Unpacking affect maintenance and its association with depressive symptoms: integrating positive and negative affects

ABSTRACT

Depression is associated with increased maintenance of negative affect (NA) and reduced – blunted and short-lived – maintenance of positive affect (PA). Studies have focused on factors associated with the maintenance of NA, specifically, the emotion regulation strategy of brooding and the capacity to hold negative affective experiences in working memory (WM). Despite its theoretical importance, less attention has been given to factors associated with the maintenance of PA in depression. This study aims to synthesise factors playing a role in the maintenance of both NA and PA. Specifically, we used self-reported assessment of PA and NA regulation and performance-based measures of NA and PA processing in WM to predict depressive symptoms severity. Participants (N = 219) completed the Affective Maintenance Task (AMT, Mikels et al., 2008), which provided performance-based measures of PA and NA maintenance, and filled out questionnaires assessing brooding, positive rumination and depressive severity. Brooding, positive rumination and AMT-based measures of positive (but not negative) affective information processing were independently associated with depressive symptoms. We highlight the unique contributions of PA processing, as well as of self-reported emotion regulation strategies in understanding depression maintenance.

KEYWORDS:

• Depression
• emotion regulation
• brooding
• positive rumination
• affective working memory

Depression is among the most prevalent mental health problems worldwide, associated with a great cost (Kessler & Wang, 2009). Clark and Watson (1991) suggest that depressed individuals experience atypical affective responses to specific events, including both high intensity of NA and a rather mild intensity of PA (or manifest anhedonia). Similarly, Davidson (1998) has argued that prolonged depressive states are due to the maintenance of NA following adverse events as well as the failure to maintain PA following positive events. Despite the dual vulnerability – enhanced NA and reduced PA – models of depression, the empirical focus has centred on the experience and regulation of NA. The current study aims to synthesise multiple factors playing a role in the dual vulnerability to depression.

Within that focus, a considerable body of work has concentrated on a specific type of regulatory strategy of NA – rumination (e.g. Nolen-Hoeksema & Morrow, 1991). A subcomponent of rumination, brooding, involves repetitively focusing on the causes, meanings and consequences of one's NA in a way that does not lead to active problem-solving but instead prolongs NA (Nolen-Hoeksema, 1991). Brooding is typically assessed by a self-report measure (Nolen-Hoeksema & Morrow, 1991) and is found to be strongly and consistently associated with depression both concurrently and longitudinally in multiple studies (e.g., Schoofs et al., 2010; Treynor et al., 2003). Relatedly, people can employ various thoughts and actions, known as savouring responses (Bryant & Veroff, 2007), to manage PA. One such response is positive rumination, defined by Feldman et al. (2008) as “the tendency to respond to positive affective states with recurrent thoughts about positive self-qualities, positive affective experience, and one's favourable life circumstances” (p. 509). Positive rumination is assessed using two subscales of the Response to Positive Affect questionnaire (RPA; Feldman et al., 2008). The combined score on this measure (e.g., Harding et al., 2014; Li et al., 2017) or one of its subscales (emotion-focused: Feldman et al., 2008; Raes et al., 2009; self-focused; Vanderlind et al., 2022) were found to be negatively associated with depression (but see Gilbert et al., 2017). Importantly, it was found that positive rumination uniquely contributes to depressive symptoms above and beyond the contribution of brooding (Feldman et al., 2008; Raes et al., 2009). While brooding and positive rumination have been proposed as distinct regulatory responses that may be associated with vulnerability to depression, both involve a repetitive thinking style that prolongs the affective experience. Thus, it is not surprising that Burke et al. (2018) show that adolescents reporting greater depressive symptoms were characterised by higher levels of reported brooding coupled with lower levels of positive rumination.

Biases in the capacity to hold affective content in working memory (WM) assessed using various performance-based measures, have also been postulated to be involved in the maintenance of NA and failure to maintain PA (LeMoult & Gotlib, 2019). WM is a limited capacity system that temporarily maintains active content to support goal-directed behaviour such as planning and problem-solving (Baddeley, 1986). WM holds on to affective content when it is helpful to one's current goals and ignores affective content that distracts from these goals (Kircanski et al., 2012). Like in studies assessing regulatory strategies in depression, most performance-based studies have examined the processing of negative affective stimuli, and relatively fewer studies examined the processing of positive affective stimuli (for a review, see Kircanski et al., 2012).

The maintenance of affective goal-relevant content in WM has been examined using several tasks, most commonly the N-back task (Levens & Gotlib, 2010) and the affective maintenance task (AMT, Mikels et al., 2008). In the emotional N-back task, participants are presented with a series of negative, positive and neutral stimuli (typically sad, happy or neutral faces) in sequential order and are requested to answer whether the current item matches the one that was presented N items earlier in the sequence. Levens and Gotlib (2010) found that depressed individuals were more accurate on negative trials and less accurate on positive trials relative to healthy controls, suggesting that depression is associated with enhanced maintenance of negative and reduced maintenance of positive content in WM compared to healthy controls. This bias in maintaining affective content in WM has been found to persist beyond the current depressive episode (Levens & Gotlib, 2015). Interestingly, using this task when with negative, positive and neutral words split into separate blocks showed that, as compared to controls, depressed individuals exhibited reduced accuracy and longer response time only in the positive block but not in the negative (or neutral) blocks (Zhang et al., 2018).

In the N-back task, participants are instructed to maintain the content of the stimuli. However, recent models suggest that the content of WM includes not only the stimuli themselves but also the affective experience, independently from the visual or other types of content (Mikels & Reuter-Lorenz, 2019). In the AMT, participants are required to maintain (over a brief period) the affect elicited from viewing an emotional stimulus and then to compare the intensity of that affect to the intensity of the affect evoked by a second stimulus of the same valence (Mikels et al., 2008). Subjective intensity ratings that are provided later by each participant for each stimulus are then used to establish the accuracy of comparisons. Using the AMT, Gruber et al. found that individuals remitted from depression (with no residual symptoms) did not differ from controls in maintaining NA or PA (Gruber et al., 2013). In this study, however, low levels of current depressive symptoms did not allow for the assessment of the association between current depressive state and WM affective maintenance.

Despite theoretical postulations (Gotlib & Joormann, 2010), to date, no study has attempted to synthesise multiple factors playing a role in the maintenance of NA and PA. These two constructs may work in concert so that regulation strategies, such as brooding and positive rumination, might depend on the engagement of WM processes maintaining these affective states in WM. Therefore, our aim was to assess both regulatory strategies (brooding and positive rumination) and WM affective maintenance that may play a role in affective experience exhibited in depression (i.e., lingering NA and short-lived PA). We hypothesised that savouring strategies that prolong PA (such as positive rumination) and WM-based processing of PA (as assessed by the AMT) contribute to the severity of depression above and beyond the analogous measures of NA (i.e., brooding and WM-based processing of NA assessed by the AMT).

Methods

Participants

We recruited participants via Amazon Mechanical Turk (MTurk) and performed the experiment via Qualtrics Software (Provo, UT). Participation in the study was restricted to MTurk users above 18 years of age residing in the United States and fluent in English. To guarantee data quality, we followed documented recommendations (Kennedy et al., 2020). Therefore, only MTurk users with a history of providing good-quality responses (acceptance ratio of ≥ 90%) were invited to participate.

Based on studies by Gruber et al. (2013) and Burke et al. (2018), we expected a small effect size for our hypothesis. Based on G*power (Faul et al., 2009), to detect a small effect size with a linear regression model with a power of .80, a sample size of 200 was required. Consistent with exclusion rates in similar studies, we oversampled by about 25% (N = 265).

Exclusion criteria were: (1) inadequate responses to five questions that discriminated between attentive and inattentive participants; (2) improper answer to an open-ended question presented at the end of the experiment; (3) a non-unique GPS coordinates location. Accordingly, 43 participants who met these criteria and 3 who failed to complete the intensity rating task (described below) according to instructions were excluded. Thus, the final sample included 219 participants.

General procedure

Following a short description of the study, participants provided informed consent. Next, they completed the AMT, followed by the intensity rating task. Finally, participants completed the self-report measures in random order and several demographic questions. Completion of the study took about 15 minutes, and participants were debriefed and compensated.

Affect maintenance task (AMT)

Following Mikels et al. (2008), participants were shown two different images, one at a time, and were instructed to hold the intensity of their emotional response for image 1 and compare it with the intensity of their emotional response for image 2. There were 40 trials presented in random order: 20 trials of positive and 20 trials of negative stimuli comparisons. Each trial starts with a fixation point of 500 ms, followed by image 1, which remains on the screen for 2 s. Following an intertrial interval of 8 s, image 2 is presented for 2 s. Participants were then requested to indicate which image was more intense. Before test trials were presented, participants performed two practice trials.

Stimuli. 40 negative and 40 positive images were chosen from the International Affective Picture System (IAPS; Lang et al., 1999) for the creation of 40 trials while replacing 11 images used in the original AMT due to cultural differences. Based on Mikels et al. (2005) norms, the intensity ratings of negative and positive images did not differ (negative: M = 3.62, SD = 0.78; positive: M = 3.65, SD = 0.70; t(78) = -0.14, p = n.s). Images were divided into trials in advance to make sure the difference in intensity between the two stimuli did not differ between trials (M_{negative difference}= 1.18, SD = 0.51; M_{positive difference} = 1.07, SD = 0.48; t(38) = 0.70, p = n.s).

Subjective intensity rating task

Following the AMT, participants completed the intensity rating task. In this task, participants rated the intensity of their affective response to each of the 80 images presented in the AMT trials. Images were presented one at a time in a random order, and participants rated their response on a visual analog scale, which was labelled from “not at all” to “extremely”. Three participants who showed no variance in their responses (a consistent extreme rating score) were excluded from the analysis.

Self-report measures

Participants completed the Beck Depression Inventory-II (BDI-II; Beck et al., 1996), a 21-item self-report measure of depression with item 9 (suicidal ideation) excluded. Participants also completed the brooding subscale of the RRS (Treynor et al., 2003), which includes five items, and the emotion-focused and self-focused positive rumination subscales of the Response to Positive Affect questionnaire (RPA; Feldman et al., 2008), which includes nine items. In our sample, Cronbach's α for the BDI-II scale, the brooding subscale and the positive rumination subscales were good to excellent (.96, .86, .92, respectively).

Results

A response on each AMT trial was considered accurate when there was a match between participants’ response to the AMT trial and the subjective intensity ratings of the associated images. For example, an accurate response was coded when the participant selected image X as more intense than Y on an AMT trial, and rated image X as more intense than Y on the subjective intensity rating task. Total accuracy scores were calculated for each participant for positive and negative trials and reflect the percentage of trials in which the participant answered accurately from the total 20 trials presented for each valence.

Descriptive statistics

Participants’ ages ranged from 20 to 71 years of age (M = 41.24, SD = 12.2); 43% females, 82.2% were White/Caucasian, 7.3% were African American, 4.6% were Asian and the rest were multi-racial. BDI-II scores ranged from 0 to 52, 17.4% of the sample scored 21 or above, which represents moderate to severe depressive symptoms.

Descriptive statistics and bivariate correlations for study variables are presented in Table 1. The mean total accuracy score for participants (i.e. the average of negative and positive accuracy scores) was 0.80 (SD = .11), which was greater than chance (0.5), t(218) = 40.05, p < .001, Cohen's d = 2.73). In addition, t-tests for paired samples revealed that accuracy in negative trials (M = .82, SE = .01) was significantly higher than accuracy in positive trials (M = .78, SE = .01, t(218) = 3.46, p < .001, Cohen's d = 0.23). Paired samples t-test revealed that negative images were rated as significantly more intense than positive images on the subjective intensity rating task (t(218) = 24.20, p < .001, Cohen's d = 1.64).

Table 1. Descriptive statistics and correlations for study variables.

	M	SD	1	2	3	4	5	6	7
1. BDI	9.49	11.56	–
2. RRS brooding	9.59	3.66	.699***	–
3. RPA PR	21.16	6.25	–.439***	–.183**	–
4. Negative accuracy	0.82	0.13	–.040	–.098	–.109	–
5. Positive accuracy	0.78	0.13	–.181**	–.194**	–.100	.472***	–
6. Negative intensity ratings	58.74	15.27	–.053	.121†	.369***	–.188**	–.029	–
7. Positive intensity ratings	38.54	16.86	–.099	–.005	.390***	–0.172*	.123†	.709***	–

Note. N = 219. * p < .05; ** p < .01; *** p < .001; † p < .08.

BDI, Beck Depression Inventory-II; RRS, Ruminative Response Scale; RPA PR, Response to Positive Affect questionnaire – positive rumination subscales. Negative and positive accuracies as measures in the AMT.

As seen in Table 1, brooding was positively, and positive rumination negatively correlated with depression scores. In contrast to our expectations, the correlation between negative accuracy scores and depression scores was not significant. However, the correlation between the positive accuracy scores and depression scores was negative and significant. The correlation between brooding and negative accuracy scores was not significant, but the correlation between brooding and positive accuracy scores was negative and significant (although weak). Positive rumination did not show a significant correlation with either positive or negative accuracy scores. The negative and positive accuracy scores, as well as the negative and positive intensity ratings, were positively and significantly correlated.

Hypotheses examination

A hierarchical linear stepwise regression was conducted with BDI-II as the predicted variable. Brooding and negative accuracy standardised scores were entered into the model in the first step, and positive rumination and positive accuracy standardised scores were entered in the second step. As seen in Table 2, brooding contributed significantly to the model and was positively associated with depression scores. However, negative accuracy standardised scores did not contribute to the model. The additional contribution of the second step to the model was significant (F(2,214) = 28.83, p < .001, Cohen's f² = .28), and it added 10.85% to the explained variance. In this step, both positive rumination and positive accuracy standardised scores significantly contributed to the model, such that greater positive rumination and positive accuracy standardised scores were negatively associated with depression scores over and above the contributions of brooding and negative accuracy standardised scores.

Table 2. Hierarchical regression analysis prediction depression scores from brooding, positive rumination and accuracy scores.

	β	SE	t
Brooding	.70	.57	14.35**
Negative accuracy	.03	.57	.58
Step 1. R^2= 48.87%, F (2,216) = 103.24**
Brooding	.62	.52	13.71**
Negative accuracy	.04	.57	.74
Positive rumination	–.33	.52	–7.44**
Positive accuracy	–.11	.58	–2.21*
Step 2. R^2= 59.73%, F (4,214) = 79.34**

* p < .05, ** p < .001.

Negative and positive accuracies as measures in the AMT.

Discussion

We examined factors playing a role in the maintenance of NA and PA and their association with depressive symptoms. Specifically, we focused on two constructs: regulation strategies and information processing, which are responsible for maintaining these affective states in WM. Our findings provide empirical support to theoretical views of the joint role of NA and PA in depression by demonstrating that brooding, positive rumination, and WM-based maintenance of PA (but not NA) are independently associated with the severity of depressive symptoms.

Our findings are consistent with prior work that shows that repetitive thinking style is not always detrimental (Harding & Mezulis, 2017) and specifically with evidence showing that brooding and positive rumination regulation strategies are differentially associated with depressive symptoms (e.g. Burke et al., 2018). While both regulatory strategies involve a preservative cognitive style, their effect on one's affective experience and psychological outcomes diverge. Yet, the possible underlying shared mechanism that explains their joint role in predicting depressive symptoms is still unclear.

Our data regarding WM-based biases suggest that difficulty in maintaining positive, but not negative, affective experiences in WM had a significant and unique contribution to depressive symptoms above and beyond self-reported regulatory strategies. Taken together, these data are consistent with models of depression that postulate that reduced sensitivity to reward is one of its key features. Indeed, trial-by-trial probability analyses of reward responsiveness revealed that whereas depressed participants were responsive to the delivery of single rewards, they were unable to integrate the values of rewards over time to generate a persistent response bias towards the more rewarded cues in the task (Pizzagalli et al., 2008). These results further support the view of depression as associated with difficulty in maintaining positive affective experiences.

Interestingly, maintaining NA in WM was not associated with depressive symptoms. These findings highlight the independence between factors associated with the maintenance of NA and PA. Given that information processing is multi-componential, differential functions appear to be associated with positive and negative affective information (Miyake et al., 2000). For example, difficulty in disengaging from negative information has been suggested to underlie the enhanced processing of negative materials exhibited in depression (Koster et al., 2011). Therefore, it could be that impaired disengagement from NA in WM may underlie the lingering experience of NA whereas maintenance of PA in WM may underlie the short-lived PA experience evident in depression. In conclusion, it is possible that cognitive biases underlying mechanisms of depression maintenance differ as a function of the affective content (negative vs. positive) being processed. Thus, difficulty in disengagement may depend on the valence of the stimuli which is the focus of such disengagement.

An alternative explanation for that maintaining PA in WM but not NA was associated with depressive symptoms is that whereas norms for the intensity of the stimuli used in the AMT showed no difference between negative and positive stimuli (Mikels et al., 2005), intensity ratings provided by our participants indicated that negative stimuli were more intense than positive ones. This may imply that in our study, PA was milder and, therefore, more difficult to maintain. In contrast, NA was more intense and, therefore, easier to maintain, limiting the impact of individual differences.

While emotion regulation strategies are typically measured using self-report scales, information processing biases are measured using performance-based measures. Self-report measures provide valuable insights into individuals’ subjective experiences and perceptions of affect maintenance. On the other hand, performance-based measures offer an objective assessment of behavioural and cognitive processes related to affect maintenance. Integrating these methods allows us to capture a more comprehensive understanding of affect maintenance, considering both the conscious appraisal of emotional experiences and the implicit cognitive processes involved.

Several limitations of the present study need to be noted. First, data were collected through an online format, which could impact participants’ engagement. Second, since the present study is cross-sectional, no causal inferences can be made regarding the relationship between brooding, positive rumination, affective WM impairments and depressive symptoms. Future research may examine the causal role of WM biases by conducting a longitudinal investigation, as well as by examining whether training these biases impact depression severity. Third, while we assessed only brooding and positive rumination regulatory strategies, it is recognised that individuals are characterised by a personalised “toolbox” of regulatory strategies (Daches & Mor, 2015). Similarly, there are various processes, other than WM maintenance, that are suggested to predict depression. These processes include but are not limited to, attentional disengagement and interpretation biases (LeMoult & Gotlib, 2019). Thus, an investigation including a larger set of biases is needed to provide a comprehensive understanding of the cognitive mechanisms underlying depression.

To summarise, although numerous investigations have attempted to quantify the contribution of specific factors to depressive symptoms, this is the first study to use an adequately powered sample to reliably examine key factors that share the same theoretical basis (maintenance of affective experience) associated with depressive symptomology. The present study shows that biases in maintaining PA in WM are associated with depressive symptoms over and beyond brooding and positive rumination regulatory strategies. Our data also highlight differential pattern of processes related to depression maintenance: whereas brooding and positive rumination were associated with depression severity, only positive, but not negative, WM-related performance-based measure had a unique contribution to depressive severity. Future studies may seek to clarify the combined effect of regulatory strategies and information processing biases toward both negative and positive affective experiences in depression. A further refinement of the dual-vulnerability model of depression is needed to contribute to the prevention and treatment of this painful condition.

Acknowledgments

The authors would like to thank Christian Waugh for his advice and assistance in implementing the affective maintenance task and Lior Ronn for his help in data collection.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.