Abstract
Accumulating evidence for the unique social, behavioral, and physical health benefits of positive emotion and related well-being constructs has led to the development and testing of positive psychological interventions (PPIs) to increase emotional well-being and enhance health promotion and disease prevention. PPIs are specifically aimed at improving emotional well-being and consist of practices such as gratitude, savoring, and acts of kindness. The purpose of this narrative review was to examine the literature on PPIs with a particular focus on positive emotion outcomes. We evaluated the evidence on the effects of PPIs on positive emotion specifically, and discussed the range of evidence regarding the relative responsiveness of emotion measures to PPIs in order to gain a better understanding of the specific emotional pathways through which PPIs influence psychological and physical well-being. We conclude with recommendations for best evaluating effects of PPIs on positive emotion outcomes.
With some early exceptions (Fordyce, 1977, 1983; Lichter et al., 1980), historically, the emphasis in psychology has been on reducing depression, anxiety, or other negative affective constructs. Starting in the late 1990s, interest in positive emotion burgeoned with the rise of positive psychology (Seligman & Csikszentmihalyi, 2000), and with theoretical and empirical work by Folkman (1997) and Fredrickson (1998) bringing more attention to positive emotions as uniquely beneficial for health and well-being, independent of the effect of negative emotions. Accumulating evidence for social, behavioral, and physical health benefits of positive emotion and related constructs (e.g., optimism, life satisfaction; Lyubomirsky et al., 2005; Pressman & Cohen, 2005) has led to the development and testing of positive psychology interventions (PPIs). PPIs are made up of practices such as gratitude, savoring, and acts of kindness and are specifically aimed at improving well-being (Schueller et al., 2014). Some PPIs focus on a single skill and others are multicomponent interventions that teach a collection of positive skills. Various delivery modes have been tested including in person delivered to individuals (e.g., Moskowitz et al., 2017) or groups (e.g., Chaves et al., 2017), and self-guided via Internet (e.g., Addington et al., 2019; Peters et al., 2017) or with telephone support (e.g., Hausmann et al., 2018; Huffman et al., 2016).
Across several meta-analyses (Bolier et al., 2013; Chakhssi et al., 2018; Curry et al., 2018; Davis et al., 2016; Dickens, 2017; Hendriks et al., 2019; Kirby et al., 2017; Sin & Lyubomirsky, 2009; Zeng et al., 2015), PPIs have small to medium effects on well-being. However, the broad category of well-being is inconsistently operationalized across studies, making it difficult to determine whether PPIs have stronger effects on particular aspects of well-being compared to others. For example, decreases in depression are frequently used as one indicator of well-being (Bolier et al., 2013; Hendriks et al., 2019; Sin & Lyubomirsky, 2009), consistent with the field’s historical focus on negative emotions. Diverse positive constructs are usually grouped together, combining positive emotion with outcomes such as meaning and purpose, life satisfaction, and optimism (Chakhssi et al., 2018; Sin & Lyubomirsky, 2009), or with other cognitive and affective appraisals of one’s life as a whole (Bolier et al., 2013; Hendriks et al., 2019).
What cannot be gleaned from most of these prior meta-analyses is the extent to which PPIs influence positive emotion specifically, apart from other positive constructs. We define positive emotions as subjective positively valenced feelings that range from happy, calm, and satisfied to excited and thrilled, and differentiate them from other positively valenced constructs like optimism, life satisfaction, and meaning. These effects are critical to distinguish, given that positive emotion is one of the primary pathways through which PPIs are thought to benefit health and well-being (Lyubomirsky & Layous, 2013; Schueller et al., 2014).
The positive pathways to health theoretical model (Moskowitz, Addington, & Cheung, 2019) posits that engaging in the positive activities in PPIs increases the frequency of positive emotion which, in turn, has a range of proximal effects such as providing a timeout from stress (Lazarus et al., 1980); prompting more adaptive coping strategies (Folkman, 1997); broadened attention and cognition and increased behavioral action tendencies (Fredrickson, 1998); reduced emotional reactivity to daily stress and strengthened social relationships, which all lead to reduced stress. In turn, this reduction in stress predicts better physiological functioning (e.g., quicker autonomic recovery after a stressful event; Fredrickson, 1998; Pressman & Cohen, 2005; Pressman et al., 2019) and greater engagement in preventive health behaviors (Bassett, Schuette, et al., 2019; Hoogwegt et al., 2013), which ultimately lead to improved physical and psychological well-being such as less depression and anxiety, more life satisfaction, and increased meaning and purpose. The effects flowing from increased positive emotion—proximal effects, reduced stress, improved physiological function, and improved health behaviors—are hypothesized mediators of the effects of PPIs on physical health. Individual characteristics such as type of stress, baseline levels of depression and well-being, sociodemographic characteristics, and dispositional or personality factors constitute one class of potential moderators. Other potential moderators include dosage and frequency of activities (Lyubomirsky & Layous, 2013), the particular positive activity and match to individual (Schueller, 2012), and delivery mode.
The purpose of the present article is to review the literature on PPIs, with a particular focus on positive emotion outcomes. We seek to evaluate the evidence on the effects of PPIs on positive emotion specifically, and to gauge the range of evidence regarding the relative responsiveness of emotion measures to PPIs in order to gain a better understanding of the specific emotional pathways through which PPIs influence psychological and physical well-being. Although our focus is on positive emotion, we include effects on negative emotion to explore the extent to which PPIs, as theorized, have larger effects on positive emotion than on negative emotion. A number of variables will likely influence whether a PPI affects emotion, including sample demographics and other characteristics, type of control condition, and emotion assessment methods and operationalization. We discuss each of these in more detail in what follows.
Methods
Review Approach
To select studies for this review, a medical librarian searched Ovid MEDLINE, Embase, PsycInfo, and Scopus in November 2019. Keywords and subject headings (when available) were used to locate English language publications on positive psychology interventions and positive emotion. The full list of search terms is in the supplemental material. Positive emotion terms included both specific positive emotions as well as scales or measures used to assess those emotions. After deduplication in EndNote, a total of 737 records remained for review. In addition, we reviewed studies included in previous meta-analyses (Bolier et al., 2013; Chakhssi et al., 2018; Curry et al., 2018; Sin & Lyubomirsky, 2009) and narrative reviews of PPIs (Hernandez et al., 2018; Pressman et al., 2019). The authors reviewed the titles and abstracts, selected 55 for full text review, and retained 31 studies for inclusion to represent a range of samples (e.g., clinical, student, general population), self-reported positive emotion outcome measures, PPI activities (e.g., multicomponent, single component), and delivery modes (e.g., in person [individual and group], telephone, online).
Emotion outcomes
We focus and organize our review around the emotion measures that appeared most frequently in our search, grouping them in three categories: (a) dimensional measures (e.g., PANAS; Watson et al., 1988); (b) discrete emotion measures (e.g., Differential Emotions Scale; Izard, 1977); and (c) measures that are “blended” in that they incorporate a broader set of affective constructs like meaning and purpose in addition to emotion items (e.g., Subjective Happiness Inventory; Seligman & Csikszentmihalyi, 2000).
According to dimensional models (Larsen & Diener, 1992; Russell, 1980), emotion can be classified along two orthogonal axes: valence and activation. The valence dimension captures the continuum of how pleasant versus unpleasant an emotion is, from negative to positive. The activation dimension captures the continuum of how arousing or energizing an emotion is, from low activation (e.g., drowsy, lethargic) to high activation (e.g., alert, frenetic). These two dimensions can be conceptualized as a circumplex (Larsen & Diener, 1992) which captures a majority of the variance in subjective emotional experience. Watson and Tellegen (1985) proposed that the best way to view the two dimensions of valence and activation is to rotate the dominant poles so they range from high activation pleasant to low activation unpleasant. According to this approach, each emotion can be understood as a combination of these two dimensions, resulting in the following four quadrants of emotion: high-activation positive emotion (e.g., excited, elated), high-activation negative emotion (e.g., nervous, angry), low-activation positive emotion (e.g., calm, content), and low-activation negative emotion (e.g., sad, bored).
In contrast, discrete measures of emotion rest on the assumption that there is a fundamental set of basic emotions that can be differentiated based on their physiological, behavioral, and subjective experience components (Izard & Buechler, 1980). Discrete emotion measures tap a set of theorized basic states thought to be fundamental to the subjective experience. These basic emotions are thought to be consistent across cultures (Ekman, 1992) and include emotions such as interest, joy, surprise, sadness, anger, disgust, contempt, fear, shame, and guilt (Izard, 1977).
At the broader level of well-being, positive psychology research has distinguished between hedonic and eudaimonic outcomes. Positive emotion falls in the hedonic category, whereas eudaimonic well-being includes meaning and purpose, a sense of mastery, and fulfillment (Ryan & Deci, 2001; Ryff, 1989). A number of outcome measures in PPIs combine both hedonic (high positive emotion, low negative emotion, and life satisfaction) and eudaimonic (meaning and purpose) constructs, and as these are frequently the primary outcome in PPI studies, we include a brief review of these combined measures in addition to our focus on measures that tap only emotions. Examination of the effects of PPIs on these blended measures allows us to differentiate the effects of PPIs on positive emotion alone versus the effects of PPIs on hedonic and eudaimonic well-being more generally.
In addition to information on the emotion measures, we also include information about the time frame over which participants are asked to retrospectively report their emotional experience. These range from current state to past day, past week, and past month. Finally, our review contains information on the operationalization of emotion scores, as some studies use cut point or the balance of positive to negative affect instead of assessing aggregate (sum or average) positive and negative emotion scores separately as dependent variables.
Other Study Characteristics
In addition to the emotion outcome, we also collected information about characteristics of each study that likely play a role in efficacy of the PPIs. These variables include sample characteristics, whether participants were randomly assigned to the PPI, as well as size and the strength of the control condition. Strength of the control condition is a particularly important consideration in interpretation of intervention effects (Freedland et al., 2019). No treatment or waitlist control conditions are used to determine whether an intervention works at all—does engaging in the PPI improve emotion outcomes better than not engaging in any intervention? “Treatment as usual” determines whether an intervention works compared to a clinically relevant comparison—does the PPI have a bigger effect on emotion change when compared to a standard treatment (e.g., antidepressants)? The most rigorous controls for PPIs are attention and time matched, which address questions of mechanism—how or why an intervention works—and account for nonspecific effects such as facilitator contact and social support. For example, comparison to other activities that have the same amount of contact with study staff but do not include the hypothesized active ingredients (positive activities) of the program would be included in this most stringent group (Freedland et al., 2019).
Results
Dimensional Measures of Emotion
The PANAS (Watson et al., 1988) captures high-activation positive and negative emotions and is one of the most frequently used emotion measures in both observational and intervention studies. It consists of 10 high-activation positively valenced items (interested, excited, strong, enthusiastic, proud, alert, inspired, determined, attentive, and active) and 10 high-activation negatively valenced items (distressed, upset, guilty, scared, hostile, irritable, ashamed, nervous, jittery, afraid). Most PPI studies using the PANAS operationalize positive affect (PA) as the sum or average of the positive emotion items and the sum or average of negative emotion items, and examine positive and negative affect separately.
Details of 14 PPIs with PANAS as the emotion outcome are shown in Table 1. Overall, it appears that PPIs are associated with increased PANAS PA (e.g., Huffman et al., 2016; Shao et al., 2016) if the intervention is compared to minimal control—a waitlist or treatment-as-usual condition—but not if the control is more active or attention-matched (e.g., Hausmann et al., 2017; Mohammadi et al., 2018; Müller et al., 2016). For instance, Chaves et al. (2017) used a very active time- and attention-matched condition—cognitive behavioral therapy (CBT)—and found that whereas both the intervention and the active CBT control demonstrated significant increases on PANAS positive emotion, effects did not differ between PPI and the active CBT control condition. This suggests that for improving PA as measured with the PANAS, PPIs are superior to not engaging in any intervention, but may not be more effective when compared to more active control activities. There is a similar pattern for effects on PANAS negative affect such that in comparison to waitlist, treatment-as-usual, or assessment-only controls, PPIs have a significant impact on negative affect as well, but again, not if the control condition is more active or attention-matched.
Selected Positive Psychological Intervention Studies with Self-Reported Positive Emotion Outcome.
CBT = cognitive behavioral therapy; FU = follow-up; HIV = human immunodeficiency virus; mDES = modified Differential Emotions Scale; NR = not reported; PA = positive affect; PANAS = Positive and Negative Affect Schedule; PPI = positive psychological intervention; NA = negative affect; SPANE = Scale of Positive and Negative Experience; TAU = treatment as usual; VA = Veteran’s Administration
Rather than looking at a sum or average score on positive and negative affect separately, some researchers calculate “affect balance” based on the ratio of positive to negative affect on the PANAS (Drozd et al., 2014; Hausmann et al., 2018). Similar to the results reported before, effect of the PPI appears to depend on the strength of the control condition. In one study comparing a multicomponent self-guided online-delivered PPI to a waitlist control, intervention participants showed significant increases in PANAS positive to negative affect balance (Drozd et al., 2014). In contrast, in another study that relied on a more stringent attention-matched control condition, there were no between-group differences in PANAS affect balance of a phone-delivered PPI compared to a neutral activities control (Hausmann et al., 2018).
In summary, a number of PPI studies report significant improvements on PANAS positive affect within the intervention group (Celano et al., 2018; Chaves et al., 2017; Müller et al., 2016), but few show significantly greater improvements compared to attention-matched controls. Effects are generally the same for negative emotion, that is, PPIs reduce PANAS negative within the intervention, but the effects do not differ significantly from more active controls (e.g., Hausmann et al., 2018). The studies that have the PANAS as a primary emotion outcome differ on other important factors such as delivery mode (e.g., self-guided with telephone support, in-person group delivery, self-guided workbook), gender make-up of the sample, and normative trajectories of change in affect in the sample (e.g., women with cervical cancer, community participants, acute coronary syndrome, chronic pain), but it is not yet clear the extent to which any of these variables moderate the effect of PPIs on the PANAS.
Discrete emotion measures
Examples of discrete emotion measures used as outcomes in PPIs are the Differential Emotions Scale (Izard, 1977), the Affect-Adjective Scale (Diener & Emmons, 1984), the Scale of Positive and Negative Emotion (Diener et al., 2010), and the Brief Mood Introspection Scale (Mayer & Gaschke, 1988), which, in contrast to the PANAS, tap both high- and low-activation positive and negative affect (e.g., happy, pleased, joyful, contented, calm, worried, angry, frustrated, sad).
Details of 12 PPI studies that use a discrete emotion outcome measure are shown in Table 1. Note that three of the studies are listed twice in this section (Basset, Cohn et al., 2019; Cohn et al., 2014; Moskowitz et al., 2017) because they included multiple retrospective time frames for reporting emotion—assessing both past day emotion and past week emotion.
Similar to the PANAS findings, studies that have tested the effects of PPIs on discrete emotion measures have shown effects on positive emotion when compared with waitlist or treatment-as-usual controls (Fredrickson et al., 2008; Moskowitz, Cheung et al., 2019; Nelson et al., 2016; Schotanus-Dijkstra et al., 2019), but not when compared with more active conditions (Cheung et al., 2016; Hwang et al., 2017; Moskowitz et al., 2017) that control for time spent on activities and attention from facilitators. Interestingly, unlike the PANAS findings, the effects of PPIs seem to be limited primarily to increasing positive emotion, and there do not seem to be comparable decreases in negative emotion as assessed with discrete emotion measures.
In addition, interesting differences emerge when comparing the time frame of assessment using discrete emotion measures. There is some suggestion that intervention effects are more likely on discrete measures of positive emotion when using shorter time frames (e.g., past day or current moment; Fredrickson et al., 2008; Peters et al., 2017), but less so for longer time frames (e.g., past month; Hwang et al., 2017). For example, one of first studies to test effects of a PPI on a discrete emotion measure was done by Fredrickson and colleagues in a test of loving kindness meditation (Fredrickson et al., 2008). The modified Differential Emotions Scale (mDES), which augments the original DES with additional positive items (e.g., amusement, gratitude, love, pride; Fredrickson, 2013), was used to assess emotion daily for the 9 weeks of the program, and responses were aggregated to form weekly averages of positive and negative emotion. Participants in the loving kindness condition had significant increases on positive emotion with the past day retrospective time frame compared to the waitlist control. There were no significant effects of the loving kindness intervention on negative emotion.
In a study that assessed both past day and past week on the mDES, 159 people newly diagnosed with HIV were randomized to a multicomponent PPI or to an attention-matched interview and questionnaire condition (Moskowitz et al., 2017). Although there was no difference between intervention and control on mDES positive (or negative) emotion assessed over the past week, participants in the intervention showed higher levels of positive emotion over the past day assessed using the Day Reconstruction Method (Kahneman et al., 2004). There were no significant effects on either past week or past day negative emotion.
Summary for discrete emotion measures
As in the studies that have PANAS as the emotion outcome, the PPIs that report effects on discrete emotion measures differ on a number of factors including sample, delivery mode, and strength of control condition, so it is premature to draw definitive conclusions. However, there is some indication that, in adequately powered studies, PPIs are likely to significantly influence positive emotion, particularly when compared to minimal control conditions (e.g., Moskowitz, Cheung et al., 2019; Schotanus-Dijkstra et al., 2019). In contrast to findings with the PANAS, few studies that use discrete emotion outcomes find effects of PPIs on negative emotion, supporting the notion that PPIs may have specific, targeted effects on positive but not negative emotion, as assessed by discrete emotion measures.
Measures that combine hedonic and eudaimonic well-being
Table 1 includes details of seven studies that use blended hedonic–eudaimonic outcome measures.
The Steen Happiness Index
The Steen Happiness Index (Seligman et al., 2005), later updated with additional items and renamed the Authentic Happiness Index, was explicitly designed to be sensitive to influence of PPIs and assesses pleasure, engagement, and meaning. Participants are asked to select the statement from each group that best describes the way they have been feeling for the past week, including today. Emotion items include bipolar emotion terms such as sorrow/joy, boredom/fascination, shame/pride, and bad/unbelievably great mood, but the scale also includes additional constructs such as social engagement as well as meaning and purpose. Studies that use the Authentic Happiness Index as an outcome usually find significant effects of the PPI. Mongrain et al. (2011) randomized 719 adults to compassionate activities for 1 week or to an early childhood memory control. Participants in the compassion condition experienced greater increases on the Steen Happiness Index compared to the control over the 6-month follow-up period. Similarly, Proyer et al. (2016) randomized 113 adults to write about nine beautiful things (three in human behavior, three in nature or the environment, and three beauty in general) every evening for a week or to a control condition in which they wrote about early memories. Assessments showed significant increases on the Authentic Happiness Inventory in the intervention condition at all follow-up points. Neither study assessed negative emotion specifically but Proyer et al. (2016) reported that post-intervention participants had decreased depression at post and 1 week follow-up.
The Oxford Happiness Inventory
The Oxford Happiness Inventory (Argyle et al., 1989) is a dispositional measure that also mixes a number of constructs including life satisfaction, optimism, and sense of control, as well as some items that particularly reflect hedonic (e.g., “I am very happy,” “I often experience joy and elation”) and eudaimonic affect (e.g., “I don’t have a particular sense of meaning and purpose in my life [reverse coded]” “I feel that life is very rewarding”). In a randomized trial in patients who had recent heart surgery, Nikrahan et al. (2016) tested three versions of a group PPI compared to a waitlist control condition. When the three intervention arms were combined for analysis, there were no differences on the Oxford Happiness Inventory from baseline to 7 weeks. However, by the 15-week assessment, the intervention participants reported increases compared to a waitlist control. Mohammadi et al. (2018) tested a group PPI in 62 cardiac patients. The control condition was an attention-matched cardiac education class. Results showed significant between-group differences in improvements from pre to postintervention on the Oxford Happiness Inventory. In contrast, the authors found no group differences on the PANAS, suggesting the possibility that effects on the Oxford Happiness Inventory were driven by the eudaimonic content instead of the hedonic (high-activation positive emotion) items captured by the PANAS.
Hedonic–eudaimonic measures summary
Measures such as the Oxford Happiness Inventory or the Steen Hapiness Index combine hedonic and eudaimonic components so it is not possible to distinguish PPI effects on emotion from effects on other facets of well-being. In addition, this group of measures also includes some that are more dispositionally focused (e.g., Oxford Happiness Inventory), that is, they ask participants how they are generally, without a specific time frame referenced. In theory, we would expect less change in dispositional well-being and for eudaimonic well-being to change more slowly in response to a relatively brief PPI. Explicit exploration of effects of PPIs on hedonic compared to eudaimonic measures is an important area for future research focus.
Discussion
Positive emotion is one of the key mechanisms through which positive psychology interventions (PPIs) are hypothesized to influence psychological and physical health (Moskowitz, Addington, and Cheung, 2019; Schueller et al., 2014). Although a number of meta-analyses have demonstrated that PPIs have broadly beneficial effects, most do not differentiate positive emotion specifically or tease apart effects on positive emotion from those on other theorized pathways (Bolier et al., 2013; Chakhssi et al., 2018; Curry et al., 2018; Sin & Lyubomirsky, 2009). In one exception, Curry et al. (2018) conducted a meta-analysis of acts of kindness interventions and, in a subanalysis of individual measures, found that across five studies that used the PANAS, the effect sizes varied widely from −.46 to .70. This wide variability highlights the fact that even if studies deliver the same type of intervention and use the same outcome measure, there are likely many other factors that drive the efficacy of the intervention.
Our goal in this review was to provide an overview of types of emotion outcomes used in tests of PPIs, evaluate the evidence on the effects of PPIs on positive emotion specifically, and examine whether there is sufficient evidence to determine whether different measures were more or less responsive to PPIs. Based on this review, we can draw several conclusions regarding emotion outcomes in PPIs. First, PPIs often have a significant influence on positive emotion, especially when looking at change within the intervention group or when evaluating the intervention versus a minimal control condition such as a treatment-as-usual or waitlist. One of the most frequently used measures to evaluate the efficacy of PPIs is the PANAS, which assesses higher activation emotions. Specific PPIs such as loving kindness, for example, might be expected to impact lower activation positive emotions in particular (“peaceful positive emotions”; Zeng et al., 2015), and exclusive reliance on the PANAS may be missing important effects on lower activation emotions that are not tapped by the measure. On the other hand, some measures may be overly inclusive in that they combine hedonic (emotion) and eudaimonic (meaning and purpose) items, obscuring effects on either type of well-being individually. For example, Mohammadi et al. (2018) tested effects of a PPI on the Oxford Happiness Inventory and the PANAS. Whereas they found significant improvements on a measure that combines emotion and eudaimonic items—the Oxford Happiness Inventory—there were no group differences on the PANAS, perhaps suggesting that eudaimonic items were driving the effects of the intervention. It is important for researchers to select measures that can precisely assess the outcome that is targeted by the PPI, to add to our understanding of whether interventions can improve these different aspects of psychological well-being. Although there are a number of interventions that specifically target meaning and purpose (Breitbart et al., 2015; Westerhof et al., 2010), few have measured both positive emotion and meaning and purpose separately to begin to parse out any differential effects on emotion.
Another source of variability in PPI outcomes and research that utilizes self-reported emotion measures more generally is the time frame for retrospective reporting. The emotion assessments reviewed here vary on a spectrum from measuring momentary emotion (e.g., “How happy are you feeling right now”) or past day emotion (e.g., “How happy did you feel in the past day”), to assessing emotions experienced over longer durations, such as past week emotion (e.g., “How happy did you feel over the past week”). Furthermore, emotion measures vary in terms of whether they capture the individual’s state levels of emotion versus more trait/dispositional tendencies to experience emotion (e.g., “I consider myself a happy person”). Ecological momentary assessment (EMA) or daily diary measures of emotion allow for more fine-grained examination of emotion processes such as intraindividual variability (Ong & Ram, 2017) or responsivity to daily stress (Mroczek et al., 2015) that may better capture the emotion changes in response to PPIs. For example, using a daily diary assessment, Mroczek et al. (2015) demonstrated that lower levels of positive emotional reactivity in response to a daily stressful event are predictive of a lower risk of mortality. These findings suggest that reactivity or other operationalizations of daily emotional experience calculated from daily diaries or ecological momentary assessment (Ong & Ram, 2017) may be an important focus of measurement of PPI effects for future studies, particularly if physical health is one of the more distal outcomes of interest.
This is not to say that broader time frames for self-reported emotion are not useful. There is evidence that shorter retrospective time frames for self-reported emotion may rely more on emotion experience, whereas for longer recall periods participants rely more on their beliefs about what they likely felt based on context rather than the actual emotion experienced (Robinson & Clore, 2002). This review suggests that there is evidence for PPIs’ influence on shorter (e.g., Fredrickson et al., 2003; Moskowitz et al., 2017; Peters et al., 2017) as well as longer time frames (e.g., past week; e.g., Auyeung & Mo, 2019; Huffman et al., 2016; Moskowitz, Cheung et al., 2019). So not only do PPIs influence emotional experience, they influence an individual’s beliefs and expectations as well as awareness about the experience of PPIs (“I just completed this program that taught me how to have more positive emotion, therefore I must be experiencing more positive emotion”). As with many interventions or treatments, effect of PPIs on self-reported emotion outcomes may be at least partially driven by these expectancy effects (Kirsch, 1997). PPIs often explicitly aim to increase positive emotion, so it is perhaps surprising that effects on self-reported positive emotion outcomes are not even stronger. Both emotional experiences and beliefs are likely to have an important impact on psychological and physical health. Investigators need to be cognizant of what memory processes may be at play with measures using different retrospective time frames and understand that these differences lead to distinctive types of information regarding impact of an intervention.
This review makes clear that it is critically important for researchers to begin with a theoretical model to guide the selection of emotion measure and decisions about how best to operationalize emotion. What are the hypothesized mechanisms of effect of the PPI under study? Which affective states, specifically, would be expected to change in response to the intervention? For example, results from a large population-based observational study demonstrate that PANAS PA, and in particular the item “active,” predicts lower risk of mortality (Petrie et al., 2018). Based on these findings, researchers testing PPIs that include a physical activity component (e.g., Celano et al., 2018), or other activities hypothesized to increase higher activation positive emotions, may consider including the PANAS to be sure to capture these increases. Similarly, in observational studies, optimism is a significant predictor of better health behaviors in patients following acute coronary syndrome (Millstein et al., 2016). Researchers who are particularly interested in how PPIs can improve such health behaviors should make sure to include measures of optimism to best capture the hypothesized mechanisms for beneficial health effects in this population (e.g., Duque et al., 2019; Huffman et al., 2016).
Researchers should also take characteristics of their population into account in selecting emotion measures. In a recent study that reported on the preliminary efficacy of a PPI tailored specifically for people with bipolar disorder, the authors gave careful thought to the selection of outcomes most appropriate for this population—differentiating between high- and low-activation emotions as well as how much participants valued different emotional states (Painter et al., 2019). The intervention content focused on low-activation positive emotions in particular (e.g., calm, rested, relaxed, peaceful, serene), because high-activation positive emotions may be a symptom of, or trigger for, mania.
Other potential measures of emotion include psychophysiology (e.g., skin conductance), behavioral observation of facial expression or posture, vocal pitch, and neuroimaging. These non-self-reported measures of emotion also have drawbacks such as imprecise match to subjective experience (Cacioppo et al., 2000), low convergence among the measures (Mauss & Robinson, 2009), and high participant burden. Indirect measures of emotion—those that do not require direct verbal self-report of an internal emotional state such as the affect misattribution procedure (Lee et al., 2019) or the Implicit Positive and Negative Affect Test (IPANAT; Quirin et al., 2009)—hold promise in laboratory studies but are untested in the field in response to PPIs. For example, in-lab manipulation of affect based on film clips results in changes on the IPANAT not reflected in self-report that correlated with cardiovascular reactivity (van der Ploeg et al., 2016). New approaches to emotion assessment using passive sensors in smartphones (e.g., accelerometer, microphone, calls, location, SMS) are showing reasonable correlation with self-reported mood (Servia-Rodríguez et al., 2017), but also have not yet been tested in response to PPIs and may provide an important new avenue for investigation of PPI effects on emotion.
Schueller et al. (2014) suggest that PPIs have effects through increases in positive emotion, positive cognitions, and/or positive behaviors, so in addition to specifying which type of positive emotion is hypothesized to improve for a given PPI, researchers should consider which cognitions and behaviors may be influenced and measure these outcomes as well. Beyond the measure and operationalization of self-reported emotion, the studies reviewed here also vary on a number of other characteristics that likely influenced the effect of the PPI, such as demographic characteristics of the sample, the type of stress they were experiencing, length and content of the intervention, make-up of the control conditions, and length of follow-up. Careful theoretically based selection and reporting of the emotion outcome measures in PPI research will move the field to a place where we can start looking at these potential moderators of intervention effects and toward answers to the question of which PPIs have the biggest impact on emotional well-being and for whom.
Supplemental Material
Appendix_-_Search_Terms_for_Moskowitz_Measuring_Positive_Emotion_Outcomes_in_Positive_Psychology_Interventions – Supplemental material for Measuring Positive Emotion Outcomes in Positive Psychology Interventions: A Literature Review
Supplemental material, Appendix_-_Search_Terms_for_Moskowitz_Measuring_Positive_Emotion_Outcomes_in_Positive_Psychology_Interventions for Measuring Positive Emotion Outcomes in Positive Psychology Interventions: A Literature Review by Judith T. Moskowitz, Elaine O. Cheung, Melanie Freedman, Christa Fernando, Madelynn W. Zhang, Jeff C. Huffman and Elizabeth L. Addington in Emotion Review
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
