Individual differences in spontaneous self-affirmation and mental health: relationships with self-esteem, dispositional optimism and coping

ABSTRACT In two online studies, we test whether spontaneous self-affirmation (measured by the Spontaneous Self-Affirmation Measure, SSAM) predicts better mental health and coping and the role that self-esteem and dispositional optimism play in these relationships. Study 1 (N = 110) was cross-sectional. In study 2 (N = 192) we collected the mental health measures one month post-baseline. Consistent with pre-registered hypotheses, the SSAM predicted less anxiety, depression and avoidant coping, and greater wellbeing and non-avoidant coping; however, relationships involving self-esteem and optimism varied with the reported source of self-affirmation measured by the SSAM (strengths, values, social relations). Overall, the findings are generally consistent with the hypothesis that spontaneous self-affirmation tends to function as a resource that fosters positive coping with threats.

To date, most research on self-affirmation has been experimental, much of which has employed manipulations requiring participants to reflect upon their most important values. This has been used to good effect to understand and influence responses to a wide range of threats, notably in education and physical health (Cohen & Sherman, 2014;Easterbrook et al., 2021;Harris & Jessop, in press;Harris et al., 2020).
An important, but hitherto little researched, complement to experimental research involves understanding how self-affirmation functions as a spontaneous response to threats in everyday life. In this context, researchers have begun to assess the nature and implications of individual differences in tendency to report self-affirming naturally in their everyday lives (Harris et al., 2019;Pietersma & Dijkstra, 2012).
Spontaneous self-affirmation has the potential to function as a defensive response to everyday threats, but also as a positive response that fosters constructive coping (Harris et al., 2019). Much remains to be discovered about the factors that influence which tendency predominates in whom and when.
In numerous studies, Harris and colleagues have examined how the tendency to report responding to threats by self-affirming spontaneously -as measured by the Spontaneous Self-Affirmation Measure (SSAM; Harris et al., 2019) -predicts a range of outcomes, including measures of responsiveness to threat and risk information, self-perception and self-function, and responses to health threats. In competitive analyses with measures of positive self-regard, including self-esteem, spontaneous self-affirmation emerged as an independent predictor of many outcomes (Harris et al., 2019). This included responses to messages about physical health, where spontaneous self-affirmation functioned in ways that were similar to experimentally induced self-affirmation by predicting greater openmindedness and readiness to engage in behavior change.
As a next step in exploring the correlates and potential implications of individual differences in spontaneous self-affirmation, and how it relates to other facets of selffunctioning and identity, we turn our attention in this paper to the relationship between spontaneous self-affirmation, mental health and wellbeing. As we outline below, there are grounds for expecting it to relate to such experiences. Indeed, there is some evidence that induced self-affirmation can buffer and may even promote psychological wellbeing, such as the experience of happiness, life-satisfaction and self-actualization, but little evidence concerning the role that self-affirmation (whether induced or spontaneous) plays in preventing, exacerbating or coping with mental distress, such as anxiety and depression. To inform our understanding of the ways in which spontaneous self-affirmation predicts responding to everyday threats and challenges that could diminish mental health, we therefore examined how it predicts measures of both mental distress and psychological wellbeing and tested what happens when we control for its relationships with two wellestablished predictors of distress and wellbeing, trait self-esteem and dispositional optimism.

Spontaneous self-affirmation, self-esteem, and optimism
The tendency to engage in spontaneous self-affirmation correlates positively with both trait self-esteem and dispositional optimism (Harris et al., 2019;Lannin et al., 2021;Pietersma & Dijkstra, 2012). These dispositions are, in turn, associated positively with mental health and wellbeing (e.g., Achat et al., 2000;Conversano et al., 2010;Lancastle & Boivin, 2005;Sowislo & Orth, 2013;Steiger et al., 2014;Taylor & Broffman, 2011). Consequently, a research question at the heart of this research programme is to establish the extent to which these correlates of spontaneous self-affirmation help confer protection against mental distress and foster wellbeing. However, the SSAM comprises a higher order factor with three first-order factors that differ in the strength of their relationships with self-esteem and dispositional optimism (Harris et al., 2019), and this may have implications for the relationship between the SSAM and mental health. The factors represent three sources that people may use to self-affirm: a focus on personal strengths and attributes, on important personal values, and on valued social relationships. These sources, in turn, are the ones that have been mainly used to manipulate self-affirmation experimentally. The typical respondent reports using all three sources, but the relative importance and frequency attached to each varies from individual to individual.
A relative focus on personal strengths is more strongly associated with self-esteem than is a relative focus on either of the other SSAM factors (Harris et al., 2019). Consequently, in the current studies we test the strengths of the paths between each firstorder factor and mental health, enabling us to depict this relationship at a more molecular level than that solely comprising relationships involving the higher-order factor.

Spontaneous self-affirmation, mental distress and psychological wellbeing
Our initial working hypothesis is that in general (i.e., for most people, most of the time) spontaneous self-affirmation functions as a resource that helps foster positive coping with threats. If so, being higher in this tendency should typically help offset the negative impact of daily stressors, threats and hassles and help protect mental health and wellbeing.
Several factors underpin this working hypothesis. First it is a direct extension from research showing the benefits of experimentally manipulated (i.e., non-spontaneous) selfaffirmation for physical health (e.g., Epton et al., 2014) and for psychological wellbeing (e.g., Howell, 2017;Schüz & Schüz, 2017). Second, it is consistent with the theoretical understanding of how engaging in self-affirmation enables the individual to go beyond defensive responding to threats (e.g., by better contextualizing them; Sherman & Cohen, 2006). Third, as described above, it is consistent with the fact that the tendency to engage in spontaneous self-affirmation correlates positively with dispositions, such as trait selfesteem and dispositional optimism, that potentially protect mental health.
It also has a measure of direct empirical support. For example, two SSAM items were used in a nationally representative US survey (the Health Information National Trends Survey 4, Cycle 3; HINTS). Emanuel et al. (2018) found that those scoring higher on those SSAM items typically reported more positive affect, such as happiness and hopefulness, and less negative affect, such as sadness and anger, (though not less anxiety). Taber et al. (2015) found that cancer survivors in the HINTS sample scoring higher on the SSAM items reported greater happiness and hopefulness. In research using the full set of SSAM items, Harris et al. (2019) found that being higher in spontaneous self-affirmation predicted less reported experience of the negative basic emotions (e.g., anger, sadness and fear). The SSAM was also significantly negatively associated with measures of depression and state anxiety; however, it did not predict trait anxiety (see, Harris et al., 2019, supplemental materials). Of the SSAM factors, all correlated significantly positively with happiness; strengths and social relations with meaningful existence; and strengths alone with satisfaction with life and (negatively) with manifest anxiety. Relationships that existed after controlling for self-esteem and the other SSAM factors included strengths with happiness and satisfaction with life, values with meaningful existence (negative), and social relations with happiness and meaningful existence. Across three studies focusing exclusively on wellbeing, Jessop et al. (2022) found that the SSAM positively predicted wellbeing, both cross-sectionally and longitudinally.
Spontaneous self-affirmation, as measured by the SSAM, has therefore been shown to correlate with some, but not all, indices of mental health and to do so typically in the ways predicted by the working hypothesis. Nevertheless, findings need replicating and more needs to be discovered about these relationships and the roles that traits associated with mental health, such as self-esteem and dispositional optimism, may play in them. 1

The SSAM and dispositional coping
The working hypothesis is predicated on the assumption that those who spontaneously self-affirm typically have access to (and beliefs about possessing) resources that can be deployed to constructively offset psychological threats (Cohen & Sherman, 2014). In support of this, Harris et al. (2019) presented data showing that, after controlling for its relationship with self-esteem, spontaneous self-affirmation predicted non-defensive, control-based and systematic responding to threats. (Of note, however, as revealed in their supplemental materials, the SSAM first-order factors showed a somewhat diverse pattern of relationships.) To extend our understanding of how spontaneous self-affirmation is linked to coping with threats, in the studies reported here we examined the relationships between the SSAM and its factors and two inventories of dispositional coping that are widely used in mental health research, the Brief COPE Inventory (Carver, 1997) and the Cognitive Emotion Regulation Questionnaire (CERQ, Garnefski et al., 2001).
Coping has been defined as the "constantly changing cognitive and behavioral efforts to manage specific external and/or internal demands that are appraised as taxing or exceeding the resources of the person" (Lazarus & Folkman, 1984, p. 141). There are many strategies available to people for coping with threats and stressors. The COPE and CERQ identify different strategies that individuals may characteristically deploy to cope. In any given situation these may prove beneficial or counter-productive (e.g., Greenaway et al., 2014).
We used both the COPE and CERQ to ensure breadth of coverage. The COPE measures 14 coping strategies. It focuses on diverse ways of responding characteristically, such as by problem solving. The CERQ assesses 9 types of emotion-related thoughts following a negative event. Between them the COPE and CERQ enabled us to assess a comprehensive set of strategies available to an individual to respond to threats and stressors. However, in order to make sensible predictions that we could test statistically, it was necessary to reduce these strategies into meaningful clusters within inventories. In doing this we were guided both by assumptions about the nature of these strategies and by cues in the research of others. We identified the COPE strategies of emotional support, instrumental support, planning, positive reframing, active coping, and acceptance, as one cluster, and of denial, substance use, behavioral disengagement, venting, self-distraction, and self-blame, as another (see, Gloria & Steinhardt, 2016). We considered the former to be a cluster of more positive, non-avoidant or constructive responses in general (cluster 1) and the latter as a cluster of less positive, more avoidant and less constructive responses in general (cluster 2), although it should be noted that this is a convenient shorthand and is not intended to beg questions about any given strategy being adaptive or maladaptive inherently or in a specific context. Rather we were concerned to identify a set of strategies that clustered together, which, broadly, were associated positively and negatively (respectively) with mental health, and that we could use in analyses involving the SSAM and its factors. Likewise, subject to the same caveat, we identified the CERQ strategies of acceptance, putting into perspective, refocusing on planning, positive refocusing, and positive reappraisal, as a cluster (cluster 1) of typically less avoidant, more constructive responses, and the strategies of rumination, self-blame, other-blame, and catastrophizing, as a cluster (cluster 2) of typically more avoidant, less constructive responses (see also, Garnefski et al., 2001).
We hypothesized that spontaneous self-affirmation, as measured by the SSAM, would correlate positively with the less avoidant, more constructive coping clusters and negatively with the more avoidant, less constructive coping clusters of each inventory. Consistent with this, Harris et al. (2019, supplemental materials), found that the SSAM correlated positively with the CERQ strategies of putting in perspective, positive refocusing, positive reappraisal, acceptance, and planning, negatively with self-blame and catastrophizing, and not at all with other-blame. The SSAM first-order factors also predicted several of the positive strategies (e.g., positive reappraisal) but none of the negative ones. We know of no previous data involving the SSAM and COPE.
These hypotheses are based on the operational assumption that spontaneous selfaffirmation is likely to be associated with the appraisal process and the perceptions of the resources available to cope with threat, such that those higher in spontaneous selfaffirmation will tend to possess self-perceptions of having the ability to cope that influence their (secondary) appraisals of stressors and their coping resources in ways that facilitate coping adaptively. This assumption is derived principally from the theoretical perspective that self-affirming enables the individual to adopt a less defensive and more adaptive attitude to threats, as suggested by the findings of experimental research on self-affirmation, supplemented by our initial findings using the SSAM (Harris et al., 2019; i.e., our working hypothesis).

The current studies
In this paper we present data from two online studies. Study 1 is a cross-sectional study in which we first tested our hypotheses about the links between the tendency to engage naturally and spontaneously in self-affirmation and indices of mental distress and wellbeing, how spontaneous self-affirmation predicts dispositional coping, and the role that self-esteem and dispositional optimism play in these relationships. Ahead of study 2 we pre-registered these hypotheses before testing them in a longitudinal study (see supplemental online material, SOM). Therefore, study 1 tests these hypotheses contemporaneously and study 2 does so prospectively.
In each study, we measured both the relative lack of mental distress (depression and anxiety) and the relative presence of positive mental states and attributes (life satisfaction, happiness, sense of having a worthwhile life, flourishing and a positive affect balance). To avoid begging questions about the relationship between them, such as whether or not they are opposite ends of a continuum (e.g., Iasiello et al., 2020), we pre-registered separate hypotheses for the summary indices of mental distress, on the one hand, and psychological wellbeing, on the other (https://osf.io/zfyeg/?view_only=).

Study 1
We first tested our hypotheses cross-sectionally. Hypotheses 1a and 1b predicted that spontaneous self-affirmation would correlate negatively with the measures of anxiety and depression (1a) and positively with the wellbeing measures (1b). Hypotheses 2a-2c predicted that the relationships between spontaneous self-affirmation and anxiety and depression would be eliminated by controlling for self-esteem (2a), but that those between spontaneous self-affirmation and the wellbeing measures would not be eliminated by controls for either self-esteem (2b) or dispositional optimism (2c). 2 Hypothesis 2d predicted that spontaneous self-affirmation would correlate positively with the more "adaptive" (nonavoidant) coping strategies (cluster 1) and negatively with the "less adaptive" (avoidant) coping strategies (cluster 2). Hypothesis 3 predicted that greater use of the non-avoidant coping strategies and/or lower use of the avoidant coping strategies would mediate the relationships between spontaneous self-affirmation and the mental distress measures and between spontaneous self-affirmation and the wellbeing measures.
We also report findings relevant to a further hypothesis (hypothesis 4), concerning the role of social connection in spontaneous self-affirmation: that spontaneous selfaffirmation will be associated with higher feelings of compassion, love and connection to others (see note 1).

Participants
Overall, N = 142 participants consented to participate in the study and completed the demographic measures (n = 140 for age), of whom n = 120 responded to at least one other measure and n = 117 completed the SSAM and self-esteem measures. Of these n = 7 did not respond to most of the other measures. The main analyses were therefore conducted on the remaining n = 110 (78%) participants. These participants did not differ from the rest of the sample (n = 32) in either age (M = 30.8 vs M = 27.6 years respectively), F(1, 138) = 1.46, p = .229, or gender (scored 1 = female, 2 = male), F(1, 139) < 1, p = .378. They were typically female (n = 85, 77%), white (n = 100, 91%), in full time employment (n = 59, 54%) or a student (n = 31, 28%), and aged between 18 and 85 (M = 30.8, SD = 14 years). They were an opportunity sample recruited through the University Participation Pool and one of the author's Facebook and Twitter accounts.

Measures
For all measures below, higher scores equal more of the construct being measured. Mean responses were calculated for multiple item measures. Demographic measures included age, gender, ethnicity and employment status. 3

Distress
The Center for Epidemiological Studies Depression scale (CES-D; Radloff, 1977;α = .93) was used to measure levels of depression during the previous week (20 items, e.g., "I was bothered by things that usually don't bother me"). Responses were given on a 4-point scale, 1 (rarely or none of the time), 2 (some or a little of the time), 3 (occasionally or a moderate amount of the time), 4 (most or all of the time). The short version of the Overall Anxiety Severity and Impairment Scale (OASIS; Norman et al., 2011;α = .91) was used to measure anxiety during the previous week (5 items, e.g., "In the past week, how often have you felt anxious?"). Responses were given on a 5-point scale, 1 (no anxiety; little or none; none), 2 (infrequent; mild), 3 (occasional; moderate), 4 (severe; frequent), 5 (constant; extreme; all the time). (Labels varied with the item.) Anxiety was also measured with the item developed by the UK Office for National Statistics (ONS; Allin & Hand, 2017;VanderWeele et al., 2020; "Overall, how anxious did you feel yesterday?"). Responses were given on a 11-point scale ranging from 1 (not at all) to 11 (completely).

Wellbeing
We measured indices of hedonic (affect balance, happiness), evaluative (life satisfaction), and eudaimonic (flourishing, worthwhile life), wellbeing (VanderWeele et al., 2020). The Flourishing Scale (Diener et al., 2010;α = .88) was used to measure flourishing (8 items, e.g., "I lead a purposeful and meaningful life"). Responses were given on a 7-point scale 1 (strongly disagree), 2 (disagree), 3 (slightly disagree), 4 (neither agree nor disagree), 5 (slightly agree), 6 (agree), 7 (strongly agree). The Modified Differential Emotions Scale (MDES; Fredrickson et al., 2003) was used to measure positive affect (10 items, "I have felt amused, fun loving, or silly"; α = .91) and negative affect (7 items, "I have felt sad, downhearted, or unhappy"; α = .86) over the previous 7 days. 4 Responses were given on a 5-point scale ranging from 1 (never) to 5 (most of the time). Affect balance was calculated as positive minus negative affect, so positive scores indicate a positive balance (e.g., Veilleux et al., 2020). The items developed by the ONS (Allin & Hand, 2017;VanderWeele et al., 2020) were used to measure happiness ("Overall, how happy did you feel yesterday?"), life satisfaction ("Overall, how satisfied are you with your life nowadays?"), and sense of having a worthwhile life ("Overall, to what extent do you feel the things you do in your life are worthwhile?"). Responses were given on a 11-point scale ranging from 1 (not at all) to 11 (completely).

Coping
Dispositional coping styles were measured using the Brief Coping Orientations to Problems Experienced (Brief COPE; Carver, 1997) and the Cognitive Emotion Regulation Questionnaire-Short (CERQ-Short; Garnefski & Kraaij, 2006) scales. For the COPE the subscales measuring emotional support, instrumental support, planning, positive reframing, active coping, and acceptance were combined to form cluster 1 (non-avoidant coping, α = .91) and the subscales measuring denial, substance use, behavioral disengagement, self-distraction, self-blame and venting, were combined to form cluster 2 (avoidant coping, α = .81). Each Brief COPE subscale comprised two items (e.g., "I've been turning to work or other activities to take my mind off things"). Responses were given on a 4-point scale, 1 (I haven't been doing this at all), 2 (I've been doing this a little bit), 3 (I've been doing this a medium amount), 4 (I've been doing this a lot). For the CERQ-Short the subscales measuring acceptance, putting into perspective, refocusing on planning, positive refocusing, and positive reappraisal, were combined to form cluster 1 (non-avoidant coping, α = .86) and the subscales measuring rumination, self-blame, other-blame, catastrophizing) were combined to form cluster 2 (avoidant coping, α = .80). Each CERQ-Short subscale comprised two items (e.g., "I've been turning to work or other activities to take my mind off things"). Responses were given on a 5-point scale, anchored at 1 (almost never) and 5 (almost always).

Procedure
The recruitment message invited participants to take part in a study exploring "coping styles and other potential factors relating to anxiety, depression and well-being". It contained a hyperlink to the survey program in Qualtrics, where the information sheet and consent form were presented. All responses were anonymous. At the end of the survey participants were debriefed and given a unique participant reference number to enable them to remove their data before analysis (together with a date after which that would not be possible). Participants could also enter a prize draw to win £25 (approximately $31). The survey took approximately 20-30 minutes to complete.

Results
Structural equation modeling (SEM) was conducted using the lavaan package in R (Rosseel, 2012). Other analyses were conducted using SPSS (version 27).
Two features of the data obliged us to depart from some of the planned analyses (see SOM). First, when analyzed as latent variables, mental distress and wellbeing were so negatively correlated that the pattern of findings in one model was essentially an inverted replication of the pattern in the other. When combined into a single latent mental health factor, however, depression dominated it to the extent that it was mainly a measure of depression. Consequently, to provide a more informative and nuanced set of findings we modeled all outcomes simultaneously as independent outcomes.
Second, the initial SEM analyses tested the SSAM as a general spontaneous selfaffirmation factor with three indicators (a focus on strengths, values, and social relationships). However, the associations of the three SSAM factors with self-esteem, particularly of focusing on strengths, were so distinctive in this domain that if we attempted to constrain the model by forcing them to predict overall spontaneous self-affirmation it failed to converge. Consequently, we conducted the SEM analyses testing hypotheses 2a-2c and 3 using the first-order factors as independent predictors in the model, instead of the overall SSAM.
Hypotheses 1a and 1b. The zero-order correlations relevant to the hypotheses can be found in Table 1. Consistent with the hypotheses, the SSAM correlated significantly negatively with depression and OASIS anxiety (but not ONS anxiety) and positively with wellbeing.
As described above, rather than testing these hypotheses using the overall SSAM, we used the three subscales as independent predictors in the models (Table 2). Reading across, Table 2 displays the results of the SEM analyses and shows how the three SSAM factors predicted each outcome before adding the covariates (RSES and LOT-R), with each covariate separately, and with both covariates simultaneously. SSAM strengths predicted lower depression and greater wellbeing on all indices. Interestingly, strengths predicted more anxiety (on both measures) once the covariates were added, at which point it no longer predicted the wellbeing measures, suggesting that its relationships with these outcomes were a function of its relatively positive relationship with optimism and (especially) self-esteem. The other SSAM factors did not predict any outcomes until the introduction of the covariates, at which point SSAM relations was a positive predictor of flourishing, life satisfaction and (marginally) happiness.
To test the role that self-esteem and dispositional optimism played in these relationships, we undertook equivalent SEM analyses to those for the mental health measures using the three SSAM first-order factors and the coping clusters as the outcomes. Thus, reading across, Table 3 shows how the three SSAM factors predicted each cluster before adding the covariates, with each covariate separately, and with both covariates simultaneously.

Cluster 1
On the COPE, SSAM values positively predicted non-avoidant coping, with and without the covariates. No other variable predicted non-avoidant coping. On the CERQ, SSAM strengths positively predicted non-avoidant coping. This relationship was rendered marginal by adding the covariates. (The RSES was also a marginal positive predictor.) No other variable predicted non-avoidant coping.

Cluster 2
On the COPE, SSAM strengths negatively predicted avoidant coping. This relationship was eliminated by adding the covariates; RSES was then the sole predictor (negative), with the LOT a marginal predictor. No variable predicted CERQ avoidant coping in any model. 5

Discussion
As predicted, the SSAM correlated negatively with depression and anxiety (on one of the two measures) and positively with wellbeing. The findings for wellbeing replicate those of Jessop et al. (2022) and extend them by showing that, in this domain, the associations of the three SSAM factors are distinctive. Indeed, in this study they were so distinctive that we were compelled to conduct the planned SEM analyses using the first-order factors as independent predictors, rather than using the overall SSAM. That is, the three sources of self-affirmation (particularly of focusing on strengths) tended to have divergent relationships with the covariates (particularly self-esteem) and hence with the various mental health and coping variables.
In the SEM analyses, before controlling for self-esteem and optimism, only spontaneous self-affirmation through strengths predicted better mental health and wellbeing. Controlling for the shared variance with self-esteem and optimism changed things. The positive relationships between a focus on strengths and mental health were accounted for by the variance this factor shared with these dispositions. In contrast, with the covariates in the model, social relations was a positive predictor of several indices of wellbeing, whereas spontaneous self- affirmation through values was not. Consequently, self-esteem and optimism have different implications for the wellbeing of those who tend to self-affirm spontaneously, being associated with benefits for those who tend to self-affirm through strengths but lower wellbeing for those who tend to self-affirm using their social relations.
The data also supported the hypothesis that the SSAM would correlate positively with non-avoidant coping; however, it only correlated significantly negatively with avoidant coping on the COPE. In SEM analyses, SSAM strengths predicted more non-avoidant coping on the CERQ but less avoidant coping on the COPE. In both cases these relationships were also largely a product of its shared variance with self-esteem. SSAM values predicted more non-avoidant coping on the COPE, with and without the covariates, but not on the CERQ. SSAM social relations did not predict coping in any model. Overall, these relationships are consistent with the working hypothesis, which proposes that self-affirming enables the individual to adopt a less defensive and more adaptive response to threats. However, they suggest a more nuanced picture in which the source of the self-affirmation matters.
As a measure, the SSAM is designed to be modular and to be used flexibly. That is, we do not presuppose that the three factors should always load onto a higher-order factor or even that the factors exhaust the sources used to self-affirm. Indeed, this is not the first time that a focus on strengths, in particular, has been shown to be distinctive: Webb et al. (2020) recently found that people preoccupied with their weight expected and experienced less negative feelings when they weighed themselves if they tended to report self-affirming through strengths rather than through values or social relations.
Little is known about whether and when different sources of self-affirmation may have different implications (see, Cohen & Sherman, 2014). The current findings suggest focussing on strengths may be somewhat distinctive. Clearly, however, these findings require replication.

Study 2
In Study 2 we took an equivalent set of measures on a separate sample, but this time collecting the distress and wellbeing measures approximately one month after measuring the SSAM and covariates. The purpose of this was to not only test how well the findings replicated but also to provide a more robust assessment of the relationships between the variables by eliminating any tendency of cross-sectional measurement to inflate or otherwise distort these relationships (e.g., by making the relationships salient).

Participants
Overall, N = 336 participants consented to participate in the study and completed the demographic measures, of whom n = 320 completed the T1 measures. Of these, n = 196 began responding to the T2 survey, with n = 192 (57%) completing sufficient measures to be included in the main SEM analyses. These participants did not differ from the rest of the sample (n = 144) in either age (M = 32.4 vs M = 32.1 years respectively), F(1, 334) < 1 , p = .863, or gender, X 2 (2, N = 336) = .351, p = .839. They were typically female (n = 125, 65%), British (n = 157, 82%), in full time employment (n = 66, 34%) or a student (n = 81, 42%), and aged between 18 and 81 (M = 32.4, SD = 14.9 years). They were an opportunity sample recruited through the University Participation Pool and the researchers' social media accounts.

Measures
Recruitment to the study occurred in two consecutive years. Measures were the same as those in study 1. The measures of coping and individual differences were taken at T1. The measures of distress and wellbeing were taken at T2. Scale alphas ranged from .69 (CERQ avoidant) to .94 (CES-D). As in study 1, higher scores equal more of the construct being measured.

Procedure
The procedure at time 1 was essentially as described for study 1. Approximately 28 days later, participants were emailed the link to the time 2 questionnaire, at the end of which they were debriefed, told how to remove their data if they so wished and the date after which that would not be possible, and given the option to enter a prize draw to win £25 (approx. $33).

Results
Hypotheses 1a and 1b. These hypotheses were supported. The SSAM correlated significantly negatively with the distress and positively with the wellbeing outcomes (Table 1).
To facilitate comparisons with Study 1, we analyzed the data in the same way, using the three subscales as independent factors in the model and predicting all outcomes simultaneously but as independent outcomes. Hypotheses 2a-2c. As in study 1, SSAM strengths correlated more positively with selfesteem (r = . 58) and dispositional optimism (r = . 49) than did SSAM values (r = . 31 and r = . 29 respectively) or social relations (r = . 18 and r = .21 respectively) (all comparisons p < .001). Self-esteem and LOT-R correlated r = .80, which is sufficiently high that caution must be exercised in interpreting the coefficients for these in the final model. Table 2 displays the results of the SEM analyses. It shows how the three SSAM factors predicted each outcome with and without the covariates (RSES and LOT-R). As in study 1, SSAM strengths predicted lower depression and greater wellbeing on all indices. However, unlike study 1, it predicted less anxiety (on both measures). On the OASIS, strengths predicted less rather than more anxiety once the covariates were added (in contrast to study 1), at which point (consistent with study 1) it no longer predicted wellbeing. As in study 1 (with one exception), the other SSAM factors did not predict any outcomes until the introduction of the covariates (the exception being that social relations significantly predicted happiness). As in study 1, SSAM relations positively predicted wellbeing after introducing the covariates (in study 2 predicting affect balance, happiness and [marginally] life satisfaction).
Hypothesis 2d. This was again supported: the SSAM correlated significantly positively with the non-avoidant clusters of the COPE (r = .29, p < .001) and CERQ (r = .40, p < .001) and significantly negatively with the avoidant clusters on both coping measures (r = −.28, p < .001; r = −.20, p = .008, respectively). Table 3 displays the results of the SEM analyses and shows how the three SSAM factors predicted each cluster with and without the covariates.

Cluster 1
As in study 1, SSAM values positively predicted non-avoidant coping on the COPE, with and without covariates. Unlike study 1, however, with the covariates in the model SSAM strengths (negative) and SSAM social relations (positive) approached significance as predictors of COPE non-avoidant coping. Unlike study 1, the LOT also predicted it (positively). On the CERQ, as in study 1, SSAM strengths positively predicted nonavoidant coping; however, unlike study 1, this relationship was eliminated by adding the covariates. SSAM values positively predicted CERQ non-avoidant coping, unlike study 1, and did so with and without covariates.

Cluster 2
On the COPE, as in study 1, SSAM strengths negatively predicted avoidant coping. However, unlike study 1, this relationship remained significant after adding the covariates. With the covariates, SSAM values predicted avoidant coping (positively) as did the RSES (negatively), unlike study 1. Also unlike study 1, on the CERQ SSAM strengths negatively predicted avoidant coping, (rendered marginal by the covariates) and the LOT also predicted it negatively. In study 2, the RSES did not predict CERQ coping.
Hypothesis 3. Mediation analyses were conducted using the lavaan package in R (Rosseel, 2012). To reduce complexity and aid interpretation, we analyzed mediation separately for mental distress and wellbeing and for the COPE and CERQ. Paths were estimated using bootstrapping (5,000 samples). The tables (Tables 4-7) display direct paths (from coping to specific outcome) and indirect paths (from SSAM factor to specific outcome via coping) where the path was significant in at least one study. If no path is displayed, this indicates the path was non-significant in both studies.

Mental distress
On the COPE, non-avoidant coping did not predict distress (Table 4). On the CERQ, nonavoidant coping significantly predicted less depression in both studies but OASIS anxiety only in study 2 and ONS anxiety only in study 1. In study 2, but not in study 1, values predicted less depression via CERQ non-avoidant coping.
On both the COPE and CERQ, avoidant coping significantly predicted all three distress outcomes and did so positively (Table 5). On both the COPE and CERQ, SSAM strengths predicted less depression and anxiety (both measures) via less avoidant coping. All these findings were replicated in the study 1 data. In study 2, but not study 1, values predicted more depression and OASIS anxiety via CERQ avoidant coping.

Wellbeing
On the COPE and CERQ, non-avoidant coping significantly and positively predicted all outcomes ( Table 6). The only finding replicated in study 1 was for affect balance on the CERQ. On the COPE and CERQ avoidant coping predicted all wellbeing outcomes and did so negatively (Table 7). All these relationships were replicated in the study 1 data.
On the COPE there were significant indirect paths through non-avoidant coping for values with life satisfaction, having a worthwhile life, and flourishing. On the CERQ there were significant indirect paths through non-avoidant coping for strengths with life satisfaction, having a worthwhile life, and flourishing and values with all outcomes. Each source predicted greater coping and better wellbeing. None of these relationships were found in the study 1 data (Table 6).
On both the COPE and CERQ there were significant indirect paths through avoidant coping involving strengths (Table 7): strengths was associated with less avoidant coping and thus greater wellbeing (all five outcomes). These paths involving the COPE were also found in study 1; on the CERQ, however, the only significant indirect path in study 1 involving strengths was for affect balance, albeit other indirect paths involving strengths approached significance. Of note, there were also significant indirect paths on the CERQ involving values through avoidant coping: these predicted lower wellbeing on several outcomes. These relationships were not found in study 1.

Discussion
In essentials, the findings of study 2 largely replicate those of study 1. The data supported the predictions that the SSAM would correlate positively with indices of wellbeing and non-avoidant coping and negatively with indices of mental distress and avoidant coping. As in study 1, in the SEM analyses SSAM strengths predicted the mental health outcomes, largely because of its relationship with the covariates, especially self-esteem, whereas Questionnaire; 95% Confidence Intervals in brackets. a direct path = non-avoidant coping to outcome; indirect path = SSAM factor to outcome via non-avoidant coping. ^p < .10. *p < .05. **p < .01.
SSAM social relations predicted several indices of wellbeing, but only once controls for self-esteem were introduced; again, self-affirmation through values did not predict mental health, with or without the covariates.
With regard to coping, in the SEM analyses the findings involving the COPE were substantially replicated. In study 2 the relationships involving the CERQ were also more consistent with those found on the COPE than they had been in study 1. Thus SSAM strengths predicted non-avoidant coping on the CERQ and less avoidant coping on both the CERQ and the COPE; the latter relationship was maintained after adding the covariates. SSAM values predicted more non-avoidant coping on both the CERQ and the COPE, with and without the covariates. Consistent with study 1, SSAM social relations generally did not predict coping (though it approached significance for non-avoidant coping on the COPE). Unlike study 1, however, SSAM values positively predicted avoidant coping on the COPE once self-esteem was controlled for. 7  95% Confidence Intervals in brackets. a direct path = avoidant coping to outcome; indirect path = SSAM factor to outcome via avoidant coping. *p < .05. **p < .01. ***p < .001.  Mediation analyses indicated mediation through both forms of coping on both coping measures. However, study 2 revealed more evidence of mediation than did study 1; these typically involved better outcomes (less distress; greater wellbeing), although the path from values through avoidant coping was an exception, predicting more distress and lower wellbeing. The most robust finding (in terms of replication across studies and inventories) involved strengths being associated with less distress and greater wellbeing via reduced avoidant coping.
In study 1, strengths predicted more anxiety after controlling for the covariates. In study 2, it predicted less rather than more anxiety after controlling for the covariates. We have no explanation for this inconsistency, albeit the findings of study 2 are more consistent with our working hypothesis.
The strong positive relationship between the RSES and LOT-R in this study makes it hard to interpret the effects of these variables in the combined model; they are best interpreted when entered alone (see , Table 3).

General discussion
In two studies we have established that the tendency to report spontaneously selfaffirming in response to threat is, as predicted, associated with less reported anxiety and depression as well as greater wellbeing. Again, as predicted, it is also associated with more non-avoidant and less avoidant coping. However, when further unpacked, the relationship between spontaneous self-affirmation and mental health clearly varies with the reported source of the self-affirmation. These sources, in turn, differ in their relationships with coping.
As proposed, trait self-esteem plays a role in the relationships between spontaneous self-affirmation and mental health. However, this role varies with the source of selfaffirmation. Harris et al. (2019) hypothesized that a focus on personal strengths as a source of self-affirmation would be distinctive because it would correlate with selfesteem more strongly than would the other sources of self-affirmation and this was clearly the case in the two studies reported here. In both samples, before controlling for selfesteem, the tendency to focus on strengths invariably predicted less mental distress and greater wellbeing, whereas the tendency to focus on values or social relations did not. However, controlling for self-esteem typically eliminated the positive relationships between strengths and reported mental health, which is consistent with the view that self-esteem plays a key role in relationships involving a focus on strengths. In contrast, in both studies, after controlling for self-esteem social relations positively predicted indices of wellbeing, but not distress, whereas a focus on values predicted neither. As hypothesized, self-esteem appears to play a different (and less central) role in the relationships involving the other two sources of self-affirmation measured by the SSAM.
It is interesting that self-esteem potentially "disrupts" the benefits for mental health that are known to accrue from thinking about one's positive social relations (e.g., Jetten et al., 2011). We suspect that this is in part because of the implications of a focus on strengths and of social relations for the sense of connection and belonging: to reflect on strengths is to consider characteristics that typically set one apart from others, whereas to reflect on social relations is to consider the very things that connect us to others. This is manifested in the data we collected on feelings of social connection, where a focus on social relations was associated more with feelings of compassion, love and connection to others than was a focus on strengths. For those higher in self-esteem, therefore, these two tendencies may conflict and eliminate the benefits for mental health that are associated with a more wholehearted focus on either. This process may also account for the lack of association between values and mental health, with or without controls for self-esteem. Whether reflecting on values makes social connections salient depends on the value (some values connect us to others, others are neutral, yet others set us apart). Furthermore, a focus on values has only a small to moderate correlation with selfesteem. Therefore, neither the self-esteem nor the connection paths to mental health are particularly strongly associated with the tendency to reflect on values.
Obviously, the diverse patterns revealed in these studies, with their different implications for mental health, create complexity, given that most participants report focusing on all three sources of self-affirmation. Nevertheless, as the correlations involving the overall SSAM in these studies and in the studies of wellbeing reported by Jessop et al. (2022) reveal, aggregated across individuals the correlation is positive with mental health and consistent with our working hypothesis.
The unusually large relationship between the SSAM items and the dispositional optimism item reported in the HINTS studies (e.g., Emanuel et al., 2018;Taber et al., 2015) led us to examine the extent to which optimism played a role in the relationships between spontaneous self-affirmation and mental health independently of self-esteem. In the current studies, controlling for dispositional optimism eliminated fewer of the relationships between strengths and the mental distress and wellbeing measures than did controlling for self-esteem. Moreover, in competitive analyses, optimism and selfesteem predicted various outcomes, but self-esteem did so more often and typically more strongly. Overall, these findings suggest some role for optimism stripped of selfesteem in the relationships between spontaneous self-affirmation and mental health but that, of the two, self-esteem is the more important. 8 The second element of the studies we report here builds on the work of Harris and colleagues (Harris et al., 2019) concerning how, if at all, the SSAM and its first order factors relate to other ways of coping. In the current studies we extended our research programme by testing whether the SSAM and its factors consistently predicted responses on two widely used inventories of coping with threats to mental health. They did.
In both studies the SSAM correlated positively with indices of non-avoidant coping and negatively with indices of avoidant coping. However, SEM analyses showed that the different sources of self-affirmation assessed by the SSAM had diverse relationships with coping and again the findings replicated across both studies. Again, all three sources had a divergent set of relationships before introducing the covariates. Again, a focus on strengths had a set of relationships that were generally eliminated by controlling for selfesteem, a focus on values had a set of relationships that were not, and a focus on social relations rarely predicted dispositional coping. Specifically, in both studies SSAM values positively predicted the non-avoidant coping cluster of the COPE and SSAM strengths positively predicted the non-avoidant coping cluster of the CERQ; however, whereas prediction by values was not eliminated by controls for the covariates in either study, in study 2 prediction by strengths was. In both studies SSAM strengths negatively predicted the avoidant coping cluster of the COPE; this was eliminated by controls for the covariates in study 1, but not in study 2. Indeed, study 2 tended to uncover more relationships between the SSAM factors and coping, including a positive relationship between values and non-avoidant coping on the CERQ (with and without covariates) but a positive relationship between values and avoidant coping on the COPE (after adding the covariates). Notably the LOT and RSES were only occasional predictors of coping. The differences in findings between the two studies presumably reflect the greater power of study 2; the reasons for the differences in findings between the two coping inventories are unclear but may reflect their differential emphases on behavior (COPE) and cognition (CERQ).
In the analyses of mediation through coping, the most robust indirect "effects" (in terms of both direction of effect and consistency across studies and coping measures) involved strengths through avoidant coping: strengths predicted less avoidant coping and thereby less distress and greater wellbeing (in study 2). 9 Although other indirect effects were found, they were less consistent across studies and measures. In all cases involving non-avoidant coping, however, the source predicted less distress and greater wellbeing. In contrast, at least in study 2, values sometimes predicted more distress and lower wellbeing via avoidant coping. 10 Although we did not have specific predictions about which of the sources would predict distinctively, the findings are consistent in broad terms with our pre-registered hypotheses that the SSAM would predict more non-avoidant and less avoidant coping and concerning mediation of outcomes through coping. Alongside the findings of Harris et al. (2019) a picture is emerging of how the SSAM and its factors sit alongside other ways of coping and are largely associated with more constructive (in the current studies, less avoidant and more non-avoidant) ways of coping.
In their seminal work, Spencer and colleagues Steele et al., 1993) argued that self-esteem underpins individual differences in resiliency to self-threats because those higher in self-esteem possess more self-evaluative resources with which to self-affirm. The findings of our research programme support, qualify and extend this. Those higher in self-esteem report tending to favor focusing on their strengths to selfaffirm, whereas those lower in self-esteem tend to favor using values and (especially) their social relations. In this paper we have demonstrated that these sources of self-affirmation diverge in their relationships with mental health, wellbeing and coping.
In experimental work some attention is being paid to ways in which the source of the self-affirmation may influence processes and perhaps even outcomes (e.g., Chen & Boucher, 2008;Iles et al., 2021;Schimel et al., 2004;Shnabel et al., 2013;Zhu & Yzer, 2021). For example, there is evidence that affirmations may be particularly effective when they describe or emphasize belonging socially (Shnabel et al., 2013). Of note, the current studies are consistent with the possibility that manipulations that ask people to reflect upon their strengths and attributes may induce different processes or have different consequences from those that ask people to reflect upon their values and principles or their social relations (Harris et al., 2019;Webb et al., 2020).
Either way, when investigating spontaneous self-affirmation, one implication is that researchers should be sensitive to situations in which trait self-esteem is consistently and strongly related to the focal outcomes. Where this is the case, there is potential for a focus on strengths, which is more strongly related to self-esteem, to deviate in its relationships from the other sources assessed by the SSAM, which will require researchers to be alert to the possibility that they should use the subscales as predictors alongside the overall SSAM. More generally, researchers using the SSAM should be sensitive to any evidence in their data that the three subscales relate to the principal outcomes in different ways, where possible using analyses that enable them to test for this and its implications for the overall SSAM factor (e.g., by using the subscales as indicators of a higher-order SSAM factor in SEM analyses). In this regard, uncovering the moderators that influence when the SSAM does and does not function well as a higher-order factor would be useful.
Naturally, in drawing conclusions from these studies we should be alert to their limitations. The samples are opportunity samples of predominantly white British participants. There was attrition in Study 2, which could also affect generalizability, albeit the retained participants (57%) did not differ demographically from those who dropped out. Although we pre-registered our hypotheses and analyses, we were obliged to deviate from some elements of these, which meant that some hypotheses were not tested exactly as planned and there were some departures from the proposed analyses. (These are summarized in the SOM.) Nevertheless, although caution should be exercised when interpreting any individual finding, especially if statistically marginal, we are struck by the extent to which the findings replicated in essentials across the studies and reassured by the fact that they did.
What if any are the practical implications? Research on self-affirmation has the potential to enhance our understanding of mental health and may even (in due course) provide one means of helping people to cope with the threats and challenges that foster mental health problems. Of course, we should always be careful about extrapolating from correlational data to inform interventions, especially relatively early on in a research program. However, one potential implication is that practitioners keen to encourage their clients to self-affirm (assuming they do not do so naturally and spontaneously) might find this easier to achieve if they match the means to the client's self-esteem: those higher in self-esteem may be encouraged to focus on their strengths, whereas those lower in self-esteem may be encouraged to focus on their social relationships. This may have the advantage of matching the client's natural inclinations and resources (Harris et al., 2019). Practitioners not willing to take any risks when encouraging people to self-affirm might encourage them to focus on their values. However, these are tentative suggestions about which we must be very cautious.
Overall, the findings of the current studies extend our understanding of the mental health correlates and potential implications of individual differences in spontaneous selfaffirmation and how, in this domain, spontaneous self-affirmation relates to other aspects of self-functioning and identity. They show that the tendency to report responding to threats by self-affirming is typically associated with better mental health indices and coping strategies, but also that the picture is nuanced when it comes to a focus on specific sources of self-affirmation and the role that self-esteem (and to a lesser extent dispositional optimism) play in these relationships. Overall, they are generally consistent with the hypothesis that spontaneous self-affirmation tends to function as a resource that helps foster positive coping with threats. Next steps in the research process include replication and testing of the extent and implications of the mental health correlates of spontaneous self-affirmation, clarifying the mechanisms involved, and testing for moderators. The use of designs that enable us to shed light on potential causal mechanisms, such as cross-lagged designs (e.g., Lannin et al., 2021) would be beneficial.