Participants completed the CAPS-5 based on their past-month symptoms to assess current PTSD. All interviews were administered by a trained graduate student, audio-recorded, and independently scored by a second trained graduate student to determine reliability. Discrepancies were resolved through consensus with a doctoral-level psychologist with expertise in trauma and PTSD. Inter-rater reliability was ICC = 0.99.

Psychiatric Comorbidity The Structured Clinical Interview for DSM-5, Research Version (SCID-5-RV; First et al., 2015) was used to determine past and current psychiatric diagnoses. Modules A (Mood Episodes); D (Mood Disorders); and F (Anxiety Disorders) were administered. Diagnostic interviews were administered by a trained graduate student. A subset of interviews (n = 38) was coded by another trained graduate student to determine reliability. Inter-rater reliability ranged from ICC = 0.80–1.0.

Negative Thought Participants rated four subdomains of negative thought at each observation. Participants were prompted with the instructions, “For the following questions, please keep in mind the trauma that the study has focused on so far. In the past 10 min…” and rated items on a 7-point Likert scale ranging from 1 (Not at all) to 7 (Very much). The four items were: “I was blaming myself for the trauma” (Self-Blame), “I was blaming someone else for the trauma” (Other-Blame), “I was thinking negatively about myself” (Negative Self-Attributions), and “I was thinking negatively about other people or the world” (Negative Other- and World-Attributions). Items were selected to reflect DSM-5 symptoms and mirror frequently used measures of trauma-related negative beliefs (e.g., PTCI; Foa et al., 1999). Negative thought was measured in two ways. First, we summed all four items to create an overall composite of negative thought. McDonald’s Omega was used to calculate reliability for this composite within-(ω_W = 0.66) and between-persons (ω_B = 0.70). Second, we examined individual subdomains to assess differences in types of negative thought (see Supplement Table 1 for descriptive statistics).

Perceived Situational Stress At each survey, we measured participants’ appraisals of stressful situations using the single item, “In the past 10 min, my situation was stressful,” rated on the same 7-point Likert scale described above. Psychometric work suggests that though internal reliability cannot be estimated with a single item, single items of emotional experience show concurrent and predictive validity comparable to their multi-item counterparts in experience sampling designs (Song et al., 2023).

We first assessed for associations between PTSD severity and demographic and clinical characteristics using bivariate correlations, independent samples t-tests, and analysis of variance (ANOVA). For main analyses, we examined mean levels (H1), variability (H2), inertia (H3), and reactivity (H4) in negative thought and their associations with PTSD severity (measured as total CAPS-5 score). Due to potential differences in subdomains of negative thought, we also conducted exploratory analyses by individually modeling Self-Blame, Other-Blame, Negative Self-Attributions, and Negative Other- and World-Attributions. Analyses were conducted in SPSS Version 28 and Mplus Version 8.7 using full-information maximum likelihood estimation with robust standard errors to handle missing data. This approach estimates parameters from the full sample with cases of incomplete data included in computations so that all available information is used in obtaining optimal parameter estimates (Arbuckle, 1996).

Variability in negative thought was computed using the person-specific SD¹ (iSD) of negative thought across all observations, yielding one estimated score for each participant. PTSD-related differences in variability were assessed using Pearson’s correlation coefficients. For all other dynamics (i.e., mean levels, inertia, reactivity), we used multilevel models because of the nested nature of the data (Level 1 observations within Level 2 participants). We initially sought to treat the negative thought composite and all subdomains continuously. Examination of descriptive statistics and assumptions of multilevel regression models for Self-Blame and Other-Blame showed unacceptably high levels of skewness and kurtosis and non-normal distribution of residuals and random effects. After unsuccessful attempts at variable transformations, we dichotomized these two subdomains to reflect absence versus presence of blame. Scores of 1 = Not at all were coded as absent (0) and scores ≥ 2, indicating at least some blame, were coded as present (1). Thus, models were analyzed using multilevel linear regressions for the negative thought and negative attribution subdomains and multilevel logistic regressions for the blame subdomains. All output scripts for the main and subdomain multilevel models analyzed in Mplus and described below are available on the Open Science Framework: https://​osf.​io/​mvde5/​?​view_​only=​157d068605e044e4​b09d0a67d056b89c​

Multilevel modeling accounts for data dependencies and allows for disaggregation of within- and between-subjects effects while handling missing data and time-varying intervals between observations (Nezlek, 2012; Wang & Maxwell, 2015). An uncorrelated residual covariance structure was used for all multilevel models. Mean levels of negative thought were assessed as a means-as-outcomes model with PTSD severity entered at L2. Inertia was assessed using the autocorrelation in a slopes-as-outcomes model by regressing current negative thought (outcome) on negative thought at the previous observation (predictor) at L1, and whether PTSD severity (L2) impacts this relationship (cross-level interaction). Reactivity was assessed by testing if perceived situational stress is associated with concurrent shifts in negative thought using a slopes-as-outcomes model. In this model, we regressed current negative thought on situational stress at time t (L1), and if PTSD severity (L2) impacts this relationship (cross-level interaction). Negative thought at the last observation (t-1) was entered as another L1 predictor to adjust for prior levels of negative thought. For all lagged models, data were analyzed within-day such that the final assessment on one day was not used to predict the first assessment on the next EMA day. In all models, categorical and continuous L1 predictors were person-mean-centered and estimated as random effects and L2 predictors were grand-mean-centered (Yaremych et al., 2023). Robustness checks adjusting for person-specific mean levels of negative thought in analyses of negative thought variability, inertia, and reactivity were conducted to examine if effects were an artifact of individual differences in mean levels of negative thought (Dejonckheere et al., 2019). Finally, following suggestions from reviewers, we conducted two additional sets of robustness checks to determine if any significant cross-level effects of PTSD withstood covariation for the presence of a Major Depressive Disorder diagnosis, specifically, and at least one psychiatric disorder, generally. All significant findings reported below remained unchanged after robustness checks adjusting for the presence of psychiatric disorders. Thus, we opted to focus those models that solely include PTSD severity, adjusted for person-specific mean negative thought throughout all formulas, text, and tables.

Demographic and clinical characteristics for this sample have been previously reported but are reproduced here (Pugach et al., 2023). Consistent with our stratified recruitment procedures, PTSD severity did not differ in age, gender, race/ethnicity, education, religion, and romantic relationship status (see Table 1). The index trauma type breakdown in the full sample was sexual assault (42.5%); physical assault (23.8%); natural disaster, accident, or fire (16.3%); serious illness, injury, or death (16.3%); and combat exposure (1.3%). PTSD severity was not associated with index event type (F[4,75] = 0.83, p = 0.508), but PTSD severity was higher among participants with at least one comorbid psychiatric diagnosis (p < 0.001). In total, participants completed 2158 surveys (M = 26.98 surveys per person), or 63.1% of all surveys received (SD = 19.7). PTSD severity was not associated with survey completion rate, r =  − 0.04, p = 0.699.²

There are several descriptive findings worth highlighting (see Supplement Table 1). First, at least some negative thought was endorsed at 684 of 2158 observations (31.7%), with Self-Blame least frequently endorsed (n = 127, 5.9%) and Negative Attributions—Self most endorsed (n = 507, 23.5%). Second, the intraclass correlation coefficient (ICC) value in an unconditional multilevel model quantifies the proportion of variance between-subjects, with 1-ICC reflecting the proportion within-subjects. The ICC value of 0.43 showed that 43% of the variance in negative thought is between-person and 57% is within-person, suggesting that both state and dispositional factors influence variability in negative thought. Third, visual inspection of momentary fluctuations in negative thought showed that while people with higher PTSD severity reported higher average levels of negative thought than those with lower PTSD severity, most people experienced considerable variability in negative thought in daily life (Fig. 1). Finally, within- and between-persons correlations (Supplement Table 2) show that people who experience more negative thought also experience more perceived situational stress, and that moments characterized by increased negative thought are characterized by more perceived situational stress.

We next compared whether PTSD severity was associated with mean levels of negative thought. We estimated a means-as-outcomes model with PTSD severity (total CAPS-5 score) at L2 predicting overall negative thought:

Level 1:

Level 2:

The L1 β_0j parameter represents each person’s average level of negative thought across observations. The γ₀₀ parameter reflects negative thought at the average level of PTSD severity and the γ₀₁ parameter represents the effect of deviations in PTSD severity on negative thought. Results showed a significant effect of PTSD severity on negative thought, b = 0.07, SE = 0.01, p < 0.001, 95% CI [0.04, 0.10], suggesting that those with higher PTSD severity reported higher average levels of momentary negative thought, consistent with H1. Subdomain analyses showed a similar effect of PTSD severity for Self-Blame, Other-Blame, Negative Self-Attributions, and Negative Other- and World-Attributions (see Supplement Table 3).

Negative Thought Variability (H2) We used a Pearson’s correlation to examine the linear relationship between PTSD severity and variability (i.e., iSD) of negative thought. There was a significant positive association between PTSD and negative thought variability, r = 0.47, p < 0.001, 95% CI [0.28, 0.63], consistent with H2. After adjusting for mean levels of negative thought using a partial correlation, the association between PTSD severity and variability in negative thought was reduced but still significant, r = 0.24, p = 0.030. Exploratory analyses by subdomain showed a similar pattern: PTSD severity was positively associated with iSD for all subdomains (ps =  < 0.001); however, only the relationship between the iSD of Negative Self-Attributions and PTSD severity were significant after adjusting for mean levels of negative thought (Supplement Table 4).

Level 1:

Level 2:

At L1, the β₁ parameter represents the autocorrelation of negative thought over successive observations at each person’s mean level of negative thought. This value conceptually represents inertia, or how self-perpetuating negative thought is across time. The γ₁₀ parameter at L2 reflects the slope of the autocorrelation and the γ₁₁ parameter represents the effect of PTSD severity on inertia. All other values are as described in the means-as-outcomes model (H1) above. Results show that negative thought inertia was significant across observations, b = 0.20, SE = 0.04, p < 0.001, 95% CI [0.12, 0.28], see Table 2. Inconsistent with H3, PTSD severity did not predict negative thought inertia, p = 0.432 (p = 0.666 in unadjusted model). Exploratory analyses showed an identical pattern of findings for all negative thought subdomains. An inertia effect was present for Self-Blame, Negative Self-Attributions, and Negative Other- and World-Attributions, but not for Other-Blame. No cross-level interactions effects were significant (see Supplement Table 5).

Level 1:

Level 2:

At L1, the β₂ parameter represents how much perceived situational stress is associated with concurrent shifts in negative thought, adjusted for negative thought at the previous observation. The γ₂₀ parameter represents the slope of reactivity and the γ₂₁ parameter represents the effect of PTSD severity on reactivity. All other parameters are consistent with the inertia model above. Results show that after adjusting for previous levels of negative thought and person-specific mean levels of negative thought, there was a significant effect of perceived situational stress on concurrent negative thought, b = 0.54, SE = 0.08, p < 0.001, 95% CI [0.38, 0.70], see Table 2. The effect of PTSD severity was significant, b = 0.02, SE = 0.01, p = 0.031, 95% CI [0.002, 0.04] such that those with higher PTSD severity experienced greater shifts in negative thought associated with perceived stressful situations, consistent with H4 (unadjusted model p = 0.001, 95% CI [0.012, 0.046]). Exploratory analyses showed a similar pattern of findings for reactivity across subdomains. The effect of PTSD severity was significant in all unadjusted models (ps range from < 0.001 to 0.015), but analyses adjusted for person-mean levels showed that the cross-level interaction of PTSD severity was only significant for Negative Self-Attributions (p = 0.029, Supplement Table 6).

Consistent with our predictions, trauma-exposed adults reported marked fluctuation in their endorsement of negative thought on an hour-to-hour basis (Fig. 1). Variability in these fluctuations was greater among individuals with higher PTSD severity, even after adjusting for mean levels, suggesting that people with higher PTSD severity experience more frequent and intense moments of negative thought compared to individuals with less severe PTSD. This finding is consistent with current symptoms descriptions in DSM-5-TR (APA, 2022) and with research showing that variability in other cognitive processes (i.e., rumination) is associated with psychopathology (Bean & Ciesla, 2024; Kiekens et al., 2024). At the same time, findings suggest that despite higher mean levels of negative thought, people with higher PTSD severity have many moments when they are not experiencing negative thought like blame or negative attributions. Well-timed cognitively focused interventions could potentially leverage moments when patients are not blaming or thinking negatively about themselves, others, or the world to identify more adaptive patterns of thinking.

That people with higher PTSD severity display heightened negative thought reactivity provides some preliminary context for how negative thought unfolds in daily life. Extensive theory on cognitive vulnerability—the idea that negative thoughts emerge in stressful situations—has shown that cognitive reactivity characterizes people at risk for depression and precipitates its onset and recurrence (Beck, 1967). Here, we found that people with higher PTSD severity reported heightened negative thought during moments of perceived situational stress, even after adjusting for mean levels of negative thought, suggesting that cognitive reactivity is a distinct feature of PTSD. Findings align with cognitive theory of PTSD (Ehlers & Clark, 2000) and idiographic network analysis research showing that negative thought is a prominent maintenance factor for PTSD across people (Bridges-Curry, 2025; Reeves & Fisher, 2020). That is, heightened cognitive reactivity could reflect misperceptions of ongoing threat in PTSD and trouble processing disconfirming information. In turn, negative thoughts could become more easily reinforced and contribute to symptom chronicity. Other studies examining the temporal structure of PTSD have found that hyperarousal symptoms (e.g., startle response) are most central to maintaining PTSD, however (Greene, 2021), suggesting that there is heterogeneity across people in how PTSD is maintained.

Contrary to our hypothesis, we did not find that negative thought inertia varied based on PTSD severity. This stands in contrast to recent work finding that ruminative inertia is elevated in individuals with depression and anxiety (Bean & Ciesla, 2024; Bean et al., 2020, 2021), and extensive research documenting a robust association between rumination and PTSD (Moulds et al., 2020). We did find that negative thought was inert on average, suggesting that negative thought generally persists for trauma-exposed adults over time. This could be taken to mean that negative thought inertia is not a unique feature of PTSD severity, but rather that once negative thought begins, it is generally difficult for most people to dismiss. Further, although inertia may not necessarily be stronger in those with more severe PTSD, it may be more problematic because negative thought occurs more frequently in this population.

In sum, findings suggest that heightened PTSD severity is characterized by negative thought patterns that are less stable and more readily activated by perceived stress, but not more resistant to change over time. Conceptually, this pattern suggests that individuals with more severe PTSD experience larger and more extreme activation of negative posttraumatic thought, particularly during stressful moments, with negative thought exhibiting comparable return to baseline regardless of PTSD severity. Our findings have implications for future studies on the cognitive dynamics of PTSD and the development and adaptation of interventions for PTSD.

That research on negative cognition in PTSD is over-represented by static, between-subjects, and cross-sectional assessments raises questions about how these processes unfold within people over time. Our results show that when negative thought is repeatedly probed, trauma-exposed individuals vary considerably in how much they espouse negative thought at different moments in time. The tendency for negative thought to vary occurs on relatively short timescales, shown here on an hour-to-hour basis. This variability is driven both by state and dispositional factors and isolating within-subjects effects provided unique information over and above person-specific averages. The continued adoption of ambulatory methods that repeatedly track people over time and statistical approaches that disaggregate within- and between-subjects processes will help advance our understanding of cognitive processes involved in PTSD. Our findings produced some recommendations for this future research.

Our hypotheses were informed by existing research on the cognitive dynamics of depression. Research in this area has often adopted daily measures of rumination adapted from the Ruminative Responses Subscale of the Response Styles Questionnaire (Nolen-Hoeksema et al., 2008). These questions tap the tendency to ruminate, including thoughts about causes, symptoms, and reactions to negative mood. By contrast, our assessment of negative thought focused on specific content domains prevalent in PTSD (blame, negative attributions; Foa et al., 1999). Despite the conceptual overlap between rumination and negative thought, they are distinct in operationalization and measurement, which may add context to our findings. Rumination is a process that exacerbates and prolongs depression throughout a variety of mechanisms, one of which is by enhancing negative thought (Nolen-Hoeksema et al., 2008). Rumination might “cut across” different types of negative thought to support a proclivity toward maladaptive thinking, particularly among individuals inclined towards negative mood. Assessments of rumination, particularly over longer intervals (e.g., day-to-day; Bean et al., 2020, 2021), could better reflect the perseverative process thought to keep individuals “cognitively stuck” (Koval et al., 2012). By contrast, our dense sampling schedule showed that momentary endorsement of negative thought content was infrequent, especially for Self-Blame and Other-Blame, which were only present on 5.9% and 7.1% of occasions, respectively. The low frequency of negative thought over this period may have contributed to difficulty detecting the cognitive dynamics of interest. Likewise, this sampling frame may have also introduced challenges for model estimation. Although we took steps to address low frequency of endorsement in extreme cases (e.g., dichotomizing the blame variables and using logistic regression), findings suggest that the cognitive dynamics of negative thought such as variability and inertia might be better captured over less frequent intervals (e.g., multi-hour, day-to-day). Future research measuring both rumination and negative thought in daily life will help clarify the relationship between these cognitive processes.

Finally, the negative thought dynamics examined here have transdiagnostic implications. Findings join a growing body of research showing that variability, inertia, and reactivity in cognitive processes are associated with depression, anxiety, and NSSI (Bean & Ciesla, 2024; Bean et al., 2020, 2021; Cole et al., 2021; Kiekens et al., 2024). Though the focus of our study was on trauma-exposed adults varying in PTSD severity, most people with PTSD reported at least one current comorbid disorder, consistent with national incidence data (96.5% PTSD comorbidity rate with at least one other disorder; Gadermann et al., 2012). Notably, results held after adjusting for MDD comorbidity, specifically, and psychiatric comorbidity, generally. These findings raise two possibilities. First, whereas previous research on cognitive dynamics has largely focused on rumination in depressed samples, the dynamics of negative posttraumatic thought assessed here could be specific to PTSD. Negative cognitive processes such as negative thought are central cognitive features of PTSD (Ehlers & Clark, 2000). Coupled with the high comorbidity shared with other disorders observed following trauma exposure, it is possible that PTSD is the primary driver of these dynamics in trauma exposed individuals. In other words, the cognitive dynamics of negative thought could be due to shared comorbidity. Second, these cognitive dynamics may be features of and share similar vulnerabilities for elevated scores on broader dimensions that cut across PTSD and other psychiatric disorders (e.g., fear, distress), consistent with transdiagnostic conceptualizations of psychopathology (Kotov et al., 2021). A transdiagnostic perspective would help explain emerging research documenting dynamic features of cognition that characterize depression, anxiety, NSSI, and now PTSD. Unfortunately, our tests for covariation used a categorical measure of psychiatric disorder diagnostic comorbidity that may be underpowered compared to the dimensional measure of PTSD, leaving us unable to rule out method effects. Future work in trauma exposed samples would benefit from including transdiagnostic samples and assessment tools to better understand if the cognitive dynamics of negative posttraumatic thought serve as a transdiagnostic or disorder-specific marker of PTSD. Additionally, research should also consider longitudinal designs that can speak to whether cognitive dynamics serve as markers for psychopathology or confer risk for its development and maintenance following trauma exposure.

Findings suggest that standard delivery of cognitive interventions for negative thought (e.g., weekly outpatient treatment with little-to-no between-session support; Schleider et al., 2021) might be suboptimal for the timescale on which it varies in daily life. Alternative formats that increase the frequency of cognitive interventions could better match how negative thoughts fluctuate in daily life. Optimizing the timing of intervention delivery could help patients gain insight into their own cognitive processes (e.g., when and in what contexts negative thought typically occurs) and learn to use therapy skills when needed most. Massed delivery formats (e.g., massed CPT), wherein interventions are delivered at an accelerated pace of multiple times per day over consecutive days, have been shown to reduce PTSD symptoms with large effects comparable to standard delivery methods and hold promise for improving treatment access, efficiency, and retention (Galovski et al., 2022; Held et al., 2022). Integrating ambulatory assessments into massed treatment opens the door for intriguing questions, such as whether changes in the temporal dynamics of negative thought track or precede symptom improvement. For example, negative thoughts may become less variable, “sticky,” and reactive as patients learn to skillfully challenge them in therapy. In addition to more frequent therapist intervention, just-in-time adaptive interventions (JITAI) also hold promise in reducing negative thought by providing tailored, technology-delivered support in the moment as needed (Nahum-Shani et al., 2018). JITAIs may be self-initiated by the user during high-stress moments or automatically triggered via passive sensing (e.g., ambulatory physiological assessment; Wisco et al., 2024 under review) to promote skills use in real-time. JITAIs may be used in conjunction with standard psychotherapy or as stand-alone interventions capable of increasing treatment efficacy and global accessibility to treatment, and several mobile health applications already exist to assist in the treatment of PTSD (Rodriguez-Paras et al., 2017).

Finally, subdomain analyses highlight the importance of examining and treating specific negative thought types. Though the temporal dynamics identified here were generally consistent across subdomains, subdomains differed in in their frequency of endorsement. Endorsement of Negative Self-Attributions and Negative Other- and World-Attributions were most common (23.5% and 19.7% of observations, respectively), whereas Self-Blame and Other-Blame were endorsed infrequently. Cognitive dynamics for Negative Self-Attributions were most robust, as this was the only subdomain to survive adjustment for person-mean levels of negative thought for variability and reactivity. Findings are particularly relevant for treatment studies seeking to identify within-person mechanisms of change (Alpert et al., 2023), given discrepant findings regarding which subdomains are altered in treatment (Dillon et al., 2020; Holliday et al., 2018; Schumm et al., 2015). Findings suggest that targeting Negative Self-Attributions may be particularly important given its frequency and role in demarcating PTSD severity. However, given potential person-to-person differences in the subdomains of negative thought endorsed, personalized approaches that identify which thoughts are most relevant for a given person and how these thoughts change in response to trauma-focused treatment also hold great value (Brown et al., 2019).

There are a few limitations of our study that can inform future work. First, though the racial/ethnic diversity of this sample is a strength, this was largely a community sample of young females. Future work will need to examine if results generalize to other sociodemographic groups and settings where the prevalence and domains of negative thoughts might differ (e.g., in treatment seeking individuals or individuals with primary diagnoses other than PTSD). Second, in terms of design, lack of personalization around our assessment interval (~ 9 am to 11:30 pm) precludes capturing the experiences of those who wake up and go to bed either earlier or later. Future designs might customize the sampling period to each participant’s schedule to avoid this issue, in addition to employing a sampling schedule that is better suited to capture negative thought in daily life (i.e., a longer sampling period within less frequent within-day assessment).

Third, our measures of self-reported situational stress and negative thoughts left us unable to parse the effects of objective stress exposure from an individual’s subjective appraisals. In other words, perceptions of situational stress could be influenced by negative thought. The confounding of stressful stimuli and response is prominent in the stress literature. One possible solution follows the use of gold-standard life stress interviews (e.g., Hammen et al., 1987), which use multiple independent raters blind to participant diagnoses and responses to rate episodic life stress. Interview methods yield a measure of inter-rater reliability while capturing idiosyncratic stress appraisals of each rater. Application of this method to ambulatory designs is limited by inefficiency and impracticality, though some have proposed feasible solutions for integration (Harkness & Monroe, 2016). Another solution consistent with recent recommendations (Wright et al., 2019) is to incorporate passive sensing methods to objectively monitor and capture changes in one’s environment (e.g., GPS location, voice activity) or stress response (e.g., accelerometer or physiological sensing, text messages). Developing plausible and scalable measures of stress for ambulatory methods that enable researchers to parse distinct components of stressors and stress responses will be an important goal for future research.

Finally, another limitation is the contemporaneous method we used to assess negative thought reactivity. We operationalized reactivity as the within-subjects associations between concurrent perceived situational stress and negative thought, adjusted for negative thought in the previous survey. This precluded us from establishing a precise temporal relationship between perceived stress and negative thought. At minimum, we can be confident that results establish hour-to-hour covariation between perceived situational stress and negative thought, and that this association exists beyond the tendency for negative thought to persist from one moment to the next. Ambulatory designs examining the temporal features of negative thought will need to carefully measure contextual characteristics to clarify their influence on negative thought over time.