Momentary PERMA: An Adapted Measurement Tool for Studying Well-Being in Daily Life

At the heart of most operationalizations of well-being lies a distinction between two fundamental components of well-being: hedonia and eudaimonia. Hedonia is the experience of positive emotions or the “good feeling” (Henderson, et al., 2014). Eudaimonia refers to virtuous living and “functioning well” (Boehm et al., 2016). By definition, hedonia and eudaimonia are interrelated where feeling good can be experienced in tandem with functioning well.

In the field of positive psychology, some frameworks defining well-being have only targeted one component of well-being, either hedonic well-being or eudaimonic well-being. For example, Ryff’s theory of Psychological Well-Being (Ryff, 1989) is a multifaceted framework of well-being that expands on the elements comprising eudaimonia (e.g., purpose in life, autonomy, positive relationships, etc.). On the other hand, Diener et al.’s theory of Subjective Well-Being (1999) captures cognitive and affective evaluations—elements related to hedonia and feeling good. More recently, however, there has been a shift towards understanding well-being by combining elements of hedonia and eudaimonia in one model. For instance, Seligman’s (2011, 2018) PERMA framework of well-being includes elements that comprise both hedonia (e.g., positive emotions) and eudaimonia (e.g., meaning, accomplishment, etc.). As an example, a person who experiences a high sense of meaning (eudaimonia) through a volunteering experience might also be experiencing negative emotions (hedonia) at the time of volunteering at a hospital due to exposure to other people’s ill-being. Assessing this person’s well-being through a framework merely focused on eudaimonia will provide a different assessment of how they are doing compared to an assessment through a framework solely focused on hedonia. Thus, having a picture of both hedonia and eudaimonia simultaneously provides more nuances to our understanding of well-being.

In sum, theoretical perspectives on what entails a “good life” have grown exponentially in the past two decades (Rombaoa & Heshmati, 2023). We believe that no one theory can be deemed as the “superior” framework in explaining psychological well-being. Rather, clarifications need to be made around how specific theoretical approaches to well-being may be differentially advantageous for operationalizing this multiplex construct as it relates to people’s lives. A first step towards this clarification is identifying the utility of these theoretically driven operationalizations of well-being as it applies to people’s real-world experiences. Whereas it is important to understand what makes up well-being, it is equally important to recognize the nature of change in these elements as they unfold in daily lives; while some elements may be more stable in the way they are experienced in people’s lives, some others may be more susceptible to change and be state-like in nature rather than trait-like. Hence, assessments of well-being need to be in-line with the types of questions being asked with regards to the good life: Do we want to understand people’s overall sense of well-being (e.g., a nation’s general quality of life), or do we want to understand the day-to-day experiences of living a high quality life (e.g., fluctuations in experiences of positive emotions in their workplace).

Global assessments of well-being have been a focus of research for decades, and have become more prevalent with the advent of the field of Positive Psychology (Seligman & Csikszentmihalyi, 2000). These assessments have been shown to be useful for both researchers and clinicians in understanding the degree of an individual's overall quality of life and ways to increase people’s general sense of well-being. World leaders and governments have also shown interest in these assessments to monitor their citizen’s quality of life (Diener, et al., 2009; Stiglitz, et al., 2009). Well-being global assessments are typically focused on the totality of one’s positive experiences in life that requires retrospection via self-reports in survey-based measures.

Despite the ease that comes with answering seemingly straightforward questions, the emphasis on global assessments introduces some limitations in measurements of well-being. First, the reliance on retrospective recall in global measures, introduces unreliable and biased responses; for instance, people are more prone to remember negative events and feelings in their lives when in a negative state (Clark & Teasdale, 1982). Second, global assessments prevent the examination of people’s dynamic moment-to-moment or day-to-day changes in feelings and experiences as a sequence of events over the course of time. Finally, the contextual governance of experiences is neglected in global measures, eliminating an in-depth understanding of how well-being varies in the context of others or in interaction with the environment as life is lived in real-time and real-world settings (Shiffman et al., 2008). These limitations have cast some doubt in the validity of global self-report measures (Hudson et al., 2020; Kahneman, 1999; Schwarz & Strack, 1999). Experiences of well-being, however, unfold in a momentary manner in people’s daily lives in their naturalistic context. Thus, examinations of people’s experiences of well-being as they live their lives day-to-day and moment-to-moment in real-world contexts, is imperative to a more precise understanding of well-being.

Capturing well-being in real-time and real-world contexts is an important new direction to further our understanding of PERMA. People’s well-being undergoes dynamic changes governed by their contextual living circumstances and interactions with their immediate and extended environment as it unfolds over a sequence of events over both short and longer time scales (Donaldson et al., 2021). It is these momentary experiences that accumulate to generate global evaluations of one’s well-being (Heshmati et al., 2022). As a result, our understanding of the dynamic changes in well-being and its indicators over time and across contexts are lost when we use merely global assessments of well-being. By using experience sampling methods (Csikszentmihalyi & Larson, 1987) such as Ecological Momentary Assessment (EMA; Stone & Shiffman, 1994) we can move beyond the limitations of global assessments and take a dynamical modeling approach to measure how indicators of well-being change in people’s momentary and daily experiences.

Specifically, we can recognize patterns in people’s moment-to-moment experiences of elements of well-being and examine how those varying patterns of individual differences relate to more long-term and stable reports of well-being. For instance, when using momentary reports of feeling loved – a measure of one of the PERMA elements, positive relationships – over short time-scales, we can examine love in terms of (a) mean levels: a person’s average levels of felt love across time, as well as (b) intra-individual variability (IIV): how much a person’s levels of feeling loved fluctuate around their mean, over day-to-day momentary experiences. With this, we can spot time-structured change patterns that amount to how much a person’s experience of positivity in their day-to-day relationships is related to longer term experiences of well-being (Oravecz et al., 2020).

Thus far, using experience sampling methods, researchers have made great advancements in the study of affect–an important component of subjective well-being. Through investigations of within-person change in positive and negative emotions in shorter time scales and its manifestation as a momentary state in daily life, research has demonstrated links between mood variability and more stable states and traits such as depressive states (McConville & Cooper, 1996), psychological health (Gruber et al., 2013; Goicoechea et al., 2023), physical health (Hardy & Segerstrom, 2017), emotional eating (Heshmati et al., 2023), lifestyle (Rombaoa et al., 2023); and age (Rocke & Brose, 2013; Larson et al., 1980). In fact, in a study conducted by Eid and Diener (1999), the validity of hedonic variability, as measured by positive and negative affect IIV across time, was examined in relation to average hedonic levels and personality traits. Eid and Diener (1999) found that hedonic variability, especially variability in positive emotions, is “sufficiently stable [and distinct] to be considered a psychological trait” (Eid & Diener, 1999; p. 662).

Whereas examining within-person characteristics of momentary changes of affect has provided insight into hedonic well-being, less is known regarding within-person changes of eudaimonic elements of well-being. Moreover, within-person dynamics in both hedonia and eudaimonia, as a multidimensional framework, has not been examined in relation to more stable and global reports of well-being. For this purpose, we developed a momentary measure of PERMA (mPERMA)–as a momentary and multidimensional measure of well-being– by adapting the PERMA-Profiler (Butler & Kern, 2016) items for the Ecological Momentary Assessment paradigm (EMA; Stone & Shiffman, 1994).

The process we adopted in developing a state-measure of PERMA (i.e., mPERMA) was in line with the suggested procedures by Horstmann and Ziegler (2020). Horstmann and Ziegler (2020) suggest that for constructing measures suitable for experience sampling methods (i.e., state measures), researchers need to answer three questions that inform item construction and scale development. First, we needed to consider what is the construct that is being measured. We aimed at measuring both hedonic and eudaimonic elements that make up well-being. The PERMA theory (Seligman, 2011, 2018) incorporates both aspects and has a validated trait-level measurement instrument, the PERMA-Profiler (Butler & Kern, 2016), therefore we chose this tool as a basis for ours.

Second, we considered the intended purpose of the measurement. Our goal was to capture experiences of well-being in people’s daily lives in real-world contexts and in-situ. The PERMA-Profiler validated three items per PERMA element for measuring PERMA as a trait (Butler & Kern, 2016), which can be considered parsimonious given the complexity of PERMA. To make them appropriate for momentary assessments where participants respond to items multiple times per day, we adapted the phrasing for reflecting daily life states of well-being. That is to say that mPERMA adopts the dimensionality of the PERMA-Profiler scale to capture the five PERMA elements, but adapts the trait-level language of the PERMA-Profiler items (e.g., “In general, how often do you feel joyful?”) to momentary based assessments of well-being (e.g., “How joyful do you feel right now?”). Additionally, the measure should include items measuring each element of well-being appropriate to different situations of a person’s life. For instance to capture Engagement, we have situational-specific items such as “When I noticed the text message, I was absorbed in what I was doing” or “When I noticed the text message, I felt excited and interested in the things around me.”

Third, we considered the targeted population. The mPERMA measure is targeted for the general population, specifically adults 18 years and older. There has been a plethora of research studies focusing on different elements of well-being in adults’ lives, both on the trait and state level. Specifically, PERMA-Profiler was originally validated using an adult population (Butler & Kern, 2016) and subsequent research validating this measure with participants 18 years and older (e.g., Bartholomaeus et al., 2020; Donaldson et al., 2021; Umucu et al., 2020) and tested globally (e.g., Giangrasso, 2021; Ryan et al., 2019; Wammerl et al., 2019). However, it is essential to acknowledge that the sample used in the current study for the purpose of validation of the measurement was limited to college students 18–22 years of age. This decision was primarily driven by the time-intensive nature of our measurements and the convenience of using a college student sample which is often more easily accessible for research involving detailed measurement designs. We elaborate on this further in the limitations section of this paper.

In the current study, we first aimed to test the factor structure of mPERMA, a momentary scale developed to measure the PERMA building blocks in real-time and real-world contexts. After validation of the measure, we examined how within-person dynamics of each of the PERMA building blocks—measured via mPERMA—were associated with global measures of SWB, the Flourishing Scale (FS; Diener et al., 2009) and the Emotional Well-Being subscale of the Short Form Health Survey (SF-36; Ware & Sherbourne, 1992). We selected the FS as our global assessment of psychological well-being because through a single well-being score it assesses various self-perceived elements of well-being (e.g., relationships, self-esteem, purpose, etc.) from a global perspective and has been shown to be strongly associated with other global well-being scales (see Diener et al., 2009). Additionally, the FS is proposed to be aligned with advancements in the psychological theories of human flourishing that have emerged in recent years. The authors have incorporated insights from multiple contemporary theories, including those proposed by Ryff(1989), Ryff & Singer (1998), and Ryan and Deci (2000). These theories emphasize the importance of universal human psychological needs, such as competence, relatedness, and self-acceptance, which the FS is proposed to effectively capture (Diener et al., 2009). However, it should be noted that FS was not shown to be associated with autonomy, a component of eudaimonic well-being present in Ryff’s (1989) Psychological Well-Being framework and hence this element is missing from our examination of eudaimonic well-being. In addition to the FS that captures the global eudaimonic aspects of well-being, we used the Emotional Well-Being subscale of the SF-36 to capture the global hedonic dimensions of well-being.

Specifically, we examined mPERMA in terms of mean levels and its dynamics of IIV. Previous research examining the dynamical characteristics of well-being elements demonstrated that IIV in these elements (e.g., positive emotions; Eid & Diener, 1999) are sufficiently distinct to be considered as unique traits. We used this dynamical characteristic of mPERMA (i.e., IIV) to explore whether there are systematic associations between this and more stable well-being measures. Specifically, we tested the association of mean levels and fluctuations (IIV) in each PERMA element in relation to globally assessed levels of hedonic (i.e., emotional well-being) and eudaimonic (i.e., Flourishing) well-being.

Previous reports of within-person dynamic correlates with physical health and mental health indicators have shown that higher variability in affect is negatively associated with health indicators both short-term and long-term, while higher mean levels of positive affect is related to better mental health (e.g., Hardy & Segerstrom, 2017; Ram & Gerstorf, 2009). Moreover, a study examining intra-individual variability in components of subjective well-being (e.g., life satisfaction, positive emotions, negative emotions) in relation to global assessments of psychological states (e.g., social support) found that higher IIV is negatively correlated to social support and positively associated with adverse life events (Gadermann & Zumbo, 2007). A more recent study examining momentary experiences of love in everyday life in relation to more global assessments of psychological well-being (e.g., flourishing, emotional well-being, gratitude) found that higher mean levels of momentary feelings of love were positively associated with psychological well-being (Oravecz, et al., 2020). Thus, we hypothesized that higher mean levels and lower fluctuations (IIV) in the momentary experiences of the Positive emotions element of PERMA would be associated with higher levels of emotional well-being. Similarly, we expected that higher mean levels and lower IIV in the other four elements of the PERMA building blocks (i.e., E, R, M, A) would be associated with higher levels of Flourishing. By articulating building blocks of well-being in terms of dynamic characteristics we can identify various ways through which well-being may be manifested and is experienced in a day-to-day manner that is related to and may lead to higher levels of global well-being.

In this section, we report how we determined our sample size, all data exclusions (if any), and all measures in the study. The data collected in this study and the respective analytic codes are publicly available and can be accessed through the Open Science Framework (OSF; ​​https://​osf.​io/​emyzp/​?​view_​only=​3370dc87c5914b85​870a83ea22e3b53b​).

Participants consisted of a total of 160 undergraduate students at a major public university in the North East, United States. Participants were recruited by a convenience sampling method through the university’s research website. Sixty-eight percent of the participants were female and 32% of the participants were male. Participants’ age ranged from 18 to 22 years old. Seventy-four percent of the participants were White/Caucasian, 6% of them were Black/African-American, 9% of them were Asian or Pacific Islander, and 4% of them were Hispanic/Latino. The remaining 1% identified themselves as other races. This project has been approved by The [blinded] Human Subjects Protection Program and IRB protocol number: STUDY00006362. Informed consent was obtained from all participants before they were involved in the study.

This study is part of a larger study which includes a 56-day intervention study using a signal-contingent EMA design. In the current study we used the EMA data associated with the first 14 days of the larger study, the period before the intervention began. The aim of the study was voiced as measuring psychological well-being in everyday life while also completing some small exercises. On the first day of the study, participants completed an introductory survey on the online Qualtrics survey platform (Qualtrics, Provo, UT) which included the Flourishing scale, Short Form Health Survey, demographics questions, and other psychological scales on a computer in the lab. Starting from the next day, participants received the EMA surveys via text messages to their smartphones six times a day, semi-randomly during the day, for 14 days. The semi-random schedule of the text messages was in a way that surveys were sent to participants’ phones randomly within six designated two-hour blocks across the day with no less than half an hour between each survey. In these surveys, participants gave structured self-reports on their momentary well-being (momentary PERMA). The surveys were scheduled based on the participants’ waking hours, between the time in the morning that they reported waking up until the time they reported they usually went to bed. To reduce the participant workload in the momentary assessments, two items out of the three items for each element were randomly selected and implemented within each momentary assessment and one item was kept missing systematically using the random missingness paradigm. Therefore, in each session participants answered 10 items (2 × 5). As missing data comes from the design, the missingness mechanism is considered completely at random (MCAR, Rubin, 1976). We used maximum likelihood methods to impute the missing data as it produces unbiased estimates of population parameters (Davey, 2009; Enders, 2010; McKnight et al., 2007). Participants were prompted to respond to EMA surveys 6 times a day totaling to 84 survey prompts. On average, participants responded to 75 of the 84 prompts (~ 89% compliance rate). participants were incentivised with payments for their time based on the number of surveys they completed by the end of the study duration.

The sample size was determined based on the larger intervention study with recruiting sufficient samples for each of the three experimental groups. We tried to adhere to general sample size guidelines for conducting factor analysis in the multilevel framework (Comrey & Lee, 1992; Kline, 1994; Maas & Hox, 2005). From the recruited sample of 165, 5 were excluded due to incomplete data (below 80% completion of all surveys) and participant attrition. The analysis was conducted on the final sample of N = 160 with a total of 10,979 observations.

The data analysis for this study involved two steps. First, multilevel factor analysis (MFA) was conducted to validate the five-factor structure of mPERMA. The multilevel approach was chosen due to the nested nature of the EMA measurements that were clustered within-person. Second, a Bayesian regression analysis was conducted to investigate the association between global levels of flourishing and within-person means and the dynamical characteristics of mPERMA elements such as fluctuations of the five mPERMA elements. The factor analysis was conducted in the R statistical software (Version 4.2.2., R Core Team, 2022). Specifically, the lavaan package (Rosseel, 2012) was used for MFA, and the psych (Revelle, 2021), stats (Bolar, 2019), and ggplot (Wickham et. al, 2016) packages were used for descriptive statistics. JASP (JASP Team, 2022) was used to conduct the Bayesian regression analysis.

For each mPERMA element, we calculated a mean value for every individual separately. Mean levels correspond to the mean value of each individual’s scores over measurements across time. Figure 1 illustrates two individuals' mPERMA scores over time. The straight black line indicates the mean level of each example person’s data across time. The plot on the left demonstrates person A with a high mean score in positive emotions across the study time, and the plot on the right shows Person B with a low mean score in their positive emotions.

Besides mean levels, we also calculated people’s fluctuations around the mean (IIV) individually for each person. The IIV values correspond to standard deviations of each person’s scores over measurements across the study duration. Figure 2 shows two example individuals who demonstrate different patterns in IIVs for the Positive emotions element of mPERMA: Person A (left plot) demonstrates high variability and person B (right plot) shows low variability respectively. Note that although mean levels (straight black line) of both people are pretty similar, their experiences of positive emotions fluctuate differently over time.

There were a total of 10,979 observations for 160 participants with an average of 75 observations per participant. Descriptive statistics for each item and its corresponding intra-class correlations are reported in Table 1. As shown in Table 1, the ICC values are between 0.20 and 0.62 indicating a considerable amount of between-level variance among individuals. Typically, the norm for state measures is to show ICCs ranging from 0.20 to 0.50 (Podsakoff et al., 2019; Sherman et al., 2015; Sun & Vazire, 2019). Variations in the ICCs demonstrated in Table 1 indicate that some of the mPERMA elements may have better sensitivity to situational or contextual factors (e.g., Engagement and Positive emotions) compared to others with higher ICCs (e.g., Meaning and Relationship) that may be more attributable to the person. Further descriptive statistics for the IIVs and mean levels of each PERMA element is presented in the supplementary material.

Results from the MFA analysis are summarized in Table 2. As shown in Table 2, the level one CFA model fitted the data properly (RMSEA = 0.017, SRMR = 0.018, CFI = 0.996, TLI = 0.994) indicating that we had a solid within-level factor model, thus we continued to test the between-level variance factor model. The fit statistics for the null model, as expected, showed a poor fit indicating that there was a between-level variance that we need to take into consideration (RMSEA = 0.162, SRMR = 0.695, CFI = 0.548, TLI = 0.526). This encouraged us to fit the independence model to test whether this variance is important enough to define a between-level model. The independence model indicated an acceptable model fit for two of the fit indices (RMSEA and SRMR), and the CFI and TLI fit indices were also close to a good fit (RMSEA = 0.074, SRMR = 0.027, CFI = 0.913, TLI = 0.902). Therefore, for quality assurance, we moved onto the fourth and fifth steps of the MFA modeling to test group level modeling. In the saturated model, we got an acceptable model fit and moved on to test the hypothesized model (RMSEA = 0.024, SRMR = 0.018, CFI = 0.996, TLI = 0.989).

The hypothesized model included a five-factor structure for both within and between models. Of note, we updated the scaling of our latent factors via manifest items that produced interpretable factor loadings. The results showed that the hypothesized model fits the data well (RMSEA = 0.021, SRMR = 0.021, CFI = 0.994, TLI = 0.992). The standardized factor loadings are reported in Table 3. The factor loadings in both within and between levels were significantly larger than zero (p < 0.00). Factor correlations are shown in Table 5. At the within-level, factor correlations range from highest magnitude of 0.498 (Positive emotion with Relationship) to lowest magnitude of 0.129 (Engagement with Meaning). At the between-level, factor correlations range from highest magnitude of − 0.880 (Meaning with Accomplishment) to lowest − 0.302 (Engagement with Meaning).Results indicate that the items “I am feeling positive” (Positive emotion), “I lead a purposeful and meaningful life” (Meaning), and “I am making progress towards accomplishing my goals” (Accomplishment) show standardized factor loadings > 1 (Table 3) and negative between-level residual variances (Table 4). The between-level variance for “I am feeling positive” (p = 0.076), and “I lead a purposeful and meaningful life” (p = 0.670) were not significantly different from 0, however, “I am making progress towards accomplishing my goals” was (p = 0.017). Additional variances that were not significantly different from 0 include “I was absorbed in what I was doing” (Engagement; p = 0.699), “I feel helped and supported by others” (Relationship; p = 0.497, and “What I do in my life is valuable and worthwhile” (Meaning; p = 0.399).

We conducted post hoc analyses to explore the three “Heywood cases” (e.g., estimated negative variances, standardized loadings > 1) and assess additional diagnostics regarding model misspecification (Kolenikov & Bollen, 2012). There were three estimated residual variances that were negative (Table 4): for items P2 (− 1.729, p = 0.076), M1 (− 0.174, p = 0.670) and A1 (− 0.756, p = 0.017). The first two were not significantly different from zero, however, the estimated residual variance for the third item (A1) was significant (Table 5). It is not realistic to have negative variances in the population, thus, there is a need to explore and identify causes. These Heywood cases often serve as error indicators to conduct deeper investigations into the parameter estimation process, data quality, model specification, etc. Common causes include outliers, small samples, or model misspecification (Bentler & Chou, 1987; Chen et al., 2001; Kolenikov & Bollen, 2012). Practical guidelines suggest that some Heywood cases are not problematic after conducting diagnostic assessment (Bentler & Chou, 1987). Our data has a large sample to parameter ratio (greater than rule of thumb guidelines of 5 to 1; Bentler & Chou, 1987). To assess model specification, we conducted post hoc analyses with alternative model structures to assess factor structure at the between and within levels (see Table 6).

We conducted a post hoc multilevel exploratory factor analysis to identify the factor structure at both levels. First, we fit a modified model with the non-significant residual variances in the hypothesized model fixed to 0. Model fit does not appear to be improving with these modifications. We conducted post hoc exploratory analysis to assess the factor structure on the between-level with 5-factor structure at the within-level. Huang (2017) suggests that higher order factor structures may be simpler or different from within-level factor structures, thus, we compared model fit across increasing factor structures (See Table 6). Exploratory factor analysis of the total covariance matrix guided measurement model specifications in the exploratory MFA. Overall, the factor correlations are higher in the between-level than the within-level, however, the 5-factor solution indicates better fit looking at the model fit indices and indicates lowest SRMR at the between-covariance level (SRMR_between = 0.093). Post hoc explorations indicate that a 5-factor structure is preserved in both the within- and between-levels.

We conducted a Bayesian regression analysis with person-specific characteristics (i.e., mean levels and IIV) of each mPERMA element once in relation to Flourishing and a second time in relation to Emotional Well-Being. We report the correlations among mPERMA person-specific characteristics in relation to Flourishing along with Emotional Well-being in Table 1 of the Supplemental Material.

Although there have been steps made towards better understanding and validating building blocks of well-being on a global level, research on daily experiences of well-being as it unfolds in people’s naturalistic environments has been scarce. The current study aimed to bridge this gap by examining the PERMA building blocks of well-being in people’s day-to-day lives through an EMA design. We followed the steps suggested by Horstmann and Ziegler (2020) to create a measure of PERMA suitable for EMA methods. We dub this new iteration of the scale “mPERMA” (momentary PERMA). First, we tested the factor structure of mPERMA through MFA. We then used the daily life data captured via mPERMA to examine the within-person mean-levels and dynamics of change (e.g., fluctuations) in each PERMA element and their relation to people’s globally assessed hedonic (i.e., Emotional Well-Being) and eudaimonic (i.e., Flourishing) well-being.

As hypothesized, the MFA for mPERMA revealed a five-factor structure, at both within- and between-levels, indicating a sound measure of the five PERMA building blocks at a momentary level. Specifically, all three items adapted to EMA assessments measuring each of the five PERMA elements loaded significantly onto the respective PERMA factor. Moreover, mPERMA measures indicated adequate reliability in both within and between levels of analysis. In other words, both aggregated and disaggregated measures have the same factor structure with slightly different factor loadings. This finding implies that well-being can be assessed through various elements including both hedonic and eudaimonic aspects of well-being as a momentary experience in daily life. Additionally, considering that the reliability of the full scale is also relatively high [coefficient omega of 0.97 (between) and 0.78 (within), McDonald, 1999]), researchers and practitioners can use either the subscale scores individually or the full scale score as a whole, having flexibility to choose the scoring approach that best suits their research questions and client goals. This approach paves the way for further applied investigations of well-being as is experienced on a momentary basis and in a contextually and ecologically valid manner across different populations and in different life circumstances.

Moreover, the estimated factor correlations at the between-level demonstrated several interesting pairwise comparisons that shed light on the discriminant validity of the facets. Among the facets demonstrating greater discriminant validity, Engagement consistently emerged as a distinct factor. Specifically, the pairwise correlations between Engagement and Relationship, Engagement and Meaning, and Engagement and Accomplishment indicated relatively lower factor correlations, suggesting a clearer differentiation between these facets. Conversely, facets with lower discriminant validity exhibited higher factor correlations with each other. Notably, the pairwise comparisons of Meaning and Accomplishment, Positive Emotion and Relationship, and Relationship and Meaning demonstrated stronger factor correlations, indicating a greater overlap or shared variance between these facets. In sum, these results suggest that Engagement consistently represents a distinct facet with relatively lower correlations with other facets, supporting its discriminant validity. On the other hand, facets such as Meaning, Accomplishment, Positive Emotion, and Relationship demonstrate somewhat higher correlations, indicating a certain degree of overlap in the variance they capture.

To further our understanding of how different elements of well-being unfold over time in day-to-day life, we then used mPERMA to obtain within-person dynamic characteristics of each of the PERMA elements assessed over 14 days and 75 momentary time-points. Using the within-person mean levels and dynamic characteristics (i.e., intra-individual variability) we were able to examine what dynamics of change may be associated with global well-being for each PERMA well-being element. Using a Bayesian regression analysis, we first examined these associations in relation to people’s global reports of their eudaimonic well-being as measured by the Flourishing scale. Next, we examined all five mPERMA elements with regards to people’s hedonic well-being as measured by the Emotional Well-Being scale.

As hypothesized, we found that eudaimonic elements of PERMA (e.g., Relationship, Meaning, and Accomplishment), measured on a momentary level, in fact were meaningfully associated with Flourishing—a measure of eudaimonic well-being. Moreover, with regards to hedonic well-being (Emotional Well-Being), the momentary Positive emotion element of PERMA indicated the only meaningful association. This finding provides further evidence that while various elements of well-being can be assessed individually on a momentary basis in people's daily lives and natural environments, they also are meaningfully associated with the corresponding aspect of people’s global reports of well-being. Most importantly, for each of these elements, the dynamic characteristic of change that is related to global reports of well-being is different.

Specifically, we found that in relation to global reports of eudaimonic well-being (i.e., Flourishing), people’s within-person mean levels of Relationship and Meaning are meaningfully and positively associated with Flourishing. That is, regardless of within-person fluctuations, the higher a person’s overall average of moment-to-moment Relationship and Meaning is, the higher their global assessments of Flourishing would be. These results suggest that in order for individuals to experience overall high levels of eudaimonic well-being, it is crucial for them to monitor their daily and momentary experiences of positive relationships (i.e., feeling loved and supported; Heshmati et al., 2019; Heshmati & Donaldson, 2020) and meaning in life (i.e., sense of purpose and value; Cohen & Cairns, 2012; Steger & Kashdan, 2007). It is important to note that maintaining consistently high levels of meaning and positive relationships is key, regardless of the fluctuations in these experiences throughout one’s daily life. This finding is in line with previous research examining love in daily life, with similar results: within-person averages of moments of feeling loved in daily life was significantly associated with psychological well-being (Oravecz et al., 2020); this is while fluctuations in those experiences were not significantly associated with well-being.

On the other hand, our results indicate that it’s the within-person fluctuations in experiences of Accomplishment, rather than mean levels, that is meaningfully associated with Flourishing; more importantly, this association is negative. In other words, higher levels of fluctuation in momentary experiences of accomplishment are related to lower levels of eudaimonic well-being. When contextualizing this finding in people’s daily lives, we can infer that having a steady flow of accomplishment and feelings of success towards one's goals is more relevant for people’s overall sense of eudaimonic well-being than having too many ebbs and flows in feeling accomplished and successful in handling responsibilities and achieving goals. It may be that the high points experienced in this state does not offset the low points and therefore a gradual steady pace is more helpful, regardless of how on average they feel momentarily accomplished. We, however, do caution the reader that these implications are based on correlational associations and examination of causal relationships is needed in future studies.

Additionally, we found that Positive emotion – a hedonic element of PERMA– was the only element that highly mapped onto people’s global reports of hedonic well-being as measured by the Emotional Well-Being scale. Specifically, within-person mean levels of Positive emotion were meaningfully associated with Emotional Well-Being, such that the higher on average people’s momentary experiences of positive emotions, the higher their reports of global emotional well-being would be; this is despite how much fluctuations was existent in their momentary experiences of Positive emotions. Contrary to our expectations, variability in positive emotions in daily life was not meaningfully related to global reports of Emotional Well-Being. Prior research examining intra-individual variability in positive and negative affect in relation to depressive symptoms (McConville & Cooper, 1996) and health (Gruber et al., 2013) have found significant relationships. However, this association might not entirely be true for positive emotions and emotional well-being. Most recently, Oravecz et al., (2020) examined intra-individual characteristics of feeling loved in a momentary manner as it related to emotional well-being. Love was treated as a micromoment of positivity resonance in which people would experience moments of positive emotions with another person. Similar to our findings regarding Positive emotions, Oravecz et al. (2020) also found that out of all the different within-person characteristics of momentary love (e.g., mean levels, IIV, inertia), only within-person mean levels of felt love was meaningfully associated with emotional well-being. While our analytic approach does not allow for causal inference, this finding may suggest that people’s global evaluations of their emotional well-being is more sensitive to the average level of their daily life experiences of those emotions compared to the ebbs and flows. In other words, as long as people experience overall higher levels of positive emotions in their daily life, it doesn't matter that at times they might be low on those emotions. In fact, this might be an indication that when positive emotions are low for people, they are experiencing some level of negative emotions (Russel, 2003) which on its own is an important indicator of well-being. Unfortunately, in the current study, we did not include a measure of negative emotions and therefore cannot confirm this inferred conclusion. We suggest that future studies explore the role of within-person experiences of negative emotions along with positive emotions in relation to global Emotional Well-Being.

Notably, we found no evidence for the relationship between within-person patterns of Engagement in daily life and Flourishing; nor with Emotional Well-Being. Based on previous studies that have also examined Engagement on a momentary basis and its relation to other elements of PERMA (Heshmati et al., 2020), momentary engagement in daily life is not a central element in the network of PERMA. This is particularly true for the specific age group that comprised the participants of this study —early adults in the US. Rather, due to the central role of Positive emotions and Relationship in early adults’ network of PERMA well-being (Heshmati et al., 2020), this finding may suggest that Engagement is indirectly related to individuals’ global well-being. In other words, it may be that increases in Engagement are related to increases in positive emotions which in turn relates to Emotional Well-Being. Similarly, momentary experiences of Engagement may be associated with momentary experiences of positive relationships which in turn relates to Flourishing. Indirect associations among these momentary elements of PERMA warrant future research.

It should be noted that some items in the mPERMA item pool do not explicitly include the “present moment” wording. However, in our opinion, this should not be a reason as to whether participants rated these items based on their momentary experiences on the states measured. We believe that several factors contribute to capturing such momentary evaluations on all items of mPERMA. Firstly, the initial instructions for the participants emphasize the importance of responding based on momentary experiences. Secondly, the removal of general wording pertaining to overall life evaluations such as items that begin with “in general…” helps focus participants on their present experiences. Lastly, the nature of the assessments and the frequency of prompts throughout the day ensures that participants provide responses in various temporal contexts. While we understand that specifically for eudaimonic items (e.g., purpose) demonstrate lower iSDs and higher ICCs, we interpret these findings as indicators of the stability and lesser variability of these specific well-being elements. In fact, our expectations regarding the stability of certain PERMA elements due to their inherent nature were supported by these results.

While the present study has afforded opportunities for more contextualized and ecologically valid examinations of well-being, we also want to acknowledge that the current study has its limitations. This study is limited in that we used a college sample of early adults to conduct this EMA study, as conducting these types of intensive data collections (6 times a day for 14 days) through technology is more feasible with a population that is comfortable with having smartphones with them at all times. Moreover, the nature of our analysis was a correlational approach and is limited in making causal inferences. Future studies are warranted to explore the causal associations among mPERMA elements in daily life and their effects on global assessments of well-being. Furthermore, our study was only limited to the investigation of direct relationships. We also encourage future research to explore indirect relationships among these constructs, in addition to the direct associations. For instance, the role of momentary experiences of engagement and flow in daily life may be further clarified with investigations of indirect mechanisms through which flow state may be related to overall reports of well-being. Additionally, this study encountered potential Heywood cases (negative residual variances) in the model estimation that can affect the analysis. Future studies should conduct rigorous assessments to diagnose the source of these Heywood cases and impact on analysis (Bentler & Chou, 1987; Kolenikov & Bollen, 2012). This study was a first to examine the dynamics of daily PERMA through intensive data and we hope that it provides the foundation for future research to expand the examination of well-being as it is experienced in daily life in other age groups and to more representative samples.

The scientific study of well-being has taken big steps in the past decade with more and more research focused on cross-validation of prevalent well-being measures (Donaldson et al., 2020; Goodman et al., 2018; Joshanloo, 2019). Most of these scientific endeavors, however, have been limited to global assessments of well-being in an attempt to understand people’s feelings and evaluations of their lives. These global assessments require retrospection in self-report measures, introducing unreliable and biased responses. Additionally, global well-being measures lack ecological validity, neglecting life as it is lived in real-time and real-world settings (Shiffman et al., 2008). All in all, while advancements in the understanding of how people think and feel about their lives have been made, less research has focused on how well-being is experienced in people’s lives (Dolan et al., 2017). To bridge this gap, this study took a step toward expanding the scientific understanding of well-being by examining how experiences of a good life unfold in people’s day-to-day lives and how these experiences relate to global evaluations of well-being. We used Ecological Momentary Assessment, as a study design that surpasses recall biases while also providing ecologically valid information regarding people’s experiences of well-being. Through this design, we were able to provide a snapshot of the within-person mean levels and dynamics of change ( IIV) in various elements of well-being (i.e., Positive Emotions, Engagement, Relationships, Meaning, Accomplishment) in people’s own daily contextual environments.

Exploring the dynamic of change in momentary experiences of PERMA well-being has implications for applied research in two ways: (1) When people think about well-being, they are most likely to take a unidimensional approach to experiencing and enhancing well-being. However, our research demonstrates that different elements of well-being (P-E-R-M-A) experienced in daily life are related to different aspects of people’s global evaluations of well-being, which underscores the importance of richness in how well-being is experienced in daily life. (2) When intervention scientists or practitioners prescribe practices for increasing global levels of life satisfaction and well-being in people’s daily lives, they usually focus on the types of activities that merely enhance average levels of various elements of well-being from one time to the next. However, this study highlights the importance of dynamics of change in each element of well-being as experienced in daily life and how those dynamics might be related to people’s overall sense of well-being.

For instance, a most recent meta-analysis on the effectiveness of interventions in enhancing eudaimonic well-being (van Dierendonck & Lam, 2022) found that targeted psychological practices and programs that were theoretically linked to Ryff’s Psychological Well-Being (Ryff, 1989) dimensions can improve people’s eudaimonic well-being. This conclusion was however made based on pre- and post- change scores from intervention studies, with no information on what the dynamics of change looked like for each dimension of psychological well-being. The current study suggests that for various elements of well-being it may be more beneficial if interventions target the appropriate dynamic of change for that element as it relates to global well-being. For example, to increase positive emotions, meaning, and positive relationships in people, interventions should target increasing within-person average levels of these momentary experiences. However, for the accomplishment element, interventions should focus on decreases in extreme fluctuations in these experiences, perhaps by keeping people’s sense of achievement and success at a steady level. Thus, as a next step to this recent meta-analysis on effectiveness of psychological well-being interventions (van Dierendonck & Lam, 2022), future research can examine daily dynamic change as a result of well-being interventions to ultimately inform psychological practices that are most effective in enhancing people’s daily experiences of a good life.

To our knowledge, this research is the first to study the factor structure of momentary assessments of the PERMA building blocks to be used in EMA studies of well-being, while highlighting the importance of the dynamics of daily experiences of these building blocks in relation to global hedonic and eudaimonic assessments. This study also continues the conversation around the validity of PERMA in relation to other well-being measures, as presented in the Journal of Positive Psychology debate between Goodman et al. (2018) and Seligman (2018) followed by Donaldson et al. (2020).