Testing the Mindfulness-to-Meaning Theory in Daily Life

The mindfulness-to-meaning theory (MMT) describes the processes through which mindfulness leads to enhanced eudaimonic wellbeing (indirectly via mediating processes such as increased decentering, reappraisal, positive affect, and savoring), but little is currently known about how these processes impact one another over short time periods (e.g., across several hours). The current study tested the MMT by measuring these variables repeatedly as they occur naturalistically in daily life. Three hundred and forty-five community members aged 18–65 completed surveys on smartphones six times per day over 7 days, assessing their current levels of decentering, reappraisal, positive affect, savoring, and wellbeing, as part of a larger study. Multilevel structural equation modeling in Mplus was used to analyze the nested data with mediation models. There was a significant indirect effect through the proposed MMT pathway at the within-person level, with all variables measured concurrently. Lagged mediation examining prospective effects indicated that the full indirect MMT pathway was not significant in predicting later wellbeing, though some individual indirect pathways were significant prospectively. Follow-up analyses testing alternative temporal ordering suggested bidirectional effects of savoring and positive affect in explaining the mutual association between decentering and wellbeing. Overall, this study found support for hypothesized MMT processes in daily life and measured over short time periods, with evidence for bidirectional effects for some processes. However, reappraisal showed inconsistent effects, requiring further study and replication using ecological momentary assessment designs.

Mindfulness, rooted in traditional Buddhism, is an intentional, non-judgmental awareness of experience in the present moment (Kabat-Zinn, 1990). Mindfulness is effective in the treatment of various psychological disorders (see Grecucci et al., 2015;McConnell & Froeliger, 2015) and is becoming increasingly prominent in the general population as a way of cultivating a healthy mind (Garland et al., 2017a;Holzel et al., 2011). There are numerous ways to conceptualize mindfulness: trait mindfulness refers to the relatively stable tendency to think and behave mindfully in everyday life, which can be distinguished from naturalistic mindfulness states (e.g., paying attention to eating on a particular occasion) and formalized practice (e.g., focusing on breathing) (Pepping & Duvenage, 2016). It is thought that repeated formalized practice increases one's natural disposition to be mindful in everyday life by leading to more frequent naturalistic mindful states (Garland et al., 2017a). However, it is not currently well understood how momentary mindful states in daily life may lead to positive mental health outcomes.
One observed outcome of mindfulness is higher eudaimonic wellbeing (referred to as wellbeing throughout this paper), which may be defined as a sense of meaning, purpose, and vitality that fosters a flourishing life (Bloch et al., 2017;Garland et al., 2015a). This concept is distinct from hedonic wellbeing, which is focused on pleasure but not meaning systems. To better understand the connection between mindfulness following stressful experiences and eudaimonic wellbeing, Garland et al. (2015a) proposed a cognitive-emotional process model called the mindfulness-to-meaning theory (MMT). Although MMT includes numerous potential mechanisms, we focus here on four key components: decentering, reappraisal, positive affect, and savoring. These components capture the core mediating outcomes that link mindfulness to meaning and are considered key mechanisms in the MMT, whereas a range of other features may be considered sub-hypotheses in the model (e.g., broadened attention, Garland et al., 2015b).
In brief, MMT posits that this process begins with decentering, which is characterized by an objective and psychologically distanced perspective on internal experiences specifically (e.g., viewing thoughts as just passing mental experiences, rather than "buying into" them; Bernstein et al., 2015). It is possible to decenter from both positive and negative emotions (Naragon-Gainey & DeMarree, 2017a), and the psychological distance created by decentering may allow people to respond to different emotional states more flexibly. Decentering is often considered the key "meta-mechanism" that facilitates successive processes, and therefore plays a crucial role in how mindfulness engenders meaning Shapiro et al., 2006). Decentering interrupts the default activation of cognitive schemas and habitual reactions by broadening one's attention to different elements of present experience, as well as facilitating positive reappraisal (Grecucci et al., 2015;Shapiro et al., 2006). Positive reappraisal (referred to as reappraisal throughout this paper) is an adaptive process that involves changing how one thinks about a negative emotional stimulus to be more benign, meaningful, or growth promoting, and this altered perspective may then increase positive emotions. In MMT, particular emphasis is placed on reappraisal as the core cognitive mechanism that perpetuates changes in positive emotion and construction of meaning from experience (Garland et al., 2017a). Greater experiences of positive emotion provide opportunities to reflect on and savor the positive emotions that arise. Savoring of positive experiences is facilitated by reappraisal, as attentional resources that would otherwise be involved in rumination are "freed up" and can be redirected to pleasurable stimuli (Bryant & Smith, 2015;Cheung & Ng, 2020). Savoring is thought to trigger networks of associations and meaning with daily life events. This contextualizes events into a deeper meaning system, which encourages a sense of meaning/purpose that leads to fulfilment (Garland et al., 2015a). Additionally, MMT postulates that these processes unfold in a self-reinforcing "upward spiral" over time, mutually influencing one another to promote wellbeing. Specifically, MMT predicts that meaning develops through repeated reappraisal and savoring, stimulated by decentering. As meaning continually develops, positive affect is increased as tendencies to reappraise and savor are reinforced, promoting this cycle again. This self-reinforcing system among the processes increases flourishing and greater engagement with life, leading to increased wellbeing (Garland et al., 2015a). Bryant and Smith (2015) also suggested that one may appreciate and savor experiences of meaning, stimulating the process over again.
MMT posits that its processes operate at several different timescales: dispositional tendencies, long-term temporal dynamics, and short-term momentary processes. At the dispositional level, MMT suggests that one's tendency to be mindful would be positively associated with their overall experience of wellbeing, and that this association could be explained by individual differences in tendencies to decenter, reappraise, and savor positive affect. Indeed, past research has primarily tested subsets of MMT processes using cross-sectional self-report measures of dispositional tendencies (e.g., Chu & Mak, 2020;Garland et al., 2017a;Rogge & Daks, 2021;Tan et al., 2021). Longitudinal studies, with time scales of weeks to years, have also generally supported the MMT, offering greater evidence for the temporal ordering of variables (e.g., Cheung & Ng, 2020;Garland et al., 2017b;Hanley & Garland, 2014, but see Tan, 2019. For example, in a recent test of MMT, Hanley et al. (2021) tested several components in the hypothesized order (i.e., decentering, broadened awareness, reappraisal, and wellbeing), comparing participants who received mindfulness training to those who did not, using five assessments over the course of 6 to 7 years. They found predicted effects over time in the hypothesized process order, with less support for alternative temporal orderings. Although mediation was supported for simple indirect effects, the full indirect pathway was not significant.
Thus, there is initial support for the temporal effects of some MMT processes over the course of months or years. However, MMT suggests that the timescale of these processes could unfold quite quickly, over minutes or hours (Garland et al., 2015a;Hanley et al., 2021). Furthermore, researchers have highlighted that ecological measurement of dispositional mindfulness is needed to understand how mindfulness and related variables manifest in daily life (e.g., Kaplan et al., 2018). An ecological momentary assessment (EMA) design, or repeated intensively sampled momentary assessments conducted in daily life, is ideally suited for addressing both of these issues. Furthermore, EMA designs allow for temporal tests of within-person processes that capture how levels of variables on one occasion affect their levels at the next occasion of measurement (Hamaker, 2011). Such findings can identify potential causal processes that could be targeted in prevention or intervention programs. Importantly, numerous EMA studies indicated that states of mindfulness and wellbeing-as well as hypothesized mediating processes like decentering, reappraisal, positive affect, and savoring-fluctuate over the course of hours, likely due in part to different affective and contextual experiences (e.g., Du et al., 2018;Enkema et al., 2020;Jose et al., 2012;Suelmann et al., 2018).
It is unclear how these processes jointly operate in daily life as they have not been assessed collectively. However, EMA and daily diary studies have generally reported theory-consistent associations between isolated pairs of MMT constructs, including mindfulness and positive affect (e.g., Brockman et al., 2016;Du et al., 2018;Goldberg et al., 2020), reappraisal and positive affect (e.g., Colombo et al., 2021;Nezlek & Kuppens, 2008;Pavani et al., 2016), and mindfulness and wellbeing (Goldberg et al., 2020). Most studies have found bidirectional effects, when time lags have been examined. However, Brockman et al. (2016) found daily reappraisal use predicted lower levels of daily mindfulness. Fewer studies have looked at savoring in daily life, but savoring mediated the impact of daily positive events on momentary positive affect (Jose et al., 2012), and savoring enhanced positive emotions related to positive daily events (Doorley & Kashdan, 2021). Therefore, there is some support for bivariate associations among these processes in daily life, but they have not been tested as a larger whole model consistent with MMT.
The aim of the current study was to determine the temporal effects of MMT processes in daily life over the course of several hours, through serial mediation modeling of EMA data. We hypothesized findings would be consistent with MMT predictions, such that there would be positive and significant within-person concurrent associations (i.e., associations on a given occasion) and mediation among MMT variables (see Fig. 1a, with MMT key pathways shown along the exterior). We also hypothesized the presence of an indirect effect when variables were lagged in the order specified by the model, testing two different lagged intervals (see Fig. 1b and c). Last, as an exploratory analysis, we tested alternative lagged mediation models where the order of these variables was reversed (see Fig. 2) to examine the specificity of temporal ordering in MMT and possible bidirectional effects consistent with an "upward spiral."

Participants
This study conducted secondary data analysis on a sample of 379 community members aged 18 to 65 from the USA collected in 2017-2019; no other published manuscripts from this dataset report this set of variables together or examine their indirect effects Park & Naragon-Gainey, 2020). Individuals who were currently seeking or receiving psychological treatment were over-sampled to increase variability in depression and anxiety symptoms. We therefore consider this sample to be a community sample with elevated rates of psychopathology, intermediate between a community and clinical sample. Participants were recruited through advertisements on online sites such as Craigslist, newspapers, and flyers in public spaces such as coffee shops, universities, and mental health clinics. People were excluded from participating if they did not speak English and/or were diagnosed with untreated dementia, schizophrenia, or a cognitive impairment, as this may have caused differences in baseline cognitive tasks (not analyzed in this study).
Of the 379 participants who participated in the baseline portion of the study, 356 enrolled in the EMA portion of the study. Of these, an additional 11 participants were excluded from EMA analyses because they submitted fewer than 30% valid EMA reports, consistent with prior EMA research (e.g., Moran et al., 2017). The final sample for these analyses did not differ from the 34 participants who completed the baseline study but did not have EMA data (i.e., 23 did not enroll in the EMA study and 11 had too few reports) in terms of sex, age, education level, and employment status. However, there were significant differences on race and income between the two groups (ps < 0.05). Specifically, participants not included in EMA analyses were more likely to report low household income (< $10,000 per year), more likely to identify as Black or African American, and less likely to identify as White or Asian.
The final sample consisted of 345 participants (67.0% female; mean age = 34.52, SD = 14.03, range = 18-65). In the final sample, 66.4% of participants identified as White, 13.6% as Black or African American, 12.8% as Asian, 6.7% as more than one race, and 0.6% as Native American or Alaska Native. Most of the sample was currently working (33.9% part-time and 24.9% full-time), and 24.9% were unemployed. In addition, 31.9% of participants were full-time students, and 6.1% were part-time students. Multiple selections were possible, and 19% of the sample reported both working and in education (either full-or parttime). Most participants had a gross household income of less than $40,000 annually (less than $10,000 = 30.5%; $10,000-$20,000 = 16.5%; $20,000-$40,000 = 22.6%; $40,000-$60,000 = 11.0%; $60,000-$80,000 = 6.1%; more than $80,000 = 13.4%). A majority of the participants (60.6%) reported having some experience with meditation or mindfulness practice, with a reported mean duration of 32.1 months (SD = 54.9, range = 1 to 410). Based upon a semi-structured diagnostic interview, 41.6% of the sample met criteria for one or more emotional disorders, with the most common diagnoses being social anxiety disorder (24.9%), generalized anxiety disorder (20.4%), and a unipolar depressive disorder (11.4%). About half of the sample (49.6%) reported currently receiving therapy or taking psychiatric medication.
Individual EMA reports were removed if they were completed outside of the required 30-min response window or completed extremely quickly. After removing invalid reports, there were a total of 11,954 completed reports out of 14,490 possible reports. Thus, a mean of 82.5% of the reports (SD = 14.7%) was submitted and valid, ranging from 31 to 100% of reports across participants.

Procedure
Eligibility was determined through email or phone screening. Once a baseline appointment was scheduled and informed consent was obtained, participants completed a 3to 4-h baseline assessment in the lab, which included measures of heart rate variability, cognitive tasks, a semi-structured diagnostic interview (Anxiety and Related Disorders Interview Schedule for DSM-5, Brown & Barlow, 2014), and self-report surveys on the computer. The baseline data were not analyzed in the current study.
Participants were invited to enroll in a 7-day EMA protocol that commenced within 4 days of the baseline appointment, and these EMA questions are the focus of the current study. A research assistant demonstrated how to complete the EMA surveys, answered any questions, and reviewed example items with all participants. Participants completed brief surveys six times per day on a smartphone device for 7 days, yielding up to 42 surveys per person. This prompt frequency is common for EMA assessments of this duration, to appropriately capture dynamic experiences without being too burdensome to participants (Degroote et al., 2020;Wrzus & Neubauer, 2022). Surveys were sent between 9am and 9 pm via text message links to Qualtrics through the SurveySignal system (Hofmann & Patel, 2015). Survey times were random within each 2-h block, except that they were constrained to be a minimum of 60 min apart. Participants were instructed to complete the surveys within 30 min of receiving the text message in order to be considered as valid. A reminder text message was sent if the survey was not completed within 20 min. Research assistants continuously screened data, and if any problematic responding was recognized, participants were contacted immediately via phone or email. Participants were paid $1.50 for each valid survey and an additional $15 bonus if fewer than 9 of 42 surveys were missed. In total, maximum compensation was $78. As an extra incentive to complete the surveys, each completed survey also entered participants into a drawing for one of four iPads. This study was approved by the Institutional Review Board at the University at Buffalo.

Measures
EMA surveys consisted of several items taken from prior EMA publications, or based on and adapted from existing self-report trait measures of these constructs and theory to be suitable for EMA. Table 1 shows a list of items analyzed in the current study and details about the source of each item. Note that measures were selected for the larger study and therefore are not always ideal representations of MMT processes; in particular, the reappraisal item does not assess positive reappraisal specifically but rather reappraisal generally (see "Limitations and Future Research"). For each survey, participants were provided with instructions that read: "What follows is a series of questions that asks about your recent experiences. Please answer these with respect to the past 30 min, including this moment." Participants rated their response on a 5-point scale from 1 (very slightly or not at all) to 5 (extremely). Reappraisal and savoring were assessed with single items. Decentering items 2 and 3 were reverse scored, and then summed with decentering item 1 to create a composite score with a maximum of 15; the assessment included a fourth decentering item, but structural and validity analyses indicated that this item performed poorly relative to the other items . Decentering items purposefully did not specify a particular valence of emotion, as this study was interested in decentering experiences in general (i.e., from both negative and/ or positive experience) (see DeMarree & Naragon-Gainey, 2022 for a discussion of valence in the assessment of decentering). Four positive affect items were summed to create a composite score with a maximum of 20, and two wellbeing items were summed for a maximum score of 10. Multilevel internal consistency analyses (see Geldhof et al., 2014) revealed acceptable reliability for scores on decentering (within-person omega = 0.62) and positive affect (within-person omega = 0.82). There was also a strong correlation between the two wellbeing items (within-person r = 0.56). To examine the convergent validity of these items, some of which were novel, we computed the correlation between their between-person variance and validated trait measures of the same construct administered at baseline (i.e., Reappraisal from the Thought Control Questionnaire (Wells & Davies, 1994); Decentering from the Experience Questionnaire (Fresco et al., 2007); Positive affect from the Positive and Negative Affect Schedule (Watson et al., 1988); Savoring from the Emotion-Focused Rumination Scale of the Responses to Positive Affect Questionnaire (Feldman et al., 2008); Wellbeing from the Questionnaire for Eudaimonic Wellbeing (Waterman et al., 2010). All convergent correlations were significant at p < 0.001: the correlation for reappraisal = 0.22, decentering = 0.43, positive affect = 0.69, savoring = 0.40, and wellbeing = 0.47. In addition, the construct validity of decentering was supported given a moderate and significant correlation (r = 0.43; p < 0.001) with trait mindfulness (assessed with the Five Factor Mindfulness Questionnaire; Baer et al., 2006).

Data Analyses
The data have a two-level nested structure, with survey responses (n = up to 42) nested within participants (N = 345). As such, multilevel structural equation modeling (MSEM) was used to test a number of serial mediation models. Intraclass correlations (ICCs) were also consulted to identify the need for multilevel modeling, as these values indicate the proportion of variability attributed to between-persons (vs. within-person). Indicators were modeled as observed variables, rather than factors, given that some variables were measured with single items and to increase model parsimony. MSEM improves upon traditional multilevel modeling because MSEM separates the within-and betweenlevel variance completely by creating orthogonal latent variables (i.e., controls within-person level variability when analyzing the between-person level, and vice versa) (Sadikaj et al., 2019). This separation of variance at each level means that within-person variables are interpretable as person mean-centered that remove individual differences. Although between-person effects are not the primary focus of the current study and are not as well-suited for mediation analysis due to the lack of temporal information, we report between-person correlations and mediation analyses of the MMT in the online supplement for interested readers (Appendix 1; results suggest mixed support for the MMT pathways but little evidence for hypothesized indirect effects).
Multilevel mediation was used to examine the MMT in several ways, modeling the key MMT pathway (i.e., decentering to reappraisal to positive affect to savoring to wellbeing), and including all direct and indirect paths among them. First, we examined within-person concurrent associations and mediation (i.e., associations on a given occasion), specified in the order hypothesized by the MMT. Next, we examined two plausible lagged models to test the temporal precedence of effects. In the first model, only the outcome variable-wellbeing-was assessed at the next report (T + 1). The second model covered larger periods of time, with the first predictor variable-decentering-at T, all mediators at T + 1, and wellbeing at T + 2. Last, given that variables in MMT are suggested to relate in a bidirectional manner (upward spiral) but this has not been tested previously, two alternative lagged models were also tested, corresponding to the primary lagged models but with the order of variables reversed. Specifically, the alternative models consisted of (1) wellbeing at T and the other variables (i.e., savoring, positive affect, reappraisal, decentering) at T + 1, and (2) wellbeing at T, savoring, positive affect, and reappraisal at T + 1, and decentering at T + 2.
Bayesian estimation in Mplus Version 8.4 (Muthén & Muthén, 1998 was used to correctly model asymmetric distributions for indirect effects, as bootstrapped confidence intervals are not currently available in MSEM with other types of estimation (such as maximum likelihood). Non-informative priors were used to freely estimate parameters, given that informative priors need to be generated from results of previous studies and there are no within-person studies of this model to draw from (see Muthen & Asparouhov, 2012). Parameter estimates are presented with 95% credibility intervals, which is the interval that contains 95% of probable values (i.e., the effect has 95% probability of falling within this range). As such, the effect is interpreted as "significant" if the 95% credibility interval does not contain zero (Muthen & Asparouhov, 2012). Bayesian estimation does not provide the standard SEM fit indices that maximum likelihood and other estimators yield, but we report the posterior predictive p-value (PPP) and its 95% confidence interval. PPP values near 0.5 indicate excellent model fit, as do PPP confidence intervals that do not include zero (Asparouhov & Muthén, 2020).
Lagged variables for decentering, reappraisal, positive affect, savoring, and wellbeing were created to predict level of variables at the next time point from the prior time point. The time elapsed between reports varied across people and occasions, since they were sent at quasi-random intervals, so the time lag was included as a covariate in all lagged paths to account for variation in time between reports. The average time between reports at T (i.e., the current report) and T + 1 (i.e., the next report) was 3.67 h (SD = 4.22), while the average time passed between reports at T and T + 2 was 7.56 h (SD = 5.54). Additionally, the report number (i.e., 1-42) was included as a covariate in all models to account for changes in reporting style (e.g., fatigue, boredom, or changed understanding of items) or levels of variables over the course of the week. Autoregressive paths were included for regressions that involved variables at different time points (i.e., lagged associations) to control for prior levels of the same variable, thereby predicting residualized change over time. For lagged models, between-person variances and covariances were freely estimated. Table 2 presents within-person correlation matrices, means, and standard deviations, as well as ICCs for each variable. ICCs ranged from 0.39 to 0.59, which confirms the need for multilevel analysis, as there was substantial variance at both within-and between-person levels (Hox, 2002). At the within-person level, correlations were positive and mostly statistically significant (p < 0.01), with small to medium effects. The association between wellbeing and positive affect was large (r = 0.64), whereas the association between decentering and reappraisal was not significant (r = 0.02, p = 0.30).

Concurrent Mediation Model
Associations among variables within-persons at the same report are presented in Fig. 1, panel A, and the model was an excellent fit to the data (PPP = 0.75; 95% CI = − 28.44 to 9.23). Regression paths in the hypothesized key pathway (i.e., decentering to reappraisal to positive affect to savoring to wellbeing) were positive and significant, as was the direct effect of decentering on wellbeing. Unstandardized indirect effect estimates and 95% credibility intervals are presented in Table 3. All possible indirect pathways were positive and statistically significant, including the MMT full indirect pathway of interest through the key variables (although small in magnitude). This reflects partial mediation and suggests that reappraisal, positive affect, and savoring (jointly and individually) explain some, but not all, of the relationship between decentering and wellbeing on a given occasion.

Primary Lagged Mediation Models
Decentering T scores were positively associated with Wellbeing T+1 scores (r = 0.15, p < 0.001) and Wellbeing T+2 scores (r = 0.09, p < 0.001), so we proceeded to conduct tests of lagged mediation. Note that autoregressive effects and the time interval between reports were controlled for in all lagged paths, though for clarity they are not presented in the figures.
We first tested a model with all variables assessed at T, except wellbeing which was assessed at T + 1. This model was not a good fit to the data: PPP = 0.00, 95% CI = 5156.33 to 5339.05. Regression paths are presented in Fig. 1, panel  B, and unstandardized indirect effect estimates and credibility intervals are presented in Table 3. Decentering T did not significantly predict Reappraisal T , but the other hypothesized regression paths were positive and significant, including the lagged direct effect of Decentering T to Wellbeing T+1 , as well as Savoring T to Wellbeing T+1 . The full MMT indirect pathway of interest was not statistically significant, but all indirect pathways that excluded reappraisal were positive and significant. Thus, partial mediation of the lagged association between decentering and wellbeing was observed for positive affect and savoring.
Next, we tested a model with decentering assessed at T; reappraisal, positive affect, and savoring at T + 1; and wellbeing at T + 2. This model also did not provide a good fit to the data: PPP = 0.00, 95% CI = 8854.86 to 9209.36. Regression paths are presented in Fig. 1, Table 3. The pattern of results was consistent with the previous lagged mediation model. Specifically, key pathways and the lag-2 direct effect of decentering on wellbeing were positive and significant, except that Decentering T scores did not predict Reappraisal T+1 scores. The full MMT indirect pathway of interest through key variables was also not statistically significant. However, indirect pathways including positive affect and/or savoringand excluding reappraisal-were significant. This suggests that positive affect and savoring partially account for the relationship between decentering and wellbeing, even when

Alternative Lagged Models with Reverse Temporal Ordering
Two alternative models tested whether observed effects are specific in temporal ordering, both of which did not provide a good fit to the data based upon PPP (model 1:  were generally significant and positive. However, there was not a direct effect of Wellbeing T scores on Decentering T+1 scores, and Reappraisal T+1 was inversely associated with Decentering T+1 . Although not a key path, it is also noteworthy that Wellbeing T scores were negatively associated with Reappraisal T+1 . All indirect effect paths, including the main indirect effect and pathways with reappraisal, were significant, indicative of full mediation.
In the second mediation model, variables were again in reverse order but now temporally separated by two reports, with wellbeing at T, savoring, positive affect, and reappraisal at T + 1, and decentering at T + 2. Regression paths are presented in Fig. 2, panel B, and unstandardized indirect effects are presented in Table 4. Among key pathways, all were positive and significant, including the lagged direct effect of Wellbeing T on Decentering T+2 , except that the Reappraisal T+1 to Decentering T+2 association was not significant. The primary indirect pathway of interest was not statistically significant, but all pathways that excluded reappraisal were statistically significant. Taken together, these alternative models suggest bidirectional effects of savoring and positive affect, but not reappraisal, in explaining the mutual association of wellbeing and decentering.

Discussion
This study found some support for hypothesized processes in the MMT in daily life, demonstrating how decentering engenders meaning throughout the course of the day. Specifically, the hypothesized indirect effect from decentering to wellbeing was supported for within-person concurrent analyses. These results replicate and extend prior cross-sectional and longitudinal tests of the MMT to suggest that MMT processes are associated as expected when assessed in one's daily life, supporting the model's ecological validity. However, this within-person concurrent analysis does not allow for a strong test of mediation because temporal precedence cannot be established. Lagged analyses that specify temporal ordering revealed a more complex picture. That is, when temporally parsed over the course of the day, decentering predicted increases in wellbeing, through an indirect effect via positive affect and savoring, but not via reappraisal. Furthermore, alternative temporal models in reverse order revealed bidirectional effects between decentering and wellbeing, positive affect and savoring, and savoring and wellbeing, supporting the idea of an upward spiral among these MMT processes.

What is the Role of Reappraisal?
The theory-consistent associations between decentering, positive affect, savoring, and wellbeing in the current study are reflective of past studies, including traditional longitudinal designs (e.g., Chu & Mak, 2020;Hanley et al., 2021;Tan et al., 2021). Regarding mediation, the pattern of indirect effects is largely similar to findings of Garland et al. (2017a), where dispositional mindfulness was linked to increased reappraisal, savoring of positive experience, and meaning in life for cancer patients. However, when analyzed at a short time scale in the current study, reappraisal did not play a role in linking decentering to meaning several hours later. Specifically, although reappraisal predicted greater positive affect, it was not related to prior decentering or subsequent wellbeing as expected. For example, reappraisal was frequently (a) not significantly related to decentering, (b) not significantly related to wellbeing, (c) not present in significant indirect pathways connecting decentering to wellbeing, and (d) inversely predicted by wellbeing and inversely predictive of decentering in alternative models. This is intriguing given the emphasis placed on reappraisal as the core mechanism in MMT (Garland et al., 2015a), and empirical support found in some prior longitudinal studies over time periods of months or years (e.g., Cheung & Ng, 2020;Hanley et al., 2021;Garland et al., 2017b). However, past evidence is somewhat mixed, as Tan, (2019) found that reappraisal did not influence meaning after 2 weeks, and Brockman et al. (2016) found that daily reappraisal predicted lower levels of daily mindfulness. Conceptually, Bryant and Smith (2015) suggested that mindfulness and savoring may be linked in the absence of reappraisal. Indeed, MMT extensions have concurred with this idea, at least on brief time scales (Garland et al., 2015b), and have noted that reappraisal may primarily exert its influences over longer time scales as its cumulative impacts manifest (Hanley et al., 2021). In this way, positive affect and savoring may play larger roles in how decentering promotes wellbeing day-today while effects of reappraisal are not manifested strongly over short time intervals, consistent with our results. Turning to measurement considerations, it is important to consider the relative infrequency with which reappraisal is employed in daily life. In particular, and consistent with some prior EMA studies (e.g., McMahon & Naragon-Gainey, 2020), reappraisal use was relatively low in the current sample (overall mean = 1.80 out of 5, SD = 0.96). This is likely because reappraisal is most relevant in response to aversive or stressful situations, which are generally infrequent on an average day for most people (Garland et al., 2015a). Furthermore, reappraisal is relatively cognitively effortful (Keng et al., 2013), so people may instead choose to use lower effort strategies like distraction (e.g., Sheppes & Meiran, 2008). In addition, it is important to consider the possibility that the single EMA item used in the current study ("I changed the way I thought about what caused my feelings") may not have adequately captured reappraisal or may have been interpreted in different ways by participants. For example, the item was not specific to positive reappraisal, as hypothesized by MMT theory, and did not specify how people changed the way they were thinking, which could conceivably include maladaptive processes like increased self-blame or frustration. Of note, the reappraisal item was significantly but not strongly correlated with the baseline measure of reappraisal (r = 0.22), with poorer levels of convergence relative to the other EMA variables in this study. Thus, it will be important to replicate this finding with validated and multi-item measures of reappraisal in daily life before drawing strong conclusions.

MMT Upward Spiral
The current study examined the extent to which these processes may have mutual effects when considered together, consistent with MMT theory (e.g., Garland et al., 2015a). We found that just as current decentering predicted later wellbeing via greater intermediate positive affect and savoring, so too did current wellbeing predict later decentering via greater intermediate savoring and positive affect. This suggests that the processes enhance one another, consistent with predictions of MMT's upward spiral. In particular, savoring and positive affect, savoring and wellbeing, and reappraisal and savoring all had bidirectional associations, consistent with some past studies that examined these pairs in isolation (e.g., Doorley & Kashdan, 2021;Jose et al., 2012) and theoretical predictions (e.g., Bryant & Smith, 2015;Garland et al., 2015a). Overall, these findings suggest a benefit to learning multiple skills, and that the skills may mutually support one another, ultimately promoting meaning and wellbeing in daily life. This is consistent with the Mindfulness-Oriented Recovery Enhancement therapeutic approach, which integrates mindfulness, reappraisal, and savoring practices (Garland et al., 2015b). As the primary model and models with reversed directionality both produced significant indirect pathways, we should remain cautious of assuming specific temporal ordering of these processes, at least in daily life scenarios. This may reflect the complex intertwined nature of the processes in daily life, while a clearer pattern appears over longer trajectories, as outlined in MMT. These results may suggest that even implementing small changes in strategy use may promote an immediate sense of wellbeing that, over time, promotes the continuous cycle of positive affect and savoring. This allows for maintenance and endurance of positive affect and wellbeing, which may offset negative affect and symptoms of people suffering from clinical disorders, for example.

Limitations and Future Research
Numerous limitations should be considered when interpreting these results. First, although EMA designs provide rich temporal data and enhance ecological validity, they require the use of short questionnaires to be feasible, and few such measures have been validated for EMA use. In particular, single items may be problematic if they are not clear, have poor validity, or do not measure all aspects of a given construct. As such, our findings should be replicated where possible with validated EMA measures of MMT processes, including a measure of positive reappraisal specifically. Additionally, there are limitations associated with using archival data. For example, although we examined shorter time scales than previous studies that are likely closer to natural variations in these processes, it is reasonable to think they may happen even faster (i.e., seconds to minutes; Hanley et al., 2021). Laboratory studies may be better suited to capture this timescale, and the current study was not able to detect effects that dissipated more quickly than a few hours given the pre-determined time intervals of assessment. Like prior studies, we did not comprehensively test all aspects of the MMT (e.g., broadened attention, working memory, response inhibition) due to the large number of components and lack of availability of some measures. Furthermore, mindfulness is relevant to other outcomes related to psychological health (Keng et al., 2011), and models could be extended to account for variables such as resilience and symptoms of psychological disorders (e.g., reduction of depression and improvements in functioning). It is also noteworthy that the indirect pathway hypothesized by the MMT was quite small in the current study, even when it was statistically significant. However, a small effect size is common in multiple serial mediation models like this, and such effects can still be theoretically or clinically meaningful.
Next, it is important to consider potential limitations regarding generalizability of results from this sample. Although the analyzed sample included people with a range of incomes and races, this sample differed from those without EMA data, which therefore may limit generalizability of the EMA results. Additionally, it is unclear how well results would generalize beyond Western contexts. Similarly, the sample features a large proportion of participants who met criteria for a current psychological disorder, and it is unknown whether MMT operates differently for people with emotional disorders.
In addition, mindfulness practice was not taught in the current study, making comparisons to past cross-sectional and longitudinal studies difficult, as past studies often measured changes based on mindfulness meditation practice. Further complicating comparisons to past work, some research suggests that mindfulness experience impacts mindfulness processes and understanding of mindfulness concepts/measures (Baer & Carmody, 2008). Although 61.5% of participants in this study reported practicing mindfulness at some point in their life, there was a lot of heterogeneity in this experience (range of 0-34 years), which may further complicate comparisons with past work.
Lastly, as highlighted by Garland et al. (2015a), MMT has a bounded scope in that it is only theorized within the context of stressful or aversive experiences. The current study examined how MMT variables naturally occurred throughout the day, rather than targeting occasions of stress or negative affect. Future EMA studies could target stressful events and gather contextual information that may function as moderators (e.g., social context, cognitive resources).
Author Contribution TS: conducted background research, conceptualized study aims, analyzed the dataset, and wrote the first draft and edited subsequent versions of this manuscript. KD: obtained study funding, conceptualized the study and design, recruited participants, collected and cleaned data, provided critical edits, and reviewed multiple versions of the manuscript. KNG: obtained study funding, conceptualized the study and design, recruited participants, collected and cleaned data, and contributed to analyses, as well as editing and writing multiple versions of the manuscript.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions This research was supported in part by NCCIH/ NIH grant 1R21AT009470, "Decentering in Daily Life: Underlying Mechanisms and Impact on Well-Being," to Kristin Naragon-Gainey and Kenneth G. DeMarree. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Declarations
Ethics Approval This study was approved by the Institutional Review Board at the University at Buffalo and has been performed in accordance with ethical standards.

Informed Consent Statement
Informed consent was obtained from all individual participants.

Conflict of Interest
The authors declare that they have no conflict of interest.
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:// creat iveco mmons. org/ licen ses/ by/4. 0/.