Worry and rumination enhance a positive emotional contrast based on the framework of the Contrast Avoidance Model

Abstract

The Contrast Avoidance Model (CAM) suggests that worry increases negative affect and decreases positive affect. CAM also suggests that in response to a positive event, higher worry enhances the probability of experiencing greater decreased negative affect and increased positive affect (positive emotional contrasts; PECs). Consequently, worrying may be reinforced by repeated PECs. However, no study has tested whether rumination enhances PECs. Also, emotional specificity in these processes has not been considered. Therefore, we tested whether both rumination and worry enhanced PECs related to specific emotions. After resting baseline, participants with pure generalized anxiety disorder (GAD group, n = 91), pure depression symptoms (depression group, n = 91), and non-GAD and non-depressed healthy controls (HCs, n = 93) engaged with randomly assigned induction tasks (either worry, rumination, or relaxation), and then watched an amusement video. Regardless of group, both worry and rumination increased sadness and fear and decreased amusement more than relaxation from baseline. However, worry increased fear more than rumination, and rumination increased sadness more than worry. Although all inductions led to PECs during the video, worry enhanced fear PECs more than rumination, and rumination enhanced sadness PECs more than worry. The GAD group who worried experienced the most salient PECs of amusement relative to other groups.

1. Introduction

Worry and rumination are two commonly studied forms of repetitive negative thought that have been found to lead to various emotional and behavioral problems (Calmes & Roberts, 2007; Muris, Roelofs, Rassin, Franken, & Mayer, 2005). Worry is a defining feature of generalized anxiety disorder (GAD; American Psychiatric Association, 2013), whereas rumination is one of the major contributors to major depressive disorder (MDD; American Psychiatric Association, 2013). Worry is defined as a chain of thoughts and associated negative feelings over anticipated future threats (Borkovec, Robinson, Pruzinsky, & DePree, 1983). Rumination is defined as a chain of thoughts and negative feelings over depressive symptoms and their negative implications (Nolen-Hoeksema, 1991). Rumination diverges from worry in its focus, which is more on past negative events such as losses or failures, whereas worry is considered more future-oriented (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Worry has been studied primarily in relation to GAD and other anxiety disorders and rumination has been studied predominantly in the literature of depressive disorders.

Despite their different origins, a considerable number of studies have shown convergence between the two repetitive negative thoughts. Worry and rumination interchangeably occurred in individuals with GAD and MDD (Blagden & Craske, 1996; McLaughlin, Borkovec, & Sibrava, 2007; Starcevic, 1995; Szkodny & Newman, 2019; Watkins, 2008; Watkins, Moulds, & Mackintosh, 2005). In addition, a high correlation between worry and rumination was reported in both non-clinical and clinical samples (Segerstrom, Tsao, Alden, & Craske, 2000; Szkodny & Newman, 2019). Furthermore, worry and rumination shared the same latent structures and factor loadings (Segerstrom et al., 2000; Topper, Molenaar, Emmelkamp, & Ehring, 2014). Based on these findings, researchers have posited that there might be a common underlying mechanism in these two repetitive negative thoughts.

The Contrast Avoidance Model (CAM) has been suggested as one such mechanism (for a more complete review, see Newman & Llera, 2011; Newman, Llera, Erickson, Przeworski, & Castonguay, 2013). CAM suggests that regardless of diagnostic characteristics (i.e., GAD and MDD), worry increases and sustains negative emotion and decreases positive emotion. This heightened negative affect/reduced positive affect facilitates the avoidance of a further sharp increase in negative emotion or decrease in positive emotion when encountering a negative event. This phenomenon is referred to as the avoidance of a negative emotional contrast (NEC). Similar to worry, rumination also increases and sustains negative emotions (Nolen-Hoeksema et al., 2008; Nolen-Hoeksema, 2000). Thus, rumination might also be motivated by avoidance of NECs.

In fact, there is empirical evidence for CAM as a mechanism of worry and rumination. For example, laboratory studies found that in both high and low worriers, worry (vs. relaxation) increased negative affect, and this facilitated the avoidance of NECs upon exposure to fear and sadness videos (Llera & Newman, 2014, 2010). Momentary assessment studies also showed that worry reduced the likelihood of experiencing NECs (Newman et al., 2019, 2022) and blunted their effects (Crouch, Lewis, Erickson, & Newman, 2017). Furthermore, GAD or depressed participants (vs. controls) were more likely to report intentionally inducing negative affect via worry or other strategies to prevent NECs (Llera & Newman, 2017; Newman, Rackoff, Zhu, & Kim, 2023). Moreover, two experimental studies examined both worry and rumination in individuals with GAD and MDD. Both studies found that worry and rumination facilitated the avoidance of NECs in response to negative emotional exposure or a stressful cognitive task (Jamil & Llera, 2021; Kim & Newman, 2022). Thus, CAM appears to apply to both perseverative thought styles.

CAM also posits that worrying is motivated by the probability of experiencing a positive emotional contrast (PEC). A PEC is defined as a sharp increase in positive emotion (positive affect PEC) and/or a sharp decrease in negative emotion (negative affect PEC; Newman et al., 2013, 2019, 2022). For example, those with pathological worry (vs. non-worriers) were more likely to endorse thinking that it was better to expect the worst because if the worst did not happen or turned out better than expected, they would be pleasantly surprised (Borkovec & Roemer, 1995; Borkovec, Hazlett-Stevens, & Diaz, 1999; Llera & Newman, 2017). Because most worries do not come true (91% of the time in one study; LaFreniere & Newman, 2020), chronic worriers are likely to encounter frequent negative and positive affect PECs. Also, worry and its associated negative affect are theorized to increase the probability of a PEC, even if the person encounters a positive or benign event unrelated to the worry. Thus, the theory suggests that worrying is likely to be naturally reinforced by increased likelihood of PECs. This may also be the case for rumination.

There is empirical support for the PEC tenets of CAM. Compared to relaxation and neutral inductions, prior worry led to greater subsequent decreased negative affect (PEC) upon exposure to a humorous video (Llera & Newman, 2014). In an ecological momentary assessment (EMA) study, naturalistic worry (vs. not worrying) predicted a higher likelihood of a negative affect PEC in the next hour among those with and without GAD (Newman et al., 2019). In another EMA study, mood and worry were assessed before and after social interactions in those with GAD (Newman et al., 2022). Although on average participants experienced mood improvement following their interactions suggesting that most of these interactions were benign or positive, this effect was moderated by worry. In particular, if they were not worried before the interaction, participants experienced an NEC (both decreased positive mood and increased negative mood) from before to after the interaction. On the other hand, if they were worried before the interaction, they were more likely to experience a PEC (both decreased negative affect and increased positive affect). These findings suggest that in those with GAD, whereas not worrying was both positively and negatively punished, worrying was both positively and negatively reinforced. Furthermore, compared to non-anxious controls, those with GAD and depression were more likely to endorse intentionally worrying or intentionally inducing negative affect to increase the likelihood of experiencing a PEC (Llera & Newman, 2017; Newman, Rackoff, Zhu, & Kim, 2023). Although multiple studies showed that higher depressive symptoms were associated with larger decreased negative affect and increased positive affect in response to positive events (Khazanov, Ruscio, & Swendsen, 2019; Panaite, Koval, Dejonckheere, & Kuppens, 2019), no prior study to our knowledge examined how rumination contributed to these effects, leaving open the question of whether rumination and the negative emotional state induced by rumination enhances the possibility of negative and positive affect PECs.

It has also been suggested that the types of emotion most induced by worry and rumination are different from each other. For example, worry induced anxiety and fear more than sadness, and rumination elicited more sadness than anxiety and fear (Borelli, Hilt, West, Weekes, & Gonzalez, 2014; McLaughlin et al., 2007; Zetsche, Ehring, & Ehlers, 2009). Based on these findings, Kim and Newman (2022) examined CAM regarding different types of emotion and found that prior worry facilitated greater avoidance of an NEC of fear (vs. sadness) upon exposure to a fearful video, and prior rumination led to greater avoidance of an NEC of sadness (vs. fear) upon exposure to a sadness video. However, no prior study examined whether a PEC would demonstrate similar emotional specificity (i.e., greater reduced fear (vs. sadness) upon exposure to a positive stimulus following prior worry, and greater reduced sadness (vs. fear) upon exposure to a positive stimulus following prior rumination).

Therefore, the current study attempted to explore whether PECs to a positive stimulus would be more likely following rumination and worry and would demonstrate emotional specificity depending on whether the person had worried or ruminated. We also tested whether PECs occurred transdiagnostically. Therefore, we screened participants to form a pure GAD group (GAD group), pure depressive symptoms group (depression group), and non-GAD and non-depressed healthy control groups (HCs). In addition, we tested whether there was emotional specificity to PECs. Examining these research questions may allow a more comprehensive understanding of the potential reinforcement motivation and maintenance factors behind worry and rumination, and ultimately benefit the identification of viable treatment targets for more personalized interventions. Our hypotheses were as follows:

• Hypothesis 1a. Worry and rumination would increase fear and sadness from resting baseline more than relaxation.

• Hypothesis 1b. Because prior worry and rumination would lead to greater increased fear and sadness from baseline to induction relative to relaxation, worry and rumination would result in a sharper decrease in fear and sadness from induction to amusement video (i.e., enhancement of negative affect PEC by worry and rumination).

• Hypothesis 2a. Worry would increase fear more than rumination and rumination would increase sadness more than worry from resting baseline (i.e., emotional specificity of worry and rumination).

• Hypothesis 2b. Because worry would increase fear from baseline to induction more than rumination, worry would result in a sharper decreased fear from induction to amusement video than rumination. Because rumination would increase sadness from baseline to induction more than worry, rumination would lead to greater decreased sadness from induction to amusement video than worry (i.e., emotional specificity of worry and rumination).

• Hypothesis 3a. Worry and rumination would decrease amusement from resting baseline to induction more than relaxation.

• Hypothesis 3b. Because worry and rumination would lead to greater decreased amusement from baseline to induction relative to relaxation, this would lead to sharper increased amusement from induction to amusement video (i.e., enhancement of PEC by worry and rumination as measured by positive affect).

We did not have hypotheses about the effects of group but left this as an exploratory question due to the paucity of studies examining group differences in PECs, and no prior PEC studies that included a depression group.

2. Method

2.1. Research design

Data were collected as part of a larger research project examining effects of worry and rumination on both NECs and PECs. Findings on NECs were published elsewhere (i.e., Kim & Newman, 2022). To test the main and interaction effects of group (i.e., GAD group, depression group, and HCs) and induction (i.e., worry, rumination, and relaxation) on emotions (i.e., fear, sadness, and amusement), we used a mixed design with two between-subject factors (i.e., induction and group) and time as a repeated-measures factor.

2.2. Participants

In total, 275 participants (78.18% females; M_age = 18.61, SD_age = 1.12) were recruited from a subject pool of a university located in the eastern U.S.¹ There were 76.36% White, 10.55% Asian, 5.82% Hispanic, 4.36% African American, and 2.91% other participants. They were assigned to the GAD group if they met full diagnostic criteria on the Generalized Anxiety Disorder Questionnaire-IV (GAD-Q-IV; Newman et al., 2002) and scored below 14 on the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996). They were included in the depression group if they scored 20 or above on the BDI-II, and did not meet diagnostic criteria on the GAD-Q-IV. They were included as HCs if they neither met diagnostic criteria on the GAD-Q-IV nor scored 14 or above on the BDI-II. This resulted in 91 GAD participants, 91 depressed, and 93 HCs. Among them, 90 participants received the worry induction, 90 ruminated, and 95 received the relaxation induction.

2.3. Screening measures

2.3.1. Generalized Anxiety Disorder Questionnaire-IV (GAD-Q-IV; Newman et al., 2002)

The GAD-Q-IV is a nine-item self-report measure that assesses GAD symptoms based on diagnostic criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV and DSM-5; American Psychiatric Association, 2013; American Psychiatric Association, 1994). Diagnosis can be determined using a dimensional cutoff or based on whether individuals met full diagnostic criteria. We used criterion scoring because one study found that compared to the dimensional scoring system, criterion scoring led to the best sensitivity (89%) and specificity (82%; Moore, Anderson, Barnes, Haigh, & Fresco, 2014). Internal consistency in the current sample was good (Cronbach’s α = .86).

2.3.2. Beck Depression Inventory-II (BDI-II; Beck et al., 1996)

The BDI-II is a 21-item self-report measure assessing symptoms of major depressive disorder. It showed good convergent and discriminant validity (Beck et al., 1996; Steer, Ball, & Ranieri, 1999) and high retest reliability (Beck, Steer, & Garbin, 1988). A cutoff score of 18 showed high sensitivity (94%) and specificity (92%), with a high correct classification rate (92%; Arnau, Meagher, Norris, & Bramson, 2001). We used a slightly more rigorous cut score as suggested by the original development study (20 or above for moderate to severe depression and 13 or less for minimal depression; Beck et al., 1996) to screen participants. Internal consistency was good in the current sample (Cronbach’s α = .92).

2.3.3. Subjective Emotion Scales and Manipulation Check measures

To measure subjective fear, sadness, and amusement, we used three items from the emotion scales developed by Gross and Levenson (1995). For the manipulation check of inductions, we used three items measuring the intensity of worry, rumination, and relaxation. Each of these items was rated on a 9-point Likert scale ranging from 0 (not at all) to 8 (extremely). Internal consistency with these six items was good (Cronbach’s α = .84).

2.4. Procedure

Consenting participants were randomly assigned to either worry, ruminate, or relax stratified by group. They then completed a brief demographic questionnaire and a 5-minute resting baseline which allowed them to become acclimated to the research setting. They subsequently completed baseline measures of subjective emotion. Then, they engaged with worry, rumination, or relaxation (described below). Following this, they completed emotion measures and the manipulation check. This was followed by the amusement video exposure. Next, participants once again rated their emotional states.²

2.4.1. Worry, rumination, and relaxation inductions

For worry and rumination inductions, we adapted the worry induction method developed by Borkovec and Inz (1990), which directed participants to engage with personally relevant worrisome scenarios. Following Borkovec et al. (1983) and Nolen-Hoeksema (1991), worry was defined as “a chain of uncontrollable thoughts and images and doubts about things that might happen in the future” and rumination was defined as “passively and repetitively thinking about possible causes, implications, and consequences of past stressful events and negative feelings as opposed to its solutions.” Before the experiment, definitions of worry and rumination were explained to participants including examples adopted from previous studies³ (e.g., Nolen-Hoeksema et al., 2008; Stöber & Joormann, 2001; Tallis, Davey, & Bond, 1994; Trapnell & Campbell, 1999). Throughout the study, definitions of worry and rumination were displayed on a separate sheet of paper as a reference. Based on these definitions, participants wrote five scenarios that made them the most worried or ruminative. Then, they engaged with each of those scenarios for one minute (total of five minutes) and rated their levels of worry, rumination, and relaxation on a 9-point Likert scale ranging from 0 (not at all) to 8 (extremely). Scenarios were retained if the target cognitive process was higher than 4 and at least 3 points higher than the non-target cognitive process.⁴ The same standard was applied to temporal orientation.⁵ The pre-selected most worrisome or ruminative scenario was presented before the induction task to ensure that participants engaged with the correct cognitive process. For the relaxation induction, participants learned diaphragmatic breathing and five muscle group progressive-muscle relaxation before the experiment. They later engaged with relaxation during the induction (Bernstein & Borkovec, 1973; Borkovec & Costello, 1993).

2.4.2. Amusement video exposure

For the amusement exposure, we used a video clip validated by Gross and Levenson (1995), which was a fake orgasm scene from “When Harry Met Sally.”

2.5. Analytic plan

Analyses were performed using SPSS (IBM SPSS Version 25.0). Descriptive statistics and baseline diagnostic measures were analyzed via two one-way MANOVAs. In each model, either group or induction was included as an independent variable, and age, GAD symptom counts, continuous GAD-Q-IV scores, BDI-II scores, and baseline fear, sadness, and amusement were included as dependent variables. Differences in gender across the groups and induction conditions were tested by the Chi-square test. For the manipulation check of the inductions, we ran a two-way MANOVA examining the main effect of induction type and the interaction between induction type and group on levels of worry, rumination, and relaxation. For the amusement video manipulation check, a paired samples t-test was implemented by separately comparing amusement scores to fear and sadness scores. This was followed by a one-way ANOVA testing group differences in elicited amusement. All post hoc analyses for ANOVA or MANOVA were implemented via Tukey’s HSD with p < .05. To test the main hypotheses, we conducted random intercept modeling examining the two-way interaction between induction and time, and group and time (i.e., induction × time; group × time) and the three-way interaction between induction, group, and time (i.e., induction × group × time) at each of two time trends.⁶ Analyses included the intercept (i.e., either fear, sadness or amusement) at each time trend as random effects (Level 1: within-individual variance/random intercept; Level 2: between-individual variance/differences across the inductions and groups over time). All model parameters were estimated by restricted maximum likelihood with variance components. Following significant results, we conducted simple slope analyses and slope comparison t-tests as suggested by Howell (2012). For each of three emotions (i.e., fear, sadness, and amusement), these processes were repeated twice for two different time trends (a total of six multilevel models). The induction and group variables were coded multicategorically.⁷ For each time trend, the earlier time point was coded as zero and the latter was coded as one.

3. Results

3.1. Power analysis

A Monte Carlo simulation estimated the power of the current sample (R package SIMR; Green & MacLeod, 2016). We conducted 1,000 simulations with a target effect size of Cohen’s d = 0.5 for all possible interaction terms. The smallest power was 85.20%, 95% CI [82.85, 87.34], indicating that the current sample size was sufficient for planned analyses.

3.2. Descriptive statistics and baseline scores

3.2.1. Differences across groups

The overall effect of group on descriptive statistics and baseline scores was significant, F(14, 528) = 127.56, p < .001, Wilk’s Λ = .05, d = 3.68. Specifically, there was no effect of group on age. As expected, there were significant group differences in GAD symptom counts and dimensional GAD-Q-IV scores. The GAD group reported greater GAD symptom counts and dimensional scores than the depression group (p < .001) and HCs (p < .001). The depression group also had greater GAD symptom counts and dimensional scores than HCs (p < .001). There were also significant group differences in baseline BDI-II scores. The depression group had higher BDI-II scores than the GAD group (p < .001) and HCs (p < .001). The GAD group also had higher BDI-II scores than HCs (p < .001). There were significant group differences in baseline fear and sadness. Fear at baseline in the GAD (p = .040) and depression group (p = .027) was higher than fear in HCs. The GAD and depression groups were not different in their fear levels (p = .987). Similarly, sadness in the depression (p < .001) and GAD groups (p = .023) was greater than sadness in HCs. There was no difference between the depression and GAD groups in baseline sadness (p = .138). Amusement did not differ across the groups. Results from Chi-square analysis indicated a non-significant group difference in the gender of the participants, χ² (2, N = 275) = 1.44, p = .487 (See Table 1 for descriptive statistics).

Table 1.

Descriptive statistics and baseline emotions.

Category	GAD Group (N = 91) M (SD)	Depression Group (N = 91) M (SD)	HCs (N = 93) M (SD)	F	p	d
Age	18.78 (1.30)	18.64 (1.27)	18.41 (.66)	2.60	.076	0.28
Gender (Female/Male)	75/16	69/22	71/22	-	-	-
GAD-Q-IV (Sx. counts)	9.20 (1.88)	5.17 (3.61)	1.32 (1.94)	118.43 ***	< .001	1.87
GAD-Q-IV (Continuous)	9.32 (1.29)	5.06 (3.01)	1.32 (1.26)	359.73 ***	< .001	3.26
BDI-II	8.52 (3.03)	25.52 (6.47)	4.23 (3.67)	541.12 ***	< .001	4.03
Baseline Fear	.48 (1.08)	.50 (1.08)	.13 (.68)	4.28 *	.015	0.36
Baseline Sadness	.59 (1.15)	.90 (1.41)	.16 (.56)	10.55 ***	< .001	0.56
Baseline Amusement	.98 (1.76)	.84 (1.55)	1.06 (1.61)	.42	.661	0.11

Note. GAD-Q-IV (Sx. counts) = Number of symptoms on the Generalized Anxiety Disorder Questionnaire for DSM-IV; GAD-Q-IV (Continuous) = Dimensional score of Generalized Anxiety Disorder Questionnaire for DSM-IV (In this study, the categorical standard was used to screen participants); BDI-II = Beck Depression Inventory-II; GAD Group = participants who met full diagnostic criteria on the GAD-Q-IV and scored below 14 on the BDI-II; Depression Group = participants who scored 20 or above on the BDI-II and did not meet diagnostic criteria on the GAD-Q-IV; HCs = participants who neither met diagnostic criteria on the GAD-Q-IV nor scored 14 or above on the BDI-II;

^*p < .05,

^**p < .01,

^***p < .001 (two-tailed);

Statistically significant results are bolded.

3.2.2. Differences across induction conditions

Overall differences between inductions on age, GAD symptom counts, dimensional GAD-Q-IV scores, BDI-II scores, and emotions were not significant, F(14, 528) = .74, p = .738, Wilk’s Λ = .96, d = 0.28. Specifically, there were nonsignificant differences in age, F(2, 270) = .26, p = .770, d = 0.09, GAD symptom counts, F(2, 270) = .41, p = .667, d = 0.11, dimensional GAD-Q-IV scores, F(2, 270) = 1.08, p = .340, d = 0.18, BDI-II scores, F(2, 270) = .49, p = .615, d = 0.12, baseline fear, F(2, 270) = .18, p = .837, d = 0.06, sadness, F(2, 270) = .82, p = .440, d = 0.16, and amusement scores, F(2, 270) = 1.06, p = .347, d = 0.18. Results also indicated non-significant gender differences across the induction conditions, χ² (2, N = 275) = 3.74, p = .154.

3.3. Manipulation check

There was an overall effect of induction, F(6, 522) = 100.43, p < .001, Wilk’s Λ = .22, d = 2.15, and no induction by group interaction, F (12, 779) = .93, p = .518, Wilk’s Λ = .96, d = 0.24.

3.3.1. Manipulation-check of worry induction

There was an effect of the worry induction on levels of worry, F(2, 263) = 112.52, p < .001, d = 1.85. It led to more worry (M = 4.32, SD = 1.88) than rumination (M = 2.81, SD = 2.08; p < .001) and relaxation (M =.52, SD = 1.12; p < .001) inductions. The rumination induction also led to more worry than the relaxation induction (p < .001). In addition, there was no interaction with group, F(4, 263) = 1.55, p = .190, d = 0.31, indicating the worry manipulation was successful across groups.

3.3.2. Manipulation-check of rumination induction

There was a significant main effect of induction on rumination levels, F(2, 263) = 174.50, p < .001, d = 2.30. The rumination induction led to more rumination (M = 5.29, SD = 1.96) than the worry (M = 2.78, SD = 2.18) (p < .001) and relaxation (M =.45, SD =.97; p < .001) inductions. The worry induction also induced more rumination than the relaxation induction (p < .001). There was no induction by group interaction effect, F(4, 263) = .28, p = .889, d = 0.13, showing that the manipulation of rumination was effective across all groups.

3.3.3. Manipulation-check of relaxation induction

There was an effect of induction on relaxation levels, F(2, 263) = 141.95, p < .001, d = 2.08. The relaxation induction elicited more relaxation (M = 6.29, SD = 1.95) than worry (M = 2.04, SD = 1.70; p < .001) and rumination inductions (M = 2.18, SD = 2.12; p < .001). The worry and rumination inductions were not significantly different from one another (p = .889). In addition, there was no induction by group interaction, F(4, 263) = .53, p = .717, d = 0.18, indicating the manipulation of relaxation was successful across all groups.

3.3.4. Manipulation-check of amusement video exposure

The amusement video elicited more amusement (M = 5.38, SD = 2.14) than fear (M =.18, SD =.71), t(271) = 35.83, p < .001, d = 4.35, and sadness (M =.17, SD =.68), t(271) = 38.22, p < .001, d = 4.64. There was no difference between fear and sadness scores, t(271) = .21, p = .834, d = 0.03, indicating that the amusement video had its intended effect.

3.4. Analysis of fear

3.4.1. Fear score from baseline to induction time trend (T1)

There was a significant time-by-induction interaction at the first time trend, F(2, 263) = 55.51, p < .001, d = 1.30. There was no significant two-way, F(2, 263) = 1.47, p = .232, d = 0.21 or three-way interaction with group, F(4, 263) = 1.72, p = .146, d = 0.32. Fear increased from baseline to worry and from baseline to rumination. Fear did not significantly change from baseline to relaxation (see Table 2). From baseline to worry, fear increased more than from baseline to rumination, t(356) = 4.24, p < .001, d = 0.45, and more than from baseline to relaxation, t(360) = 9.22, p < .001, d = 0.97. Fear also increased more from baseline to rumination than from baseline to relaxation, t(360) = 4.39, p < .001, d = 0.46 (see Fig. 1).

Table 2.

Means of fear, sadness, amusement scores at baseline, induction, and amusement video exposure, and simple slopes by the two-way interaction between time and induction from baseline to induction (T1) and induction to amusement video exposure (T2) time trends.

Emotion Type	Induction Type	Baseline M (SD)	Induction M (SD)	Amusement Exposure M (SD)	Baseline to Induction (T1)				Induction to Exposure (T2)
					β	t	p	d	β	t	p	d
Fear	Worry	.32 (1.02)	2.89 (2.37)	.23 (.85)	.58	9.45 ***	< .001	1.41	−.60	−10.02 ***	< .001	1.50
	Rumination	.37 (.83)	1.40 (2.10)	.14 (.63)	.31	4.34 ***	< .001	0.65	−.38	−5.43 ***	< .001	0.81
	Relaxation	.25 (.74)	.15 (.59)	.16 (.65)	−.07	−.99	.322	0.15	.01	.12	.906	0.02
Sadness	Worry	.54 (1.04)	2.67 (2.42)	.21 (.73)	.50	7.65 ***	< .001	1.14	−.57	−9.23 ***	< .001	1.38
	Rumination	.66 (1.32)	3.68 (2.56)	.26 (.88)	.60	9.96 ***	< .001	1.49	−.67	−11.99 ***	< .001	1.79
	Relaxation	.49 (1.15)	.18 (.68)	.04 (.25)	−.16	−2.18	.030	0.32	−.14	−1.87	.063	0.28
Amusement	Worry	1.12 (1.75)	.17 (.66)	5.19 (2.25)	−.34	−4.86 ***	< .001	0.73	.84	20.30 ***	< .001	3.03
	Rumination	1.00 (1.63)	.17 (.57)	5.37 (2.12)	−.33	−4.59 ***	< .001	0.69	.86	22.46 ***	< .001	3.36
	Relaxation	.86 (1.38)	.82 (1.34)	5.59 (2.04)	−.02	−.22	.829	0.03	.81	18.72 ***	< .001	2.77

Note.

^***p < .001 (two-tailed);

Statistically significant results are bolded.

Fig. 1.

Fear and sadness from baseline to induction and induction to amusement video exposure.

3.4.2. Fear scores from induction to video exposure time trend (T2)

There was a time-by-induction interaction at the second time trend, F (2, 263) = 50.05, p < .001, d = 1.23. There were no two-way, F(2, 263) = 2.42, p = .27, d = 0.36 or three-way interactions with group, F(4, 263) = 2.08, p = .084, d = 0.36. Fear decreased from both worry and rumination to the amusement video, and did not significantly change from relaxation to amusement video (see Table 2). From worry to the amusement video, fear decreased more than from rumination to the amusement video, t(356) = −3.98, p < .001, d = 0.42, and more than from relaxation to the amusement video, t(360) = −9.51, p < .001, d = 1.00. Fear also decreased more from rumination to the amusement video than from relaxation to the amusement video, t(360) = −5.09, p < .001, d = 0.54. (see Fig. 1).

3.5. Analysis of sadness

3.5.1. Sadness scores from the baseline to induction time trend (T1)

There was a significant time-by-induction interaction at the first time trend, F(2, 263) = 69.09, p < .001, d = 1.45. There was no significant two-way, F(2, 263) = 1.08, p = .340 d = 0.18 or three-way interaction with group, F(4, 263) = 2.14, p = .076, d = 0.36. Sadness increased from baseline to worry and rumination inductions. However, sadness decreased from baseline to relaxation (see Table 2 for simple slopes). Sadness increased more from baseline to rumination than from baseline to worry, t(356) = −2.19, p = .029, d = 0.23, and more than from baseline to relaxation, t(360) = 9.95, p < .001, d = 1.05. Sadness also increased more from baseline to worry than from baseline to relaxation, t(360) = 7.83, p < .001, d = 0.83. (see Fig. 1).

3.5.2. Sadness scores from induction to video exposure time trend (T2)

There was an overall time-by-induction interaction at the second time trend, F(2, 263) = 72.98, p < .001, d = 1.49, and time-by-group interaction, F(2, 263) = 6.85, p = .001, d = 0.46. There was no three-way interaction, F(4, 263) = 2.20, p = .070, d = 0.37.

Sadness decreased from both rumination and worry to amusement video. There was no significant change in sadness from relaxation to amusement video (see Table 2). From rumination to amusement video, sadness decreased more than from worry to amusement video, t(356) = 2.48, p = .014, d = 0.26, and more than from relaxation to amusement video, t(360) = −8.38, p < .001, d = 0.88. Sadness also decreased more from worry to amusement video than from relaxation to amusement video, t(360) = −11.13, p < .001, d = 1.17 (see Fig. 1).

There was also a significant decrease in sadness across all groups from induction to amusement video, β = −.46, t(181) = −6.99, p < .001, d = 1.04; β = −.53, t(173) = −8.29, p < .001, d = 1.26; β = −.44, t(187) = −6.73, p < .001, d = 0.98. However, the MDD group decreased more than HCs, t(358) = −2.59, p = .010, d = 0.27. The GAD group decrease was non-significantly in-between the MDD, t(352) = 1.42, p = .158, d = 0.15, and HCs decrease, t(366) = −1.12, p = .265, d = 0.12. To understand these findings, we examined group differences at the beginning and ending points of the sadness slope. There was a significant group difference only at the beginning point (i.e., during the inductions), F(2, 269) = 5.42, p = .005, d = 0.40, and not during the amusement video, F(2, 269) = 2.84, p = .060, d = 0.29. On average, during the inductions, the depression group (M = 2.80, SD = 2.73) had significantly greater sadness than HCs (M = 1.59, SD = 2.21; p = .003). Sadness in the GAD group (M = 2.14, SD = 2.52) was non-significantly in-between the depression group (p = .181) and HCs (p = .282).

3.6. Analysis of amusement

3.6.1. Amusement scores from baseline to induction time trend (T1)

There was a time-by-induction interaction on amusement, F(2, 263) = 9.57, p < .001, d = 0.54. There were neither two-way, F(2, 263) = .711, p = .492, d = 0.15 nor three-way interactions with group, F(4, 263) = .492, p = .742, d = 0.17. Amusement decreased from baseline to worry and rumination but did not significantly change from baseline to relaxation (see Table 2). From baseline to worry, t(360) = −3.24, p = .001, d = 0.34, and from baseline to rumination, t(360) = −2.91, p = .004, d = 0.31, amusement decreased more than from baseline to relaxation. Amusement slopes from baseline to worry and from baseline to rumination were not significantly different from each other, t(356) = −.46, p = .647, d = 0.05. (see Fig. 2).

Fig. 2.

Amusement from baseline to induction to amusement video exposure.

3.6.2. Amusement scores from induction to video exposure time trend (T2)

At the second time trend, there were no significant two-way interactions between time and induction, F(2, 263) = .89, p = 412, d = 0.16, and time and group, F(2, 263) = 1.00, p = 368, d = 0.17. There was a significant three-way interaction, F(4, 263) = 2.91, p = .022, d = 0.42. There was a significant increase in amusement across all groups and all induction conditions (see Table 3). However, the GAD group in the worry condition had a greater increase in amusement than the depression group and HCs in the worry condition. The increase in amusement in the depression group and HCs in the worry condition were not significantly different from each other. We did not find any effect of group on change in amusement from rumination and relaxation inductions to the amusement video (see Table 4 and Fig. 2).

Table 3.

Means of amusement scores at induction and amusement video exposure, and simple slopes for the three-way interaction between time, induction, and group from induction to amusement video exposure (T2) time trend.

Induction Type	Group	Induction M (SD)	Amusement Exposure M (SD)	Induction to Exposure (T2)
				β	t	p	d
Worry	GAD	.17 (.59)	6.07 (1.86)	0.91	16.59 ***	< .001	4.32
	Depression	.26 (.86)	4.67 (2.24)	0.80	9.56 ***	< .001	2.63
	HCs	.09 (.52)	4.82 (2.42)	0.81	10.98 ***	< .001	2.72
Rumination	GAD	.26 (.68)	5.13 (2.17)	0.84	11.92 ***	< .001	3.05
	Depression	.07 (.26)	5.83 (2.14)	0.89	14.39 ***	< .001	3.81
	HCs	.17 (.65)	5.17 (2.05)	0.86	12.72 ***	< .001	3.31
Relaxation	GAD	1.07 (1.36)	5.47 (2.18)	0.78	9.38 ***	< .001	2.44
	Depression	.35 (.71)	5.77 (2.32)	0.85	12.44 ***	< .001	3.19
	HCs	1.03 (1.68)	5.52 (1.63)	0.81	10.66 ***	< .001	2.73

Note. GAD = participants who met full diagnostic criteria on the GAD-Q-IV and scored below 14 on the BDI-II; Depression = participants who scored 20 or above on the BDI-II and did not meet diagnostic criteria on the GAD-Q-IV; HCs = participants who neither met diagnostic criteria on the GAD-Q-IV nor scored 14 or above on the BD-II;

^***p < .001 (two-tailed);

Statistically significant results are bolded.

Table 4.

Simple slope comparison by manipulation type and group on amusement slopes from induction to amusement video exposure (T2) time trend.

Dependent Variable	Time Trend	Induction Type	GAD Group vs. Depression Group				GAD Group vs. HCs				Depression Group vs. HCs
			t	df	p	d	t	df	p	d	t	df	p	d
Amusement	Induction to Amusement Exposure	Worry	2.56 *	110	.012	0.49	2.10 *	122	.038	0.38	−.51	116	.613	0.29
		Rumination	−1.55	116	.123	0.29	−.23	118	.820	0.04	1.35	114	.179	0.25
		Relaxation	−1.59	118	.114	0.29	−.13	118	.894	0.02	1.54	120	.126	0.28

Note. GAD = participants who met full diagnostic criteria on the GAD-Q-IV and scored below 14 on the BDI-II; Depression = participants who scored 20 or above on the BDI-II and did not meet diagnostic criteria on the GAD-Q-IV; HCs = participants who neither met diagnostic criteria on the GAD-Q-IV nor scored 14 or above on the BD-II;

^*p < .05 (two-tailed).

Statistically significant results are bolded.

3.6.3. Follow-up analyses on the interaction between the worry induction and GAD from induction to video exposure time trend (T2)

To determine what contributed to the three-way interaction at T2, we examined differences at the beginning (i.e., induction phase) and ending points (i.e., amusement video exposure phase) of the amusement slopes in the worry condition. At the beginning point (i.e., during the worry induction), there was no significant group difference, F(2, 87) = .48, p = .620, d = 0.20. The group difference was significant only at the ending point, F(2, 87) = 3.65, p = .030, d = 0.59. Specifically, the GAD group was significantly more amused than the depression group (p = .047) during the amusement video. There was no significant difference between the GAD group and HCs (p = .067), and the depression group and HCs (p = .962) (see Fig. 2).

4. Discussion

A tenet of CAM is that worry increases negative emotion and decreases positive emotion. Another tenet of CAM is that because of the impact of worry on negative and positive emotions, it enhances negative and positive affect PECs. Such PECs may reinforce ongoing engagement with worry and exacerbate symptoms of anxiety. Previous studies found evidence supporting the effect of worry on positive emotions (Llera & Newman, 2014) and the enhancement of a PEC involving reduced negative and increased positive emotion via worry (Llera & Newman, 2014; Newman et al., 2019, 2022). However, no study to date examined these phenomena concerning rumination, which is another form of repetitive negative thought. Furthermore, it was unclear whether the enhancement of PECs occurred transdiagnostically in those with GAD, depression, and non-anxious/non-depressed controls. In addition, although a previous study (Kim & Newman, 2022) found emotional specificity between rumination and worry in NECs regardless of group, this has remained untested in PECs. To this end, we experimentally manipulated worry, rumination, and relaxation before watching an amusement video clip. We recruited individuals with pure GAD, pure depression symptoms, and non-GAD/non-depressed healthy controls to test whether these effects were transdiagnostic or specific to certain groups.

Results largely supported our hypotheses. In terms of change from baseline, regardless of group, both worry and rumination led to increased fear and sadness (Hypothesis 1a) and decreased amusement (Hypothesis 3a). However, relaxation did not increase fear or decrease amusement, and it decreased sadness from baseline. Importantly, and as predicted, both worry and rumination increased fear and sadness and decreased amusement to a greater extent than did relaxation (Hypothesis 1a). These findings extend the CAM notion that worry dampens positive emotion to rumination. Also consistent with the second hypothesis, for all participants, worry increased fear more than rumination, and rumination increased sadness more than worry (Hypothesis 2a). Therefore, as predicted, there was some emotional specificity in the effects of worry and rumination.

As also predicted in our first hypothesis, both prior worry and prior rumination (vs. relaxation) led to a sharper decrease in negative emotion during the amusement video regardless of group (Hypothesis 1b). This suggests a causal role of both worry and rumination in enhancing PECs transdiagnostically via a decline in negative emotions in response to a positive event, extending previous research on worry to rumination (Llera & Newman, 2014; Newman et al., 2022). For example, Llera and Newman (2014) found an experimental effect of prior worry (vs. relaxation and a neutral induction) on decreased fear, sadness, tension, and anger during a funny video. Also, Newman et al. (2022) found that those who worried (vs. those who did not worry) prior to a social interaction evidenced decreased negative affect from before to after that interaction.

We also found support for our emotional specificity hypothesis (Hypothesis 2b). In particular, because there was induction specificity regarding the degree to which fear and sadness increased (from baseline to induction), there was specificity regarding the degree to which these emotions decreased (from induction to amusement video exposure). Similar to findings from a study on NECs (Kim & Newman, 2022), worry increased fear from baseline more than rumination. As a result, those who worried had the most heightened PECs regarding decreased fear in response to the amusement video. Also, because rumination increased sadness more than worry, the decrease in sadness in response to the amusement video was most salient in those who had ruminated compared to other inductions across all groups. Thus, PECs entailed emotional specificity of the degree of negative reinforcement received. As with prior NEC studies, these effects occurred regardless of diagnostic status. However, prior studies also found that those with depression or GAD preferred to worry or ruminate to avoid a negative contrast and increase a positive contrast more than non-anxious/non-depressed controls (Jamil & Llera, 2021; Llera & Newman, 2017; Newman, Rackoff, Zhu, & Kim, 2023). Taken together, this may explain why rumination has been more strongly linked to depression and why worry is more relevant to GAD. Whereas depression is more of a disorder of chronic sadness, GAD is more closely associated with chronic anxiety. Thus, those with depression might be more motivated to ruminate to increase the degree of experiencing a PEC related to a reduction in their chronic sadness. However, those with GAD might be more motivated to worry to experience a more salient PEC in their chronic anxiety.

Analysis of amusement demonstrated a more complicated picture than we predicted in our third hypothesis (Hypothesis 3b). Although both worry and rumination reduced amusement from baseline more than relaxation regardless of group, amusement was increased to the same degree across all induction conditions in response to the amusement video. This was consistent with two prior experimental studies that did not find greater enhancement of happiness or amusement from funny videos in those who had worried compared to prior relaxation and neutral inductions (Llera & Newman, 2014, 2010). However, unlike those studies, we found that in those with GAD who worried, there was a stronger amusement PEC from induction to video exposure than HCs and the depression group who engaged with worry. This finding was not due to lower amusement in the GAD group in response to the worry induction. Instead, it was due to their enhanced amusement response to the amusement video, which was more heightened than the amusement of the depression group and HCs who worried. This is consistent with one prior EMA study that found that prior worry (vs. not worrying) led to increased positive affect from before to after social interactions in a sample of individuals with GAD (Newman et al., 2022). This provides important insights into why individuals with GAD prefer worrying. The more saliently positive affect PECs are experienced, the more likely the person’s positive beliefs regarding worry are reinforced. In addition, effects of worry and rumination on negative affect PECs regardless of group was contrasted with the positive affect PEC that was only exhibited in the GAD group who worried. This may indicate that positive and negative affect PECs are qualitatively different processes. A sharp decrease in negative emotion in response to an unexpected positive event can serve as negative reinforcement (e.g., relief). On the other hand, a sudden increase in positive emotion can be conceptualized as positive reinforcement (e.g., reward). Future research should directly test different motives associated with the two PECs,

We also had another unexpected finding. Although sadness decreased from induction to amusement video across all groups, the decline of sadness in the depression group was steeper than that in HCs regardless of the induction type. Also, the depression group’s high sadness during the induction contributed to their steeper decrease. The depression group’s consistently greater sadness implies that they may have greater proneness to maintain sadness. In previous studies, no matter what emotion-regulation strategy they engaged with, depressed participants (compared to non-depressed) were more likely to maintain their sadness, and they preferred to feel sad and remain low in energy (Millgram, Joormann, Huppert, & Tamir, 2015; Newman, Rackoff, Zhu, & Kim, 2023; Yoon, Verona, Schlauch, Schneider, & Rottenberg, 2020). At the same time, findings that the depression group also showed the steepest decline in sadness in response to the amusement video suggests the possibility that such sustained sadness is motivated by increasing the probability of a steeper negative affect PEC. This is consistent with another paper on self-reported contrast avoidance in depressed participants (Newman, Rackoff, Zhu, & Kim, 2023). Future research should continue to examine what motivates depressed individuals to maintain sadness.

It is of note that two prior studies (Llera & Newman, 2014, 2010) neither found decreased amusement from baseline to worry, nor a GAD-specific effect of prior worry on subsequently increased amusement during an amusement video exposure. This may be due to differences in methodology. Our study had a much larger sample of GAD participants (n = 91 vs. 38–48) giving us more power to detect differences. Moreover, the discrepancy might be due to differing worry induction methods used in the current and previous studies. Previous studies asked participants to think about their most worrisome topic and worry about it as intensively as they could during both a practice period and the experiment. However, in our study, participants first wrote down all of their current worry or ruminative topics and practiced each one for one minute and only those that met certain intensity criteria and were highest on their target emotion were used to ensure that their topic was charged with as intense thoughts as possible throughout the induction tasks. These differences might have resulted in significant decreases in amusement from worry and rumination in our study and may have allowed us to detect the GAD group’s more robust effect on the enhancement of an amusement PEC via worry. Another methodological difference that could account for differential outcomes was different video clips. Whereas the amusement video from the prior studies was a black and white candy factory scene from “I Love Lucy”, the video in the current study was a fake orgasm scene from “When Harry Met Sally.”

Our study has clinical implications. In particular, intervention might target ongoing monitoring and greater awareness of the cost associated with maintaining negative mood from repetitive negative thinking compared to the brevity of the experience of a PEC. For example, in participants with GAD, a 10-day intervention that monitored distress and interference associated with worry, as well as the ultimate outcome of worry, showed greater reduced worry at posttreatment and follow-up than a thought log alone (e.g., LaFreniere & Newman, 2016). Also, even though those high in contrast avoidance endorsed sustained negative mood to enhance the likelihood of a positive emotional contrast, they reported that they were uncomfortable engaging with a relaxed or euthymic state due to the increased likelihood of an NEC (Llera & Newman, 2017; Newman, Rackoff, Zhu, & Kim, 2023). Thus, positive emotions are not savored. In fact, those with MDD and GAD report they are more inclined to dampen their positive moods, compared to persons who are less depressed or anxious (Bryant, 2003; Eisner, Johnson, & Carver, 2009; Feldman, Joormann, & Johnson, 2008; Min’er & Dejun, 2001). Furthermore, negative contrast sensitivity mediated the effects of both depression and GAD symptoms on relaxation-induced anxiety (Kim & Newman, 2019). Therefore, savoring and mindfulness interventions focused on sustaining positive and euthymic moods could be helpful. Such intervention led to significantly greater reductions in worry, dampening of positive emotions, and depression symptoms, as well as greater increases in positive emotions, savoring, prioritizing positivity, and optimism compared to active control (LaFreniere & Newman, 2023a, 2023b). Furthermore, cognitive restructuring targeting positive beliefs about PECs may be a viable intervention as well.

The present study was not without limitations. Although our focus was on fear, sadness, and amusement, there are many other types of emotions that were not measured. Examining positive emotions with lower physiological arousal, such as euthymia or contentment, might allow a more comprehensive understanding of the mechanism of PECs. In addition, we recruited college students using self-report measures and they were predominantly White and women. Therefore, it would be important to replicate this study among treatment-seeking populations with more diverse backgrounds.

In summary, our study demonstrated that PECs were enhanced by both worry and rumination. Although PECs may feel momentarily good, relying on worry and rumination is not a healthy emotional coping strategy because worry and rumination make people feel miserable for an extended time and this is at the core of their mental health problems. Thus, future research should continue to explore the effects of worry and rumination on emotional contrasts and identify effective treatment modalities.

Abbreviations:

PEC

positive emotional contrast

CAM

contrast avoidance model

GAD

generalized anxiety disorder

MDD

major depressive disorder

HCs

healthy controls

NEC

negative emotional contrast

EMA

ecological momentary assessment

negative affect PEC

sharp decrease in negative emotion

positive affect PEC

sharp increase in positive emotion