Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Oct 1.
Published in final edited form as: Emotion. 2019 Jun 27;20(7):1234–1243. doi: 10.1037/emo0000627

Emotion differentiation moderates the effects of rumination on depression: A longitudinal study

Daphne Y Liu 1, Kirsten E Gilbert 1, Renee J Thompson 1
PMCID: PMC6933110  NIHMSID: NIHMS1030298  PMID: 31246044

Abstract

Elevated trait rumination is associated with and predicts onset of major depressive disorder, but not all people with elevated trait rumination develop major depressive disorder. We hypothesize that certain emotional processes weaken the rumination-depression link, protecting against increases in depression. In this prospective longitudinal study, we examined one such process, emotion differentiation—the ability to discern specific emotions. Because higher negative emotion differentiation (NED) facilitates down-regulating negative emotions and the content of rumination tends to be negative, we predicted that NED, but not positive emotion differentiation (PED), would moderate the rumination-depression association, such that rumination would only predict increases in depression when negative emotions are less, not more, differentiated. Over one week of experience sampling, 65 community-dwelling adults (M = 38.4 years) repeatedly reported their emotions, from which we computed NED and PED. Participants completed self-report measures of rumination and depression at baseline and a measure of depression six months later. Regression analyses suggested that the combination of NED and PED, but not a unique contribution of either NED or PED, interacted with rumination to predict significant changes in depression, after controlling for mean emotion. Specifically, rumination predicted significant increases in depression when emotion differentiation was lower, but not higher. Findings demonstrate longitudinal benefits of emotion differentiation in adults and suggest emotion differentiation as a promising avenue for studying major depressive disorder.

Keywords: emotion differentiation, depression, rumination, experience sampling, longitudinal

Have you ever caught yourself dwelling on an embarrassing situation or an unpleasant interaction? This type of thinking is referred to as rumination, the process of repetitively focusing on the nature, causes, and consequences of one’s problems and feelings (Nolen-Hoeksema, 1991). Rumination was first proposed by Nolen-Hoeksema as a dispositional tendency to respond to depression, in which individuals “repetitively focus[] on the fact that one is depressed; on one’s symptoms of depression; and on the causes, meaning, and consequences of depressive symptoms” (Nolen-Hoeksema, 1991, p. 569). Therefore, rumination is often referred to as depressive rumination.1

Rumination has been linked to various negative psychological outcomes, particularly depression (Nolen-Hoeksema, 1991; Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Ruminative tendencies are positively associated with depressive symptoms and the development of major depressive disorder, both cross-sectionally and longitudinally (Treynor, Gonzalez, & Nolen-Hoeksema, 2003; Watkins, 2008). Theory and research suggest that rumination exacerbates negative thinking, interferes with problem solving, and prevents constructive behaviors (Nolen-Hoeksema et al., 2008). In particular, the brooding component of rumination—passive perseveration on one’s depression and unachieved standards—has been theoretically conceptualized and empirically demonstrated to be the most maladaptive component of rumination and is consistently related to depression (Armey et al., 2009; Nolen-Hoeksema et al., 2008; Treynor et al., 2003).

Despite trait rumination being implicated in the etiology of major depressive disorder, people with high trait rumination do not always develop major depressive disorder. We hypothesize that certain psychological processes may interrupt the mechanisms underlying the rumination-depression link and weaken this association. We propose that high emotion differentiation is one such process. Emotion differentiation refers to one’s ability to discern specific emotions (Barrett, Gross, Christensen, & Benvenuto, 2001). People high in emotion differentiation make more fine-grained distinctions between emotions of different types and use different and specific emotion words to characterize how they feel (e.g., sad versus frustrated). For high differentiators, by virtue of their high ability to distinguish specific emotions, changes of various specific emotions from moment to moment tend to be independent of other each (e.g., feel sadder but as frustrated). People low in emotion differentiation are less able to discern the nuances across different emotions and characterize their emotions using the same set of emotional words across multiple occasions. For low differentiators, changes of various specific emotions from one occasion to another tend to be correlated with each other (e.g., feel sadder and more frustrated).

From the feelings-as-information prospective (Schwarz, 1990), emotion is a critical source of information. Higher emotion differentiation provides more fine-grained information about the situation, which may facilitate emotion regulation (Gross, 2015; Kashdan, Barrett, & McKnight, 2015). When people experience more (versus less) differentiated intense negative emotions, they engage in more frequent emotion regulation strategies (Barrett et al., 2001). A keen awareness of specific emotions may also allow individuals to evaluate the link between a triggering event and emotional reaction at a meta-cognitive level, serving as valuable guidance for selecting appropriate and context-specific emotion regulation strategies (Gross, 2015). Moreover, identifying and labeling specific negative emotions serves a direct regulatory function to reduce negative emotion (Kircanski, Lieberman, & Craske, 2012). High differentiators may find it easier to clearly label how they feel, resulting in less intense negative emotions. Taken together, we posit these adaptive qualities of high emotion differentiation prevent individuals from becoming stuck in the ruminative cycle and may weaken the association between rumination and increases in depression.

Emotion differentiation is frequently examined separately for negative and positive emotion. Low negative emotion differentiation (NED) is associated with various negative psychological outcomes, including major depressive disorder (Anand, Chen, Lindquist, & Daughters, 2017; Demiralp et al., 2012; Kashdan, Ferssizidis, Collins, & Muraven, 2010; Pond et al., 2012; Zaki, Coifman, Rafaeli, Berenson, & Downey, 2013). Among individuals with borderline personality disorder and low (but not high) NED, trait rumination is associated with nonsuicidal self-injury (NSSI) urges and acts (Zaki et al., 2013), demonstrating that the negative effect of rumination is specific to low NED. Starr, Hershenberg, Li, and Shaw (2017) also found rumination examined at both the momentary and daily level to be more strongly associated with within-person variations of depressive symptoms when NED was low versus high in college student and veteran samples. These findings indicate that individuals with low NED are more vulnerable to behavioral and psychological costs of rumination, possibly because they are less skillful at effectively down-regulating negative affect relative to those with high NED. Research is still needed to examine longer-term implications of NED on depression.

Compared to NED, positive emotion differentiation (PED) has received less empirical attention and demonstrated less consistent implications for well-being. Of the 34 published empirical studies examining emotion differentiation, 16 assessed PED and wellbeing, among which only six demonstrated adaptive features of high PED with the rest reporting null findings (n = 8) or demonstrating maladaptive features (n = 2) of high PED.2 In terms of adaptive outcomes associated with PED, for example, high (versus low) PED predicts less avoidance or impulsive coping in response to stress (Tugade, Fredrickson, & Barrett, 2004) and protects against impulses for maladaptive behaviors related to borderline personality disorder pathology (Dixon-Gordon, Chapman, Weiss, & Rosenthal, 2014). In terms of maladaptive outcomes associated with PED, high (versus low) PED has also been associated with reduced salutary effect of savoring positive experiences on depressive symptoms (Starr et al., 2017), possibly because experiencing positive emotions in a differentiated manner could diminish and narrow positive emotional experiences (Dixon-Gordon et al., 2014; Starr et al., 2017). Given these equivocal findings, it is important that researchers continue to elucidate the associations between PED and wellbeing, including examining their associations over time.

Building from this literature, we examine whether emotion differentiation moderates the association between trait rumination and prospective increases in depression in an adult sample. We extend research by Starr et al. (2017) by using a community sample and focusing on individual differences in trait rumination and depressive symptoms. In addition, this is the first study to examine how emotion differentiation is associated with prospective changes in depression over a period of months. We assessed emotion differentiation using a well-established method, experience sampling (ESM), that has high ecological validity (Kashdan et al., 2015).

The emotion differentiation literature has found greater and more consistent negative implications associated with undifferentiated negative than positive emotions, including major depressive disorder (e.g., Demiralp et al., 2012). Additionally, it appears that high NED facilitates down-regulation of negative emotion, whereas, when PED has a significant effect, preliminary evidence suggests that high PED regulates maladaptive behaviors reinforced by positive emotion (e.g., Tugade et al., 2004) or that it diminishes positive emotional experience (Starr et al., 2017). The content of rumination is often negative (Nolen-Hoeksema et al., 2008), posing a high demand for down-regulating negative emotion. Thus, we expect NED to play a more crucial role in intervening ruminative processes than PED. We hypothesized that rumination would only predict increases in depression when negative emotions are less, not more, differentiated. We did not expect PED would moderate the rumination-depression association considering less relevance of differentiating positive emotions when individuals perseverate on negative affect and previous null findings of the association between PED and major depressive disorder (Demiralp et al., 2012). Because mean intensity of negative emotion is related to emotion differentiation and depression (Demiralp et al., 2012; Erbas et al., 2014), we examined whether any significant findings were accounted for by mean intensity of negative or positive emotion.

Method

Participants and Procedure

Seventy-nine community adults participated in a study that examined the everyday emotional experiences of adults (Bailen, Wu, & Thompson, 2018; Gilbert, Tonge, & Thompson, 2019; Thompson & Boden, 2019). The current paper focuses on the 65 participants (60% female) who completed the entire study, including the six-month follow-up survey (i.e., 14 participants [17.7%] did not complete follow-up). We used G*Power 3.1 (Faul, Erdfelder, Buchner, & Lang, 2009) to conduct power analysis of number of participants needed for a medium-sized R2 increase (i.e., F2 = 0.15) based on our primary analyses. In light of our original hypotheses, we entered 0.05 for alpha error probability, 0.80 for power, 1 for number of tested predictors (i.e., NED/PED × rumination interaction), and 5 for number of predictors (T1 depression, mean negative/positive emotion, T1 rumination, NED/PED, and NED/PED × rumination interaction). Results showed that we need 55 participants to detect a medium-sized of R2 increase for multiple regression analyses; thus, the sample size of the study provided acceptable power to test our hypotheses. These 65 participants were on average 38.4 years old (SD = 14.5, range = 20–71). All were native English speakers and had a racial/ethnic distribution as follows: 67.7% White (6.6% Hispanic), 18.5% Black, 7.7% biracial, 4.6% Asian, 1.3% Middle Eastern. Most participants had earned a Bachelor’s degree or higher (76.9%) and were employed part- or full-time (89.3%), and half (50.8%) were married or living with a romantic partner.

This study consisted of two laboratory sessions (T1), a week of ESM, and a six-month follow-up survey (T2). The two baseline laboratory sessions were scheduled before and after the ESM period. At baseline, participants completed self-report questionnaires and an ESM tutorial, including a practice survey. Participants chose their preferred 12-hour period to receive surveys on their iPhone or a provided iPod Touch 4. Each day, participants received eight prompts, each occurring randomly within a 90-min time window. Participants completed a mean of 73% of all surveys (SD = 22%; range = 20–100%). Participants were debriefed and financially compensated at the second baseline laboratory session. Approximately six months later, participants (N = 65) completed self-report measures online and received financial compensation. The research protocol was approved by the university institutional review board at Washington University in St. Louis.

Measures

Emotion differentiation.

At each ESM prompt, participants reported their current feelings (“I feel [emotion] right now”) on a five-point scale (0 = not at all, 4 = extremely). Negative (i.e., frustrated, hostile, sluggish, sad, disappointed, dull, nervous) and positive emotions (i.e., happy, calm, excited, relaxed, enthusiastic, content) that represent the affective circumplex (e.g., Barrett & Russell, 1999) were included. NED and PED were calculated by computing the average intraclass correlations (ICC) with consistency between negative and positive emotions, respectively, followed by Fisher’s r-to-z transformation. Higher ICC represents greater resemblance of ratings of distinct emotions across occasions and, therefore, lower emotion differentiation. To aid interpretation, scores were subtracted from one so greater values represent higher emotion differentiation (e.g., Erbas et al., 2014; Tugade et al., 2004).3

Mean negative and positive emotion.

We calculated a mean negative and positive emotion score for each participant by averaging the participant’s ratings for the negative and positive emotions, respectively. Internal reliability was excellent for negative emotions (α = 0.93) and positive emotions (α = 0.94).

Depressive symptoms.

Depressive symptoms at T1 and T2 were measured by the 20-item Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977), which assesses depressive symptoms over the preceding week. The CES-D was developed to assess depressive symptoms of community samples and has adequate reliability and validity (Eaton, Smith, Ybarra, Muntaner, & Tien, 2004). Internal reliability was good at T1 (α = 0.88) and excellent at T2 (α = 0.90).

Rumination.

We administered the 5-item Brooding subscale of the original 22-item Ruminative Responses Scale (Nolen-Hoeksema & Morrow, 1991; Treynor et al., 2003) to measure trait rumination. Participants reported how frequently they engaged in ruminative thoughts and behaviors in response to negative emotions using a 4-point scale (1 = never; 4 = always). The Brooding subscale is free of depression content and has demonstrated excellent psychometric properties (Treynor et al., 2003). Internal reliability was acceptable (α = 0.76).

Results

Preliminary Analyses

We first examined demographic and clinical differences between participants who did and did not complete the six-month follow-up session. These groups did not differ in gender, χ2(1) = 0.47, p = 0.49, age, Welch’s t(19.2) = −0.61, p = 0.55, race, χ2(4) = 2.65, p = 0.62, or level of education, χ2(5) = 7.02, p = 0.22. They reported comparable levels of rumination, Welch’s t(16.9) = 0.12, p = 0.90, NED, Welch’s t(15.4) = 0.18, p = 0.86, PED, Welch’s t(16.5) = −0.48, p = 0.64, and T1 depression, Welch’s t(15.6) = −0.39, p = 0.70.

We present descriptive statistics and zero-order correlations of emotion differentiation, rumination, and depression (see Table 1). Based on the clinical cut-offs of CES-D (≥ 16; Radloff, 1977), 29.2% and 38.5% of the sample reported depressive symptoms that indicate risk for clinical depression at T1 and T2, respectively. Zero-order Pearson correlations revealed that NED and PED were significantly positively associated. Participants were on average better at differentiating negative than positive emotions, t(63) = 8.52, p < 0.001, d = 1.06. Rumination was significantly inversely associated with NED and PED. T1 depression was significantly inversely associated with NED, but not PED. However, T2 depression was not significantly correlated with NED or PED. Lastly, rumination was significantly positively correlated with T1 and T2 depression.

Table 1.

Mean (M), standard deviation (SD), and zero-order correlations of study variables

Variable M (SD) 2 3 4 5 6 7 8 9
1. ESM NED 0.32a (0.30) 0.57*** 0.87*** 0.36** −0.23 0.16 −0.46*** −0.31* −0.08
2. ESM PED 0.01a (0.33) - 0.90 −0.56*** −0.16 0.13 −0.34** −0.19 −0.17
3. ESM General emotion differentiation 0.33 (0.56) - −0.15 −0.21 0.16 −0.45*** −0.27* −0.14
4. ESM Valence discrepancy of differentiation 0.31 (0.30) - −0.06 0.02 −0.09 −0.10 0.11
5. ESM Mean NE 0.49 (0.41) - −0.27* 0.39** 0.64*** 0.55**
6. ESM Mean PE 1.66 (0.67) - −0.21 −0.44*** −0.37**
7. T1 Rumination 10.28 (3.10) - 0.56*** 0.38**
8. T1 Depressive symptoms 10.89 (8.53) - 0.62***
9. T2 Depressive symptoms 13.45 (9.80) -
Open in a new tab

Note. ESM = measured through experience sampling method; general differentiation = the sum of negative and positive emotion differentiation (i.e., NED + PED); NE = negative emotion; PE = positive emotion; T1 = measured at baseline laboratory session one or two; T2 = measured at 6-month follow-up session; valence discrepancy of differentiation = subtracting positive from negative emotion differentiation (i.e., NED – PED).

a

Means displayed are z-transformed, reverse-coded ICCs.

*

p < 0.05,

**

p < 0.01,

***

p < 0.001

Primary Analyses

To test our hypothesis that rumination would be more strongly associated with increases in depression when NED is lower (not higher), we conducted two hierarchical linear regression analyses predicting T2 depression. One analysis examined NED, and the other examined PED (see Table 2 for full results). All predictors were standardized. In Step 1, we entered T1 depression, mean negative emotion (mean positive emotion), trait rumination, and NED (PED). In Step 2, we entered NED by rumination interaction (PED by rumination interaction).

Table 2.

Hierarchical Regression Models of Emotion Differentiation and Rumination as Predictors of T2 Depressive Symptoms

Model 1a Model 2a Model 3b
Steps/Predictors b (SE) t R2 ΔR2 p b (SE) t R2 ΔR2 p b (SE) t R2 ΔR2 p
Step 1 0.44 0.41  0.47
 Intercept 13.5 (0.94) 14.3 <0.001 13.5 (0.97) 13.9 <0.001 13.5 (0.93) 14.4 <0.001
 T1 Depressive symptoms 4.46 (1.37) 3.26 0.002 5.33 (1.28) 4.16 <0.001 3.98 (1.44) 2.77 0.008
 ESM Mean NE 2.48 (1.24) 2.00 0.05 - - - 2.46 (1.23) 2.01 0.05
 ESM Mean PE - - - −1.15 (1.09) −1.05 0.30 −1.21 (1.04) −1.16 0.25
 T1 Rumination 0.95 (1.24) 0.77 0.44 0.38 (1.23) 0.31 0.76 0.88 (1.24) 0.71 0.48
 ESM NED 1.61 (1.08) 1.50 0.14 - - - 2.50 (1.23) 2.03 0.05
 ESM PED - - - −0.44 (1.04) −0.42 0.68 −1.52 (1.15) −1.33 0.19
Step 2 0.51 0.07 0.007 0.46 0.05 0.02 0.55 0.08 0.01
 Intercept 12.2 (1.00) 12.3 <0.001 12.6 (1.00) 12.6 <0.001 12.2 (0.99) 12.4 <0.001
 T1 Depressive symptoms 4.52 (1.29) 3.49 <0.001 5.34 (1.23) 4.34 <0.001 3.90 (1.36) 2.87 0.006
 ESM Mean NE 2.68 (1.17) 2.28 0.03 - - - 2.63 (1.16) 2.27 0.03
 ESM Mean PE - -  - −1.31 (1.05) −1.25 0.22 −1.47 (0.99) −1.49 0.14
 T1 Rumination 0.52 (1.18) 0.44 0.66 −0.18 (1.21) −0.15 0.88 0.31 (1.19) 0.26 0.80
 ESM NED 1.56 (1.02) 1.53 0.13 - - - 2.38 (1.18) 2.02 0.05
 ESM PED - - - −0.21 (1.00) −0.21 0.84 −1.76 (1.44) −1.22 0.23
 T1 Rumination × ESM NED −2.75 (0.98) −2.81 0.007 - - - −1.55 (1.38) −1.12 0.27
 T1 Rumination × ESM PED - - - −2.65 (1.09) −2.43 0.02 −1.76 (1.44) −1.22 0.23
Open in a new tab

Note. ESM = measured through experience sampling method; NE = negative emotion; NED = negative emotion differentiation; PE = positive emotion; PED = positive emotion differentiation; T1 = measured at baseline laboratory session one or two; T2 = measured at 6-month follow-up session.

a

Models 1 and 2 are primarily analyses testing our hypotheses for NED and PED, respectively.

b

Model 3 is post-hoc analyses examining combined and unique contribution of NED and PED in the same model.

For the model examining NED (see Model 1, Table 2), there were no significant main effects of rumination or NED. As hypothesized, the rumination by NED interaction significantly predicted T2 depression, b = −2.75, p = 0.007, 95% CI [−4.71, −0.79]. Importantly, we included T1 depression in the model, so the interaction significantly predicted prospective change in depression. We also included mean negative emotion to rule out the possibility that our findings were better explained by mean negative emotion. We used unstandardized beta weights and +/− 1.0 standard deviation (SD) values of NED to predict T2 depression (see Figure 1). Simple slope analyses showed that rumination significantly predicted increases in depression when NED was low, b = 3.27, p = 0.03, 95% CI [0.41, 6.14], but not high, b = −2.23, p = 0.18, 95% CI [−5.49, 1.03]. A subsequent Johnson-Neyman test showed that rumination and T2 depression was significantly positively associated when NED was 0.77 SD below the mean, significantly negatively associated when NED was 1.86 SDs above the mean, and not related when NED was between 0.77 SD below the mean and 1.86 SDs above the mean.

Figure 1.

Figure 1.

Open in a new tab

Follow-up depressive symptoms as a function of baseline trait rumination controlling baseline depression and mean negative or positive emotion, as moderated by (a) negative and (b) positive emotion differentiation. Shaded regions delineate the 95% confidence bands for the simple slopes.

For the model including PED (see Table 2, Model 2), neither rumination nor PED was a significant predictor of T2 depression. Contrary to our hypothesis, PED significantly interacted with rumination to predict T2 depression, b = −2.65, p = 0.02, 95% CI [−4.83, −0.46]. As in the NED model, these models controlled for T1 depression and mean positive emotion. However, simple slopes analyses revealed that rumination did not significantly predict T2 depression at either low PED, b = 2.47, p = 0.10, 95% CI [−0.46, 5.40], or high PED, b = −2.82, p = 0.12, 95% CI [−6.37, 0.72] (see Figure 1). A subsequent Johnson-Neyman test showed that rumination and T2 depression were significantly positively associated when PED was 1.39 SDs below the mean, significantly negatively associated when PED was 1.86 SDs above the mean, and not related when PED was between 1.39 SDs below the mean and 1.86 SDs above the mean.

Post-Hoc Analyses

Given that NED and PED were positively correlated, it is possible that findings for one emotion differentiation term (e.g., PED) could be driven by its shared variance with the other (e.g., NED). To test this, we conducted an additional hierarchical regression model including both rumination by NED and rumination by PED interactions (see Model 3, Table 2 for full model description). Results showed that neither interaction was significant. Importantly, the interaction terms are collinear (variance inflation factors: rumination by NED interaction = 2.09; rumination by PED interaction = 2.13), and the addition of both interactions significantly improved the model, ΔR2 = 0.08, p < 0.001. These findings suggest that their combined contribution is significant.

In light of the above findings, and because NED and PED are positively correlated and individually showed similar pattern of findings, we suspected it may not be meaningful to treat NED and PED separately. As such, for each participant, we computed a sum score of NED and PED (i.e., NED + PED) to examine their combined effect into a single construct, what we will refer to as general differentiation. Further, to examine whether it is meaningful to distinguish NED and PED, we computed a difference score of NED and PED (i.e., NED - PED) for each participant to index the extent to which the participant is more greatly differentiating negative than positive emotions (i.e., valence discrepancy of differentiation).4 We conducted a hierarchical linear regression model including general differentiation and valence discrepancy of differentiation and their moderation with rumination in predicting longitudinal depressive symptoms (see Table 3 for the full model and results). Results showed that general differentiation significantly interacted with rumination to predict T2 depression, b = −2.94, p = 0.004, 95% CI [−4.90, −0.98], whereas valence discrepancy of differentiation did not interact with rumination to predict T2 depression, b = 0.01, p = 0.99, 95% CI [−2.43, 2.45]. Simple slope analyses showed that rumination significantly predicted increases in depression when general differentiation was low, b = 3.26, p = 0.03, 95% CI [0.42, 6.11], but not high, b = −2.62, p = 0.12, 95% CI [−5.90, 0.66] (see Figure 2). A Johnson-Neyman test showed that rumination and T2 depression were significantly positively associated when general differentiation was 0.79 SDs below the mean, significantly negatively associated when general differentiation was 1.46 SDs above the mean, and not related when general differentiation was between 0.79 SDs below the mean and 1.46 SDs above the mean.

Table 3.

Hierarchical Regression Model of General Emotion Differentiation and Rumination as Predictors of T2 Depressive Symptoms

Step/Predictors b (SE) t R2 ΔR2 p
Step 1 0.47
 Intercept 13.5 (0.93) 14.43 <0.001
 T1 Depressive symptoms 3.98 (1.44) 2.77 0.008
 ESM Mean NE 2.46 (1.23) 2.01 0.05
 ESM Mean PE −1.21 (1.04) −1.16 0.25
 T1 Rumination 0.88 (1.24) 0.71 0.48
 ESM General differentiation 1.08 (1.08) 1.00 0.32
 ESM Valence discrepancy of differentiation 1.91 (0.97) 1.98 0.05
Step 2 0.55 0.08 0.01
 Intercept 12.21 (0.99) 12.36 <0.001
 T1 Depressive symptoms 3.90 (1.36) 2.87 0.006
 ESM Mean NE 2.63 (1.16) 2.27 0.03
 ESM Mean PE −1.47 (0.99) −1.49 0.14
 T1 Rumination 0.32 (1.18) 0.27 0.79
 ESM General differentiation 1.34 (1.02) 1.31 0.20
 ESM Valence discrepancy of differentiation 1.66 (0.94) 1.78 0.08
 T1 Rumination × ESM General differentiation −2.94 (0.98) −3.01 0.004
 T1 Rumination × ESM Valence discrepancy of differentiation 0.01 (1.22) 0.01 0.99
Open in a new tab

Note. ESM = measured through experience sampling method; general differentiation = the sum of negative and positive emotion differentiation; NE = negative emotion; PE = positive emotion; T1 = measured at baseline laboratory session one or two; T2 = measured at 6-month follow-up session; valence discrepancy of differentiation = subtracting positive from negative emotion differentiation.

Figure 2.

Figure 2.

Open in a new tab

Follow-up depressive symptoms as a function of baseline trait rumination controlling baseline depression and mean negative and positive emotion, as moderated by general emotion differentiation. Shaded regions delineate the 95% confidence bands for the simple slopes.

These results suggest that general differentiation significantly moderated the association between rumination and increases in depression. Additionally, valence discrepancy of differentiation did not emerge as a significant moderator. Therefore, it is more appropriate to interpret the results in terms of the umbrella construct of general differentiation.

Discussion

Trait rumination has consistently been associated with subsequent increases in depression (Nolen-Hoeksema et al., 2008). We examined whether high emotion differentiation weakened this association over six months. As hypothesized, rumination predicted significant prospective increases in depression when negative emotions were less, but not more, differentiated, and this finding was not better explained by mean negative emotion. However, the interactive effect of rumination and PED was also significant, showing similar result patterns to the interactive effect of rumination and NED, which was in contrast to our hypothesis. Follow-up analyses provided evidence that interpreting NED or PED findings alone paints an incomplete picture. Instead, it is the combination of NED and PED, or general differentiation, that is most important in understanding how emotion differentiation is associated with long-term psychological outcomes.

Our findings are consistent with growing evidence that emotion differentiation serves as a resilience factor. Previous findings show that high NED (and sometimes high PED) weakens associations between various risk factors (e.g., intense negative emotions, rumination, borderline personality disorder pathology) and negative outcomes (e.g., alcohol consumption, aggression, NSSI, urges for maladaptive behaviors) among a variety of samples, including underage social drinkers, undergraduate students, adults with borderline personality disorder, and veterans (Dixon-Gordon et al., 2014; Kashdan et al., 2010; Pond et al., 2012; Zaki et al., 2013). This moderation pattern provides compelling evidence that emotion differentiation creates a sort of resilience against negative outcomes (e.g., depression) when facing risk factors (e.g., high rumination).

We proposed separate hypotheses for NED and PED based on the existing literature. However, given the positive correlation between NED and PED and how they similarly interacted with rumination to predict changes in depression, we examined the combined and unique contributions of NED and PED. We found that the combined contribution of NED and PED, or general differentiation, significantly moderated the association between rumination and changes in depression. In contrast, differentiating emotions of one valence more than the other did not predict changes in depressive symptoms. We discuss our theorization of potential mechanisms underlying these findings below.

Given the role of negative affect in depressive rumination and the implication of NED in down-regulating negative emotion in relation to depression, we were surprised to find that this moderation was not specific to NED. However, we computed emotion differentiation scores using data across a week of ESM. That is, we did not assess momentary emotion differentiation. Consequently, our NED and PED scores may both represent participants’ general, trait-like, rather than their momentary, ability to differentiate emotions. People’s ability to differentiate emotions may vary momentarily, and momentary emotion differentiation is uniquely associated with momentary psychological processes (e.g., Tomko et al., 2015). It is possible that trait-like emotion differentiation has a general adaptive value and protects against increases in depression. Had we assessed state emotion differentiation, rumination, and depressed mood, we might have found that differentiating negative (rather than positive) emotions plays a more critical role in disrupting rumination and improving emotion in the moment. This speculation should be tested in future research.

Starr et al. (2017) demonstrated that daily and momentary rumination were more strongly associated with within-person changes in daily and momentary depressive symptoms when NED was low (versus high) in college student and veteran samples. The present study extends this research by assessing trait rumination, focusing on between-person differences in a community sample representing a wide age range, and examining longitudinal changes of depressive symptoms. We also assessed effects of NED, PED, and general differentiation. Taken together, these results provide support for the idea that emotion differentiation is a consistent moderator of the rumination-depression link for both state-like assessment within persons and trait-like examination across individuals.

In the current study, NED was positively associated with PED. Effect sizes of the association between NED and PED in the literature have ranged from small (Barrett et al., 2001; Dixon-Gordon et al., 2014; Erbas, Ceulemans, Koval, & Kuppens, 2015; Erbas, Sels, Ceulemans, & Kuppens, 2016; Lennarz, Lichtwarck-Aschoff, Timmerman, & Granic, 2018; Sheets, Bujarski, Leventhal, & Ray, 2015; Starr et al., 2017) to moderate (Boden, Thompson, Dizén, Berenbaum, & Baker, 2013; Emery, Simons, Clarke, & Gaher, 2014; Erbas et al., 2015, 2016; Starr et al., 2017) to large (Selby et al., 2014). The weighted effect size of the Pearson correlations between NED and PED reported in these 13 studies was 0.31. In this context, the association between NED and PED observed in the present study was on the higher end of this range. Despite frequent positive correlations between NED and PED, most existing research has examined NED and PED separately without considering their combined contribution, and many studies only computed NED but not PED. It is therefore unclear whether many findings for NED can be extended to PED or general differentiation. Future research should parse apart when NED, PED, and/or general differentiation are associated with various well-being indices across contexts.

We posit two non-exclusive mechanisms that may underlie the moderating effect of general differentiation. First, high ability to differentiate specific emotions might aid in interrupting the ruminative cycle via emotion regulation. Rumination often worsens mood, which further increases rumination, forming a vicious cycle that increases risk for depression (Moberly & Watkins, 2008; Nolen-Hoeksema et al., 2008). Higher emotion differentiation allows for more fine-grained inferences about one’s emotions and better emotion labeling, thereby facilitating effective and context-specific emotion regulation. These benefits protect individuals from getting stuck in rumination and subsequently developing depression. Future experimental research can examine the link between emotion differentiation and emotion regulation by manipulating emotion differentiation (e.g., Cameron, Payne, & Doris, 2013) and measuring how effective participants are in regulating their negative emotion in response to, for example, sadness-inducing emotional stimuli.

Secondly, we posit that emotion differentiation may be associated with concreteness of self-focused processing (Moberly & Watkins, 2006; Schaich, Watkins, & Ehring, 2013; Watkins, 2004), with lower emotion differentiation representing less concrete, more abstract processing. Abstract self-focused processing has been shown to lead to poorer emotional and psychological outcomes compared to concrete self-focused processing (Moberly & Watkins, 2006; Schaich et al., 2013; Watkins, 2004). Similar to emotion differentiation, trait rumination appears to be associated with adverse outcomes, only among those with abstract (not concrete) self-focus. Additionally, low NED and abstract thinking appear to function under the same mechanisms, including increasing negative emotion reactivity (Starr et al., 2017; Watkins, Moberly, & Moulds, 2008) and impeding problem solving (Watkins & Moulds, 2005). Hence, it is meaningful to examine whether emotion differentiation is related to concreteness of processing to clarify the nature of the link between low emotion differentiation and poor well-being.

Regarding clinical implications, results demonstrated emotion differentiation might be an important potential intervention target in the prevention and treatment for depression. Future researchers should directly test whether people major depressive episodes more remit more quickly among those with high versus low emotion differentiation. Given previous evidence of the salutary effects of training concrete thinking (Watkins et al., 2008), cultivating emotion differentiation is likely to benefit psychological well-being. There is evidence of the success of manipulating emotion differentiation in the lab among college students (Cameron et al., 2013) and enhancing emotion differentiation through self-monitoring using ESM among patients with major depressive disorder (Widdershoven et al., 2019). Future research should directly examine the efficacy of enhancing emotion differentiation on improving well-being among clinical samples, such as those with major depressive disorder.

We note several limitations. First, owing to the correlational nature of the current study, we can only speculate the mechanisms through which emotion differentiation predicts prospective changes in depression. Second, positive emotions appear to be more similar to each other than were negative emotions, possibly because English language has fewer positive than negative emotion words (Averill, 1957). This may be partially why our participants tended to have higher NED than PED (see Erbas et al., 2016 for similar results). Participants’ general low levels of PED (i.e., floor effect) may explain why rumination did not predict changes in depression when PED was either low or high defined as one SD above and below the mean. Third, we did not assess momentary rumination and depression during ESM, which would allow us to examine the momentary associations of rumination and NED on momentary depression. Lastly, although several participants exhibited clinically significant levels of depressive symptoms, we assessed depressive symptoms in a community sample using a self-report measure. The use of clinical cutoffs on self-report measures to indicate the presence of a major depressive episode is associated with elevated false positive and false negative errors compared with diagnostic clinical interviews (see Bredemeier et al., 2010, for a discussion). Thus, we cannot draw confident conclusions about the applicability of the current findings to clinically depressed populations. Future research should follow individuals at risk for major depressive disorder and examine whether rumination and emotion differentiation interact to predict occurrence of major depressive episodes indexed via clinical interviews.

Conclusion

Further elucidating the well-established association between rumination and depression, we examined the moderating role of emotion differentiation using a prospective longitudinal design. Trait rumination only predicted increases in depression over six months for low, but not high, emotion differentiation. A closer examination suggested that results are best explained by the combined contribution of NED and PED, or general differentiation. Findings demonstrate high emotion differentiation as a resilience factor for depressive symptoms and a promising avenue for research on major depressive disorder.

Acknowledgments

This research was supported by the Spencer T. Olin Fellowship, Washington University in St. Louis, to DYL and by K23 MH115074–01 to KEG. We thank Tammy English, Mike Strube, Haijing Wu Hallenbeck, and Natasha Haradhvala Bailen for their feedback on the manuscript. We also thank Jordan Davis for her help with managing the study.

Footnotes

1

The current paper focuses on rumination in the context of negative affect, not positive affect (i.e., positive rumination; Feldman et al., 2008).

2

Because researchers have used different statistical models for studying PED, combining different statistical outcomes into a common effect size indicator is beyond the central aims of the current paper.

3

Two participants’ ICCs for positive emotion were negative, reflecting measurement error. We changed them to zero as recommended by Cohen, Cohen, West, and Aiken (2003) before performing transformation (see Boden, Thompson, Dizén, Berenbaum, & Baker, 2013, for similar procedures). We ran the regression model for PED excluding these two participants. The significance of the results was not affected; the PED-rumination interaction term continued to be significant, b = −2.64, p = 0.02.

4

Positive valence discrepancy values indicate NED being greater than PED, negative values indicate PED being greater than NED, and zero indicates NED being equal to PED. We subtracted PED from NED because we hypothesized that NED plays a larger role in the rumination-depression association than PED. However, subtracting NED from PED also represents their discrepancy. The correlation between NED minus PED and PED minus PED is −1.

References

  1. Anand D, Chen Y, Lindquist KA, & Daughters SB (2017). Emotion differentiation predicts likelihood of initial lapse following substance use treatment. Drug and Alcohol Dependence, 180, 439–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Armey MF, Fresco DM, Moore MT, Mennin DS, Turk CL, Heimberg RG, … Alloy LB (2009). Brooding and pondering: Isolating the active ingredients of depressive rumination with exploratory factor analysis and structural equation modeling. Assessment, 16(4), 315–327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Averill JR (1957). A semantic atlas of emotional concepts. Catalog of Selected Documents in Psychology, 5, 330. [Google Scholar]
  4. Bailen NH, Wu H, & Thompson RJ (2018). Meta-emotions in daily life: Associations with emotional awareness and depression. Emotion. [DOI] [PubMed] [Google Scholar]
  5. Barrett LF, Gross J, Christensen TC, & Benvenuto M (2001). Knowing what you’re feeling and knowing what to do about it: Mapping the relation between emotion differentiation and emotion regulation. Cognition & Emotion, 15(6), 713–724. [Google Scholar]
  6. Barrett LF, & Russell JA (1999). The structure of current affect: Controversies and emerging consensus. Current Directions in Psychological Science, 8(1), 10–14. [Google Scholar]
  7. Boden MT, Thompson RJ, Dizén M, Berenbaum H, & Baker JP (2013). Are emotional clarity and emotion differentiation related? Cognition and Emotion, 27(6), 961–978. [DOI] [PubMed] [Google Scholar]
  8. Bredemeier K, Spielberg JM, Silton RL, Berenbaum H, Heller W, & Miller GA (2010). Screening for depressive disorders using the Mood and Anxiety Symptoms Questionnaire Anhedonic Depression Scale: A receiver-operating characteristic analysis. Psychological Assessment, 22(3), 702–710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cameron CD, Payne BK, & Doris JM (2013). Morality in high definition: Emotion differentiation calibrates the influence of incidental disgust on moral judgments. Journal of Experimental Social Psychology, 49(4), 719–725. [Google Scholar]
  10. Cohen J, Cohen P, West SG, & Aiken LS (2003). Applied multiple regression/correlation analysis for the behavioral sciences. (3rd ed). Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers. [Google Scholar]
  11. Demiralp E, Thompson RJ, Mata J, Jaeggi SM, Buschkuehl M, Barrett LF, … Jonides J (2012). Feeling blue or turquoise? Emotional differentiation in major depressive disorder. Psychological Science, 23(11), 1410–1416. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dixon-Gordon KL, Chapman AL, Weiss NH, & Rosenthal MZ (2014). A preliminary examination of the role of emotion differentiation in the relationship between borderline personality and urges for maladaptive behaviors. Journal of Psychopathology and Behavioral Assessment, 36, 616–625. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Eaton WW, Smith C, Ybarra M, Muntaner C, & Tien A (2004). Center for epidemiologic studies depression scale: Review and revision (CESD and CESD-R). In M. E. Maruish (Ed.), The use of psychological testing for treatment planning and outcomes assessment: Instruments for adults (pp. 363–377). Mahwah, NJ: Lawrence Erlbaum Associates Publishers. [Google Scholar]
  14. Emery NN, Simons JS, Clarke CJ, & Gaher RM (2014). Emotion differentiation and alcohol-related problems : The mediating role of urgency. Addictive Behaviors, 39(10), 1459–1463. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Erbas Y, Ceulemans E, Koval P, & Kuppens P (2015). The role of valence focus and appraisal overlap in emotion differentiation. Emotion, 15(3), 373–382. [DOI] [PubMed] [Google Scholar]
  16. Erbas Y, Ceulemans E, Lee Pe M, Koval P, & Kuppens P (2014). Negative emotion differentiation: Its personality and well-being correlates and a comparison of different assessment methods. Cognition and Emotion, 28(7), 1196–1213. [DOI] [PubMed] [Google Scholar]
  17. Erbas Y, Sels L, Ceulemans E, & Kuppens P (2016). Feeling Me, Feeling You : The Relation Between Emotion Differentiation and Empathic Accuracy. Social Psychology and Personality Science, 7(3), 240–247. [Google Scholar]
  18. Faul P, Erdfelder E, Buchner A, & Lang A-G (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160. [DOI] [PubMed] [Google Scholar]
  19. Feldman GC, Joormann J, & Johnson SL (2008). Responses to positive affect: A self-report measure of rumination and dampening. Cognitive Therapy and Research, 32(4), 507–525. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gilbert KE, Tonge NA, & Thompson RJ (2019). Associations between depression, anxious arousal and manifestations of psychological inflexibility. Journal of Behavior Therapy and Experimental Psychiatry, 62, 88–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gross JJ (2015). Emotion regulation: Current status and future prospects. Psychological Inquiry, 26(1), 1–26. [Google Scholar]
  22. Kashdan TB, Barrett LF, & McKnight PE (2015). Unpacking emotion differentiation: Transforming unpleasant experience by perceiving distinctions in negativity. Current Directions in Psychological Science, 24(1), 10–16. [Google Scholar]
  23. Kashdan TB, Ferssizidis P, Collins RL, & Muraven M (2010). Emotion differentiation as resilience against excessive alcohol use: An ecological momentary assessment in underage social drinkers. Psychological Science, 21(9), 1341–1347. [DOI] [PubMed] [Google Scholar]
  24. Kircanski K, Lieberman MD, & Craske MG (2012). Feelings into words: Contributions of language to exposure therapy. Psychological Science, 23(10), 1086–1091. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Lennarz HK, Lichtwarck-Aschoff A, Timmerman ME, & Granic I (2018). Emotion differentiation and its relation with emotional well-being in adolescents. Cognition and Emotion, 32(3), 1–7. [DOI] [PubMed] [Google Scholar]
  26. Moberly NJ, & Watkins ER (2006). Processing mode influences the relationship between trait rumination and emotional vulnerability. Behavior Therapy, 37, 281–291. [DOI] [PubMed] [Google Scholar]
  27. Moberly NJ, & Watkins ER (2008). Ruminative self-focus and negative affect: An experience sampling study. Journal of Abnormal Psychology, 117(2), 314–323. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Nolen-Hoeksema S (1991). Responses to depression and their effects on the duration of depressive episodes. Journal of Abnormal Psychology, 100(4), 569–582. [DOI] [PubMed] [Google Scholar]
  29. Nolen-Hoeksema S, & Morrow J (1991). A prospective study of depression and posttraumatic stress symptoms after a natural disaster: The 1989 Loma Prieta earthquake. Journal of Personality and Social Psychology, 61(1), 115–121. [DOI] [PubMed] [Google Scholar]
  30. Nolen-Hoeksema S, Wisco BE, & Lyubomirsky S (2008). Rethinking rumination. Perspectives on Psychological Science, 3(5), 400–424. [DOI] [PubMed] [Google Scholar]
  31. Pond RS, Kashdan TB, DeWall CN, Savostyanova A, Lambert NM, & Fincham FD (2012). Emotion differentiation moderates aggressive tendencies in angry people: A daily diary analysis. Emotion, 12(2), 326–337. [DOI] [PubMed] [Google Scholar]
  32. Radloff LS (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1(3), 385–401. [Google Scholar]
  33. Schaich A, Watkins ER, & Ehring T (2013). Can concreteness training buffer against the negative effects of rumination on PTSD? An experimental analogue study. Journal of Behavior Therapy and Experimental Psychiatry, 44(4), 396–403. [DOI] [PubMed] [Google Scholar]
  34. Schwarz N (1990). Feelings as information: Informational and motivational functions of affective states In Higgins ET & Sorrentino RM (Eds.), Handbook of Motivation and Cognition (pp. 527–561). New York: Guilford Press. [Google Scholar]
  35. Selby EA, Wonderlich SA, Crosby RD, Engel SG, Panza E, Mitchell JE, … Le Grange D (2014). Nothing tastes as good as thin feels: Low positive emotion differentiation and weight-loss activities in anorexia nervosa. Clinical Psychological Science, 2(4), 514–531. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Sheets ES, Bujarski S, Leventhal AM, & Ray LA (2015). Emotion differentiation and intensity during acute tobacco abstinence: A comparison of heavy and light smokers. Addictive Behaviors, 47, 70–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Starr LR, Hershenberg R, Li YI, & Shaw ZA (2017). When feelings lack precision: Low positive and negative emotion differentiation and depressive symptoms in daily life. Clinical Psychological Science, 5(4), 613–631. [Google Scholar]
  38. Thompson RJ, & Boden MT (2019). State emotional clarity and attention to emotion: A naturalistic examination of their associations with each other, affect and context. Cognition and Emotion. [DOI] [PubMed] [Google Scholar]
  39. Tomko RL, Lane SP, Pronove LM, Treloar HR, Brown WC, Solhan MB, … Trull TJ (2015). Undifferentiated negative affect and impulsivity in borderline personality and depressive disorders: A momentary perspective. Journal of Abnormal Psychologysychology, 124(3), 740–753. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Treynor W, Gonzalez R, & Nolen-Hoeksema S (2003). Rumination reconsidered: A psychometric analysis. Cognitive Therapy and Research, 27(3), 247–259. [Google Scholar]
  41. Tugade MM, Fredrickson BL, & Barrett LF (2004). Psychological resilience and positive emotional granularity: Examining the benefits of positive emotions on coping and health. Journal of Personality, 72(6), 1161–1190. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Watkins ER (2004). Adaptive and maladaptive ruminative self-focus during emotional processing. Behaviour Research and Therapy, 42(9), 1037–1052. [DOI] [PubMed] [Google Scholar]
  43. Watkins ER (2008). Constructive and unconstructive repetitive thought. Psychological Bulletin, 134(2), 163–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Watkins ER, Moberly NJ, & Moulds ML (2008). Processing mode causally influences emotional reactivity: Distinct effects of abstract versus concrete construal on emotional response. Emotion, 8(3), 364–378. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Watkins ER, & Moulds M (2005). Distinct modes of ruminative self-focus: Impact of abstract versus concrete rumination on problem solving in depression. Emotion, 5(3), 319–328. [DOI] [PubMed] [Google Scholar]
  46. Widdershoven RLA, Wichers M, Kuppens P, Hartmann JA, Menne-Lothmann C, Simons CJP, & Bastiaansen JA (2019). Effect of self-monitoring through experience sampling on emotion differentiation in depression. Journal of Affective Disorders, 244, 71–77. [DOI] [PubMed] [Google Scholar]
  47. Zaki LF, Coifman KG, Rafaeli E, Berenson KR, & Downey G (2013). Emotion differentiation as a protective factor against nonsuicidal self-injury in borderline personality disorder. Behavior Therapy, 44(3), 529–540. [DOI] [PubMed] [Google Scholar]

RESOURCES