Cognitive reappraisal is renowned as an adaptive emotion regulation strategy associated with better mental health and wellbeing (Aldao et al., 2010; Gross & John, 2003; Hu et al., 2014). Reappraisal involves reinterpreting a situation in a way that changes its meaning to modify the valence, intensity, or duration of an emotional response. For example, when faced with a stressful situation, reappraisal might involve thinking about it in ways that help one stay calm, such as considering how a cancelled event will free up time for other important activities. However, reappraisal does not always have the presumed benefits. In a longitudinal survey that tracked the mental health of a representative sample of Australians, people’s self-reported reappraisal use was not associated with depression and anxiety symptoms across the early acute lockdown phase of the COVID-19 pandemic in 2020 (Dawel et al., 2021). Overall, the literature now points to the possibility that reappraisal may be beneficial only for some people and in some situations (Ford et al., 2017; Ford & Troy, 2019). Therefore, a key aim for the field is building a precise scientific model that can be used to predict for whom, when, and what reappraisal is helpful, to generate personalized emotion regulation guidance and interventions (Doré et al., 2016). Using data from Dawel et al. (2020) and a new wave of data collected in March 2021, the primary aim of the present study was to investigate if the relationship between how much people use reappraisal (frequency of reappraisal use) and mental health and wellbeing is moderated by factors that might influence how well reappraisal achieves its intended emotion regulation goals (reappraisal effectiveness). Note, the term reappraisal always refers to cognitive reappraisal in this article.
Cognitive Reappraisal and Psychological Health and Wellbeing
Meta-analyses and multiple individual studies indicate greater reappraisal use is associated with a variety of adaptive psychological outcomes, including lower depression and anxiety symptoms, higher positive affect, and better interpersonal functioning (Aldao et al., 2010; Gross & John, 2003; Hu et al., 2014). However, on average, the effects associated with reappraisal use are small (e.g., r = − .05 to − 0.17, Aldao et al., 2010) and several individual studies have found no association between reappraisal use and mental health (e.g., for depression, Arditte and Joormann, 2011; De France et al., 2019; for PTSD, Khan et al., 2021). In the COVID-19 context, evidence linking reappraisal and mental health has also been mixed. Some observational studies found greater reappraisal use was associated with decreased pandemic-related mental health impacts (Kuhlman et al., 2021; Preti et al., 2021; Tyra, Ginty et al., 2021; Tyra, Griffin et al., 2021) and two intervention studies found that reappraisal training was able to reduce stress in parents during the pandemic (Preuss et al., 2021) and momentary negative affect about COVID-19 in adults across the globe (Wang et al., 2021). However, other research found evidence for this association was mixed (Low et al., 2021; Preti et al., 2021) or that reappraisal use was not associated with psychological outcomes (Dawel et al., 2021; Dicker et al., 2022; Tyra, Ginty et al., 2021; Tyra, Griffin et al., 2021; Venanzi et al., 2022).
The clear variation in these findings aligns with recent arguments that no emotion regulation strategy is universally helpful (or harmful), highlighting interactions between person-level and situation-level factors within the context of any given regulation strategy (Doré et al., 2016; Ford & Troy, 2019; Sahi et al., 2022). For example, the situation-strategy fit hypothesis suggests that a strategy is likely to be effective only to the extent that it meets the needs of the specific situation (Haines et al., 2016; Troy et al., 2013). Supporting this argument, there is considerable evidence that the benefits of reappraisal vary depending on people’s goals, characteristics, and circumstances (e.g., Diefenbach et al., 2022; Gruber et al., 2012; McRae, Ciesielski et al., 2012). In some instances, reappraisal may even have negative outcomes. Many people report feeling worse after unsuccessful attempts at reappraisal, which may occur due to poor reappraisal skills or limited realistic possibilities for reappraising the situation (Ford & Troy, 2019). Therefore, our original finding that reappraisal use was not associated with depression or anxiety (Dawel et al., 2021) may have been due to the conflation of benefits for some participants with negative effects for others. Moderation analyses have the potential to reveal such opposing effects and to identify the circumstances under which reappraisal use may promote mental health and wellbeing. In addition, it is likely that reappraisal use alone is not sufficient, and that the effectiveness of a person’s reappraisal strategy—how well it achieves its intended emotion regulation goals—also plays an important role.
Potential Moderators
It is clear that the effectiveness of reappraisal as an emotion regulation strategy varies across individual traits and circumstances (Gross & John, 2003; Kuhlman et al., 2021; Preuss et al., 2021). Factors likely to moderate the specific association between reappraisal and mental health and wellbeing are those that most strongly impact reappraisal effectiveness. These include an individual’s reappraisal ability—their ability to effectively use reappraisal to modify their emotional response. Importantly, individual differences in reappraisal ability have been found to correlate with psychological wellbeing (McRae, Ciesielski et al., 2012) and the quality (e.g., more detailed and plausible alternative ways of thinking about a situation) of individual reappraisals impacts their effectiveness (Southward et al., 2022). In the current study, our broad hypothesis was that the association between reappraisal and mental health and wellbeing would be strongest at the level of a moderating variable where reappraisal is predicted to be most effective (e.g., a stronger association at high compared to low self-efficacy, or at low compared to high neuroticism). We selected eight individual difference variables available in our dataset that could be expected to influence reappraisal effectiveness. These were self-efficacy, emotion differentiation, difficulty identifying feelings, neuroticism, stress, SES, an inclination for analytic thinking, and age. The rationale developed from the literature for selecting each of these variables is outlined below.
For self-efficacy, Goldin et al. (2012) found that higher self-efficacy for reappraisal use predicted improvement in social anxiety symptoms with cognitive behaviour therapy. This finding suggests that higher self-efficacy may be associated with greater reappraisal effectiveness. Our study did not include a measure of self-efficacy specific to reappraisal. However, we did include a general measure of self-efficacy (Pearlin Mastery scale; Pearlin and Schooler, 1978) and general self-efficacy tends to be related to self-efficacy in specific domains (Luszczynska et al., 2005). We therefore hypothesized that the association between reappraisal use and mental health and wellbeing would be stronger at higher levels of self-efficacy.
There are also good theoretical and empirical reasons to predict that people’s emotional experiences and abilities will influence reappraisal effectiveness. The affect-as-information theory argues that emotionally skilled people can use their emotions to guide adaptive responses (Schwarz, 1990), which may include an ability to effectively reappraise a situation. Two such skills are the ability to identify feelings clearly and the ability to precisely differentiate between similar emotions. For example, differentiating anger from contempt rather than conflating them into “feeling bad”. More generally, people experiencing intense emotions, such as those with high trait neuroticism or experiencing high stress, may have difficulty reappraising their situation (e.g., because they have fewer cognitive resources available; Sheppes, 2014). For example, Troy et al. (2010) found reappraisal ability was associated with reduced depression in people with lower but not higher stress levels (but cf. Langer et al., 2021). We therefore hypothesized that the association between reappraisal and mental health and wellbeing would be stronger at high levels of emotion differentiation and low levels of difficulty identifying feelings, neuroticism, and stress.
SES was also included as a potential moderator as prior research has found greater reappraisal is associated with reduced depression (Troy et al., 2017) and anxiety (Hittner et al., 2019) for people with low but not high SES. The underlying theoretical argument is that reappraisal is more helpful in uncontrollable situations that cannot be modified by other means (e.g., problem-solving, Troy et al., 2013, 2017); low SES is associated with lower control over one’s circumstances (Kraus et al., 2012). While the COVID-19 pandemic is an uncontrollable event for everyone, those with fewer socioeconomic resources are likely to be most affected. This construct of controllability of one’s external situation differs from self-efficacy, which refers to one’s perceived ability to take effective action to achieve tasks or goals. We therefore hypothesized that the association between reappraisal and mental health and wellbeing would be stronger at lower SES levels, with SES serving as a proxy for controllability.
Another potential moderator we considered is an inclination for analytic thinking. Reappraisal ability is related to general cognitive abilities (McRae, Jacobs et al., 2012; Mohammed et al., 2022; Orgeta, 2009; Toh and Yang, 2022; but cf. Gil et al., 2022) and engages brain regions associated with executive function in the prefrontal cortex, downregulating responses in emotion-associated regions, including the amygdala (Goldin et al., 2008; Steward et al., 2021). Thus, we hypothesized the association between reappraisal and mental health and wellbeing would be stronger for people who were more inclined to engage in analytic thinking.
Finally, reappraisal effectiveness may improve with age, potentially due to practice and skill development, as has been observed in adolescents (Decicco et al., 2012, 2014; McRae et al., 2012). To our knowledge, only one study has investigated associations between age and reappraisal effectiveness across the adult lifespan, finding no significant differences (Livingstone & Isaacowitz, 2021). Based on the potential for skill development, we predicted the association between reappraisal and mental health and wellbeing would be stronger with increasing age. We analyzed age first as a continuous variable, and then in age terciles because the brain regions involved in executive function, which are critical to reappraisal (Goldin et al., 2008; Steward et al., 2021), deteriorate across the last few decades of life (Nyberg et al., 2010), potentially offsetting learning benefits for reappraisal in older adults (i.e., it was possible reappraisal benefits might only be seen for middle-aged compared to young adults).
Additional Outcome Measures
Dawel et al. (2021) investigated associations of reappraisal with depression and anxiety. However, these data are from a larger study that includes other measures of potentially relevant psychological outcomes. Namely, wellbeing (WHO-5; Topp et al., 2015), loneliness (DJGLS; Gierveld & Van Tilburg, 2006), and functional impairment caused by COVID-19 across work and social domains (WSAS; Mundt et al., 2002). Regarding wellbeing, experimental evidence indicates reappraisal is more effective for increasing positive affect than decreasing negative affect (McRae, Ciesielski et al., 2012). Gutiérrez-Cobo et al. (2021) also found reappraisal measured before the pandemic predicted higher engagement in positive activities during COVID-19 lockdowns, which ultimately benefited affective happiness levels. Thus, reappraisal may have promoted positive aspects of psychological wellbeing more strongly than it decreased negative affect and distress during the pandemic. Other recent work has demonstrated a strong association between emotion regulation strategy use (including reappraisal) and loneliness (Preece et al., 2021). Perceived loneliness is associated with an increased risk of depression, anxiety, suicide, dementia, and premature mortality (Holt-Lunstad et al., 2015; Leigh-Hunt et al., 2017), and can impact social and other areas of functioning (Cacioppo & Cacioppo, 2018). Therefore, the development of interventions to reduce loneliness remains a high priority. Finally, we considered whether reappraisal may have helped people cope with impacts related to the pandemic specifically. The COVID-19 pandemic represents an uncontrollable and unexpected event, and it has been argued that reappraisal may be particularly important for regulating emotions in circumstances that are uncontrollable (Troy et al., 2013, 2017). We therefore predicted that reappraisal would be associated with higher wellbeing and lower levels of loneliness and functional impairment.
Present Study
The present study aimed to elucidate for whom and what reappraisal use is adaptive during times of crisis. We anticipated that, by investigating multiple potential moderators (self-efficacy, emotion differentiation, difficulty identifying feelings, neuroticism, stress, SES, inclination for analytic thinking, age) across a range of psychological outcomes (depression, anxiety, loneliness, functional impairment, wellbeing), we might build a more nuanced framework for personalizing reappraisal interventions. Data are from The Australian National COVID-19 Mental Health, Behaviour and Risk Communication Survey, comprised of seven fortnightly waves of data collection from late March to June 2020, and an eighth follow-up wave in March 2021. Our prospective design controlled for levels of the outcome variables at wave 1, to predict these outcomes at two later time points (wave 7 at three months and wave 8 at 12 months). We opted to analyze outcomes at both these waves because it is important to understand how changes in the effects of interest vary over short and longer periods. We also included gender and age as covariates in our models, except where age was the moderator variable, because both are associated with mental health and wellbeing (gender; Solmi et al., 2022; Steel et al., 2014). Although the total evidence suggests emotion regulation strategies are similarly related to psychological outcomes across men and women (Nolen-Hoeksema, 2012), some previous findings indicate there may be gender differences in emotion regulation use (McLaughlin et al., 2011; Nolen-Hoeksema, 2012; Preston et al., 2022; Scheibe et al., 2015; Urry & Gross, 2010), the underlying mechanisms of regulation (McRae et al., 2008), and moderator effects (Jiang et al., 2022; Preston et al., 2022). Table 1 summarizes our hypothesized relationships between reappraisal and the five outcome variables and Table 2 summarizes our hypotheses about how these associations will be moderated. To scaffold readers through the complexity of our results, we also pre-emptively summarize the core findings that map onto each hypothesis in the final columns of Tables 1 and 2.
Table 1
Hypothesized associations between reappraisal and outcome variables
Outcome variable | Measure | Association predicted with reappraisal | Association found with reappraisal |
|---|---|---|---|
Depression | PHQ-9 | Negative | None |
Anxiety | GAD-7 | Negative | None |
Loneliness | DJGLS | Negative | Negative |
COVID-induced functional impairment | WSAS | Negative | Positive |
Wellbeing | WHO-5 | Positive | Positive |
Table 2
Hypothesized moderation effects
Moderator | Measure | Stronger association predicted at… | Stronger association found at (for outcome)… |
|---|---|---|---|
Self-efficacy | Pearlin Mastery Scale | Higher self-efficacy | Lower self-efficacy (wellbeing) |
Emotion differentiation | ICC | Higher emotion differentiation | — |
Difficulty identifying feelings (DIF) | Toronto Alexithymia Scale-DIF subscale | Lower difficulty identifying feelings | Higher difficulty identifying feelings (loneliness, impairment, wellbeing) |
Neuroticism | BFI-10-neuroticism subscale | Lower neuroticism | Higher neuroticism (wellbeing) |
Stress | Mean stress rating | Lower stress | Higher stress (wellbeing, impairment) |
SES | Household income | Lower SES | — |
Analytic thinking | Adapted Cognitive Reflection Task (CRT) | Correct responses (requiring analytic thinking) | Correct responses (loneliness, wellbeing) |
Age (continuous) | Age in years | Older age; OR | Younger age (impairment) |
Age (categorical) | Tercile age groups | In middle-aged vs. young adults |
Method
Transparency and Openness
The study is pre-registered at: https://osf.io/xq63d/?view_only=b5a8e7c708c8497fa319262a775aa04c. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study, and we follow Journal Article Reporting Standards (Kazak, 2018). Data were analyzed using IBM SPSS Statistics and R (R Core Team, 2022). Due to the sensitive nature of some of the data collected in this study, the ethics protocol does not allow open access to the full dataset. However, data can be provided directly to other researchers who propose studies in line with the original protocol, and the authors are committed to providing the data used in the current study to any researcher who wishes to check or reanalyse our findings.
Sample and Survey Design
Samples were drawn from the longitudinal Australian National COVID-19 Mental Health, Behaviour and Risk Communication Survey (ANU COVID-19 MHBRCS; Dawel et al., 2020) on the basis that participants had at least one of the outcome measures available at wave 7 or 8 (Table 3 and Supplement S1). The first wave of data was collected in March 2020. Subsequent waves were collected fortnightly thereafter until June 2020, with an eighth wave collected in March 2021 to assess longer-term impacts of the COVID-19 pandemic. The full study protocol and survey measures are provided at: https://psychology.anu.edu.au/research/projects/australian-national-covid-19-mental-health-behaviour-and-risk-communication-survey. Although the sample size was determined by the available data, we also conducted post-hoc power analyses to ensure our moderation models had sufficient power to test our hypotheses (Supplement S2). Reliability for the measures is reported in Supplement S3.
Table 3
Sample Demographics
Wave 7 | Wave 8 | |
|---|---|---|
N | 752 | 512 |
% Male | 52.9 | 53.5 |
% Female | 46.9 | 46.3 |
Age M (SD) | 50.7 (16.0) | 52.3 (15.6) |
Age range | 18–87 | 19–87 |
Cognitive Reappraisal
Cognitive reappraisal was measured at waves 2, 4, and 6 using the 6-item cognitive reappraisal subscale of the Emotion Regulation Questionnaire (ERQ-reappraisal; Gross and John, 2003). Participants rated their agreement with statements describing to what extent they had used certain strategies for regulating emotions over the last two weeks (e.g., “I controlled my emotions by changing the way I was thinking about the situation I was in”) from Strongly Disagree (1) to Strongly Agree (7). Item scores were summed separately for each wave1 and then averaged across waves 2, 4, and 6 to produce an average total reappraisal score, which was used in all models. Higher scores indicate greater reappraisal use.
Mental Health and Wellbeing Outcome Measures (for Waves 1, 7, and 8)
For these measures, item scores were summed to produce a total score at each wave. Higher scores indicate higher levels of depression or generalized anxiety symptoms, loneliness, COVID-induced functional impairment, or wellbeing.
Patient Health Questionnaire-9 (PHQ-9; Spitzer et al., 1999). Depression symptoms were measured using the 9-item PHQ-9. Participants reported how frequently they experienced symptoms associated with depression over the last two weeks. Response options were Not at all (0), Several days (1), More than half the days (2), and Nearly every day (3).
Generalized Anxiety Disorder-7 (GAD-7; Spitzer et al., 2006). Generalized anxiety symptoms were measured using the 7-item GAD-7. Participants reported how frequently they experienced symptoms associated with generalized anxiety disorder over the last two weeks. Response options were the same as for the PHQ-9.
De Jong Gierveld Loneliness Scale (DJGLS; Gierveld & Van Tilburg, 2006). Loneliness was measured using the 6-item DJGLS. Participants rated how much the items applied to them over the last two weeks, from No! (1) to Yes! (5). Items 4, 5, and 6 were reverse coded.
Work and Social Adjustment Scale (WSAS; Mundt et al., 2002), used to measure level of functional impairment due to COVID-19. Participants rated the level of impairment COVID-19 had caused them over the last two weeks across five life domains (ability to work, home management, social leisure activities, private leisure activities, ability to form and maintain close relationships) from Not at all (0) to Very severely (8).
World Health Organisation-5 Wellbeing Index (WHO-5; Topp et al., 2015). The 5-item WHO-5 assessed how frequently participants experienced signs of psychological wellbeing over the last two weeks. Response options were All of the time (5), Most of the time (4), More than half of the time (3), Less than half of the time (2), Some of the time (1), and At no time (0).
Moderator Measures
Self-efficacy (Pearlin Mastery Scale; Pearlin and Schooler, 1978). Self-efficacy was measured at wave 4 with the 7-item Pearlin Mastery Scale, using a 7-point scale from Strongly agree (1) to Strongly disagree (7). Item scores were summed to produce a total score.
Emotion differentiation (Barrett et al., 2001). Participants rated how strongly they felt five negative emotions (angry, anxious, disgusted, fearful, sad) from Not at all (0) to Extremely strongly (6) in response to 12 images depicting the health and work (6 images) and social connection (6 images) impacts of the COVID-19 pandemic (also see Dawel et al., 2023). Participants viewed one image of each type at each wave from waves 2 to 7. Emotion differentiation was calculated as the intraclass correlation coefficient (ICC3,k)2 of each participant’s negative emotion ratings across the 12 images. An ICC was calculated only if participants responded to at least 10 images. ICCs were reverse-scored (see Erbas et al., 2014; Kalokerinos et al., 2019) so that higher values indicated higher differentiation.
Difficulty Identifying Feelings subscale of the Toronto Alexithymia Scale (TAS-DIF; Leising et al., 2009). Difficulty identifying feelings was measured at waves 1 and 7 with the 7-item TAS-DIF. Item scores were summed in each wave, then averaged to produce a total score.3
Neuroticism subscale of the Big Five Inventory-10 (BFI-10-N; Rammstedt and John, 2007). Neuroticism was measured at wave 2 with the BFI-10-N, using a 5-point scale from Disagree strongly (1) to Agree strongly (5). The first item was reverse scored and then added to the second item score to produce a total score.
Stress. Stress was measured using a single purpose-developed item. Participants were asked to indicate “Over the last 2 weeks, to what extent have you experienced stress in your life (at home and work)?” from Not at all (1) to Extremely (6). Item scores were summed separately for waves 2, 4, and 6 and then averaged to produce a total score.
SES (household income). SES was measured as self-reported household income at wave 1. Participants were asked, “Before tax is taken out, what is the present income for your household?”. Household was defined as “a person living alone or a group of people (including family, spouse/partner, children, group household) who usually live together and share or pool resources (e.g., money, food) in some way)”. Response options were 0-$300 (1), $300-$575 (2), $576-$1075 (3), $1076-$1700 (4), $1701–2400 (5) or More than $2400 (6).
Cognitive Reflection Task Item (CRT; Frederick, 2005). Inclination for analytic thinking was measured at wave 6 with a modified CRT item that replaced lilipads with virus immunity. Participants were asked, “Suppose that, in a population, people are starting to gain immunity from the virus. Every day, the number of people who have immunity doubles. If it takes 48 days for half the population to become immune, how many days would it take for the entire population to become immune?”. Participants responded by entering a number. Responses were scored as correct (1) or incorrect (0).
Age. Age in years was reported at wave 1. To test our age hypothesis that the association between reappraisal and outcome measures would be stronger in middle-age compared to young adulthood, we split our wave 7 and 8 samples into terciles by age in years (young adults at w7: < 42 years, N = 253; w8: < 43 years, N = 168; middle-aged adults at w7: 42–59 years, N = 244; w8: 43–60 years, N = 164; and older adults at w7: > 59 years, N = 254; w8: > 60 years, N = 179).
Data Analysis Overview
We first examined bivariate relationships between all measures. No multicollinearity was present between the predictor (i.e., cognitive reappraisal), moderator, and covariate measures.4 As our outcome measures were all non-normally distributed (Supplement S4), we report Spearman’s ρ correlations for the 90 models outlined in Fig. 1. We then performed moderator analyses using the PROCESS (v4.1; Hayes, 2022) path analysis macro in SPSS (version 25; IBM Corp., 2017). Each model tested if one of the nine moderator variables changed the relationship between cognitive reappraisal and one of the five outcome measures at either wave 7 or wave 8, after accounting for scores on that outcome measure at wave 1, gender, and age (unless the moderator variable was age). Each of the nine moderator variables was entered one at a time for each of the five outcome variables.
Fig. 1
Moderation Models
×
![]()
Outliers were identified via regression analyses for each model. Data points were excluded from a model if they exceeded at least two of Mahalanobis (p < .001), Cook’s, or Leverage distanced cut-off scores. Cut-off scores varied across models, depending on the number of participants and continuous predictor variables included. Less than 3% of data were identified as outliers in each model. Results in the main text are for models with outliers removed. Models with and without outliers are reported in full in Supplement B, showing retaining outliers resulted in only minor variations. The alpha significance level was Bonferroni-corrected for the five outcome measures to α = 0.01 for all analyses, with the exception that moderator effects were investigated further if they met a more liberal criterion of α = 0.05 to ensure no significant underlying simple effects were concealed. The decision to Bonferroni-correct for the five outcome measures but not the nine moderator tests aimed to appropriately limit Type I errors without over-inflating Type II errors (Perneger, 1998).
Results
Bivariate Correlations
Figure 2 presents Spearman’s ρ correlations separately for wave 7 and 8. As in Dawel et al. (2021), reappraisal was not significantly correlated with depression or anxiety. However, reappraisal was significantly correlated with the other three outcome measures, which were not investigated in Dawel et al. (2021). Specifically, higher reappraisal was associated with higher functional impairment due to COVID-19, as well as higher wellbeing and lower loneliness in both the wave 7 and 8 samples. Note, the direction of association for functional impairment was opposite to that hypothesised. The moderator and covariate variables mostly correlated with the outcome measures in the expected directions, except there were very few significant correlations for income. The correlations among the outcome measures were also consistently significant in the expected directions. Namely, depression, anxiety, loneliness, and impairment were positively correlated with one another and negatively correlated with wellbeing. We also observed non-perfect correlations within each of the outcome measures from Wave 1 to Wave 7 and 8, indicating there was change in these variables over time (also see Batterham et al., 2021 for latent growth modelling showing change in depression and anxiety over time in this sample).
Fig. 2
Spearman’s ρ Correlations Between Measures
Note. CRT coded as 0 = incorrect, 1 = correct. Gender coded as 1 = male, 2 = female. Note, there were too few non-binary participants to analyse this group separately. * p < .05. ** p < .01. *** p < .001
×
![]()
Moderation Analyses
Table 4 summarises the results of the moderation models, which are provided in full in Supplement B. Overall, the model R2 values indicate they explained 24.4% to 58.3% of the variance in the outcome measures, all ps < 0.001. Relationships between reappraisal and the final wave outcome scores were significant in several instances for wellbeing (w7 = 8/9 models; w8 = 5/9) and functional impairment (wave 7 = 2/9 models; w8 = 5/9), and in one instance for anxiety at wave 7 (self-efficacy model). Reappraisal was never a significant direct predictor of final wave scores for depression or loneliness.
Relationships between the moderation and outcome variables were also generally consistent with the correlational findings. Higher self-efficacy was significantly associated with lower depression, anxiety, loneliness, and impairment, and higher wellbeing. Associations for emotion differentiation, age, and inclination for analytic thinking were in the same direction, but weaker and less consistently significant. Conversely, higher difficulty identifying feelings, neuroticism, and stress were all significantly associated with higher depression, anxiety, loneliness, and impairment, and lower wellbeing (except for neuroticism in the wave 8 depression model). The full models in Supplement B show results for the covariates were also generally consistent with the correlational findings. Wave 1 scores invariably predicted final wave (w7 or w8) scores for the outcome measures, all ps < 0.001.
Table 4
Summary of Moderator Models Predicting Outcomes at Waves 7 and 8
Outcome measure effect | Moderator variable | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
i. Self-efficacy | ii. Emotion differentiation | iii. Diff. ID feelings | iv. Neuroticism | v. Stress | vi. SES | vii. Analytic thinking | viii. Age in years (continuous) | ix. Age tercile (young/middle/older) | |||
PHQ-w7 | |||||||||||
Reappraisal | 0.046 | -0.004 | 0.008 | 0.001 | -0.044 | -0.009 | 0.022 | -0.003 | -0.048 | ||
Moderator | -0.167*** | -2.592** | 0.310*** | 0.287** | 2.019*** | -0.050 | -0.677 | -0.042*** | mid vs. young = 0.771*, mid vs. old = -0.709 | ||
Moderator*reappraisal | 0.000 | -0.076 | 0.003 | -0.007 | -0.017 | 0.005 | -0.112 | -0.003 | mid vs. young = 0.109, mid vs. old = 0.049 | ||
PHQ-w8 | |||||||||||
Reappraisal | 0.014 | -0.001 | -0.005 | 0.003 | -0.036 | -0.005 | 0.030 | 0.012 | -0.040 | ||
Moderator | -0.159*** | -2.042 | 0.247*** | 0.245* | 1.854*** | -0.043 | -0.776 | -0.010 | mid vs. young = 0.691, mid vs. old = 0.146 | ||
Moderator*reappraisal | 0.004 | 0.104 | -0.005 | -0.019 | -0.033 | -0.009 | -0.135 | -0.001 | mid vs. young = 0.095, mid vs. old = 0.045 | ||
GAD-w7 | |||||||||||
Reappraisal | 0.063** | 0.012 | 0.035 | 0.033 | -0.020 | 0.008 | 0.023 | 0.014 | 0.010 | ||
Moderator | -0.150*** | -2.961*** | 0.341*** | 0.438*** | 2.100*** | -0.068 | -0.755* | -0.044*** | mid vs. young = 0.684*, mid vs. old = -0.795* | ||
Moderator*reappraisal | 0.000 | 0.049 | 0.004 | -0.008 | -0.020 | -0.003 | -0.042 | -0.002 | mid vs. young = 0.055, mid vs. old = 0.007 | ||
GAD-w8 | |||||||||||
Reappraisal | 0.061* | 0.032 | 0.036 | 0.045 | 0.006 | 0.025 | 0.034 | 0.035 | 0.008 | ||
Moderator | -0.141*** | -2.451* | 0.235*** | 0.445*** | 2.078*** | -0.075 | -0.749* | -0.018 | mid vs. young = 0.699, mid vs. old = -0.354 | ||
Moderator*reappraisal | 0.000 | 0.085 | -0.003 | -0.009 | -0.006 | 0.007 | -0.026 | -0.002 | mid vs. young = 0.094, mid vs. old = 0.023 | ||
DJGLS-w7 | |||||||||||
Reappraisal | 0.000 | -0.013 | -0.019* | -0.015 | -0.020* | -0.010 | -0.009 | -0.011 | -0.016 | ||
Moderator | − 0.073*** | -0.908** | 0.104*** | 0.158*** | 0.448*** | -0.072* | -0.298* | -0.010** | mid vs. young = 0.486***, mid vs. old = 0.162 | ||
Moderator*reappraisal | 0.002* | 0.030 | -0.003* | -0.007 | -0.013 | 0.007 | -0.024 | 0.000 | mid vs. young = -0.005, mid vs. old = 0.008 | ||
DJGLS-w8 | |||||||||||
Reappraisal | 0.010 | 0.002 | -0.005 | 0.007 | -0.014 | 0.004 | 0.020 | 0.008 | 0.001 | ||
Moderator | -0.082*** | -0.508 | 0.084*** | 0.193*** | 0.514*** | -0.128** | -0.504** | -0.012* | mid vs. young = 0.633***, mid vs. old = 0.214 | ||
Moderator*reappraisal | 0.002* | 0.021 | -0.004* | -0.005 | -0.010 | -0.004 | -0.101*** | -0.001 | mid vs. young = 0.014, mid vs. old = 0.000 | ||
WSAS-w7 | |||||||||||
Reappraisal | 0.162** | 0.091 | 0.124** | 0.087 | 0.079 | 0.035 | 0.025 | 0.038 | -0.080 | ||
Moderator | -0.288*** | -2.940 | 0.505*** | 0.481** | 2.904*** | -0.285 | -0.978 | -0.086*** | mid vs. young = 1.640*, mid vs. old = -1.140 | ||
Moderator*reappraisal | -0.009 | -0.064 | 0.016* | -0.010 | 0.089* | 0.005 | 0.014 | -0.001 | mid vs. young = 0.173, mid vs. old = 0.120 | ||
WSAS-w8 | |||||||||||
Reappraisal | 0.267*** | 0.101 | 0.205** | 0.197** | 0.162** | 0.147* | 0.152* | 0.175** | 0.095 | ||
Moderator | -0.382*** | -8.705*** | 0.510*** | 0.648** | 2.714*** | -0.181 | -1.684 | -0.071** | mid vs. young = 2.137*, mid vs. old = -0.772 | ||
Moderator*reappraisal | -0.012 | 0.041 | 0.011 | -0.015 | 0.105* | 0.019 | -0.084 | -0.010* | mid vs. young = 0.292, mid vs. old = -0.124 | ||
WHO-w7 | |||||||||||
Reappraisal | 0.121*** | 0.149*** | 0.146*** | 0.128*** | 0.159*** | 0.103** | 0.089* | 0.114*** | 0.190*** | ||
Moderator | 0.197*** | 1.158 | -0.235*** | -0.604*** | -1.445*** | 0.076 | 1.057** | 0.015 | mid vs. young = -0.252, mid vs. old = 0.242 | ||
Moderator*reappraisal | -0.007* | -0.032 | 0.017*** | 0.025* | 0.069** | 0.002 | 0.125 | -0.002 | mid vs. young = -0.010, mid vs. old = -0.122 | ||
WHO-w8 | |||||||||||
Reappraisal | 0.104** | 0.116* | 0.130*** | 0.091* | 0.149*** | 0.100** | 0.064 | 0.108** | 0.134 | ||
Moderator | 0.185*** | 0.072 | -0.155*** | -0.623*** | -1.278*** | 0.258* | 1.139* | 0.017 | mid vs. young = -0.730, mid vs. old = -0.451 | ||
Moderator*reappraisal | -0.009* | -0.088 | 0.022*** | 0.026 | 0.091*** | 0.007 | 0.166 | -0.004 | mid vs. young = 0.029, mid vs. old = -0.112 | ||
Figure 3 illustrates the nature of significant moderation effects from Table 4 at high (1 SD above the mean), average (mean), and low (1 SD below the mean) levels of the relevant moderator variable, or for correct and incorrect responses to the cognitive reflection task for the analytic thinking moderator variable. Of the nine potential moderators, significant moderation effects were most consistently found for self-efficacy, difficulty identifying feelings, and stress, primarily in the context of wellbeing. However, these effects were all in the opposite direction to that hypothesised, with significant associations emerging primarily when other vulnerabilities were present rather than under conditions where strengths were expected to improve reappraisal ability. Figure 3A shows that higher reappraisal was significantly associated with higher wellbeing when self-efficacy was average or low, and when difficulty identifying feelings, stress, and neuroticism were average or high. These effects were evident at wave 7 for all four of these moderators and wave 8 for all except neuroticism. Figure 3B also shows that higher reappraisal was significantly associated with lower loneliness at wave 7 when difficulty identifying feelings was high.
For functional impairment, which was associated with reappraisal in the opposite direction to that predicted (i.e., higher reappraisal was associated with higher impairment), we also observed stronger relationships at higher levels of vulnerability. Figure 3C shows that higher reappraisal was significantly associated with higher functional impairment at wave 7 when difficulty identifying feelings was average or high, at wave 8 when age was average or young, and at wave 8 when stress was average or high. There was also a trend in the same direction for stress at wave 7.
Fig. 3
Simple Slopes Analysis for Significant Moderation Effects for Wellbeing, Loneliness, and Functional Impairment
×
![]()
The only moderation effects that were in the direction we hypothesized were for analytic thinking. Higher reappraisal was significantly associated with higher wellbeing (analytic reasoning in Fig. 3A) and lower loneliness (analytic reasoning in Fig. 3B) at wave 8 for participants who correctly responded to the cognitive reflection question correctly but not for those who responded incorrectly.
Discussion
The present study produced several key findings. First, we found no association between reappraisal and depression or anxiety, contrary to meta-analytic evidence (which is mostly cross-sectional; Aldao et al., 2010; Hu et al., 2014) but robustly confirming the lack of association found in Dawel et al., 2021) and other systematic longitudinal evidence (Riepenhausen et al., 2022). Second, while greater reappraisal was associated with higher wellbeing and reduced loneliness as predicted, it was also associated with more rather than less pandemic-induced functional impairment. Third, moderation effects were mostly in the opposite direction to that predicted, overall indicating that reappraisal may be most helpful for improving wellbeing in people with additional vulnerabilities. Finally, greater reappraisal was associated with higher wellbeing at 3 months and lower loneliness at 12 months in people who demonstrated an inclination for analytic reasoning but not for those who did not. Altogether, our results highlight and further support the complex nature of the relationship between reappraisal and mental health.
Reappraisal and Wellbeing
Reappraisal tended to be more strongly associated with psychological wellbeing than other outcome measures (e.g., mean Spearman’s ρ for wave 7 and 8 = 0.184 for wellbeing compared to -0.100 for loneliness and 0.122 for functional impairment) and moderation effects were most consistently evident in the wellbeing models. Specifically, greater reappraisal was associated with higher wellbeing at average to low levels of self-efficacy and average to high levels of stress, neuroticism, and difficulty identifying feelings. These results are consistent with a recent systematic review (Riepenhausen et al., 2022) that found reappraisal predicted wellbeing in adverse circumstances and argued reappraisal may promote resilience in the face of stressors. The COVID-19 pandemic can be conceptualised as such a stressor. This set of findings is also important because self-efficacy (Chen et al., 2000; Luszczynska et al., 2005) and neuroticism (Rantanen et al., 2007; Wortman et al., 2012) are relatively stable; thus, it may be helpful for therapy to focus on developing skills that are protective for these individuals.
Reappraisal and Negative Outcomes
Reappraisal was differentially associated with the four maladaptive outcome measures. We found the predicted negative association of reappraisal with loneliness but, as previously found in Dawel et al. (2021), no association with depression or anxiety. Dawel et al. (2021) explore the potential reasons for this lack of association in detail. One additional potential explanation is that some people’s emotion regulation goals aimed to maintain current mood rather than decrease negative affect. Clinical depression has previously been associated with regulating to maintain or even increase negative affect (Millgram et al., 2015). Therefore, our results may have confounded maintainers with individuals attempting to decrease negative or increase positive affect. To test this hypothesis, future research should measure people’s emotion regulation goals in addition to emotion regulation use and effectiveness.
Counter to our initial predictions, the present study also revealed a positive association between reappraisal and functional impairment. People who reported COVID-19 had caused greater impairment in their social and work functioning reported using reappraisal more—not less as we had predicted. However, with hindsight it makes sense that people who experienced worsened circumstances due to COVID-19 would attempt to use reappraisal more. At the time our data was collected, social and traditional media strongly encouraged people to reframe their experiences to manage the impacts of COVID-19 (e.g., Polizzi & Lynn, 2020). The finding of a positive relationship between reappraisal and functional impairment is also consistent with our recent argument that the relationship between reappraisal use and psychosocial outcomes can work in both directions (Dawel et al., 2021). The uncontrollable negative circumstances of the pandemic may have motivated people to try to reframe their situation more frequently, potentially improving overall psychological wellbeing but not the level of functional impairment as we measured it. It is possible our measure of functional impairment captured the external reality of people’s circumstances more strongly than their perceptions of impairment. For instance, the level of functional impairment in the work domain due to job loss is likely to be high irrespective of how much reappraisal is used. Consistent with this argument, moderation analyses suggested that greater reappraisal was associated with increasing levels of impairment at higher levels of stress and difficulty identifying feelings, and with younger age, indicating those most vulnerable to COVID-induced adversity tended to use reappraisal more.
Clinical Implications
The results contribute to further understanding the complexity in how cognitive reappraisal use and effectiveness impact psychological health across different individuals and contexts. Such nuanced knowledge can be used clinically to determine for whom and when cognitive reappraisal is likely to be beneficial and inform the development of personalized and situationally-suited interventions. The results suggest that in assessing whether to offer reappraisal training clinicians may wish to consider their clients’ inclination for analytic thinking or whether they have vulnerability factors such as low self-efficacy or difficulty identifying feelings. It may also be appropriate to encourage clients who have high self-efficacy or who are adept at identifying feelings to draw on these strengths to effectively regulate their emotions, rather than providing a reappraisal intervention. The specific nature of any psychosocial stressors that a client is currently facing should also be considered.
An important consideration is that successful reappraisal may not always be desirable (Ford et al., 2019; Ford & Troy, 2019; van’t Wout et al., 2010). While reappraisal interventions consistently reduce negative affect, this may reduce motivation to take problem-solving action (Ford et al., 2019). For example, Van’t Wout et al. (2010) found participants instructed to use reappraisal were more likely than uninstructed participants to accept unfair offers in the ultimatum game experimental paradigm, disadvantaging them. This is consistent with our finding that reappraisal is related to greater functional impairment, as reappraising and accepting circumstances may impede problem-solving that could have reduced impairment (e.g., taking action to improve one’s financial or social circumstances in the pandemic context).
Limitations and Future Directions
The present study measured reappraisal use and investigated moderators that were theorised to influence reappraisal effectiveness. It is critical that future work also measure reappraisal effectiveness directly. For instance, by asking participants about how reappraisals impacted their emotions. It is also critical that future work measure reappraisal self-efficacy over and above general self-efficacy. While we found that reappraisal was more beneficial for individuals with low general self-efficacy, reappraisal-specific self-efficacy may positively influence outcomes, particularly if measured on a short time scale in relation to specific reappraisal events (e.g., using Ecological Momentary Assessment methods; Erbas et al., 2022). It would also benefit future work to measure participants’ emotion regulation goals, as this may provide insight into why reappraisal is not always associated with reduced depression and anxiety. One possibility is that people focused their reappraisals in the pandemic context on increasing positive affect rather than decreasing negative affect, leading to increases in wellbeing but not decreases in depression and anxiety. Alternatively, it may be that reappraisal is not effective at reducing depression and anxiety in contexts where there are strong situational reasons for negative affect to be an appropriate response. Another open question concerns the extent to which each measure captured state or trait like properties. This is particularly relevant given that this study took place during COVID-19 lockdown in Australia because there were acute and extreme impacts on people’s individual circumstances, which had flow-on impacts for psychological wellbeing (Batterham et al., 2021; Dawel et al., 2020). Specifically, this period was associated with an unprecedented lack of controllability at a global level, poorer mental health outcomes (Khan et al., 2022) and increased loneliness (Cooper et al., 2021; Losada-Baltar et al., 2022). The generalisability of this study is limited in that our data may not reflect typical relationships between reappraisal, outcome, and moderator variables. However, in times of increasing uncertainty and global and environmental instability, it is critical to understand what strategies are useful for psychological outcomes in times of duress. Future research should investigate the relationships between these variables outside of times of acute global crisis. Additionally, while the current study used a robust longitudinal design, experimental research is needed to establish the potential causative relationships between reappraisal use and wellbeing or functional impairment, particularly for impairments attributable to clear external and potentially unchangeable stressors (e.g., pandemic, recession, incarceration etc.). An intriguing finding concerns the analytic reasoning task. However, the present study relied on a single item to measure this construct. Future research would benefit from more rigorously investigating the role of analytic reasoning in moderating emotion regulation benefits. Additional longitudinal research using larger population samples may also help elucidate the unique contributions of moderating variables.
Conclusions
Overall, our findings further challenge the assumed simplistic causal relationship between greater reappraisal use and better psychosocial outcomes (e.g., Aldao et al., 2010; Gross & John, 2003; Hu et al., 2014). Our large sample longitudinal data make clear there are circumstances when reappraisal use is not linked to depression and anxiety levels. This result highlights a critical need to understand the situations and individuals for whom reappraisal use reduces these mental health symptoms and to identify alternative interventions that are more effective in crisis situations like the pandemic. However, the present study also indicates that reappraisal use in such situations is potentially beneficial for wellbeing, particularly for people experiencing personal vulnerabilities such as low self-efficacy, high neuroticism or stress, and greater difficulty identifying feelings. The present study provides guidance as to when intrapersonal emotion regulation is effective. However, there is a growing focus on the role of interpersonal emotion regulation (e.g., Hofmann, 2014) and it will also be important to understand for whom and what interpersonal strategies work best. Overall, the present study supports emergent arguments that we need a nuanced person-situation-strategy fit approach to emotion regulation. Understanding the use, effectiveness, and goals of emotion regulation strategies at an individual and situational level promises to enable clinicians to provide enhanced, more personalized interventions, and support people to select regulation strategies that will work for them across different contexts.
Acknowledgements
We thank Patrice Ford for assistance with study administration. We also thank the members of the Australian National COVID-19 Mental Health, Behaviour and Risk Communication Survey team for their contribution to this project.
Declarations
Ethics Approval
This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of the relevant university (protocol 2020/152).
Consent to participate
Informed consent was obtained from all individual participants included in the study.
Consent to publish
The participant consent procedure included consent to publish data without any individually identifying information.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
