Brief report: perseverative cognition and behaviors during the early months of the COVID-19 pandemic

Abstract

Objective

The onset of the COVID-19 pandemic engendered a host of different behavioral responses in key areas such as resource purchasing and compliance with COVID regulations. This study explores why participation in these behaviors may have varied by examining the role of perseverative cognition (i.e., worry, rumination).

Method

A representative sample of U.S. adults (N = 230, M_age = 45.4, 50% female) was recruited online early in the COVID-19 pandemic (June 2020), approximately three months into safety-related school and business shutdowns. Participants completed a series of questionnaires on COVID-19-related worry, non-specific rumination, and a range of behaviors observed during the early stages of the pandemic (e.g., excessive shopping, purchasing guns, following isolation guidelines). Exploratory principal axis factoring and reliability studies of the newly created COVID-19 worry and behavior scales were performed, followed by an examination of the associations between worrying, ruminating and COVID-19-related early behaviors.

Results

COVID-related worry and reflective rumination were positively associated with purchasing behavior (i.e., purchasing household items, items for security, and items for entertainment). COVID-related worry was also positively associated with compliance behavior (i.e., watching the news, following government directives strictly, and staying home but contacting family and friends remotely). Brooding rumination was not significantly associated with either type of behavior.

Conclusions

Perseverative cognition can help explain patterns in COVID-related behaviors, what motivated differences in these behaviors and, thereby, help explain how people coped during this trying time.

Keywords:

• Worry
• rumination
• COVID-19
• purchasing
• compliance
• coping

REVIEWING EDITOR:

• Daryl O’Connor, University of Leeds, United Kingdom

SUBJECTS:

• Introductory Psychology
• General Psychology
• Cognitive Psychology
• Environmental Psychology
• Health Psychology
• Social Psychology
• Psychological Science

Introduction

Within the intricate tapestry of behavioral responses to the unprecedented challenges posed by the COVID-19 pandemic, this paper explores the dynamic interplay between pervasive worry and a spectrum of behaviors. These behaviors, ranging from excess shopping and stockpiling to meticulous adherence to government safety guidelines, reveal nuanced dynamics shaping our collective response to the profound challenges of the time. The diversity in behavioral responses during the pandemic prompts an exploration of how different forms of perseverative cognition, specifically worry and rumination, may differentially influence individuals in times of heightened threat (Brosschot et al., 2006). In the Perseverative Cognition Hypothesis, Brosschot et al. (2006) explain that perseverative cognition (PC) represents repetitive thinking that extends the cognitive representation of a stressor, potentially contributing to variations in responses. The Generalized Unsafety Theory of Stress was subsequently developed to expand on how stress responses remain activated (Brosschot et al., 2016). In this theoretical model, Brosschot et al. (2016) describe that unconsciously, people have a feeling of unsafety and their default response is the stress response. Those who are unable to turn this response off also do not perceive safety, even when safety cues are present (Brosschot et al., 2016). Prior research has examined the impact of PC on behavior (Clancy et al., 2016, 2020). Specifically, the role of worry and rumination in health behaviors (e.g., physical activity, proper diet, sleep) has been studied as a possible route to poorer health outcomes (Clancy et al., 2016, 2020). These studies, although related to health, can inform our understanding of how PC may influence broader behavioral phenomena, such as those observed in the current study.

In a stressful situation, people may worry or ruminate when they perceive a lack of controllability (Brosschot et al., 2006). Worry, a form of repetitive thinking and psychological problem-solving, involves attempts to preemptively navigate the negative impact of an impending, uncertain stressor (Borkovec et al., 1983, 1998; Brosschot et al., 2006; Clancy et al., 2016). In parallel, rumination entails repetitive negative thoughts about past stressors and reflects an inward-focused, unproductive problem-solving approach aimed at understanding one’s negative emotions (Brosschot et al., 2006; Clancy et al., 2016; Hong, 2007; Nolen-Hoeksema et al., 2008). Reflective rumination, a subtype of rumination, denotes purposeful inward engagement for cognitive problem-solving to alleviate depressive symptoms, presenting a potentially adaptive coping style (Treynor et al., 2003, p. 256). Conversely, brooding rumination, another subtype, involves passively comparing one’s current situation with an unachieved standard, suggesting a potentially maladaptive coping strategy (Treynor et al., 2003, p. 256). The current investigation seeks to unravel how these distinct forms of perseverative cognition may shape and differentiate behavioral responses during times of heightened threat, such as the COVID-19 pandemic.

The onset of the COVID-19 pandemic presented specific stressors, such as stress relating to loss of income and to the unpredictability of the length of quarantine (Park et al., 2020). The general public responded with varying behaviors related to fear, worry, and rumination, such as changes in purchasing and compliance behaviors (Clemens et al., 2020; Coifman et al., 2021; Kemp et al., 2021). Observable responses included panic buying (Yuen et al., 2020), increased news consumption regarding COVID-19 (Stainback et al., 2020), and participation in preventive health behaviors (Liu, 2020). While not uniformly health-risk or health-promoting, these behaviors offer valuable insights into the diverse coping strategies employed during the pandemic. PC, identified as a potentially harmful cognitive pattern, exhibits problem-solving functions (Watkins, 2008) and extends beyond health behaviors to encompass everyday responses to stressful situations. Worry and rumination are closely related forms of repetitive thinking, not only in terms of appraisal but also in terms of problem-solving confidence and effort (Watkins et al., 2005). Therefore, this study further examines this relationship through COVID-specific worry and multiple subtypes of rumination. The behaviors under study are neutral in that they are not inherently health-risk or health-promoting. The associations between these forms of PC and behavior were examined to determine the reasons behind divergent responses to a common stressor, COVID-19. Notably, no behavior-specific hypotheses were posited, as the correlations between pandemic behaviors and PC were exploratory.

Method

Participants and procedure

Two hundred and thirty participants were recruited online from a nationally representative sample in June 2020 via the Qualtrics research service arm (https://www.qualtrics.com/research-services/online-sample/). Informed consent was obtained from all participants who completed a series of questionnaires in a larger Responses to COVID-19 study. In order to begin answering the study’s questionnaires, participants consented to participate in the study at the start of the survey. The researchers of this study did not obtain identifying information from participants recruited through the online research service, except for demographic information asked in the survey itself. All participant demographics can be found in Table 1. This survey study was determined and confirmed to be exempt from University of California-Irvine Institutional Review based on the Exempt Self-Determination Process. Please reference https://research.uci.edu/human-research-protections/do-you-need-irb-review/self-exempt/ for further details on the Institutional Review Board’s requirements. A G*power analysis (Faul et al., 2009) for a linear multiple regression was performed, specifying a medium effect size (f² = 0.15), a power of 0.80, and an alpha of 0.05 with 14 predictors. This analysis revealed a minimum recommended sample size of 135 participants. Of note, a meta-analysis on perseverative cognition and health behaviors reported very heterogeneous effect sizes, particularly in studies measuring worry (Clancy et al., 2016).

Table 1. Participant demographics.

Age (n = 229)^a	Mean	SD
18–87	45.40	17.41
Gender (N = 230)	Frequency	(%)
Male	115	50.00
Female	115	50.00
Ethnicity (N = 230)	Frequency	(%)
Asian or Asian American, including Chinese, Japanese, and others	8	3.48
Black or African American	18	7.83
Hispanic or Latino, including Mexican American, Central American, and others	38	16.52
White, Caucasian, Anglo, European American; not Hispanic	151	65.65
Native American, Eskimo, Aleut	2	0.87
Mixed; Parents are from two different groups	9	3.91
Other	4	1.74
Education (N = 230)	Frequency	(%)
Middle School	2	0.87
High School	43	18.70
College/Junior College	82	35.65
Graduate or Professional School	98	42.61
Other	5	2.17
Income (N = 230)	Frequency	(%)
<$25,000	35	15.22
$25,000–<$50,000	44	19.13
$50,000–<$75,000	31	13.48
$75,000–<$100,000	31	13.48
$100,000–<$150,000	39	16.96
>$150,000	40	17.39
Don’t know—not sure	2	0.87
Decline to respond	8	3.48

SD = standard deviation; n = number of included participants; N = total sample size.

^aOne age outlier was excluded from the age demographic analysis.

Measures

Covariates

Covariates included sex, age, education level, income level, ethnicity (using question 13 from the Multigroup Ethnic Identity Measure; Phinney, 1992), depression (using the Center for Epidemiological Studies-Depression CESD-10 scale; Radloff, 1977), and the following yes-no COVID-related questions: C1) Have you tested positive for COVID-19? (Either for an active virus or in an antibody test) and C2) If you have not been tested, do you suspect you might have COVID-19? C3) Do you know someone personally* who has been diagnosed with COVID-19? *personally means someone that you have directly communicated with within the last month (e.g., in person, on the phone, texted). C4) Do you know someone personally who has been hospitalized with COVID-19? C5) Do you know someone personally who has died from COVID-19? These variables were controlled for because they could have impacted how worried a participant was about COVID-19 infection and how much they ruminated during the pandemic.

COVID-19-related worry scale

To measure COVID-19-related worry, participants answered a 10-item scale developed by our laboratory. This scale was adapted from prior pandemic-related research studies and questions from COVID-related questionnaires that were being quickly developed early in the pandemic (Conway et al., 2020; Maunder et al., 2004; Reynolds et al., 2008; Rubin et al., 2014).

Responses were measured using a 5-point Likert scale ranging from 1 (not at all) to 5 (extremely). In this sample, the measure achieved adequate reliability (α = 0.89). The “COVID-19-related worry scale” asked, “During the COVID-19 pandemic, to what extent have you been worried about the following:” (1) Social isolation, (2) Getting infected with COVID-19, (3) Transferring COVID-19 to someone else, (4) Experiencing xenophobia and racism due to COVID-19, (5) Not having enough household supplies or food, (6) Paying bills, (7) Unemployment, (8) Getting adequate healthcare, (9) Being away from family, and (10) Being with family.

General rumination during the COVID-19 pandemic

General rumination during the COVID-19 outbreak was measured using the 10-item short-form version of the ruminative responses scale (RRS-10; Treynor et al., 2003). The RRS-10 is a valid and reliable measure of rumination that includes brooding and reflective rumination subscales (Treynor et al., 2003). Responses were measured using a 4-point Likert scale ranging from 1 (almost never) to 4 (almost always). For both factors, the scale items in this sample had good internal consistency (Brooding α = 0.86, Reflection α = 0.86).

COVID-19 pandemic behavior scale (CPBS)

A 10-item scale was created to measure common behaviors people participated in during the pandemic and the frequency of these behaviors. This scale was partly inspired by a CNBC news article published at the start of the pandemic (Whitten, 2020) that reported on changes in consumer behavior including stockpiling of food and household supplies, as well as purchases related to entertainment, connection with others, and protection from others. It was also inspired by behaviors of the general public at that time. The behaviors are not explicitly positive or negative in terms of their outcomes, but simply things that people were doing during a period of lockdown and high regulation to prevent disease spread.

Items on the “COVID-19 pandemic behavior scale” (CPBS) were assessed using a 5-point Likert scale from 1 (never) to 5 (a great deal). Both factors of this scale (explained more below) achieved adequate reliability in this sample (purchasing behavior α = 0.82, compliance behavior α = 0.79). The CPBS asked, “How often have you done any of these activities since the COVID-19 pandemic began?” in regards to the following behaviors: (1) Volunteering to help others (either individually or in a mutual aid group), (2) Purchasing large quantities of household supplies and/or food, (3) Purchasing items for security (cybersecurity software, home security, or buying guns/ammunition, etc.), (4) Purchasing items for entertainment, (5) Watching the news, (6) Staying on social media, (7) Following government directives regarding COVID-19 very strictly, (8) Staying home but still contacting family/friends through social media/calling/texting, (9) Not talking to anyone (isolating oneself) and staying home, and (10) Spending time with family (being physically present with family) and/or with friends.

Data analysis

An exploratory principal axis factor analysis was performed on the CPBS and the COVID-19-related worry scale. One factor was extracted for the worry scale, and two factors resulted for the CPBS. Descriptive statistics (e.g., mean, SD, and range) and Cronbach’s alpha scores were calculated for each scale and subscale (see “Measures” section) (Cronbach and Shavelson, 2004). Pearson’s r correlations were performed to test the associations between our variables of interest. Hierarchical multiple linear regressions were performed to predict compliance and purchasing factor scores on the CPBS. The first step included worry, brooding, and reflective rumination. Next, age, sex, ethnicity, income level, depression, and education level were included. Lastly, the five COVID-related covariates were included (C1–C5 described above).

Results

Factor analyses

Exploratory principal axis factoring of the COVID-19-related worry measure was performed. After viewing the scree plot, one factor was extracted, and conceptualized as representing worrying aspects of the COVID-19 pandemic (e.g., possible infection, loss of job/income, isolation). Additionally, exploratory principal axis factoring of the CPBS was performed. In addition to exploring the dimensionality of a new measure, this test was used to perform data reduction on the list of behaviors to provide a more parsimonious test of whether rumination and worry predict COVID-19-related behaviors. After assessing the scree plot, a two-factor solution was reached with one factor representing purchasing behaviors during the pandemic (items 2, 3 and 4) and another representing compliance with government directives to minimize the spread of the pandemic (items 5, 7 and 8). The CPBS items (6), (9) and (10) did not load cleanly onto a factor, and were dropped. Item 1 of the CPBS was dropped to maximize the interpretability of the purchasing factor.

Main analyses

Descriptive statistics were computed for the key predictor and predicted variables (Table 2). The average participant scored near the midpoint of every measure, other than for the compliance behavior measure, which had a higher mean. Scores on most measures were positively correlated. The two factors of the RRS (brooding and reflection) were highly correlated. Although this suggests that they may be better conceived of as unidimensional in this sample, both factors were retained for analysis because of the initial findings separating them (Treynor et al., 2003) and potentially differential relationships with the predicted variables, based on their differences in adaptiveness.

Table 2. Summary statistics.

Variable	Mean	SD	Range	Compliance	Purchasing	Brooding	Reflection
Compliance	3.80	1.02	1–5
Purchasing	2.57	1.15	1–5	0.264**
Brooding	2.16	0.81	1–4	0.023	0.517**
Reflection	2.03	0.79	1–4	0.001	0.576**	0.845**
Worry	2.56	1.03	1–5	0.191**	0.572**	0.554**	0.560**

^†p < 0.10, *p < 0.05, **p < 0.01.

Hierarchical multiple linear regressions

The main predictor variables accounted for much of the variation in the purchasing factor (Table 3, Model 1, R² = 42.2%, p < 0.001). COVID-19-related worry and the reflective rumination factor drove this effect. They were also uniquely significant after introducing demographic and COVID-related covariates (Model 2/3: Worry β = 0.384/0.378, ps < 0.001; Reflection β = 0.441/0.462, ps = 0.001). The brooding rumination factor coefficient estimate was minimal and did not reach statistical significance in any model. The COVID-related covariates did not reach significance.

Table 3. Hierarchical multiple linear regressions predicting COVID-19 purchasing behavior.

Predictor	Criterion variable Purchasing behavior (M1)			Criterion variable Purchasing behavior (M2)			Criterion variable Purchasing behavior (M3)
		95% CI			95% CI			95% CI
	B	Lower	Upper	B	Lower	Upper	B	Lower	Upper
Worry	0.409**	0.270	0.547	0.384**	0.251	0.517	0.378**	0.243	0.513
Brooding	0.002	−0.271	0.275	0.174	−0.095	0.442	0.155	−0.117	0.428
Reflection	0.541**	0.262	0.819	0.441**	0.181	0.702	0.462**	0.191	0.733
Age				−0.002	−0.007	0.002	−0.002	−0.007	0.003
Sex (male = 1)				−0.212^†	−0.445	0.022	−0.190	−0.428	0.048
Education				0.206**	0.069	0.343	0.201**	0.063	0.338
Ethnicity				−0.084	−0.192	0.024	−0.082	−0.192	0.028
Income				0.048	−0.014	0.109	0.048	−0.014	0.110
CESD-10				−0.030**	−0.051	−0.010	−0.030**	−0.052	−0.009
C1							−0.003	−0.763	0.757
C2							0.090	−0.327	0.507
C3							0.293	−0.087	0.672
C4							−0.297	−0.789	0.195
C5							−0.048	−0.558	0.462
R^2	0.422			0.523			0.530
ΔR^2	0.422**			0.101**			0.007

β = unstandardized coefficients; CESD-10 = Center for Epidemiological Studies-Depression-10; C1–C5 represent the COVID-related covariates.

†p < 0.10; *p < 0.05; **p < 0.01.

Unlike the purchasing factor, the main predictor variables failed to account for much of the variation in the behavioral compliance factor, although the model was statistically significant (Table 4, Model 1, R² = 5.4%, p = 0.006). COVID-19-related worry drove this effect and was also uniquely significant after introducing demographic and COVID-related covariates (Model 2/3: β = 0.294/0.289, ps < 0.001). Both rumination factor coefficient estimates were minimal and did not reach statistical significance in any model. Knowing someone who was diagnosed with COVID-19 positively predicted compliance behavior.

Table 4. Hierarchical multiple linear regressions predicting COVID-19 compliance behavior.

Predictor	Criterion variable Compliance behavior (M1)			Criterion variable Compliance behavior (M2)			Criterion variable Compliance behavior (M3)
		95% CI			95% CI			95% CI
	β	Lower	Upper	B	Lower	Upper	β	Lower	Upper
Worry	0.279**	0.123	0.435	0.294**	0.138	0.450	0.289**	0.131	0.447
Brooding	−0.002	−0.309	0.305	0.227	−0.088	0.542	0.203	−0.115	0.522
Reflection	−0.199	−0.512	0.115	−0.211	−0.517	0.095	−0.159	−0.475	0.158
Age				0.008**	0.002	0.013	0.008**	0.002	0.014
Sex (Male = 1)				0.117	−0.157	0.391	0.132	−0.146	0.410
Education				0.142^†	−0.018	0.302	0.135	−0.026	0.296
Ethnicity				0.035	−0.092	0.162	0.033	−0.096	0.162
Income				0.035	−0.037	0.108	0.035	−0.037	0.108
CESD-10				−0.035**	−0.060	−0.011	−0.038**	−0.063	−0.013
C1							0.122	−0.766	1.010
C2							−0.069	−0.556	0.418
C3							0.485*	0.042	0.928
C4							−0.233	−0.809	0.342
C5							−0.147	−0.743	0.448
R^2	0.054			0.152			0.171
ΔR^2	0.054**			0.098**			0.019

β = unstandardized coefficients; CESD-10 = Center for Epidemiological Studies-Depression-10; C1-C5 represent the COVID-related covariates.

†p < 0.10; *p < 0.05; **p < 0.01.

Discussion

The results of this study indicate that COVID-19-related worry is positively associated with both purchasing behavior and compliance behavior during the pandemic. Additionally, more general reflective (but not brooding) rumination was positively associated with purchasing behavior. These results are similar to previous research (Clemens et al., 2020; Coifman et al., 2021; Kemp et al., 2021). However, neither subtype of rumination was associated with compliance behavior. To add to prior research, the distinction between ruminative subtypes reveals that reflective rumination and worry can have similar motivating effects on behavior.

Individuals high in worry showed a higher level of behavioral engagement compared to those high in rumination, as higher worry was significantly associated with both higher purchasing and compliance behavior. People might worry or ruminate when they feel unable to cope with a threatening situation (Brosschot et al., 2006). The neutral behaviors we examined showed an association with PC, suggesting that these behaviors could have been ways of taking back control in an unprecedented situation (i.e., coping). A possible reason why high worry was linked with behavioral engagement could be that those who worry tend to have less confidence in their problem-solving techniques, thus pushing them to keep generating anticipatory solutions to their stressor (Borkovec et al., 1998; Davey, 1994; Davey et al., 1992; Hong, 2007). However, in regards to rumination, Burwell and Shirk (2007) found that brooding rumination was associated with disengaging coping strategies such as denial, and reflective rumination was associated with adaptive coping strategies such as problem solving in an adolescent sample. Similarly, Treynor et al. (2003) suggested that reflective rumination may be a more adaptive ruminative subtype compared to brooding rumination. Likewise, the results of the present study suggest that compared to brooding rumination, reflective rumination could indicate a stronger drive to participate in behavioral coping techniques. However, even though these behaviors could be problem-solving responses to PC, this does not necessarily mean that worry or rumination are beneficial (Marroquín et al., 2010).

Limitations and future directions

The study is cross-sectional, and therefore, the resulting associations do not imply causal relationships. Since the data under study was collected during an unprecedented major global threat and was collected at a specific time in the early months of the pandemic, the generalizability of these findings may be limited to times of pandemic and other major disasters. However, that being said, the results may point to how worry and rumination can influence people’s behaviors during responses to these major threats and disasters.

In regards to the data analyses, the worry, brooding, and reflective rumination variables were all placed in the same regression to allow for comparison of the behaviors they most strongly predicted. However, a limitation was that the brooding and reflective rumination variables were highly correlated, which could have impacted the reliability of the results. Additional analyses to specifically examine which worries led to which behaviors could not be performed because multiple testing of data could invalidate the results. Therefore, a future study could use free-response questions to investigate the reasoning behind perseveration and associated behaviors. An additional limitation was that measures for general worry and COVID-19-specific rumination were not used in this study. Therefore, a future study can include these measures to determine whether they are associated with a similar pattern of behavior as was observed in the present study.

In conclusion, the current study illuminates the nuanced relationship between PC and purchasing and compliance behaviors during times of crisis. By identifying the differential impacts of subtypes of PC on behavior, we underscore the adaptive nature of certain coping strategies in response to heightened threat. Our findings emphasize the importance of understanding individuals’ coping mechanisms and motivations, shedding light on whether these strategies ultimately aid or hinder them. Moving forward, continued research in this area can provide valuable insights into effective interventions and support systems for individuals facing similar challenges.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The dataset is available from the corresponding author upon reasonable request.

Additional information

Funding

This work was funded by the Undergraduate Research Opportunities Program at the University of California, Irvine.

Notes on contributors

Leeanne Qussiny

Leeanne Qussiny, B.S., obtained her Bachelor of Science in Biological Sciences from the University of California, Irvine (UCI). She performs research in Psychological Science with the Stress, Emotion, and Physical Health Lab at UCI under the supervision of Professor Sarah D. Pressman. To attain a deeper understanding of the outside factors impacting physical health, her research interests include the physiological, emotional, and behavioral responses to stress.

Cameron R. Wiley

Cameron R. Wiley, M.A, M.A. is a 5th year doctoral candidate in the Psychological Science department at the University of California, Irvine with a concentration in Health Psychology. His program of research examines how the regulation of emotion and stress influences cardiovascular functioning and health, the physiological mechanisms (e.g., autonomic, immune) that facilitate this connection, how these psychophysiological interactions differ across racial/ethnic groups to contribute to cardiovascular health disparities, and the roles of emotional and social well-being factors in mitigating those disparities.

Logan T. Martin

Logan T. Martin, M.A., is a graduate student in Affective Science at the University of California, Irvine in the department of Psychological Science. His research interests focus on meta-emotional beliefs, improving measurement and methods in psychological science, and the interconnections between emotion and health.

Sarah D. Pressman

Sarah D. Pressman, Ph.D., is a Professor of Psychological Science and the Associate Dean of Undergraduate Education at the University of California, Irvine. She has published extensively on the beneficial effects of emotional and social well-being for physical health and the mechanisms by which these experiences “get under the skin” to influence biology.