Decentering Predicts Attenuated Perseverative Thought and Internalizing Symptoms Following Stress Exposure: A Multi-Level, Multi-Wave Study

Abstract

While research identifies a growing list of risk factors for anxiety and depression, it is equally important to identify potential protective factors that may prevent or reduce vulnerability to developing internalizing psychopathology. We hypothesized that forms of perseverative thinking, such as rumination and worry, act as mechanisms linking negative life experiences and prospective symptoms of anxiety and depression. More specifically, we investigated whether decentering, the meta-cognitive capacity to adopt a distanced perspective toward one’s thoughts and feelings, serves as a protective factor at various points along this mediational pathway. A sample of 181 undergraduate students were recruited and assessed at five time points over a 12-week period. Multilevel modeling indicated that decentering was associated with an attenuated impact of (1) negative events on prospective depressive symptoms; (2) negative events on prospective brooding, and (3) brooding, pondering and worry on prospective internalizing symptoms. Multilevel moderated mediation analyses provided partial support for the hypothesis that perseverative thinking would mediate the longitudinal associations between negative life events and internalizing symptoms, with decentering attenuating risk at several connections of the indirect pathways. The strongest support was provided for moderated mediation models in which decentering was associated with attenuated relationships between negative events, brooding, and symptoms of depression. This study is the first to elucidate the role of decentering as a protective factor against anxiety and depressive symptoms at different points in the path from stress to perseverative thought to internalizing symptoms. Decentering therefore may be a critical target for clinical intervention to promote resilience against anxiety and depression.

Depression and anxiety are the most prevalent forms of psychopathology, with high rates of comorbidity between the two conditions (Brown et al., 2001; Kessler et al., 2005). A well-established risk factor for the onset of both anxiety and depressive disorders is exposure to negative life events (Faravelli et al., 2012; Finlay-Jones & Brown, 1981; Kendler et al., 2003; McLaughlin et al., 2010). In light of this robust association, a substantial body of research has investigated various cognitive, affective, and neurophysiological mechanisms that might explain this relationship (Eisenberger & Cole, 2012; Hankin, 2010; McLaughlin & Hatzenbuehler, 2009). One potential cognitive mechanism that might account for the stress-psychopathology link is perseverative thinking (PT), which may exacerbate and prolong the physiological and psychological effects of stress (Hamilton & Alloy, 2016; Mennin & Fresco, 2013b; Ottaviani et al., 2016; Stange, Alloy, et al., 2017). Nevertheless, further research is needed to probe the specific pathways by which stress might lead to perseverative thinking and internalizing symptoms, and to elucidate any protective factors that might reduce its impact. This knowledge could then help to identify modifiable targets for prevention and treatment.

Perseverative thinking is a transdiagnostic process that has been demonstrated to contribute to the development of anxiety and depressive disorders (Ehring & Watkins, 2008; Nolen-Hoeksema & Watkins, 2011). PT is characterized by repetitive, negative patterns of thinking. Negative or stressful experiences can represent a subjective failure to progress towards a goal, thus creating a discrepancy between one’s desired state and one’s current state (Martin & Tesser, 1996; Watkins & Nolen-Hoeksema, 2014). In this conceptualization, PT can occur in response to negative events as an attempt to resolve such discrepancies (Watkins & Nolen-Hoeksema, 2014a), particularly when experienced in the presence of negative affective states (Chan et al., 2013; Hawksley & Davey, 2010; Meeten & Davey, 2011). One form of PT is worry, a negative and uncontrollable chain of thoughts and images concerning future events whose outcome is uncertain (Borkovec et al., 1983). Worry is a defining feature of generalized anxiety disorder (GAD; American Psychiatric Association, 2013), and has been shown to predict increases in anxiety and depressive symptom severity in adults (Hong, 2007; Ryum et al., 2017) and adolescents (Young & Dietrich, 2015). Another form of PT is rumination, which involves negative and repetitive patterns of thinking concerning the causes and consequences of one’s sad mood (Nolen-Hoeksema, 1991). Two primary components of rumination have been identified: brooding, involving more negative and perseverative content, and pondering, a potentially less maladaptive form of PT that involves reflection on past events with curiosity (Treynor et al., 2003). Individuals who engage in rumination tend to be at greater risk for depressive symptoms and episodes (Abela & Hankin, 2011; Alloy et al., 2017; Nolen-Hoeksema et al., 2008; Stange et al., 2014, 2016). Ruminative responses also prospectively predict elevated symptoms of anxiety in adults and adolescents (McLaughlin & Nolen-Hoeksema, 2011; Roelofs et al., 2009). Furthermore, a number of prospective studies implicate rumination as a mediator of the association between negative life events and internalizing symptoms in both adults and adolescents (Hamilton et al., 2015; Michl et al., 2013; Young & Dietrich, 2015), whereas the role of worry as a mediator in this link requires further investigation (Olatunji et al., 2010).

Prior work examining PT as a mediator of stress-internalizing symptom relationships mostly has measured these questions at the between-subject level – that is, examining whether trait levels of PT explain why individuals with more stress (relative to others) experience more symptoms. However, if constructs are causally related, they should not only be related across individuals, but also within individuals over time (Stange et al., 2019). That is, fluctuations in the antecedent (e.g., negative life events) should correspond with fluctuations in the outcome – when individuals experience more stress than usual, they should experience increases in PT and symptoms. Few studies to our knowledge have examined whether within-subject increases in stress correspond with future symptoms via increased PT. If supported, this would provide further evidence consistent with a causal relationship between these constructs¹.

Given that all individuals periodically experience times of increased stress from negative experiences, it is important to identify protective factors that may reduce the likelihood of experiencing PT and internalizing symptoms during these periods. Such protective factors may have important implications for the prevention of depression and anxiety. One such factor that has relevance to PT and internalizing disorders is decentering, the meta-cognitive tendency to take a distanced perspective toward one’s internal experiences. Decentering involves observing one’s thoughts and feelings as temporary, objective events as opposed to reflections of the self or reality that are necessarily true (Fresco et al., 2007a; Safran & Segal, 1996). Decentering is inversely associated with a range of psychological symptoms, including anxiety and depression (Bernstein et al., 2015, 2019; Fresco, Moore, et al., 2007a; Stange, Hamilton, Fresco, et al., 2017). Decentering also is linked with less severe responses to stress (Feldman et al., 2010a), and with reduced engagement in forms of self-referential processing like PT (Banerjee et al., 2018; Lebois et al., 2015; Mennin & Fresco, 2013a). According to the metacognitive processes model, decentering is composed of three interrelated processes – meta-awareness, disidentification from internal experience, and subsequently reduced reactivity to thought content (Bernstein et al., 2015). Thus, decentering may assist in disengaging from PT by first bringing awareness to one’s worrying or ruminating, then allowing negative thoughts and emotions associated with PT (e.g. “I am afraid”) to be perceived as separate from one’s self (e.g. “I have the feeling of being afraid”), leading to reductions in the tendency for negative thought content to worsen negative affect. Neuroimaging studies have shown decentering and self-referential processing to be associated with the same underlying neural networks (King & Fresco, 2019; Scult et al., 2019), adding a neurobehavioral account of how decentering may assist in reducing perseverative thinking and its associated distressing symptoms. Findings from clinical trials reveal that decentering is a mechanism of symptom change in both cognitive-behavioral therapy and mindfulness- and acceptance-based treatments, with improvements in decentering predicting reductions in anxiety and depressive symptoms, worry and rumination (Bieling et al., 2012; Farb et al., 2018; Fresco et al., 2007b; Gu et al., 2015; Hayes-Skelton et al., 2015; Hoge et al., 2015; Mennin et al., 2018; O'Toole et al., 2019; in press). Mounting evidence indicates that decentering represents a modifiable intervention target, and promotion of decentering also may contribute to the prevention of perseverative thinking, as well as internalizing symptoms and disorders. However, exactly how and at what point decentering protects against the impact of stress on internalizing symptoms requires further investigation. For example, decentering might reduce the likelihood of prolonged engagement in PT after negative events. Alternatively, decentering might reduce the likelihood of PT persisting to the extent that it leads to the development of depression or anxiety.

The primary aim of the current study was to investigate whether decentering is associated with muted associations between two putative risk factors (negative life events and perseverative thinking) and internalizing symptom severity over time, within a non-clinical sample of young adults. We used a person-centered approach to test whether comparatively higher levels of negative life events would predict prospective anxiety and depressive symptom severity, with increases in perseverative thinking explaining this relationship. Broadly, we hypothesized that when individuals experienced more negative life events than their usual amount, they would report subsequent increases in worry and rumination, as well as higher levels of anxiety and depressive symptoms, and that individuals with greater decentering would be protected from these effects (see Figure 1). This hypothesis is rooted in the diathesis-stress component of cognitive theories of vulnerability to depression, which proposes that vulnerable individuals are expected to experience increases in depression when they encounter more stress than usual (i.e., proximal increases in stress) (Abela & Hankin, 2008).

Figure 1.

Conceptual figure representing hypothesized relationships between negative life events, perseverative thinking, and internalizing symptoms, with decentering attenuating the strength of each of the three links.

More specifically, we hypothesized that individuals with a greater tendency to decenter would experience a weaker relationship between increases in stress and symptoms of depression and anxiety (Hypothesis 1), a weaker relationship between increases in stress and subsequent PT (brooding, pondering, and worry; Hypothesis 2), and a weaker relationship between increases in PT and subsequent symptoms of depression and anxiety (Hypothesis 3). This design enabled us to examine when (at which point in the pathway from stress to PT to symptoms) decentering may have greatest utility. Finally, in the event that the individual links in Hypotheses 1–3 (see Figure 1) were supported, we examined whether a mediation pathway from stress to PT to symptoms would be weaker among individuals with greater decentering, either via mitigated responses to stress in terms of PT (Hypothesis 4), or via mitigated responses to PT in terms of symptoms (Hypothesis 5).

Method

Participants and procedure

Participants were undergraduate students at a large urban university, recruited using flyers on campus and from undergraduate psychology classes. To be included in the study, participants were required to have normal or corrected-to-normal vision and be fluent in English. Students received psychology course credit or were compensated in cash for participation. All participants provided written informed consent approved by the University’s Institutional Review Board. The sample included 181 participants (56.7% female, 66.2% Caucasian/White, 19.1% African American/Black, 13.3% Asian/Pacific Islander, 0.6% Native American, 5.1% other race, and 8.3% Hispanic/Latino), with a mean age of 21.98 years (SD = 5.90). This report tests a new set of research questions using data drawn from a study reported elsewhere (Stange, Hamilton, Fresco, et al., 2017).

Participants completed a set of self-report questionnaires at baseline, and at four follow-up assessments (Times 2–5) spaced three weeks apart. This time frame was chosen to allow for sufficient variability in negative life events, while enabling modelling of short-term fluctuations in internalizing symptoms (Stange, Hamilton, Olino, et al., 2017). All questionnaires used in the present analyses were administered at baseline. During the follow-up assessments, participants completed measures of rumination, worry, anxiety and depressive symptoms, and exposure to negative life events in the prior three weeks. At Time 5, participants were interviewed to verify that life events reported met a priori criteria and occurred within the correct three-week interval.

Participants were required to have completed at least one of the four follow-up assessments to be included in the present analyses, which yielded a final sample size of 157. Participants included in the analyses had higher levels of brooding (p = .04, d = .47) than participants who were excluded from analyses (n = 24) but did not differ on any other study variables or demographic characteristics (ps > .09, ds = .01–.37).

Measures

Life Events Scale (LES) and Interview (LEI).

The LES (Alloy & Clements, 1992) is composed of 134 items, representing major, moderate, and minor life events in a range of domains relevant to college students (e.g., school, finances, family, social and romantic relationships). Participants were asked to indicate whether or not they had experienced each of the events in the previous three weeks. To address potential reporting biases, the LEI is administered by an interviewer, who evaluates the validity of responses in the LES according to a priori criteria for event definition. Reported events are excluded if they do not meet the event definition criteria or if they did not occur within the relevant interval. For example, for the event “did poorly on or failed an exam or major project in an important class (i.e., grade less than C),” the LEI probes require that the exam or project be worth at least 20% of the final grade to be classified as major. Interviewers also must verify the timing of events by anchoring events to time of week, year, and time during the semester or school break, as well as in relation to the timing of other major events. Interviews were completed by phone or in person by a trained graduate student. During the LEI in the present study, 4.2% of events endorsed on the LES were disqualified; 0.2% of the final events used in analyses were added by the interviewer; 0.1% of the final events used had been reclassified by the interviewer. The LES and LEI have demonstrated excellent reliability and predictive validity (Alloy & Clements, 1992; Stange, Hamilton, Olino, et al., 2017). The internal consistency of the LES was good (αs = .84–.91 at Times 2–5) in the current study.

Beck Depression Inventory (BDI; Beck et al., 1996).

The BDI-II is a widely used self-report measure of depressive symptom severity. It includes 21 items probing cognitive, affective, and somatic symptoms of depression during the previous three weeks. The BDI-II has demonstrated good internal consistency and validity in undergraduate samples (Storch et al., 2004; Dozois et al., 1998). In the present study, αs = .87 - .90 at Times 1–5.

Beck Anxiety Inventory (BAI; Beck et al., 1988).

Anxiety symptom severity during the previous three weeks was measured by the BAI, a 21-item self-report inventory. The BAI has been found to have good validity and internal consistency (Creamer et al., 1995; Fydrich et al., 1992). In the present study, αs = .91 - .93 at Times 1–5.

Ruminative Response Scale (RRS; Treynor, Gonzalez & Nolen-Hoeksema, 2003).

The RRS is a 10-item self-report measure that assesses the extent of engagement in two components of rumination: brooding and reflective pondering. The measure consists of five brooding items (e.g., “think about a recent situation, wishing it had gone better”) and five reflection items (e.g., “analyze recent events to try to understand why you are depressed”), which are scored on a Likert scale ranging from 1 (almost never) to 4 (almost always). It has excellent internal consistency and validity (Armey et al., 2009; Treynor et al., 2003). In the present study, α = .74 for brooding, and α = .76 for reflective pondering at Time 1. At Times 2–5, the RRS contained the same items with instructions modified to assess ruminative responses over the previous three weeks (at Times 2–5, αs = .74–.84 for brooding, and αs = .77–.87 for pondering).

Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger & Borkovec, 1990).

The PSWQ is a 16-item measure of trait worry that captures the generality, excessiveness, and uncontrollability characteristics of pathological worry (e.g., “my worries overwhelm me” and “I worry all the time”). Items are rated on a 1 (not at all typical of me) to 5 (very typical of me) Likert scale. Among samples of undergraduates, the PSWQ repeatedly has demonstrated good internal consistency and good test-retest reliability over intervals as long as ten weeks (Meyer et al., 1990; Fresco et al., 2002). In this sample, the PSWQ demonstrated excellent internal consistency (α = .95) at Time 1. For the follow-up assessments at Times 2–5, the PSWQ contained the same items with instructions modified to probe worrying over the previous three weeks (αs = .93–.94 at Times 2–5).

Experiences Questionnaire (EQ; Fresco et al., 2007a).

The decentering subscale of the EQ is an 11-item self-report measure of trait-like decentering. It assesses an individual’s tendency to mentally distance oneself from one’s thoughts and feelings (e.g., “I can observe unpleasant feelings without being drawn into them”). Participants respond to each item on a Likert scale ranging from 1 (never) to 5 (all the time). Higher scores indicate greater levels of decentering. The subscale has been shown to have good validity and internal consistency (Bernstein et al., 2015; Fresco et al., 2007a, b; Teasdale et al., 2002). In the current study, α = .88.

Statistical Analyses

Multilevel modeling (MLM) (Raudenbush & Bryk, 2002) was used to test the primary hypotheses, given the nested structure of the data (multiple observations of negative life events, real-world PT, and internalizing symptoms within each person). MLM is advantageous in terms of maximizing data usage because it can flexibly handle cases with missing data, so participants with missing data (e.g., participants who miss a follow-up visit) are not eliminated from the data analyses. Analyses were conducted with the Mplus 6.12 statistical software package (Muthén & Muthén, 2010), which allowed for use of full information maximum likelihood (FIML) estimation of data, using maximum likelihood estimation with robust standard errors.

To test the hypothesis that higher levels of life stress (compared to one’s usual levels) would predict higher levels of weekly internalizing symptoms (of depression and anxiety), particularly for individuals with lower levels of decentering (Hypothesis 1), negative life events (which were person-centered at Level 1) served as the focal predictor of weekly internalizing symptoms. Idiographic (person-centered) approaches such as this may test theoretical vulnerability-stress models more accurately than is possible with nomothetic (sample mean-centered) approaches (Abela & Hankin, 2008; Stange et al., 2019). Because responses to life events are likely to occur in close proximity to the events themselves, we tested associations that were temporally concurrent (i.e., life events over the last three weeks predicted PT and symptoms that were measured at the same time points). A random intercept and random slope were included in each model to allow individuals to vary in the relationship between life events and internalizing symptoms. Decentering (which was standardized between participants at Level 2) served as a between-subject moderator of these slopes, and as a predictor of symptoms (i.e., intercept). Covariates at Level 2 included baseline levels of symptoms and each participant’s mean level of negative life events. Significant interactions were probed by testing the simple slopes of life events on internalizing symptoms at ± 1 standard deviation from the mean of decentering.

Next, MLM also was used to test the hypothesis that higher levels of life stress (compared to usual levels for oneself) would predict higher levels of weekly PT (brooding, pondering, and worry) particularly for individuals with lower levels of decentering (Hypothesis 2). Negative life events (person-centered at Level 1) served as the focal predictor of weekly PT (either brooding, pondering, or worry). A random intercept and slope were included in each model as an estimate of the relationship between life events and PT for each person. Decentering served as a between-subject moderator of these slopes, and as a predictor of random intercepts of PT. Covariates at Level 2 included trait levels of PT and each participant’s mean level of negative life events (which were standardized between participants at Level 2).

To test the hypothesis that increases in PT (compared to usual for oneself) would predict higher levels of future weekly internalizing symptoms (of depression and anxiety), particularly for individuals with lower levels of decentering (Hypothesis 3), PT (which was person-centered at Level 1) served as the focal predictor of weekly internalizing symptoms. A random intercept and slope were included in each model as an estimate of the relationship between PT and internalizing symptoms for each person. Decentering served as a between-subject moderator of these slopes, and as a predictor of random intercepts of symptoms. Covariates at Level 2 included baseline levels of symptoms and each participant’s level of trait PT.

In the event that the individual links in Hypotheses 1–3 (Figure 1) were supported, we planned to conduct multilevel moderated mediation tests, to examine whether negative life events would lead to internalizing symptoms via PT, with decentering moderating these indirect pathways either via the link between life events and PT (Hypothesis 4) or the link between PT and symptoms (Hypothesis 5). First, multilevel mediation models were constructed (with Level 1 variables), modeling paths between life events, perseverative thinking, and internalizing symptoms, with indirect effects (with bootstrapped confidence intervals) representing the strength of these relationships at the within-subject level. Next, in each model, decentering was added as a moderator of one of these links (based on the results of Hypotheses 1–3). Indirect effects then were computed with decentering set to +/− 1 SD from the sample mean (Aiken & West, 1991), so that the indirect effects represented the within-subject pathway from stress to perseverative thought to internalizing symptoms, for individuals at high and low levels of decentering (Muthén & Muthén, 2010).

Finally, we employed false discovery rate (FDR) correction for the interactions tested in each of Hypotheses 1–5, and we indicate which analyses survived this FDR correction (Benjamini & Hochberg, 1995).

Results

Including the baseline assessment, a total of 570 observations were completed (Mean = 4.63, SD = 1.12, Range = 2–5). The mean number of life events across participants and waves was 8.05 per occasion (SD = 6.43); the mean BDI score across participants and waves was 7.24 (SD = 6.52); at baseline, 24% of the sample had BDI scores suggesting at least mild depression (BDI > 13), and 11% had at least moderate depressive symptoms (BDI > 20).

Person-centered variables were centered so that each person’s mean was 0; the standard deviations were as follows: negative life events (SD = 3.81), brooding (SD = 2.25), pondering (SD = 2.43), and worry (SD = 8.55). Components of PT were correlated as follows: for trait variables, brooding was correlated with pondering (r = .49, p < .001) and worry (r = .61, p < .001), and pondering was correlated with worry (r = .42, p < .001); for person-centered variables, brooding was correlated with pondering (r = .67, p < .001) and worry (r = .51, p < .001), and pondering was correlated with worry (r = .47, p < .001).

The intra-class correlation (ICC) for an empty model predicting BDI was .67, indicating that 67% of the variance in depressive symptoms occurred at the between-subject level (Level 2), whereas 33% of the variance occurred at the within-subject level (Level 1). The ICCs for the other level-1 variables were as follows: negative life events (.76), BAI (.65), brooding (.53), pondering (.51), worry (.46).

There were no racial or ethnic differences in study variables (ps > .06). Relative to men, women reported higher levels of negative life events (t = 2.44, p = .02, d = 0.39), brooding (t = 2.51, p = .01, d = 0.40), worry (t = 4.02, p < .001, d = 0.65), and symptoms of depression (t = 2.65, p = .01, d = 0.41) and anxiety (t = 3.73, p < .001, d = 0.58), but they did not differ on decentering (t = 1.45, p = .15, d = 0.23) or pondering (t = 1.45, p = .14, d = 0.24).

Hypothesis 1: Does Decentering Predict an Attenuated Relationship between Negative Life Events and Internalizing Symptoms?²

As hypothesized, there was a significant interaction between decentering and negative life events predicting symptoms of depression (Table 1), such that negative life events predicted symptoms of depression more strongly among individuals with low decentering (B = 0.63, SE = 0.09, p < .001) than among individuals with high decentering (B = 0.33, SE = 0.09, p < .001; Figure 2a). In contrast, the interaction between decentering and negative life events predicting symptoms of anxiety was not significant, although there was a significant positive main effect of life events on anxiety (Table 1). Of the two analyses conducted for this hypothesis, the decentering × life events interaction predicting symptoms of depression remained significant after false discovery rate (FDR) correction (threshold q = .025).

Table 1.

Multilevel Models with Decentering Moderating Relationships between Person-Centered Negative Life Event Exposure and Internalizing Symptoms at Each Wave

Predicting Depression Symptoms				.71
BDI (Intercept)	0.48	0.57	.40
Fixed Effects
Within-Subject (Level 1)
Life Events Scale (LES) Slope	0.48	0.07	<.001	.26
Between-Subject (Level 2)
BDI (Baseline)	0.47	0.07	<.001	.25
LES Mean	0.25	0.07	<.001	.09
Decentering	−0.94	0.40	.02	.04
Decentering × LES	−0.15	0.06	.01	.05
Random Effects
BDI (Random Intercept)	10.76	1.88	<.001	.20
LES (Random Slope)	0.12	0.03	<.001	.11
Predicting Anxiety Symptoms				.75
BAI (Intercept)	0.51	0.58	.38
Fixed Effects
Within-Subject (Level 1)
Life Events Scale (LES) Slope	0.50	0.10	<.001	.16
Between-Subject (Level 2)
BAI (Baseline)	0.42	0.05	<.001	.35
LES Mean	0.24	0.07	.001	.08
Decentering	−0.80	0.47	.09	.02
Decentering × LES	−0.07	0.09	.40	<.01
Random Effects
BAI (Random Intercept)	15.41	2.81	<.001	.19
LES (Random Slope)	0.41	0.16	.01	.05

Note. BDI = Beck Depression Inventory; LES = Life Event Scale; BAI = Beck Anxiety Inventory. R² for full model represents pseudo R², the proportion of variance explained by the model relative to an unrestricted (intercept-only) model containing no predictors (Kreft & Leeuw, 1998; Singer, 1998). ΔR² = partial R² of given predictor (proportion of variance in outcome predicted) after accounting for covariates. Level 2 predictors were centered at Level 2; Level 1 predictors were centered at Level 1 (except for variables in meaningful units, i.e., BDI and BAI).

Figure 2.

Cross-level interaction between decentering and person-centered negative life events predicting weekly (a) symptoms of depression, and (b) brooding. Decentering plotted at +/− 1 SD from the sample mean.

Hypothesis 2: Does Decentering Predict an Attenuated Relationship between Negative Life Events and Perseverative Thought?

As hypothesized, there was a significant interaction between decentering and negative life events predicting weekly brooding (Table 2), such that negative life events predicted brooding more strongly among individuals with low decentering (B = 0.15, SE = 0.04, p < .001) than among individuals with high decentering (B = 0.06, SE = 0.03, p = .02; Figure 2b). In contrast, the interactions between decentering and negative life events predicting pondering and worry were not significant, although there was a significant positive main effect of life events on both pondering and worry (Table 2). Of the three analyses conducted for this hypothesis, the significant decentering × life events interaction predicting brooding did not survive the post-hoc FDR correction threshold.

Table 2.

Multilevel Models with Decentering Moderating Relationships between Person-Centered Negative Life Event Exposure at Each Wave and Perseverative Thinking at Each Wave

Model (Predictor)	B/γ	SE	p	ΔR 2
Predicting Weekly Brooding				.64
Brooding (Intercept)	7.26	0.96	<.001
Fixed Effects
Within-Subject (Level 1)
Life Events Scale (LES) Slope	0.11	0.03	<.001	.09
Between-Subject (Level 2)
Trait Brooding (Baseline)	1.61	0.28	<.001	.20
LES Mean	0.10	0.03	<.001	.08
Decentering	−0.32	0.21	.13	.02
Decentering × LES	−0.04	0.02	.04	.03
Random Effects
Brooding (Random Intercept)	2.74	0.76	<.001	.09
LES (Random Slope)	0.002	0.04	.96	<.01
Predicting Weekly Pondering				.37
Pondering (Intercept)	8.90	0.34	<.001
Fixed Effects
Within-Subject (Level 1)
Life Events Scale (LES) Slope	0.15	0.04	<.001	.10
Between-Subject (Level 2)
Trait Pondering (Baseline)	2.16	0.22	<.001	.42
LES Mean	0.09	0.03	.001	.06
Decentering	0.07	0.24	.78	<.01
Decentering × LES	−0.03	0.04	.40	<.01
Random Effects
Pondering (Random Intercept)	4.89	0.97	<.001	.16
LES (Random Slope)	0.01	0.02	.74	<.01
Predicting Weekly Worry				.54
Worry (Intercept)	35.53	0.96	<.001
Fixed Effects
Within-Subject (Level 1)
Life Events Scale (LES) Slope	0.35	0.13	.005	.05
Between-Subject (Level 2)
Trait Worry (Baseline)	5.99	0.72	<.001	.34
LES Mean	0.33	0.08	<.001	.11
Decentering	0.04	0.67	.95	<.01
Decentering × LES	−0.11	0.08	.15	.01
Random Effects
Worry (Random Intercept)	38.78	5.53	<.001	.27
LES (Random Slope)	0.14	0.21	.49	<.01

Note. LES = Life Event Scale. R² for full model represents pseudo R², the proportion of variance explained by the model relative to an unrestricted (intercept-only) model containing no predictors (Kreft & Leeuw, 1998; Singer, 1998). ΔR² = partial R² of given predictor (proportion of variance in outcome predicted) after accounting for covariates. Level 2 predictors were centered at Level 2; Level 1 predictors were centered at Level 1 (except for variables in meaningful units, i.e., BDI and BAI).

Hypothesis 3: Does Decentering Predict an Attenuated Relationship between Perseverative Thought and Internalizing Symptoms?

There were significant interactions between decentering and brooding predicting symptoms of depression and anxiety (Table 3), such that brooding predicted higher symptoms among individuals with low decentering (BDI: B = 1.04, SE = 0.12, p < .001; BAI: B = 1.12, SE = 0.18, p < .001), but not among individuals with high decentering (BDI: B = −0.04, SE = 0.22, p = .86; BAI: B = 0.33, SE = 0.29, p = .26; Figures 3a and 3b).

Table 3.

Multilevel Models with Decentering Moderating Relationships between Person-Centered Perseverative Thinking and Internalizing Symptoms at Each Wave

Model (Predictor)	B/γ	SE	p	ΔR 2
Brooding Predicting Depression Symptoms				.67
BDI (Intercept)	2.58	0.70	<.001
Fixed Effects
Within-Subject Level (Level 1)
Brooding Slope	0.50	0.13	.001	.10
Between-Subject (Level 2)
BDI (Baseline)	0.50	0.05	<.001	.43
Trait Brooding (Baseline)	1.69	0.51	.001	.08
Decentering	−0.94	0.43	.03	.03
Decentering × Brooding	−0.54	0.12	<.001	.13
Random Effects
BDI (Random Intercept)	12.28	1.56	<.001	.32
Brooding (Random Slope)	0.23	0.17	.17	.01
Brooding Predicting Anxiety Symptoms				.67
BAI (Intercept)	1.76	0.92	.06
Fixed Effects
Within-Subject Level (Level 1)
Brooding Slope	0.72	0.18	<.001	.11
Between-Subject (Level 2)
BAI (Baseline)	0.51	0.04	<.001	.55
Trait Brooding (Baseline)	0.88	0.64	.17	.01
Decentering	−0.75	0.55	.18	.01
Decentering × Brooding	−0.40	0.17	.02	.04
Random Effects
BAI (Random Intercept)	20.08	2.78	<.001	.28
Brooding (Random Slope)	0.64	0.36	.08	.02
Pondering Predicting Depression Symptoms				.64
BAI (Intercept)	2.09	0.89	.02
Fixed Effects
Within-Subject Level (Level 1)
Pondering Slope	0.48	0.14	<.001	.08
Between-Subject (Level 2)
BDI (Baseline)	0.56	0.07	<.001	.33
Trait Pondering (Baseline)	0.75	0.53	.16	.01
Decentering	−1.23	0.45	<.01	.05
Decentering × LES	−0.36	0.10	.001	.09
Random Effects
BDI (Random Intercept)	13.46	1.98	<.001	.18
Pondering (Random Slope)	0.41	0.18	.02	.04
Pondering Predicting Anxiety Symptoms				.65
BAI (Intercept)	1.90	0.95	<.05
Fixed Effects
Within-Subject Level (Level 1)
Pondering Slope	0.71	0.16	<.001	.13
Between-Subject (Level 2)
BAI (Baseline)	0.50	0.05	<.001	.43
Trait Pondering (Baseline)	0.98	0.62	.11	.02
Decentering	−1.05	0.53	<.05	.03
Decentering × LES	−0.40	0.15	<.01	.05
Random Effects (Level 2)
BAI (Random Intercept)	21.49	2.91	<.001	.29
Pondering (Random Slope)	0.54	0.22	.02	.04
Worry Predicting Depression Symptoms				.69
BDI (Intercept)	2.05	0.77	<.01
Fixed Effects
Within-Subject Level (Level 1)
Worry Slope	0.13	0.06	.03	.03
Between-Subject (Level 2)
BDI (Baseline)	0.57	0.05	<.001	.50
Trait Worry (Baseline)	0.73	0.47	.12	.02
Decentering	−0.98	0.46	.03	.03
Decentering × LES	−0.10	0.04	.02	.05
Random Effects (Level 2)
BDI (Random Intercept)	11.60	2.08	<.001	.19
Worry (Random Slope)	0.05	0.02	.02	.05
Worry Predicting Anxiety Symptoms				.69
BAI (Intercept)	2.10	0.90	.02
Fixed Effects
Within-Subject Level (Level 1)
Worry Slope	0.19	0.07	<.005	.05
Between-Subject (Level 2)
BAI (Baseline)	.51	0.05	<.001	.44
Trait Worry (Baseline)	0.49	0.67	.47	<.01
Decentering	−1.14	0.47	.02	.04
Decentering × LES	−0.14	0.06	.01	.04
Random Effects (Level 2)
BAI (Random Intercept)	18.65	2.90	<.001	.24
Worry (Random Slope)	0.05	0.02	<.005	.05

Note. BDI = Beck Depression Inventory; BAI = Beck Anxiety Inventory. R² for full model represents pseudo R², the proportion of variance explained by the model relative to an unrestricted (intercept-only) model containing no predictors (Kreft & Leeuw, 1998; Singer, 1998). ΔR² = partial R² of given predictor (proportion of variance in outcome predicted) after accounting for covariates. Level 2 predictors were centered at Level 2; Level 1 predictors were centered at Level 1 (except for variables in meaningful units, i.e., BDI and BAI).

Figure 3.

Cross-level interactions between decentering and person-centered (a) brooding predicting symptoms of depression, (b) brooding predicting symptoms of anxiety, (c) pondering predicting symptoms of depression, (d) pondering predicting symptoms of anxiety, (e) worry predicting symptoms of depression, and (f) worry predicting symptoms of anxiety. Decentering plotted at +/− 1 SD from the sample mean.

The interactions between decentering and pondering predicting symptoms of depression and anxiety were significant and marginally significant, respectively (Table 3). Pondering predicted higher symptoms among individuals with low decentering (BDI: B = 0.84, SE = 0.13, p < .001; BAI: B = 1.11, SE = 0.19, p < .001), but not among individuals with high decentering (BDI: B = 0.12, SE = 0.20, p = .55; BAI: B = 0.31, SE = 0.25, p = .22; Figures 3c and 3d).

Finally, there were significant interactions between decentering and worry predicting symptoms of depression and anxiety (Table 3), such that worry predicted higher symptoms among individuals with low decentering (BDI: B = 0.23, SE = 0.05, p < .001; BAI: B = 0.34, SE = 0.07, p < .001), but not among individuals with high decentering (BDI: B = 0.03, SE = 0.09, p = .72; BAI: B = 0.05, SE = 0.11, p = .60; Figures 3e and 3f).

Of the six analyses conducted for this hypothesis, all six significant interactions survived the post-hoc FDR correction (threshold q = .05).

Hypotheses 4 and 5: Does Decentering Moderate the Indirect Pathways from Negative Life Events to Perseverative Thought to Internalizing Symptoms?

Next, we conducted a series of multilevel moderated mediation models to examine whether PT would mediate the relationship between higher levels of negative life events than usual and symptoms of depression and anxiety, and whether this pathway would be stronger for individuals with lower levels of decentering. For Hypothesis 4, we specifically examined whether decentering would moderate the link between higher than usual levels of negative life events and brooding (given evidence for this interaction above in tests of Hypothesis 2), within the context of these mediation pathways predicting depression and anxiety. Consistent with Hypothesis 4, there was a significant conditional indirect effect of life events on symptoms of depression via brooding that were stronger among individuals low in decentering, compared to individuals high in decentering (Table 4). However, the conditional indirect effect of life events on symptoms of anxiety via brooding was only marginally significant. Of the two analyses conducted for this hypothesis, the significant indirect effect predicting symptoms of depression survived the post-hoc FDR correction (threshold q = .025).

Table 4.

Multilevel Moderated Mediation Models with Perseverative Thinking Mediating Relationships between Fluctuations in Negative Life Events and Internalizing Symptoms, with Decentering Moderating the Relationships between Negative Life Events and Perseverative Thinking

Model (Predictor)	B/γ	SE	p	ΔR 2
Stress → Brooding → Depression, with Decentering Moderating the Stress → Brooding Link
Brooding (Intercept)	7.26	0.94	<.001
BDI (Intercept)	−1.15	1.58	.47
Fixed Effects
Within-Subject (Level 1) - Predicting Brooding
Life Events Scale (LES) Slope	0.10	0.02	<.001	.16
Within-Subject (Level 1) - Predicting Depression Symptoms
Brooding Slope	0.35	0.11	<.005	.07
Life Events Scale (LES) Slope	0.35	0.07	<.001	.16
Between-Subject (Level 2) - Predicting Levels of Brooding
Trait Brooding (Baseline)	1.60	0.27	.06	.21
LES Mean	0.10	0.03	<.001	.08
Decentering	−0.31	0.21	.13	.02
Between-Subject (Level 2) - Predicting Levels of Depression Symptoms
BDI (Baseline)	0.37	0.06	<.001	.22
Trait Brooding (Baseline)	0.80	0.43	.06	.03
LES Mean	0.16	0.06	.01	.05
Decentering	−0.35	0.31	.27	.01
Decentering × Brooding	−0.04	0.02	.01	.03
Random Effects
BDI (Random Intercept)	0.87	2.52	.73	<.01
Brooding (Random Intercept)	2.76	0.74	<.001	.10
LES predicting Brooding (Random Slope)	0.002	0.03	.96	<.01
LES predicting BDI (Random Slope)	0.20	0.08	.02	.05
Brooding predicting BDI (Random Slope)	0.08	0.03	<.01	.05
Conditional Indirect Effect of LES on BDI via Brooding
Indirect Effect (High Decentering)	0.02	0.01	.06
95% CI: .000 to .044
Indirect Effect (Low Decentering)	0.05	0.02	.01
95% CI: .014 to .090
Stress → Brooding → Anxiety, with Decentering Moderating the Stress → Brooding Link
Brooding (Intercept)	7.26	0.97	<.001
BAI (Intercept)	−6.88	1.91	<.001
Fixed Effects
Within-Subject (Level 1) - Predicting Brooding
Life Events Scale (LES) Slope	0.11	0.04	<.001	.05
Within-Subject (Level 1) - Predicting Anxiety Symptoms
Brooding Slope	0.62	0.13	<.001	.15
Life Events Scale (LES) Slope	0.34	0.08	<.001	.12
Between-Subject (Level 2) - Predicting Levels of Brooding
Trait Brooding (Baseline)	1.61	0.28	<.001	.20
LES Mean	0.10	0.03	<.001	.08
Decentering	−0.32	0.21	.12	.02
Between-Subject (Level 2) - Predicting Levels of Anxiety Symptoms
BAI (Baseline)	0.39	0.07	<.001	.19
Trait Brooding (Baseline)	−0.57	0.65	.38	.01
LES Mean	0.18	0.09	.05	.03
Decentering	0.11	0.62	.86	<.01
Decentering × Brooding	−0.04	0.02	.06	.03
Random Effects
BAI (Random Intercept)	2.63	4.96	.60	<.01
Brooding (Random Intercept)	2.77	0.77	<.001	.09
LES predicting Brooding (Random Slope)	.002	0.04	.97	<.01
LES predicting BAI (Random Slope)	0.20	0.13	.14	.02
Brooding predicting BAI (Random Slope)	0.15	0.04	.001	.10
Conditional Indirect Effect of LES on BAI via Brooding
Indirect Effect (High Decentering)	0.04	0.02	.03
95% CI: .003 to .070
Indirect Effect (Low Decentering)	0.09	0.04	.01
95% CI: .030 to .147

Note. BDI = Beck Depression Inventory; LES = Life Event Scale; BAI = Beck Anxiety Inventory. R² for full model represents pseudo R², the proportion of variance explained by the model relative to an unrestricted (intercept-only) model containing no predictors (Kreft & Leeuw, 1998; Singer, 1998). ΔR² = partial R² of given predictor (proportion of variance in outcome predicted) after accounting for covariates. Level 2 predictors were centered at Level 2; Level 1 predictors were centered at Level 1 (except for variables in meaningful units, i.e., BDI and BAI).

To test Hypothesis 5, we specifically examined whether decentering would moderate the link between increases in PT and internalizing symptoms, within this mediation pathway. Consistent with this hypothesis, there were significant conditional indirect effects of life events on symptoms of depression and anxiety via brooding among individuals low in decentering, but not among individuals high in decentering (Table 5). With respect to pondering, in the moderated mediation models predicting symptoms of depression and anxiety, there were significant conditional indirect effects of life events on symptoms of depression and anxiety via pondering among individuals low in decentering, but not among individuals high in decentering (Table 5). The model with worry predicting symptoms of depression failed to converge, which did not support the hypothesis. However, supporting Hypothesis 5, in the model with worry predicting symptoms of anxiety, there were significant conditional indirect effects of life events on symptoms of anxiety via worry among individuals low in decentering, but not among individuals high in decentering (Table 5). Of the six analyses conducted for this hypothesis, the five significant interactions survived the post-hoc FDR correction (threshold q = .05).

Table 5.

Multilevel Moderated Mediation Models with Perseverative Thinking Mediating Relationships between Fluctuations in Negative Life Events and Internalizing Symptoms, with Decentering Moderating the Relationships between Perseverative Thinking and Internalizing Symptoms

Model (Predictor)	B/γ	SE	p	ΔR 2
Stress → Brooding → Depression, with Decentering Moderating the Brooding → Depression Link
Brooding (Intercept)	8.61	0.27	<.001
BDI (Intercept)	−1.57	1.03	.13
Fixed Effects
Within-Subject (Level 1) - Predicting Brooding
Life Events Scale (LES) Slope	0.13	0.06	.02	.03
Within-Subject (Level 1) - Predicting Depression Symptoms
Brooding Slope	0.35	0.11	.001	.07
Life Events Scale (LES) Slope	0.35	0.07	<.001	.16
Between-Subject (Level 2) - Predicting Levels of Brooding
Trait Brooding (Baseline)	1.85	0.20	<.001	.39
LES Mean	0.11	0.02	.001	.19
Decentering	−0.30	0.22	.17	.01
Between-Subject (Level 2) - Predicting Levels of Depression Symptoms
BDI (Baseline)	0.32	0.07	<.001	.14
Trait Brooding (Baseline)	0.73	0.39	.07	.03
LES Mean	0.20	0.06	.001	.08
Decentering	0.91	0.72	.21	.01
Decentering × Brooding	−0.18	0.07	.02	.05
Random Effects
BDI (Random Intercept)	0.86	2.81	76	<.01
Brooding (Random Intercept)	3.17	0.57	<.001	.19
LES predicting Brooding (Random Slope)	0.01	0.04	.23	<.01
LES predicting BDI (Random Slope)	0.19	0.08	.02	.04
Brooding predicting BDI (Random Slope)	0.09	0.03	.005	.06
Conditional Indirect Effect of LES on BDI via Brooding
Indirect Effect (High Decentering)	0.02	0.02	.26
95% CI: −.009 to .057
Indirect Effect (Low Decentering)	.07	.03	.03
95% CI: .005 to .103
Stress → Brooding → Anxiety, with Decentering Moderating the Brooding → Anxiety Link
Brooding (Intercept)	8.61	0.27	<.001
BAI (Intercept)	−5.28	1.39	<.001
Fixed Effects
Within-Subject (Level 1) - Predicting Brooding
Life Events Scale (LES) Slope	0.13	0.05	<.01	.05
Within-Subject (Level 1) - Predicting Anxiety Symptoms
Brooding Slope	0.66	0.14	<.001	.14
Life Events Scale (LES) Slope	0.33	0.08	<.001	.11
Between-Subject (Level 2) - Predicting Levels of Brooding
Trait Brooding (Baseline)	1.85	0.20	<.001	.39
LES Mean	0.11	0.02	.001	.19
Decentering	−0.30	0.22	.16	.01
Between-Subject (Level 2) - Predicting Levels of Anxiety Symptoms
BAI (Baseline)	0.37	0.07	<.001	.17
Trait Brooding (Baseline)	−0.55	0.57	.33	<.01
LES Mean	0.19	0.08	.01	.04
Decentering	2.42	1.19	.04	.03
Decentering × Brooding	−0.28	0.10	<.005	.06
Random Effects
BAI (Random Intercept)	1.98	4.49	.66	<.01
Brooding (Random Intercept)	3.16	.57	<.001	.19
LES predicting Brooding (Random Slope)	0.01	0.03	.80	<.01
LES predicting BAI (Random Slope)	0.16	0.10	.13	.02
Brooding predicting BAI (Random Slope)	0.14	0.04	.001	.08
Conditional Indirect Effect of LES on BAI via Brooding
Indirect Effect (High Decentering)	0.05	0.03	.13
95% CI: −.013 to .073
Indirect Effect (Low Decentering)	0.13	0.06	.04
95% CI: .018 to .179
Stress → Pondering → Depression, with Decentering Moderating the Pondering → Depression Link
Pondering (Intercept)	8.88	0.34	<.001
BDI (Intercept)	−1.19	1.01	.24
Fixed Effects
Within-Subject (Level 1) - Predicting Pondering
Life Events Scale (LES) Slope	0.15	0.04	<.001	.10
Within-Subject (Level 1) - Predicting Depression Symptoms
Pondering Slope	0.25	0.10	.02	.05
Life Events Scale (LES) Slope	0.37	0.08	<.001	.14
Between-Subject (Level 2) - Predicting Levels of Pondering
Trait Pondering (Baseline)	2.08	0.24	<.001	.36
LES Mean	0.09	0.03	<.001	.06
Decentering	0.09	0.24	.72	<.01
Between-Subject (Level 2) - Predicting Levels of Depression Symptoms
BDI (Baseline)	0.40	0.06	<.001	.25
Trait Pondering (Baseline)	0.08	0.37	.83	<.01
LES Mean	0.23	0.06	<.001	.10
Decentering	.002	0.58	.99	<.01
Decentering × Pondering	−0.12	0.06	<.05	.03
Random Effects
BDI (Random Intercept)	3.35	2.84	.24	.01
Pondering (Random Intercept)	5.45	0.87	<.001	.23
LES predicting Pondering (Random Slope)	0.01	0.02	.69	<.01
LES predicting BDI (Random Slope)	0.08	0.03	.01	.05
Pondering predicting BDI (Random Slope)	0.19	0.08	.02	.04
Conditional Indirect Effect of LES on BDI via Pondering
Indirect Effect (High Decentering)	0.02	0.02	.20
95% CI: −.013 to .046
Indirect Effect (Low Decentering)	0.05	0.03	.05
95% CI: −.010 to .088
Stress → Pondering → Anxiety, with Decentering Moderating the Pondering → Anxiety Link
Brooding (Intercept)	8.88	0.34	<.001
BAI (Intercept)	−3.85	1.12	.001
Fixed Effects
Within-Subject (Level 1) - Predicting Pondering
Life Events Scale (LES) Slope	0.15	0.04	<.001	.10
Within-Subject (Level 1) - Predicting Anxiety Symptoms
Pondering Slope	0.55	0.12	.001	.14
Life Events Scale (LES) Slope	0.34	0.09	<.001	.10
Between-Subject (Level 2) - Predicting Levels of Pondering
Trait Pondering (Baseline)	2.15	0.22	<.001	.42
LES Mean	0.18	0.03	.001	.21
Decentering	0.09	0.24	.72	<.01
Between-Subject (Level 2) - Predicting Levels of Anxiety Symptoms
BAI (Baseline)	0.37	0.06	<.001	.22
Trait Pondering (Baseline)	−0.35	0.37	.34	<.01
LES Mean	0.18	0.07	<.01	.05
Decentering	1.62	0.80	.04	.03
Decentering × Pondering	−0.27	0.09	<.005	.06
Random Effects
BAI (Random Intercept)	0.37	3.26	.91	<.01
Pondering (Random Intercept)	5.41	0.87	<.001	.23
LES predicting Pondering (Random Slope)	0.006	0.02	.68	<.01
LES predicting BAI (Random Slope)	0.19	0.14	17	.01
Pondering predicting BAI (Random Slope)	0.16	0.04	<.001	.11
Conditional Indirect Effect of LES on BAI via Pondering
Indirect Effect (High Decentering)	0.04	0.02	.08
95% CI: −.017 to .072
Indirect Effect (Low Decentering)	0.12	0.05	.01
95% CI: .006 to .190
Stress → Worry → Anxiety, with Decentering Moderating the Worry →Anxiety Link
Worry (Intercept)	35.87	1.04	<.001
BAI (Intercept)	−4.63	1.74	<.01
Fixed Effects
Within-Subject (Level 1) - Predicting Worry
Life Events Scale (LES) Slope	0.38	0.12	<.005	.07
Within-Subject (Level 1) - Predicting Anxiety Symptoms
Worry Slope	0.16	.04	<.001	.11
Life Events Scale (LES) Slope	0.30	0.07	<.001	.12
Between-Subject (Level 2) - Predicting Levels of Worry
Trait Worry (Baseline)	6.00	0.75	<.001	.33
LES Mean	0.29	0.10	.005	.06
Decentering	0.26	0.77	.75	<.01
Between-Subject (Level 2) - Predicting Levels of Anxiety Symptoms
BAI (Baseline)	0.48	0.04	<.001	.52
Trait Worry (Baseline)	−0.50	0.84	.51	<.01
LES Mean	0.23	0.07	<.005	.08
Decentering	3.96	1.53	<.01	.05
Decentering × Worry	0.16	0.04	<.001	.11
Random Effects
BAI (Random Intercept)	1.06	2.12	.62	<.01
Worry (Random Intercept)	35.87	1.04	<.001	.90
LES predicting Worry (Random Slope)	0.22	0.17	.19	.01
Worry predicting BAI (Random Slope)	0.01	0.003	.05	.08
Conditional Indirect Effect of LES on BAI via Worry
Indirect Effect (High Decentering)	0.02	0.02	.25
95% CI: −.042 to .014
Indirect Effect (Low Decentering)	0.11	0.04	<.005
95% CI: −.011 to .047

Note. BDI = Beck Depression Inventory; LES = Life Event Scale; BAI = Beck Anxiety Inventory. R² for full model represents pseudo R², the proportion of variance explained by the model relative to an unrestricted (intercept-only) model containing no predictors (Kreft & Leeuw, 1998; Singer, 1998). ΔR² = partial R² of given predictor (proportion of variance in outcome predicted) after accounting for covariates. Level 2 predictors were centered at Level 2; Level 1 predictors were centered at Level 1 (except for variables in meaningful units, i.e., BDI and BAI).

Discussion

The main purpose of this study was to investigate to what extent, and at what points in the pathway from negative life events to internalizing symptoms, decentering acts as a protective factor against two established risk factors (negative life events and perseverative thought) for anxiety and depressive symptoms. Decentering had a buffering effect at several different points in the pathway, including between negative life events and prospective depressive symptoms, between negative life events and brooding, and between three forms of PT (brooding, pondering and worry) and internalizing symptoms. Furthermore, brooding mediated relationships between negative life events and internalizing symptoms, and this mediation pathway was buffered by greater levels of decentering at both points in the indirect pathway to symptoms of depression. Together, these findings demonstrate that decentering is associated with reduced likelihood of brooding in response to stressful events, and with reduced likelihood of developing anxiety and depressive symptoms after engaging in PT. Thus, decentering may represent a promising target for clinical intervention to enhance resilience after stressful events.

As hypothesized, when individuals were exposed to more negative life events than usual, they engaged in greater PT and experienced elevated levels of internalizing symptoms. Consistent with prior literature, higher levels of negative life events than usual were associated with prospective levels of symptoms of depression and anxiety (Michl et al., 2013; Vinkers et al., 2014). However, higher levels of decentering were found to be associated with a weaker relationship between exposure to negative life events and depressive, but not anxiety, symptoms, indicating that individuals with greater decentering were protected from future depressive symptoms even during times when they experienced more negative events than usual. Our analyses also showed higher levels of negative life events to be predictive of future levels of brooding, pondering, and worry over time, with decentering moderating the link from negative life events to brooding, but not to pondering or worry. As decentering involves awareness of and detachment from one’s internal experiences, it thereby may reduce the effects of thought content on other mental processes, including engaging in negative forms of cognitive elaboration (Bernstein et al., 2015), such as brooding.

Brooding in particular involves an abstract processing style, characterized by a focus on general and decontextualized information, for example, the meaning, causes and consequences of events (e.g. “Why is this happening to me?”; Watkins, 2008; Watkins & Roberts, 2020). An abstract processing style contributes to reduced sensitivity to environmental cues, which can include potential reward and changes in action contingencies. As a consequence, ruminators are less susceptible to corrective learning and the disconfirmation of negative beliefs (Reilly et al., 2019). In being aware and stepping back from one’s thoughts and feelings, decentering may help individuals disengage from an abstract style of rumination, and instead observe one’s thoughts and feelings within the context of a specific moment (e.g. “I am having this thought in this moment”). Taking a self-distanced perspective also may facilitate more adaptive means of reflecting on negative experiences, for example, through reconstruing one’s emotions and the meaning of events, as opposed to recounting the details of the experience (Ayduk & Kross, 2010; Kross & Ayduk, 2008). That is, adopting a decentered perspective may facilitate cognitive reappraisal, by allowing one to notice and question the accuracy of negative thought content, thereby reducing the need to cognitively elaborate or emotionally react. Indeed, in an open trial of Emotion Regulation Therapy, improvements in decentering predicted improvements in cognitive reappraisal, an important facilitator in symptom change in anxiety and mood disorders (Forkmann et al., 2014; O’Toole et al., 2019; Smits et al., 2012). The current findings lend support for the role of decentering in determining whether exposure to negative experiences will lead to adaptive or maladaptive forms of self-reflection. However, not all hypotheses in the present study were supported. For example, decentering did not moderate the prospective relationships between stress and anxiety, and between stress and pondering or worry. As such, future research should verify whether these findings can be replicated to clarify why decentering might not influence these associations.

We also examined the influence of decentering in the association between perseverative thinking and subsequent internalizing symptoms. When people engaged in more PT than usual, they were likely to have higher levels of anxiety and depressive symptoms; however, they were protected against worsening symptoms if they had greater tendency to decenter. This buffering effect of decentering was consistent for all three forms of PT (brooding, pondering, and worry). According to relational frame theory (Hayes et al., 2001), humans tend to link or “fuse” thoughts with emotions, such that having an anxious thought elicits as much anxiety as experiencing the anxiety-provoking situation. Thus, fusion with internal experiences promotes greater emotional reactivity to internal experiences. Decentering processes might reduce the likelihood of developing anxiety and depressive symptoms by buffering emotional reactivity to negative thought content (Bernstein et al., 2015), as occurs with worry and brooding. For example, through the process of meta-awareness or awareness of one’s internal experiences, attention is drawn away from one’s thoughts to the present moment, thereby reducing emotional reactivity. Furthermore, disidentification from internal experiences such as negative thought content may diminish the emotional salience of those thoughts. Thus, by adopting a decentered perspective, engaging in perseverative thinking will not necessarily generate the anxious or depressed feelings that typically follow.

A second objective of the paper was to link together the above results to test a potential causal pathway, whereby perseverative thinking explains the relationship between negative life events and prospective levels of anxiety and depressive symptoms, with these relationships attenuated by decentering. Several moderated mediation models provided support for these hypotheses, although the evidence was somewhat mixed. The well-recognized association between negative life events and symptoms of anxiety and depression was mediated by worry and rumination. Negative life events were likely to lead to brooding, particularly for individuals who reported lower levels of decentering, which, in turn, precipitated higher depressive and anxiety symptoms, and more so for individuals with lower levels of decentering. Higher levels of decentering failed to buffer the impact of negative life events on pondering and worry; however, decentering did predict diminished relationships between worry and pondering and anxiety symptoms in these mediational models. Critically, although negative experiences may induce perseverative thinking, our results suggest that “all is not lost” if perseveration does increase. That is, decentering still may have utility in preventing perseveration from leading to depressive and anxiety symptoms. Thus, decentering appears to play a role in protecting individuals from both the direct and indirect impacts of negative life experiences, and, in turn, may promote resilience against anxiety and depression.

The negative life experiences that precipitate anxiety and depression often occur in an unpredictable and incontrollable way. Determining mechanisms and moderators in the link between negative events and psychopathology is therefore crucial to identifying targets for intervention that may prevent the onset of psychological disorder. Our findings identified two potential treatment targets, including perseverative thinking and decentering. Both cognitive and “third-wave” behavioral therapies have been shown to reduce perseverative thinking (Chambers et al., 2008; Mennin et al., 2018; van Aalderen et al., 2012; Watkins et al., 2011; for a review, see Querstret & Cropley, 2013) and improve anxiety and depression (Cuijpers et al., 2011; McCarney et al., 2012; Stewart & Chambless, 2009; Vøllestad et al., 2012). The extent to which these existing interventions can prevent the development of PT and anxiety and depression remains to be established. However, a recent study found cognitive-behavioral training targeting perseverative thinking may be effective in preventing the onset of depression and anxiety, even in non-clinical, never-depressed samples (Topper et al., 2017).

Our findings suggest that decentering is a protective factor that potentially can mitigate the deleterious impacts of two established risk factors for anxiety and depression (negative life events and PT), and identify at which specific points decentering may or may not be useful. Training in decentering might reduce the tendency to engage in brooding following recent stressors, or reduce emotional reactivity to repetitive negative thought content, thereby attenuating the development of internalizing symptoms. Decentering has been demonstrated to be a mediator of change across mindfulness-based and cognitive-behavioral treatments, with increases in decentering corresponding with decreases in worry, rumination, anxiety and depressive symptoms, depression relapse, impairments of functioning and quality of life (Feldman et al., 2010b; Fissler et al., 2016; Hayes-Skelton et al., 2015; Hoge et al., 2015; Mennin et al., 2018). Findings from clinical studies suggest that improvements in decentering may not only contribute to reducing the frequency of rumination or worry, but also to dampening one’s emotional response to repetitive thoughts (Feldman et al., 2010; Mennin et al., 2018). For example, acceptance-based behavior therapy focuses away from eliminating worry in generalized anxiety disorder, and uses mindfulness and acceptance techniques to strengthen decentering, and thereby, reduce emotional reactivity to internal experiences (Hayes-Skelton et al., 2015; Roemer et al., 2008). Decentering appears to play a significant role in promoting resilience to developing depressive and anxiety symptoms and as a therapeutic mechanism of change. Preventative interventions could be tailored for individuals who have experienced recent negative stressors and are therefore at greater risk of developing negative and repetitive patterns of thinking, and anxiety and depressive symptoms.

The present study had several strengths, including use of a longitudinal multi-wave design, allowing person-centered modeling of negative experiences, perseverative thinking and internalizing symptoms. However, results should be interpreted in light of a number of limitations. Although statistically significant, the effects of decentering in moderating the link between PT and internalizing symptoms were relatively small. Replication using larger samples and in clinical samples is needed to determine potential clinical impact. As a non-clinical sample of students was used, there were relatively low mean levels of internalizing symptoms. However, there was ample variability in levels of negative life events, worry, rumination, and anxiety and depressive symptoms (both between and within subjects over time; ICC range of .46–76) to allow the modeling of relationships between these constructs in the present sample.

Next, variables aside from decentering that may also contribute to explaining the variance in PT and internalizing symptoms warrant exploration in future studies. For example, stress appears to interact with gender, avoidant coping, self-esteem, negative cognitive style, metacognition and cognitive reappraisal ability to predict internalizing symptoms (Blalock & Joiner, 2000; Eisenbarth, 2012; Spada et al., 2008; Stange et al., 2013; Troy et al., 2010), whereas the association between rumination and depressed mood is amplified by negative cognition (Ciesla et al., 2011; Ciesla & Roberts, 2007; Hamlat et al., 2015; Stange et al., 2015). Further investigation using clinical samples is needed to improve the generalizability of our findings, although it is important to note that college-age students are at heightened risk for depression and anxiety (Duffy et al., 2019). Although the use of a life events interview to verify the occurrence of negative events was a strength of the current study (Liu, 2013), for feasibility reasons, the interview only was completed at the last wave, which could have resulted in recall biases. Furthermore, to limit the number of analyses in the current manuscript, we utilized a composite measure of negative life events; future research could further probe the specificity of life events by sub-classifying them as major versus minor, dependent versus independent, and episodic versus chronic events. There are circumstances where a single, enduring (and potentially traumatic) event may continue to exert influence on individuals and heighten risk for depression. Although this may manifest in the occurrence of subsequent life stressors (consistent with a stress generation approach) and therefore be reflected by higher levels of life events than usual, the context and severity of individual life events remains important to consider in depression risk.

Although within-subject factors were measured on several occasions every three weeks, fluctuations in these constructs may occur more quickly and transiently, and these fluctuations may fail to be captured within the timeframe in which they were evaluated. Although this was a prospective study, because responses to life events are likely to occur in close proximity to the events themselves, the associations tested in this report were concurrent in nature – that is, life events over the prior three weeks were measured (and analyzed) at the same time points as perseverative thought and current symptoms. Although our results held when accounting for lagged versions of the outcome variables (representing residual changes in PT and symptoms), future reports also could test whether such associations exist for relationships that are lagged over longer periods of time (e.g., whether life events have longer-term effects on PT and symptoms several weeks later). Future studies using daily diary and ecological momentary assessment methods might improve accuracy in modelling the relationships between negative life events, perseverative thinking, and anxiety and depressive symptoms, and confirm whether decentering would have a similar protective effect at a more fine-grained level of granularity.

In addition, alternative statistical approaches to modelling time series data might be considered in future studies, to account for lagged outcome variables while minimizing bias in model estimates (Falkenström et al., 2017). Like other psychological attributes, decentering may have both a stable component and one that fluctuates over time and across context (Shoham et al., 2017). As such, an assessment of state decentering or more recent use of self-distancing techniques might elucidate how decentering might in the moment reduce PT following negative experiences, and, in turn, prevent PT from giving rise to psychopathological symptoms. Future research also may consider using a behavioral measure of decentering rather than relying on self-reports (Shepherd et al., 2016). Some further important questions regarding the utility of decentering remain, including how much ability or training in decentering is necessary for beneficial effect, and how decentering should best be taught and practiced in order to safeguard against the adverse effects of negative experiences and PT, and to reduce emotional reactivity to negative thought content.

Negative life experiences can naturally challenge one’s capacity to cope, activating negative and maladaptive patterns of cognitive reflection and emotional reaction. However, the findings of this study suggest that decentering, or the tendency to adopt a distanced perspective on one’s thoughts and emotions, may be helpful in buffering the negative impacts of two robust vulnerability factors for the development of internalizing psychopathology. This is the first study to examine when decentering may and may not be helpful in the pathway from negative life experiences to perseverative thinking to symptoms. Decentering appears to be helpful at several different junctures, particularly, in being associated with a buffered impact of negative life events on brooding, and with a buffered impact of of brooding, pondering and worry on depressive and anxiety symptoms. As such, decentering appears to be an important protective factor and potential target for clinical intervention. Training individuals in decentering may contribute to the prevention of anxiety and depression, and therefore, decentering may be an important skill to cultivate and strengthen.

Highlights.

• Negative life events and perseverative thinking are two well-established risk factors for anxiety and depression.

• Perseverative thinking mediates the longitudinal association between negative life events and internalizing symptoms.

• Decentering predicts the attenuated effects of negative life events and perseverative thinking on internalizing symptoms.

• Decentering also predicts the attenuated effects of negative life events on brooding.

• Decentering represents an important protective factor and target for clinical intervention.