Reexamining trait rumination as a system of repetitive negative thoughts: A network analysis
Introduction
Everyone feels down sometimes. However, not everyone responds to such feelings in the same way. For example, some people tend to ruminate with repetitive, negative, and self-focused thoughts like “why am I so sad?” and “why can't I handle things better?” (Nolen-Hoeksema & Morrow, 1991). This cognitive-affective response style is strongly associated with risk, maintenance, and worsening of depressive symptoms (Nolen-Hoeksema, 2000, 1991; Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Chronic rumination increases the salience of negative events and impairs problem-solving, thus exacerbating and prolonging negative mood states (Donaldson, Lam, & Mathews, 2007; Joormann & Gotlib, 2010; Nolen-Hoeksema et al., 2008). Longitudinal, prospective, and experimental data converge: getting stuck in this mental habit of “self-critical moody pondering”, as it has been called, and getting stuck often, is a problem (Raes & Hermans, 2008; Watkins & Nolen-Hoeksema, 2014). And notably, frequent negative rumination and its consequences are not unique to depression, but are evident in anxiety disorders and related psychopathology as well (McLaughlin, Aldao, Wisco, & Hilt, 2014; McLaughlin & Nolen-Hoeksema, 2011).
Broadly considered as a core transdiagnostic feature of psychological disorders (e.g. Harvey, Watkins, Mansell, & Shafran, 2004; McLaughlin & Nolen-Hoeksema, 2011), rumination is a plausible target of treatment (Watkins, 2015). And there is preliminary evidence that rumination-focused therapies, such as mindfulness-based cognitive therapy (van Aalderen et al., 2012) and cognitive-behavioral approaches (Watkins et al., 2011), may improve treatment outcomes for depression. However, the results of such targeted interventions have been few and inconsistent, and in general, high rates of rumination predict slower treatment response, lower rates of recovery, and higher rates of relapse (Ciesla & Roberts, 2002; Schmaling, Dimidjian, Katon, & Sullivan, 2002). Given that rumination is so relevant for understanding emotional disorders and developing interventions, more work is needed to clarify the nature and malleability of rumination itself.
One limitation to extant efforts could be the conceptualization and subsequent measurement of rumination as a unitary construct. In reality, rumination is complex and multifaceted. For example, the typical definition includes components of perseveration, passivity, negativity, and self-focus. Furthermore, rumination about a negative experience can feel like a productive strategy for introspection and resolution (Papageorgiou & Wells, 2001). There have been important strides toward analyzing rumination to understand its function better. Factor analyses and other approaches have distinguished between adaptive versus maladaptive types of rumination or perseverative thinking, such as differentiating between abstract-evaluative and concrete repetitive thought (Watkins, 2008), or brooding and reflective pondering (Treynor, Gonzalez, & Nolen-Hoeksema, 2003). For example, abstract-evaluative thoughts (e.g. fixation on high-level, “why” aspects of one’ situation) and brooding (e.g. “What am I doing to deserve this?”) are more frequent and persistent in individuals with a history of major depression, those experiencing current symptoms, and those who eventually experience an episode than those without psychopathology. Such thinking patterns can increase the salience of negative thoughts and memories, delay problem solving or instrumental behaviors, and reduce cognitive and attentional flexibility (Nolen-Hoeksema et al., 2008; Watkins, 2008). In contrast, the links between mental health and more concrete thoughts (e.g. low-level, “how” details of one's situation) and reflection (e.g. analyzing one's thoughts) are less clear. Although there is evidence that such forms of self-reflection can be adaptive and benefit problem-solving, and that such thinking patterns are not elevated in individuals in remission from depressive episodes and are associated with less depression over time, they are associated with concurrent depressive symptoms and negative memory biases (Joormann, Dkane, & Gotlib, 2006; Nolen-Hoeksema et al., 2008). These models usefully categorize a person's general response style, the types of thoughts that are generally associated with mood symptoms, and the types of thoughts that typically have direct negative versus neutral or positive consequences.
However, in real life, people rarely fall into just one pattern of negative repetitive thinking. Instead, people exhibit different combinations of reflective, brooding, abstract, and concrete thoughts, for instance. Indeed, existing models fail to clarify why one person's pattern or combination of adaptive and maladaptive thoughts leaves them vulnerable to frequent, problematic rumination and associated mood symptoms, whereas another person's pattern of thinking does not. Although categorizing individual thoughts alone can be beneficial (e.g. adaptive or maladaptive, reflective or brooding, and abstract or concrete), such labels may become blurry once we consider how they all interact as a system.
We therefore propose reconsidering trait rumination from a network perspective. Currently, rumination and its sub-types are measured as single sum scores from self-report scales. As a result, component processes (e.g. items) or entire scales are treated as interchangeable and reflective of an underlying, latent response style. In some ways, this is perplexing as it is easy to generate examples of how component processes might interact (e.g., the more one criticizes oneself for failures, the more likely one is to brood about how sad one feels) and how these processes might be more or less influential on one's mood and circular thinking. Furthermore, some ostensibly neutral or even adaptive thinking styles from one factor, like reflection, may become problematic in relation to others. A network approach to reconceptualizing rumination and its components allows for these possibilities. The feature distinguishing this computational approach from factor analyses is that it characterizes rumination as a system of interacting components that do not have to have a common, underlying cause (Schmittmann et al., 2013). Accordingly, we can look for clues as to why some people get stuck in a pattern of ruminative responses and to what specific targets are ripe for intervention.
To do so, we used the items of the Ruminative Responses Scale (RRS; Nolen-Hoeksema & Morrow, 1991), the most common instrument for measuring trait rumination. There are important limitations to this approach. First, this is only one measure and thus results may vary with another scale. Second, some items do share conceptual overlap (e.g., going away to think about one's feelings and thinking about why one is feeling that way) as is typical of indices intended to reflect a conjectured latent, common cause. However, this intention notwithstanding such items need not measure the same process. Thus, although we take data-driven steps (described below) to reduce items that do appear to be measuring single processes, we still do not argue that the resulting network models completely capture the trait rumination construct. Nor do we argue that each item necessarily reflects a completely unique aspect of rumination. In fact, achieving these aims would likely require a rigorous, iterative process of including various self-report, behavioral, and other variables to devise a parsimonious and comprehensive measure for this purpose. We do argue that there is value in examining all items simultaneously as conceptual overlap is not synonymous with fungibility and simple sum scores likely occlude meaningful differences and interactions between components. The network approach is a fresh, data-driven way to gain new perspectives on how component processes or items cluster and relate to one another (individually and as sub-networks). Results can drive new hypotheses to be experimentally tested regarding different components of the rumination construct that might be most influential initiating or maintaining problematic response styles. The goal is to explore whether past approaches to quantifying rumination could be missing informative elements.
We have three primary aims: (1) explore whether and how RRS items interrelate in different ways, (2) test whether within this larger network RRS items constitute distinguishable communities (sub-networks) of processes, and (3) if so, are there particular items that function as “bridges,” i.e. processes that connect or are shared by communities. These analyses can help to highlight especially potent component processes that may foster broader vulnerability for problematic emotional and cognitive responses. Additionally, community detection and bridge analyses could highlight thought patterns that make a person more likely to tip from adaptive reflection and introspection to maladaptive brooding. Overall, this new lens on rumination is exploratory. Each analysis has its strengths and limitations and is intended to provide new hypotheses about plausible causal connections between components of rumination.
Section snippets
Participants
De-identified data came from adults (N = 403, 231 women) who enrolled in a research program concerning cognition, emotion, and exercise and who completed the RRS between 2014 and 2017 (Bernstein & McNally, 2016, 2017, 2018). Participants were between the ages of 18 and 58 (Mage = 24.59, SD = 7.27) and 102 identified as Hispanic or Latino (25.31%). The self-reported racial breakdown of the sample is as follows: 58.56% Caucasian or white, 10.67% African American or black, 18.11% Asian American or
Results
RRS items and the abbreviated names to be used in figures are included in Table 1.
Discussion
We first examined the interplay between components of a trait rumination measure (RRS) by computing a graphical LASSO network. The pairs or clusters of nodes that emerged as most strongly interconnected in the graphical LASSO were clinically interpretable. Items that intuitively seem most related mechanistically emerged with the strongest edges, such as brooding about feelings of sadness as well as loneliness or repetitive self-criticism with self-directed anger and wishing recent situations
Conflicts of interest
Authors have no conflicts of interest to disclose.
Funding Acknowledgment
This work was supported by the American Psychological Association of Graduate Students [Junior Scientist Fellowship, Psychological Science Research Grant], Psi Chi [Psychological Science Research Grant], the Association for Psychological Science Student Caucus (APSSC) [APS Student Grant], and Harvard University [Stimson Fund Research Grant, Barbara Ditmars Restricted Funds Research Award, Allport Fund Research Award], all awarded to the first author.
References (56)
- et al.
Exploring behavioral pattern separation and risk for emotional disorders
Journal of Anxiety Disorders
(2018) - et al.
The network of conscientiousness
Journal of Research in Personality
(2016) - et al.
Rumination and attention in major depression
Behaviour Research and Therapy
(2007) - et al.
Mapping network connectivity among symptoms of social anxiety and comorbid depression in people with social anxiety disorder
Journal of Affective Disorders
(2018) - et al.
An integrative network approach to social anxiety disorder: The complex dynamic interplay among attentional bias for threat, attentional control, and symptoms
Journal of Anxiety Disorders
(2016) - et al.
A network perspective on comorbid depression in adolescents with obsessive-compulsive disorder
Journal of Anxiety Disorders
(2018) - et al.
Adaptive and maladaptive components of rumination? Diagnostic specificity and relation to depressive biases
Behavior Therapy
(2006) - et al.
Rumination as a transdiagnostic factor in depression and anxiety
Behaviour Research and Therapy
(2011) - et al.
Metacognitive beliefs about rumination in recurrent major depression
Cognitive and Behavioral Practice
(2001) - et al.
Deconstructing the construct: A network perspective on psychological phenomena
New Ideas in Psychology
(2013)
Psychological treatment of depressive rumination
Current Opinion in Psychology
The efficacy of mindfulness-based cognitive therapy in recurrent depressed patients with and without a current depressive episode: A randomized controlled trial
Psychological Medicine
Unpacking rumination and executive control: A network perspective
Clinical Psychological Science
Acute aerobic exercise buffers against negative affect
Cognition & Emotion
Acute aerobic exercise hastens emotional recovery from a subsequent stressor
Health Psychology
A prospective study on how symptoms in a network predict the onset of depression
Psychotherapy and Psychosomatics
Self-directed thought and response to treatment for depression: A preliminary investigation
Journal of Cognitive Psychotherapy
The igraph software package for complex network research
InterJournal Complex Systems
Estimating psychological networks and their accuracy: A tutorial paper
Behavior Research Methods
qgraph: Network visualizations of relationships in psychometric data
Journal of Statistical Software
Personalized network modeling in psychopathology: The importance of contemporaneous and temporal connections
Clinical Psychological Science
Exploring the idiographic dynamics of mood and anxiety via network analysis
Journal of Abnormal Psychology
Bayesian model choice and information criteria in sparse generalized linear models
glasso: Graphical lasso-estimation of Gaussian graphical models. R package version
Depression sum-scores don't add up: Why analyzing specific depression symptoms is essential
BMC Medicine
Graph drawing by force directed placement
Software: Practice and Experience
Cognitive behavioural processes across psychological disorders: A transdiagnostic approach to research and treatment
Deconstructing trait anxiety: A network perspective
Anxiety, Stress & Coping
Cited by (42)
A network analytic investigation of avoidance, dampening, and devaluation of positivity
2023, Journal of Behavior Therapy and Experimental PsychiatryCrying over spilled milk? A network analysis of aversive well-being comparison, brooding rumination and depressive symptoms
2023, Journal of Affective DisordersComorbidity of loneliness and social anxiety in adolescents: Bridge symptoms and peer relationships
2023, Social Science and Medicine“Don't [ruminate], be happy”: A cognitive perspective linking depression and anhedonia
2023, Clinical Psychology ReviewA network approach to climate change anxiety and its key related features
2023, Journal of Anxiety DisordersCitation Excerpt :To ensure that none of the variables included in the network overlap conceptually, we implemented a data-driven method to identify potentially redundant pairs of variables. To do so, we followed the procedure described in recent publications (e.g., Bernstein et al., 2019; Heeren & McNally, 2018). First, we tested whether our correlation matrix was positive definite, ensuring that our variables were not a linear combination of other variables.