Avoidance of negative emotional contrast from worry and rumination: An application of the Contrast Avoidance Model

Abstract

According to the Contrast Avoidance model (CAM), worry causes increased and sustained negative affect and such negative affect enables avoidance of a future sharp increase in negative emotion. However, only pathological worriers (vs. controls) view worry as a positive coping strategy to avoid a negative emotional contrast (NEC). We examined if rumination, which is another type of repetitive negative thought, would function similarly. Individuals with self-reported symptoms of pure generalized anxiety disorder (GAD; n = 90), pure depression (MDD; n = 85), and non-anxious/non-depressed controls (HC; n = 93) were randomly assigned to conditions where they were asked to worry, ruminate, or relax. Emotional and physiological changes were measured during worry and subsequent exposure to fearful and sad videos. We also assessed participant group differences in preference for worry or rumination as a strategy to cope with negative affect during the negative emotional video exposures. Consistent with CAM, regardless of the group, both worry and rumination enabled avoidance of NEC. Whereas worry led to greater avoidance of a fear contrast, rumination led to greater avoidance of a sadness contrast. On the other hand, relaxation enhanced NEC. Skin conductance also indicated patterns in line with CAM. In the subjectively perceived preference, the GAD group reported a greater preference for worry in coping with a fear contrast than HC. However, such a salient pattern was not found for the MDD group. Treatment implications of these findings are discussed.

The contrast avoidance model (CAM; Newman & Llera, 2011) suggests that worry increases and sustains anxiety, and this increased anxiety enables subsequent avoidance of a sharp negative emotional contrast. Consequently, worry is reinforced by preventing sudden negative emotional contrasts (NEC; see Newman & Llera, 2011 for a full review). Although CAM suggests that the avoidance of NEC is a phenomenon that occurs among individuals both with and without generalized anxiety disorder (GAD), the model also posits that individuals with GAD feel more uncomfortable with NEC than non-anxious individuals, and this makes them prefer to worry compared to being in a euthymic state.

An accumulation of evidence has supported this model. In laboratory studies, individuals who engaged in worry demonstrated a greater increase in anxiety from baseline during worry (vs. relaxation), and such increased anxiety precluded further sharp increases in response to negative emotional film clips (Llera & Newman, 2014, 2010, 2017). Such findings were replicated in ecological momentary assessment studies (Crouch, Lewis, Erickson, & Newman, 2017; Newman et al., 2019). In addition, whereas worry led to similar levels of heightened anxiety and subsequent contrast avoidance in those with and without GAD (Llera & Newman, 2014, 2010), those with GAD perceived worry to be more helpful in coping with NEC and were more comfortable worrying than being in a relaxed state (Llera & Newman, 2014, 2017).

Even though CAM was initially developed as a theory focused on worry in those with GAD, it has been posited that worry and depressive rumination might be transdiagnostic constructs wherein the same emotion regulation mechanism operates (Watkins, Moulds, & Mackintosh, 2005). For example, studies have shown that rumination, like worry, increases negative affect and physiological activation (McLaughlin, Borkovec, & Sibrava, 2007; Moberly & Watkins, 2008; Nolen-Hoeksema & Morrow, 1993; Ottaviani et al., 2016). When rumination was compared to worry, the two types of mentation styles were discriminated only by the type of subjective negative affect they induced (McLaughlin et al., 2007). Whereas worry led to increased anxious affect, rumination led to increased depressed affect. Furthermore, both those with GAD (Borkovec & Roemer, 1995) and those with MDD (Papageorgiou & Wells, 2001; Watkins & Moulds, 2005) were found to have positive beliefs about worry or rumination respectively.

In fact, there is evidence that negative contrast avoidance may operate for rumination. In a recent study (Jamil & Llera, 2021), rumination as well as worry increased negative emotionality and reduced an emotional contrast in response to negative feedback on an intelligence test. The study also found that GAD (vs. nonGAD) and MDD (vs. nonMDD) groups reported worry or rumination respectively, to be more helpful in coping with the negative feedback. However, there were several issues with this study. Unlike the current study, analyses of the effects of worry or rumination on subjective and physiological contrast avoidance did not examine diagnostic group as an independent variable. Instead, it examined worry, rumination, and relaxation. Also, for analyses of contrast avoidance coping, there were only seven participants who were both nonanxious and nondepressed so comparison groups were not nonpathological (i.e., depressed persons formed 77% of the nonGAD comparison group and GAD individuals formed 67% of those in the nondepressed group). In addition, 36% of the GAD and 40% of the MDD group were the same participants who met criteria for both GAD and MDD. Yet data from these comorbid participants were included in one analysis of the independent effects of GAD and a separate analysis of the effects of MDD with no statistical control for comorbidity. Therefore, this study did not provide a fair test of whether contrast avoidance was specific to either GAD or MDD. In addition, the study did not test which specific emotion led to their effects. Considering that worry and rumination induce different types of emotions - anxiety and fear by worry and sadness by rumination (Borelli, Hilt, West, Weekes, & Gonzalez, 2014; Llera & Newman, 2014; McLaughlin et al., 2007; Zetsche, Ehring, & Ehlers, 2009), emotional contrast may operate differently depending on the types of emotions they create. This can be optimally tested by implementing a standardized emotion elicitation task, and measuring specific emotions separately.

The current study attempted to test CAM in a laboratory study that manipulated worry, rumination, and relaxation by considering emotional specificity. Also, the current study attempted to compare diagnostic group differences - pure GAD, pure MDD, and non-anxious/non-depressed controls (HC) in their perception of the utility of worry and rumination in contrast avoidance. We made a number of predictions.

First, both worry and rumination (vs. relaxation) would lead to greater negative affect and greater avoidance of NEC during the subsequent video exposure. Specifically, worry would enable avoidance of a fear contrast, whereas rumination would enable avoidance of a sadness contrast. However, relaxation would not increase subjective negative emotion and would therefore lead to a sharper upward surge in negative affect during the video exposure. We did not expect to see significant diagnostic group differences.

Second, regardless of diagnostic status, both worry and rumination would lead to a greater NS-SCR amplitude from resting baseline and prevent a sharp surge during the video exposure. On the other hand, relaxation would enhance a negative contrast by decreasing NS-SCR amplitude and then cause a sharp upward surge during the video exposure.

Lastly, we predicted that both GAD and MDD groups would report a greater preference for worry and rumination in helping them cope with a negative emotional contrast.

Examining whether negative contrast avoidance is a common underlying mechanism for both worry and rumination might benefit the development of improved treatments for repetitive negative thinking (RNT) that target contrast avoidance. Although CBT for RNT is more efficacious than non-CBT interventions (Covin, Ouimet, Seeds, & Dozois, 2008; Hanrahan, Field, Jones, & Davey, 2013; Spinhoven et al., 2018), there is little evidence that newer RNT treatments are superior to traditional CBT (Covin et al., 2008; Hanrahan et al., 2013; Spinhoven et al., 2018). Furthermore, meta-analyses showed that half of patients failed to benefit from CBT for worry (Covin et al., 2008; Hanrahan et al., 2013). Therefore, there is still substantial room to improve the efficacy of CBT for RNT.

Method

Experimental Design

This study had a three-by-three factorial design that tested both main and interaction effects between group (i.e., GAD vs. MDD vs. HC) and emotion induction (i.e., worry vs. rumination vs. relaxation) on emotional and psychophysiological responses during the negative emotional exposure (i.e., fear and sadness videos).

Participants

A total of 268 participants (216 females and 52 males; age: M = 18.60, SD = 1.55) were recruited from a university subject pool based on the Generalized Anxiety Disorder Questionnaire-IV (GAD-Q-IV; Newman et al., 2002) and Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996). Participants were included in the GAD group if they met full diagnostic criteria on the GAD-Q-IV and scored less than 14 on the BDI-II. Those in the MDD group were those who scored 20 or above on the BDI-II and did not meet diagnostic criteria on the GAD-Q-IV. Participants who met neither of the standards (i.e., not meeting the full diagnostic criteria on the GAD-Q-IV and scored less than 14 on the BDI-II) were assigned to HC. In total, 90 GAD participants, 85 MDD participants, and 93 non-anxious/non-depressed controls were recruited. Among them, 90 participants were assigned to the worry induction condition, 90 were allocated to the rumination induction and 88 individuals were assigned to the relaxation condition. Participants’ ethnic distribution was 73.88% White, 11.94% Asian, 6.34% Hispanic, 4.85% African American, and 2.99% others.

Measures

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

The GAD-Q-IV consists of nine items measuring the full diagnostic criteria of GAD as defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV and DSM-5; American Psychiatric Association, 2013; American Psychiatric Association, 1994). The measure showed high specificity (96%) and sensitivity (67%) (Newman et al., 2002).

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

The BDI-II tests the presence of depressive symptoms. A recent meta-analytic review (Wang & Gorenstein, 2013) found high internal consistency among college students (Cronbach’s α = .93) and outpatients (Cronbach’s α = .92). The measure also has good convergent and discriminant validity (Beck et al., 1996; Steer, Ball, & Ranieri, 1999).

Subjective Emotion Scale and Manipulation Check

Participants were asked to complete a self-report inventory developed by Gross and Levenson (1995), directing them to circle the number that best describes the greatest amount of each emotion felt at any time during the preceding task. We modified this scale by adding three manipulation check items measuring worry, rumination, and relaxation.

Contrast Avoidance Questionnaire-Experimental Assessment (CAQ-E; Llera & Newman, 2014)

We administered the CAQ-E (Llera & Newman, 2014) to measure the perceived helpfulness of worry and rumination in coping with the negative video exposures. The internal consistency of the CAQ-E in the current study (Cronbach’s α = 76) was slightly higher than the previous study (Cronbach’s α = 73) (Llera & Newman, 2014).

Physiological Measure

We measured non-specific skin conductance (NS-SCR) amplitude as a marker of sympathetic activity. Data were recorded at 1,000 Hz through the Biopac MP150. (Biopac Systems, Inc., Goleta, CA). Recorded waveforms were down-sampled (i.e., 50 samples/seconds) and smoothened, and artifacts were filtered using a low-pass finite impulse response filter. As outlined in Kim, Bang, and Kim (2004), SCRs lower than 10% of the maximum amplitude were excluded from each participant’s data. Since SCR amplitude is leptokurtic, we log-transformed the data as suggested by Braithwaite, Watson, Jones, and Rowe (2015).

Experimental Manipulation

Worry, Rumination and Relaxation Inductions

Based on the results of a pilot study, we adopted the induction method developed by Borkovec and Inz (1990) for worry and modified it for the rumination induction. Prior to the experiment, participants were provided definitions of worry and rumination and wrote five scenarios that might make them feel charged with the most intense worry or rumination. After this, participants practiced thinking about each scenario for one minute (i.e., five minutes for five scenarios), and rated the degree to which they felt worried, ruminative, and relaxed from each scenario on a 9-point Likert scale ranging from 0 (not at all) to 8 (extremely). In order to be eligible for the emotion manipulation, their target emotion had to be higher than 4 on the rating scale and at least 3 points greater than their non-target emotions. Temporal orientation was another screening criterion. For the worry induction, temporal orientation had to be inclined to the future (higher than 4 on the scale) whereas the temporal orientation of rumination was required to be past-oriented (less than 4 on the scale). Using these pre-screened scenarios, participants were asked to engage with their most intense worry or rumination scenario during the induction period. For the relaxation induction, participants were asked to engage with diaphragmatic breathing and five muscle group progressive-muscle relaxation (Bernstein & Borkovec, 1973; Borkovec & Costello, 1993).

Emotion Inducing Film Clips

We used two film clips which were validated by Gross and Levenson (1995). For the fear exposure, participants watched a video clip from “The Shining (Kubrick, 1980).” For the sadness exposure, a scene from “The Champ (Lovell & Zeffirelli, 1979)” was used.

Procedure

Consenting participants were randomly assigned to induction conditions stratified by group. Before beginning the experiment, participants washed their hands with non-abrasive liquid soap and were seated in a comfortable chair which was placed three feet away from a high-definition computer display (i.e., Dell 23-inch computer monitor with 1,920 × 1,080 resolution). Then, their non-dominant hand was fitted with a skin conductance monitoring device. After this, participants practiced their assigned induction task. For those assigned to either the worry or rumination induction condition, their worrisome or ruminative scenarios were screened based on the level of the target emotion and temporal orientation. The experiment began with a 5-minute acclimation period, and this was followed by ratings of emotion. Then participants engaged with the worry, rumination, or relaxation induction. For the worry and rumination inductions, participants were asked to engage with personally relevant scenarios which were screened before the experiment. For the relaxation induction, participants engaged with a combination of two different applied relaxation techniques (i.e., diaphragmatic breathing and five muscle group progressive-muscle relaxation; Bernstein & Borkovec, 1973; Borkovec & Costello, 1993). Then participants rated their emotions and completed the manipulation check. Following this, participants watched one of the two videos (either fear or sad video clip) in counterbalanced order. Subsequently, they completed emotion ratings and the CAQ-E. This process was repeated until participants had watched both videos. All instructions and stimuli were programmed using E-prime software (Schneider, Eschman, & Zuccolotto, 2002).

Analytic Plan

ANOVA was conducted on descriptive statistics, baseline scores, and manipulation checks. For the main hypotheses, we ran multilevel modeling to examine the main and interaction effects of the inductions and groups over time. This was followed by a series of simple slope analyses and slope comparison tests as outlined by Howell (2012). For the test of perceived emotion coping, generalized linear modeling was conducted. Post-hoc tests were conducted via LSD with p < .05.

Results

Power Analysis

A power analysis via Monte Carlo simulation (SIMR; Green & MacLeod, 2016) with a target effect size of Cohen’s d = 0.50 demonstrated that the minimum estimated power was 96.30%, 95% CI [94.94, 97.38], indicating that the number of participants was sufficient.

Descriptive Statistics and Baseline Scores

Descriptive Statistics

There was no significant difference between the GAD, MDD, and HC groups in age. As expected, the screening measures used in the current study were effective in forming the three participant groups (see Table 1).

Table 1.

Descriptive statistics.

Dependent Variable	GAD group (N = 90) M (SD)	MDD group (N = 85) M (SD)	Control group (N = 93) M (SD)	F	p	d
Age	18.82 (1.30)	18.89 (2.09)	18.45 (.68)	2.35	.098	0.29
GAD-Q-IV (Continuous)	9.22 (1.25)	5.37 (3.00)	1.23 (1.03)	391.66 ***	<.001	3.37
BDI-II	8.70 (2.92)	25.34 (6.64)	3.90 (3.58)	531.69 ***	<.001	4.00

GAD-Q-IV, Dimensional Score of Generalized Anxiety Disorder Questionnaire for DSM-IV (in this study, the categorical questionnaire was used as for the screening measure); BDI-II, Beck Depression Inventory-II;

^***p < .001 (two-tailed); statistically significant results are bolded with p < .05.

Baseline Emotion and NS-SCR Amplitude Scores

Both the GAD and MDD groups reported higher baseline fear scores than HC. The difference between the GAD and MDD groups was not significant. A group difference was also found in baseline sadness. The MDD group exhibited higher sadness than the GAD and HC groups. In addition, the GAD group had higher sadness scores than HC. On the other hand, baseline NS-SCR amplitude did not differ across the three groups. (see Table 2).

Table 2.

Baseline emotion and NS-SCR amplitudes by participant group.

Dependent Variable	GAD group (N = 90) M (SD)	MDD group (N = 85) M (SD)	Control group (N = 93) M (SD)	F	p	d
Fear	0.44 (0.69)	0.55 (0.97)	0.13 (0.34)	8.79 ***	<.001	0.51
Sadness	0.53 (0.86)	0.94 (1.11)	0.15 (0.43)	19.54 ***	<.001	0.77
NS-SCR	0.19 (0.30)	0.23 (0.36)	0.24 (0.41)	.21	.811	0.09

NS-SCR amplitudes are reported in non-transformed values;

^***p < .001 (two-tailed); statistically significant results are bolded with p < .05.

Manipulation Check

Manipulation Check of Worry, Rumination, and Relaxation Inductions

There were significant differences between the inductions in the degree to which they led to increased worry, F(2, 265) = 198.10, p < .001, d = 2.45, rumination, F(2, 265) = 218.36, p < .001, d = 2.57 or relaxation, F(2, 265) = 226.54, p < .001, d = 2.62. Post hoc tests showed that those in the worry induction reported higher worry (M = 4.48, SD = 1.38) than those in the rumination (M = 2.70, SD = 1.78), and relaxation conditions (M = 0.38, SD = 0.77). Similarly, rumination scores were higher among individuals in the rumination induction (M = 5.26, SD = 1.76) than those in the worry (M = 2.89, SD = 1.87), and relaxation inductions (M = 0.39, SD = 0.77). In the relaxation induction, relaxation scores (M = 6.40, SD = 1.42) were higher than in the worry (M = 1.95, SD = 1.36), and rumination conditions (M = 2.19, SD = 1.87).

Manipulation Check of Video Exposure

Each video exposure also effectively elicited its target emotion, Fear, F(2, 801) = 280.71, p < .001, d = 1.67; Sadness, F(2, 801) = 972.35, p < .001, d = 3.13. Post hoc comparison indicated that fear scores were higher in response to the fear video (M = 3.74, SD = 2.25) than the sad video (M = 1.55, SD = 1.93). Sadness scores were greater in response to the sad video (M = 5.09, SD = 1.99) than the fear video (M = 0.60, SD = 1.31).

Main Data Analysis Results

Subjective Emotion

Fear.

At the first time trend (T1: baseline to induction), there was a significant time-by-induction interaction, _{T1 × induction} F(2, 259) = 41.65, p < .001, d = 1.12. Fear increased from baseline to worry, and from baseline to rumination but did not change from baseline to relaxation (see Table 3). Slope comparisons indicated that worry led to significantly greater fear increase than rumination and relaxation, and rumination led to greater fear increase than relaxation (see Table 4). Group did not exhibit any significant two or three-way interaction effects with time and induction, _{T1 × group} F(2, 259) = .66, p = .518, d = 0.14 _{T1 × induction × group} F(4, 259) = 1.78, p = .134, d = 0.33.

Table 3.

Means of dependent variables at baseline, induction, and exposure, and simple slopes of fixed effects by induction for baseline to induction and induction to exposure time trends predicting subjective emotions.

Emotion	Induction type	Baseline	Induction	Exposure	Baseline to Induction (T1)				Induction to Exposure (T2)
		M (SD)	M (SD)	M (SD)	Slope	t	p	d	Slope	t	p	d
Fear	Worry	.50 (1.27)	2.97 (2.32)	4.02 (2.11)	.55	8.86 ***	<.001	1.32	.23	3.20 **	.002	0.48
	Rumination	.38 (1.00)	1.43 (1.95)	3.90 (2.38)	.32	4.57 ***	<.001	0.68	.50	7.62 ***	<.001	1.14
	Relaxation	.13 (.50)	.09 (.42)	3.30 (2.21)	−.04	−.49	.624	0.07	.71	13.34 ***	<.001	2.01
Sadness	Worry	.32 (.95)	2.11 (2.15)	5.29 (1.89)	.48	7.23 ***	<.001	1.08	.62	10.55 ***	<.001	1.57
	Rumination	.72 (1.48)	3.44 (2.55)	5.34 (2.08)	.55	8.76 ***	<.001	1.31	.38	5.48 **	<.001	0.82
	Relaxation	.39 (1.04)	.22 (.70)	4.62 (1.94)	−.10	−1.27	.206	0.19	.84	20.02 ***	<.001	3.02

Fear was measured in the fear video exposure condition; Sadness was measured in the sadness video exposure condition;

^**p < .01,

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

statistically significant results are bolded with p < .05.

Table 4.

Simple slope comparison by induction type (subjective emotions).

Dependent Variable	Slope	Worry vs. Rumination				Worry vs. Relaxation				Rumination vs. Relaxation
		t	df	p	d	t	df	p	d	t	df	p	d
Fear	Baseline to induction (T1)	3.90 ***	356	<.001	0.41	8.70 ***	352	<.001	0.92	4.52 ***	352	<.001	0.48
	Induction to exposure (T2)	3.05 **	356	.002	0.32	5.27 ***	352	<.001	0.56	1.83	352	.068	0.19
Sadness	Baseline to induction (T1)	2.35 *	356	.019	0.25	6.97 ***	352	<.001	0.74	8.54 ***	352	<.001	0.90
	Induction to exposure (T2)	2.78 **	356	.005	0.29	3.30 **	352	.001	0.35	6.11 ***	352	<.001	0.65

Fear was measured in the fear video exposure condition; Sadness was measured in the sadness video exposure condition;

^*p < .05,

^**p < .01,

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

statistically significant results are bolded with p < .05.

At the second time trend (induction to fear video exposure), the time-by-induction interaction was significant, _{T2 × induction} F(2, 259) = 17.08, p < .001, d = 0.74. There was an increase in fear following all three inductions (worry, rumination or relaxation; see Table 3). However, slope comparisons indicated that there was a greater increase in fear from induction to the fear video following rumination and relaxation inductions than following worry with no significant difference between rumination and relaxation (see Table 4). There were no two or three-way interactions with the group, _{T2 × group} F(2, 259) = .58, p = .559, d = 0.13 _{T2 × induction × group} F(4, 259) = .95, p = .437, d = 0.24. These trends can be found in Fig. 1.

Fig. 1.

Subjective emotion trends from baseline to induction to exposure.

Sadness.

At the first time trend, there was also a significant time-by-induction interaction for sadness, _{T1 × induction} F(2, 259) = 54.35, p < .001, d = 1.31. Sadness increased in response to worry and rumination, but not in response to relaxation (see Table 3). Slope comparison demonstrated that rumination yielded a greater increase in sadness than worry and relaxation, and worry increased sadness more than relaxation did. (see Table 4). There were no two or three-way interactions with the group, _{T1 × group} F(2, 259) = .52, p = .596, d = 0.13 _{T1 × induction × group} F(4, 259) = 1.71, p = .148, d = 0.33.

At the second time trend, the time-by-induction interaction was also significant, _{T2 × induction} F(2, 259) = 25.44, p < .001, d = 0.87. The increase in sadness during the sad video was significant following all of the three inductions including worry, rumination, and relaxation (see Table 3). However, the results from slope comparison showed that the increase in sadness was less following rumination than following worry and relaxation. Similarly, the increase in sadness was less following worry than relaxation (see Table 4). There were no two or three-way interactions with the group, _{T2 × group} F(2, 259) = 1.46, p = .235, d = 0.21 _{T2 × induction × group} F(4, 259) = 1.28, p = .277, d = 0.28. Fig. 1 depicts these trends.

Skin Conductance (NS-SCR Amplitude)

Fear Exposure Condition.

At the first time trend, the time-by-induction interaction was significant, _{T1 × induction} F(2, 222.01) = 5.29, p = .006, d = 0.44. There was an increase in NS-SCR amplitude from baseline to worry and rumination, but not from baseline to relaxation (see Table 5). NS-SCR amplitude changes in worry and rumination were greater than in relaxation with no significant differences between worry and rumination (see Table 6). Two or three-way interactions with group were not significant, _{T1 × group} F(2, 221.77) = .72, p = .486, d = 0.16 _{T1 × induction × group} F(4, 221.70) = .41, p = .804, d = 0.17.

Table 5.

Mean NS-SCR amplitudes at baseline, induction, and exposure, and simple slopes of fixed effects for baseline to induction and induction to exposure time trends predicting NS-SCR amplitudes.

Exposure	Induction type	Baseline M (SD)	Induction M (SD)	Exposure M (SD)	Baseline to Induction				Induction to Exposure
Condition		M(SD)	M(SD)	M(SD)	Slope	t	p	d	Slope	t	p	d
Fear	Worry	.20 (.28)	.28 (.28)	.24 (.30)	.56	4.17 ***	<.001	0.67	−.16	−1.73	.086	0.27
Exposure	Rumination	.19 (.29)	.40 (.56)	.26 (.32)	.50	3.85 ***	<.001	0.08	−.28	−2.38*	.019	0.37
	Relaxation	.26 (.38)	.32 (.42)	.39 (.50)	−.01	−.05	.958	0.01	.36	2.42 *	.017	0.37
Sadness	Worry	.19 (.29)	.28 (.37)	.22 (.30)	.57	3.52 **	.001	0.58	−.11	−.92	.360	0.15
Exposure	Rumination	.22 (.35)	.35 (.43)	.20 (.24)	.55	3.86 ***	<.001	0.61	−.31	−2.48*	.014	0.39
	Relaxation	.22 (.30)	.35 (.68)	.31 (.40)	−.04	−.23	.818	0.04	.39	2.19 *	.030	0.34

Means and standard deviations are reported in non-log-transformed values;

^*p < .05,

^**p < .01,

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

statistically significant results are bolded with p < .05.

Table 6.

Simple slope comparison by induction type (NS-SCR amplitudes).

Exposure	Slope	Worry vs. Rumination				Worry vs. Relaxation				Rumination vs. Relaxation
Condition		t	df	p	d	t	df	p	d	t	df	p	d
Fear	Baseline to induction (T1)	.36	301	.720	0.04	2.68 **	308	.008	0.31	2.41 *	305	.017	0.28
Exposure	Induction to exposure (T2)	.78	320	.435	0.09	−2.97**	329	.003	−0.33	−3.37**	329	.001	−0.37
Sadness	Baseline to induction (T1)	.06	304	.948	0.01	2.46 *	304	.014	0.28	2.52 *	316	.012	0.28
Exposure	Induction to exposure (T2)	1.13	314	.260	0.13	−2.32*	320	.021	−0.26	−3.21**	322	.001	−0.36

Fear was measured in the fear video exposure condition; sadness was measured in the sadness video exposure condition;

^*p < .05,

^**p < .01 (two-tailed);

Statistically significant results are bolded with p < .05.

At the second time trend, the time-by-induction interaction was also significant, _{T2 × induction} F(2, 240.11) = 10.86, p < .001, d = 0.60. The increase of NS-SCR amplitude from induction to the fear video exposure was significant following relaxation, but there was no significant change following worry and rumination (see Table 5). The fear slope was steeper following relaxation compared to worry and rumination, with no difference between worry and rumination (see Table 6). We did not find any two or three-way interactions with the group, _{T2 × group} F(2, 239.90) = 2.24, p = .109, d = 0.27 _{T2 × induction × group} F(4, 239.85) = .55, p = .698, d = 0.19. Fig. 2 shows these trends.

Fig. 2.

NS-SCR amplitude from baseline to induction to exposure.

Sadness Exposure Condition.

At the first time-trend, there was a significant time-by-induction interaction, _{T1 × induction} F(2, 224.42) = 4.51, p = .012, d = 0.40. There was an increase in the NS-SCR amplitude from baseline to worry and rumination, but not from baseline to relaxation (see Table 5). Both worry and rumination slopes were greater than the relaxation slope with no difference between worry and rumination (Table 6). Neither two-way nor three-way interaction with the group was found, _{T1 × group} F(2, 224.22) = .090, p = .914, d = 0.06 _{T1 × induction × group} F(4, 224.05) = .48, p = .750, d = 0.19.

The time-by-induction interaction at the second time-trend was also significant, _{T2 × induction} F(2, 228.23) = 7.50, p = .001, d = 0.51. The sad video exposure led to an increase in the NS-SCR amplitude following relaxation. However, prior rumination led to decreased NS-SCR amplitude upon video exposure. There was no significant change following worry (see Table 5). Slope comparison demonstrated that relaxation led to a greater increase than worry and rumination with no significant difference between worry and rumination (see Table 6). We did not find any two or three-way interactions with the group. _{T2 × group} F(2, 228.11) = .83, p = .437, d = 0.14 _{T2 × induction × group} F(4, 228.05) = 1.50, p = .204, d = 0.32. These trends can be found in Fig. 2.

Contrast Avoidance Questionnaire

Fear Coping Scores.

There was a significant two-way interaction between induction and group, _{induction × group} F(4, 255) = 2.56, p = .039, d = 0.41 in predicting fear coping scores. We found a significant between-group difference only in the worry induction. Those with GAD who worried before the fear video (M = 8.69, SD = 1.06) reported that worry was more effective in helping them cope with the fear video than HC (M = 5.73, SD = 0.99). The MDD group (M = 8.46, SD = 1.12) who worried before the fear video exposure were not significantly different from either the GAD group or HC. These results are demonstrated in Fig. 3.

Fig. 3.

Fear and sadness negative emotional contrast coping scores by induction type and group (CAQ-E).

Sadness Coping Scores.

Unlike the fear coping scores, the results of sadness coping only indicated significant main effects of induction, _induction F(2, 255) = 6.75, p < .001, d = .46. Specifically, all three participant groups reported that relaxation (M = 9.42, SD = 0.59) and rumination (M = 8.56, SD = 0.59) were more helpful than worry (M = 6.46, SD = 0.59) in coping with sadness elicited. The difference between relaxation and rumination was not significant. Fig. 3 demonstrates these trends.

Discussion

There has been scarce literature that tested the emotion dysregulation processes of worry and rumination based on the framework of CAM (Newman & Llera, 2011). In order to improve upon this, the current study compared the subjective and physiological effects of worry, rumination, and relaxation in relation to subsequent fear and sadness exposure in those with pure GAD, pure MDD, and HC.

Results supported our first hypothesis. Regardless of group, both worry and rumination led to a greater increase in sad and anxious affect from baseline and lesser NEC in sad and anxious affect in response to the sad and fearful video exposure compared to relaxation. At the same time, our findings demonstrated that there was emotional specificity in the contrast avoidance effects of worry and rumination. Although both RNT types led to contrast avoidance of both types of negative affect, the extent to which each RNT led to contrast avoidance was different depending on the type of emotion they elicited most strongly from baseline (i.e., Worry led to greater contrast avoidance of fear than rumination did, and rumination yielded greater contrast avoidance of sadness than worry did). These results are in line with previous findings that worry evoked more fear than sadness (Llera & Newman, 2014; McLaughlin et al., 2007), whereas rumination evoked more sadness than anxiety (Borelli et al., 2014; McLaughlin et al., 2007; Zetsche et al., 2009).

Results from the NS-SCR amplitude analyses also supported our second prediction. Both worry and rumination (vs. relaxation) increased NS-SCR amplitude and prevented a further sharp increase during the fearful and sad video exposure. On the other hand, relaxation amplified the negative contrast by lowering NS-SCR amplitude from baseline and yielding a sharp surge during the video exposure. Thus, the effect of contrast avoidance was not limited to subjective emotions but was also manifested in physiological reactivity. Taken together, results on subjective and physiological contrast avoidance replicate and extend findings by Jamil and Llera (2021).

As we predicted, regardless of group, our relaxation induction also led to significantly greater relaxation (evidenced by subjective and NS-SCR measures) than the worry and rumination inductions. In addition, relaxation did not increase fear or sadness from baseline, and effects of relaxation did not differ by diagnostic group. As a result, relaxation led to a subsequent sharper surge in negative affect than did worry or rumination during the fear and sadness video exposure. This suggests that although relaxation induced anxiety (RIA) could have been present in a subset of the participants—typically reported to be 15% of highly anxious populations (Luberto, McLeish, & Kallen, 2020), such an increase was not salient enough to invalidate the effectiveness of the relaxation manipulation tested here. It is also possible that the combination of two different relaxation induction methods in our study was more effective than the single relaxation practice in previous studies (Heide & Borkovec, 1983; Luberto et al., 2020).

We also examined subjective appraisals of worry and rumination as strategies to cope with negative emotional videos. Results revealed a more complex picture than we expected. In fear coping, only individuals with GAD differed from HC in their report of worry as more helpful. Those with MDD fell nonsignificantly between the GAD and HC. These results suggest that those with GAD may be more uncomfortable with a fear contrast, and they use chronic worry as a coping mechanism. This was consistent with the results of a previous study showing a significant group difference between those with GAD and non-GAD controls in their perceptions of the utility of worry in negative contrast avoidance (Llera & Newman, 2014). This suggestion is also in line with self-report findings showing that compared to non-anxious controls, those with GAD were more likely to report engaging with worry specifically to increase and sustain their negative mood to avoid NEC (Llera & Newman, 2017). Considering that the MDD group did not demonstrate such a clear preference in fear coping, this positive appraisal could be more unique to those with GAD.

On the other hand, the analysis of sadness coping scores demonstrated that regardless of group, all participants reported rumination and relaxation were significantly more helpful than worry at enabling sadness contrast coping. These results were inconsistent with the emotional and psychophysiological responses measured across the experiment in all groups. As noted above, both rumination and worry were more effective at enabling subjective and physiological contrast avoidance of sadness compared to relaxation. This inconsistency is difficult to apprehend. One possible interpretation is that as opposed to those with GAD seeing worry as uniquely more helpful than HC at enabling contrast avoidance of anxiety, those with depression were not unique in their view of rumination as helpful in avoiding contrast of sadness. Although our results on negative contrast coping differ from the findings of Jamil and Llera (2021), which demonstrated a preference for worry in GAD (vs. nonGAD) and a preference for rumination in MDD (vs. nonMDD) in coping with negative feedback, these authors neither compared pure GAD or pure MDD to purely nonanxious non-depressed controls nor examined contrast avoidance of sadness or fear specifically. Therefore, Jamil and Llera (2021) could not inform whether individuals with MDD had a unique preference for rumination for sad contrast coping. Another possible interpretation is that the simultaneous effect of rumination on increased fear, physiological arousal, and sadness in this study was less favorable than sadness alone to individuals with MDD regardless of their significant contrast avoidance effect. Recent findings suggest that individuals with MDD prefer to maintain their sadness and they select emotion regulation strategies that could facilitate the maintenance of their sadness (Millgram, Huppert, & Tamir, 2020; Millgram, Joormann, Huppert, & Tamir, 2015). In addition, those with depression tend to prefer soothing sadness to sadness with high arousal (Yoon, Verona, Schlauch, Schneider, & Rottenberg, 2020). As evidenced in the current study, the sadness induced by the rumination manipulation amplified NS-SCR amplitudes and subjective fear beyond the resting baseline. Therefore, it is possible that the combination of emotions and arousal induced in the current study could have been perceived less favorably by the depressed individuals throughout the experiment. Consequently, this could have resulted in the non-significant group differences in the sadness coping scores. On the other hand, the results of our study demonstrate that individuals with GAD do not seem to dislike highly arousing fear, considering their significantly greater preference for worry in coping with fear contrast.

Our findings have several potential clinical implications. Patients who have positive beliefs about the benefits of RNT in avoiding NEC and who actively engage with RNT to avoid NEC might benefit from a cognitive intervention targeting the short-sightedness of their positive perceptions. Considering our findings demonstrating a preference for worry (vs. relaxation) in those with GAD to enable avoidance of fear contrast, this approach may be more effective for individuals with GAD. Such an intervention could point out that although worry does enable avoidance of negative contrast, this comes at the expense of sustained heightened fear or anxiety. In addition, a behavioral intervention would be crucial to change the negative reinforcement of contrast avoidance. This intervention might be implemented by having patients repeatedly exposed to two consecutive situations that elicited negative contrast effects (e.g., relaxation training immediately followed by exposure to negative imagery). For this intervention, the type of emotion should be selected by considering the emotional specificity of RNT. As we found in the current study, repeated exposure to fear contrast may be more effective for individuals with excessive worry whereas repeated exposure to sadness contrast could be implemented for those suffering from excessive rumination.

Several limitations should be mentioned. GAD and MDD are highly comorbid disorders. Therefore, in a naturalistic setting, it would not be uncommon to have the two conditions occurring at the same time. Although the current study distinguished the two diagnostic groups for a cleaner assessment of the effects of each diagnostic group on negative contrast avoidance, recruiting comorbid samples could help illuminate whether having both disorders was associated with perceptions of worry and rumination as more helpful in coping with contrast avoidance than GAD or MDD alone. Furthermore, studies could examine naturalistic effects of worry and rumination and perceptions of these thought styles as coping strategies within a comorbid group as these unfold over time. There was also limited diversity of our samples. Participants in the current study were college students whose age and education levels were highly homogeneous. We also did not specifically target treatment-seeking samples. Thus, this study should be replicated among treatment-seeking populations with more diverse demographic backgrounds. Also, future research could include different disorders such as obsessive-compulsive disorder and posttraumatic stress disorder, considering their close association with repetitive negative thinking processes (Ehring & Watkins, 2008; McEvoy, Watson, Watkins, & Nathan, 2013). Moreover, future research could consider the inclusion of different types of emotions that were not measured in the current study. In addition to fear and sadness, a close relationship has been reported between depression, GAD, and anger (Deschênes, Dugas, Fracalanza, & Koerner, 2012; Winkler, Pjrek, & Kasper, 2005). Findings also suggest a high correlation between symptoms of obsessive-compulsive disorder and disgust (Shapira et al., 2003). The inclusion of different emotional disorders and their related emotions would help to better understand the transdiagnostic nature and emotional specificity of CAM.

In summary, the current study showed that rumination, as well as worry, play a causal role in the avoidance of NEC but they have different emotional specificity. Although the avoidance of NEC can give a sense of relief at the moment, this is not a healthy emotion regulation strategy because it requires individuals to sustain their negative affect, and this is what deteriorates their mental health. We hope our findings contribute to future research on the mechanism of RNT and help to establish a more effective intervention for individuals who are suffering from excessive worry and rumination.