Reducing Contrast Avoidance in GAD by Savoring Positive Emotions: Outcome and Mediation in a Randomized Controlled Trial

Abstract

This secondary analysis of a randomized controlled trial (RCT) examined whether contrast avoidance (CA) in Generalized Anxiety Disorder (GAD) could be reduced through savoring—purposeful engagement with positive emotions. 85 participants were diagnosed with GAD by both questionnaire and clinical interview. They were then randomly assigned to one of two seven-day smartphone-delivered ecological momentary interventions (EMIs). The SkillJoy EMI facilitated in-the-moment practice of savoring positive emotions in participants’ daily lives. The active self-monitoring control (ASM) was nearly identical to SkillJoy in ratings, activities, and language, yet omitted specific attention to positive emotion and savoring. CA was assessed by questionnaire at pre-trial and post-trial. Savoring was assessed by questionnaire at pre-trial and fifth-day mid-trial. Longitudinal linear mixed models and simple slope analyses examined CA change between and within conditions. Bias-corrected boot strapping path analysis examined mediation by savoring using individuals’ CA slopes extracted from a multilevel model as outcome. Results showed that SkillJoy led to significant reductions in CA, whereas the ASM control did not. The relation between treatment condition and reduction in CA was mediated by increases in savoring from pre- to mid-trial. It may be possible for treatment to meaningfully reduce CA in GAD, specifically through savoring practices.

1. Introduction

The function of worry has been at the heart of Generalized Anxiety Disorder (GAD) research for several decades (Newman et al., 2017). Functional theories propose that chronic worry serves some purpose for those with GAD, despite its drawbacks. Rationales for GAD treatment often attempt to address these functions directly (e.g., metacognitive therapy, intolerance-of-uncertainty therapy, etc.; Newman, Basterfield, et al., 2022; van der Heiden, 2012). Eliminating the function driving worry may not only reduce worry itself, but also prevent later relapse. One well-supported function of worry is the prevention of negative emotional contrasts. The Contrast Avoidance Model (CAM) of GAD proposes that worry increases and maintains distress over time to prevent aversive shifts toward greater negative emotion (Newman & Llera, 2011). This worry-sustained distress buffers against upsetting increases in negative emotion and decreases in positive emotion (negative emotional contrasts, or NECs). The CAM’s function for worry has been extensively studied and supported by laboratory experiments (Jamil & Llera, 2021; Kim & Newman, 2022; Llera & Newman, 2010, 2014; Skodzik et al., 2016), validated self-report questionnaire (Llera & Newman, 2017; Newman, Rackoff, et al., 2023), weekly diary (Crouch et al., 2017), and ecological momentary assessment (Baik & Newman, 2023; Newman et al., 2019; Newman, Schwob, et al., 2022; Vîslă et al., 2021). Even though worry lessens NECs for both those with and without GAD, those with GAD report a particularly heightened sensitivity and aversion toward them (Jamil & Llera, 2021; Kim & Newman, 2019, 2022; Llera & Newman, 2014, 2017; Newman, Rackoff, et al., 2023). Unlike non-GAD controls, those with GAD prefer to keep themselves in a distressed state by intentionally worrying, aiming to avoid being vulnerable to spikes in negative emotion (Llera & Newman, 2017; Newman, Schwob, et al., 2022). Within this basic research, contrast avoidance (CA) has been repeatedly suggested to be an important psychotherapy target. Yet no study thus far has examined whether clinical intervention can reduce CA specifically.

CA reduction may be particularly relevant for positive emotion interventions—a growing focus of anxiety treatment development (Carl et al., 2013; Craske et al., 2019). The CAM proposes that when worry increases negative emotion, it lowers positive emotion as well (Newman & Llera, 2011). Accordingly, worry inductions in experimental studies led to reduced positive emotions like contentment (Llera & Newman, 2014) and amusement (Kim & Newman, 2022). Reduced positive emotion was also associated with worry in daily life (Newman, Schwob, et al., 2022). Furthermore, the CAM theorizes that worrying leads to desirable positive emotional contrasts, or increased positive emotion and decreased negative emotion (Newman et al., 2019; Newman, Schwob, et al., 2022). For example, when a worry prediction does not come true (as is very often the case; LaFreniere & Newman, 2020), distress will be relieved. Also, when positive experiences followed worry, studies showed decreased negative emotion (Kim & Newman, 2022; Llera & Newman, 2014; Vîslă et al., 2021) and increased positive emotion (Kim & Newman, 2023) when compared to low or no worry. These positive contrasts may both negatively and positively reinforce the worry that preceded them (Newman, Schwob, et al., 2022). Thus, those with GAD are incentivized to be in a continually worried state rather than a positive one (Buhk et al., 2020).

Furthermore, when those with GAD are in a continually positive state, it leaves them vulnerable to sudden negative emotional shifts. This vulnerability to contrast may be uncomfortable for worriers. In fact, contrast avoidance mediated the relationship between GAD status and relaxation-induced anxiety (Kim & Newman, 2019). Moreover, those with GAD have reported preferring to “expect the worst and be pleasantly surprised” than to “expect success and be disappointed” (Llera & Newman, 2017; Newman, Rackoff, et al., 2023). A recent EMA study had a similar finding as well. Across 8 prompts per day for 8 days, a mixed GAD and depressed sample were more likely than controls to report that when they worried, they “focused on the negative, because allowing themselves to feel happy left them vulnerable to feeling terrible in the end” (Baik & Newman, 2023). Accordingly, those with GAD may dampen positive emotions by worrying in order to prevent a subsequent negative contrast. This self-protective process is known as “kill-joy thinking” (Quoidbach et al., 2010). Kill-joy thinking may be treated by encouraging an opposite approach—savoring positive emotions instead of extinguishing them.

Savoring is purposefully attending to, amplifying, and extending the duration of positive feelings (Bryant & Smith, 2015). According to meta-analysis, savoring-related practices reduced psychopathology and increased well-being across samples (Bolier et al., 2013). Savoring is incompatible with CA’s dampening of positive emotion and maintenance of distress. Furthermore, extending the duration of positive emotion may indirectly allow for exposure to negative shifts. When those with GAD dwell on positive emotions at length, they will eventually experience life’s negative contrasts. As these contrast events recur over time, worriers may learn to tolerate negative shifts. Thus, purposefully maintaining positive emotion may naturally foster habituation to contrast itself. It may also lead worriers to tolerate being vulnerable to potential future shifts, whether or not they actually occur. Through practice, worriers may learn to “drop their guard.” In other words, savoring may build acceptance of contrast vulnerability. Remaining in these positive emotional states also prevents positive contrasts, stopping them from reinforcing the worry process. Furthermore, if attention is highly immersed in pleasant affect, there may be less capacity for concurrent worry. As a result, there may be less overall worry and less distress to be reinforced by the relief of positive shifts. In either case, by increasing acceptance of negative contrasts and preventing worry reinforcement by positive contrasts, savoring may reduce CA over time—as well as the worry it feeds.

One prior study examined whether savoring practices did in fact reduce worry in GAD (the current study is a secondary analysis of that primary efficacy study). In the prior primary study, a smartphone-based ecological momentary intervention (EMI) was first developed to facilitate savoring practices in daily life: SkillJoy. Then, in a randomized controlled trial (RCT), SkillJoy was compared to a closely-related active treatment control over seven days in a GAD sample (LaFreniere & Newman, 2023). Compared to the active control EMI, SkillJoy’s savoring practices led to significantly greater reductions in worry. It also led to greater increases in positive emotions, savoring, optimism, and prioritizing positive goals and activities. Other RCTs examining the influence of positive affect therapies on anxiety have found similarly beneficial effects for both anxiety and depressive disorders (e.g., Craske et al., 2019). Yet neither SkillJoy’s primary analysis nor any other study has examined savoring’s impact on contrast avoidance.

We aimed to address this question in the current study, analyzing savoring’s effect on CA in GAD. We hypothesized that longitudinal linear mixed models would show greater reductions in CA across the RCT for SkillJoy users compared to control users. We also proposed that these reductions in CA would be explained by increases in savoring, as demonstrated by mediation in bias-corrected bootstrapping path analysis.

2. Method

This study was a secondary analysis of a randomized controlled trial comparing a savoring treatment to a closely-related, active treatment control. An institutional review board approved this study. It conformed to the Transparency and Openness Promotion guidelines. It was registered on ClinicalTrials.gov (ID: NCT05040061). Materials can be made available upon reasonable request to the authors. All data was collected prior to the COVID-19 pandemic.

2.1. Participants

85 participants (41 SkillJoy users, 44 control users) were identified through a university subject pool and recruited via email. All contacted participants met full DSM-5 criteria for GAD on the Generalized Anxiety Disorder Questionnaire-IV during pre-screening (GAD-Q-IV; Newman et al., 2002). Note that the DSM-IV and DSM-5 criteria for GAD are identical. Those who responded to the recruitment email were assessed for GAD by clinical interview a second time (the Mini-International Neuropsychiatric Interview (MINI) GAD section; Sheehan et al., 2015; Sheehan et al., 1998) by telephone. They were required to meet GAD criteria on both the MINI interview and the GAD-Q-IV to participate. The sample mean GAD-Q-IV score was 10.09 (lowest = 6.75), all surpassing the clinical cutoff of 5.7 and highly similar to those of community-based, treatment-seeking GAD samples (e.g., Newman et al., 2011). All participants spoke English, were 18 years of age or older, and were able to consent. Reported participant identities were 82.4% White, 4.7% Black, 4.7% Asian, 3.5% Multi-Racial, 3.5% Latinx, and 1.2% Middle Eastern. Regarding gender, 77 identified as women and 8 identified as men.

2.2. Measures

The Contrast Avoidance Questionnaire – Worry (CAQ-W; Llera & Newman, 2017).

The CAQ-W is a 30-item self-report measure of participants’ discomfort with and avoidance of negative emotional contrasts and of sustained positive emotion. It assesses intentional use of worry to sustain distress, dampen positive emotion, avoid negative emotional contrasts, and increase the likelihood of acute positive contrasts. Participants rate the extent to which items are true of them from (1) “not at all true” to (5) “absolutely true” (total scores ranging from 30 to 150). Example items are “I worry to guard against sudden shifts in my mood.” “Worrying is better than feeling good and then being thrown off by a negative event.” “It is better to have worried first and then get the best outcome than to expect the best all along.” Llera and Newman (2017) have shown the CAQ-W to have strong retest reliability (total scale r = .90) and robust internal consistency reliability. The scale also has good construct validity as shown by expected differences between scores of GAD and non-anxious groups, as well as convergent and discriminant correlations with measures of GAD, worry, negative emotion, sensation seeking, and fear of happiness in expected directions (Llera & Newman, 2017). Internal consistency in our study’s sample was good (α = .92). We used pre- to post-trial change in CAQ-W total score as outcome in both our efficacy and mediation models. Reliable change criterion for the CAQ-W in our sample was 15.70.

The Savoring the Moment Subscale of the Savoring Beliefs Inventory (SBI; Bryant, 2003).

The SBI is a 14-item survey measuring one’s self-perceived ability to savor. Respondents rate how true a particular statement is for them on a scale of (1) strongly disagree to (7) strongly agree. The SBI contains 3 subscales (Reminiscing, Anticipation, and Savoring the Moment). However, in this study we only employed the Savoring the Moment subscale, which measures savoring of present-moment emotional experience. We theorized that it was in-the-moment intentional engagement with positive emotion and its perpetuation that would lead to reduction of contrast avoidance (i.e., choosing to stay with vulnerability to emotional shift, to sit with and extend positive emotions, etc.). Thus, we focused on the Savoring the Moment subscale as our mediator of CA change. Importantly, neither the SBI Reminiscing nor the SBI Anticipation subscale assess the purposeful enjoyment and extension of the in-the-moment positive emotions that result from reminiscing or anticipation. Thus, Savoring the Moment was most appropriate for measuring and representing the construct we expected would generate change in CA. Example items include: “I know how to make the most of a good time” and “I can prolong enjoyment by my own effort.” Subscale scores range from 1 to 7. The subscale has good retest reliability, internal consistency reliability, and convergent and divergent validity (Bryant, 2003). Internal consistency was also good (pre: α = .80; post: .86) in the current sample.

2.3. Smartphone Software Applications

PACO: The Personal Analytics Companion (PACO Developers, 2018)

PACO: The Personal Analytics Companion (PACO Developers, 2018) is an online software package for developing and implementing ecological momentary studies on smartphone. The PACO mobile application allows for fixed, user-editable, and stratified random prompting on both iOS and Android phones. Quip (Taylor & Gibbs, 2018) is a smartphone application for creating editable documents, embedded spreadsheets, and task lists. Multiple users can edit Quip documents in real time across separate devices.

2.4. Procedure

First, we used a random number generator to randomly assign eligible participants to either the SkillJoy treatment condition or the active self-monitoring (ASM) control condition. Participants first presented to the laboratory and provided informed consent. They used PsychData to complete baseline questionnaires on a computer, followed by a computerized task for a separate study. They were then given a description and rationale for how their assigned intervention would reduce their GAD symptoms and increase positive emotions (see Supplementary Materials A). Afterward, they acquired study apps on their personal phones and were acquainted with their respective software. All participants received Quip documents with their condition’s rationale, as well as instructions on how to complete each part of the intervention, a spreadsheet for scheduling, and a blank document for notes. We then used online presentation slides to train participants on how to use their assigned app. Participants had continual access to these slides throughout the trial. Both groups were trained in study procedures, were informed of the required 5^th day phone call to check compliance, planned for 8^th day study tasks, and were given opportunities to ask questions.

Participants underwent each condition’s EMI for seven days via the PACO app. The EMIs were delivered in participants’ natural environments throughout their days. Each day participants received eight prompts that required engagement and assessment responses. A ninth prompt in the afternoon only served as a reminder for a condition-specific task. See Supplementary Materials B for all EMI prompts. Each day’s first and last prompt times were user-editable to fit user sleep schedules. Other prompts were delivered at random times within pre-determined intervals. Four daily prompts in both conditions included the PANAS-X Joviality scale. Participants also received a compliance check phone call on the morning of the fifth day of the trial. This phone call inquired about participant rate of compliance, level of effort, and if any harm had occurred. After this phone call, participants completed a battery of mid-trial questionnaires at home on their personal devices. This mid-trial battery included the SBI Savoring the Moment subscale. On the eighth day of the study, participants returned to the laboratory and completed the same questionnaires and tasks given at baseline.

2.4.1. SkillJoy Condition

SkillJoy Training:

SkillJoy participants were first trained in savoring practice exercises that resembled their intervention. They were guided through remembering and describing a positive moment from their past, as well as one good experience from their day. Next, they engaged in present-moment savoring. They first chose a small candy or dried fruit from multiple options. They were then guided through attending to the sensory, emotional, and cognitive experience of eating it in-the-moment, purposefully attempting to amplify and extend the resulting positive emotion. After, they brainstormed and scheduled enjoyable activities for the week (all rated 7 or greater on a 10-point enjoyment scale) and recorded them in Quip.

SkillJoy EMI:

Present-moment savoring prompts were delivered by stratified random prompting (three prompts). The other six prompts were delivered at fixed times distributed across the day, with two being user-editable. SkillJoy included a variety of savoring interventions modified to target GAD pathology. These interventions included 1) Enjoyable Activity Savoring. Participants were prompted at a time of their choosing (before 11:30 AM) to schedule an exact time for an enjoyable activity for the following day. They were then reminded (in this prompt) and later prompted (in a fixed reminder prompt) to practice savoring upcoming positive activities. At the end of the day, they were asked to focus on what they liked about their enjoyable activity; 2) Present-Moment Positive Evaluation. Participants received three identical daily prompts guiding them to focus on and savor what they enjoyed about the present moment; 3) Savoring recent memories. Participants received two daily prompts that encouraged reflective savoring of recent activities and events; 4) “Counting Blessings” Technique. Twice a day participants received prompts guiding them to consider and write about events that turned out better than expected and events that were enjoyable or went well. 5) Looking forward to the day’s events. During their first daily morning prompt, participants were asked to savor anticipation of an upcoming positive activity for that day.

2.4.2. Active Self-Monitoring Control EMI

Active Self-Monitoring (ASM) Training:

ASM control participants were trained using content that was identical to that of SkillJoy participants, yet it omitted all aspects focused on engaging with positive emotion. To practice reflective thinking, they were first guided to remember a day one week in the past and an event from the current day. After, they engaged in present-moment self-awareness. They chose a small candy or dried fruit and were asked to pay attention to their thoughts and feelings while eating it and to describe them. This activity resembled a mindfulness exercise, but did not explicitly direct attention to any specifically positive emotions or thoughts. Next, participants described and scheduled possible future events they had to do or planned to do during each day of the following week. They recorded them in their condition’s Quip template.

ASM EMI:

The ASM control included self-monitoring activities mirroring SkillJoy’s interventions and language, but omitting components for increasing or sustaining positive emotion. Careful attention was paid to match SkillJoy in its exact structure, wording, and tone. Within prompts, they received reminders to pay attention to their thoughts and feelings (instead of the reminders to savor positive emotion in SkillJoy). Interventions included: 1) Planning tomorrow’s activities. Each morning at a time of their choosing (before 11:30AM) participants received a prompt to schedule “the day’s major events” for the following day using their Quip schedule; 2) In-the-moment thoughts and feelings. Participants received three identical stratified random prompts to attend to their thoughts and feelings at that present moment. In addition, participants received a separate prompt that reminded them to be aware of their thoughts, feelings, and plans in general. 3) Remembering. Participants received two prompts that encouraged remembering events from their day, no matter what events these were. These prompts asked them to think of an event from their day, take 60 seconds to remember the event in detail, and complete a series of ratings about it. 4) Recording the day’s events. During the day, participants received two prompts guiding them to think about and record recent events. At the end of the day, they were asked to remember and think about their events, thoughts, and feelings from across that day. They wrote down three events that happened that day and reflected on the day’s most salient event. 5) Anticipating the day’s biggest event. In each day’s initial prompt, participants were asked to anticipate and record the day’s most important upcoming event.

2.5. Planned Statistical Analyses

According to a power analysis with RMASS II, 70 total participants were required for a two-level longitudinal linear mixed effects model of adequate power (0.80) and moderate effect (d = 0.5). To account for missing data, we used multiple imputation with 100 iterations of the Markov Chain Monte Carlo (MCMC) method, creating 10 imputed datasets. We then used longitudinal multilevel modeling to examine contrast avoidance outcome differences between conditions. All multilevel modeling results reflect analyses pooled across the imputed datasets. Longitudinal linear mixed models compared between-condition differences in pre- to post-trial change in the CAQ-W total score (contrast avoidance). This model included the intercept, linear time trend, intervention condition, and interaction between time trend and condition as fixed effects and the intercept as a random effect. Afterward, multilevel simple slopes analyses were conducted to examine the degree of change in CA within each condition.

For the mediation analysis, we employed bias-corrected bootstrapping regression-based path analysis (Preacher & Hayes, 2008). We used full information maximum likelihood (FIML) estimation to account for missingness. The reported confidence intervals are 95% bootstrap confidence intervals. Our mediation analyses resulted from a two-step process: 1) Extracting individual participants’ slopes of change in CAQ-W scores from a longitudinal linear mixed model, then 2) analyzing mediation of intervention condition’s effect on the extracted CAQ-W random effect slopes by change in savoring scores. We used multilevel modeling in the R statistical program’s lme4 and lmerTest packages to extract each individual’s slope of change in CAQ-W scores from pre- to post-trial. To do so, we specified a linear mixed effects model where CAQ-W scores were predicted by a fixed effect of intercept, a random effect of intercept, and a random effect of time trend from pre- to post-trial. The fixed effect of time trend was not specified so that its common influence would not be removed from individuals’ slopes of change. To acquire participants’ slopes of change, we then extracted the random effects of time trend on CAQ-W scores for each individual. These slopes in CA were then used as the outcome variable of our mediation model. We used Mplus statistical software to run a bias-corrected bootstrapping path analysis where savoring change from pre- to mid-trial was specified as a mediator for the relationship between intervention condition and the slopes of pre- to post-trial change in CAQ-W scores. We chose to use pre- to mid-trial change (as opposed to mid- to post-trial change) as our mediator for several reasons. Most importantly, methodologists argue it is problematic for the treatment mediator to be concurrent with the outcome variable when examining outcome mediation (Kraemer et al., 2002). To be viewed as a putative mechanism of outcome, the mediator should temporally precede the outcome, yet occur during treatment. Furthermore, studies find that early sudden gains often occur the first half of treatment, which then predict better post-trial outcomes (see meta-analysis; Shalom & Aderka, 2020).

Between-condition differences in baseline measures, degrees of missingness, and compliance rates were examined with independent samples t-tests. Reliable change index analyses used the formula of Jacobson and Truax (1991). Cohen’s d was calculated in the traditional manner for t-tests, (M₂ − M₁) ⁄ SD_pooled where $S{D}_{\text{pooled}}=\sqrt{((S{D}_1{}^2+S{D}_2{}^2)/2)}$ (Cohen, 1988). In linear mixed models and simple slope analyses, Cohen’s d was calculated with an alternative formula for multilevel models, $d=2t/(\sqrt{df})$, as recommended by Rosenthal (1994). For bootstrapping path analyses, Cohen’s d effect sizes were calculated by the formula d = B/(√(n)*SE). We report unstandardized coefficients for path betas (a, b, ab, c, and c’).

3. Results

3.1. Descriptive Statistics and Baseline Group Difference Tests

3.1.1. Missingness.

4.71% of values were missing. The number of missing values between SkillJoy users and controls did not significantly differ (t(83) = 1.10, p = .273, d = 0.24). Thus, the number of non-responses was low and approximately equal for both conditions.

3.1.2. Compliance rates.

The average intervention compliance rate was 92.63% of prompts completed. There was no difference in compliance between SkillJoy users (M = 91.85%, SD = 16.36) and active self-monitoring (ASM) controls (M = 93.36%, SD = 13.16; t(83) = 0.47, p = .639, d = 0.10). Both conditions had a mode of 100% compliance. Thus, compliance rates were high and did not meaningfully differ between conditions.

3.1.3. Baseline difference tests.

SkillJoy users and ASM controls did not differ significantly in level of contrast avoidance scores at baseline (t(83) = −1.37, p = .174, d = 0.29). The conditions also did not differ in savoring the moment scores at baseline (t(83) = 0.95, p = .346, d = 0.20). In addition, the conditions did not differ in GAD symptoms according to their continuous scores on the GAD-Q-IV (t(83) = −0.65, p = 0.520, d = −0.10).

3.2. Contrast Avoidance Outcome Results

There was a marginally significant condition by time interaction: SkillJoy users achieved a greater reduction in contrast avoidance than ASM controls across the trial (t(83) = −1.73, p = .083, d = −0.38). This partially supported our hypothesis that savoring practice would lessen contrast avoidance to a greater degree over time than a non-savoring treatment alternative (i.e., more than placebo, spontaneous remission, relevant common factors, etc.). Simple slope analyses found that SkillJoy users significantly decreased in contrast avoidance (β = −8.68, t(81) = −2.92, p = .004, d = −0.65). Yet ASM controls did not significantly decrease (β = −1.57, t(81) = −0.54, p = .590, d = −0.12). This finding supports our hypothesis: Savoring lessened CA, whereas the active control did not. 32.5% of SkillJoy users and 19% of controls surpassed the reliable change criterion over the seven days of EMI use. Overall, longitudinal linear mixed models suggested SkillJoy led to significant reductions in CA, whereas the ASM control did not.

3.3. Mediation by Savoring Results

Figure 1 summarizes the results of the longitudinal mediation analyses. Results supported the hypothesized savoring mediation model: According to a 95% bias-corrected bootstrap confidence interval, the indirect effect was significant (ab = −1.02, 95% bootstrap CI = [−2.49, −0.11], d = −0.18, R² = 0.27). This model indicated that the effect of treatment condition predicting pre- to post-trial reduction in contrast avoidance was mediated by pre- to mid-trial increases in savoring. After including mediation by savoring, the remaining direct effect was not significant (c’ = −0.70, 95% bootstrap CI = [−2.52, 1.03], d = 0.08). Note that treatment condition significantly predicted increases in savoring pre- to mid-trial (favoring SkillJoy) and greater savoring pre-mid trial significantly predicted decreases in contrast avoidance pre- to post-trial (see Figure 1)¹. SkillJoy not only led to significant reductions in contrast avoidance, but did so by increasing savoring—its theorized mechanism.

Figure 1.

Conceptual diagram of a mediation model where the relationship between treatment condition (X) and pre- to post-trial change in contrast avoidance (Y) is mediated by pre- to mid-trial change in savoring (M). Unstandardized regression coefficients: a = effect of X on M, b = Effect of M on Y, ab = indirect effect (representing mediation), c’ = direct effect of X on Y (with M included in the model), c = total effect of X on Y (with M not included in the model). ^†marginally significant at α = .10. * significant at α = .05. **significant at α = .01.

4. Discussion

The motivation to prevent future negative emotional shifts promotes excessive worry in GAD (Llera & Newman, 2017; Newman, Rackoff, et al., 2023). Weakening this contrast avoidance (CA) function of worry may be valuable for treating the disorder. The current study tested whether savoring positive emotions—a mindset counter to CA—was one means to diminish CA. To this end, we examined whether a smartphone-based intervention (SkillJoy) would reduce CA by improving the quality and frequency of savoring practices in GAD. Across 7 days in a randomized controlled trial, SkillJoy reduced CA over time to a greater degree than an active treatment control in a GAD sample. According to simple slope analyses, SkillJoy led to significant CA reductions, whereas the active control did not. Furthermore, the relationship between treatment condition and decreases in CA pre- to post-trial was mediated by pre- to mid-trial increases in savoring. Thus, the SkillJoy intervention was able to change CA. Our findings suggest this change was brought about through engagement with positive emotions.

Over a decade of basic research on contrast avoidance has underscored its influence on worry in GAD (Llera et al., 2022). Despite calls for psychotherapy to address CA, it has been an open question as to whether treatment could meaningfully change it. The current study suggests psychological intervention can indeed alter CA over time. In fact, our findings demonstrated that it was possible to change CA after only one week of consistent savoring. Results also showed that the CAQ-W, a trait measure of CA, was sensitive to change from treatment. Changing worry is certainly a worthy endeavor. Yet changing the underlying functions driving worry—like CA—may encourage long-term maintenance of worry reductions and other gains. Fortunately, according to our initial evidence, CA may respond to at least one method of modification.

In this study, that method was savoring—purposefully engaging and extending positive emotions (Bryant & Smith, 2015). Our pattern of findings suggest it was specifically savoring practices that reduced CA. First, CA was only significantly lessened by the SkillJoy condition, not the ASM control. This RCT was highly controlled, focusing solely on increasing savoring as its manipulation. The only notable difference between the two conditions was the inclusion of directives to savor—present in SkillJoy, yet absent from the control. Even the language used in the EMIs was nearly identical between conditions. The only exception was encouraging extended engagement with positive emotion in SkillJoy. Second, bias-corrected bootstrapping path analysis showed that increases in savoring scores mediated treatment condition predicting CA change. Thus, the relationship between SkillJoy and CA appeared to be explained by savoring itself. Although our highly controlled design demonstrated the efficacy of the SkillJoy condition’s unique between-condition elements, mediation adds value by statistically underscoring savoring as a mechanism. Other variables could have possibly contributed to CA reductions (such as simple increases in positive emotions), yet savoring was still revealed to be a significant mediator. Lastly, ASM did not change CA despite the fact that it did change other outcomes in the primary analyses: worry, depression symptoms, prioritizing positive activities and goals, kill-joy thinking, and optimism (LaFreniere & Newman, 2023). Even though ASM changed these other outcomes to a lesser degree than SkillJoy, ASM still had influence on them—without savoring. That was not the case with CA as an outcome. Thus, savoring—the difference between the two conditions—appeared to be the element changing contrast avoidance.

Yet why might savoring be responsible for this change? First, savoring-related interventions have reduced various symptoms of psychopathology in many studies (Bolier et al., 2013). In particular, SkillJoy aimed to foster intentional attention to positive emotion, prolonging it to the best of each user’s ability. This intention is directly counter to the self-protective mindset found in CA—one that maintains distress at length (Newman & Llera, 2011; Newman et al., 2013). As an alternative, extended savoring may strengthen client tolerance to being in a state that is vulnerable to possible emotional shifts. In one study, fear of negative emotional contrasts mediated the relationship between higher GAD symptoms and lower savoring (Malivoire et al., 2022). SkillJoy’s many repeated directives to savor (8 per day) aimed to facilitate a continuous state of being open to contrast, lessening “kill-joy” thinking (Quoidbach et al., 2010). As those with GAD “sit with” positive emotions, dwelling on them past the point where it is comfortable, they may learn to let go of bracing for negative shifts. Savoring may also facilitate habituation to such shifts when they actually occur. If savorers with GAD fluctuate from positive to negative states, they may learn that they can handle living without worry’s emotional buffer. They may realize the costs of CA outweigh its benefits if those benefits are ultimately unnecessary. Living in a prolonged state of contentedness with brief spikes in distress may be better than a prolonged state of distress that prevents such spikes. Moreover, at a basic level, it is unlikely for savoring to co-occur with kill-joy thinking within the same moment (though one may vacillate between them across time). The intention to sustain enjoyment is at odds with the intention to sustain distress.

Extended positive states may also prevent the reinforcement of worry that results from positive emotional contrasts (i.e., shifts from worry-related states to non-negative states). When those with GAD worry, it reduces positive emotion (e.g., Kim & Newman, 2022; Llera & Newman, 2014; Newman, Schwob, et al., 2022) and increases distress (e.g., Jamil & Llera, 2021; Llera & Newman, 2010; Newman et al., 2019; Newman, Schwob, et al., 2022). This distress is later relieved when worries do not come true (e.g., Kim & Newman, 2022; Vîslă et al., 2021) and positive emotions spike when positive events occur after worrying (e.g., Kim & Newman, 2022; Llera & Newman, 2014). Both of these post-worry outcomes are avenues for a positive emotional contrast. This relief (negative reinforcement) and pleasant feeling (positive reinforcement) after worry may promote worry (Newman, Schwob, et al., 2022). Yet if those with GAD savor to stay in a positive state independent of worry outcomes, it may reduce the possibility of positive contrasts and lessen worry’s reinforcement cycle. Interrupting this reinforcement may also lessen the motivation to avoid contrast. Such an urge may ultimately be rewarded when following it results in positive contrasts, yet extinguished when ignored. Due to this process, habituation to CA, and acceptance of contrast vulnerability, SkillJoy’s savoring practices may reduce CA motivation.

At the same time, savoring may achieve more than lowering CA and its resulting worry. Research suggests savorers garner positive emotional benefits as well. When CA’s distress maintenance is curtailed, it may allow for greater enjoyment and happiness long-term. Lessening CA essentially lessens kill-joy thinking responses to positive events. Lower CA may permit greater subsequent savoring, creating an upward cycle favoring well-being. Studies have also shown that naturalistic savoring after positive daily events not only improved regulation of negative emotions and symptoms of GAD and depression (Doorley & Kashdan, 2021; Khazanov et al., 2019), but also increased the intensity of positive emotions and moods over time (Doorley & Kashdan, 2021; Jose et al., 2012). This has been observed at a neurological level as well. When compared to dampening and a neutral control during a single EEG session, savoring increased neural reward reactivity to a rewarding task (Irvin et al., 2020). Furthermore, savoring that was perceived as more successful by participants led to relatively greater reward reactivity. When compared to a control task, savoring also increased visual attention to positive stimuli (Corman et al., 2020). Overall, savoring that lowers the kill-joy tendencies of CA may promote greater savoring and positive emotion later on.

The current RCT provided preliminary evidence for promising effects in GAD, yet also had limitations. First, our results cannot be generalized past one week, requiring long-term follow-up to know whether savoring has a lasting effect. Since it is possible that the effects of our 7-day treatment on CA were not maintained, a longer-term study would have been more beneficial. Note our sample was also limited in generalizability, due to it being mostly comprised of White women who were college students. In addition, momentary contrast avoidance could also have been assessed directly. Future studies would do well to examine whether savoring reduces intentional avoidance of negative contrasts in real time. Such studies may also be able to examine whether sharp shifts from savoring-supported positive emotion to negative emotion do in fact promote habituation to emotional contrast, or at least perceived acceptance of contrast. We would also like to acknowledge the slightly increased risk of Type I error by examining simple slopes for an outcome model without a fully significant time by condition interaction. Lastly, although in-the-moment ecological momentary intervention was a strength of this study, this method may not fully represent the effects of savoring in person-to-person psychotherapy. The effects of savoring in a therapist-guided psychotherapy context should also be studied.

Within such a context, psychotherapists treating GAD may be able to capitalize on the savoring effects found in this study. Whether by EMI or in psychotherapy, training and practice in savoring may fruitfully complement existing techniques that focus on negative phenomena. The current results support clinicians teaching practices for purposeful present-moment attention to positive emotion, as well as strategies for amplifying and extending that emotion in GAD. Therapists may benefit from leading active practice of savoring in session, so it can be best utilized in daily life. Clients can be guided to identify exactly what they enjoy about an experience. They can then attend to any resulting positive emotion from those factors, sitting with the emotion for as long as possible. They should do so despite any discomfort or motivation toward CA—a notable possibility in a GAD sample. Clients can then be assigned savoring activities between sessions. Savoring strategies include specifically noting, listing, and focusing on enjoyable elements of an activity, meditation on positive feelings, images, or sounds/music, recalling and recording positive memories, verbally expressing, discussing, and sharing what one enjoys with other people, and other approaches. As in SkillJoy, repeated smartphone reminders to savor throughout the day may also be helpful, aiming to counter CA’s habitual nature. Furthermore, treatment may be augmented by specifically teaching clients about contrast avoidance and kill-joy responses. Therapists can then offer worriers another approach to emotion, introducing savoring and other techniques. As demonstrated by this study, clients may then let go of their guard by holding on to their joy.

Table 1.

Means and standard deviations for savoring and contrast avoidance measures at pre-trial, mid-trial, and post-trial.

	Pre-Trial		Mid-Trial		Post-Trial
Measure	SkillJoy M (SD)	ASM M (SD)	SkillJoy M (SD)	ASM M (SD)	SkillJoy M (SD)	ASM M (SD)
SBI Savoring the Moment	3.91 (1.06)	4.14 (1.19)	4.58 (0.69)	4.05 (1.40)	--	--
CAQ-W	97.93 (22.93)	92.00 (16.63)	--	--	89.18 (23.01)	90.36 (20.49)

Note. SBI = Savoring Beliefs Inventory; CAQ-W = Contrast Avoidance Questionnaire - Worry Total Score

Highlights.

• An RCT compared savoring practices to a control on changes in contrast avoidance.

• Savoring practices significantly reduced contrast avoidance. The control did not.

• Reductions in contrast avoidance were mediated by increases in savoring.