When Feelings Lack Precision: Low Positive and Negative Emotion Differentiation and Depressive Symptoms in Daily Life

Abstract

Research suggests the ability to differentiate discrete emotions protects against psychopathology. Little is known about daily processes through which negative and positive emotion differentiation (NED, PED) influence depressive symptomatology. We examined NED and PED as moderators of associations between daily processes (negative/positive experiences, brooding, and savoring) and daily depressive symptoms. Hypotheses were tested using intensive longitudinal techniques in two samples oversampled for depression: 157 young adults (Study 1) and 50 veterans recruited from VA primary care (Study 2). In Study 1, low NED predicted stronger associations between daily brooding and depressive symptoms. In Study 2, low NED predicted stronger reactivity to daily negative events. In both studies, low PED strengthened salutary effects of positive experiences and savoring on symptoms. Largely consistent across demographically divergent samples, results suggest both low NED and PED enhance effects of daily events and perseverative self-focus on fluctuations in depressive symptoms.

Keywords

negative emotion differentiation positive emotion differentiation depression brooding savoring ecological momentary assessment daily diary

Emotion differentiation (ED) refers to the ability to identify and precisely label discrete emotional states (Barrett, Gross, Christensen, & Benvenuto, 2001; Kashdan, Barrett, & McKnight, 2015). Those who are low on this ability tend to report their emotions in broad terms of valence (“I feel good,” or “I feel upset”) rather than pinpointing concrete emotions (“I feel excited,” or “I feel afraid”). Increasing research suggests that ED has consequences for psychopathology and well-being. Emotions communicate critical information about the need to employ attentional and behavioral resources. Thus, when people are better able to discriminate between discrete emotional states they are more prepared to extract relevant information about the causes and consequences of their emotions, such as the eliciting context, cognitive and physiological correlates, and behavioral urges. This awareness also provides information needed to effectively select and deploy appropriate emotion regulation strategies (Barrett et al., 2001). Supporting this theoretical model, ED is associated with less impulsive emotional responding and more effective use of emotion regulation strategies (Barrett et al., 2001; Kashdan et al., 2015; Tugade, Fredrickson, & Barrett, 2004).

Given that appropriately experiencing and flexibly expressing emotions is central to well-being and represents a transdiagnostic process in psychopathology (Kring & Sloan, 2011), it is not surprising that difficulty discriminating between concrete emotional states—that is, low ED—has been linked to critical processes and behaviors across a wide range of disorders, including substance abuse, eating disorders, borderline personality disorder, autism spectrum disorder, and social anxiety disorder (Dixon-Gordon, Chapman, Weiss, & Rosenthal, 2014; Erbas, Ceulemans, Boonen, Noens, & Kuppens, 2013; Kashdan & Farmer, 2014; Kashdan, Ferssizidis, Collins, & Muraven, 2010; O’Toole, Jensen, Fentz, Zachariae, & Hougaard, 2014; Selby et al., 2014; Zaki, Coifman, Rafaeli, Berenson, & Downey, 2013). Some evidence also specifically links poor differentiation of negative emotions (NEs; referred to as negative emotion differentiation; NED) to depression, including self-reported depressive symptoms, major depressive disorder (MDD), and symptom severity within depressive episodes (Demiralp et al., 2012; Erbas, Ceulemans, Lee Pe, Koval, & Kuppens, 2014; Golston, Gara, & Woolfolk, 1992). This growing body of work suggests that NED could be an important contributor to depression’s etiology and maintenance. However, research on ED and depression remains limited, and no research has teased apart specific daily processes through which ED influences depressive symptoms.

Most research on ED has focused on NED. A much smaller set of studies has explored positive emotion differentiation (PED), perhaps because the limited existing research suggests that compared to NED, PED is less consistently linked to emotion regulation deficits and well-being (Barrett et al., 2001; Demiralp et al., 2012; Kashdan & Farmer, 2014; Pond et al., 2012). NE and positive emotion (PE) serve different purposes; whereas PEs build long-term resources and broaden one’s response repertoire (Fredrickson, 1998, 2001), NEs primarily function to allocate resources to avoid or mitigate immediate threats (Parrott, 2002). Thus, failure to regulate PEs may be much less costly than failure to regulate NEs (Barrett et al., 2001; Quigley & Barrett, 1999). That said, several studies have suggested that PED does play a role in psychopathological processes and coping behaviors (Dixon-Gordon et al., 2014; Selby et al., 2014; Tugade et al., 2004). However, PED has rarely been explored in the context of depression, despite the fact that there has been a sharpening focus on the role of PEs and rewarding experiences within the depression literature (Treadway & Zald, 2011). Indeed, the absence of PEs and failure to anticipate, seek, and benefit from positive experiences are increasingly considered a central and perhaps defining feature of depression (Kovacs et al., 2016; Pizzagalli et al., 2009; Rottenberg, 2007; Treadway & Zald, 2011; Watson & Naragon-Gainey, 2010; Weinberg, Liu, Hajcak, & Shankman, 2015). Although one prior study failed to document a basic association between PED and MDD (Demiralp et al., 2012), PED may still influence critical daily processes that influence depressive symptoms.

The limited previous research has assumed that, like low NED, low PED would confer risk for maladaptive outcomes (Dixon-Gordon et al., 2014; Hill & Updegraff, 2012; Selby et al., 2014; Tugade et al., 2004). This assumption is based on the logic that the ability to perceive emotions in a more sophisticated, granular manner is adaptive, regardless of whether the emotions are positive (high PED) or negative (high NED). However, it is also possible that excessively differentiating between PE states may lead to a more constrained, narrower experience of PEs. As a result, high PED may make some behaviors and experiences less emotionally rewarding. On the one hand, this may reduce risk for maladaptive behaviors that are typically reinforced by emotional rewards; supporting this notion, high PED has been associated with a reduced rate of self-destructive behaviors such as disordered eating and self-injurious behavior (Dixon-Gordon et al., 2014; Selby et al., 2014). On the other hand, PED may also constrain the emotional benefits of adaptive experiences, such as everyday uplifts and positive experiences. Notably, a core feature of depression is the absence of PE and reactivity to rewarding experiences. Thus, high PED could contribute to anhedonia and depressive symptom risk by reducing the antidepressant effects of positive activities.

In the current research, we first focus on the impact of ED on levels of depressive symptoms when positive and negative events happen in everyday life. In addition, we examine the role of ED in amplifying or decreasing depressive symptoms when employing two common emotion regulation strategies: brooding (on NE) and savoring (PE).

NED and Daily Negative Experiences

We expect that low NED will predict a stronger association between everyday negative experiences and daily depressive symptoms. As noted previously, those with poor NED have more difficulty selecting and implementing effective emotion regulation strategies (Barrett et al., 2001; Kashdan et al., 2015) and are more likely to resort to destructive behaviors when confronted with NE (Kashdan et al., 2010; Pond et al., 2012; Zaki et al., 2013). Lacking appropriate coping skills, these individuals may be more vulnerable to exacerbation of depressive symptoms when negative experiences occur. Moreover, a wide body of evidence suggests that the act of affect labeling may be regulatory in itself. Affect labeling has been linked to diminished emotional reactivity (as measured by self-report, neural activation, and autonomic response) following exposure to negative stimuli (Kircanski, Lieberman, & Craske, 2012; Lieberman et al., 2007; Lieberman, Inagaki, Tabibnia, & Crockett, 2011). As NED suggests a greater propensity toward precise labeling of emotions, those with higher NED may be protected against depressive responses following negative everyday experiences.

To our knowledge no studies have examined whether NED confers reactivity to daily stressors. However, Kashdan et al. (2014) did find that, among those with low self-esteem, low NED predicted greater neural reactivity in response to social rejection, consistent with the hypothesis that poor NED serves as a vulnerability factor that exacerbates the consequences of stressors or other kinds of negative environmental events. We thus hypothesized that lower NED would predict stronger associations between daily negative experiences and daily depressive symptoms.

PED and Daily Positive Experiences

As a corollary, we also examined whether PED would impact the association between everyday positive experiences and daily depressive symptoms. Although the depression literature has traditionally focused far more on stressors, positive experiences have long been considered important agents in reducing depressive symptoms (Lewinsohn & Graf, 1973; Lewinsohn, Sullivan, & Grosscup, 1980), and reactivity to positive experiences is being increasingly explored in the context of depression, both in daily life and within the laboratory (Bylsma, Morris, & Rottenberg, 2008; Bylsma, Taylor-Clift, & Rottenberg, 2011; Peeters, Nicolson, Berkhof, Delespaul, & deVries, 2003; Starr & Hershenberg, in press; Thompson et al., 2012). As previously explained, low PED may amplify salutatory effects of positive events on depressive symptoms by allowing for a more diffuse, less constrained PE experience when “good things” happen. Moreover, research suggests that positive affect labeling is associated with diminished self-reported pleasure (Lieberman et al., 2011). Therefore, individuals with low PED, who are not prone to labeling PEs, may be more reactive to daily positive experiences. We thus hypothesized that lower PED would predict stronger associations between daily positive experiences and lower levels of daily depressive symptoms.

Emotion Differentiation and Positive and Negative Rumination

Next, we considered how NED and PED might interact with, respectively, negative rumination (brooding) and positive rumination (savoring). In a wide body of research that includes longitudinal, daily diary, and experimental evidence, depressive rumination robustly predicts the onset and maintenance of negative mood and depressive episodes (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Brooding, or passive focus on negative consequences of symptoms, has been identified as the most depressogenic component of rumination (Treynor, Gonzalez, & Nolen-Hoeksema, 2003). Low NED may exacerbate the tendency to engage in, and the negative consequences of, brooding. Indeed, those with difficulty understanding their NEs may feel more compelled to dwell on them, and low NED may amplify the negative effects of brooding on daily depressive symptoms. For example, having a more diffuse, generalized perception of NEs may mean that intense focus on one negative feeling (e.g., I feel disappointed) may quickly spread to other NEs (I feel sad, I feel guilty, I feel anxious). In turn, brooding about a wider range of NEs may provoke correspondingly broad negative inferences about the self, world, and future, triggering or exacerbating depressive symptoms. Little previous work has examined the effects of NED and rumination, though one study showed that high NED protects against the effects of trait rumination on nonsuicidal self-injury in borderline personality disorder (Zaki et al., 2013). We thus hypothesized that low NED would be associated with trait rumination and that low NED would strengthen the association between daily brooding and depressive symptoms.

Although rumination is typically shorthand for brooding and other forms of perseverative focus on NE, rumination can also be in response to PE. For example, savoring refers to PE-focused cognitive responses that serve to increase or maintain one’s PEs (Martin & Tesser, 1996; Quoidbach, Berry, Hansenne, & Mikolajczak, 2010; Wood, Heimpel, & Michela, 2003). Although less widely studied than depressive rumination, growing research suggests that savoring may be protective against depression. Higher savoring beliefs are negatively correlated with depression, and savoring predicts decreased daily depressive symptoms (Bryant, 2003; Hurley & Kwon, 2012; Li, Starr, & Hershenberg, 2016). Analogous to our predictions for NED and brooding, we expect that the association between savoring and lower depressive symptoms will be amplified among those with low PED. For example, for those with a less differentiated perception of PE, savoring one PE (e.g., I feel cheerful) may intensify a broad range of PEs (I feel enthusiastic, I feel confident, etc.), which may translate into reduced depressive symptoms. Thus, we hypothesized that the association between savoring and lower depressive symptoms will be amplified among those with low PED.

The Present Research

We tested these hypotheses across two studies, one in a sample of young adults and another in a sample of veterans recruited from a primary care cohort being evaluated for behavioral health symptoms. Both were oversampled for depressive symptoms to allow for significant variation in daily depressive symptoms over the course of the study. Both studies relied on intensive longitudinal techniques for calculation of PED/NED and testing of hypotheses. These techniques produce real-time data collected in naturalistic settings, minimizing the need for retrospective recall and increasing generalizability. Furthermore, objectively calculating ED from momentary affect ratings, rather than asking participants to self-report on their perceived ability to differentiate emotions, reduces reliance on introspection and self-awareness. In Study 1, ecological momentary assessment (EMA) and daily diary surveys were both administered over overlapping time periods. EMA data were used to calculate ED, as EMA is better suited to capture discrete emotional states, and daily diary data were used to test study hypotheses (because daily diary allows for longer surveys and more careful assessment of hassles, uplifts, savoring, and brooding). Study 2 relied exclusively on EMA data, allowing for replication of findings across intensive longitudinal approaches.

After examining basic associations between ED and baseline depression and rumination, we tested the following hypotheses:

Hypothesis 1: Low NED will predict stronger associations between daily negative experiences and daily depressive symptoms.

Hypothesis 2: Low PED will predict stronger associations between daily positive experiences and reductions in daily depressive symptoms.

Hypothesis 3: Low NED will predict stronger associations between daily brooding and daily depressive symptoms.

Hypothesis 4: Low PED will predict stronger associations between daily savoring and reduced daily depressive symptoms.

With the exception of Hypothesis 3, which was examined only in Study 1, all hypotheses were tested in both samples, allowing for direct replication across two samples that varied considerably in depression risk, demographic characteristics, and experience sampling methodology.

Study 1

Method

Participants

We recruited 160 undergraduate psychology students. Although they should not be considered representative of clinical populations, college students are vulnerable to depressive symptoms (e.g., Garlow et al., 2008), and research suggests that findings generated in undergraduate samples generalize to clinical samples (Vredenburg, Flett, & Krames, 1993). Eligibility criteria for participation were minimum age of 18 years, access to Internet and a personal cell phone, and no English comprehension difficulties. To ensure we recruited a sample with a broad range of depressive symptoms, we conducted a screening study where potential participants completed a self-report depression measure, the Quick Inventory of Depressive Symptomatology (QIDS; Rush et al., 2003). Participants were then preferentially recruited to achieve approximately equal distribution across three categories (based on published clinical thresholds, Rush et al., 2003): no symptoms (QIDS < 6, 31% of sample), mild symptoms (QIDS score of 6–10, 33% of sample), and moderate to severe symptoms (QIDS > 10, 36% of sample). Students received extra credit and were entered into raffles based on diary compliance. Table 1 reports sample characteristics. Of the participants recruited, one did not complete diaries, five provided too few valid EMA surveys for ED calculation, and two were excluded from analyses after failing inattention checks (described later), resulting in a final sample of 152 participants. This research was approved by the University of Rochester’s Research Subjects Review Board.

Table 1.

Demographic Information and Descriptive Data for Study 1 and Study 2

Variable	Study 1	Study 2
N	157	50
Age	20.10 (1.23)	56.48 (14.06)
% female	81	12
Race
% Caucasian	57	44
% African American	6	52
% Asian	29	0
% Other	7	4
Baseline depressive symptoms	8.81 (5.19)	12.92 (6.00)
Baseline rumination	56.56 (15.03)	53.90 (15.58)
NED	0.83 (0.13)	0.74 (0.20)
PED	0.66 (0.18)	0.56 (0.21)
Aggregated daily/momentary variables
Depressive Sx/depressed mood	4.00 (4.29)	1.85 (0.81)
Negative experiences/hassles	7.83 (5.45)	2.41 (0.70)
Positive experiences/uplifts	10.71 (6.29)	3.59 (0.70)
Brooding	9.41 (3.18)	—
Savoring	1.97 (0.56)	2.53 (0.87)

Note: Values are means, with standard deviations in parentheses, unless otherwise noted. All variables (except rumination) were assessed using different measures in Study 1 and Study 2, and therefore descriptive data should not be directly compared across studies. Baseline depressive symptoms assessed with QIDS (Rush et al., 2003) in Study 1 and PHQ-9 (Kroenke et al., 2001) in Study 2. NED and PED were computed using different indicators of NEs and PEs and may also not be directly comparable. See the Method sections for Study 1 and Study 2 for more detail.

Procedure

Participants completed an initial baseline visit and once-a-day daily diaries for 14 consecutive days, beginning the evening of the baseline visit. Diaries were completed as close to bedtime as convenient. Participants completed 10.97 (78.3%) diaries on average. Number of missed diaries was not significantly related to baseline depressive symptoms, NED, or PED.

In addition to the daily diary, participants also completed a brief EMA protocol by completing short, telephone-based surveys five times per day for five days. In the current study, EMA data were used for the computation of NED/PED only; all hypotheses were tested using daily diary data. EMA surveys were administered using the telephone-based platform telEMA (Fernandez, Johnson, & Rodebaugh, 2013). Participants designated time-of-day windows (typically 12 hr) in which they were available to complete surveys. The daily window was divided into five equal intervals, and one call was placed at a random time within each of the five intervals, with the additional constraint that no two calls could be less than 1.5 hr apart (to minimize burden). Participants were able to designate multiple phone numbers and specify the number of repeat calls they would receive if they missed a call. Participants had up to 30 min to call a designated number to complete a survey if they missed a call. EMAs were typically started the day after baseline participation (thus, EMA and daily diary periods overlapped) and were always timed so that the 5-day period included three weekdays and two weekend days. Item order was randomized within blocks. During the data cleaning process, EMA data were inspected for evidence of invalid response patterns (e.g., repeatedly entering identical numbers), and suspicious data were excluded. Participants completed an average of 19.44 EMAs (78%), and 80% of the sample completed at least 18/25 EMAs. Four participants completed three or fewer valid EMAs and were excluded.

In line with Maniaci and Rogge’s (2014) recommendation, we included six inattention items in the baseline measure to identify inattentive respondents. Exclusion of inattentive respondents improves statistical power (Maniaci & Rogge, 2014). Two participants failed inattention checks consistently and were excluded from all analyses.

Measures

Baseline

Baseline depressive symptom severity was measured using the QIDS (Rush et al., 2003), a 16-item self-report questionnaire assessing the nine criterion symptom domains of MDD according to the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 1994). The QIDS was administered in the screening study for recruitment purposes and then readministered at baseline and utilized as a continuous score. Previous research has supported the psychometric properties of the QIDS (Rush et al., 2003; Rush et al., 2006). For example, the QIDS has demonstrated construct validity through correlations with depressive symptoms and diagnoses, as well as sensitivity to symptom change (Gonzalez, Boals, Jenkins, Schuler, & Taylor, 2013; Trivedi et al., 2004). Cronbach’s alpha in this study was .84. Baseline rumination was assessed using the Ruminative Response Scale (RRS; Nolen-Hoeksema & Morrow, 1991), a widely used 22-item scale prompting respondents to rate the frequency of 22 ruminative thoughts or behaviors. The RRS has excellent internal consistency and external validity (Butler & Nolen-Hoeksema, 1994; Nolen-Hoeksema & Morrow, 1991); in this sample, Cronbach’s alpha was .94.

Diary items: Daily negative and positive experiences

Daily negative (“hassles”) and positive (“uplifts”) experiences were assessed based on methods of Totenhagen, Serido, Curran, and Butler (2012). Participants were given a list of items and asked to indicate how much of a hassle and an uplift each item was on that day on a scale of 0 (none) to 3 (a great deal). Items ranged across 15 general life domains: (a) parents and family members, (b) romantic life, (c) close friends, (d) other peers, (e) social events, (f) career, (g) finances, (h) exercise, (i) health, (j) chores, (k) hobbies, (l) extracurricular activities, (m) recreation, (n) online activities, and (o) other. An average total score was then computed for hassles and uplifts, respectively. Internal consistency for these and other daily measures was computed by separately computing Cronbach’s alphas for each of the fourteen days and then calculating the mean. Mean internal consistency for hassles and uplifts were .83 and .84, respectively. Hassles and uplifts scales in this sample were significantly associated with concurrent daily depressive symptoms, PE, and NE in expected directions, supporting construct validity (see also Li et al., 2016; Starr & Hershenberg, in press).

Diary items: Daily depressive symptoms

We assessed daily depressive symptoms using a modified version of the seven-item Depression subscale of the Depression Anxiety Stress Scale (DASS; Antony, Bieling, Cox, Enns, & Swinson, 1998). Items are rated on a Likert-type scale from 0 to 3. The DASS has demonstrated strong psychometric properties (Brown, Chorpita, Korotitsch, & Barlow, 1997; Clara, Cox, & Enns, 2001). The original items were modified so that the time frame indicates the current day (e.g., “Indicate how much the statement applied to you today”). Supporting validity, aggregated mean of daily DASS depressive symptoms was significantly correlated with baseline depressive symptoms (r = .61, p < .001). Mean internal consistency across individual days was .93.

Diary items: Daily brooding and savoring

Daily brooding was assessed using the five-item Brooding subscale of the RRS (Treynor et al., 2003), with instructions modified to cover brooding over the course of the current day. Each item was rated on a 4-point Likert-type scale. This scale, which has shown excellent psychometric properties in between-persons studies (Miranda & Nolen-Hoeksema, 2007; Moberly & Watkins, 2008; Treynor et al., 2003), has been previously adapted to assess daily brooding in daily diary research (Starr, 2015). Reports of daily brooding using this measure have been associated with baseline rumination measures and concurrent depressed mood (Li et al., 2016; Starr, 2015), supporting construct validity of the daily measure. Savoring was assessed using a shortened version of the Response to Positive Affect scale (RPA; Feldman, Joormann, & Johnson, 2008), with instructions prompting participants to consider how they have responded to feeling “happy, excited, or enthused” modified to apply to the current day only. Consistent with previous research suggesting that emotion-focused (EF) and self-focused (SF) positive rumination compose a single factor (Nelis et al., 2016), two items with the highest factor loadings were taken from the EF and SF subscales respectively to create a four-item daily savoring scale (e.g., “Think about how happy you feel”), each rated on a 4-point Likert-type scale. The full RPA scale has shown adequate internal consistency as well as convergent and incremental validity (Feldman et al., 2008; Raes, Daems, Feldman, Johnson, & Van Gucht, 2010). In support of the validity of the daily application of the RPA, Li et al. (2016) showed that baseline savoring was significantly associated with aggregated mean ratings of daily savoring (r = .50, p < .01). Mean daily Cronbach’s alpha for brooding and savoring were .86 and .83, respectively.

Positive and negative emotion differentiation

In the EMA survey, we calculated current PE and NE based on ratings on standard mood adjectives used in emotion reactivity research (replicating Bylsma et al., 2011; also see Hershenberg, Mavandadi, Wright, & Thase, 2017). Seven positive mood ratings (talkative, enthusiastic, confident, cheerful, energetic, satisfied, and happy) and seven negative mood ratings (tense, anxious, distracted, restless, irritated, depressed, guilty) were assessed on each call. To calculate ED, average intraclass correlation coefficients (ICCs) for either PE or NE items were calculated for each participant across all assessments (Shrout & Fleiss, 1979; Tugade et al., 2004). For ease of interpretation, ICCs were subtracted from 1.0 to reverse the score, so that higher NED/PED scores reflect greater ED ability and lower scores reflected lower differentiation of discrete emotions. This established, well-validated method of calculating ED has been used in multiple previous studies (e.g., Hill & Updegraff, 2012; Selby et al., 2014; Tugade & Fredrickson, 2007). Mean emotion intensity (used as a covariate) was calculated by taking the mean levels of NE and PE for each participant across all EMA observations.

Data analytic approach

We used multilevel modeling (MLM) using SPSS 23.0 MIXED. MLM is a powerful statistical approach that accounts for the nested, nonindependent nature of intensive longitudinal data. Repeated measures were nested within participants. Most hypotheses were tested using cross-level interactions (known as slopes-as-outcomes models) between a Level 1 (within-subjects) predictor (daily hassles, uplifts, brooding, or savoring) and a Level 2 (between-subjects) moderator (NED or PED). MLM copes well with missing data and has greater statistical power compared to traditional analytic approaches.

All predictors were entered as fixed effects, with Level 1 variables and the intercept also modeled as random effects. Level 2 predictors were mean centered. In addition, following the recent recommendations of Bolger and Laurenceau (2013), we partitioned each Level 1 predictor into two orthogonal components: a between-subjects component, represented by the person’s grand-mean-centered aggregated mean score over the course of the full diary period (i.e., a means component, ${\overset{⌣}{X}}_{\cdot j}$ ), and a within-subjects component, represented by the person-mean-centered score (i.e., the daily deviation from the means component, ${\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}$ ). Both within- and between-subjects components were included for all Level 1 predictors, both as main effects and in cross-level interactions. The inclusion of the between-subjects components of Level 1 variables enhances the interpretability of models by ensuring that results are not artifacts of individual differences in the average level of the variable (e.g., the tendency for some people to brood more than others). However, results for the between-subjects variables are not themselves considered interpretable (Bolger & Laurenceau, 2013). Thus, although results for between-subjects variables are included in Table 1 for the sake of comprehensiveness, readers should focus primarily on results for the “within” variables (e.g., brooding_within); for simplicity, “between” variables are not described in the Results section text. It is important that this caveat only applies to the between-subjects component of Level 1 variables; Level 2 variables (e.g., NED/PED) are inherently between-subjects and are interpretable. Note that use of this relatively new analytic approach did not substantially change results, as compared to a more traditional approach where between-subjects components of Level 1 variables were not included.

We also controlled for time in all models to ensure that effects were not artifacts of temporal change. We applied a first-order autoregressive (AR[1]) model to correct for autocorrelation of residuals and an unstructured covariance matrix for random effects. Daily depressive symptoms (concurrent to predictors) were entered as the outcome in all MLM models.

Following the notation of Bolger and Laurenceau (2013), moderation models (Hypotheses 1–4) can be described with the following equations:

Level 1:

Y_{i j} = β_{0 j} + β_{1 j} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}) + β_{2 j} (Time) + ε_{i j}

Level 2:

β_{0 j} = γ_{00} + γ_{01} W_{j} + γ_{02} {\overset{⌣}{X}}_{\cdot j} + γ_{03} W_{j} {\overset{⌣}{X}}_{\cdot j} + u_{0 j}

β_{1 j} = γ_{10} + γ_{11} W_{j} + u_{1 j}

β_{2 j} = γ_{20}

These equations can be simplified into the following equation, in which the first seven terms denote fixed effects and the last three terms represent random effects (see Bolger & Laurenceau, 2013):

\begin{array}{l} Y_{i j} = γ_{00} + γ_{01} W_{j} + γ_{02} {\overset{⌣}{X}}_{\cdot j} + γ_{03} W_{j} {\overset{⌣}{X}}_{\cdot j} \\ + γ_{10} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}) + γ_{11} W_{j} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}) \\ + γ_{20} (Time) + u_{0 j} + u_{1 j} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}) + ε_{i j} \end{array}

For example, in the model constructed to test Hypothesis 3, $W_{j}$ represents NED (the Level 2 moderator), ${\overset{⌣}{X}}_{\cdot j}$ and ${\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j}$ respectively represent main effects for brooding_between and brooding_within (the Level 1 predictor), and $W_{j} {\overset{⌣}{X}}_{\cdot j}$ and $W_{j} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j})$ respectively denote NED × brooding_between and NED × brooding_within interactions. In all moderation models, of primary interest is the significance of the sixth term in Equation 5, $W_{j} ({\overset{⌣}{X}}_{i j} - {\overset{⌣}{X}}_{\cdot j})$ , the interaction between ED and the within-subjects component of the Level 1 variable. Significant interactions were probed using simple slope tests (Aiken & West, 1991; Preacher, Curran, & Bauer, 2006) at one standard deviation above and below the moderator’s mean.

Finally, because individuals prone to more extreme (i.e., higher or lower) levels of negative and positive affect will have a restricted range of emotion ratings, we also controlled for mean emotion intensity (NE intensity in Hypotheses 1 and 3, PE intensity in Hypotheses 2 and 4). Doing so was a conservative approach to rule out any concerns related to a restricted range. Note that results remained unchanged when controlling for mean emotion intensity; thus, for parsimony, we do not present the models including this additional covariate.

Results

The Study 1 EMA data set (used for the computation of ED) included 3,029 valid observations. Average ICC levels were .17 for NE and .34 for PE, resulting in mean NED and PED (1-ICC) of .83 and .66 respectively.

The Study 1 daily diary data set included 1,723 valid observations. Descriptive data for study variables are reported in Table 1.

As a preliminary step, we examined bivariate correlations between NED, PED, baseline depressive symptoms (QIDS), and baseline rumination (RRS). We found a marginally significant, negative association between NED and QIDS (r = −.16, p = .056) but no correlation between PED and QIDS (p > .05). NED was significantly correlated with baseline RRS (r = .19, p = .017). NED and PED were also significantly correlated with each other (r = .19, p = .017).

Hypothesis 1

Next, we examined whether NED moderated the association between fluctuations in daily hassles and daily depressive symptoms. As described in the Data Analytic Approach section, a model was constructed that included main effects for daily hassles and NED as well as their interaction and the effects of time. Full results are reported in Table 2. The interaction between daily hassles and NED were not significant (p > .05). Although nonsignificant, the interaction is illustrated in Figure 1a-i to facilitate comparison with Study 2.

Table 2.

Results of Multilevel Models Predicting Daily Depressive Symptoms From Daily Variables, as Moderated by Negative and Positive Emotion Differentiation

	Study 1			Study 2
Variable	b (SE)	p	95% CI	b (SE)	p	95% CI
Hypothesis 1
Intercept	4.32 (0.35)	<.001	[3.63, 5.00]	1.77 (0.15)	<.001	[1.48, 2.07]
Neg. experiences_within	0.20 (0.03)	<.001	[0.14, 0.25]	0.21 (0.03)	<.001	[0.14, 0.28]
NED	−4.16 (2.48)	.096	[−9.06, 0.74]	−0.49 (0.51)	.357	[−1.49, 0.56]
NED × neg. experiences_within	−0.16 (0.22)	.468	[−0.60, 0.28]	−0.69 (0.16)	<.001	[−1.02, −0.41]
Neg. experiences_between	0.30 (0.06)	<.001	[0.18, 0.41]	0.70 (0.19)	.001	[0.32, 1.07]
NED × neg. experiences_between	0.85 (0.38)	.027	[0.10, 1.60]	−2.12 (1.31)	.112	[−4.76, 0.52]
Time	−0.06 (0.02)	.006	[3.63, 5.00]	0.00 (0.00)	.027	[0.00, 0.00]
Time of day^a	n/a			0.00 (0.00)	.902	[0.00, 0.00]

Hypothesis 2
Intercept	4.81 (0.38)	<.001	[4.07, 5.55]	1.81 (0.15)	<.001	[1.52, 2.10]
Pos. experiences_within	−0.10 (0.02)	<.001	[−0.14, −0.05]	−0.22 (0.04)	<.001	[−0.29, −0.14]
PED	1.99 (1.97)	.313	[−1.89, 5.88]	−0.07 (0.50)	.883	[−1.08, 0.94]
PED × pos. experiences_within	0.38 (0.13)	.004	[0.12, 0.64]	0.56 (0.17)	.002	[0.21, 0.90]
Pos. experiences_between	−0.03 (0.05)	.552	[−0.14, 0.07]	−0.52 (0.17)	.003	[−0.86, −0.19]
PED × pos. experiences_between	0.07 (0.27)	.803	[−0.46, 0.59]	−0.29 (0.65)	.657	[−1.58, 1.01]
Time	−0.14 (0.02)	<.001	[−0.19, −0.09]	0.00 (0.00)	.029	[0.00, 0.00]
Time of day^a	n/a			0.00 (0.00)	.778	[0.00, 0.00]

Hypothesis 3^b
Intercept	4.24 (0.28)	<.001	[3.69, 4.79]		n/a
Daily brooding_within	0.46 (0.04)	<.001	[0.38, 0.55]
NED	−1.80 (1.95)	.357	[−5.65, 2.05]
NED × brooding_within	−0.73 (0.33)	.028	[−1.38, −0.08]
Daily brooding_between	0.80 (0.07)	<.001	[0.65, 0.94]
NED × brooding_between	−0.45 (0.50)	.368	[−1.43, 0.54]
Time	−0.07 (0.02)	<.001	[−0.11, −0.04]

Hypothesis 4
Intercept	4.73 (0.35)	<.001	[4.04, 5.42]	1.94 (0.15)	<.001	[1.64, 2.24]
Daily savoring_within	−1.54 (0.22)	<.001	[−1.97, −1.11]	−0.16 (0.05)	.004	[−0.27, −0.05]
PED	2.74 (1.84)	.139	[−0.9, 6.39]	−0.09 (0.52)	.860	[−1.15, 0.96]
PED × daily savoring_within	4.15 (1.19)	.001	[1.79, 6.51]	0.48 (0.26)	.072	[−0.05, 1.01]
Daily savoring_between	−2.28 (0.58)	<.001	[−3.43, −1.12]	−0.08 (0.13)	.541	[−0.19, 0.35]
PED × daily savoring_between	−1.21 (0.26)	.672	[−7.50, 4.85]	−1.21 (0.58)	.029	[−2.30, −0.13]
Time	−0.13 (0.02)	<.001	[−0.17, −0.08]	0.00 (0.00)	.016	[0.00, 0.00]
Time of day^a	n/a			0.00 (0.00)	.279	[0.00, 0.00]

Note: Focal moderation findings are bolded. Daily depressive symptoms (depressed mood in Study 2) was the outcome variable in all models. Positive and negative experiences defined respectively as uplifts and hassles in Study 1 and pleasantness/unpleasantness of recent activities in Study 2.

Study 2 only.

Hypothesis 3 was not evaluated in Study 2.

Fig. 1.

Daily depressive symptoms as a function of (a) recent negative experiences (Hypothesis 1, with results from [i] Study 1 and [ii] Study 2) and (b) daily brooding (Hypothesis 3, Study 1 only), as moderated by negative emotion differentiation.

Hypothesis 2

Next, we tested whether low PED moderated the association between uplift fluctuations and daily depressive symptoms. Following the approach described earlier, we constructed a multilevel model with daily uplifts, NED, and their interaction, as well as time, with daily depressive symptoms entered as the outcome. Full results are reported in Table 2. As shown there, the main effect for uplifts (but not PED) was significant, p < .001. Of note, the interaction term was significant, p = .004. The significant interaction was decomposed using a simple slope tests (Aiken & West, 1991; Preacher et al., 2006). Supporting Hypothesis 2, as illustrated in Figure 2a-i, the negative association between daily uplifts and daily depressive symptoms was stronger when PED was low (M − 1 SD), b = −0.16, SE = 0.03, p < .001; in contrast, uplifts did not significantly predict reductions in depressive symptoms when PED was high (M + 1 SD), b = −0.03, SE = 0.03, p = .403.

Fig. 2.

Findings from (i) Study 1 and (ii) Study 2 showing daily depressive symptoms as a function of (a) recent positive experiences (Hypothesis 2) and (b) recent savoring (Hypothesis 4), as moderated by positive emotion differentiation.

Hypothesis 3

We next examined whether NED predicted stronger positive associations between daily brooding and daily depressive symptoms. NED and daily brooding were entered into a multilevel model as both main effects and an interactive effect, along with time, with daily depressive symptoms as the outcome. Full results are displayed in Table 2. The main effect for brooding was significant, p < .001, but not for NED. Supporting Hypothesis 3, the interaction between daily brooding fluctuations and NED was significant, p = .028. Decomposition is illustrated in Figure 1b. As expected, reports of daily brooding were more predictive of same-day depressive symptoms at low levels of NED, b = 0.56, SE = 0.06, t(136.65) = 9.21, p < .001, as compared to high levels, b = 0.37, SE = 0.06, p < .001.

Hypothesis 4

Finally, we examined whether PED likewise predicted stronger associations between daily savoring and reduced depressed mood. Analogous to the earlier presented models, we constructed a model that included PED, daily savoring, PED × daily savoring, and time. As shown in Table 2, the main effect for savoring (but not for PED) was significant, p < .001. Moreover, supporting Hypothesis 4, the interaction was significant, p = .001 and is presented in Figure 2b-i. As predicted, daily savoring more strongly predicted lower depressive symptoms for those with low PED, b = −2.26, SE = 0.30, p < .001, compared to those with high PED, b = −0.81, SE = 0.30, p = .008.

Conclusions and limitations

These results provide broad support for the majority of hypotheses (with the notable exception of Hypothesis 1; we found no support for NED conferring stress reactivity in this sample). Results should be interpreted in the context of study limitations. First, this study utilized an undergraduate sample. On one hand, young adults are at high risk for first onset of depression (Kessler, Berglund, Demler, Jin, & Walters, 2005), and research suggests that undergraduate samples generate comparable findings to clinical samples (e.g., Vredenburg et al., 1993). That said, some have criticized the overreliance on college samples to study clinical phenomena (Coyne, 1994). A higher risk sample could arguably produce findings more applicable to clinical depression. From a demographic standpoint, as with most studies recruiting from psychology courses, our sample had a limited age range and was predominantly female. This is potentially problematic because ED increases with age (Carstensen, Pasupathi, Mayr, & Nesselroade, 2000), and depression risk (and rumination) varies considerably by gender (Nolen-Hoeksema, Larson, & Grayson, 1999). Replicating findings in a sample with markedly different demographics and higher clinical risk would provide greater confidence in the generalizability of results.

Furthermore, all Study 1 models reflect concurrent associations (i.e., the Level 1 predictor variable was reported at the same time as the outcome variable). The daily diary design (with a full-day interval between surveys), while offering numerous benefits, was not well suited for testing lagged effects. The effects of daily events tend to quickly dissipate and are sometimes countered by mood rebound effects, making lagged findings often elusive when using daily diary designs (Bolger, DeLongis, Kessler, & Schilling, 1989; Stone, Neale, & Shiffman, 1993). Because we tested only concurrent models in Study 1, it is unclear if predictor variables preceded outcomes; the reverse direction of causality remains a possibility (for example, those with low NED may be more prone to generating negative experiences when depressed).

To address these limitations, we tested hypotheses in a second study, which included a sample of older, largely male veterans recruited from a primary care cohort being evaluated for behavioral health symptoms. Veterans are at substantially higher risk for mental health problems compared to the general population (Dohrenwend et al., 2006; Seal, Bertenthal, Miner, Sen, & Marmar, 2007), thus, depressive symptoms reported in this sample may be more reflective of clinically significant pathology. This sample used an experience sampling/EMA approach in which participants were signaled multiple times per day to complete surveys in naturalistic settings at random intervals. We tested the same hypotheses as in Study 1, with the exception of Hypothesis 3 (NED × brooding), for which data were not available. In addition, because the EMA approach allowed for shorter intervals between surveys, we were able to explore lagged models (predictor variables temporally preceding outcomes) in addition to testing concurrent models, analogous to the diary design.

Study 2

Method

Participants and procedure

We recruited veterans with a range of depression severity based on scores on the Patient Health Questionnaire−9 (PHQ-9; Kroenke, Spitzer, & Williams, 2001) from a primary care cohort being evaluated for behavioral health symptoms at a northeastern Veteran Affairs Medical Center. The Behavioral Health Laboratory (BHL) within the VAMC collects screening data on new veteran patient referrals on an ongoing basis. Potentially eligible veterans were identified by their responses to the PHQ-9 included in the BHL assessment; those eligible to participate and interested in hearing more about the study were contacted by study staff. Veterans were selectively recruited to achieve an approximate distribution of one third no depressive symptoms, one third minor depression, and one third major depression, according to PHQ-9 cut-offs (Kroenke et al., 2001). Exclusion criteria included psychotic disorders and current mania. Although sampling methods specifically targeted an overselection of depression, the sample was psychiatrically and medically heterogeneous; 1 participant (2%) had a probable past manic episode, 25 participants (50%) received a probable diagnosis of posttraumatic stress disorder, 34 participants (68%) indicated significant interference from pain, and four participants (8%) were considered at risk for alcohol problems.

Interested veterans were invited to a laboratory session, during which they provided a baseline measure of depressive symptoms with repeated administration of the PHQ-9 and received instructions on the EMA portion of the study, which began the following day. We collected EMA data using Interactive Voice Recording, a phone-based system for collecting data via keypad press. Participants were called six times per day for seven days. Calls occurred on a random basis within a 12-hr block of participant designated time (e.g., 9 a.m. to 9 p.m.). If they missed the call, participants were given 25 min to call back a toll-free number to complete the survey. We worked closely with participants during the baseline assessment to make sure that they understood the items being asked of them on the phone surveys. Adherence to the protocol was monitored, and we reached out to participants to troubleshoot noncompliance. Participants received $35 for the lab session and $2.50 per EMA call (max payment $140). Participants completed an average of 69% of calls (M = 29.10, SD = 10; modal number of calls 39 out of 42), which is comparable to other EMA studies (Bylsma et al., 2011). Rate of missed surveys was unrelated to baseline depression, NED, or PED. Demographic information is displayed in Table 1; as shown, veterans were predominantly male and racially diverse. The majority (72%) were older than 50 years of age. All study procedures were approved by the Corporal Michael J. Crescenz VAMC Institutional Review Board.

Baseline

Depressive symptoms

Participants rated the frequency over the past two weeks with which they experienced each of nine symptoms of depression using the PHQ-9 (Kroenke et al., 2001). Psychometric properties of the PHQ-9 are well established (Spitzer, Kroenke, Williams, & Group, 1999; Spitzer, Williams, Kroenke, Homyak, & McMurray, 2000), and Cronbach’s alpha in this study was .83. Baseline rumination was measured with the RRS (see Study 1).

Momentary positive and negative experiences

At the time of each phone call, participants were asked to report “how you were spending your time before you took a break to take this survey.” To rate the valence of their current activity, participants used a face-valid, continuous scale, ranging from 1 (most unpleasant) to 5 (most pleasant). We refer to this scale as “pleasant activities.” To facilitate comparisons with Study 1, for Hypothesis 1 only we reverse coded the scale, so that greater scores reflect more unpleasant recent experiences (which we refer to as “unpleasant activities”).

NED and PED

On each phone call, after rating the pleasantness of their current activity, participants were asked to “Keep thinking about how you felt before you took a break to take this survey. Use a 1 to 5 scale, where 1 is ‘I didn’t feel this way at all’ and 5 is ‘I felt this way a great deal.’” We used the same adjectives as in Study 1, and PED and NED were computed using identical procedures as in Study 1.

Momentary depressed mood

Depressed mood was assessed using the rating for the single item, “I felt depressed,” on the 1 to 5 continuous scale described earlier. The use of single-item indicators of mood is relatively common in EMA/diary research (as brevity is critical for survey compliance) and is psychometrically justifiable for noncomplex constructs (Burisch, 1997; Laurenceau, Barrett, & Rovine, 2005; Starr, 2015; Starr & Davila, 2012a).

Savoring

Similar to Study 1, savoring items were adapted from Feldman and colleagues’ (2008) RPA scale, although in this study, all items were taken from the EF scale of the RPA (no items from the SF scale were administered; we still refer to this scale as savoring for consistency with Study 1, but note that in this study it reflects the somewhat narrower construct of emotion-focused savoring). Participants were asked, “How have you responded to these feelings?” (i.e., their emotion ratings) using a 1 (not at all) to 5 (a great deal) scale. We administered three emotion-focused items, “I started to think about how happy I feel,” “I started to think about how strong I feel,” and “I savored this moment,” and took their average for a total savoring score. Mean Cronbach’s alpha (computed separately for each prompt, and then averaged) was .77.

Data analytic approach

Prior to data analysis, EMA data were inspected and suspicious response patterns (e.g., large numbers of identical numeric responses) were flagged for exclusion (see McCabe, Mack, & Fleeson, 2012). The analytic approach for Study 2 was similar to that in Study 1, with the following changes to accommodate the EMA design. A continuous variable representing time passed since the first completed survey was used as the repeated measures variable and was entered as a covariate to account for temporal artifacts. To control for possible effects of diurnal mood variation, time of day was also included in all concurrent models (as shown in Table 2, this variable was not significant in any models and was dropped from lagged models to for parsimony). As in Study 1, we ran additional models including mean emotion intensity as a covariate in tests of all hypotheses; findings were not substantially impacted, and results presented here exclude this covariate for simplicity. Depressed mood was entered as the outcome variable in all MLM analyses.

We tested both concurrent and lagged models in this sample. To create lagged variables, data for each signal were shifted in our data set, so that depressed mood at each signal t could be predicted by variables at signal t − 1. To prevent overnight lags, the first signal of each day was excluded as outcomes; because this reduced the amount of data available for analysis, power was correspondingly reduced in lagged analyses. Moreover, effects are generally weaker in lagged analyses (because effects of within-day events are typically short-lived; Marco & Suls, 1993). Consequently, we consider lagged models exploratory, as they may be somewhat underpowered and should be interpreted in conjunction with concurrent models. Time lag between observations was included as a covariate in lagged models to account for nonequal intervals.

Results

The Study 2 data set included 1,455 valid observations. Table 1 displays descriptive data for major study variables. In this data set, mean NE ICC was 0.26 (M NED = 1-ICC = 0.74, SD = 0.20) and mean PE ICC was 0.44 (M PED = 0.56, SD = 0.44). Baseline depressive symptoms were significantly, negatively correlated with NED (r = −.29, p = .040), but not with PED (r = −.23, p = .116). NED was significantly, negatively correlated with baseline rumination (r = −.31, p = .033). NED and PED were also significantly correlated with each other (r = .40, p = .004).

Hypothesis 1

We next examined whether low NED predicted stronger associations between unpleasantness ratings of recent activities and depressed mood. Following the prescribed data analytic plan, we tested a model that included unpleasant activities, NED, and their interaction, as well as total elapsed time, time of day, and between-subjects components of unpleasant activities as covariates. Full results appear in Table 2. As shown, there was a main effect for negative experiences, but not for NED. It is important that the NED × unpleasant experiences interaction was significant (p < .001). Decomposition revealed that, in line with expectations, at low levels of NED, unpleasant activities were strongly related to depressed mood, b = 0.35, SE = 0.04, p < .001, whereas at high levels of NED, unpleasant activities were not significantly predictive of depressed mood, b = 0.07, SE = 0.05, p = .150 (see Figure 1a-ii for illustration).

We also tested this hypothesis using lagged data, where NED was examined as a moderator of the lagged association between unpleasant activities at the previous signal and current depressed mood.¹ The interaction was again significant, b = −0.50, SE = 0.22, p = .031. At low levels of NED, negative experiences predicted nonsignificant increases in depressed mood, b = 0.08, SE = 0.06, p = .171, whereas at high levels, negative experiences actually predicted marginal decreases in depressed mood, b = −0.12, SE = 0.07, p = .096.

Hypothesis 2

We next tested PED as a moderator of the association between pleasantness ratings of recent activities and depressed mood, by testing a model that included pleasant activities, PED, PED × pleasant activities, and time and between-subjects component covariates. As shown in Table 2, the main effect for positive experiences (but not PED) was significant. Notably, the focal interaction term was significant. We probed the significant interaction, as shown in Figure 2a-ii, and found that, consistent with hypotheses and Study 1 results, pleasant activities sharply predicted reduced depressed mood at low levels of PED, b = −0.33, SE = 0.05, p < .001, but only marginally predicted lower depressed mood at high levels of PED, b = −0.10, SE = 0.05, p = .071. Testing the same hypothesis using lagged data, we again found a significant interaction, b = 0.58, SE = 0.20, p = .020, with a similar pattern of results, with positive experiences predicting decreased later depressed mood only at low levels of PED.²

Hypothesis 4

Finally, we examined whether PED predicts strengthened associations between daily savoring and depressed mood. We constructed a multilevel model where daily savoring, PED, and their interaction predicted concurrent depressed mood, with total elapsed time, time of day, and between-subjects savoring components as covariates. Full results are listed in Table 2; as displayed there, the main effect for savoring was significant. The interaction between PED and savoring was marginally significant, p = .072 (it is interesting that this interaction was significant in supplemental analyses controlling for the covariate of mean emotional intensity, p = .046). Aligning with predictions and with Study 1 findings, daily savoring predicted significantly decreased concurrent depressed mood among those with low PED, b = −0.27, SE = 0.07, p = .001, but not among those with high PED, b = −0.06, SE = 0.08, p = .459 (see Figure 2b-ii). When substituting lagged depressed mood as the outcome, the interaction was nonsignificant (p = .113) though trended similarly such that momentary savoring was predictive of decreases in depressed mood among those with lower levels of PED.

General Discussion

Drawing from two intensive longitudinal studies with complementary designs and samples, our findings provide evidence for roles of both NED and PED in modulating within-person variations in depressive symptoms in daily life. Results largely replicated across both studies, despite substantial differences in sample demographics; although both samples were oversampled for depressive symptoms, Study 1 featured undergraduate, predominately female young adults and Study 2 included veterans recruited from a primary care cohort being evaluated for behavioral health symptoms who were predominantly males older than 50. The consistency of findings across such divergent samples supports the robustness and generalizability of our pattern of results.

First, we found evidence that NED (but not PED) was correlated with depressive symptoms, aligning with other studies that have linked low NED to depressive symptoms and disorders (Demiralp et al., 2012; Erbas et al., 2014; Golston et al., 1992). This finding is consistent with research linking depression to alexithymia, a personality trait marked in part by blunted ability to identify and describe one’s emotions (Honkalampi, Hintikka, Tanskanen, Lehtonen, & Viinamäki, 2000). Individuals with depression show a broad range of deficits in cognitive functioning, including overgeneralized autobiographical memory, negative attentional biases, reduced pro-spective imagery, and impaired executive functioning (for a review, see Joormann & Arditte, 2014), which may cause depressed people to perceive NE in a blunt, overgeneralized manner. Discrete emotional states provide a wealth of information about the nature of the emotional situation, its potential consequences, and effective regulation strategies (Parrott, 2002); decreased ability to discern nuance in NE experiences puts depressed individuals at an inherent disadvantage for managing these emotions. Furthermore, the act of affect labeling may be implicitly regulatory (e.g., Kircanski et al., 2012; Lieberman et al., 2011). That said, our data (as with other existing studies) cannot distinguish between two important possibilities: that low NED is a preexisting risk factor for the emergence of depression, or that low NED is a feature of depression that emerges with onset and improves with remission. Future research using long-term longitudinal design can help to resolve this important question.

The current research does provide evidence that ED (both negative and positive) influences the ebb and flow of depressive symptoms. First, the present study is the first to show that poor NED predicts a stronger association between daily negative experiences and daily depressive symptoms. This suggests that the effects of NE differentiation may be more salient in the context of negative experiences. This finding is consistent with the affect-as-information theory, which posits that specific, differentiated NE experiences are more adaptive than global affective states because they are subject to fewer misattribution errors (Schwarz & Clore, 1996). When people have the ability to discriminate between discrete emotional states elicited by a negative event, they are better able to identify the cause of the emotional experience and generate an adaptive response, whereas those who experience undifferentiated, global affective states cannot (Russell & Barrett, 1999). This capacity may be particularly relevant to depressive symptoms because the meaning attributed to adverse events (e.g., interpreting them as internal, global, and stable) is often associated with the etiology and maintenance of symptoms (Abramson, Metalsky, & Alloy, 1989; Hankin, Fraley, & Abela, 2005; Weiner, 1974). Moreover, greater NE differentiation is correlated with the selection and implementation of effective emotion regulation strategies (Barrett et al., 2001; Kashdan et al., 2015), and the act of labeling emotions may itself help regulate emotions (Kircanski et al., 2012; Lieberman et al., 2007; Lieberman et al., 2011). Accordingly, decreased differentiation may lead to more difficulties coping with negative experiences and their resulting emotions and increased daily depressive symptoms.

Moreover, NED moderated the associations between daily negative events and depressive symptoms in our veteran sample (Study 2), but not in our sample of young adults (Study 1). This discrepant finding may be associated with the demographic differences of the two samples. Previous research suggests that ED improves as adults age, so that poor differentiation of NEs may confer more risk for depressive symptoms in older adults (Carstensen et al., 2000). This explanation is speculative, and future work will help to address if this difference reflects the different demographics of the study samples.

Low NED also predicted stronger associations between daily brooding and concurrent depressive symptoms. Coupled with the negative correlation between NED and trait rumination, findings suggest that people with low NED suffer a rumination “double whammy,” with a greater tendency toward rumination as well as greater depressive reactivity to it. A number of studies have suggested that low NED is related to the selection and deployment of ineffective or destructive coping strategies. The current study is the first to show that maladaptive emotion regulation techniques also have more deleterious affective consequences among those with low NED. Findings align with those of Zaki et al. (2013), who showed that trait rumination was more predictive of nonsuicidal self-injury in borderline personality disorder for those with low NED. People who have trouble understanding their emotions may be more likely to get especially “stuck” when brooding about their NE, which may lead the brooding to have more of a deleterious impact on effective problem solving. In contrast, it could be the case that those with high NED may be more effective at identifying the source of their emotions, which may make ruminative self-focus more productive and less impactful on depressive symptoms.

Paralleling our NED findings, we also showed that, across both samples, low PED enhanced the ameliorative effects of both positive experiences and positive rumination (savoring) on depressive symptoms. Most of the limited previous research on PED has focused on the role of low PED in predicting problematic behaviors (e.g., Dixon-Gordon et al., 2014; Selby et al., 2014); we showed that low PED might be associated with adaptive outcomes in the context of depressive symptoms. Indeed, the extent to which low PED is harmful versus helpful may largely depend on the outcome being considered. Among those with low PED, positive experiences may trigger a wealth of PEs, rather than a specific, constrained emotion. In some contexts, these broad, undifferentiated PEs may motivate and reinforce destructive behaviors (e.g., Dixon-Gordon et al., 2014; Selby et al., 2014). However, in the context of depression, these PEs may motivate more behavioral activation, culminating in reduced depressive symptoms. Thus, PED may have trade-offs; it may be protective in some clinical contexts while adding to vulnerabilities in others. That said, research on PED remains very limited, and more investigation is decidedly needed.

Theoretical Considerations, Future Directions, and Conclusions

Can emotion differentiation be differentiated?

Notably, although research has typically assumed that low ED scores reflect an inability to cognitively discriminate between emotions, they may also reflect a genuine propensity toward experiencing multiple emotions in clusters rather than individually. If so, perseverative focus on one NE (through brooding) or PE (through savoring) may activate a network of related emotions, amplifying the affective experience and influencing depressive symptoms; likewise, positive and negative events may trigger a broader (and perhaps more intense) affective response in individuals labeled as low emotion differentiators. This may be one reason why the literature on PED is somewhat mixed; low PED may actually be composed of two counteractive components, one potentially beneficial (the propensity to experience multiple concurrent PEs) and one potentially maladaptive (deficient ability to discern discrete emotional states). In contrast, low NED may be more consistently maladaptive because both the tendency to experience large clusters of NEs and the inability to differentiate these NEs are likely to have negative consequences (and may therefore have additive effects). Although our approach to assessing NED and PED (using EMAs to calculate intraclass correlations among momentary emotions) is a widely accepted method (Selby et al., 2014; Shrout & Fleiss, 1979; Tugade et al., 2004), future researchers should develop techniques that better discriminate between the experience and the discernment of multiple concurrent emotions.

Clinical implications

Our findings on how NED and PED influence fluctuations of depressive symptoms in daily life may be helpful in increasing understanding of the dynamics of emotional experience in depression and inform the application of effective treatment targets. Evidence has pointed toward the efficacy of interventions that help individuals expand their emotion vocabulary to better identify and precisely label discrete emotions (Cameron, Payne, & Doris, 2013; Kircanski et al., 2012). This approach, known as affect labeling, suggests that better recognizing and naming discrete emotional states reduces emotional reactivity and maladaptive emotion regulation strategies (Lieberman et al., 2007). This, in turn, may facilitate the ability to manage one’s behavior or distress in response to negative experiences, thereby reducing depressive symptoms. Moreover, while working to improve NED, clinicians might help patients harness the effects of low PED as they work to increase patients’ awareness of PE states and increased sensitivity to rewards, though future empirical work is needed to see if this could be a useful application of our basic research findings.

Study limitations

When considered in isolation, each study has several limitations. Study 1 utilized an undergraduate sample; although participants were oversampled for depressive symptoms, the pathology captured is likely a nonideal proxy for clinical depression. Study 1 also used a daily diary design that relied on recall of experiences, perseverative thought, mood, and symptoms over the course of the day, and all analyses tested concurrent associations. Study 2 relied on single-item measures and a relatively small sample. However, these limitations were well balanced across the two studies. Study 2 included a sample of veterans (a higher-risk population with a wider and older age range) and shorter interval EMA sampling which allowed lagged analyses; Study 1 included more detailed daily measures, a larger sample, and a greater representation of women. Thus, the replication of results across the two studies provides greater confidence in findings. That said, several limitations remain across our study as a whole. Although both studies oversampled for self-reported depressive symptoms, neither utilized clinical interviews to assess clinical depression. Likewise, as Study 1 utilized a college sample and Study 2 included a psychiatrically heterogeneous group of veterans, neither sample can be considered exclusively representative of major depression, and future research can help to establish if results generalize. Moreover, as explained previously, our operationalization of ED cannot tease apart if low ED reflects an inability to discriminate between emotions or a genuine propensity to experience multiple emotions concurrently. In addition, both studies relied on the same emotion adjectives (derived from Bylsma et al., 2011), which were not explicitly designed to assess ED and may capture nondiscrete emotional states. Recent evidence underscores the importance of careful selection of emotion adjectives in momentary research, and this should be an important consideration in future ED research. As noted previously, our design cannot determine whether NED contributes to risk for depression or simply emerges concomitant with symptoms. Finally, our measures of daily preservative self-focus, particularly savoring, were brief and may have conceptual overlap with other related constructs (e.g., mindfulness), suggesting that future research will benefit from testing the specificity of these established associations.

Despite these limitations, the current study adds to a growing literature highlighting the importance of ED to emotional health. Indeed, evidence points to low NED as a transdiagnostic factor that contributes to a range of psychopathological conditions and behaviors including depression, eating disorders, social anxiety disorder, alcohol abuse, and aggression. In contrast, low PED may have trade-offs for mental health, depending on the clinical context. Future research is needed to better understand the nature and consequences of these intriguing constructs.

Footnotes

Acknowledgements

We thank the members of our research team, particularly Fanny Mlawer and Chistopher Anzalone, who managed data collection for Study 1, as well as the participants who generously contributed their time. We thank Erin Wright and Sara Mooar for help with data collection, for Study 2 and Dorothy McDougall and Joan Havey for administrative support, Christopher Petro for technical support, and the veterans who generously participated in this research.

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Funding

Study 1 was completed with funds provided by the University of Rochester. Study 2 was completed with funds provided by the VISN 4 Mental Illness Research, Education, and Clinical Center (MIRECC director D. Oslin) Pilot Project Funds, Corporal Michael J. Crescenz Veterans Affairs Medical Center, Philadelphia, PA. The views expressed in the article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the U.S. government.

Notes

References

Abramson

L. Y.

Metalsky

G. I.

Alloy

L. B.

(1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96, 358–372. doi:0033–295X/89/J00.15.

Aiken

L. S.

West

S. G.

(1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage.

Allison

(2015). Don’t put lagged dependent variables in mixed models. Retrieved from http://statisticalhorizons.com/lagged-dependent-variables

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Antony

M. M.

Bieling

P. J.

Cox

B. J.

Enns

M. W.

Swinson

R. P.

(1998). Psychometric properties of the 42-item and 21-item versions of the Depression Anxiety Stress Scales in clinical groups and a community sample. Psychological Assessment, 10, 176–181.

Barrett

L. F.

Gross

Christensen

T. C.

Benvenuto

(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, 713–724.

Bolger

DeLongis

Kessler

R. C.

Schilling

E. A.

(1989). Effects of daily stress on negative mood. Journal of Personality and Social Psychology, 57, 808–818. doi:10.1037/0022–3514.57.5.808

Bolger

Laurenceau

J.-P.

(2013). Intensive longitudinal methods: An introduction to diary and experience sampling research. New York, NY: Guilford.

Brown

T. A.

Chorpita

B. F.

Korotitsch

Barlow

D. H.

(1997). Psychometric properties of the Depression Anxiety Stress Scales (DASS) in clinical samples. Behaviour Research and Therapy, 35, 79–89.

10.

Bryant

(2003). Savoring Beliefs Inventory (SBI): A scale for measuring beliefs about savouring. Journal of Mental Health, 12, 175–196.

11.

Burisch

(1997). Test length and validity revisited. European Journal of Personality, 11, 303–315. doi:10.1002/(Sici)1099–0984(199711)11:4<303::Aid-Per292>3.0.Co;2–#

12.

Butler

L. D.

Nolen-Hoeksema

(1994). Gender differences in responses to depressed mood in a college sample. Sex Roles, 30, 331–346.

13.

Bylsma

L. M.

Morris

B. H.

Rottenberg

(2008). A meta-analysis of emotional reactivity in major depressive disorder. Clinical Psychology Review, 28, 676–691. doi:10.1016/j.cpr.2007.10.001

14.

Bylsma

L. M.

Taylor-Clift

Rottenberg

(2011). Emotional reactivity to daily events in major and minor depression. Journal of Abnormal Psychology, 120, 155–167. doi:10.1037/a0021662

15.

Cameron

C. D.

Payne

B. K.

Doris

J. M.

(2013). Morality in high definition: Emotion differentiation calibrates the influence of incidental disgust on moral judgments. Journal of Experimental Social Psychology, 49, 719–725. doi:10.1016/j.jesp.2013.02.014

16.

Carstensen

L. L.

Pasupathi

Mayr

Nesselroade

J. R.

(2000). Emotional experience in everyday life across the adult life span. Journal of Personality and Social Psychology, 79, 644–655.

17.

Clara

I. P.

Cox

B. J.

Enns

M. W.

(2001). Confirmatory factor analysis of the Depression-Anxiety-Stress Scales in depressed and anxious patients. Journal of Psychopathology and Behavioral Assessment, 23, 61–67. doi:10.1023/A:1011095624717

18.

Coyne

J. C.

(1994). Self-reported distress: Analog or ersatz depression? Psychological Bulletin, 116, 29–45.

19.

Demiralp

Thompson

R. J.

Mata

Jaeggi

S. M.

Buschkuehl

Barrett

L. F.

. . . Jonides

(2012). Feeling blue or turquoise? Emotional differentiation in major depressive disorder. Psychological Science, 23, 1410–1416. doi:10.1177/0956797612444903

20.

Dixon-Gordon

K. L.

Chapman

A. L.

Weiss

N. H.

Rosenthal

M. Z.

(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:10.1007/s10862–014–9423–4

21.

Dohrenwend

B. P.

Turner

J. B.

Turse

N. A.

Adams

B. G.

Koenen

K. C.

Marshall

(2006). The psychological risks of Vietnam for U.S. veterans: A revisit with new data and methods. Science, 313, 979–982. doi:10.1126/science.1128944

22.

Erbas

Ceulemans

Boonen

Noens

Kuppens

(2013). Emotion differentiation in autism spectrum disorder. Research in Autism Spectrum Disorders, 7, 1221–1227. doi:10.1016/j.rasd.2013.07.007

23.

Erbas

Ceulemans

Lee Pe

Koval

Kuppens

(2014). Negative emotion differentiation: Its personality and well-being correlates and a comparison of different assessment methods. Cognition & Emotion, 28, 1196–1213. doi:10.1080/02699931.2013.875890

24.

Feldman

G. C.

Joormann

Johnson

S. L.

(2008). Responses to positive affect: A self-report measure of rumination and dampening. Cognitive Therapy and Research, 32, 507–525.

25.

Fernandez

K. C.

Johnson

M. R.

Rodebaugh

T. L.

(2013). TelEMA: A low-cost and user-friendly telephone assessment platform. Behavior Research Methods, 45, 1279–1291.

26.

Fredrickson

B. L.

(1998). What good are positive emotions? Review of General Psychology, 2, 300–319.

27.

Fredrickson

B. L.

(2001). The role of positive emotions in positive psychology: The broaden-and-build theory of positive emotions. American Psychologist, 56, 218–226.

28.

Garlow

S. J.

Rosenberg

Moore

J. D.

Haas

A. P.

Koestner

Hendin

Nemeroff

C. B.

(2008). Depression, desperation, and suicidal ideation in college students: Results from the American Foundation for Suicide Prevention College Screening Project at Emory University. Depression and Anxiety, 25, 482–488. doi:10.1002/da.20321

29.

Golston

R. B.

Gara

M. A.

Woolfolk

R. L.

(1992). Emotion differentiation: A correlate of symptom severity in major depression. Journal of Nervous and Mental Disease, 180, 712–718.

30.

Gonzalez

D. A.

Boals

Jenkins

S. R.

Schuler

E. R.

Taylor

(2013). Psychometrics and latent structure of the IDS and QIDS with young adult students. Journal of Affective Disorders, 149, 217–220.

31.

Hankin

B. L.

Fraley

R. C.

Abela

J. R. Z.

(2005). Daily depression and cognitions about stress: Evidence for a traitlike depressogenic cognitive style and the prediction of depressive symptoms in a prospective daily diary study. Journal of Personality and Social Psychology, 88, 673–685.

32.

Hershenberg

Mavandadi

Wright

Thase

M. E.

(2017). Anhedonia in the daily lives of depressed veterans: A pilot report on experiential avoidance as a moderator of emotional reactivity. Journal of Affective Disorders, 208, 414–417.

33.

Hill

C. L.

Updegraff

J. A.

(2012). Mindfulness and its relationship to emotional regulation. Emotion, 12, 81–90.

34.

Honkalampi

Hintikka

Tanskanen

Lehtonen

Viinamäki

(2000). Depression is strongly associated with alexithymia in the general population. Journal of Psychosomatic Research, 48, 99–104.

35.

Hurley

D. B.

Kwon

(2012). Results of a study to increase savoring the moment: Differential impact on positive and negative outcomes. Journal of Happiness Studies, 13, 579–588.

36.

Joormann

Arditte

(2014). Cognitive aspects of depression. In Gotlib

I. H.

Hammen

C. L.

(Eds.), Handbook of depression (3rd ed., pp. 259–276). New York, NY: Guilford.

37.

Kashdan

T. B.

Barrett

L. F.

McKnight

P. E.

(2015). Unpacking emotion differentiation: Transforming unpleasant experience by perceiving distinctions in negativity. Current Directions in Psychological Science, 24, 10–16. doi:10.1177/0963721414550708

38.

Kashdan

T. B.

DeWall

C. N.

Masten

C. L.

Pond

R. S.

Jr. Powell

Combs

. . . Farmer

A. S.

(2014). Who is most vulnerable to social rejection? The toxic combination of low self-esteem and lack of negative emotion differentiation on neural responses to rejection. PLoS ONE, 9, e90651. doi:10.1371/journal.pone.0090651

39.

Kashdan

T. B.

Farmer

A. S.

(2014). Differentiating emotions across contexts: Comparing adults with and without social anxiety disorder using random, social interaction, and daily experience sampling. Emotion, 14, 629–638.

40.

Kashdan

T. B.

Ferssizidis

Collins

R. L.

Muraven

(2010). Emotion differentiation as resilience against excessive alcohol use an ecological momentary assessment in underage social drinkers. Psychological Science, 21, 1341–1347.

41.

Kessler

R. C.

Berglund

Demler

Jin

Walters

E. E.

(2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey replication. Archives of General Psychiatry, 62, 593–602. doi:10.1001/archpsyc.62.6.593

42.

Kircanski

Lieberman

M. D.

Craske

M. G.

(2012). Feelings into words: Contributions of language to exposure therapy. Psychological Science, 23, 1086–1091.

43.

Kovacs

Bylsma

L. M.

Yaroslavsky

Rottenberg

George

C. J.

Kiss

. . . Kapornai

(2016). Positive affectivity is dampened in youths with histories of major depression and their never-depressed adolescent siblings. Clinical Psychological Science, 4, 661–674. doi:10.1177/2167702615607182

44.

Kring

A. M.

Sloan

D. M.

(2011). Emotion regulation and psychopathology: A transdiagnostic approach to etiology and treatment. New York, NY: Guilford.

45.

Kroenke

Spitzer

R. L.

Williams

J. B.

(2001). The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine, 16, 606–613.

46.

Laurenceau

J.-P.

Barrett

L. F.

Rovine

M. J.

(2005). The interpersonal process model of intimacy in marriage: A daily-diary and multilevel modeling approach. Journal of Family Psychology, 19, 314–323. doi:10.1037/0893–3200.19.2.314

47.

Lewinsohn

P. M.

Graf

(1973). Pleasant activities and depression. Journal of Consulting and Clinical Psychology, 41, 261–268. doi:10.1037/h0035142

48.

Lewinsohn

P. M.

Sullivan

J. M.

Grosscup

S. J.

(1980). Changing reinforcing events: An approach to the treatment of depression. Psychotherapy: Theory, Research & Practice, 17, 322–334. doi:10.1037/h0085929

49.

Y. I.

Starr

L. R.

Hershenberg

(in press). Responses to positive affect in daily life: Positive rumination and dampening moderate the association between daily events and depressive symptoms. Journal of Psychopathology and Behavioral Assessment.

50.

Lieberman

M. D.

Eisenberger

N. I.

Corckett

M. J.

Tom

S. M.

Pfeifer

J. H.

Way

B. M.

(2007). Putting feelings into words: Affect labeling disrupts amygdala activity in response to affective stimuli. Psychological Science, 18, 421–428.

51.

Lieberman

M. D.

Inagaki

T. K.

Tabibnia

Crockett

M. J.

(2011). Subjective responses to emotional stimuli during labeling, reappraisal, and distraction. Emotion, 8, 307–301.

52.

Maniaci

M. R.

Rogge

R. D.

(2014). Caring about carelessness: Participant inattention and its effects on research. Journal of Research in Personality, 48, 61–83. doi:10.1016/j.jrp.2013.09.008

53.

Marco

C. A.

Suls

(1993). Daily stress and the trajectory of mood: Spillover, response assimilation, contrast, and chronic negative affectivity. Journal of Personality and Social Psychology, 64, 1053–1063.

54.

Martin

L. L.

Tesser

(1996). Some ruminative thoughts. Advances in Social Cognition, 9, 1–47.

55.

McCabe

K. O.

Mack

Fleeson

(2012). A guide for data cleaning in experience sampling studies. In Mehl

M. R.

Conner

T. S.

(Eds.), Handbook of research methods for studying daily life (pp. 321–338). New York, NY: Guilford.

56.

Miranda

Nolen-Hoeksema

(2007). Brooding and reflection: Rumination predicts suicidal ideation at 1-year follow-up in a community sample. Behaviour Research and Therapy, 45, 3088–3095.

57.

Moberly

N. J.

Watkins

E. R.

(2008). Ruminative self-focus and negative affect: An experience sampling study. Journal of Abnormal Psychology, 117, 314–323.

58.

Nelis

Luyckx

Feldman

Bastin

Raes

Bijttebier

(2016). Assessing response styles to positive affect: One or two dimensions of positive rumination in the Responses to Positive Affect questionnaire? Personality and Individual Differences, 89, 40–46.

59.

Nolen-Hoeksema

Larson

Grayson

(1999). Explaining the gender difference in depressive symptoms. Journal of Personality and Social Psychology, 77, 1061–1072.

60.

Nolen-Hoeksema

Morrow

(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, 115–121.

61.

Nolen-Hoeksema

Wisco

B. E.

Lyubomirsky

(2008). Rethinking rumination. Perspectives on Psychological Science, 3, 400–424.

62.

O’Toole

M. S.

Jensen

M. B.

Fentz

H. N.

Zachariae

Hougaard

(2014). Emotion differentiation and emotion regulation in high and low socially anxious individuals: An experience-sampling study. Cognitive Therapy and Research, 38, 428–438. doi:10.1007/s10608–014–9611–2

63.

Parrott

W. G.

(2002). The functional utility of negative emotions. In Barrett

L. F.

Salovey

(Eds.), The wisdom in feeling: pysychological processes in emotional intelligence (pp. 341–359). New York, NY: Guilford.

64.

Peeters

Nicolson

N. A.

Berkhof

Delespaul

deVries

(2003). Effects of daily events on mood states in major depressive disorder. Journal of Abnormal Psychology, 112, 203–211. doi:10.1037/0021–843X.112.2.203

65.

Pizzagalli

D. A.

Holmes

A. J.

Dillon

D. G.

Goetz

E. L.

Birk

J. L.

Bogdan

. . . Fava

(2009). Reduced caudate and nucleus accumbens response to rewards in unmedicated individuals with major depressive disorder. American Journal of Psychiatry, 166, 702–710. doi:10.1176/appi.ajp.2008.08081201

66.

Pond

R. S. J.

Kashdan

T. B.

DeWall

C. N.

Savostyanova

Lambert

N. M.

Fincham

F. D.

(2012). Emotional differentiation moderates aggressive tendencies in angry people: A daily diary analysis. Emotion, 12, 326–337.

67.

Preacher

K. J.

Curran

P. J.

Bauer

D. J.

(2006). Computational tools for probing interaction effects in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of Educational and Behavioral Statistics, 31, 437–448.

68.

Quigley

K. S.

Barrett

L. F.

(1999). Emotional learning and mechanisms of intentional psychological change. In Brandtstädter

Lerner

R. M.

(Eds.), Action and self-development: Theory and research through the life span (pp. 435–464). Thousand Oaks, CA: Sage.

69.

Quoidbach

Berry

E. V.

Hansenne

Mikolajczak

(2010). Positive emotion regulation and well-being: Comparing the impact of eight savoring and dampening strategies. Personality and Individual Differences, 49, 368–373.

70.

Raes

Daems

Feldman

Johnson

Van Gucht

(2010). A psychometric evaluation of the Dutch version of the Responses to Positive Affect Questionnaire. Psychologica Belgica, 49, 293–310.

71.

Rottenberg

(2007). Major depressive disorder: Emerging evidence for emotion context insensitivity. In Rottenberg

Johnson

S. L.

(Eds.), Emotion and psychopathology: Bridging affective and clinical science (pp. 151–166). Washington, DC: American Psychological Association.

72.

Rush

A. J.

Bernstein

I. H.

Trivedi

M. H.

Carmody

T. J.

Wisniewski

Mundt

J. C.

. . . Nierenberg

A. A.

(2006). An evaluation of the Quick Inventory of Depressive Symptomatology and the Hamilton Rating Scale for Depression: A sequenced treatment alternatives to relieve depression trial report. Biological Psychiatry, 59, 493–501. doi:10.1016/j.biopsych.2005.08.022

73.

Rush

A. J.

Trivedi

M. H.

Ibrahim

H. M.

Carmody

T. J.

Arnow

Klein

D. N.

. . . Manber

(2003). The 16–Item Quick Inventory of Depressive Symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): A psychometric evaluation in patients with chronic major depression. Biological Psychiatry, 54, 573–583. doi:10.1016/S0006–3223(02)01866–8

74.

Russell

J. A.

Barrett

L. F.

(1999). Core affect, prototypical emotional episodes, and other things called emotion: Dissecting the elephan. Journal of Personality and Social Psychology, 76, 805–819.

75.

Schwarz

Clore

G. L.

(1996). Feelings and phenomenal experiences. In Higgins

E. T.

Kruglanski

A. W.

(Eds.), Social psychology: A handbook of basic principles (pp. 433–465). New York, NY: Guilford.

76.

Seal

K. H.

Bertenthal

Miner

C. R.

Sen

Marmar

(2007). Bringing the war back home: Mental health disorders among 103,788 US veterans returning from Iraq and Afghanistan seen at Department of Veterans Affairs facilities. Archives of Internal Medicine, 167, 476–482. doi:10.1001/archinte.167.5.476

77.

Selby

E. A.

Wonderlich

S. A.

Crosby

R. D.

Engel

S. G.

Panza

Mitchell

J. E.

. . . Le Grange

(2014). Nothing tastes as good as thin feels low positive emotion differentiation and weight-loss activities in anorexia nervosa. Clinical Psychological Science, 2, 514–531.

78.

Shrout

P. E.

Fleiss

J. L.

(1979). Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin, 86, 420–428.

79.

Spitzer

R. L.

Kroenke

Williams

J. B.

Group

P. H. Q. P. C. S

. (1999). Validation and utility of a self-report version of PRIME-MD: The PHQ primary care study. JAMA, 18, 1737–1744.

80.

Spitzer

R. L.

Williams

J. B.

Kroenke

Homyak

McMurray

(2000). Validity and utility of the PRIME-MD Patient Health Questionnaire in assessment of 3000 obstetric-gynecologic patients: The PRIME-MD Patient Health Questionnaire Obstetrics-Gynecology Study. American Journal of Obstetrics and Gynecology, 183, 759–769.

81.

Starr

L. R.

(2015). When support seeking backfires: Co-rumination, excessive reassurance seeking, and depressed mood in the daily lives of young adults. Journal of Social and Clinical Psychology, 34, 436–457. doi:10.1521/jscp.2015.34.5.436

82.

Starr

L. R.

Davila

(2012a). Cognitive and interpersonal moderators of daily co-occurrence of anxious and depressed moods in generalized anxiety disorder. Cognitive Therapy and Research, 36, 655–669. doi:10.1007/s10608–011–9434–3

83.

Starr

L. R.

Davila

(2012b). Temporal patterns of anxious and depressed mood in generalized anxiety disorder: A daily diary study. Behaviour Research and Therapy, 50, 131–141. doi:10.1016/j.brat.2011.11.005

84.

Starr

L. R.

Hershenberg

(in press). Depressive symptoms and the anticipation and experience of uplifting events in everyday life. Journal of Clinical Psychology.

85.

Stone

A. A.

Neale

J. M.

Shiffman

(1993). Daily assessments of stress and coping and their association with mood. Annals of Behavioral Medicine, 15, 8–16.

86.

Thompson

R. J.

Mata

Jaeggi

S. M.

Buschkuehl

Jonides

Gotlib

I. H.

(2012). The everyday emotional experience of adults with major depressive disorder: Examining emotional instability, inertia, and reactivity. Journal of Abnormal Psychology, 121, 819–829. doi:10.1037/a0027978

87.

Totenhagen

C. J.

Serido

Curran

M. A.

Butler

E. A.

(2012). Daily hassles and uplifts: A diary study on understanding relationship quality. Journal of Family Psychology, 26, 719–728. doi:10.1037/a0029628

88.

Treadway

M. T.

Zald

D. H.

(2011). Reconsidering anhedonia in depression: Lessons from translational neuroscience. Neuroscience and Biobehavioral Reviews, 35, 537–555. doi:10.1016/j.neubiorev.2010.06.006

89.

Treynor

Gonzalez

Nolen-Hoeksema

(2003). Rumination reconsidered: A psychometric analysis. Cognitive Therapy and Research, 27, 247–259.

90.

Trivedi

M. H.

Rush

Ibrahim

Carmody

Biggs

Suppes

. . . Dennehy

(2004). The Inventory of Depressive Symptomatology, Clinician Rating (IDS-C) and Self-Report (IDS-SR), and the Quick Inventory of Depressive Symptomatology, Clinician Rating (QIDS-C) and Self-Report (QIDS-SR) in public sector patients with mood disorders: A psychometric evaluation. Psychological Medicine, 34, 73–82.

91.

Tugade

M. M.

Fredrickson

B. L.

(2007). Regulation of positive emotions: Emotion regulation strategies that promote resilience. Journal of Happiness Studies, 8, 311–333.

92.

Tugade

M. M.

Fredrickson

B. L.

Barrett

L. F.

(2004). Psychological resilience and positive emotional granularity: Examining the benefits of positive emotions on coping and health. Journal of Personality, 72, 1161–1190. doi:10.1111/j.1467–6494.2004.00294.x

93.

Vredenburg

Flett

G. L.

Krames

(1993). Analogue versus clinical depression: A critical reappraisal. Psychological Bulletin, 113, 327–344. doi:10.1037/0033–2909.113.2.327

94.

Watson

Naragon-Gainey

(2010). On the specificity of positive emotional dysfunction in psychopathology: Evidence from the mood and anxiety disorders and schizophrenia/schizotypy. Clinical Psychology Review, 30, 839–848. doi:10.1016/j.cpr.2009.11.002

95.

Weinberg

Liu

Hajcak

Shankman

S. A.

(2015). Blunted neural response to rewards as a vulnerability factor for depression: Results from a family study. Journal of Abnormal Psychology, 124, 878–889.

96.

Weiner

(1974). Achievement motivation and attribution theory. Morristown, NJ: General Learning Press.

97.

Wood

J. V.

Heimpel

S. A.

Michela

J. L.

(2003). Savoring versus dampening: Self-esteem differences in regulating positive affect. Journal of Personality and Social Psychology, 85, 566–580.

98.

Zaki

L. F.

Coifman

K. G.

Rafaeli

Berenson

K. R.

Downey

(2013). Emotion differentiation as a protective factor against nonsuicidal self-injury in borderline personality disorder. Behavior Therapy, 44, 529–540. doi:10.1016/j.beth.2013.04.008